Response to Comments: MolDX: Molecular Biomarker Testing for Risk Stratification of Cutaneous Squamous Cell Carcinoma

The following comment was submitted to Palmetto GBA, CGS, WPS, and Noridian:

In response to the proposed LCD MolDX: Molecular Biomarker Testing for Risk Stratification of Cutaneous Squamous Cell Carcinoma, Castle Biosciences is submitting the following written comments for your consideration. The DecisionDx-cSCC gene expression profile test (40-GEP) has met criteria of medical reasonableness and necessity for Medicare coverage and should be included as a covered test in the final LCD.

Although the proposed LCD correctly identifies the unmet clinical need for improved risk stratification in patients with cutaneous squamous cell carcinoma (cSCC) and identifies that the 40-GEP test as capable of metastatic risk stratification, the policy’s concluding non-coverage rationale rests on the claim that the 40-GEP has “not yet demonstrated definitive value above the combination of available clinical, pathological, and staging criteria in accurate risk stratification.” This conclusion is not supported by an evidentiary review. Missed publications and inaccurate interpretations of data present in the proposed LCD impair the ability to draw accurate conclusions. We will address these inaccuracies in the following sections:

1. cSCC is a complicated disease state with significant under and over staging due to poor accuracy and low utility of staging systems
2. 40-GEP has been robustly validated in cohorts that are representative of the intended use population
3. Metastasis is the important outcome for Medicare beneficiaries
4. 40-GEP adds statistically significant, independent quantifiable risk information to be used in conjunction with staging. Meaning multiple studies have confirmed the definitive value above the combination of available clinical, pathological and staging criteria
5. 40-GEP significantly stratifies risk of metastasis within stages and can be used within the context of other risk stratification tools to improve metastatic risk assessment
6. Correlation of GEP results with stages or other risk factors is not a statistically or scientifically supported analysis to assess independent value from test results
7. 40-GEP has demonstrated consistent incorporation into patient management plans in alignment with national guidelines recommendations
8. 40-GEP identifies ART eligible patients with very high biological risk of metastasis that benefit from ART and similarly identifies a significant population of ART eligible patients with low biological risk of metastasis that may avoid ART
9. Appropriate use of 40-GEP to inform ART decisions can provide significant healthcare cost savings for the Medicare Trust Fund
10. CMS’ Division of New Technology recently completed their review of the clinical validity and independence of the 40-GEP and concluded that the 40-GEP is a valid and independent predictor of metastasis within this high risk cSCC population
11. 40-GEP has met criteria for medical reasonableness and necessity as defined by the Medicare Program Integrity Manual and the is consistent with similar tests in other disease states
12. Summary and Conclusions

Given the missed publications and inaccurate interpretations of the data present in the proposed LCD, we request an evidentiary review that reflects the state of clinical management of high risk cSCC and in which the clinical validity and use of the DecisionDx-SCC to improve management decisions, and therefore outcomes, is represented.

I. cSCC is a complicated disease state with significant under and over staging due to poor accuracy and low utility of staging systems

As discussed, and correctly identified in the proposed LCD, patients with cSCC with one or more high risk features have an elevated risk of poor outcomes and require additional management plans following initial surgical resection. The proposed LCD also correctly identifies that current staging and risk assessment methods have low PPV and are inconsistent with one another. However, the proposed LCD does not discuss the implications of low PPV and the need for improvement in risk stratification to inform risk aligned management decisions. Correctly understanding the unmet clinical need is critical to evaluate the appropriateness and sufficiency of molecular testing to address the need. Over and under staging is a significant and costly problem for Medicare beneficiaries. Risk stratification approaches reliant on the presence or absence of clinical or pathological factors alone, including National Comprehensive Cancer Network (NCCN), American Joint Committee on Cancer 8^th edition (AJCCv8) and Brigham and Women’s Hospital (BWH) staging systems, all misclassify a significant number of patients with cSCC. At least 30% of patients who go on to experience regional and distant metastasis are initially classified low-stage at initial diagnosis by staging systems and less than one-third of patients who present initially with high stage disease by staging systems will go on to experience poor outcomes.^1–4 This means that in patients with high-risk cSCC, a substantial number of patients with lower risk disease by staging systems experience poor outcomes and are currently being under treated, and as many as 70% of patients with high stage disease do not experience a poor outcome and are being over-treated (Table 1).

The proposed LCD also overestimates the use of formal staging systems in risk assessment. The survey publication from Patel, et al. includes data that among survey participants, 89% of physicians treating cSCC use formal staging systems. The proposed LCD should also include that 87% of the physicians surveyed are primarily from academic institutions where use of staging systems is more common and does not reflect community practice.⁵ In a quote from the manuscript, “A limitation is that the panelists were identified from mainly academic tertiary referral centers and thus results may not be generalizable to all physician practice patterns.”⁵ Of note, Gronbeck, et al. reports that only 15% of dermatologist members of the American Academy of Dermatologists were practicing in academic settings in 2020.⁶ The publication from Patel, et al. also states that “a large number of respondents” reported that they do not use staging to inform management decisions (31%-76% depending on the management decision). This is a key conclusion of the paper that drives at the unmet clinical need underpinning a molecular risk stratification test. To complicate matters further, Patel et al. demonstrated that this group of physicians utilizes different staging systems for different utilities and there is even discordance here between dermatologists and other types of treating physicians. For example, for determining if radiologic imaging should be performed, the most common staging system utilized was the BWH system for dermatologists (38.3%) and AJCC for other cancer specialists (50.7%).

A similar survey publication from Litchman, et al. that is cited, but not reviewed in the proposed LCD, finds that only 56% of clinicians use staging systems to assess risk in or to inform management of patients with cSCC. Broadly speaking, while staging systems may represent the best amalgamation of clinical and pathological risk information prior to the availability of molecular testing, the limitations in its accuracy have led to low incorporation to guide all management decisions in high-risk cSCC. Implications of this for evaluation of 40-GEP is that comparisons only to staging to determine clinical utility is flawed; staging itself does not meet the unmet need in this population, making the evaluation of new tests in combination with current risk assessment tools (to see their added improvement) another important clinical perspective that is absent from the proposed LCD and that should be added to the final LCD.

The broad nature of NCCN guideline recommendations to manage treatment of patients with NCCN high risk and very high-risk disease are also insufficiently detailed in the draft LCD. This is another critical piece of information required to understand fully the unmet clinical need in this population. When reviewed and summarized fully, the following errors and omissions in the “Surveillance and Treatment Intensity are Influenced by Risk-Stratification Group” section of the draft LCD should be addressed:

(i) Mohs Micrographic Surgery (MMS) or other forms of peripheral and deep en face margin assessment (PDEMA) are available treatment options for high risk and very high risk cSCC tumors and are the preferred surgical modality for very high risk cSCC (see NCCNv1.2023 MS-7);

(ii) Consideration for adjuvant radiation therapy is available broadly for high risk and very high risk cSCC with “poor prognostic features”, and cSCC patients with a “high risk for regional or distant metastasis” (see NCCNv1.2023 cSCC-3A and footnote y);

(iii) Follow-up frequency for high risk and very high risk have such broad recommendations as to have completely overlapping ranges, and direct physicians that “frequency of follow-up should be adjusted based on risk” with no distinction for how risk influences management recommendations (see NCCNv1.2023 cSCC-6 and footnote jj);

(iv) Nodal assessment via palpation or imaging is guided by risk assessment (see NCCNv1.2023 cSCC- 6); and

(v) Use of imaging in surveillance is guided by risk assessment (see NCCNv1.2023 cSCC-6).

While adjuvant radiation therapy (ART) is discussed as a management option for patients with high-risk and very high-risk cSCC, the proposed LCD does not include a discussion of the demonstrated improvement in outcomes in patients who receive this therapy. This becomes important when a need for improved outcomes to demonstrate utility of the 40-GEP is discussed. ART is currently recommended and considered for a broad range of high risk cSCC tumors by all relevant guidelines and societies (NCCN, AAD, ASTRO, ACR) because it has demonstrated improvement in outcomes. Additionally, ART has been and is currently covered by Medicare for use in these same patients. The combination of the low PPV and lack of consistent use in staging as well as the known benefit of ART in patients with very high risk of metastasis shows clearly the unmet clinical need. Thus, improvements in patient outcomes are to be expected to follow if a test can demonstrate improved risk stratification that informs patient selection for ART.

One publication from Ruiz et al. from 2022 that is not referenced or considered in the proposed LCD demonstrates that ART is associated with a roughly 50% reduction of poor outcomes in patients with high risk of regional and distant metastasis who have BWH T2b and T3 stage cSCC tumors.⁷ This 50% reduction in the risk of regional and distant metastasis is not unexpected based upon prior studies as well as the accepted 50% reduction in metastasis rates when ART efficacy is examined across other epithelial cancers. The most striking improvement in outcomes in the Ruiz 2022 study occurred in patients with the highest risk of metastasis in which ART treatment reduced the metastatic rate from 31.0% to 17.2%, which is a clinically significant improvement (Figure 1). For this group of patients, the significant reduction in metastatic rate outweighs the potential side effects of radiation. Patients in this study with moderate level of metastatic risk only saw an 5% absolute improvement in rate of metastasis, which is not as significant clinically. The ability to identify the responders (i.e., patients with high biological metastatic risk) in a population of patients who are eligible for ART but are on the lower end of high-risk tumor stage is a specific unmet clinical need not articulated in the proposed LCD.

An accurate understanding of the disease state and the benefits provided to patients from current treatment recommendations provides the necessary framework to evaluate if a molecular diagnostic tool addresses an established unmet need. As such, we request that these sections be updated in the final LCD to contain the information and references in the table below, which support a range of the proven health impacts of cSCC treatment modalities included in guidelines (Table 2).

II. 40-GEP has been robustly validated in cohorts that are representative of the intended use population

In order to describe the clinical validity and utility of the test and answer questions raised regarding test performance, it is first necessary describe the high-quality cohort that appropriately reflects the intended use population and the high quality of data used to validate the test.

First, the proposed LCD suggests that the clinical validation studies for 40-GEP do not include the intended use population. We would like to point out that the test was validated in patients with one or more high risk factors as our intended use population. Subsequent informal discussions with MolDX raised questions regarding clinical utility, not clinical validity, of the test in patients with the highest risk tumors, identified as BWH T3 (which is the presence of all 4 BWH risk factors, including diameter of ≥.2 cm, poor differentiation, perineural invasion of ≥0.1 mm, and invasion beyond the subcutaneous fat or the presence of bone invasion), or presence of lymphovascular invasion (LVI) as patients that may be of sufficiently high risk that 40-GEP would not change patient management. As such, Castle responded to MolDX with proposed coverage criteria language which would limit coverage from such cases. As for BWH T3 patients and patients with LVI, the test was validated for use in these patients, as they are represented in the validation cohort, albeit in low numbers, and the test demonstrated risk stratification in these patients. Even so, data was published and available by BWH T stage in the supplemental data from Ibrahim et al. 2021, such that test performance excluding patients with BWH T3 tumors could be easily evaluated (see Ibrahim et al. 2021, Supplemental Figure 2).¹⁹ Only 6 patients in the cohort published by Ibrahim, et al. displayed LVI and the manuscript states, “lymphovascular invasion was rare in this cohort”. This publication was submitted to MolDX and is referenced in the proposed LCD, but it appears the supplemental data was not considered for review.

However, in order to address this concern as completely as possible, we have evaluated risk stratification in the published cohort from Ibrahim, et al., with BWH T3 and LVI excluded and found, as expected, that the performance of the test remains consistent with these patients removed. Only 15 patients in the Ibrahim et al. manuscript were either BWH T3 or had presence of LVI. After removal of these patients, this subset demonstrates consistent event rates when compared to the published cohort in Ibrahim, et al. The overall cohort risk of metastasis was 13.8% (compared to 15.0%), and a Class 1 result had a metastatic rate of 6.3% (compared to 6.6%), a Class 2A result had a metastatic rate of 18.1% (compared to 20.0%), and a Class 2B had a metastatic rate of 52.3% (compared to 52.2%) (Figure 2). It is noteworthy that these 15 patients were distributed among Class results (4 were Class 1, 9 where Class 2A, and 2 were Class 2B), confirming that the highest risk patients by clinical and pathological factors were not all classified as Class 2B and were not responsible for driving PPV of the test as suggested in the proposed LCD. Moreover, the test did demonstrate stratification of metastatic risk across these patients, with a 25% risk of metastasis in Class 1 compared to 55% in Class 2A and 50% in Class 2B, although this is a small number of patients.

The proposed LCD also requests validation to be carried out in rare subpopulations. It is typically required for a validation cohort to demonstrate expected performance in a population that is representative of the overall distribution of patient and tumor characteristics; the cohort used for the validation of the test is representative of the overall patient population who develop cSCC, which is predominantly Caucasian and male. This fact is best evidenced by comparing the demographic data from the cohort in Ibrahim et al. with studied cohorts for BWH and AJCCv8 staging, both of which are cited in the LCD. For AJCCv8 in a manuscript by Karia et al. 2018, the studied cohort was 98.3% “white race” and for BWH, another manuscript by Karia et al 2014, the studied cohort was 97% “non-Hispanic white.”^2,20 While we agree that test performance in persons of color is an important area of future research, cSCC occurring persons of color is rare and predominantly arises in association with scarring or chronic wounds, which are associated with higher risk of metastasis than other cSCC eitologies.^21,45 While we accrue additional data to support consistent performance of the test in this unique disease etiology and subset of patients, it is not ethical to withhold coverage of a test with demonstrated validity and utility from a validation cohort similar in composition to other risk stratifcation system cohorts.

Similarly, the proposed LCD requests additional validation data in patients with desmoplasia. While the presence of desmoplasia may be associated with more aggressive disease, it is also exceedingly rare and infrequently documented in cSCC pathology reports in the United States. Importantly, desmoplasia is not included as a risk factor in BWH or AJCC staging. Further, there is not currently a CAP synoptic report in effect to standardize reporting practice for primary cSCC. A CAP synoptic report is in effect for patients with node positive cSCC in the head and neck (outside the intended use population of this test), and even then, the presence or absence of desmoplasia is not included as standard reporting criteria (https://documents.cap.org/protocols/HN.SCC_1.0.0.1.REL_CAPCP.pdf). Of note, the proposed LCD reference supporting that desmoplasia is high risk in the original MolDX LCD published is removed from the version published subsequently. Notably, this reference was a publication from a German cohort published ~30 years ago. A more recent reference is included in the proposed LCD (Reference 77: Eigentler TK et al. 2017), yet again this is from an exclusively German cohort.

III. Metastasis is the important outcome for Medicare beneficiaries

Three outcomes of interest exist in patients with cSCC, including local recurrence, regional/distant metastasis, and death. Among these, metastasis is the primary outcome of interest for guiding specific management decisions for localized cSCC per NCCN. According to NCCN guidelines, decisions for ART and SLNB are clearly to be guided by risk of regional or distant metastasis (NCCN guidelines v1.2023, SCC-3A footnote y, footnote s). As such the 40-GEP was developed to predict risk of metastasis to help navigate the widest range of management options, including ART and SLNB, available to all patients with localized high-risk cSCC as outlined in guidelines for management of patients with cSCC.

However, the proposed LCD requests evaluation of 40-GEP test performance with the additional endpoint of disease specific death. This additional endpoint has been reported. Specifically, in Wysong, et al., Class 2A was reported to predict statistically significant increase in disease-specific death with a hazard ratio of 2.44 and Class 2B predicted a statistically significant increase risk of disease-specific death with a hazard ratio of 10.15 compared to patients with Class 1 results (p<0.01, p<0.0001, respectively).²⁶ Thus, while risk of metastasis the clinically relevant endpoint, the 40-GEP test has also demonstrated risk-stratification for disease specific death.

In the cohort published by Ibrahim, et al., the 40-GEP test result produced significant risk stratification for disease specific death (Figure 3). Taken together, these data reinforce the conclusion that the DecisionDx-SCC test result provides independent risk stratification information that can improve the accuracy of risk stratification for regional/distant metastasis as well as for disease specific death. Castle Biosciences has chosen to limit DecisionDx-SCC reporting to regional/distant metastasis as the broad range of risk aligned management options articulated in current guidelines are related to the treating clinicians’ perception of their patient’s risk for metastasis and not specifically to the patient’s risk for disease specific death.

IV. 40-GEP adds statistically significant, quantifiable risk information to be used in conjunction with staging

Perhaps the most concerning misinterpretation of data in the proposed LCD is the assertion that the 40-GEP has “not demonstrated definitive value above the combination of available clinical, pathological, and staging criteria in accurate risk stratification”. This statement is objectively false and demonstrates a misunderstanding of published data. Definitive value above the combination of available information has been established by the statistical independence of the test in Cox multivariable analyses in comparison to all known risk factors and staging systems, and the risk stratification within every risk group/substage of all known staging systems. Both of these points will be addressed in detail below. To correct this aspect of the proposed LCD, a re-review of the complete manuscript from Ibrahim et al. 2021 is needed, including the published supplemental data.⁷ A complete review of this manuscript and data supplement will demonstrate the robust clinical validity of the test in the context of risk factor-based stratification approaches and directly address the primary critique contained in the proposed LCD.

The 40-GEP has consistently demonstrated statistical independence from staging systems and individual risk factors through multivariable analysis in two independent validation cohorts.^19,22 Multivariable analyses within subsets of patients, including those treated with Mohs micrographic surgery²³, head and neck tumors²⁴, and immunosuppressed patients²⁵, have all shown consistent findings of statistical independence of test results. 40-GEP has shown improvement in risk prediction over every current clinical and pathologic-based risk stratification approach, including individual risk factor-based assessment, NCCN risk designation, BWH staging, and AJCC v8 staging and has demonstrated risk stratification within stages/risk groups. This means that 40-GEP adds prognostic information not obtainable from risk factors or staging and can be added to other known risk factors for more complete assessment of metastatic risk. In fact, the quantitative risk associated with each prognostic variable and described as the hazard ratio is multiplicative when more than one independent high-risk feature is detected for a particular tumor. It needs to be made abundantly clear that the 40-GEP is intended to be used with other risk assessment systems; the 40-GEP test does not invalidate or compete with independent risk information obtained from factors or staging. Instead, as demonstrated in publication and detailed here, the test result is to be used with all available information for a comprehensive, clinical, pathological, and genomic assessment of risk.

Within the proposed LCD, specific modifications were requested to the multivariate analysis. These were performed and presented during the open comment meeting. Specifically, it was requested that tumor diameter be included as a discrete variable instead of continuous, PNI be adjusted to only include > 0.1 mm, and high risk tumor locations (area H and area M) be included (Table 3). This analysis was performed in the largest, combined cohort of validation patients to date, with BWH T3 and LVI excluded. The other risk factors requested for inclusion in multivariate analysis (rapidly growing, site of prior radiation or chronic inflammation, neurologic symptoms, and rare, aggressive histologic subtypes) were not able to be included due to insufficient reporting for these features; however, these features are not routinely reported, and not incorporated into staging. As such, omission of these factors is not anticipated to materially impact the analysis or conclusions. In this updated analysis, the 40-GEP is again independent of risk factors for prediction of metastatic risk, adding new information to improve comprehensive risk assessment. Note that the relative hazard ratio of risk factors without inclusion of GEP and with inclusion of GEP are similar, indicating that the risk factors maintain importance in risk prediction in patients with cSCC and that 40-GEP can be used in combination with risk factors.

Statements in the proposed LCD suggest that clinicians do not know how to incorporate 40-GEP with risk factors to accurately estimate risk and determine appropriate management plans, but this is simply untrue. Clinicians treating cSCC have been combining risk information (from the left column of Table 1, without 40-GEP) to inform management plans for years prior to the availability of genomic risk information; just as a clinician knows how to combine thickness information with diameter information by considering the two different hazard ratios and combining their risk, the 40-GEP results can be integrated with clinical and pathological information in the same way. The source of the information - be it clinical, pathologic or genomic - does not impact the ability for clinicians to act upon it, and it is illogical to suggest that would be the case.

The proposed LCD expresses concern regarding under staging in the validation cohort that could undermine comparisons to staging due to missing data in some cases. Because there is no CAP synoptic for cSCC, pathology reports do not consistently report on all risk factors. As articulated in the Wysong et al. manuscript, “One challenge with clinicopathologic-based guidelines is that high-risk features are often undetected through initial biopsy and, therefore, often cannot be used for surgical planning. The 40-GEP can be performed on superficial biopsies, thus enabling improved surgical decision making using molecular risk refinement before full capture of histopathologic features on excisional specimens.” As part of the initial validation effort, to stage tumors as completely as possible in the Wysong et al, biopsy samples underwent secondary pathology review for assessment of risk factors, demonstrating that the completeness of staging in this cohort is likely far greater than what is typically observed in clinical practice. Unfortunately, some risk factors, such as tumor diameter and depth of invasion, are frequently not able to be accurately assessed from the diagnostic biopsy and, therefore, are not included in a final pathology report. Importantly, missing clinical and/or histopathologic tumor information is observed in other studies evaluating BWH and AJCC staging, again highlighting that this is a consequence of the disease state. In fact, publications that establish and validate AJCC and BWH staging systems have missing data for at least one factor in about 23% of patients (tumor depth or pre-operative clinical lesion size).^3,20 Of note, both of these publications are cited in the proposed LCD as supportive of the accuracy of staging systems without any critique regarding the potential for under staging, again calling into question the level of evidence comparisons between factor-based risk assessment and 40-GEP.

Importantly, the initial validation study published by Wysong, et al. confirmed that incomplete staging factor reporting for a subset of cases did not impact the conclusions or statistical independence of the test from staging systems.²⁶ Supplemental Table V in this manuscript shows a sensitivity analysis in which highest risk that could be possible in cases with missing data were imputed, and a Cox multivariable analysis was performed comparing 40-GEP to stage. In this analysis, the 40-GEP maintained statistical significance with comparable hazard ratios, indicating that the missing stage information in the cohort does not detract from the conclusion that the 40-GEP adds statistically significant and independent information to staging systems. Notably, this imputation of the highest risk variable in cases with missing staging criteria increased the binary risk classification of 164 cases by AJCC classification, but only 6 cases by BWH classification. This should be cited and discussed in the proposed LCD (full text and supplement are attached to this letter).

V. 40-GEP significantly stratifies risk of metastasis within stages and can be used within the context of other risk stratification tools to improve metastatic risk assessment

When assessing if a molecular prognostic tool adds value to existing clinical and pathological factors, the other significant factor to evaluate for molecular prognostic testing is whether the difference in risk identified by the test is sufficient to justify different clinical action. The 40-GEP has also shown significant risk stratification within various risk groups by NCCN, BWH and AJCC, published in Figures 2, 3, Supplemental Figure 2 and Supplemental Table 2 of Ibrahim, et al. 2021, and adds prognostic information not obtainable from risk factors or staging and can be added to other known risk factors for more complete assessment of metastatic risk.¹⁹ Of note, of these figures, only Figure 2 was reviewed in the proposed LCD (full text and supplement are attached to this letter). The assertion in the draft LCD that it is unclear how the test adds value above the clinical, pathological and staging information is simply unfounded in the face of the published clinical validity data.

As shown in Figure 2 of Ibrahim, et al., the 40-GEP significantly stratifies risk within NCCN high-risk and very high-risk patient groups, identifying patients with lower and higher risk of metastasis compared to the baseline level of risk in each subpopulation. Of note, the only discussion of this data in the proposed LCD is to point out that NCCN high risk patients with a Class 2B had a metastasis rate of 37.5% compared to NCCN very high risk patients with a Class 2B result who had a metastasis rate of 60.0%, identifying the importance of underlying risk factors. While it is true that underlying clinical and pathologic risk factors maintain predictive value for a patients’ risk of regional or distant metastasis the analysis in the proposed LCD does not recognize the added value of 40-GEP results within each NCCN risk group. Again, it must be emphasized that the 40-GEP adds information to staging and does not compete with existing staging systems or risk factor based assessments. A more valid choice of data to describe would be the rates of metastasis stratified by the 40-GEP within each subset (see Figure 2 from Ibrahim et al.). Within the NCCN high-risk subset, the overall cohort had a 9.8% rate of metastasis that was reduced to 4.1% with a Class 1 result, increased to 15.7% with a Class 2A result, and increased to 37.5% with a Class 2B result. Of note, NCCN high-risk patients with a Class 2B result had metastasis rate equivalent to the overall metastasis rate in NCCN very high-risk patients. Similarly, the overall metastatic risk in NCCN very high-risk patients was 23.0%, and the test identified rate of metastasis significantly reduced to 11.9% in patients with Class 1 results, similar to the overall cohort at 25.3% in Class 2A results and increased to 60.0% with a Class 2B results. Again, it is interesting and important to note that patients with NCCN very high risk and a Class 1 result had a rate of metastasis similar to the overall NCCN high risk cohort. This indicates that the 40-GEP results identify a level of risk in the context of NCCN risk groups that would qualify as reclassification.

The same, significant risk stratification by 40-GEP within the different AJCC stages (Ibrahim, et al., Supplemental Figure 2 and Supplemental Table 2). It is important to note that AJCCv8 only applies to tumors in the head and neck, and not all cSCC tumors, which is a limitation of this staging system. Even so, Class 1 results in each AJCCv8 substage identify risk equivalent with one stage lower, and patients with Class 2B results in each AJCCv8 substage are generally associated with risk one stage higher, indicating the differentiation of risk is enough to change management in the context of how risk is assessed and patients are managed today. This data was not reviewed in the proposed LCD.

Finally, this is also true when evaluating 40-GEP class results within BWH stages, as presented at the open meeting and published in the Supplemental Figure 2 and Supplemental Table 2 in Ibrahim, et al. 2021. When evaluating BWH T1 patients (which because of 40-GEP testing criteria, all have at least one NCCN high-risk or very high-risk factor) this group as a whole had a metastasis rate of 9.5% (Table 4). For the subset of BWH T1 stage patients who had a Class 1 result, only 3.3% of patients experienced an event, roughly one-third of the rate in all T1 patients. This effectively increases the NPV from 91% with stage alone to 96% with 40-GEP and BWH stage combined.

On the other hand, a Class 2 result was associated with a metastasis rate of 18.2% in the setting of BWH T1 stage disease. This means that nearly 1 in 5 patients with a Class 2 result experience metastasis, which is similar to the rate of metastasis in BWH T2a patients. Thus, T1 patients with a Class 2 result should be treated more like BWH T2a patients based on this validation data. Notably, in the BWH T1 patients, the 40-GEP test correctly identified 80% of metastases that were missed by BWH staging (15/19) with a Class 2 result. Finally, while the Class 2B result is relatively rare in this subset, 1 in 20 patients with BWH T1 disease received a Class 2B result, which is associated with a 4-fold increased risk of metastasis. Of note, this PPV in BWH T1 patients with Class 2B result is similar to the overall rate of metastasis in BWH T3 patients; this is a level of risk that warrants action based on all other risk stratification methods and physician consensus.^5,27,28

Most of the proposed LCD focuses discussion on patients at low risk and high risk for metastasis, but there is very little discussion regarding use of the test in BWH T2a patients, which represent an intermediate risk of metastasis and comprise ~40% of patients tested. Patients with BWH T2a tumors had an overall metastasis rate of 15% (Table 5); however, with a Class 1 result, the metastasis rate decreased to 9%, similar to the overall metastasis rate of patients with BWH T1 tumors, and thus, a level of risk that supports de-escalation of management. On the other hand, Class 2 results are associated with a PPV of 20% and correctly identifies more than 70% of patients within this stage that metastasize. A Class 2B result is associated with a 43% risk of metastasis, which is comparable to the overall rate of metastasis in patients with BWH T3 tumors with clearly actionable risk. Thus, there is important clinical actionability of the 40- GEP test result in patients with BWH T2a disease.

Finally, a BWH T2b patient generally has a risk of metastasis considered actionable by clinician surveys and consensus statements for surveillance imaging and use of ART. Overall, BWH T2b tumors were associated with a 33% risk of metastasis (Table 6); however, a Class 1 result reduced this by half and resulted in an NPV that is similar to the overall metastasis rate in T2a patients. In T2b patients with Class 2 results, 43% of patients experienced metastasis, and a Class 2 result captured more than 70% of metastases within T2b patients. Importantly, the Class 2B result in this population is associated with an 80% risk of metastasis. To put it in perspective, a PPV of 80% is simply unheard of in the field of prognostic indicators; this means that 80% of patients with a T2b tumor who receive a Class 2B result eventually metastasize. These patients clearly qualify for the most aggressive therapeutic options available. Also noteworthy is that the Class 2B result is not rare in this subpopulation; 11% of patients with T2b tumors in this cohort received a Class 2B result, meaning that 1 in 9 patients are identified with an 80% risk of disease progression, which is hugely impactful in the clinic.

This data is critically important, as there is substantial discussion in the proposed LCD of the potential for a Class 1 result to be misinterpreted in patients with higher stage disease as inappropriately low risk and inappropriately de-escalated. This published data makes it clear that the Class 1 identifies a level of risk equivalent to the overall population of T2a patients, for whom more intensive management decisions, such as ART, are not typically recommended.

Importantly, this demonstrates that de-escalation of management in a T2b patient with a Class 1 result to a management strategy that aligns with BWH T2a level risk is appropriate and not associated with harm, and is consistent with published clinical utility data. The clinical management algorithm developed by Farberg, et al. 2020, confirms that clinicians are able to interpret the DecisionDx-SCC test result to inform consistent and accurate management decisions and the Class 1 result in T2b patients is associated with an 18.8% risk for regional/distant metastasis, with patients routed to the “Moderate intensity” management strategy (see Figure 2 in Farberg et al. 2020).³¹ Use of this management algorithm is demonstrated in Hooper, et al. 2022, where the intensity of post-GEP management plans recommended for a T2b patient with Class 1 result (Case 1) is similar to baseline management recommendations for BWH T2a patients (Cases 2 and 3), but are not as low as the baseline management recommendations of BWH T1 patients (Cases 4-6), demonstrating appropriate de-escalation that aligns with published metastatic rates (Hooper, et al. 2022, Figure 3).³³ Thus, de-escalation of T2b patients with Class 1 results has been demonstrated to be risk-aligned and consistent with current management of patients with similar risk, therefore, cannot be deduced to have patient harm.

Incorporating 40-GEP results with staging allows refinement of risk stratification within stages. These important data are not reviewed or discussed in the proposed LCD. This is a critical flaw in the proposed LCD that invalidates the conclusions made; missing this important information masks how the published clinical validity demonstrates the ability of the test to improve the accuracy of risk stratification in the context of existing risk stratification approaches which then informs risk aligned management strategies already articulated in guidelines. Again, when the full range of clinical validity data is reviewed, it is not possible to justify the statement in the proposed LCD that it is unclear how the test adds value to clinical and pathological information currently obtained for patients diagnosed with cSCC. Not only does the data show that 40-GEP and clinicopathologic factors should be used together, expert clinicians that treat cSCC patients every day have published their clinical workflows that do just this, which will be reviewed in a later section.

VI. Correlation of GEP results with stages or other risk factors is not a statistically or scientifically supported analysis to assess independent value from test results

It cannot be stressed enough that multivariable analysis is the appropriate statistical method to determine if new information adds value to existing information. In addition, consideration of the degree of risk stratification in combination with staging to determine if actionable changes in risk are detected informs the utility of new risk information. Instead of relying on either of these standard analytical tools, the proposed LCD relies on anecdotal comparisons of scenarios where clinical pathological risk information and GEP risk information agree without consideration of implications of GEP results on metastatic risk. There are many instances in the proposed LCD where anecdotal comparison of concordance between 40- GEP results and clinical or pathological risk factors that appears to infer reduced value of the test, again with no review of the associated metastatic risk in these patients. This is not a scientifically valid way to evaluate test performance and leads to false interpretations and conclusions. The suggestion that the alignment with clinical features duplicates or discounts the prognostic information provided by 40-GEP is inappropriate when a robust multivariable analysis demonstrates added statistically significant information. We will address each proposed LCD statement below:

“Class 2A and 2B results were more likely to be found in samples from patients with a greater number of risk factors”

The alignment of Class 2A and Class 2B results in patients with a greater number of risk factors is expected and a common observation in molecular prognostic testing. For example, this demonstrated and discussed in the validation publication from OncotypeDx Breast Cancer, a widely used, well validated GEP test in breast cancer.⁴⁶ In fact, it would be concerning if validation studies demonstrated a dramatic discordance with previously identified and validated risk factors, as this would indicate that one of the prognostic factors is incorrect. The appropriate statistical method to evaluate if concordance between risk factors and risk results represents duplication of information is a multivariable Cox analysis and to look for areas of discordance (i.e., risk stratification in patients where GEP results and clinical and pathological risk assessment differ). Significance in the multivariable model demonstrates that the 40-GEP test is finding new and additive risk information, and as discussed above, the GEP adds statistically and clinically significant stratification of risk in the context of all clinically used risk assessment methods. This should be corrected in the final LCD.

“The Class 2B result is rare and occurs more often (though not exclusively) in patients with two or more risk factors, who would already likely be classified as higher risk by existing tools; only 23 patients received a Class 2B result (12 metastatic) 18 of whom had 2 or more risk factors.”

While a Class 2B result may be relatively rare, it is associated with the highest hazard ratio of any risk factor in any multivariable analysis and has a significant impact on the predicted metastatic rate of patients, even those with higher staged disease. Of note, a Class 2B result is not rare in all patient subsets; a Class 2B was result was received in 11% of BWH T2b patients in the Ibrahim, et al. study and signifies almost double the risk of metastasis as staging information alone.¹⁹ Moreover, while 18/23 patients with Class 2B results in Ibrahim, et al. had 2 or more risk factors, counting of risk factors without assessing their relative importance is the lowest resolution of risk assessment for clinical and pathological-associated risk. When looking closer at the 23 patients with Class 2B results, the Class 2B results are found in 9 T1 patients, 7 T2a patients, 5 T2b patients, and 2 T3 patients, demonstrating no real association of the Class 2B result with risk of metastasis identified by clinical and pathological risk assessment. This makes the statement that the patients were already higher risk by other tools a false representation of the data. This should be corrected in the final LCD.

“The majority of Class 1 results were identified in samples with 1-2 risk factors, confirmative of low risk of metastasis”

The statement, “The majority of Class 1 results were identified in samples with 1-2 risk factors, confirmative of a low risk of metastasis” is also false. In order to qualify for testing, all tumors must have at least 1 high risk factor and therefor have an elevated baseline risk of metastasis in order to be eligible for testing; Castle does not accept orders in patients without at elevated risk of metastasis. In addition, Class 1 results confer a 50% decrease in metastatic risk in every BWH T stage and therefore identify a much lower risk of metastasis. Thus, the test result is not confirmative of a low risk of metastasis and this statement must be removed from the final LCD.

“The majority of patients tested with the 40-GEP receive a Class 1 or 2A result, which is comparable in regards to predictive value to current staging criteria”

While it is true that the PPV of a Class 2A/2B result is similar to the PPV achieved by staging, this statement obscures the fact that the Class 2A/2B result identify independent risk information and can be used to improve the PPV when used in conjunction with staging, as detailed in prior sections. The final LCD should be modified to reflect this statistical independence of the 40-GEP and intended use in conjunction with current staging criteria.

“It would be prudent to further explore the patients with Class 2B results who were not identified as very high-risk according to all available staging criteria and risk factor analysis, but for whom a Class 2B result would enable them to benefit from more intense management. The data is not available outside a handful of clinical anecdotes.”

The suggestion in the proposed LCD that the value of the Class 2B results has not been evaluated in the context of patients not classified as very high risk by any staging system is incorrect and this data has been published outside of clinical anecdotes. Considering the NCCN high risk cohort published in Ibrahim et al., these patients are clearly not identified as very high risk by NCCN. All BWH risk factors are very high risk by NCCN, with the sole exception of tumors with a diameter of 2-4 cm, therefore, the highest BWH stage that can be included in the NCCN high risk patient population is BWH T2a tumors, which are binned as low risk in binary classification of risk by BWH. Considering AJCCv8, AJCC T1 and T2 are characterized by tumor diameter of < 4 cm and binned as low risk in binary classification by AJCC, whereas AJCC T3 and T4 have the presence of NCCN very high-risk factors. Therefore, evaluation of the subpopulation of patients that are NCCN high risk represents a subset in which no patients are classified as very high risk by any staging system (i.e., this group formally excludes BWH T2b and T3 and AJCC T3 and T4 patients). In NCCN high-risk subset, as discussed above and published in Figure 2 of the Ibrahim et al. manuscript, the 40-GEP shows significant risk stratification, and identifies patients with a Class 1 result that still have half the risk of metastasis as the overall cohort, and patients with a Class 2B that have a similar risk of metastasis as the NCCN very high-risk cohort, and high staged BWH or AJCC disease.¹⁹ This data should be thoroughly reviewed in the final LCD and this suggestion should be removed or corrected.

In summary, use of standard analytical tools to evaluate the additional prognostic value of 40-GEP and the ability to stratify in the context of other risk factors find that the test adds significant, statistically independent information that cannot be derived from clinical and pathological factors. In addition, multivariable analyses show the risk associated with a Class 2A and Class 2B result can be quantified relative to the risk of well-known risk factors and combined to more accurately assess comprehensive metastatic risk. Finally, thorough analysis of subsets in the Ibrahim, et al. publication and data supplement, exhaustively demonstrates how the test can be used in the context of all risk assessment systems to further stratify and improve metastatic risk prediction.

VII. 40-GEP has demonstrated consistent incorporation into patient management plans in alignment with national guidelines recommendations

According to the proposed LCD, “The potential clinical utility of the 40-GEP lies in its incorporation into existing risk-assessment frameworks in order to improve prognostic value and facilitate individualized risk assessment, treatment and follow-up". This is exactly how physicians have been utilizing 40-GEP since the test became clinically available in 2020. Publications from Teplitz, et al., Au et al., Litchman et al., Farberg et al., Arron et al., Hooper, et al., and Singh et al. have all consistently demonstrated the integration of 40- GEP with clinicopathologic-based risk staging to direct patient management.^27,29–34 Of these publications, only Au et al. was discussed and evaluated in its entirety in the proposed LCD. Arron et al. and Hooper et al. are reviewed but only partially summarized, with major datasets and conclusions omitted. Farberg et al. is reviewed but dismissed based on faulty assumptions and is omitted from analysis of evidence or rationale. Teplitz et al., Litchman et al., and Singh et al., are not reviewed at all. All of these publications show that (i) the test is used by clinicians to change patient management decisions, (ii) the 40-GEP result is evaluated WITH clinical and pathological risk information to guide management decisions, and (iii) treatment pathways impacted by 40-GEP are consistent from study to study and include follow up frequency, use of nodal imaging, use of surveillance imaging, use of SLNB, and use of ART. All omitted publications, which directly demonstrate consistent clinical utility by clinicians with integration of other risk factor information, should be reviewed and included in the proposed LCD and full-text articles with accompany these comments.

In Litchman, et al., two vignettes of patients with BWH T2a tumors were assessed for pre-40-GEP and post- GEP management plans, representing a clinicopathologic gray area of risk.²⁷ In this study, a Class 1 result led to changes in recommendations for management, including a decrease in follow-up intensity, avoidance of SLNB, reduction in nodal imaging, and avoidance of adjuvant radiation therapy. In patients with Class 2B results, patients were recommended to have increased follow-up, recommendation for SLNB in about 60% of physicians, use of nodal ultrasound or CT scan 1x per year in about 60% of physicians, and recommendation for adjuvant radiation therapy in about 60% of physicians. These management plan changes reflect integration of the 40-GEP results to guide changes in management modalities by risk, as outlined in NCCN guidelines.

The Hooper et al. publication demonstrates that clinicians correctly interpret 40-GEP in the context of clinicopathologic features when determining treatment plans across a number of specific treatment modalities.³³ In this study, 34 clinicians with substantial clinical experience with the 40-GEP, defined as having ordered 10 times or more since clinical availability, were presented with 6 real world patients and asked for management plans pre- and post- receipt of 40-GEP results. Consistent with the findings of Litchman, et al., the study showed physicians changed recommendations for SLNB, surveillance imaging, and ART. What is more, the 6 patient cases represented the spectrum of risk by BWH staging; the cases include patients with T1, T2a, and T2b risk as assessed by BWH staging. Of note, recommended management changes in response to each GEP result shifted with patient risk, demonstrating that clinicians do incorporate the baseline risk of the patient with the GEP result to inform management strategy. Second, the results showed consistent management changes when assessing patients with similar risk factor/BWH stage information, showing consistent interpretation of the combined clinical, pathological, and genomic risk from clinician to clinician. Even one step further, in Saleeby et al., in which patient management plans were compared pre- and post-test result, physicians reported a rank order of risk factors considered in the formation of management plans.³⁵ Results showed that the 40-GEP result was the most influential factor in determining the patient’s management plan, again demonstrating that the results are interpreted in the context of other risk information and the GEP results demonstrate significant clinical utility to clinicians managing these patients.

Several publications have outlined clinical algorithms and consensus statements regarding the integration of 40-GEP into clinical practice.^27,29,31,36 Every one of these publications has consistently interpreted 40- GEP results in the context of clinicopathologic-based risk stratification systems.

Three years ago, Farberg, et al. published a manuscript demonstrating how 40-GEP results can be integrated with BWH or AJCC stages, comprehensive risk of metastasis evaluated using available clinical, pathological, and genomic information, and alignment of risk of metastasis with management recommendations outlined in NCCN guidelines. While this publication is reviewed in summary of evidence, it is not interpreted in the analysis of evidence, likely due to the assertion in the proposed LCD that the clinical algorithm is “out of date” due to revisions in NCCN guidelines and addition of the NCCN “Very High Risk” category after the publication. Importantly, even with the update to the NCCN guidelines and the addition of the very high-risk group, none of the findings should be altered or invalidated in any way. While an additional risk group has been added, the management recommendations present in the 2020 version of NCCN guidelines still apply to all NCCN high-risk and very high-risk patients in the 2023 version of the guidelines. Both the high risk and very high-risk patient groups are still eligible for consideration for the management modalities outlined in the algorithm (see NCCNv1.2023 cSCC-3). Moreover, BWH staging and AJCC staging, the two clinical pathological risk assessment systems that are also integrated with GEP results in the Farberg manuscript, have not changed since publication. In sum, while it is true that the NCCN subdivided what was previously known as “high-risk” into “high-risk and very high-risk” categories, BWH and AJCC remain constant and the underlying risk assessment methodologies and available management pathways have not been changed from the time of this publication in 2020. As such, the Farberg et al. management pathway should be included in the analysis of evidence and interpreted in the conclusions of the final LCD.

In the time since the original request for coverage of 40-GEP and since the update to the NCCN guidelines, a new clinical algorithm has been published by three physicians, that again demonstrates consistent integration of 40-GEP with traditional staging systems and provides a clear management matrix. This publication, Singh et al., which was not cited or mentioned in the evidence review in the proposed LCD, contains a clinical management pathway that integrated GEP results with staging information to present a clinical algorithm outlining risk-aligned management recommendations.³⁴ Moreover, this publication demonstrated the implementation of 40-GEP in cases showing the use of a management pathway, consistent with other management pathways, and the real-world utility in real-world patients.

Thus, the statement in the proposed LCD that claims “While it has been shown that clinicians are willing to adopt the 40-GEP and to potentially act on the results, it is not clear what the recommended management changes should be and there is no outcomes data to support changes in management” is not a true statement and is not supported by literature referenced above. There is consistency in how the test results are used by treating clinicians, both in the specific management modalities impacted and in how the results are incorporated with clinical and pathological factors. Further, in the recently finalized LCD MolDX: Prognostic and Predictive Molecular Classifiers for Bladder Cancer, the value of a molecular classifier to guide multiple management decisions is described clearly. This LCD states, “For patients with bladder cancer, an array of treatment possibilities exists at all stages of disease. Clinicians must consider not only the potential treatment options but must also make an individualized risk to benefit assessment to determine how to treat a specific patient. Diagnostic tests that aid in this assessment are expected to change physician management in a way that improves patient outcomes.” This evaluation of clinical utility certainly applies to the 40-GEP, as demonstrated in detailed analyses above, and the same standard should be applied for Bladder Cancer and cSCC.

Not only have experts come to consensus on the appropriate utility of 40-GEP within clinical practice, but publications have shown that ordering physicians understand the utility of the test and are willing to - and certainly do - utilize 40-GEP in their clinical practices across the country. First, and importantly, over 3,600 physicians have ordered the 40-GEP test for decision making with their high risk cSCC patients. The American Association of Medical Colleges (AAMC) reports there are just over 12,500 active dermatologists in the US (https://www.aamc.org/data-reports/workforce/data/active-physicians-us-doctor-medicine-us- md-degree-specialty-2019). This means that more than 20% of dermatologists have evaluated the validity and utility of the DecisionDx-SCC test and incorporated the test into their practice.

VIII. 40-GEP identifies ART eligible patients with very high biological risk of metastasis that benefit from ART and similarly identifies a significant population of ART eligible patients with low biological risk of metastasis that may avoid ART

As noted earlier, the literature and all guidelines and societies support the use of ART in patients diagnosed with localized high risk cSCC. And more recently, the Ruiz 2022 publication demonstrated that while there is a 50% reduction in metastatic rates, there are patients who are high risk based upon clinicopathologic selection but are at biological low risk who receive only an ^~5% absolute benefit that does not warrant the associated cost and morbidity associated with ART.¹²

As discussed above, the 40-GEP test provides clinically significant and independent improvements to risk- stratification accuracy and when combined with staging systems improves the accuracy and clinical utility when making risk-stratified decisions within the context of current treatment guidelines. Separately, we questioned whether the 40-GEP can identify patients who have improved response to ART based on 40-GEP identification of high biological risk of metastasis. We have now demonstrated, as presented to CGS during the open meeting, data from our combined cohort of 954 patients with cSCC that the 40-GEP test can identify patients who are likely to have a clinically significant response to ART as well as those who do not benefit from this intervention.

As provided previously to MolDX and discussed above, the figure below (adapted from data published in Ibrahim et al.) shows the benefit of integrating BWH stage and 40-GEP result in patients with high risk but lower stage disease (Figure 4). Specifically, the 40-GEP identifies patients with lower risk T1 or T2b stage but a very high risk of metastasis by a Class 2B result that have risk of metastasis similar to the overall population risk of BWH T2b and T3 tumors, the subset of patients that demonstrated benefit of ART in the Ruiz et al. study and who are recommended for ART in surveys and consensus groups.^5,32 Additionally, the test identifies patients with BWH T2b disease that have a risk of metastasis similar to the overall population of T2a tumors, for whom ART is available, but is of uncertain benefit.

We examined our overall cohort study of 954 patients to evaluate the ability of the 40-GEP to identify patients who are eligible for ART under NCCN guidelines and would or would not benefit from ART (Figure 5). In order to control for selection bias for patients who underwent ART for those that did not undergo ART, we used the “optimal full” matching approach for a 1:1 ratio (MatchIt_4.5.3 package in R v4.3.0) for the significant clinicopathologic variables. Then, in order to use all possible matched pairs without excluding any patients, we performed x10,000 iterations of random pairs of ART treat and untreated patients.

While all of these patients are eligible to be considered for ART treatment based on guidelines, the 40- GEP Class result did indeed identify a group of patients that saw a clinically significant benefit from ART treatment. Specifically, as shown in the figure below, in the Class 2B graph on the right, ART treatment (blue curve) is associated with a marked improvement in reduced rates of the cumulative probability of metastasis compared to the patients not treated with ART (red curve). As expected, this benefit was not seen in patients that received a lower risk DecisionDx-SCC result (Class 1 and Class 2A graphs below). This aligns with published literature, including Ruiz et al. mentioned above, that ART benefits patients the most when they have an elevated risk of regional and distant metastasis.⁷

40-GEP Class 2B results identified patients more likely to respond to ART. It is striking that all of these patients are eligible to be considered for ART treatment per NCCN guidelines (since only high and very high risk NCCN patients are eligible for 40-GEP), a relatively small portion of these patients actually saw a benefit from ART. Additionally, there is steep rise in the red curve of metastatic events in the Class 2B cohort which corresponds to the rapid rate of metastasis in these patients not treated with ART– this highlights the urgency to act on the Class 2B result where most metastatic events are seen within the first two years (Figure 5).

Putting this data together, ART is known to be most effective in improvement of outcomes when directed to patients with the highest risk of metastasis, and clinical and pathologic risk factors alone are insufficient to accurately guide ART decision making. The 40-GEP provides clinically actionable risk stratification information that identifies cSCC patients in any stage with a high metastatic risk who can be considered for and who benefit from ART.

IX. Appropriate use of 40-GEP to inform ART decisions can provide significant healthcare cost savings for the Medicare Trust Fund

While ART is a primary treatment intervention in patients with high risk of metastasis, and all NCCN defined high risk and very high risk cSCC qualify for consideration for ART use, ART carries significant complication risks as well as high cost. And, as noted earlier, there is significant under-treatment and over- treatment if one relies solely on the clinicopathologic based staging systems.

In order to model healthcare cost savings for the Medicare Trust fund, we analyzed medical claims data for the use of ART for all patients with a new diagnosis of localized cSCC. The IQVIA dataset identified 731,482 patients who were diagnosed in this time period (data on file). This compares to annual incidence rates of 1,800,000 (https://www.skincancer.org/skin-cancer-information/squamous-cell-carcinoma; accessed July 22, 2023) or coverage for 41% of the overall annual incidence. In addition, 88% of patients were 65 years of age or older. Patients who received ART with a cSCC diagnosis were identified as patients who received ART within the 12 month time period of June 2021 through May 2022 and within 12 months of diagnosis (data on file). According to this dataset, ~29,600 patients in the United States Medicare population received ART for cSCC treatment during this 12-month time period. (data on file)

Following consultation with both radiation oncology and Mohs surgeon experts regarding use of specific ART modalities to specifically impact locoregional metastasis (including nodal metastasis), we limited ART modalities to traditional radiation therapy (XRT), Intensity-modulated radiation therapy (IMRT), Image- guided radiation therapy (IGRT) and Intensity-modulated proton therapy (IMPT). We then analyzed published literature for the cost for a course of treatment for each of these modalities for cSCC and/or head-neck SCC.^37-39 The table below shows the cost for a course of treatment for each of the four modalities and their relative use (data on file). These data were then used to calculate the weighted average cost of ART at $60,693. Thus, the estimated cost of ART for a single year of treatment in Medicare patients with cSCC who were treated based on clinical and pathological risk factor assessment alone was ~$1.8 billion.^37–39

Assuming distribution of Class results as demonstrated in validation publications in these patients,¹⁹ use of the 40-GEP Class 2 result to rule-in ART and use of the 40-GEP Class 1 result to avoid ART, then use of the 40-GEP could be expected to reduce the cost of ART to $897 million.

Put differently, taking the current reimbursement rate for the 40-GEP test, the net healthcare savings for 100 Medicare beneficiaries who had their ART ruled-out with a Class 1 test would be $2.2 million, a strong cost extraction benefit for any test. Again, this analysis as performed only evaluates one aspect of the multiple demonstrated clinical utilities of the DecisionDx-SCC test.

X. CMS’ Division of New Technology recently completed their review of the clinical validity and independence of the 40-GEP and came to the conclusion that the 40-GEP is a valid and independent predictor of metastasis within this high risk cSCC population

As outlined in NCCN guidelines, and as is consistent across all types of cancer treatment, management decisions for patients with cSCC are risk-aligned, based on the treating clinician’s best evaluation of their patient’s risk of poor outcomes – including regional and distant metastasis.⁴⁰ Treatment decisions are made on a patient-by-patient basis informed by risk factors and or staging. NCCN guidelines recommend consideration of frequency of follow-up, use of surveillance imaging, nodal imaging, SLNB, and ART for patients with high risk or very high risk factors, without further restrictive criteria to guide decisions. Many patients are over or under treated based on these risk factor-based assessment systems that have well- known limitations in accuracy and are based on retrospective cohorts.⁴

The comparator for molecular risk-stratifying tests should be current risk stratification systems, not improvement in patient outcomes and is also consistent with Medicare’s definition of medically reasonable and necessary criteria.⁴¹ This follows the precedent set forth by several other MolDx LCDs which cover prognostic tests in other cancers including, but not limited to:

• LCD MolDX: Prostate Cancer Genomic Classifier
• LCD MolDX: Breast Cancer Index® (BCI) Gene Expression Test
• LCD MolDX: Prognostic and Predictive Molecular Classifiers for Bladder Cancer

Each of these policies includes an identical quote which should also be applied to the DecisionDx-cSCC test:

“Numerous prior Medicare coverage decisions have considered the evidence in the hierarchical framework of Fryback and Thornbury,[⁴²] where Level 2 addresses diagnostic accuracy, sensitivity, and specificity of the test; Level 3 focuses on whether the information produces change in the physician's diagnostic thinking; Level 4 concerns the effect on the patient management plan and Level 5 measures the effect of the diagnostic information on patient outcomes. To apply this same hierarchical framework to analyze an in vitro diagnostic test, we utilized the ACCE Model Process for Evaluating Genetic Tests.[⁴³] The practical value of a diagnostic test can only be assessed by taking into account subsequent health outcomes. When a proven, well-established association or pathway is available, intermediate health outcomes may also be considered. For example, if a particular diagnostic test result can be shown to change patient management and other evidence has demonstrated that those patient management changes improve health outcomes, then those separate sources of evidence may be sufficient to demonstrate positive health outcomes from the diagnostic test.”

Importantly, ACCE criteria to evaluate molecular testing, as has been cited in MolDX LCDs for bladder, breast and prostate cancer, does not require direct improvement patient outcomes to demonstrate clinical utility that meets the Medicare requirement for medical reasonableness and necessity.^13-15 Instead, these other MolDX LCDs state that when a test impacts management strategies that are included in guidelines and known to impact patient outcomes, the indirect improvement in patient outcomes can be inferred. Thus, the improvement in risk aligned management of patients through surveillance imaging, nodal assessment, frequency of follow-up and use of adjuvant radiation therapy, consistently supported in published clinical utility studies to be impacted by 40-GEP and included in national guidelines as management options for treatment of these patients, meets the Medicare level of evidence to be considered medically reasonable and necessary.

By demonstrating improved risk assessment over current risk stratification systems, 40-GEP has demonstrated clinical utility and positive net health impact by identifying which patients need which level of treatment, which meets the analysis laid out in the quote above. The Hooper et al. and Saleeby et al. studies exhibit this exact point.^19,35 In the Hooper et al. study, physicians indicated that the DecisionDx- SCC test result could be utilized in concert with staging and clinicopathological risk factors in a variety of scenarios to change patient management in a risk-aligned manner. Further, Saleeby et al., a multi-center prospective clinical utility study reporting only Medicare-eligible patients, reported that the DecisionDx- SCC result influenced the patient’s management plan in 24% of patients which is even higher than the 18% reported change in ART management reported in the Gore et al. publication supporting the prostate cancer test covered by LCD MolDX: Prostate Cancer Genomic Classifier (listed above).^35,44 In the field of prognostic molecular diagnostics, the burden of clinical utility has consistently been to demonstrate significant risk stratification that informs proven treatment modalities either on par with or superior to current risk stratification systems. The 40-GEP has met this level of evidence. It is not ethical, causes disease state disparity, and is harmful to cSCC patients to not follow the same precedent that has been established for prognostic tests and that has been established and followed in a variety of other disease states. Table 7 below details the criteria to be considered medically reasonable and necessary, as outline in the Medicare Program Integrity Manual, and described how the test met these requirements.

XII. Summary and Conclusions

In conclusion, the final LCD should take into close consideration how high risk cSCC is currently managed in the US. Clinicians treating cSCC are experts in the management of cutaneous malignancies, adept at incorporating multiple clinical, pathological, and genomic risk factors to inform risk-aligned treatment plans. In fact, these clinicians are already responsible for combining multiple risk factors into a comprehensive risk assessment to guide patient management; addition of genomic information as another risk factor with statistical analyses integrates directly into the current management paradigm. Physicians have demonstrated this both in practice and in clinical utility studies, that test results are being used appropriately with staging information to make individual patient management decisions.^33–35

Requiring improvement in outcomes to demonstrate medical reasonableness and necessity is a new evidentiary bar that has not been required for MolDX coverage of any other prognostic GEP, setting up an inappropriate disparity in healthcare between cutaneous malignancy and other malignancies, such as breast, prostate, and bladder cancer. Recently published data shows 40-GEP can be used to guide ART decisions that have proven impact on health outcomes for patients with a high risk of regional and distant metastasis.¹² Therefore, improvement in the accuracy of risk stratification for patients with cSCC has inherent improvement in patient outcomes as this directs those with high risk of metastasis to a therapy known to improve outcomes. Even so, new data also demonstrates that 40-GEP may be used to select for patients with high enough risk of metastasis to demonstrate improvement in outcomes when treated with ART (Class 2B), and patients with low risk of metastasis and no demonstrated improvement after ART therapy, suggesting de-escalation is appropriate in these patients.

A critical limitation of the proposed LCD is that there is incomplete consideration for how high-risk SCC is currently managed in the US. Clinicians treating SCC are skin cancer experts in both academic and community practice who incorporate multiple clinical, pathological, and now genetic risk factors to inform risk-aligned treatment plans in routine patient management. Over 3,600 clinicians experienced in treating patients with high-risk SCC have determined DecisionDx-SCC to be medically reasonable and necessary for more than 7,000 patients to inform their management decisions.⁹ Further, ART is associated with a marked reduction in metastasis rate for patients with a Class 2B DecisionDx-SCC result (see Figure 5 above); this aligns with published literature that ART benefits patients the most when they have an elevated risk of regional and distant metastasis.⁷

Physicians have demonstrated both in practice and in clinical utility studies, that test results are being used appropriately and consistently with risk factor and/or staging information to make risk-aligned individual patient management decisions.^9–11 Specifically, DecisionDx-SCC can be used to guide ART decisions⁹ that have proven impact on health outcomes for patients with a high risk of regional and distant metastasis.¹² Therefore, improvement in the accuracy of risk stratification for patients with SCC and one or more risk factors has inherent improvement in patient outcomes as this directs those with high risk of metastasis to a therapy known to improve outcomes. Finally, DecisionDx-SCC meets the standard of clinical utility established in other MolDX LCDs in Prostate and Breast cancers.^{8–11,13–15}

As an additional note, several publications are referred to in the written comments here that were missed in the initial evidence review performed in preparation of the proposed LCD. All references are listed below and full texts are attached to this letter in a .zip file as these publication are crucial for a complete understanding of the cSCC field and the data supporting the DecisionDx-SCC test. Castle trusts that a thorough review of these publications will be completed, and this evidence will be included in the final LCD. We believe that in so doing and after review of the comments from Castle Biosciences and other stakeholders, the evidence will support reconsidering the rationale for determination and finalization of the LCD with positive coverage of DecisionDx-SCC allowing Medicare beneficiaries to have access to such a useful tool in cSCC patient management.

Many of the analyses in the proposed LCD focus on using number of risk factors as a measure of overall metastatic risk (only 1 risk factor, and 2 or more risk factors) as published in Ibrahim et al. 2021. At the time of the validations, given that staging is not routinely incorporated, this data was provided in order to provide a framework for evaluating prognostic value of 40-GEP for clinicians who do not use BWH or AJCC staging. However, the proposed LCD critically highlights “the importance of the complete clinical picture beyond the absolute number of risk factors presented in patient reports.” As an example, having 2 or more risk factors does not necessarily connote the highest level of risk; it is possible to have BWH T1, T2a and T2b patients that are classified as having 2 or more risk factors, depending on what the risk factors are. Castle agrees with the authors of the proposed LCD that the complete clinical picture is important when interpreting 40-GEP results in the context of the known risk factors for a patient – in fact, much of this letter is focused on clinical validity and clinical utility data that demonstrates consistent and accurate risk aligned management decisions when using clinical, histopathological and now genomic information. As a result of this reporting feedback from MolDX, we are committed to updating our patient report at the time of publication of the combined performance and validation cohort of nearly 900 patients and will update this information to provide improved clinical, pathological, and staging context as suggested in the proposed LCD. Of note, this cohort is referenced in this letter from a recent “Late breaking abstract” presented at the 2023 American Academy of Dermatology national meeting.²²

Finally, we believe the coverage criteria that has been proposed, which excludes BWH T3 patients (i.e., patients with all four of the following: diameter of ≥ 2 cm, poor differentiation, perineural invasion of ≥0.1 mm, and invasion beyond the subcutaneous fat or the presence of bone invasion) and patients with lymphovascular invasion remains appropriate to avoid testing in patients with limited clinical utility due to a high predicted rate of metastasis by clinical and pathological factors alone. In addition, due to requests for additional validation data in rare subsets that cannot be provided, it may be appropriate to add exclusion of testing in patients with desmoplasia or in a wound with chronic scarring until additional data is available to support validity of testing in those patient subsets. Again, it is not ethical to withhold the test from the majority of patients, in which the test has achieved robust validation, in order to verify performance in rare indications.

We look forward to a collaborative dialogue moving forward, which is critical to the medical policy development process, and welcome questions and discussion regarding the large amounts of data provided here to answer queries raised in the proposed LCD. Castle Biosciences strongly urges to on this proposed LCD to rectify the missing or inappropriate analyses and misinterpretations made here to finalize an LCD that will allow for your beneficiaries to have access to such a useful tool in high risk SCC patient management. The 40-GEP addresses an important unmet clinical need, meets the level of evidence requirement by Medicare Program Integrity Manual (MPIM) and the standards set for other MolDX LCDs for molecular prognostic testing in prostate, bladder and breast cancers to be considered reasonable and necessary.

Figures, tables, and references were provided for review.

Thank you for your thoughtful comment. The policy language has been modified to incorporate many suggested changes and this document is aimed to address the above topics in a point-by-point manner.

1. cSCC is a complicated disease state with signifcant under and over staging due to poor accuracy and low utility of staging systems
2. 40-GEP has been robustly validated in cohorts that are representative of the intended use population
3. Metastasis is the important outcome for Medicare beneficiaries
4. 40-GEP adds statistically significant, independent quantifiable risk information to be used in conjunction with staging. Meaning multiple studies have confirmed the definitive value above the combination of available clinical, pathological and staging criteria
5. 40-GEP significantly stratifies risk of metastasis within stages and can be used within the context of other risk stratification tools to improve metastatic risk assessment
6. Correlation of GEP results with stages or other risk factors is not a statistically or scientifically supported analysis to assess independent value from test results
7. 40-GEP has demonstrated consistent incorporation into patient management plans in alignment with national guidelines recommendations
8. 40-GEP identifies ART eligible patients with very high biological risk of metastasis that benefit from ART and similarly identifies a significant population of ART eligible patients with low biological risk of metastasis that may avoid ART
9. Appropriate use of 40-GEP to inform ART decisions can provide significant healthcare cost savings for the Medicare Trust Fund
10. CMS’ Division of New Technology recently completed their review of the clinical validity and independence of the 40-GEP and concluded that the 40-GEP is a valid and independent predictor of metastasis within this high risk cSCC population
11. 40-GEP has met criteria for medical reasonableness and necessity as defined by the Medicare Program Integrity Manual and the is consistent with similar tests in other disease states

Points addressed:

I. cSCC is a complicated disease state with significant under and over staging due to poor accuracy and low utility of staging systems.

Comments on the LCD state that cSCC is a complicated disease state with significant under and over staging due to poor accuracy and low utility of staging systems. The overall prognosis for patients with cSCC is excellent,¹ with nodal metastasis developing in 1.9 to 5.2% of cases with an overall mortality of 1.5-3.4%.¹ Nevertheless, patients with metastatic disease are more likely to experience poor outcomes and it is critical to identify these patients for appropriate management. ¹ Several staging systems and guidelines exist and have been reviewed in the LCD; namely, the National Comprehensive Cancer Network (NCCN) Guidelines, the American Joint Committee on Cancer 8th edition (AJCC8) staging system for Head and Neck Tumors, and the Brigham and Women’s Hospital (BWH) tumor classification system.

Given limited positive predictive value (PPV) of current risk stratification approaches in identification of patients at highest risk of poor outcomes such as metastasis, there is value in improved risk stratification. However, to demonstrate clinical utility in risk stratification, a new biomarker must perform with added risk stratification (clinical validity) to currently available risk stratification measures (i.e., the combination of staging and clinical plus pathological factors) within a guideline-based standard of care therapeutic framework (i.e., NCCN) to meaningfully reclassify patients with respect to clinical decision thresholds in a manner that improves patient outcomes. To date, this has not been demonstrated for the 40-GEP.

While the 40-GEP has demonstrated the independent ability to risk-stratify cSCC patients according to their risk of metastasis^2,3, the relatively poor performance of existing staging systems (BWH and AJCC8) in comparison to the 40-GEP demonstrated in the clinical validation by Ibrahim et al is not in line with independent studies that have evaluated performance of BWH and AJCC8 staging. The sensitivity of BWH and AJCC8 staging noted in Ibrahim et al (30.2% for BWH T2b/T3 and 38.1% AJCC8 T3/T4) is significantly lower than in other publications, as noted in the LCD. Ibrahim et al notes the 40-GEP to have a sensitivity of 77.8% for a Class 2 result and 19.0% for a Class 2B result. A comparison of performance of BWH and AJCC8 staging across 680 head and neck SCCs by Ruiz et al describes a much higher sensitivity/specificity of 0.78/0.85 for AJCC8 and 0.73/0.93 for BWH staging.⁴ The cohort described in Ibrahim et al is an expansion of the cohort validated in Wysong et al, which acknowledges that the percentage of metastases occurring in low T stages was higher than previously reported. ^2,3

Furthermore, Ruiz et al demonstrate homogeneity (outcome similarity within tumor class) and monotonicity (outcome worsening with increasing tumor class) for BWH and AJCC8, and similar discriminative ability between the two systems, as shown by a c-statistic of 0.86 and 0.81, respectively. ⁴ For tumors staged by BWH, the following 10-year cumulative incidence (CIN) of nodal metastasis was observed by stage: T1 (CIN 0.2%), T2a (CIN 5%), T2b (CIN 24%), and T3 (CIN 60%).⁴ For tumors staged by AJCC8, the 10-year CIN of nodal metastasis according to stage was: T1 (0.4%), T2 (12.2%), T3 (14.1%), and T4a/T4b (42.6%), respectively.⁴ In comparison, 40-GEP performance data presented in the clinical validation by Ibrahim et al demonstrated a metastatic rate of 6.6% for Class 1, 20.0% for Class 2A, and 52.2% for Class 2B results in a cohort composed of an intended use population consisting of NCCN high and very high-risk patients.³ Ibrahim et al found the specificity of BWH T2b/T3 and AJCC8 T3/T4 to be 89.6% and 84.3%, respectively. ³ Whereas, the specificity for any Class 2 result was 55.5% and 96.9% for a Class 2B result.³ While Ibrahim et al depicts the PPV of a Class 2B 40-GEP result to be higher (52.2%), it is directly compared to the PPV of binary stages BWH T2b/T3 (33.9%) and AJCC8 T3/T4 (30%). The PPV for high stage BWH and AJCC8 is not separated by T stage for a more direct comparison to the expanded, three stage 40-GEP. Along these lines, the PPV of the 40-GEP using binary classification (Class 2 vs Class 1) notes any Class 2 result having a PPV of 23.6%, which is comparable to AJCC8 and BWH “high stage” PPV. Ibrahim et al describe the results of Multivariate Cox Regression Analyses, wherein 40-GEP classes are included in separate models alongside BWH T-stage and AJCC8 T-stage.³ While the 40-GEP retains statistical significance alongside either BWH or AJCC8 staging, the respective hazard ratios for each stage/class presented in Supplemental Table 3 of Ibrahim et al should be validated within a larger cohort, as for example, BWH T3 stage is not significant in terms of risk for metastasis in the multivariate model, whereas BWH T2b is significant with a hazard ratio of 3.21, indicative of a potential lack of statistical power.

Overall, these performance metrics do not support claims regarding overwhelmingly poor accuracy of available staging systems when compared to the 40-GEP. While the 40-GEP significantly stratifies tumors according to risk of metastasis, its clinical validity and utility must be assessed in a comprehensive clinical context by demonstrating the ability to improve the standard of care and to meaningfully change clinical management to eventually improve patient outcomes. Lack of physician utilization of management guidelines, such as those published by NCCN or staging, does not in itself render a biomarker reasonable and necessary if the biomarker has not demonstrated clinical validity and clinical utility that results in improvement of the standard of care.

II. 40-GEP has been robustly validated in cohorts that are representative of the intended use population.

III. Metastasis is the important outcome for Medicare beneficiaries.

The intended use population for the 40-GEP consists of cSCC patients with one or more high risk factors, aligning with the NCCN high risk and very high-risk groups. Recent validation of NCCN risk stratification in 10,196 tumors from 8727 patients demonstrated a 0.5% incidence of nodal metastasis and 0.1% incidence of distant metastasis in the NCCN high-risk category.⁵ The high-risk group showed a greater risk of local recurrence (1.5%), whereas the very high-risk group demonstrated a greater risk of local recurrence (9.4%), nodal metastasis (7.3%), and distant metastasis (3.9%).⁵ In patients treated with Mohs micrographic surgery (MMS), which is the preferred treatment modality for very high-risk tumors and a primary treatment option for high-risk tumors, these rates are significantly lower.⁵ Risk reduction through MMS is also supported by a study by Soleymani et al who reported the lowest rates of poor outcomes compared to previously published high-risk cohorts in high-risk cSCCs treated with MMS alone. ⁶

The relatively low underlying rate of metastasis in the NCCN high risk group questions the clinical utility of the 40-GEP in NCCN high risk patients, as hazard ratios for metastasis of 2.41 (for Class 2A result) to 6.22 (Class 2B result) do not demonstrate clear clinical utility in terms of significant changes in clinical management in a risk category with a baseline metastatic rate of <1%. Notably, NCCN guidelines state that the high-risk group is at increased risk of local recurrence and the very high-risk group is at increased risk of both local recurrence and metastasis.⁷

While comments on the LCD state that the 40-GEP has been robustly validated in cohorts that are representative of the intended use population, metastasis rates from the NCCN validation by Stevens et al are significantly lower than metastasis rates presented in the cohort used to develop and validate the 40-GEP. The clinical validation of the 40-GEP in Ibrahim et al demonstrates a 9.8% rate of metastasis in the NCCN high-risk group compared with a 0.5% incidence of nodal metastasis and 0.1% incidence of distant metastasis in the NCCN high-risk category in the NCCN validation conducted by Stevens et al.^3,5 Ibrahim et al shows 23.0% metastasis in the NCCN very high-risk group, compared with 7.3% cumulative incidence of nodal metastasis and 3.9% cumulative incidence of distant metastasis in the NCCN validation conducted by Stevens et al. ^3,5 This difference in rates of metastasis in the validation cohorts for the 40-GEP and the NCCN validation questions the generalizability of the patient population that led to the development of the 40-GEP and likely necessitates PPV adjustment to account for the lower prevalence of metastasis outside of the 40-GEP validation. Of note, the prospective study conducted by Saleeby et al reported one patient with metastasis out of 90 (1.1%) NCCN high and very high-risk patients, although the study was not complete at time of publication.⁸ Given that 85% of the cSCC metastases occur within two years and 93% within three years and this cohort has only been followed up for a median of 22.5 to 41.7 weeks, follow up data is expected.^8,9

As noted in the LCD, cSCC incidence is highest in non-Hispanic white populations, greater in males compared to females, and increases with age for a mean age at diagnosis of approximately 71 years. Ultraviolet (UV) radiation due to cumulative and chronic sun exposure is the most important environmental risk factor, occurring more frequently in those with less pigmented skin types. In individuals with highly pigmented skin, cSCC is more frequently found on non-sun exposed areas and is often associated with chronic inflammation, chronic wounds, or scarring. cSCC of this etiology is more difficult to treat and exhibits a higher recurrence risk. Comments received agree that test performance in persons of color is an important area of research. While cSCC is rare in persons of color, it predominantly arises in association with scarring or chronic wounds, which are associated with higher risk of metastasis than other cSCC etiologies and validation studies are expected to be inclusive and demonstrate test performance in these patients to avoid disparities in care.

IV. 40-GEP adds statistically significant, independent quantifiable risk information to be used in conjunction with staging. Meaning multiple studies have confirmed the definitive value above the combination of available clinical, pathological and staging criteria.

V. 40-GEP significantly stratifies risk of metastasis within stages and can be used within the context of other risk stratification tools to improve metastatic risk assessment.

Comments on the prospective LCD make it clear that the 40-GEP should be used and evaluated in combination with current risk assessment tools (to see their added improvement). NCCN guidelines direct risk assessment after biopsy into either the low, high, or very high-risk groups and refer to other staging systems, such as AJCC8, as additional resources of reference. The American Academy of Dermatology recommends tumor stratification of localized cSCC using NCCN guidelines to provide practical clinical guidance for management and follow up and demonstrates preference for BWH in terms of the most accurate prognostication tool in localized cSCC.¹⁰

The analyses presented in 40-GEP publications to date have not demonstrated clinical validity and clinical utility in this clinical context, as the multivariate analysis with staging was not conducted within the NCCN high-risk and very-high risk groups separately, but rather in the entire intended use population which includes both high-risk (largely stage 1 tumors with lower risk for metastasis and higher risk of local recurrence) and very high-risk patients (generally higher stage with greater risk of metastasis) and so the benefit of NCCN risk stratification is lost. Given the low metastasis rate in the NCCN high risk category and questionable clinical utility, the clinical validity and clinical utility of the 40-GEP in addition to staging remains to be demonstrated in the NCCN very high-risk group. Importantly, any added risk stratification (clinical validity) within staging must demonstrate clinical utility with respect to clinical-decision thresholds within the framework of national guidelines.

Submitted comments state that 40-GEP test results are to be used with all available information for a comprehensive, clinical, pathological, and genomic assessment of risk and clarify that the quantitative risk associated with each prognostic variable or staging system is multiplicative when more than one independent high-risk feature is detected for a particular tumor. For example, a patient with a BWH T1 or T2a cSCC AND a Class 2A 40-GEP test result has a likelihood of progression that is 2.8 times that of BWH T1 or T2a cSCC AND a Class 1 test result. Given that the baseline metastatic risk for most BWH T1 patients is <1% and most would not be considered NCCN very-high risk for management,^4,5 there is questionable clinical utility in terms of management changes for patients with <1% vs <3% risk of metastasis. These patients would most likely experience reassurance with their current management plan.

The impact of a Class 2B result in lower stage patients confers an increased metastatic risk of 4.7-fold, 4.2-fold, and 5.2-fold in BWH T1, AJCC T1, and NCCN high-risk patients, respectively. Again, with low baseline metastasis risk in NCCN high-risk patients who are staged as BWH T1 and AJCC T1 (<1% in Ruiz et al 2019), ⁴ a Class 2B result may not lead to clinically impactful management changes, given the above-mentioned hazard ratios and resultant overall risk.

The incidence of metastasis in BWH T2a tumors was found to be 5% in Ruiz et al, ⁴ which is lower than the 15% metastasis event rate (25/164 tumors) in Table 5 of the submitted Comment. In the cohort of T2a tumors presented in Table 5, T2a/Class 1 tumors account for 46% of the T2a cohort (75/164) with a metastasis event rate of 9% (7/75); T2a/Class 2A tumors account for 54% of the T2a cohort with a metastasis event rate of 18% of (15/82 tumors). Finally, T2a/Class 2B tumors account for 4.3% of the T2a cohort (7/164) with a metastasis event rate of 43% rate (3/7 tumors). The clinical impact of management changes between a 15% (T2a baseline pre-40-GEP), 9% (T2a/Class 1), 18% (T2a/Class 2A) metastasis event rate is not clear (all metastasis rates are below the 20% actionability risk threshold outlined in Singh et al) ¹¹. The submitted comment presents unpublished data suggesting that Class 2B results are associated with marked reduction in the cumulative probability of metastasis in patients treated with ART compared to the patients not treated with ART. This benefit was not seen in patients that received a lower risk 40-GEP result (Class 1 and Class 2A graphs). This indicates potential value of the T2a/Class 2B result obtained in seven patients (4.3% of the T2a tumors) with a higher rate of metastasis (43% event rate) with respect to selection for ART. Nevertheless, 4 of the 7 tumors with a Class 2B result did not metastasize and there is potential for over-treatment, underscoring the importance of multidisciplinary decision-making. It is important to note that of the 25 metastases that occurred in this cohort of T2a patients, only 3 (12%) occurred in patients with a Class 2B result. The majority (88%) of metastases in this cohort of T2a tumors occurred in tumors with Class 1 and Class 2A 40-GEP results. This is important when considering reduced intensity management plans based on 40-GEP results. Finally, NCCN risk stratification validation by Stevens et al demonstrated a split of T2a tumors into high-risk and very-high risk categories with 70.5% (814/1155) of T2a tumors categorized as NCCN high-risk (a group with <1% metastatic rate overall) and 29.5% of T2a tumors categorized as very-high risk, providing an additional element of risk stratification that was not factored into this analysis. It would be helpful to examine the additional impact of 40-GEP risk stratification within stage T2a tumors alongside other pertinent clinical and pathologic risk factors within each relevant NCCN risk group with respect to clinical decision thresholds.

Referenced publications suggest inconsistencies in management approaches with respect to the 40-GEP. Au et al describe the clinical course of two patients with cSCC, retrospectively tested with the 40-GEP to demonstrate how the test could have altered clinical decision-making, if used prospectively.¹² The first patient presented (Case 1) is a 65-year-old male with poorly differentiated cSCC staged as BWH T2a who had undergone MMS wherein subsequent analysis of the last layer of nonmarginal tissue was positive for cSCC. The patient declined any further treatment and was recurrence-free for four years; he was retrospectively found to have a Class 1 40-GEP result. This example is highlighted as a biologically less aggressive tumor (due to the retrospective 40-GEP Class 1 result) that did not recur despite incomplete surgical clearance. This case is presented as an example of a patient wherein 40-GEP results could have been used prospectively in decision-making to avoid additional treatment, such as ART. However, evidence that a patient with the above-mentioned clinical and pathological risk factors and incomplete surgical clearance should be managed less aggressively due to a Class 1 GEP result has not been provided, aside from this anecdote. In fact, NCCN guidelines recommend additional treatment (re-resection or ART) for patients with incomplete surgical clearance. This approach is also at odds with the proposed framework presented in Singh et al¹¹ with respect to ART, wherein guidance for a Class 1 result is “no action if high risk and negative margins.” Additional clinical evidence is warranted to support changes in guidelines and clinical surveys of physician decision-making such as Hooper et al¹³ and Litchman et al¹⁴ are insufficient, if the decisions are not rooted in clinical evidence.

Additional case examples of 40-GEP directed de-escalation of management include Case 1 in Hooper et al, a case study of an 81-year-old male with moderately differentiated cSCC and risk factors including invasion beyond subcutaneous fat, perineural and lymphovascular invasion.¹³ This patient would not qualify for 40-GEP testing given proposed coverage criteria due to presence of the above risk factors, which would deem him sufficiently high-risk, such that the impact of a Class 1 result would not meaningfully change management away from more intense therapies (such as ART), given that hazard ratios of all risk factors are to be combined for clinical decision-making. Hooper et al found that 50% of the surveyed clinicians would change treatment strategy for this patient with respect to ART, SLNB, and nodal assessment given the Class 1 result. This is conflicting with proposed coverage criteria and could result in under-treatment as a result of the 40-GEP (Hooper et al).

Saleeby et al describe clinician perception of metastasis risk and overall management intensity changes with the 40-GEP in a pre- and post- test management plan comparison in 58 patients in a prospective cohort.⁸ Of 58 patients, 51 (88%) received a Class 1 result and 7 (12%) received a Class 2A result. There were no Class 2B results. For 37 of the 51 patients with a Class 1 result (73%), the clinicians did not change their pre-GEP perception of < 5% risk of metastasis. Perceived risk reduction from 5-10% to < 5% occurred for 12 patients and 1 patient experienced a change in perception from 10-30% pre-test risk to < 5% post-test.⁸

The clinical utility of a risk reduction from 5-10% to < 5% is not defined and the precise changes in management around clinical decision-making thresholds given a change in perceived risk from < 5% or 5-10% to 10-30% are unclear. The majority of patients in this cohort had no change in perceived risk, or experienced minimal perceived change (i.e. < 5 % vs 5-10%). Therefore, even though the publication cites a 31% GEP-driven change in risk perception, it is unclear whether this change in risk perception led to meaningful change in management that ultimately led to improvement in patient outcomes. The study notes that the overall GEP-driven change in intensity of management was 22.4% with changes mostly between moderate and low intensity categories. It would be helpful to examine patient-level management changes as a result of the 40-GEP to better understand their clinical impact. For example, if patients experienced decreased follow-up intensity due to 40-GEP results, what was the decreased frequency for Class 1 vs Class 2A results? In patients with increased management intensity due to a Class 2A result, what was the increased frequency of follow-up, imaging, and details of additional therapy, such as ART? It is worth noting that 77.6% of patients did not have an overall GEP-driven change in intensity of management reported. While the comment received from Castle Biosciences highlights that results from Saleeby et al demonstrate that the 40-GEP result was the most influential factor in determining a patient’s management plan, there were no Class 2B results in the cohort, and the hazard ratio of a Class 2A result is equivalent to or lower than that of other clinical and pathological risk factors shown in Table 3 of the submitted Comment. Hence, it is inaccurate to conclude that the 40-GEP is the most influential factor given the guidance that a 40-GEP hazard ratio should be combined with hazard ratios of other risk factors to form a complete assessment of risk.

VI. Correlation of GEP results with stages or other risk factors is not a statistically or scientifically supported analysis to assess independent value from test results.

The correlation of GEP results with stages or other risk factors relates to the ability of the 40-GEP to reclassify patients in a clinically meaningful way when compared to standard of care staging and clinicopathological risk factors. The informative value of the biomarker needs to be assessed by the number of patients needed to test to achieve a clinically impactful change in management not derived from standard of care measures, i.e. mere reassurance with or minor changes in a clinical management plan are insufficient to demonstrate clinical utility. As noted above, preliminary results from Saleeby et al demonstrated a 22.4% of GEP-driven change in intensity of management, from low to moderate or vice versa⁸. The specifics of this change on the patient-level and the clinical impact are unclear. There were no patients with a Class 2B result in the 58 patients assessed for GEP-driven management change, highlighting that the Class 2B result is rare. This is consistent with data from Hooper et al, wherein Class 2B results represented 2.9% of overall testing. ¹³ The number of patients needed to test to achieve a significant management change due to a Class 2B result is unclear.

The potential value of the Class 2B result, due to its high PPV, likely lies in very high-risk NCCN patients with higher tumor stages to enable more precise identification of patients at the highest risk of metastasis, who might in turn benefit from therapies such as ART. The submitted comment from Castle Biosciences examines 40-GEP stratification within 46 patients with T2b stage disease, wherein the overall metastasis rate is 33% (15/46) (compared to 24% incidence of metastasis noted in Ruiz et al).⁴ Of the 15 patients with metastasis, 3 received Class 1 results (20% misclassified as lower risk), 8 received Class 2A, and 4 received Class 2B results. The metastasis event rate with each 40-GEP result was as follows: 17% (3/18) with a Class 1 result, 35% (8/27) with Class 2A result (similar to baseline T2b in this cohort) and 80% (4/5) with a Class 2B result. Class 1 and a Class 2A results in T2b staged patients would align with a “moderate” intensity management strategy in the framework presented by Farberg et al, which would recommend consideration of ART, chemotherapy, or immunotherapy for patient care.¹⁵ Thus, there is no difference in management between a Stage T2b Class 1 result (17% metastasis rate) and Stage T2b Class 2A result (35% metastasis rate) given a baseline metastasis rate of 33% in this cohort of Stage T2b patients. However, patients with Stage T2b disease and Class 2B results have a much higher incidence of metastasis (4/5 patients) and may merit a meaningful management change, i.e. “high intensity” per the framework in Farberg et al. It would be informative to examine the clinicopathological details of patients with stage T2b tumors and a Class 2B result to determine the added value in clinical decision making. As such, it would be prudent to know whether other clinical and pathological risk factors not captured in staging but that would have been part of a multidisciplinary consultation identified these patients as having an overall higher risk of metastasis compared to other stage T2b tumors, and whether the addition of the Class 2B result would have resulted in a specific change in management (such as recommendation of ART).

VII. 40-GEP has demonstrated consistent incorporation into patient management plans in alignment with national guidelines recommendations.

Multiple publications form Castle Biosciences have proposed approaches for clinical use of the 40-GEP.^12-17 However, these are often inconsistent and management intensity classification approaches such as low, moderate, and high intensity have not been validated in additional cohorts or supported by independent experts outside of Castle Biosciences. While there is potential for the 40-GEP to help identify patients who are most at risk for metastasis and the test could be used to select patients for specific treatment options such as ART, this has not yet been demonstrated in the published peer-reviewed literature. The NCCN expert panel on cSCC states that molecular tests that offer risk stratification for cutaneous squamous cell carcinoma have not yet shown to have additional prognostic benefit over traditional staging in NCCN "very high risk" squamous cell carcinoma, the patient group in which additional prognostic stratification would be of potential value.

VIII. 40-GEP identifies ART eligible patients with very high biological risk of metastasis that benefit from ART and similarly identifies a significant population of ART eligible patients with low biological risk of metastasis that may avoid ART. Additional comments have called for use in special populations (such as immunosuppressed patients).

Submitted comments have raised a need for improved metastasis risk stratification in special populations such as immunosuppressed patients and for specific therapeutic indications, such as selection of patients who would likely benefit from ART. These indications are out of scope of this policy request and not reflected in the peer-reviewed publications available at the time of drafting this policy. The submitted comment begins to address the question of whether the 40-GEP can identify patients who are likely to respond to ART based on 40- GEP identification of high biological risk of metastasis. Unpublished results submitted as part of the Comment suggest that patients with a 40-GEP Class 2B result are more likely to respond to ART than patients with a Class 1 and Class 2A result. While these are promising results, it will be important to demonstrate the additive value of the Class 2B result in selection of patients for ART in addition to standard of care measures that include NCCN stratification, staging and clinical and pathological factors that are components of the multidisciplinary decision to perform ART. It is also important to note that across all BWH stages in the data presented in this comment (Tables 4-6), the majority of metastases occur in tumors with Class 1 and Class 2A 40-GEP results and the Class 2B result accounts for minority metastases in each stage and so management decisions with must be made carefully. We welcome a reconsideration request with peer-reviewed publications that demonstrate the value of the 40-GEP in these populations and indications.

IX. Appropriate use of 40-GEP to inform ART decisions can provide significant healthcare cost savings for the Medicare Trust Fund.

Medicare policy generation relies on evidentiary review, requiring evidentiary support for a successful coverage decision and does not consider cost or potential cost-savings in making policy determinations.

X. CMS’ Division of New Technology recently completed their review of the clinical validity and independence of the 40-GEP and concluded that the 40-GEP is a valid and independent predictor of metastasis within this high risk cSCC population.

We are not aware of any division of CMS reviewing the clinical validity of this test or coming to any conclusion about its medical value. The determination of coverage is based on the medical value of the test in fulfilling the reasonable and necessary criteria and is provided by Medicare Administrative Contractors and the Coverage and Analysis group within CMS.

If this comment is regarding a successful Advanced Diagnostic Laboratory Test (ADLT) application, section 1834A of the Social Security Act established this program with separate reporting and payment requirements. For ADLT designation, CMS regulations require that the test must be covered under Medicare Part B, offered and furnished only by a “single laboratory,” and not sold for use by any other laboratory except that single laboratory or a “successor owner.” In addition, the test must meet either Criterion (A) (analysis of multiple biomarkers of DNA, RNA, or proteins) or Criterion (B) (cleared or approved by the FDA). See 42 C.F.R. § 414.502. Statements alluding to this process as one that establishes the medical appropriateness of a test, or that it even considered such evidence are false and wholly without merit.

XI. 40-GEP has met criteria for medical reasonableness and necessity as defined by the Medicare Program Integrity Manual and the is consistent with similar tests in other disease states.

Policy generation is dependent on the context of the clinical and scientific problems presented. Thereby, different disease states, intended uses, and solutions are likely to have different evidentiary requirements for meeting the reasonable and necessary thresholds.

Comments from Castle Biosciences state that the comparator for molecular risk-stratifying tests should be current risk stratification systems, not improvement in patient outcomes. This may be appropriate in some circumstances, but we do not believe that to be the case here. Past Medicare decisions have aimed to measure the effect of the diagnostic information on patient outcomes such that if a particular diagnostic test result can be shown to change patient management and other evidence has demonstrated that those patient management changes improve health outcomes, then those separate sources of evidence may be sufficient to demonstrate positive health outcomes from the diagnostic test. While literature published by Castle Biosciences proposes changes in clinical management intensity using 40-GEP results, these management changes have not independently been shown to improve health outcomes, as the proposed utilization schemas have not been validated, nor have they been described nor adopted by experts other than those affiliated with Castle Biosciences. For example, Singh et al presents an algorithm for incorporation of 40-GEP test results into treatment decisions for high-risk cSCC patients, wherein the publication states that due to the low metastatic rate associated with a Class 1 result, it is recommended by the authors that patients with this test result may forgo imaging.¹¹ This recommendation is not evidence-based and may not be applicable to all patients with a Class 1 result, as the metastatic rate can vary based on underlying risk and hazard ratios for risk factors should be combined for more accurate risk assessment. The Comment received from Castle Biosciences demonstrates a 20% rate of metastasis in Stage T2b patients with Class 1 results, who would be in a “moderate intensity” category in the framework presented by Farberg et al and would still be considered for imaging according to that framework. The recommendation against imaging in these patients has not been validated and is conflicting with the framework presented in Farberg et al. Therefore, while management changes have been proposed by authors affiliated with Castle Biosciences, evidence has not yet demonstrated that these changes in patient management improve health outcomes. It is also important to note that across all BWH stages in the data presented in this comment (Tables 4-6), the majority of metastases occur in tumors with Class 1 and Class 2A 40-GEP results and the Class 2B result accounts for minority of metastases in each stage and so management decisions must be made carefully.

The American Academy of Dermatology currently recommends tumor stratification of localized cSCC using NCCN guidelines to provide practical clinical guidance for management and follow up and demonstrates preference for BWH in terms of the most accurate prognostication tool in localized cSCC. The analyses presented in 40-GEP publications to date have not demonstrated clinical validity and clinical utility in a clinical context that uses NCCN risk stratification AND staging plus clinical and pathological factors, as the multivariate analysis with staging was not conducted within the NCCN high risk and very high risk groups separately, but rather in the entire intended use population which includes both high risk (largely stage 1 tumors with low risk for metastasis and higher risk of local recurrence) and very high risk patients (generally higher stage with greater risk of metastasis) and so the benefit of NCCN risk stratification is lost. Given the low metastatic rate in the NCCN high risk category and questionable clinical utility, the clinical validity and clinical utility of the 40-GEP in addition to staging remains to be demonstrated in the NCCN very high-risk group. Importantly, any added risk stratification (clinical validity) within staging must demonstrate clinical utility with respect to clinical-decision thresholds within the framework of national guidelines.

Therefore, given the data presented for molecular biomarker testing for risk stratification of cSCC, it has not been clearly demonstrated that this test is reasonable and necessary, as a fully established chain of evidence from the performance of the test to the improvement of the patient's condition, is lacking.

References

1. Wysong A. Squamous-cell carcinoma of the Skin. N Engl J Med. 2023;388(24):2262-2273. doi:10.1056/NEJMra2206348.
2. Wysong A, Newman JG, Covington KR, et al. Validation of a 40-gene expression profile test to predict metastatic risk in localized high-risk cutaneous squamous cell carcinoma. J Am Acad Dermatol. 2021;84(2):361-369. doi:10.1016/j.jaad.2020.04.088
3. Ibrahim SF, Kasprzak JM, Hall MA, et al. Enhanced metastatic risk assessment in cutaneous squamous cell carcinoma with the 40-gene expression profile test. Future Oncol. 2022;18(7):833-847. doi:10.2217/fon-2021-1277
4. Ruiz ES, Karia PS, Besaw R, Schmults CD. Performance of the American Joint Committee on Cancer Staging Manual, 8th Edition vs the Brigham and Women's Hospital Tumor Classification System for cutaneous squamous cell carcinoma. JAMA Dermatol. 2019;155(7):819-825. doi:10.1001/jamadermatol.2019.0032
5. Stevens JS, Murad F, Smile TD, et al. Validation of the 2022 National Comprehensive Cancer Network risk stratification for cutaneous squamous cell carcinoma. JAMA Dermatol. 2023;159(7):728-735. doi:10.1001/jamadermatol.2023.1353
6. Soleymani T, Brodland DG, Arzeno J, Sharon DJ, Zitelli JA. Clinical outcomes of high-risk cutaneous squamous cell carcinomas treated with Mohs surgery alone: an analysis of local recurrence, regional nodal metastases, progression-free survival, and disease-specific death. J Am Acad Dermatol. 2023;88(1):109-117. doi:10.1016/j.jaad.2022.06.1169
7. NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines). Squamous Cell Skin Cancer. Version 1.2023. https://www.nccn.org/professionals/physician_gls/pdf/squamous.pdf. Accessed May 15, 2024.
8. Saleeby E, Bielinski, K., Fitzgerald, A., Siegel, J., & Ibrahim, S. A prospective, multi-center clinical utility study demonstrates that the 40-gene expression profile (40-GEP) test impacts clinical management for Medicare-eligible patients with high-risk cutaneous squamous cell carcinoma (cSCC). SKIN The Journal of Cutaneous Medicine. 2022;6(6):482–496.
9. Venables ZC, Autier P, Nijsten T, et al. Nationwide incidence of metastatic cutaneous squamous cell carcinoma in England. JAMA Dermatol. 2019;155(3):298-306. doi:10.1001/jamadermatol.2018.4219.
10. Kim JYS, Kozlow JH, Mittal B, Moyer J, Olenecki T, Rodgers P. Guidelines of care for the management of cutaneous squamous cell carcinoma. J Am Acad Dermatol. 2018;78(3):560-578. doi:10.1016/j.jaad.2017.10.007.
11. Singh G, Tolkachjov SN, Farberg AS. Incorporation of the 40-gene expression profile (40-GEP) test to improve treatment decisions in high-risk cutaneous squamous cell carcinoma (cSCC) patients: case series and algorithm. Clin Cosmet Investig Dermatol. 2023;16:925-935. doi:10.2147/ccid.S403330
12. Au JH, Hooper PB, Fitzgerald AL, Somani AK. Clinical utility of the 40-gene expression profile (40-GEP) test for improved patient management decisions and disease-related outcomes when combined with current clinicopathological risk factors for cutaneous squamous cell carcinoma (cSCC): case series. Dermatol Ther (Heidelb). 2022;12(2):591-597. doi:10.1007/s13555-021-00665-y
13. Hooper PB, Farberg AS, Fitzgerald AL, et al. Real-world evidence shows clinicians appropriately use the prognostic 40-gene expression profile (40-GEP) test for high-risk cutaneous squamous cell carcinoma (cSCC) patients. Cancer Invest. 2022;40(10):911-922. doi:10.1080/07357907.2022.2116454.
14. Litchman GH, Fitzgerald AL, Kurley SJ, Cook RW, Rigel DS. Impact of a prognostic 40-gene expression profiling test on clinical management decisions for high-risk cutaneous squamous cell carcinoma. Curr Med Res Opin. 2020;36(8):1295-1300. doi:10.1080/03007995.2020.176328
15. Farberg AS, Hall MA, Douglas L, et al. Integrating gene expression profiling into NCCN high-risk cutaneous squamous cell carcinoma management recommendations: impact on patient management. Curr Med Res Opin. 2020;36(8):1301-1307. doi:10.1080/03007995.2020.1763284
16. Arron ST, Blalock TW, Guenther JM, et al. Clinical considerations for integrating gene expression profiling into cutaneous squamous cell carcinoma management. J Drugs Dermatol. 2021;20(6):5s-s11. doi:10.36849/JDD.2021.6068
17. Teplitz R, Prado G, Litchman GH, Rigel DS. Impact of gene expression profile testing on the management of squamous cell carcinoma by dermatologists. J Drugs Dermatol. 2019;18(10):980-984.

The following comment was submitted to WPS and Noridian:

The National Comprehensive Cancer Network^® (NCCN^®) appreciates the opportunity to comment on the Proposed Local Coverage Determination (LCD) Molecular Biomarker Testing for Risk Stratification of Cutaneous Squamous Cell Carcinoma as it relates to NCCN's mission of improving and facilitating, quality, effective, equitable, and accessible cancer care so all patients can live better lives. NCCN is supportive of the proposed LCD and is pleased to provide information and resources.

NCCN Background

As an alliance of 33 leading academic cancer centers in the United States that treat hundreds of thousands of patients with cancer annually, NCCN^® is a developer of authoritative information regarding cancer prevention, screening, diagnosis, treatment, and supportive care that is widely used by clinical professionals and payers alike. The NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines^®) are a comprehensive set of guidelines detailing the sequential management decisions and interventions that currently apply to 97 percent of cancers affecting patients in the United States.

NCCN Guidelines^® and Library of Compendia products help ensure access to appropriate care, clinical decision-making, and assessment of quality improvement initiatives. The NCCN Drugs & Biologics Compendium (NCCN Compendium^®) has been recognized by CMS and clinical professionals in the commercial payer setting since 2008 as an evidence-based reference for establishment of coverage policy and coverage decisions regarding off-label use of anticancer and cancer-related medications. NCCN was recognized by CMS in 2016 and renewed in 2021 as a qualified Provider Led Entity (PLE) for the Medicare Appropriate Use Criteria (AUC) Program for the development of AUC and the establishment of policy and decision-making for diagnostic imaging in patients with cancer.

NCCN imposes strict policies to shield the guidelines development processes from external influences. The "firewall" surrounding the NCCN Guidelines includes: financial support policies; panel participation and communication policies; guidelines disclosure policies; and policies regarding relationships to NCCN's other business development activities. The guidelines development is supported exclusively by the Member Institutions' dues and does not accept any form of industry or other external financial support for the guidelines development program. The NCCN Guidelines are updated at least annually in an evidence-based process integrated with the expert judgment of multidisciplinary panels of expert physicians from NCCN Member Institutions. The NCCN Guidelines are transparent, continuously updated, available free of charge online for non-commercial use and are available through a multitude of health information technology (HIT) vendors.

NCCN Biomarkers Compendium^®

The NCCN Biomarkers Compendium^® is intended to be a resource for payers, providers, and health care entities navigating the rapidly changing evidence-base for medically necessary biomarker testing in oncology. The NCCN Biomarkers Compendium^® contains information derived directly from the NCCN Guidelines to support decision-making around the use of biomarker testing in patients with cancer. The NCCN Biomarkers Compendium is updated continuously in conjunction with the NCCN guidelines to stay evergreen. The goal of the NCCN Biomarkers Compendium is to provide essential details for testing methodologies which have been approved by NCCN Guideline Panels and are recommended within the NCCN Guidelines. Included in the Biomarkers Compendium are testing methodologies that measure changes in genes or gene products and used for the purposes of diagnosis, screening, monitoring, surveillance, prediction, and prognostication.

Launched in 2012 and updated in 2021 to have shared data fields with other Compendia in the Library, the information within the Biomarkers Compendium is extracted directly from Guideline algorithms, principles pages, and footnotes, and all entries are reviewed and approved by the Guideline Panel pathologist or other panel member with expertise in the area. Information within the Biomarkers Compendium focuses on the biology or abnormality being measured rather than on commercially available tests or test kits, with methodologic information provided only if included in the parent Clinical Practice Guideline.

The Biomarkers Compendium is intended to be user friendly in an easy to navigate format for payers and providers alike. The displayed fields can be customized to the user and include links to the Guideline Page on which the biomarker is addressed. Compendium entries can be printed, providing the summary of the recommendations with details regarding each field within the Compendium, in a format that is concise and easy to read. With this resource, the NCCN Biomarkers Compendium aims to ensure that patients have coverage and access to appropriate biomarker testing based on the evaluations and recommendations of NCCN Panel Members.

NCCN Guidelines and Compendia as a Tool to Keep Coverage Determinations Evergreen

Recognizing the value of continuously updated evidence-based guidelines, numerous Medicare Administrative Contractors have cited NCCN Guidelines within their proposed and final coverage policies as a mechanism to evergreen coverage regarding drug and biologic coverage. These policies include Local Coverage Article: Billing and Coding: Chemotherapy and Local Coverage Determination Luteinizing Hormone-Releasing Hom1one (LHRH) Analogs, LCD Chemotherapy Drugs and their Adjuncts, and Proposed Genomic Sequence Analysis Panels in the Treatment of Solid Organ Neoplasms among others. NCCN is pleased to offer these guidelines and related compendia products as a resource for coverage to ensure it remains evergreen.

NCCN Guideline References within the LCD and Areas for Clarification

NCCN thanks for referencing NCCN content within the proposed LCD's summary of evidence. The NCCN expert panel on Squamous Cell Skin Cancer agrees with the determination of non-coverage for molecular tests that offer risk stratification for cutaneous squamous cell carcinoma based upon existing evidence which has not yet shown these tests to have additional prognostic benefit over traditional staging in NCCN "very high risk" squamous cell carcinoma. This is the group in which additional prognostic stratification would be of potential value. NCCN would like to further note thanks that the most up to date guideline is referenced within the proposed LCD. However, NCCN recognizes the evidence for biomarker tests in oncology is frequently evolving and our guidelines update as the evidence warrants changes in optimal practice. NCCN Guidelines are updated on a continual basis, with a minimum of one update annually, though more frequent updates commonly occur as the scientific evidence evolves. NCCN notes that the most frequent versions across our library of clinical practice guidelines can always be found online, free of charge, at NCCN.org.

NCCN again thanks for including NCCN content within the proposed LCD. NCCN appreciates the opportunity to comment on the Molecular Biomarker Testing for Risk Stratification of Cutaneous Squamous Cell Carcinoma. NCCN is happy to serve as a resource and looks forward to working together to advance access to high-quality cancer care for beneficiary population.

Thank you for your comment in support of this policy.

The following comment was submitted to CGS:

As a dermatologist with 20 years in practice, it is clear that most patients with cutaneous squamous cell carcinoma need only surgical clearance of the lesion of interest to enjoy long term favorable outcomes and have low risk of disease progression. However, for those patients with concerning risk factors, typically around 45 patients per year in my practice, the clinical decisions for post-surgical management are not as straightforward. I was interested to learn about the data surrounding the SCC prognostic GEP test and its ability to complement current methods of assessing risk of poor outcomes for patients with cutaneous squamous cell carcinoma. The clinically validated test classifies patients as having a low, moderate, or high risk of metastasis within three years of initial SCC diagnosis based on gene expression in the primary tumor (Wysong, et al 2020, Ibrahim, et al 2021). So, in addition to tumor staging or risk classification by the National Comprehensive Cancer Network, the SCC GEP test can provide supplemental prognostic information based on the tumor’s risk of metastasizing, which would add another layer of information to risk assessment. It is important to note in the LCD language that the test is not intended to replace or compete with traditional risk stratification methods, but rather, to supplement with additional information. With information from this additional tool, healthcare providers can better manage their patients in a risk-appropriate manner.

Importantly, the SCC GEP test is not a test that should be ordered for patients with low-risk SCC, but rather is intended for use with SCC patients with one or more risk factors (i.e., high-risk SCC). The proposed LCD language should remove reference to any results “confirmative of low risk”, as only high-risk patients are able to receive testing with the SCC GEP test.

One paper that was cited but not reviewed by the proposed LCD is the clinical utility study by Litchman et al that evaluates how physicians infer risk from specific factors, and of note, is similar in both methods and results to the paper by Patel et al that was reviewed in detail in the proposed LCD. The proposed LCD infers that Litchman et al was not reviewed due to lack of outcomes, however, this Patel et al does not have outcomes and is reviewed and included.

Litchman et al appears to be dismissed by statements implying it was not reviewed due to lack of outcomes data within it, however, this standard is not held to articles included that discuss validation and value of BWH or AJCC staging. It is noteworthy that there are no studies with outcomes that support BWH or AJCC staging, and therefore, the SCC GEP test has a comparable level of evidence as risk stratification tools currently employed in the clinic. This should be transparently reviewed and discussed in the proposed LCD.

Litchman et al, includes findings around use of specific risk factors that guide patient management is similar to findings in Patel et al. Both studies conclude that the clinicopathologic risk factors assessed as highest risk for patients to guide management were perineural invasion, immunosuppression, invasion beyond the subcutaneous fat, and location on the scalp. Of particular interest, in Litchman et al, Class 2B GEP scored the highest for perception of risk, above perineural invasion, and Class 2A GEP ranked 6^th for perception of risk, just behind location on the scalp. The similarity of results support that the survey conducted by Litchman et al is a reliable assessment of clinician behavior, and further support its inclusion in the LCD.

Finally, where the Patel et al study finds consensus among those using staging, the BWH T2b tumor is indicative of clinically actionable risk for consideration of additional therapy modalities, including radiologic imaging for detection of lymph node or distant metastasis, SLNB, post-operative ART, post-operative systemic therapy, and increased follow up, the Litchman study finds similar consensus regarding impact of a Class 2B GEP to management of a BWH T2a tumor (Vignette 1 in the Litchman, et al). From validation studies, we know that a BWH T2a tumor may be considered higher risk, particularly in the cases presented in Litchman in which the tumors are > 1.0 cm and located on the scalp, which is not considered high risk in any staging system but is in fact perceived as high risk by dermatologists managing disease in both studies. Also note from validation studies that BWH T2a tumors with Class 1 GEP carries risk similar to the overall population of BWH T1 tumors and with Class 2B GEP carries risk similar to BWH T3 tumors. Thus, these are clinically meaningful cases for evaluation of impact of test results in management changes. In fact, the pre-GEP test management plans show significant clinical uncertainty, with a mix of avoid/consider recommendations from clinicians across sentinel lymph node biopsy, nodal imaging, adjuvant radiation, and adjuvant chemotherapy. This publication shows how SCC GEP results are combined with risk factor information to guide management decisions by identifying levels of risk that are already used to guide management changes today. Analysis of evidence and rationale that ties the increased and decreased metastatic risk prediction from SCC GEP results to current metastatic risk prediction by BWH and resultant, appropriate management changes based on this risk assessment with consistent patient treatment recommendations across physician algorithm publications and clinical utility studies should be laid out fully detailed in the proposed LCD.

The Litchman et al study also found a Class 1 GEP resolved uncertainty, with high consensus among the 162 participating clinicians for de-escalation with a Class 1 GEP with consensus for reduction of sentinel lymph node biopsy (75-80%), nodal imaging (70-80%), adjuvant radiation therapy (70-80%), and adjuvant chemotherapy (80-90%), the change in which all reached statistical significance compared to baseline management. Moreover, a Class 2B GEP significantly increased intensity of management, with a ~60% increase in recommendations for sentinel lymph node biopsy, 50-60% increase in recommendation for nodal imaging (either US or CT 1-4x per year), 50% increase in recommendation of adjuvant radiation therapy, and 40% increase in recommendation for adjuvant chemotherapy. All of the management changes with Class 2B GEP also achieved statistical significance over baseline management plans. Thus, the GEP test results clearly impact management in a risk-aligned manner.

I utilize this test for the patients I see with high-risk SCC and this test has complemented my methods for stratifying individual patient risk, so that I can provide a more personalized approach to the management of each patient. I strongly support the utilization of this test in high-risk SCC and coverage of the test is necessary to provide the best possible care I can provide to my patients.

Thank you for your thoughtful comment. We share your concern for patients with cutaneous squamous cell carcinoma and in your drive to improve risk stratification and clinical management. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to CGS:

I am a Mohs surgeon. The Castle GEP tests have become an integral part of my practice in recent years.

I see many Medicare beneficiaries as part of my practice. I am writing in support of the Castle Biosciences 40-gene prognostic test for cutaneous squamous cell carcinoma. Published evidence on the validation of this test have demonstrated that patients with a low-risk Class 1 result have about half the metastatic risk of the overall cohort, patients with a Class 2A result have added risk equivalent to the presence of other commonly accepted high-risk factors, and patients a Class 2B result have ~3 times the risk of metastasis to the overall cohort. Moreover, this added risk stratification is independent of the presence of other high-risk factors; these tests provide information that cannot be assessed through traditional risk assessment methods for these patients.

This demonstrates that the test meets Medicare requirements to improve prognostication in relation to the current standard of care and to provide clinically actionable information that clinicians use to change management decisions. Importantly, the risk stratification performance of the test is maintained in patients grouped by risk. As an example, if considering use of BWH staging, the test performs significant risk stratification in T1, T2a, and T2b staged tumors.

Most patients with cutaneous squamous cell carcinoma do very well, but some patients have very aggressive disease, and it is important to identify them as quickly as possible. In my own practice, I use this test with high-risk patients to better determine their treatment among the broad range of options available. In a multidisciplinary setting, the 40-gene test is useful for discussions with our radiation oncologist to determine who should get radiation and with our head and neck surgeon as we make decisions about who should get sentinel lymph node biopsy and screening/surveillance imaging.

The test may be used to be more aggressive with treatment or to de-escalate care depending on the specific patient. This use of the test is informed by published literature and is consistently implemented in practice. I believe the test can save both patients and Medicare money and reduce risk. I implore Medicare to continue coverage for the 40 GEP SCC test. It has met the level of evidence required to support coverage and continued accessibility to Medicare patients is necessary to provide the best care available.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to CGS:

We are a group of dermato-oncologists with a particular expertise in the care of organ transplant and immunosuppressed patients with a high burden of cutaneous squamous cell carcinoma and other cutaneous malignancies. We are writing in our personal capacity as physicians who care for these high risk patients.

We are writing to advocate for MolDX to reconsider their non-coverage of the DecisionDx-SCC test and revise their current draft policy to cover this test, particularly in immunosuppressed subjects. Immunosuppressed patients with this dangerous cutaneous malignancy can benefit from more closely risk-aligned care informed by test results.

A high number of immunosuppressed patients will develop cSCC, with a much higher incidence compared to immunocompetent patients (ranges between 65-100 fold greater for solid organ transplant recipients [O’Reilly Zwald, 2011]), along with reports of a more aggressive tumor behavior (Elchouche 2019, Manyam, 2015, 2017, Stoff 2010). A recent meta-analysis on around 140,000 cSCC patients reported immunosuppressed patients are associated with significantly higher local, nodal, and disease-specific death [Zakhem, 2022]. However, as noted in the LCD proposal, immunosuppression is not considered as a risk factor in current staging systems (AJCC v8 and BWH).

The incidence of cSCC in organ transplant recipients is over 800/100,000 person years in the US (Garrett, 2017), nearly double the population-based incidence of all cancers combined in the US (SEER data 2013-17, https://www.cancer.gov [cancer.gov]). Treating these patients is complex, as the variety of immunosuppressive drugs they have taken over the years may influence both the incidence and aggressiveness of their cSCC.

Since treating physicians must be judicious in selecting treatments for these high-risk patients, having a test based on the tumor's molecular signature has been an invaluable addition to the field of cSCC. As the draft policy rightly states, the test is for those tumors where additional precision in risk stratification can help us in making patient-based recommendations for escalation of care or justify active surveillance in cases where the risks of treatment may outweigh the benefits.

DecisionDx-SCC has demonstrated statistical significance in patients with immunosuppression: A recent oral presentation at the World Congress of Dermatology, a subset analysis of 241 patients with immunosuppression was presented, of which, nearly 70% were transplant recipients. In this cohort, the DecisionDx-SCC test significantly stratified metastatic risk. When DecisionDx-SCC was integrated into a model with AJCC8 or BWH staging systems and immune status, inclusion of the test significantly increased the likelihood ratio for prediction of metastatic outcome (Arron et al. WCD July 2023 presentation; manuscript in preparation).

Many of our patients will have multiple primary high-risk cSCC; the burden of disease is matched by the burden of treatment. We are often faced with decision-making in the absence of evidence, particularly with regard to the utility of sentinel lymph node biopsy or postoperative radiation. NCCN guidelines for postoperative treatment of localized high/very-high risk cSCC with clear surgical margins are extremely broad: to recommend multidisciplinary tumor board and consider adjuvant radiation in the presence of specific risk factors indicative of high-stage disease. Many of us participate in multidisciplinary tumor boards, and there are frequent cases where additional information would guide decisionmaking.

In summary, we believe that there is value in the DecisionDx-SCC test for our for immunosuppressed patients, as it significantly improves risk prediction and is a useful tool for risk-aligned patient management plans. Please reconsider non-coverage of the DecisionDx-SCC test and revise the current draft policy to cover this test.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to CGS:

I am writing to express my disagreement with the LCD regarding non-coverage for the Castle GEP prognostic test for squamous cell carcinoma (SCC). At a high level, the clinical validity of the test in comparison to staging approaches is incompletely summarized in the LCD, statistical analyses and their relevant meaning are not appropriately interpreted, and assumptions that do not reflect the published evidence guide the evaluation of clinical utility.

As with every physician who treats patients with high-risk cutaneous neoplasms, I am tasked with assimilating diverse types of data to direct impact patient management decisions after I complete my surgical removal of a patient’s tumor (pathology reports, imaging reports, complex medical history, particular patient preferences, etc). To that end, current guidelines lay out very broad management pathways based on my clinical assessment of a patient’s risk of poor outcomes, and I have very broad latitude for the particular management strategy that I pursue for each patient. As such, all available risk information, be it from clinical, pathological, or genomic information regarding a tumor’s risk for metastasis, should be integrated and used to guide independent medical judgement as to the appropriate course of treatment for each patient.

The Castle GEP test has demonstrated impact in physician management of patients, in both clinical impact studies and in an ongoing prospective study (Saleeby et al., Hooper et al., Litchman et al., Teplitz et al.) Importantly, these studies demonstrate agreement from physicians on how the test should be used in the context of specific management decisions that are outlined by guidelines, including use of adjuvant radiation therapy, use of imaging to guide surveillance, and intensity of follow-up. I use the final clinical and pathologic risk factors identified in combination with the Castle GEP to pursue risk aligned management decisions for my patient.

Failure to cover this critically important tool, particularly when evidence is missed that supports the clinical value of the test, is a disservice to clinicians who already incorporate this tool as part of clinical practice, to the overall Medicare system which will continue to incur costs of overtreatment in high stage patients, and ultimately to patients. I am hopeful that more thorough review of the evidence supporting this test will result in approval for Medicare coverage for the benefit of these patients.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to CGS:

As a medical oncologist, I often see patients with late-stage cutaneous malignancies and high risk cutaneous squamous cell carcinoma (cSCC) that has progressed from local disease treatable with definitive surgery to advanced stage disease that requires multidisciplinary care and systemic approaches to management. Although I manage patients at late stage, it is crucial for MolDX to understand that every patient referred to my care was seen by other physicians during a window of opportunity to provide risk aligned management at an earlier time point.

In this effort, I work closely with dermatology and dermatologic surgery colleagues who are the primary points of contact for patients with cSCC as they are the physicians tasked with diagnosis and initial surgical management for cSCC patients. What is often overlooked is that dermatologists and dermatologic surgeons are crucial to establish the intensity of management that follows the diagnosis and definitive surgery. Following primary surgical management, the dermatologists and dermatologic surgeons are responsible for determining critical management strategies that align to their patient’s risk of poor outcomes. These decisions include which patients should obtain more frequent clinical follow-up, intensity of nodal screening as part of surveillance and referral to multidisciplinary care or adjuvant radiation therapy (ART). Given the fact that most cSCC are actually low risk and lack significant risk factors associated with higher risk of metastasis, too many tumors with high biological risk of metastasis are missed and are not referred to me until the disease has advanced substantially and at a time point too late for most management interventions to be effective.

As a result of this, I am frequently referred patients who present with recurrent unresectable cSCC or metastatic cSCC who could have been managed with a higher intensity of treatment to avoid these poor outcomes had they been more accurately risk stratified earlier in their care. Instead of waiting for a patient to develop bulky nodal metastasis or unresectable recurrence, patients with a high risk of poor outcomes could be better identified and watched more closely in clinic, surveilled with imaging, have their case reviewed in multidisciplinary tumor board, and/or be considered for adjuvant radiotherapy. This window of opportunity is increasing in importance with the recent approval of immune checkpoint blockade for advanced stage cSCC tumors as well as promising phase 2 neoadjuvant therapy data in even earlier stage tumors.^{1, 2} We are learning that checkpoint inhibition has a clear role in cSCC and, as a result, in my practice I am using cemiplimab for these late-stage patients and improving their outcomes.

Use of a prognostic molecular tools, such as the 40-GEP test that has been validated by Castle Biosciences to predict risk of metastasis in patients beyond prognosis provided by clinical and histopathologic risk factors, is critically important to identify which of these patients harbor biologically aggressive tumors that can be managed more closely to catch metastatic events at an earlier, more treatable time point. As in studies with similar tools for melanoma, use of imaging surveillance in patients with high risk scores can identify subclinical metastases when tumor burden is smaller and more likely to respond to therapy.³ In these studies, use of molecular prognostication is associated with higher survival rate in patients that receive testing compared to patients that do not receive testing. As such, the 40-GEP has an indirect chain of evidence that it will improve outcomes and should be covered by Medicare for patients with high risk SCC.

References were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to CGS:

As a surgical oncologist who specializes in managing high-risk malignancies, I frequently take care of patients with high-risk cutaneous squamous cell carcinoma (cSCC). These patients can be tricky to appropriately manage as cSCC lacks definitive staging and the associated tailored management to staging. Thus, it is paramount to use all clinical, pathologic, and molecular tools available to accurately predict risk of disease progression in these patients to guide management. These decisions include which patients should obtain more frequent clinical follow-up, intensity of nodal screening, including use of imaging surveillance or SLNB, and referral to multidisciplinary care or adjuvant radiation therapy (ART). Given that most cSCC have low biological risk for disease progression and the lack of significant risk factors and associated risk of metastasis, it is too often that presumed low-risk patients by staging assessment harboring aggressive disease go missed until progression has already occurred.

A GEP test available from Castle Biosciences has demonstrated robust separation of metastatic risk in patients with cSCC with high risk factor(s), both in the overall population of patients, and within specific risk subsets of disease. What is noteworthy is that the LCD accepts T2b/T3 Brigham and Women’s staged tumors as having a valid level of risk to drive patient management decisions based on physician algorithms, surveys, and retrospective studies. Indeed, published literature and the LCD both seem to consider staging as a validated tool to predict risk of metastasis and to implement that risk assessment to guide changes in management within guidelines.¹ Of note, there are no prospective studies, nor studies with outcomes that validate Brigham and Women’s staging and the use of this system to guide management decisions.

The GEP high risk results (Class 2B), across all risk subsets of patients, identifies a risk of metastasis similar to T2b/T3 Brigham and Women’s staged tumors, however, the LCD does not acknowledge this risk stratification as reasonable or valid to act upon, citing the lack of studies with outcomes.² In the same turn, a GEP Class 1 result in a T2b Brigham and Women’s staged tumor identifies an equivalent to T2a Brigham and Women’s stage or below, which is currently used to guide less intensive therapy. Again, there are no studies with outcomes used to establish that BWH T2a tumors have a low enough risk of metastasis to forgo more intensive therapy. However, when use of a Class 1 result to de-escalate care has substantial discussion of the potential harms of de-escalation. Given the equivalent rate of metastasis in these patients to those with T2a Brigham and Women’s staged tumors, and no studies with outcomes assessing appropriateness of management in these patients, the LCD is requesting a level of evidence to implement GEP testing that is not available now in current standard of care.^3,4

The level of evidence required to support GEP testing should be equivalent to that required to support current standard of care. The LCD should reconsider the requirement for GEP testing to demonstrate impact on patient outcomes given the current status quo for management in this disease state has not demonstrated improvement in outcomes. Even more, this level of evidence is not required in other advanced tests used in other disease states (see bladder cancer, breast cancer, prostate cancer, uveal melanoma LCDs). At a minimum, it is incumbent on MolDX administering this policy to clearly articulate a sound rationale for higher burden of evidence disproportionately placed upon this specific test and disease state.

GEP testing is an important advancement that improves upon the current standard of care for managing cSCC patients, and I urge you to replace this policy with on that fully articulates the value of GEP testing in the framework of the current standard of care and consistent with what is required in other disease states. Thank you for your time and consideration of these comments.

References were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to CGS:

As a board-certified dermatologist and fellowship-trained/board-certified Mohs surgeon I have over 13 years of clinical experience. In my current practice, I manage patients with cutaneous malignancies with a focus on the primary surgical treatment of their lesions – often with Mohs surgery. After the completion of my patient’s surgical and immediate post-operative care I am tasked with guiding their subsequent management both in my dermatologic surgery clinic as well as in close collaboration with my clinical dermatology colleagues. When appropriate, I am also able to route patients to other referral specialists that include a local multidisciplinary tumor board, radiation oncologists, surgical oncologists, or medical oncologists. The care of patients with high-risk SCC (HR-cSCC) tumors can present significant challenges due to the well-known limitations in predictive accuracy of existing staging approaches that use the same set of limited clinical and pathologic risk factors. Increased prognostic information related to the key driver of my post-surgical management – risk of regional and/or distant metastasis – helps me to triage my patients to effective management strategies in a more risk-aligned fashion.

I was disappointed in the level of evidence review and conclusions contained within the recent proposed LCD: MolDX: Molecular Biomarker Testing for Risk Stratification of Cutaneous Squamous Cell Carcinoma that resulted in proposed non-coverage of the 40-GEP test for squamous cell carcinoma metastatic risk stratification. From my perspective, one of the critically missing aspects in the LCD is the perspective of the clinicians treating these patients with HR-cSCC and the importance of personalized management approaches. In every NCCN guideline, in one of the opening pages, there is an opening caveat statement that highlights the importance of clinical context, “Any clinician seeking to apply or consult the NCCN Guidelines is expected to use independent medical judgment in the context of individual clinical circumstances to determine any patient’s care or treatment.” To me, this is a core tenet of what it means to provide high quality cancer care to my patients. More relevant to this letter, and what is therefore missing from the policy regarding 40-GEP, is that it does not accurately identify the clinical scenario in which DecisionDx-SCC has demonstrated utility. Rather than acknowledging the uncertainty in risk prediction in these high risk cSCC patients that is poorly defined by a myriad of clinical and pathological features, the clinical validity analysis focuses primarily on the “either-or” scenario between risk factor-based assessment and the 40-GEP test while underappreciating the “combination” value of these two approaches.

As a Mohs surgeon I think carefully about the patients who I treat surgically, and it is obligatory to evaluate the known clinical and pathological risk factors available to me while I decide perioperative, operative, and post operative management plans. The advent of the 40-GEP does not negate the value of the approach that incorporates these risk factors into risk aligned management plans and so the clinical goal is to incorporate GEP testing with risk factor-based assessment instead of evaluating one approach vs. the other in isolation. In practical application, for example, while a multivariable analysis published in Ibrahim, et al. shows that deep invasion, poor differentiation and a Class 2A result each have a hazard ratio of ~2, this does not mean that a Class 2A has no value in a patient with either poor differentiation or deep invasion due to the presence of a risk factor that already carries a similar weight hazard ratio. Instead, the risk of the two factors would be combined; a Class 2A in a patient with poor differentiation or deep invasion would portend a risk similar to a patient that had the presence of both of these risk factors. In this way, physicians are able to use risk information from multiple sources, including genetic testing, to get a complete picture of patient risk.

Another clinical example for combination of genetic and pathological risk information can be found in the risk factor “perineural invasion” (PNI, also known as “perineural involvement” in NCCN), a risk factor that is broadly agreed upon as being a marker of worse clinical prognosis. At present, PNI is incorporated into NCCN guidelines as a high risk and very high risk factor; PNI of any nerve is considered a high risk factor while PNI of a nerve lying deeper than the dermis or measuring ≥0.1mm is considered a very high risk factor. PNI is also considered as a staging factor in both AJCCv8 and BWH staging; PNI of a nerve measuring ≥0.1mm is considered a risk factor for metastasis. In this setting, the presence of any PNI differentiates between a low-risk and a high-risk cSCC, however only the presence of large caliber nerve involvement registers as a slightly higher AJCC/BWH stage. If PNI is the only clinical or pathologic risk factor noted for a case, then the presence of this concerning risk factor still correlates with the lower-end of risk by staging (AJCC T2 and BWH T2a, respectively). In practice, the presence of PNI confers only a small amount of prognostic clarity for your patients in HR-cSCC, and yet clinicians use this information to inform their surgical and post-surgical management strategies. The presence of PNI alerts to a HR-cSCC, but it is frequently unclear how much worse the prognosis might be for the patient. The clinical application of the 40-GEP occurs in this moment, after the clinical or pathologic risk factors have been identified and categorized, when the doctor managing the patient’s HR-cSCC is looking to obtain more prognostic clarity. As such, it is imperative that the policy evaluate the clinical validity of the 40-GEP to as an adjunct to staging to improve the prognostic accuracy of risk-factor-based approaches such as NCCN/AJCC/BWH. Again, the goal here is to improve our current risk stratification and not simply replace clinical and pathologic factors with 40-GEP alone.

The other aspect of the policy that I wanted to address is the insinuation that there could be potential for inappropriate de-escalation of care based on the 40-GEP, “Patient misclassification and under-treatment as a result of a false negative result could cause significant harm and needs to be addressed.” While it is true that the 40-GEP does not report 100% sensitivity/specificity and therefore has the potential to misclassify patients, this statement misses the mark clinically as it does not invoke the potential significant harm of continuing risk factor-based prognosis which ALSO has known limitations in accuracy and misclassifies patients on a regular basis. This statement suggests that risk stratification in HR-cSCC already bins patients accurately and routes them to rigidly defined post-operative management plans when the field is actively looking for ways to improve the accuracy of risk stratification and build greater consistency and risk thresholds for our management decisions.

The policy compares 40-GEP directly against an idealized, and almost sterilized, version of current clinical management for HR-cSCC and fails to appreciate how 40-GEP can improve several pain points for physicians managing these patients. What is missed from the policy, therefore, is the current landscape where patients self-triage to higher or lower intensity management on a regular basis and physicians escalate/de-escalate care based on the presence or absence of their favorite risk factor (whether or not it is incorporated into staging). The 40-GEP is compelling for HR-cSCC because it can improve the risk stratification that we use already in clinic, it provides an objective data point that can be used to guide patient discussions around important post-surgical management plans and has demonstrated improvement in risk aligned management. It is for these reasons that I believe that the 40-GEP should be considered as a covered test in the final policy.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA, WPS, and Noridian:

I am a board-certified dermatologist and micrographic surgeon.

This letter summarizes my perspective on the noncoverage draft of MolDX: Molecular Biomarker Testing for Risk Stratification of Cutaneous Squamous Cell Carcinoma for Castle Biosciences’ 40-GEP (DecisionDx-SCC) test. As a practicing that specializes in the management of high-risk cutaneous neoplasms, I strongly believe that this test should be covered under Medicare, contrary to the conclusions in the current proposed LCD.

Risk stratification systems based on clinical and pathologic factors are the initial method for risk stratification of cSCC patients in standard practice. However, these systems have well-known limitations, leading to both understaging and overstaging of patients (Tschetter 2020, Ruiz 2019, Marrazzo 2018, Canueto 2018, Haisma 2016, Karia 2014, Jambusaria 2013). That is, many patients that will eventually metastasize have lower estimation of risk based on clinical and pathological staging factors, and would actually benefit from increased nodal surveillance, imaging, and potentially even adjuvant radiation therapy. On the other hand, the majority of patients identified as highest risk do not metastasize and are unlikely to benefit from the aforementioned management strategies. Nonetheless, these patients identified by clinicopathologic factors as having high risk of poor outcomes are routinely subjected to imaging, advanced nodal surveillance, or adjuvant radiation therapy (ART) which is associated with significant health care costs and can be associated with significant morbidity.

The 40-GEP has been clinically validated and shown to provide independent prognostic information, regardless of the staging systems used. (Wysong 2020, Ibrahim 2021, Arron 2022) Through a multivariate analysis, the 40-GEP test has been shown to be independent of any other risk- stratification systems. (Ibrahim 2021) As a Mohs surgeon, I am particularly interested in the outcome of patients treated with this effective surgical approach, as was requested in the proposed LCD. The independence of the 40-GEP result from staging systems continues in the subset of patients who were all treated with Mohs surgery as presented at the 52^nd Annual Meeting of the American College of Mohs Surgeons. (Table 1, Wysong, et al. 2023a) By enhancing the accuracy of prognostic assessments in this cohort, the 40-GEP test improves the care provided to patients undergoing Mohs surgery.

The statistical improvement in accuracy of risk prediction when the 40-GEP is combined with clinicopathologic based staging is demonstrated by an ANOVA analysis comparing accuracy of the staging systems alone and in combination with staging, as presented at Annual Meeting of the American Academy of Dermatology. (Wysong, et al. 2023b) In this analysis, integrating the GEP test results with staging systems further refines metastatic risk prediction, as demonstrated by a statistically significant improvement in the likelihood ratio. The inclusion of the 40-GEP test significantly improves the accuracy of the AJCC8 and BWH staging systems, leading to more precise prognostic assessments.

To restate, we deploy increased intensity management strategies as part of our standard of care in the management of cSCC and we make decisions on a range of management options including, for example, who receives ART, who we should escalate from clinical node assessment to increased nodal surveillance with imaging modalities – all of which are already supported by guidelines for patients with high risk cSCC. A critical flaw in the draft LCD is the premise that there is no benefit or uncertain efficacy of these management strategies to improve patient outcomes. It is important to note here that these higher intensity management strategies – follow-up frequency, nodal surveillance, ART – are established interventions in current clinical practice, articulated clearly in national guidelines and have the ability to improve patient health outcomes when they are provided to the right patients – those with the highest risk for cSCC disease progression. (Ruiz 2022, Ruiz 2017) The primary clinical issue is the poor decision- making value obtained from clinical and pathologic factors alone. Thus, it is a crucial need in the clinic to have additional risk stratification information to make better aligned decisions for these patients to identify the right level of therapy, known to improve patient outcomes, for the right patient.

As noted in comment, two cases were published as a case series cited by the draft LCD. (Au 2022) After definitive surgical management and taking stock of the clinical and pathological risk factors present, both patients were similar in age, immunosuppressed status-post solid organ transplant and who both have poorly differentiated SCC tumors on the left temple that fall in the category of “small” cSCC tumors (< 2cm diameter). As a result of this specific constellation of clinical and pathologic risk factors, both patients are considered to be on the lower end of AJCC v8 and BWH risk assessment (T1 and T2a, respectively) and yet both have at least one very high risk factor by NCCN (poor differentiation).

While 40-GEP was not available at the time that these patients were seen clinically, when these patient’s tumors were tested during the clinical validation study, one patient received a Class 1 result, and one patient received a Class 2B result. Due to the independent nature of the 40-GEP test through multivariate analysis, the risk inferred by a Class 2A and Class 2B result can be added to the risk inferred by other clinicopathologic risk factors to improve clinical risk assessment of regional/distant metastasis. In the patient with the Class 2B result, the patient’s risk of metastatsis can be calculated to be 10.2 times greater than the baseline risk of an AJCC T1 stage tumor/BWH T2a tumor (Ibrahim et al. 2022). The patient who received the Class 2B result went on to experience a recurrence and metastasis of his cSCC and passed away due to disease progression. In stark contrast, the patient whose tumor returned a Class 1 result had no evidence of disease after 4-years of clinical follow-up.

Interestingly, prior to the existence of the 40-GEP test, I was more clinically concerned about the patient whose tumor received the Class 1 result – likely influenced by my intraoperative assessment of the patient’s tumor. In fact, I recommended adjuvant ART to this patient, which he and his family declined after a complex discussion of the risks/benefits of ART and his personal preferences. Had the 40-GEP test information been available to me at the time of medical decision-making following Mohs surgery, I would have been able to refer the patient with the Class 2B result for surveillance imaging and multidisciplinary care conversation – that would include consideration of ART – which when provided to patients with a high risk of metastasis is known to significantly improve patient outcomes. (Ruiz, et al. 2022). Equipped with the Class 1 result for my other patient, I would have been able to inform the shared decision making with my patient and his family around the need for ART and would have recommended – based on validation data – that the patient could safely forgo ART in favor of the recommended close clinical follow-up. These management decisions are supported by published clinical utility studies in which 196 clinicians have indicated utility of Class 1 and Class 2B result to de-escalate and escalate clinical management, respectively, based on improved estimation of risk of metastasis. (Hooper, et al. 2022; Litchman, et al. CMRO 2020)

The 40-GEP test has proven valuable in identifying subsets of patients with low-stage tumors that carry a significant risk of metastasis. This information is crucial for Mohs surgeons who are concerned about missing patients with low-stage disease, but high metastatic risk not captured by evaluation of the clinicopathologic factors alone. Data presented at Fall Clinical NPPA Annual Conference in June 2023 highlights the ability of 40-GEP Class 2A/2B results to identify 70% of patients originally identified with low stage disease that went on to experience a metastasis in the validation cohort for 40-GEP. (Somani, et al. Presented at Fall Clinical NPPA 2023) Identifying these patients allows us to find those who would benefit from a risk-aligned escalation of their treatment intensity (e.g., consideration of more frequent follow-up and consideration of surveillance imaging).

It is important to discuss assertions made by the proposed LCD makes in regards to missing information including tumor depth and diameter in the validation studies for 40-GEP that potentially undermines the value of additional risk stratification in this cohort. “Understaging” of primary cSCC is a known reality of the current clinical scenario. The foundational academic studies used to establish and validate current staging systems themselves also contain significant levels of data missingness on key factors; this is best evidenced by the fact that there was 25% missingness of clinicopathologic risk factor data in a recent high-impact study comparing AJCC and BWH staging systems that was cited in the proposed LCD section titled, “Approaches to Risk-Stratification and Staging” (Karia, et al. 2018). Importantly, the factor with the highest degree of missingness in this study was tumor depth, (150/636) which is a factor incorporated into the BWH staging system. Yet, the conclusions of this paper are accepted, and included in the LCD as foundational to current approaches to risk stratification and staging. Suffice it to say, the presence of missing clinical or pathologic risk factors does not in any way invalidate the cohort where 40-GEP was developed and validated that is grounded in the clinical reality where it is not always possible to obtain information on every clinical risk factor for patients with SCC.

Moreover, it is critical to understand that the staging in the Castle validation cohort, contained a meticulous effort to obtain complete clinical staging information from all cases in the cohort. (Wysong 2020) The study was grounded in 100% on-site monitoring of samples included in the study to ensure a comprehensive assessment of risk factors known both before and after definitive surgery and independent assessment by independent pathologists were conducted to verify and complete missing histopathologic information. In the event of discordance between the second and original pathologist, the highest risk factor result was used. However, the tumor depth and diameter, assessed as missing in the LCD, cannot be retrospectively obtained from independent pathology review. Because of the thoroughness of monitoring and data collection, the cohort used to validate 40-GEP is high quality, and the staging reflected in the study is better than what is routinely achieved in the clinic today. The staging obtained in these specimens certainly reflects a higher level of clinical and pathological information than would have been available to clinicians at the time they were making treatment decisions. Finally, sensitivity analysis performed in the initial validation study confirmed no statistical impact of missingness of data on the significance of 40-GEP clinical validation results. (Wysong 2020) Thus, the potential for understaging in the validation cohort articulated by the LCD merely reflects the actual clinical scenario for cSCC tumors within the intended use of 40-GEP and does not reduce the published clinical validity or utility of this test or the application to real-world clinical practice.

The current literature that supports risk stratification and staging based on clinical and pathologic factors alone is based on retrospective data sets that contain a significant amount of missing data on some of the risk factors incorporated into staging. The DecisionDx-SCC test is the only test that accurately identifies cases deemed low risk by staging, and its validity has been demonstrated in various cohorts, including those treated with Mohs surgery. As skin cancer experts, my colleagues and I are trained to integrate the 40-GEP test results into existing risk assessments based on clinicopathologic factors to make more informed decisions for our patients. The 40-GEP test has met the thresholds for analytic validity, clinical validity and clinical utility for patients and it is for these reasons that I use the 40-GEP test to inform management decisions for my patients who meet the intended use of the test.

In conclusion, the limitations of current staging systems and the cases I have presented emphasize the need for improved risk stratification in high-risk SCC patients. Failure to cover the 40-GEP test in the LCD would negatively impact the care of Medicare beneficiaries. I strongly urge you to reconsider and include this test in Medicare coverage.

Figures, tables, and references were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to WPS:

I am writing in regard to a subject of great importance to me, and to the patients for whom I provide care.

Patients diagnosed with cutaneous squamous cell carcinoma (SCC) may be classified as low or high risk for poor outcomes based on clinical characteristics or tumor staging systems. As a clinician, my experience helps guide my patient management decisions regardless of whether I assess only clinical and pathological features of the patient and tumor or utilize a tumor staging system. Nevertheless, some patients are classified as low risk for disease progression and go on to develop metastasis. There is an unmet clinical need for a more accurate method of predicting a patient’s risk of metastasis from SCC, so patients can be managed appropriately to help prevent metastasis which can lead to death. The 40-gene expression profile (40-GEP) test has been clinically validated in a population of patients with SCC and one or more risk factors, and the test has been shown to have significantly independent prognostic value for metastatic risk assessment. For patients diagnosed with SCC and having one or more risk factors (the 40-GEP intended use population), the 40-GEP test can be used alongside clinicians’ preferred methods of risk classification and within national guidelines to improve metastatic risk assessment and determine the best management plan for the patient.

The current draft LCD incorrectly states that a Class 1 result is more often associated with a tumor with 1 risk factor, and therefore confirms a low risk of metastasis. However, this is a misinterpretation of the data. First, this test is not indicated for patients with low-risk SCC or patients who have an SCC without identifiable risk factors; all patients eligible for GEP testing have an elevated risk of metastasis and according to guidelines are eligible for more intensive treatment or follow-up (Wysong, et al JAAD 2020; Ibrahim, et al Future Oncology 2021; NCCN Guidelines for Squamous Cell Carcinoma v1.2023). Tumors that are T1 by BWH staging included in the study (those who have no BWH risk factors) have presence of other risk factors considered high risk or very high risk by NCCN in order to qualify for testing. Thus, there are degrees of risk even embedded in T1 patients and this resonates for my own clinical practice where I am at times concerned about a subset of patients who are classified as T1 by BWH staging and AJCCv8, but who have other well-known risk factors that are currently included in the NCCN or were included in AJCCv7 staging (eg: tumor depth >2mm but less than 6mm threshold in BWH, perineural invasion in a small nerve, or even multiple small nerves, below the 0.1mm threshold in BWH, or location on the ear or non-glabrous lip), (AJCC Cancer Staging Manual 7^th Edition, NCCN Guidelines for Squamous Cell Carcinoma v1.2023).

There is significant value in improving the accuracy of metastatic risk prediction for patients with both NCCN high risk factors and very high-risk factors. In this context, it is crucial to rely on a correct interpretation of the multivariate analysis that demonstrates independence of risk prediction of the 40-GEP from clinical and pathologic risk factors and SCC staging. Moreover, validation studies consistently demonstrate that a Class 1 result is associated with half the metastatic risk of the cohort, even in lower stage tumors. For example, in Ibrahim et al. Future Oncology 2021, BWH T1 tumors were associated with an overall metastasis rate of 9.5%. However, Class 1 tumors were associated with a metastasis rate of 3.4% (Supplementary Table 2). This is a clinically meaningful reduction in risk of metastasis (which can translate into cost savings in terms of surveillance for such patients) and does not represent confirmation of an already low risk of metastasis. This added value of Class 1 results ignores the fact that in this same study, 41% of patients with BWH T1 tumors received a Class 2A or Class 2B result, which was associated with a 1.5x and 4x higher risk of metastasis than the overall T1 patient population, respectively.

The LCD should be revised to include correct analysis of the impact of Class 1 results in the context of lower-risk disease, which includes both the baseline level of risk in these patients, the reduction of metastasis risk inferred by a Class 1 result, and the accurate reporting of the number of lower stage patients that receive a Class 2A/2B result from published papers that validate the 40-GEP test.

Overall, the draft LCD fails to recognize the marked limitations of current staging approaches in SCC that are based on the presence or absence of a limited set of risk factors that do not completely characterize a patient’s metastatic risk. Recent presentations confirm that patients initially diagnosed with T1 tumors make up ~30-50% of the metastatic events (Somani et al. presented at Fall Clinical Derm for NP/PA, June 2023). In practice, I encounter T1 tumors that harbor multiple known NCCN high-risk factors yet are considered low-stage by current BWH/AJCC staging. Clinicians are clearly continuing to use the risk stratification provided by systems that use clinical and pathologic factors and for patients diagnosed with SCC with one or more risk factors. The 40-GEP is a much-needed prognostic test to help refine my patients’ risk stratification to inform my downstream management decisions that include decisions such as frequency of follow-up, referral to multidisciplinary care and consideration of imaging or adjuvant radiation therapy. I strongly encourage the authors of this draft policy to more closely scrutinize the accuracy of risk stratification provided by current staging approaches that rely only on pathology and clinical information. I believe that published literature has already demonstrated the ability of the 40-GEP to meaningfully improve the accuracy of risk prediction using NCCN risk factors or BWH/AJCC staging and in combination with the clear actionability of the test result supports positive coverage under the final LCD.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to WPS:

The DecisionDx-SCC test by Castle Biosciences, Inc. can provide additional risk stratification of patients with high-risk cutaneous squamous cell carcinoma (cSCC) who may be categorized as having a high-stage tumor without bone invasion (e.g., AJCC8 stage T3, BWH stage T2b) or as localized, “very high risk” by NCCN guidelines. Management recommendations for this heterogenous group of patients are admittedly broad within current guidelines and having a test that can ‘fine-tune’ each patient’s individual risk assessment allows for more personalized care that is better aligned with actual risk of poor outcomes. Within patients designated by NCCN or staging as “very high risk”, only some of these patients have a high biological risk for metastasis, while others may have a lower risk for metastasis and, therefore, should be managed differently. When physicians treating patients with cSCC are able to direct our existing –and proven beneficial - clinical management strategies to the patients who are at the greatest risk of disease progression we can provide the greatest benefit and outcome improvement for our patients.

The draft LCD appears to make the assumption that all of these patients have an extremely high risk of metastasis already designated, and therefore are currently treated with the highest intensity therapy possible; however, this reflects a poor understanding of patient management in this disease state. First, there are many patients staged at very high risk by NCCN, AJCC8 or BWH staging systems that do not have a high biological risk of metastasis and who are currently overtreating for their disease. This is due to the well-known limitation of clinicopathologic risk stratification systems with low overall PPV (Ruiz et al. JAMA Derm, 2019 and Jambusaria et al. JAMA Derm, 2013). It is also the case that there are many patients with high stage disease who are not managed with the full intensity of treatment available (eg: frequent follow-up, multidisciplinary tumor board, nodal imaging, adjuvant radiation therapy, etc). Further, the LCD reflects a myopic view of clinical management strategies currently employed as the reference survey from Patel et al. 2022 and includes an overwhelming majority of physicians based in academic practice as opposed to a representative combination of academic and community practice. The current issues of both overtreatment and undertreatment just mentioned are more accentuated in community practice – where a risk-factor based approach to risk stratification is most commonly utilized – and it is important for Medicare to consider the clinical practice patterns in the community in addition to academic centers as many Medicare beneficiaries will never be seen in academic practices.

Ibrahim et al. (2021) recently demonstrated that Castle’s DecisionDx-SCC test further stratified patients with high-risk SCC as having low (Class 1), moderate (Class 2A), or high (Class 2B) risk for metastasis. This refinement in risk stratification was demonstrated within tumor staging systems and within NCCN high-risk and very high-risk subsets within the validation cohort in that paper. This fact does not appear to be reflected in the draft LCD. For high-stage tumors (AJCC8 T3/T4 or BWH T2b/T3), metastasis rates were 12.5% or 15.8% for Class 1 results, 35.7% or 36.7% for Class 2A results, and 83.3% or 71.4% for Class 2B results, respectively. Whereas, without the test, these high-stage tumors had an overall metastasis rate of 30.0% for AJCC8 T3/T4 and 33.9% for BWH T2b/T3. Thus, in these high-stage tumors, the test can identify patients with less than half the risk suggested by clinical and pathological factors alone with a Class 1 result, and patients with twice the risk of metastasis, that is, twice the PPV, with a Class 2B result. Thus, the test can provide supplementary information about metastatic risk that can help guide decisions for escalating or de-escalating management of patients with high-stage or high-risk tumors. This information is included in Supplemental Table 2 of the Ibrahim et al. 2021 paper but does not seem to be included in the evidence review summarized in the draft LCD.

While the suggested intended use in the LCD excludes AJCC8 T4 and BWH T3 from testing, note that the breakdown of data published in Ibrahim, et al (Supplementary Table 2), allows for assessment of the test performance when these patients are excluded; the conclusions and significance of risk stratification do not change. Importantly, the DecisionDx-SCC test can help with management decisions for not only patients having low-stage tumors who may need more intense management due to risk for metastasis, but also for patients having high-stage tumors who may need more or less intense management based on metastatic risk. Assumptions in the LCD that Class 2B results occur more frequently in tumors with high stage by risk factor assessment and are therefore not useful are not based in the evidence, as a >30% improvement in PPV is certainly clinically actionable, and clinical utility studies demonstrate physicians do change management in patients with high stage tumors in a risk-aligned manner with GEP results. Thus, statements about the clinical validity limitations of DecisionDx-SCC should be removed from the LCD or substantially revised to reflect the complete picture of published evidence from the Ibrahim et al study.

In conclusion, I urge to finalize this LCD with positive coverage for the Castle test. This test is a critical component to risk aligned management for patients with SCC with one or more risk factors.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to WPS:

I am a board certified dermatologist with a subspeciality certification in cutaneous oncology and micrographic surgery. I spend all of my professional time treating and managing patients with skin cancer. As such, a large number of patients in my practice are seen for the treatment of squamous cell carcinoma. According to the Skin Cancer Foundation, cutaneous squamous cell carcinoma affects 1.8 million Americans every year with an estimated 15 thousand deaths associated with this disease. The incidence is increasing at a rapid rate in this primarily elderly population, and it is critical for physicians treating this disease to have established, effective options for guidance of treatment decisions.

Current methods to assess risk in these patients are limited in accuracy. Because of this, there exists inconsistent use of risk stratification from practice to practice, and inconsistent treatment plans (based on assessment of risk) from practice to practice. New tools, such as the 40-GEP test provided by Castle Biosciences, are objective advances in personalized medicine that provide individual risk stratification to guide treatment decisions. These treatment decisions have been happening for these patients every day, across the country, in the absence of accurate risk stratification tools. Yet, the current proposed LCD seems unaware of the reality clinicians treating these patients face. The draft LCD sets up unrealistic standards for the 40-GEP to achieve coverage, including creating an established and uniformly adopted system of consistent treatment plans in a disease state where this type of adoption and consistency are lacking. What is more, the current body of evidence for risk stratification based on clinical and pathologic factors is based on retrospective cohorts without demonstration of specific outcome improvement and the threshold for improving the current paradigm of risk stratification must be grounded in the quality of evidence that currently exists in a particular disease state. In this light, the 40-GEP tool (and the published evidence supporting the test), represents huge progress in being able to more accurately predict a patient’s individual risk of metastasis and that can help better risk-align management strategies for patients with this impactful cutaneous malignancy. The test shares the same level of evidence, if not higher, than the level of evidence supporting clinical and pathologic factors for risk stratification and furthermore, can be used in combination with existing risk stratification systems due to the demonstrated independence from other systems in Ibrahim et al.

Of course, I take the incorporation of a new product into my practice very seriously and have read in great detail the validation papers published related to the 40-GEP test by Wysong, et al., as well as another validation paper by Ibrahim et al. I have found the quality of these studies to be excellent. I was engaged with the results and compelled to utilize the test for patients that I find have complicated histories and/or are currently presenting with risk factors that are challenging to create effective treatment plans for. The 40-GEP has been instrumental in assisting me with these cases. I use the 40-GEP test result in a model of shared decision making together with my patients. ­­­The resulting changes in management decisions I feel have greatly impacted my patients in a positive nature. Additionally, it has allowed me to manage my patients in a more cost-effective manner for both them and their insurance who would otherwise be charged for adjuvant treatment that they may not have needed. I know having Medicare coverage for DecisionDx-SCC in the future will only increase the benefits seen by my patients.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA and WPS:

I appreciate the opportunity to discuss my perspective on your draft LCD.

The DecisionDx-SCC GEP test reveals important risk information that is not captured in current systems of risk stratification that are based on clinicopathologic factors alone. Noncoverage of DecisionDx-SCC will limit access of the test to my Medicare patients and is detrimental to patient care. Despite careful review of the clinical and pathologic risk factors that I routinely evaluate, I frequently encounter patients for whom actual metastatic risk is uncertain, and who benefit from the additional risk stratification for their SCC diagnoses provided by DecisionDx-SCC. It is in these patients where elimination of the DecisionDx-SCC test will result in an under or an over treatment of my patients.

A key question that arises when treating patients in my clinic is how to effectively and accurately assess the likelihood of metastasis from their disease. The current tools we employ, although useful, are imperfect. One major limitation is that these tools rely on binary risk factors (e.g., present vs. absent), which dilutes our ability to assess risk along a continuum. For example, in the context of staging, only perineural invasion (PNI) of a nerve with a diameter 0.1 or greater is technically counted as positive PNI; therefore PNI would be assessed as negative if there was involvement of a nerve measuring 0.01 mm or if there was involvement of a nerve measuring 0.09 mm, even though the difference in risk represented is clinically different in these two scenarios. Further, the staging systems inappropriately weigh all risk factors equally and the presence of any single risk factor can elevate a tumor to a higher stage. Finally, the majority of patients predicted to have a high risk of metastasis by current risk assessment systems will not experience a metastatic event. This poses a challenge in determining the appropriate treatment pathway for each individual patient, and specifically, when it is safe to de-escalate intensity of management or forgo a specific post-surgical treatment. These limitations are the reasons why many clinicians do not use formal staging systems, but instead, rely on the assessment of individual risk factors to assess risk. The current proposed LCD overestimates the use of formal staging systems by relying on a single survey composed of primarily academic, tertiary centers that disproportionately use these systems and ignoring other studies comprised primarily of those that diagnose and manage more then 90% of all patients diagnosed with SCC. (Patel 2022, Litchman 2020)

It is important to note that treating dermatologists often lack complete information about the presence or absence of risk factors when making critical decisions due to the absence of standardized reporting requirements (e.g., lack of CAP synoptic) by pathologists for primary cutaneous SCC and because most patients in the US are treated with Mohs micrographic surgery (MMS) where the focus is primarily on margin control and not on complete identification of all pathologic factors incorporated into staging. These shortcomings highlight the clinical need for an objective tool to identify patient risk of metastasis more accurately.

Current risk assessment systems that use only clinical and pathologic risk factors lack the granularity to accurately identify patients with low biological and high biological risk of metastasis. In the expanded validation cohort published by Ibrahim at all, the test is shown in the context of BWH and AJCC subsets and demonstrates significant stratification of risk even within these stages. (Ibrahim 2021) This illustrates the utility of the tool; because it is statistically independent of other risk factors and staging systems and can be combined with all known information about the patient to provide a more complete, comprehensive risk assessment than is otherwise available.

As demonstrated in the figure below, NCCN guidelines have a broad range of management options available for patients with one or more NCCN high risk or very high-risk factors, with wide ranges for follow-up intervals, nodal assessment options (from palpation to imaging to SLNB) and consideration or recommendation of adjuvant radiation therapy (ART). (NCCN Guidelines for Squamous Cell Skin Cancer v1.2023) It is critically challenging that the broad range of options are informed by limited guidance in the guidelines themselves on the most appropriate course of action for each clinical scenario. While surgery remains the primary treatment modality for all patients, it becomes essential to determine when/if additional interventions following definitive surgery are medically necessary, including the introduction of multidisciplinary consultation and adjuvant radiation therapy in the appropriate context. NCCN guidelines include the statement that, “The NCCN Guidelines are a statement of evidence and consensus of authors regarding their views of currently accepted approaches to treatment. Any clinician seeking to apply or consult the NCCN Guidelines is expected to use independent medical judgement in the context of individual clinical circumstances to determine any patient’s care or treatment.” Thus, the guidelines highlight the importance of evaluating the individual patient, cost/benefit consideration of treatment modalities in the context of risk, and shared decision-making with the patient. In this setting, objective risk information to guide decision-making is a critical unmet need.

As an example, in my own practice, I recently cared for a patient with cSCC staged as BWH T2b within the NCCN very high-risk group that would, according the NCCN guidelines, be recommended for adjuvant radiation therapy. However, following surgical removal of the tumor with negative margins and multidisciplinary consultation, questions remained whether or not the patient needed ART and use of imaging in their surveillance plan. During shared decision-making, the patient and the patient’s family expressed concern about the burden ART treatment would be to the patient, as ART treatment typically requires 20-25 treatments. If NCCN and BWH are the only risk-stratification tools available, I have no additional clinical data to inform my decision making, and as a result, would not be comfortable making evidence-based recommendations for my patient to avoid ART in this scenario.

It is in these types of clinical scenarios where DecisionDx-SCC has been immensely helpful for me and my patients in the model of shared decision making that incorporates both GEP results and clinicopathologic factors. The DecisionDx-SCC test enables further subdivision of patients based on their risk of metastasis and provides a breakdown of their individual risk levels in the context of the existing clinicopathologic risk factors. In the case presented, the patient received a Class 1 result and based on this GEP result interpreted in the context of the patient’s other known risk factors, I was able to use shared decision making to reach an evidence-based clinical plan together with my patient and his family to avoid ART in favor of watchful waiting. This is a scenario where DecisionDx-SCC avoided overtreatment of a patient with a lower biological risk of metastasis than would be anticipated for a patient with the same clinical and pathologic risk factors and for whom less aggressive treatment was appropriate based on the added GEP information. At now 15 months of follow-up, this patient is doing well with no evidence of disease. Including GEP results into my risk stratification approach results in an improvement in the accuracy of my risk prediction that directs specific management changes for the benefit of my patients.

This utility is not just an anecdote from my own practice. The study published by Hooper, et. al. used real-world cases and collected management data from clinicians who clinically order the test before and after DecisionDx-SCC results. Case 3 in the study represents a patient similar to mine, with a patient over the age of 80 with a BWH T2b cSCC in the head and neck region. In this study, approximately 35% of clinicians surveyed de-escalated adjuvant radiation therapy based on a Class 1 result. (Hooper 2022). Without the Decision Dx-SCC, I would lack an objective measure to allow safe de-escalation of care when appropriate.

The LCD mistakenly asserts that it is unclear how to combine the test result with staging or known risk assessment systems and infers an inappropriate assessment of risk by clinicians using the test, with concern that a Class 1 result will be interpreted as inappropriately low risk in the setting of multiple risk factors. These assertions are unfounded; data published in Ibrahim, et al. show not only the independence of the test result, but also the expected metastasis rates within different risk assessment systems when GEP results are layered on top of known information. (Ibrahim 2021) In addition, clinical utility studies including Hooper et al. 2022 and Litchman et al. 2020 show specific management changes, including adjuvant radiation therapy, and how they are escalated and de-escalated in response to test results in the context of a patients known clinical and pathologic risk factors. (Hooper 2022, Litchman 2020) Thus, concerns raised in the LCD are unfounded, are not informed by experts in treating this disease, do not reflect the reality of clinical practice, and are not informed by the published data.

Finally, my multidisciplinary team uses Castle Biosciences GEP as one of our key risk factors in making patient centric care recommendations. I urge MolDX to re-evaluate the errors and inaccuracies present in this draft LCD and finalize with DecisionDx-SCC as a covered service for Medicare beneficiaries with SCC and one or more risk factors to inform the range of downstream management decisions after definitive surgery has concluded.

Figures and references were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to WPS:

The proposed LCD MolDX: Molecular Biomarker Testing for Risk Stratification of Cutaneous Squamous Cell Carcinoma contains many inaccuracies that poorly reflect the clinical care of these patients, and as such draws faulty conclusions about the value of the SCC prognostic gene profile assay (DecisionDx-SCC, Castle Biosciences).

For example, the section that describes treatment and surveillance intensity by risk stratification group is lacking in detail. The discussion of treatment pathways offered by NCCN guidelines is incomplete and focuses only on treatment options for NCCN very high-risk tumors and the “recommended” course of treatment. There is no discussion of treatment options for NCCN high risk tumors, which are notably in the same treatment path as NCCN very high-risk tumors and sets up a false sense that most aggressive treatment options are reserved only for NCCN very high-risk tumors. This is compounded by the fact that no treatment options that are listed in the LCD for “consideration” or discussion as part of a patient management strategy, which is misleading as areas for “consideration” of a particular therapy are clearly outlined in NCCN. For example, in the case of adjuvant radiation therapy (NCCN v1.2023 - SCC-3/3A), this specific management decision can be guided by the identification of “other poor prognostics features”, or if a tumor has a “high risk for regional or distant metastasis” which both rely on the interpretation of the treating physician and outline why a prognostic test like DecisionDx-SCC is needed to better identify a patient’s individual risk of metastasis. The LCD lacks any discussion of the use or even consideration of use of adjuvant radiation therapy in these patients, which takes place today for patients with high-risk or very high-risk tumors alike.

This focused interpretation of the management pathways they outlined from the current NCCN guidelines for Cutaneous Squamous Cell Carcinoma is not appropriate, and the LCD should more completely outline the clinical decision points following definitive surgery of high-risk SCC where there is clearly substantial room for treating physician interpretation of a patient’s risk for regional and distant metastasis to inform management decisions that improve outcomes for our patients. This is essential for accurately portraying the treatment decisions being made by physicians today for patients with high-risk SCC. A more comprehensive reading of the NCCN guidelines and the broad treatment pathways available to the entire set of high risk and very high-risk patients informs understanding for why improved metastatic risk stratification is essential to provide risk aligned management for patients with high risk SCC. Proper and complete interpretation of NCCN guideline treatment recommendations also strongly supports my perspective that the DecisionDx-SCC test should receive positive coverage in the final LCD as it is clinically validated and clinically useful for the patients with high-risk SCC and informs my management decisions in these challenging cases.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to WPS:

The DecisionDx-SCC test developed by Castle Biosciences, Inc. has been validated using clinical standard operating procedures and is clinically available for testing primary tumors from patients diagnosed with cutaneous squamous cell carcinoma (SCC) with one or more factors that are risky for metastasis. While it is not intended for testing of all patients with SCC, it is indicated for predicting metastatic risk in patients that are already at elevated risk for metastasis and can further stratify into low, moderate, or high risk for developing metastasis. The Castle test analyzes gene expression of the primary tumor and the test results have demonstrated significant prognostic value for determining metastatic risk with improved accuracy compared to traditional methods of risk analyses. Importantly, the test adds information to available risk factors and risk assessment systems, and doctors like myself interpret the result in the context of the specific risk factors/stage. The suppositions in the proposed LCD that the test result does not add significant value to staging systems is not supported by the evidence and is a critical flaw in the evaluation of the test. Further, the insinuation that the 40-GEP test results would be interpreted outside the context of the known clinical/pathologic risk factors is not supported in the published clinical utility data (Farberg, et al. CMRO, Hooper et al. Skin) and is not how I use the 40-GEP test for my patients.

When comparing the accuracy of tests or risk factors, it is important to remember that comparing the accuracy (I.e., PPV or NPV) only provides context for the clinical actionability. For example, a similar PPV between two systems means that a similar proportion of patients with a high-risk result will experience a poor outcome, and patients should be treated with that likelihood of progression in mind. It does not, however, assess the value of combining the information and therefore misses how clinicians are using and will continue to use the test in clinical practice. Thus, the similar PPV between from Castle 2 (2A and 2B result combined) and BWH staging does not mean that these systems are equivalent in their risk stratification or are identifying the same patients as high risk. In fact, this is not the case, as demonstrated by the multivariate analysis showing independence of prediction in multiple studies (Wysong, et al. JAAD, Ibrahim et al. Future Oncology). In fact, data showing breakdown of BWH T stage and GEP class call show that the test identifies 75% of metastatic events in BWH T1/T2 tumors that are assumed to have lower metastatic risk, and 1 in 5 patients with a Class 2A or Class 2B result experience metastasis. (Ibrahim et al. 2021, Supplementary Table 2). This data was not discussed in the LCD, though the paper was cited, which improperly characterizes the value of data in this publication. Thus, the statement in the LCD that the test does not add value because of improper interpretation of similar “out-of-clinical-context" accuracy metrics should be corrected. Again – the clinical need in this disease state is to improve upon current staging with the addition of a highly validated gene expression profile test and NOT to remove all staging considerations based on clinical and pathologic factors; it is unclear why the draft LCD does not evaluate the test’s ability to improve on the accuracy of staging to then inform established management strategies and instead focuses on direct comparisons of accuracy between the two approaches outside the context of where clinicians use the test results.

As a healthcare provider and patient advocate, I support the use of this test for the intended use population for obtaining additional information about an individual’s risk for metastasis for better-informed clinician-patient decision-making. The advantage of using this molecular tool along with tumor staging or assessment of clinical and pathologic factors considered high risk for poor outcomes, while following national guideline recommendations, is inherently self-evident for making established and risk-aligned management decisions to achieve the most desirable patient outcomes. This tool is critical for improving patient care and should be a covered service for Medicare beneficiaries.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to WPS:

According to the Skin Cancer Foundation, cutaneous squamous cell carcinoma affects 1.8 million Americans every year with an estimated 15 thousand deaths associated with this disease. The incidence is increasing at a rapid rate in this primarily elderly population, and it is critical for physicians treating this disease to have established, effective options for guidance of treatment decisions.

Current methods to assess risk in these patients are limited in accuracy. The result of this is inconsistent use of risk stratification from practice to practice, and inconsistent treatment plans (based on assessment of risk) from practice to practice. New tools, such as the 40-GEP test provided by Castle Biosciences, are objective advances in personalized medicine that provide more information, better than what is currently available, to guide treatment decisions. These treatment decisions have been happening for these patients every day, across the country, in the absence of accurate risk stratification tools, yet the current proposed LCD seems unaware of the reality clinicians treating these patients face. The draft LCD sets up unrealistic standards for the 40-GEP to achieve coverage, including creating an established and uniformly adopted system of consistent treatment plans in a disease state where this type of adoption and consistency are lacking. What is more, the current body of evidence for risk stratification based on clinical and pathologic factors is based on retrospective cohorts without demonstration of specific outcome improvement and the threshold for improving the current paradigm of risk stratification must be grounded in the quality of evidence that currently exists in a particular disease state. In this light, the 40-GEP tool (and the published evidence supporting the test), represents huge progress in being able to more accurately predict a patient’s individual risk of metastasis and that can help better risk-align management strategies for patients with this impactful cutaneous malignancy. The test shares the same level of evidence, if not higher, than the level of evidence supporting clinical and pathologic factors for risk stratification and furthermore, can be used in combination with existing risk stratification systems due to the demonstrated independence from other systems in Ibrahim et al.

Of course, I take the incorporation of a new product into my practice very seriously and have read in great detail the validation papers published related to the 40-GEP test by Wysong, et al., as well as another validation paper by Ibrahim et al. I was engaged with the results and compelled to utilize the test for patients that I find have complicated histories and/or are currently presenting with risk factors that are challenging to create effective treatment plans for. The 40-GEP has been instrumental in assisting me with these cases - I use the 40-GEP test result in a model of shared decision making together with my patients. The resulting changes in management decisions I feel have greatly impacted my patients in a positive nature and I know having Medicare coverage for DecisionDx-SCC in the future will only increase the benefits seen by my patients.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to WPS:

Improvements in prediction of risk for patients with cutaneous squamous cell carcinoma is a critical unmet need. Currently, the best staging system that has been developed to date is the Brigham and Women’s Hospital staging system, which counts the presence of 4 specific risk factors to identify groups patients with distinct metastatic risk. While this is a straightforward and easy system used to stage SCC, it has substantial room for improvement in accuracy for predicting metastatic risk. In fact, the average reported positive predictive value of the BWH staging system is between 25 and 40%, meaning that the majority of patients identified as high risk do not experience disease progression. In some recent studies, approximately 30% of BWH lower stage patients (T1 and T2a) go on to have a metastatic event. What is more, some of the staging factors are not always apparent after diagnostic biopsy or even definitive Mohs surgery where the focus is often on margin control of the tumor and not on identification of specific clinical or histologic risk factors. For example, tumor depth measurements (included in BWH staging) are not always possible to establish histologically from tangentially cut Mohs sections without permanent sections from a Mohs debulk specimen.

The introduction of the 40 gene expression profile has demonstrated improvement in the prediction of my patients’ biological risk of metastasis and has refined risk estimations in a clinically meaningful way. In Supplemental Figure 2, published by Ibrahim and colleagues in Future Oncology in 2021, it is evident that within each BWH stage, the metastatic risk associated with a Class 1 is similar to the metastatic risk of disease one stage lower, while the metastatic risk of a Class 2B far exceed the of the general population of T2b/T3 tumors, which is demonstrated by multiple studies cited in the LCD to be a clinically actionably level of risk. (Patel 2022, Arron 2021). Thus, the evidence has established that the 40 gene expression profile can assist in determining which of the low stage patients are more likely to have a metastatic event AND that this level of risk is considered clinically actionable. When combined together, the likelihood of placing patients in the correct high or low risk bucket increases, which has clear positive implications for improved net health outcomes in these patients. Finally, the test is validated on the diagnostic biopsy specimens that are routinely obtained as part of the diagnosis of primary cutaneous SCC and is able to provide independent risk stratification information from the critical diagnostic sample used to establish the diagnosis for patients with high risk SCC. Interpretation of this important data directly answers questions raised in the LCD regarding how to interpret risk of metastasis in the context of staging and should be discussed in detail in the LCD. Given the strength of this data, I request to include discussion of this data in the LCD, remove statements that the added value to current risk factors or staging is unclear, and finalize the policy with this important test as covered for Medicare beneficiaries.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to WPS:

Gene expression profiling and other molecular tools have been shown to provide valuable prognostic information for management of patients with many different types of cancer (e.g., breast,^1–3 prostate,^4,5 and thyroid cancer;^6,7 uveal melanoma;^8,9 and melanoma ^10-12). The 40-gene expression profile (40-GEP) test by Castle Biosciences, Inc. has been clinically validated for identifying patients with localized, high-risk cutaneous squamous cell carcinoma (cSCC) as being either at low, moderate, or high risk for developing metastasis within 3 years of initial diagnosis.¹³ The test is intended for use with cSCC patients who have one or more risk factors (i.e., high-risk localized, but not advanced disease). Ibrahim et al.¹³ reported that the test stratified a subset of patients in a clinical validation cohort as low risk for metastasis (40-GEP Class 1) and this subset had metastasis rates near those of the general cSCC patient population (i.e., patients with no risk factors, 2-6%),^14–19 while they also reported stratification of a high-risk subset (40-GEP Class 2B) with metastasis rates that were >50%. The authors also described and demonstrated the advantages of combining results from the 40-GEP test with tumor staging and traditional clinicopathologic risk factor assessment for fine-tuning risk classification in high-risk cSCC. More specifically, the test was able to further refine risk stratification within individual and binary tumor stages (i.e., AJCC8 T1-T4 or BWH T1-T3; and binary low- and high-stage AJCC8 T1/T2 and T3/T4 or BWH T1/T2a and T2b/T3) as well as national guideline (NCCN) classifications for high and very high risk, as shown in multivariate analysis. Importantly, this analysis shows that the test adds information to staging systems. This point seems to be missed by the draft LCD that makes frequent allusions to the fact that clinical and pathological factors remain important predictors of risk. This is absolutely true, and in no publication or intended use document does the test claim to replace staging or to be used outside of the context of staging. Multiple figures and tables in the Ibrahim et al. publication demonstrate that within other risk assessment systems, the test stratifies within risk subsets, adding further information and granularity on risk that is otherwise unobtainable. The test should be used with, not instead of other risk factor information to provide the most comprehensive assessment of a patient’s true risk.

Use of the test to further refine risk for patients staged as T1 to determine if they are, indeed, low risk for metastasis and patients staged as AJCC8 T3 or BWH T2b to determine if they are, indeed, high risk for metastasis remains essential given that tumor staging fails to identify over 30% of tumors as high stage a time of initial diagnosis that actually develop metastasis^20–22 and more than 60% of those staged as high risk at time of initial diagnosis do not actually go on to metastasize.^23–25 It should also be noted that the 40-GEP test provides valuable information for those patients who are considered to have tumors at an intermediate stage of risk (i.e., AJCC8 T2 or BWH T2a). Further refinement of risk for these patients is equally as important for their risk-appropriate management as it is for patients with tumors classified at a lower or higher T stage (T1 and T2b). Review of the important data published in the Ibrahim et al study – including the published paper supplement - and inclusion in the LCD is important to clearly articulate how the test adds value to staging based on the clinical and pathologic risk factors.

References were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to WPS:

According to the Skin Cancer Foundation, cutaneous squamous cell carcinoma affects 1.8 million Americans every year with an estimated 15 thousand deaths associated with this disease. The incidence is increasing at a rapid rate in this primarily elderly population, and it is critical for physicians treating this disease to have established, effective options for guidance of treatment decisions.

Current methods to assess risk in these patients are limited in accuracy. The result of this is inconsistent use of risk stratification from practice to practice, and inconsistent treatment plans (based on assessment of risk) from practice to practice. New tools, such as the 40-GEP test provided by Castle Biosciences, are objective advances in personalized medicine that provide more information, better than what is currently available, to guide treatment decisions. These treatment decisions have been happening for these patients every day, across the country, in the absence of accurate risk stratification tools, yet the current proposed LCD seems unaware of the reality clinicians treating these patients face. The draft LCD sets up unrealistic standards for the 40-GEP to achieve coverage, including creating an established and uniformly adopted system of consistent treatment plans in a disease state where this type of adoption and consistency are lacking.

What is more, the current body of evidence for risk stratification based on clinical and pathologic factors is based on retrospective cohorts without demonstration of specific outcome improvement and the threshold for improving the current paradigm of risk stratification must be grounded in the quality of evidence that currently exists in a particular disease state.

In this light, the 40-GEP tool (and the published evidence supporting the test), represents huge progress in being able to more accurately predict a patient’s individual risk of metastasis and that can help better risk-align management strategies for patients with this impactful cutaneous malignancy. The test shares the same level of evidence, if not higher, than the level of evidence supporting clinical and pathologic factors for risk stratification and furthermore, can be used in combination with existing risk stratification systems due to the demonstrated independence from other systems in Ibrahim et al.

Of course, I take the incorporation of a new product into my practice very seriously and have read in great detail the validation papers published related to the 40-GEP test by Wysong, et al., as well as another validation paper by Ibrahim et al. I was engaged with the results and compelled to utilize the test for patients that I find have complicated histories and/or are currently presenting with risk factors that are challenging to create effective treatment plans for. The 40-GEP has been instrumental in assisting me with these cases - I use the 40-GEP test result in a model of shared decision making together with my patients. The resulting changes in management decisions I feel have greatly impacted my patients in a positive nature and I know having Medicare coverage for DecisionDx-SCC in the future will only increase the benefits seen by my patients.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA and Noridian:

I am a double-board certified Dermatologist and Mohs. My perspective has been shaped based on my own clinical research and clinical experience managing patients with high-risk cutaneous SCC which I will share with you today. The current draft LCD that non-covers the DecisionDx-SCC misses the mark in several areas, and the purpose of my written comments is to provide my rationale for why the 40-GEP test has met Medicare’s coverage criteria.

While reading the draft LCD, a few main issues stood out that I would like to address:

• Incomplete literature throughout the draft LCD. Several important studies that are critical to understanding the clinical utility of the 40-GEP test are not cited at all or are incorrectly or insufficiently reviewed.
• Missed connection between the 40-GEP published clinical validity data to the published clinical utility data.
• Incomplete understanding of how cSCC patients are actually managed and the associated decision points in order to determine the appropriate management pathway according to a patient’s risk.
• Confusion within the draft LCD about the efficacy of specific management strategies on patient outcomes.
• Misapplication of the ACCE hierarchical framework to evaluate the 40-GEP by requesting non-standard statistics such as monotonicity and homogeneity despite availability of data.

The LCD does not connect the 40-GEP published clinical validity data to the published clinical utility data. What I mean by this is that the multiple validation studies emphasize how DecisionDx-SCC improves the accuracy of risk stratification in the context of staging approaches^1–3 and then the clinical utility studies demonstrate that clinicians are able to combine their evaluation of patient specific clinical and pathologic risk factors together with GEP results to inform specific risk aligned management decisions.^4–6 The statement that “It is unclear how GEP results can be consistently or accurately be interpreted” is the result of incomplete and flawed evidence review and is not grounded in the published evidence.

Discussion of NCCN guidelines is limited in the draft LCD. A physician’s metastatic risk assessment informs a broad range of management changes for SCC patients with high-risk and very high-risk factor tumors (SCC-3 and SCC 3-A).⁷ The draft LCD in some places appears to characterize the management interventions as completely ineffective to improve health outcomes, and therefore, incorrectly concludes that no health outcomes can be informed by improved risk stratification with GEP testing. Specifically, there are statements that SLNB or ART may not improve outcomes – which is in direct contrast to the broad use of these interventions in patients and the fact that these interventions are covered by Medicare today. However, and concerningly, not a single study regarding efficacy of ART is referenced in the draft LCD. I request that the following studies be thoroughly reviewed and included in this section of the draft LCD.

• Ruiz et al. JAAD. 2022 – not cited, despite showing 50% reduction in recurrence with ART treatment of high risk cSCC patients.⁸
• Patel et al. Cancer Med. 2022 – included in other sections but not included in this section despite consensus to use radiologic imaging, SLNB, ART, and increased follow-up in AJCC T3 or BWH T2b due to metastatic risk.⁹
• Zhang et al. JEAVDV. 2021 – not cited despite a meta-analysis of 20 studies resulting in positive impact of ART on cSCC patient outcomes.¹⁰
• Sahovaler et al. JAMA Otolaryng. 2019 – not cited despite a meta-analysis of 21 studies finding ART to improve both overall survival and disease specific death in cSCC patients.¹¹

For imaging as a management strategy, the draft LCD essentially ignores this treatment modality and similar to ART, not a single study regarding imaging is even references. I request that the following studies be thoroughly reviewed and included in this section of the LCD.

• Ruiz et al. JAAD. 2017 – not cited, despite showing significant reduction in poor outcomes for patients that received imaging.¹²
• Mahajan et al. JAAD. 2020 – not cited, despite showing prognostic value of imaging.¹³

More than half of the section related to treatment strategies and intensities is given to consideration of a single survey published by Patel, et al. that includes 156 cSCC-treating physicians. However, important results of this study are omitted from review (as are other survey-based manuscripts). This survey is used to establish that “most” dermatologists and cancer-specialists use staging, and omits the very important datapoint that “a large number of respondents reported they do not use staging systems to help consider whether to perform radiologic imaging, SLNB, post-operative radiation therapy (PORT), adjuvant systemic therapy, or increased follow-up.”⁹ The treatment of patients with high-risk SCC in the US occurs at clinic centers in community practice and academic centers.¹⁴ A critical limitation of the LCD is that it focuses references exclusively to academic medical practice that hide the full range of risk stratification approaches used in the US - meaning risk factor based in addition to formal staging – and as a result minimizes the unmet clinical need in high-risk SCC.

Litchman et al. CRMO 2022, a survey similar to Patel, et. al. and is not discussed in the draft LCD, in which 162 dermatologists rank GEP results with risk factors. Of note, Class 2B was the most impactful risk factor for cSCC metastasis. In order of impact, physicians rank the most impactful risk factors as the following: 40-GEP Class 2B, Perineural invasion (>0.1 mm), Immunocompromised, Depth of invasion, 40-GEP Class 2A. Moreover, this manuscript characterizes how clinicians incorporate the test results to impact specific management options outlined in NCCN guidelines to be considered/recommended based on the risk of metastasis.

The Analysis of Evidence section of the draft LCD asserts, “It is unclear how GEP results can be consistently or accurately interpreted”. This evidence is available in Farberg et al. CMRO 2020, however, the draft LCD reviewed and inappropriately dismissed in LCD as “out of date” due to changes in NCCN guidelines in 2021. However, the changes in guidelines only divided the “high risk” features in this paper into “high risk” and “very high risk” and primary treatment for high risk and very high risk is the same. The update to the guidelines did not change any treatment recommendations and therefore does not invalidate the algorithm that directly answers this question posed in the LCD. There is no basis for failure to completely consider the clinical utility findings contained in Farberg et al. 2020 based simply on the minor shift in NCCN risk groups, with no other changes to risk assessment, including staging systems, or recommendations for management options. Also note this figure assesses the risk of metastasis from validation studies when combined with BWH staging and provides a combined risk assessment to align intensity of management strategies, as outlined by NCCN guidelines. As such, it is clear from this publication how GEP results can be used in the context of clinicopathologic risk factors that result in changes in patient management.

Multiple additional studies and consensus panels have derived nearly identical clinical pathways for management of patients within treatment pathways available in NCCN guidelines.^4–6,16 This consistency from clinical impact studies and expert panels demonstrates that the 40-GEP results are interpreted in the context of known risk factors. The next section points to the specific publications where this has been shown definitively.

Singh et al. 2023 is similar to Farberg et al. 2020 and is updated to include new NCCN high risk versus very high risk. This clinical management algorithm directs management across multiple treatment pathways including surveillance imaging, SLNB, ART, clinical follow-up, and multidisciplinary tumor board. Please refer to the published Figure 1 of the Singh et al Manuscript.

Arron et al. 2020 convened an expert consensus panel to publish recommendations for high risk results and are nearly identical to the preceding two published management algorithms. Again, they take several management recommendations from the guidelines and match them to outcomes of the 40-GEP to provide a risk aligned treatment plan. Please refer to both components of the published Table 2 of the Arron et al. manuscript.

The published clinical utility data demonstrates that clinical use is consistent with algorithms and expert panel recommendations – specifically that the DecisionDx-SCC test results are interpreted appropriately and consistently. Litchman et al. 2020 utilized 2 patient vignettes, both with a BWH T2a stage, and gathered recommendations from 168 physicians on management changes with 40-GEP. Essentially, this study shows that for the same patient with a BWH T2a cSCC – a Class 1 result from the 40-GEP is associated with decreased management intensity and a Class 2B result is associated with increased management intensity. Again, the class result is interpreted in the context of the patient’s tumor to inform appropriate management decisions, and this is aligned with the clinical management pathways recommended in the three papers mentioned above.

Another study that is not completely reviewed in the proposed LCD is from Hooper et al 2020 and published in Cancer Investigation. Although the draft LCD states that it would be prudent to provide “greater detail the patients for whom the 40-GEP led to a change in management” and “the specific change in management” to augment the clinical utility evidence supporting DecisionDx-SCC, there is incomplete evidence review of published clinical utility data that addresses this exact question in Hooper et al. 2020. The LCD cites the clinical ordering data from this paper, but it does not discuss the most impactful data from the figure that describes the clinical utility aspect of the manuscript (please refer to published Figure 3 and Figure 4 from Hooper et al. manuscript). This paper evaluates both overall management strategy changes as well as specific management changes for 6 real world cases of patients tested with DecisionDx-SCC.⁴ These management changes directed by GEP result and BWH T stage align with the management algorithm in Farberg, et al. CMRO 2020. For each case, the impact of the GEP test result changes based on the underlying risk of the case as it is interpreted in the context of the clinicopathologic risk factors. Importantly, these figures discuss both overall management intensity as well as how GEP result interpreted in the context of clinical and pathologic risk factors informs specific changes in management by treating clinicians (please refer to published Figure 3 from Hooper et al. manuscript).

In conclusion, the recent NCCN addition of high risk and very high risk subsets does not negate the clinical management guideline published by Farberg et al. 2020, Arron et al 2021, and Singh et al 2023. Clinical utility studies show that these management guidelines are implemented by clinicians interpreting test results with clinical and pathological information (Litchman et al. 2020, Hooper et al. 2022). These manuscripts directly address requests from the draft LCD for specific management changes that integrate risk factor assessment and GEP results. Notably, NCCN management pathways are substantially similar after the introduction of the “very high risk” category have not changed since the publication.

Finally, the draft LCD requests odd, specific statistical analyses that are used to evaluate comparable staging systems that are based on similar or overlapping factors to demonstrate superiority, notably distinctiveness, homogeneity, and monotonicity. These statistical analyses are not typically applied to molecular testing as they are intended to be used with staging, and because they do not use overlapping factors, do not compete with staging as an either/or risk assessment strategy. This is supported by the fact that the ACCE hierarchical framework, which is applied and frequently cited in other local coverage determinations such as Breast, Bladder and Prostate cancers, does not evaluate these metrics. In this hierarchical framework, analytical and clinical validity are calculated by sensitivity, specificity, positive predictive value, and negative predictive value, as has been published supporting the DecisionDx-SCC. Moreover, these analyses are easily calculated from the published available data, should the contractor assessing the validity of the test choose to evaluate.

Homogeneity is defined as outcome similarity within a risk group, monotonicity is defined as outcome worsening with increasing risk group, and distinctiveness is defined as differences between risk groups. The most straightforward way to address distinctiveness is by Kaplan Meier analysis, which shows the clear and statistically significant separation between test results of the 40-GEP across validation studies, including subset analyses (see Ibrahim et al. for examples).

Homogeneity can be determined by evaluating the proportion of metastatic events that occur in low-risk subsets and their statistical comparison (performed by McNemar’s test). In this analysis conducted using published data, the GEP Class 1 result had fewer metastatic events than were captured by patients in BWH T1/T2a stages. Note that T3 tumors were excluded from this analysis, per the proposed intended use population indication in the LCD. What is most interesting in this analysis of homogeneity is that the GEP test outperformed BWH staging with a p value <0.0001.

While the LCD claims the potential for under-staging in the validation cohort, which may account for higher proportion of events in BWH T1/T2a tumors than observed in other published studies, the LCD fails to appreciate that:

1. to meet 40-GEP testing criteria, cohorts typically consist of BWH T1 tumors that also have at least one other high risk factor not recognized by BWH staging, making this cohort enhanced for metastatic risk compared to the general SCC population in which other studies are conducted which include BWH T1 tumors that lack other clinical or pathologic risk factors,
2. the PPV of BWH staging in our cohort is similar to other published cohorts, suggesting under- staging is unlikely (Ruiz et al JAMA Dermatol 2019)
3. sensitivity analysis to evaluate the potential impact of under-staging due to missing clinical data showed no significant impact (Wysong et al. JAAD 2020, supplemental data). This analysis should be included in the LCD. In this analysis, the missing features were imputed as highest risk possible, and impact on risk stratification and statistical comparisons was assessed. When the highest risk factors were imputed for missing data, only 6 patients moved from T1/T2a classification to T2b/T3 classification, and the statistical findings did not change.
4. There are other studies cited in the draft LCD that describe the statistical accuracy of currently used staging systems (AJCC/BWH) that include significant missing clinical data (Karia et al. 2018, and Jambusaria et al. 2014).^17–20 Of note, there is no critique about the potential for under staging in these cohorts from the authors of the draft LCD.

The final LCD should be updated with these important analysis checks, and inferences of the impact of under-staging on the findings of the validation should be removed based on the published analyses refuting this claim and the frequency of missing clinical variables in other high impact studies that are both used clinically and cited without criticism in the draft LCD.

Similarly, monotonicity can be calculated from available, published data (excluding BWH T3). In this analysis, we compare the proportion of poor outcomes captured by BWH high stage disease to the proportion of poor outcomes captured by GEP Class 2 results. As with the homogeneity analysis, high risk test results have improved performance over BWH staging. Again, these comparisons were able to be made from published data that was available well before the time the LCD was drafted. The same comparisons, with highly similar findings, can be calculated for AJCC staging.

The LCD should remove the request for these statistics from the LCD, and/or include the above tables as they can be easily constructed from published data. Moreover, this data further supports all prior analyses that the test adds value even in the context of staging. Suggestion that the added value is unclear is not supported and the conclusion should be updated to reflect that the DecisionDx-SCC test significantly improves risk stratification over use of clinical and pathological risk factor assessments alone.

Overall, it is exquisitely clear how to use this test in a way that aligns the published clinical validity with the published clinical utility to inform the risk aligned management strategies already included in guidelines. Data confirms that the test is clinically validated and that treating clinicians are using 40-GEP results appropriately, and non-coverage of this test will result in net harm to Medicare beneficiaries. I do hope they consider positive coverage after a more comprehensive data review as we all continue to try and do what is best for our patients diagnosed with SCC.

Figures, tables, and references were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Noridian:

The SCC GEP test offered by Castle Biosciences is an important risk stratification tool for patients with cutaneous squamous cell carcinoma. The proposed LCD fails to completely evaluated the evidence supporting the test and fails to correctly interpret the data presented, and as such, should be revised to cover the DecisionDx-SCC test following re-review of misinterpreted evidence.

First, I am a board certified dermatopathologist and it is well-known in dermatopathology that current histopathologic risk factors that are incorporated into staging approaches and the NCCN guidelines are frequently challenging to report from diagnostic biopsy specimens, which make up the majority of diagnostic material for high-risk SCC cases. While it is true that excisions can often contain more histopathologic risk factor information, many high-risk SCC cases are treated with Mohs Micrographic surgery and a tumor debulk is not always sent for review. As such, additional histopathologic risk factors are not always able to be identified for high-risk SCC cases based on surgical modalities used in primary treatment of the tumor (Farberg, et. al. Dermatology and Therapy, 2022). It is also important to recognize the lack of standardized CAP synoptic reporting for primary cutaneous SCC which leads to a marked diversity of reporting styles and histopathologic risk factors reported on diagnostic biopsy reports. Also, in my experience most board-certified dermatologists and dermatopathologists don’t stage cutaneous SCC due to these challenges and a lack on consensus on staging systems. Given these challenges, it is important to revise the draft LCD to acknowledge the limited ability of histologic evaluation to estimate the underlying risk of regional or distant metastasis for patients with high-risk SCC. The current draft appears to elevate some specific histologic factors, such as desmoplasia, however, it remains the case that such risk factors are rarely observed and only a very limited subset of risk factors has been incorporated into staging. Thus, comparison to these risk factors does not impact population-level conclusions regarding test performance. Therefore, comparisons of accuracy must be compared to existing methods of risk stratification as opposed to hypothetical amalgams of histologic risk factors that have yet to be defined.

Second, the value of the multivariate analysis seems to be not fully appreciated or understood in the LCD. Two independent multivariate analyses have demonstrated that the GEP results are significant predictors of risk in the context of other risk factors, or in the context of traditional staging systems. This means that, when considered the relative predictive power of all other risk information, the GEP results add information that complements known factors. When multiple factors are significant, the analyses suggest using the risk factors together to provide a complete risk assessment of the tumor. Therefore, the multiple statements that it is unclear or not demonstrated that the test adds value to risk factor assessment are false, and not supported by evidence review. In fact, strict reading of the evidence and statistical analyses finds the opposite conclusion. These statements should be retracted from the LCD and replaced with supportive statements that correctly interpret the multivariate significance. For example, the LCD MolDX: Molecular Risk Stratification for Melanoma, the multivariate analyses showing significance of GEP results with higher hazard ratios than traditional staging factors concludes,

“This study demonstrated that the Melanoma DecisionDx test was an independent predictor of both recurrence and distant metastasis-free survival based on multivariate analysis (including Breslow thickness, ulceration, age, and SLNB status; with the GEP having the highest Hazard Ratio of 2.9). The NPV of the GEP test was 92% (90-94% CI) for recurrence-free survival and 93% (91-95% CI) for distant metastasis-free survival, both outperforming SLNB.”

Further in the paragraph it states,

“[DecisionDx-Melanoma] was demonstrated to be an independent predictor of RFS, DMFS, and OS in multivariate analysis, including AJCC 8th ed. staging (staging was also an independent predictor). Other studies had similar findings. Combining AJCC staging or other clinicopathological factors and GEP was also found to improve risk stratification.”

Given the similar findings for the SCC GEP test, these supportive statements should be included, along with acknowledgement that the test demonstrates added value to current staging approaches. It is important to recognize the limitations of histopathology to completely explain the risk of disease progression in cutaneous squamous cell carcinoma. It is important to identify the clinical and pathologic factors associated with poor outcomes in this disease state, however, there is also a clear need to refine current risk stratification approaches. The DecisionDx-SCC test has demonstrated the ability to improve risk stratification in the context of individual risk factors as well as staging systems, and for high-risk SCC can inform management decisions for patients that are based on their actual risk of regional/distant metastasis.

Reference was provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Noridian:

I am writing to provide my perspective on the Castle SCC 40-GEP test and the value of gene expression profile testing for patients diagnosed with squamous cell carcinoma (SCC). The test provides information about the individualized risk of SCC and sheds light on proper management of the patient.

SCC is a complex disease. The risk assessment has been heterogeneous, resulting in the lack of management guidelines. The current management recommendations are based on clinical and pathological factors only, however, they lack individualized risk stratification. This has led to underestimation or overestimation of the risk, directly and indirectly increasing healthcare cost on SCC management. The Castle 40-GEP, as a consistent and accurate predictor of risk that can be combined with the full patient history, clinical and pathological findings, is a tool that can improve the current standard of care.

For example, the use of the Castle 40-GEP can help to determine which patient would require adjuvant radiation therapy. Patients with high-risk SCC may benefit from adjuvant radiation therapy by reducing the risk of metastasis and recurrence, hence lower mortality. Castle 40-GEP study can evaluate the individual risk of the SCC and provide risk stratification for consideration of such adjuvant therapy. A publication from June of 2021 in the Journal of Drugs in Dermatology shows how to incorporate Castle 40-GEP into practice. Based on the current data, the expert panel determined that if a patient’s GEP result is a Class 2B (highest risk of metastasis), referral to a radiation oncologist is warranted. For patients with a Class 1 (lowest metastatic risk), this referral can be avoided. Utilizing the GEP result along with other high-risk clinical and pathologic factors provides an additional level of objective evidence when making this decision.

I find Castle 40-GEP test a powerful tool in helping clinicians like me to better understand the risk of a specific SCC. This can help to guide management for biologically high-risk patients in order to reduce mortality, while sparing patients from unnecessary treatments. Medicare coverage to maintain the accessibility of this tool for my patients is of critical importance. I strongly support the coverage of this test.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Noridian:

I am writing to discuss the need to reverse the noncoverage policy related to gene expression profile (GEP) testing in patients with cutaneous squamous cell carcinoma (cSCC).

There are several issues with the proposed LCD that I have listed in short below. These points highlight issues that need to be rectified prior to finalization of the policy.

• The validation cohort of 420 patients with cSCC with one or more risk factors was insufficiently reviewed in the LCD (Ibrahim et al 2022). This study provides significant data supporting the use of the GEP with staging. However, this was not reviewed in the noncoverage policy, perhaps because the supplemental tables and figures were overlooked.
• The real-world clinical utility study was insufficiently reviewed in the LCD (Hooper et al 2022). This study provides significant data supporting the impact of combining risk factors with the results of GEP, and showcases a comprehensive approach integration of clinical, pathological, and genetic information to guide patient management. However, this was not reviewed in the noncoverage policy. Only data evaluating testing trends in clinically tested patients is reviewed, overlooking the most crucial data presented in this paper.
• Evaluation of consistency of treatment modalities informed by the GEP test is lacking. Multiple publications have outlined how to incorporate the GEP results with risk factors to make management decisions, but thoughtful consideration of this data is absent, including Singh et al 2023, which was completely absent from review.

I kindly request that MolDx thoroughly review the SCC literature and consider the points raised regarding independent class calls, their significance within the context of staging and risk factors, and the comprehensive management strategies associated with each class. By considering these factors, MolDx can ensure that coverage policies reflect the evolving understanding of cSCC management and provide patients with access to the most appropriate diagnostic tools and treatment options.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Noridian:

Traditional risk assessment methods fall short of identifying all patients with SCC at high risk for metastasis; the low accuracy of individual risk factors and formal staging systems have resulted in low uptake for formal staging in the community.^1–5 The use of molecular tools such as gene expression profiling have demonstrated clinical utility in breast cancer, prostate cancer, melanoma, and uveal melanoma,^6–8 to identify patients at high risk for metastasis based on gene expression in the primary tumor and add more objective information during decision-making for risk-adjusted patient management. These tests represent the cutting edge of personalized medicine.

The 40-gene expression profile test (40-GEP/ DecisionDx-SCC) has been clinically validated and is clinically available from Castle Biosciences, for testing primary tumors from patients with cSCC and >1 high-risk factor.⁹ Results from the test provide independent prognostic information to improve the accuracy of risk stratification for a patient’s risk of cSCC regional or distant metastasis within three years of diagnosis (the time during which most metastases from cSCC occur).^1,10 The risk stratification provided by Castle’s test is used to complement traditional clinicopathologic risk assessment and tumor staging systems to provide the patient and provider with the most informative risk profile to determine the optimal management and treatment pathways for optimal outcomes. There are many claims in the proposed LCD that evidence is lacking to support positive coverage, however this is the result of an incomplete or very narrow interpretation of the published literature. More specifically, the LCD suggests that there is uncertainty of the value the test adds to staging, when this is comprehensively covered in the validation study published by Ibrahim et al, is included in physician management algorithms, including Farberg, et al., and Singh, et al., and is addressed directly and seen to impact clinical utility in Hooper, et al., and Litchman, et al. Further the LCD states that outcomes data is needed to support coverage, despite other tests with similar utilities achieving coverage with similar levels of evidence in other disease states (e.g., breast and prostate cancer). This sets up inequality in affordable access to care between disease states that is against the spirit of innovative medicine.

Another aspect that is missing from the draft LCD is the lack of recognition of the varied reporting practices for histopathologic risk factors in cSCC. There is not a standard “synoptic” report that is required for primary biopsy diagnosis of a cSCC – unlike in cutaneous melanoma. This lack of standardization leads to variable reporting of histologic risk factors that are incorporated into staging and NCCN guidelines. Instead of critiquing the validation data of DecisionDx-SCC for potentially including under-staged patients in the proposed LCD, the final policy should recognize that every effort was made to obtain staging information in the DecisionDx-SCC validation cohort, that reporting of histologic risk factors is not required or uniformly observed in clinical practice, and that there is missing clinical or histopathologic risk factor tumor data in foundational manuscripts (cited in the LCD) that support staging systems such as BWH (pre-operative tumor size) and AJCCv8 (tumor depth).^4,11 These statements will better articulate and increase the unmet clinical need for improved risk stratification in cSCC and highlight the prognostic value of DecisionDx-SCC GEP results interpreted in the context of the reported clinical and histopathologic factors to inform risk aligned decision making.

This test should be available for patient testing with affordable coverage. By more accurately identifying patient risk of poor outcomes, the most appropriate treatment can be implemented to reduce the likelihood of poor outcomes. This provides health as well as economic benefits, as more accurate identification of metastatic risk early in disease can allow for timely surveillance and appropriate intervention at a time when response could be optimal, while avoiding potentially costly treatment for more aggressive disease.

References were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Noridian:

I am writing to address concerns regarding the description of patient management in the proposed Local Coverage Determination (LCD). The current description, contained within a single paragraph, fails to capture the complexity of managing patients with cutaneous squamous cell carcinoma (cSCC). While it is essential to recognize that NCCN guidelines provide a framework for patient management and outline potential treatment modalities, there is significant grey area in treatment modalities and when they are considered or recommended that is missed in the generic review. Almost every treatment modality in the NCCN guidelines for cSCC relies on a physician’s impression of a patient’s level of risk to determine the appropriate level of treatment within each treatment modality. These nuances are missing from the proposed LCD, which impairs the ability to correctly place prognostic testing in the treatment paradigm and understand the utility of the test.

For example, for high-risk and very high-risk NCCN patients, quarterly, biannual, or yearly follow up are all options for these patients and the only guidance that NCCN provides is “Frequency of follow-up should be adjusted based on risk.” (Footnote jj, NCCNv1.2023). Notably, the ranges for follow-up offered for high-risk and very high-risk patients are so broad and non-specific as to be completely overlapping and can be clearly improved by more granular individual risk information.

Additionally, the LCD should give appropriate attention to ART, not only by mentioning when it is recommended but also by discussing when it should be considered. Again, NCCN guidelines include ART as a possible treatment for both high-risk and very high-risk patients. Guidelines suggest to “consider ART” for both patient groups and offer a few specific risk factors that should direct a physician to consider ART including “high-risk for regional or distant metastasis”, however no additional guidance is provided to determine that risk. (Footnote y, NCCNv1.2023)

There is no discussion in the LCD at all regarding use of imaging for nodal assessment or surveillance imaging, though both are included in guidelines recommendations for consideration or use in patients with very high-risk tumors and are supported by evidence that imaging positively impacts outcomes (Ruiz JAAD 2017). NCCN guidelines include recommendations for imaging for nodal assessment at diagnosis and direct that “Imaging modality and targeted area should be at the discretion of the treating team based on suspected extent of disease” (Footnote g, NCCNv1.2023). Moreover, imaging can be considered to surveil for disease progression “when clinical exam is insufficient” and is also “at the discretion of the treating team based on suspected extent of disease” (SCC-6, Footnote g, NCCNv1.2023).

All these management plans for patients with cSCC included in guidelines are open to interpretation by clinicians and informed by risk assessment. Unfortunately, due to well-documented insufficiency of current risk assessment tools, there is need for additional information to identify which patients to treat or follow more closely. Without additional risk assessment tools, clinicians often hesitate to decrease patient management, and tend to overtreat patients with underlying low risk tumor biology.

The prognostic gene assay 40-GEP available from Castle Biosciences has been validated to improve stratification of cSCC patients regarding metastatic probability. The validation of the assay extensively compares the assay to all available information, including risk factors or staging systems. Clinical utility studies published show reliably consistent impact of the assay results on patient management plans in NCCN guidelines.

Because of the complexities associated with treating patients with cSCC, additional tools like the Castle prognostic gene assay are needed to determine probability of metastasis. Physicians and patients need access to this test to make the best decisions for their care and delineate between options and intensity of care. I urge you to revise the LCD to accurately depict the intricate nature of patient management in cSCC and revise the determination to coverage for the Castle prognostic assay.

Reference was provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA and Noridian:

Thank you for allowing the opportunity to provide comments regarding the draft LCD MolDX: Molecular Biomarker Testing for Risk Stratification of Cutaneous Squamous Cell Carcinoma.

As a surgical oncologist, I take care of patients with high risk squamous cell carcinomas (SCC) and many with locally advanced disease on a regular basis. In my practice I use the available tools for risk stratification that include careful evaluation of clinical and pathologic risk factors – and more recently – I have integrated genomic testing with the DecisionDx-SCC test. Based on my clinical experience and independent review, I thought that it was important to present on behalf of my patients because I disagree with Proposed LCD. This letter outlines how I incorporate the DecisionDx-SCC test into my clinical practice and the published data that supports the appropriateness to do so. Finally, I also included a recent case from my clinical practice in which I utilized the DecisionDx-SCC test to determine the appropriate treatment for one of my SCC patients with now 22 months of clinical follow-up.

By way of background, while I have clinically distinguished my practice to focus on cutaneous neoplasms, during my training and currently in my surgical oncology practice, I not only see patients with skin cancer but also tumors involving other organ systems. In this way, I can compare evidentiary thresholds in cutaneous oncology to a range of other cancers. When looking across disease states, it is clear that delineation of specific management changes with subsequent improved outcomes is not required to be considered medically reasonable and necessary, either in the clinic or by Medicare. Proposed LCD puts forth an unreasonably high threshold for clinical utility for DecisionDx-SCC that is not required for prognostic tools in other disease states, nor for risk stratification tools currently used in SCC. For example, in other MolDX LCDs covering similar prognostic tests in prostate and breast cancer, tests have positive coverage policies WITHOUT demonstrating a direct benefit to patient outcomes following clinically valid risk stratification. Here clinical utility is established when a clinically validated risk stratification tool is proven to inform specific patient management decisions that have proven benefit to patient outcomes - this definition has been used consistently across disease states to establish the threshold for MolDX clinical utlity.¹ However, the same clinical utility threshold is not being applied in the draft LCD for the SCC prognostic test and instead the proposed LCD appears to necessitate long-term outcome data or a randomized trial to establish any clinical utility for molecular testing in SCC. Requiring a different level of evidence for molecular tests for SCC than is applied to molecular tests for prostate and breast cancer establishes disparity and inequity between disease states that is unethical.

The evidentiary bar for clinical utility in risk stratification should also be equivalent to the evidence available for the existing risk stratification systems in the SCC disease state. Specifically, the LCD references the three current clinical and pathologic based risk stratification approaches that are in use across the US currently: NCCN high risk/very high risk groups, AJCCv8 staging for head and neck, and Brigham and Women’s Hospital (BWH) staging. Based on the arbitrarily high evidentiary bar for clinical utility set in the proposed LCD, these risk stratification systems would not be considered to have met the Medicare standard for clinical utility. In contrast to their approach to the DecisionDx-SCC test, the discussion actually highlights the clinical utility of risk stratification based on clinical and pathologic risk factors. The review of these systems makes sense as NCCN/AJCC/BWH risk stratification approaches are clinically useful and are being used in clinical practice, DESPITE THE FACT that they are based on retrospective data sets, lack demonstrated improvement in patient outcomes and are not specified by guidelines to inform specific management changes. Even without long term outcome data supporting NCCN, AJCCv8 or BWH, current management decisions for patients with SCC are informed by the presence or absence of specific high risk clinical and pathologic factors with risk-aligned treatment decisions made on modality-by-modality and patient-by-patient basis (as acknowledged in guidelines). The LCD relies on different metrics for evaluating the clinical utility of clinical/pathologic risk stratification and gene expression profile risk stratification. The bar for establishing clinical utility should be consistently applied. While NCCN, AJCC and BWH do certainly provide risk stratification, as noted above, there is clearly room for improvement in accuracy of risk stratification. As such, a well-validated test that is demonstrated to inform risk aligned management decisions is inherently clinically useful in SCC. The DecisionDx-SCC test has clearly met the threshold for clinical utility in this SCC disease state where none of the clinical and pathologic systems of risk stratification have long term evidence with outcome improvement through their clinical application. DecisionDx-SCC has been well-validated for clinical use including robust analytic validity and one of the largest SCC patient sets available in the US.^2-4 Clinical Validation has now been demonstrated in close to 1000 paHents.^3–5 Importantly, the published clinical validity data from Ibrahim et al demonstrates that the Class 2A and Class 2B results are independent of all other risk-prediction systems used in current clinical practice.^3,4 The test not only meets or exceed accuracy of other staging systems, but is an independent predictor of metastasis. This is crucial, because it means that treating clinicians such as myself can combine DecisionDx-SCC with clinicopathologic-based staging, which improves the accuracy of risk prediction over other risk assessment approaches.⁵ The importance of this analysis is unappreciated and is a superior, and statistically supported, method to evaluate clinical validity over the comparison to the accuracy to staging that is favored.

The LCD assumes that there are clearly-defined, specific management pathways for patients with SCC, when in fact, as discussed above, management decisions exist on a spectrum informed by risk factors on a patient-by-patient basis. All patients eligible for the 40-GEP test fall into either the NCCN high risk or the NCCN very high risk group where treating clinicians must establish risk aligned management plans after definitive surgical treatment. Even though there are two risk groups in the NCCN guidelines for cutaneous squamous cell carcinoma, the outlined post-surgical management pathways do not distinguish between the two groups and are informed -at least in part- by the treating clinicians understanding of the patient’s risk of regional and distant metastasis (See SCC-3 and SCC-3A, NCCNv1.2023).⁶ The SCC NCCN guidelines also acknowledge that patient-centered decision making is a key component of high-quality cancer care (see Principles of Treatment and Patient Education sections of NCCN 1.2023).⁶ This concept is missing from the draft LCD. The next sections include examples of the types of decisions that physicians caring for patients are tasked to make after definitive surgery of a patients’ SCC. These decisions are based on the physician’s understanding of the patient’s risk of metastasis and should be influenced by the risk assessment made from the combination of clinical, pathologic and gene expression profile information.

Follow Up frequency – Per NCCN guidelines, all patients eligible for DecisionDx-SCC are eligible for a follow-up schedule ranging from quarterly to yearly, but this is a broad suggestion meant to be offered based on individual risk and per the physician’s discretion. We have to make a decision which takes into account that a follow up schedule that is too intense/frequent can be expensive and a burden to the patient while if we follow up too infrequently, we could miss early disease progression where it can be treated most effectively.

Nodal Evaluation/Surveillance – Again, all patients eligible for the 40-GEP test are directed to a range of nodal surveillance/evaluation including: nodal palpation, ultrasound imaging, CT imaging, PET imaging, or sentinel lymph node biopsy. And once again, NCCN guidelines ask physicians to determine the appropriate methodology based on a patient’s individual risk of metastasis. It is also clear from other cancer types as well as other skin cancers that when surveillance imaging is targeted to the patients with the highest risk of disease progression that you can identify smaller tumors sooner that have favorable treatment responses in other disease states.^7–9 While SCC-specific research would be preferred here, the SCC field is typically lagging behind other fields by a few years.

Adjuvant Radiation Therapy - Per NCCN, all patients eligible for 40-GEP test fall into either the “consider” or “recommend” ART category, however, again, there is only broad direction from NCCN guidelines on who to treat with ART. Due to the burden and potential side effects of ART treatment, we need to be selective when considering which patients should receive ART treatment. Similar to the other treatments discussed above, NCCN guidelines rely on the physician’s understanding of the patient’s risk of regional and distant metastasis to direct treatment for patients. (SCC NCCN guidelines - SCC-3 and SCC-3A footnote y) Because I am tasked with establishing a comprehensive assessment patient’s risk of metastasis to make the management decisions listed above, I utilize the 40-GEP test along with existing staging methods to inform my risk aligned management decisions. In fact, in my practice, I typically follow a workflow with my high-risk SCC patients that I have included below and that I discussed during the open comment meeting. This workflow allows me to combine risk stratification information from both BWH staging and DecisionDx-SCC results to make the most informed decisions with my patients. This additional level of risk stratification allows for identification of patients for escalation and de-escalation of care for patients that would have been under or over treated if only using BWH. Importantly, this workflow is consistent with the clinical algorithms from Farberg et al. 2020 and Singh et al 2023 which also contain clinical management algorithms that are either not included or discounted in the data analysis section of draft LCD.^10,11

I would like to provide an example case from my clinical practice where I followed this workflow to deescalate management intensity with a DecisionDx-SCC Class 1 and that aligns with the published clinical utility of the test.^10–13 Because the extensive clinical validation, the results of DecisionDx-SCC are robust, accurate and consistent, meaning that use of the test to inform risk-aligned management is appropriate. This patient came to my clinic and presented as the following:

• male with large fungating central chest mass
• Clinical size: 4.2 x 4.0 cm
• Diagnosis: Poorly differentiated SCC, lesion extends to every margin
• Clinical/Pathologic Staging: BWH T2b

In Ibrahim et al. manuscript that BWH T2b patients have a risk of metastasis of 32.6%, however not all T2b tumors carry the same risk of disease progression. This means that, given the low positive predictive value of BWH staging, if all T2b patients in this cohort received intensive management, nearly 70% would be overtreated.^14,15 Prior to GEP testing, my initial plan was to consult with a multidisciplinary tumor board, consider pre-op axial imaging with CT or PET/CT (which would have increased the total cost of the patient’s care), and I would have offered an SLNB (which would have added time, cost, and risk to the patient’s care). Additionally, I would have referred this patient to a radiation oncologist for adjuvant radiotherapy following surgical clearance of the tumor and scheduled clinical follow-up 2-4x per year. When I received the DecisionDx-SCC result back as a Class 1, I was able to better assess this patient’s risk of metastasis. From the Ibrahim et al. study, a Class T2b patient with a Class 1 result carries a risk of metastasis of 18.8%. This represents a nearly 50% reduction in expected risk of metastasis over 3years for this patient and more closely matches the expected cohort risk for a BWH T2a tumor where I would not typically recommend ART or many of the other previously considered management choices. After receiving the Class 1 result and putting this independent piece of information in the context of the patient’s other known risk factors, I was able to update his treatment plan accordingly after his surgery. To that end, we had a personalized discussion regarding management plans where we discussed both his BWH T stage as well as the Castle GEP result and together, we decided to proceed with a wide local excision with wound VAC. Once negative margins were confirmed, we utilized a split-thickness skin graft and moved forward with active surveillance 1-2x per year follow up with thorough clinical nodal exams. Since this patient is still at an elevated risk of metastasis (albeit reduced from pre-GEP estimation), we are continuing close clinical follow-up with thorough nodal exam, but we were able to avoid pre-operative imaging, SLNB, and ART that are all considered for a patient with a BWH T2b tumor based on clinical and pathologic risk factors alone.

Finally, the next question is, how do I know that this was the right treatment pathway for my patient? My certainty is informed both by the shared decision-making approach which entailed a comprehensive discussion of clinical, pathologic and genomic risk factors and his subsequent clinical course with no evidence of recurrence or metastasis. Since initial surgical removal of his SCC, this patient has remained disease free for now 22 months at the time of his last clinic visit. For this patient, the Class 1 test result helped to de-escalate care by accurately identifying T2a-like risk in a patient with a BWH T2b tumor and we have been able to safely avoid unnecessary ART therapy using this shared decision-making approach.

I was able to treat this patient in a risk aligned manner to BWH-T2a-like risk, and I’m confident in the chosen management strategy because of the superior accuracy metrics of BWH when combined with GEP as opposed to just BWH staging alone.⁵

DecisionDx-SCC combined with staging gives the most accurate assessment of metastatic risk. Without the GEP result clinicians are consistently overtreating patients who present with BWH T2b tumors due to the low PPV of staging and this overall is a well-known clinical challenge for doctors who manage high-risk SCC tumors. Again, although it is presented not all patients with T2b tumors benefit from the high intensity management strategies that are considered in NCCN guidelines, routinely performed for these patients, and covered by Medicare. Shared decision making with patients who have T2b tumors and a Class 1 DecisionDx-SCC test result is an evidence-based approach informed by published data and that leads to consistent and accurate risk-aligned management decisions. Although not discussed in my written comments, the same workflow to inform management decisions that combines clinical, pathologic and genomic risk information to produce more accurate risk stratification clearly applies to patients with BWH T1 and T2a tumors.

In conclusion, patient management strategies related to follow-up, imaging, nodal assessment, adjuvant therapy, etc. are established and are known to have a positive impact on outcomes, but treating clinicians are unclear which patients to treat because of low PPV of risk assessment systems that rely on clinical and pathologic factors alone. Guidelines recommend that these management strategies be informed by individual risk and balanced with cost/benefit in a shared decision-making model. Clinicians trained to manage high risk SCC interpret GEP results in the context of the known risk factors or staging systems, just as they current interpret multiple risk factors and consistent with data from clinical utility studies.^12,13

I hope that my comments support reconsideration of the 40-GEP test and support positive coverage inclusion in the final LCD.

Figure and references were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Noridian:

Cutaneous squamous cell carcinoma (cSCC) is most often a low risk cutaneous malignancy that can be definitively cured by surgical excision with clear margins. However, some cSCC tumors are at high risk for poor outcomes (e.g., nodal and distant metastasis) and doctors are alerted to this possibility by the presence of high-risk features seen in the clinic or under the microscope reviewed by a pathologist during diagnosis or a Mohs surgeon during surgical removal. Most cSCC tumors that have poor outcomes are high-stage tumors (e.g., Brigham and Women’s Hospital [BWH] classification T2b or T3);^1–4 although, over 30% of tumors that go on to metastasize are initially classified as low-stage (e.g., BWH T1 or T2a).^4–6 Specifically, a substantial proportion of unfavorable outcomes for patients with cSCC develop in cases staged initially as BWH T2a.⁷ Importantly, patients having primary tumors with only a single very high-risk factor (either poor differentiation, invasion beyond subcutaneous fat, perineural invasion ≥0.1mm, or diameter ≥ 2 cm) are classified as BWH T2a. BWHT2a tumors are a gray area with substantial uncertainty in risk and appropriate treatment, as all of these tumors (with the exception of those with a tumor diameter ≥ 2 cm but < 4 cm) are classified as NCCN very high risk, but according to risk assessment and consensus groups, may not have actionable level of risk for some higher intensity treatment pathways (such as ART).

This is a space where additional prognostic information about the primary tumor, which can be derived from molecular tools such as the 40-gene expression profile (40-GEP) test.⁸ The 40-GEP only benefit the decision-making process for planning the best patient follow-up and surveillance regimen because it adds to, and does not take away from, what is already available. Clinical scenarios involving BWH T2a tumors are absent from discussion in the LCD, with discussion of the value of test results only discussed in cases where Class 1 results confirm low risk or Class 2A/2B or Class 2B results confirm high risk. Oversimplification of the disease state and failure of the LCD to appropriately characterize the current management challenges clinicians face due to the well-known limitations of clinical and pathologic staging approaches conceals the clinical value and utility of the 40-GEP test. This complexity needs to be accurately represented in the final policy and the policy authors should turn back to the published validation data where the 40-GEP has been shown to improve the accuracy of risk stratification when the 40-GEP test result is interpreted in the context of the patient’s known risk factors or BWH stage (Figure 2, and Supplemental Figure 3 of Ibrahim et al).⁸

The uncertainty in patient management is not well captured in the draft policy posted, which undermines the clinical problem and therefore limits assessment of the 40-GEP to solve that clinical problem. In one instance the LCD mischaracterizes the NCCN clinical pathway for SCCs by stating:

“Current NCCN management guidelines for local cSCC recommend MMS or other forms of peripheral and deep en face margin assessment (PDEMA) for very high-risk tumors, with consideration of adjuvant radiation therapy (RT) in cases with poor prognostic features.”

However, based on current NCCN guidelines (v1.2023) Mohs or other PDEMA forms are not applicable only for very high-risk tumors but also for tumors with high risk factors (see SCC-3).⁹ Clinical practice patterns reflect this statement as well, as Mohs surgery is broadly adopted across the US for SCC with high and very high risk factors. More importantly for the 40-GEP test, given the fact that the NCCN establishes that high risk SCC includes tumors with at least one high risk or very high risk factors, all of the downstream management pathways that follow in the current NCCN guidelines are able to be considered or recommended to patients in the intended use population for 40-GEP. The proposed LCD focuses discussion on the NCCN very high risk group disproportionately in the discussion and this obfuscates clinical instances where this test is clinically useful to inform the broad range of management decisions available to NCCN high risk AND very high risk tumors.

There are numerous instances in NCCN guidelines where there is significant ambiguity and/or clinical judgement leeway for physicians to identify appropriate management strategies for patients with high risk SCC (for example, follow-up frequency, imaging, ART). In each of these instances, the types of specific management approaches advised by guidelines are formally based on the treating clinician understanding of their patient’s risk of disease progression. The LCD appears to acknowledge this fact in the Surveillance and Treatment Intensity are Influenced by Risk-Stratification Group section and states, “Suggested follow up frequency and intensity vary according to risk group.” However, the LCD misses the downstream ramifications of this statement; when the current treatment paradigm directs physicians to change management based on their understanding of their patient’s risk for poor outcomes, risk stratification tools (including 40-GEP) that improve the accuracy of risk prediction lead to clear changes in risk aligned management strategies. Essentially, if a tool can adjust a clinician’s assessment of the “risk group” for a patient they are treating, the tool by definition then informs specific risk aligned management strategies already articulated as reasonable and necessary within guidelines.

Moreover, adjuvant radiation therapy can be considered for high risk and very high risk cSCC with negative margins with PNI or other poor prognostic features, with no guidance on which prognostic features to assess for guidance. “Suggested follow up frequency and intensity vary according to risk group” as included in the LCD oversimplifies the clinical scenario in which follow-up recommendations for high risk and very high risk are overlapping and offered by ranges, with no guidance for selection of intervals (3-6 month for 3 years, then 6-12 months for 3 years, then 6-12 months for life). Consideration for imaging is included if clinical exam is insufficient, to be guided by suspicion of extent of disease, to be assessed by histology (for high risk histological features) with no guidance on what these features are or how they should be interpreted. This is not included as a management modality in the LCD. All of these management “recommendations” have substantial room for interpretation of patient risk to guide management decision with very little guidance given on how to select the appropriate patient for escalated guidance. This uncertainty needs to be reflected in the LCD with high fidelity to appropriately position the unmet clinical need the test fills, as well as to answer questions in the guidelines for specific management modalities impacted; these are the management decisions shown to be impacted in clinical utility studies by the test.

As an example of the complexity of this risk assessment and treatment recommendations in this disease state, there is marked heterogeneity of treatment for patients with T2a tumors as BWH positions all of these tumors on the lower end of the risk spectrum and NCCN very high risk positions most of these tumors in the very high risk category. This is one area with a clear unmet clinical need for improvement in risk stratification and the published clinical validity and clinical utility data demonstrate the ability of the 40-GEP test to address this current gap for Medicare patients. As demonstrated in validation publications, the 40-GEP test provides clinically meaningful separation of risk in BWH T2a patients, with Class 1 patients having significantly reduced risk of metastasis and Class 2B patients having significantly increased risk of metastasis.⁸ When compared to baseline risk of metastasis, 50% of patients receive a result that will impact management decisions (82/164), with the majority of these results being Class 1 and eligible for de- escalation. This is supported by clinical utility studies, in which treating clinicians reach consensus on risk-aligned, de-escalation decisions in patients with T2a tumors and Class 1 results.^10,11

Having a more objective test that provides supplementary data about the biology of the tumor should be considered essential, along with current staging systems and national guidelines for providing the best management of the patient. 40-GEP should be a covered service.

References were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Noridian:

As a board-certified dermatologist and Mohs surgeon, I am writing to express my concerns regarding the current coverage policy for the DecisionDx-SCC test under LCD and to advocate for positive coverage under the final LCD. Additionally, I would like to discuss how the gene expression profile provided by this test can be effectively incorporated with current NCCN high-risk or very high-risk factors, AJCC Version 8 staging, and BWH staging to inform risk-aligned management decisions, as these data are specifically requested and yet not reviewed or summarized in sufficient detail.

The DecisionDx-SCC test, based on gene expression profiling, has shown great promise in accurately assessing the risk of recurrence and metastasis in patients with cutaneous squamous cell carcinoma (cSCC) through two published validation cohorts (Wysong et al, 2020, and Ibrahim et al. 2021). This non-invasive test analyzes the expression of specific genes associated with aggressive tumor behavior, enabling a more precise determination of the patient's risk profile.

By incorporating the DecisionDx-SCC test result with existing risk factors such as NCCN high-risk or very high-risk factors, AJCC Version 8 staging, and BWH staging, dermatologists and Mohs surgeons can make well-informed decisions regarding the specific management decisions frequency of follow-up, nodal imaging surveillance, and consideration of adjuvant radiation therapy. These options are well articulated enumerated in the published literature, and I find the confusion expressed regarding guidance of specific management strategies to be with basis.

The gene expression profile obtained from the DecisionDx-SCC test provides valuable information that complements the current risk stratification systems. By identifying patients with a high-risk gene expression signature, we can better identify those who may require more intensive surveillance, including more frequent follow-up visits and nodal imaging surveillance as outlined in clinical management algorithms (Farberg et al. 2020, Arron et al. 2021, Singh et al. 2023). Additionally, this test can aid in the identification of patients at elevated risk for metastasis for whom might benefit from adjuvant radiation therapy, which can significantly improve outcomes in select cases.

The integration of the DecisionDx-SCC test result with established staging systems allows for a more personalized and risk- aligned approach to patient management. The test provides clinicians with objective data to guide decision-making and enables a more tailored treatment plan, minimizing unnecessary interventions for low-risk patients while optimizing care for those at higher risk.

I kindly request a reconsideration of the coverage policy for the DecisionDx-SCC test under LCD. The positive coverage of this gene expression profiling test would promote evidence-based medicine, facilitate risk-aligned management decisions, and ultimately improve patient outcomes.

Thank you for your time and consideration. I am confident that expanding coverage for the DecisionDx-SCC test will have a positive impact on patient care, providing dermatologists and Mohs surgeons with a valuable tool to guide risk-aligned management decisions.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

The current version of AJCC SCC staging is limited to tumors on the head and neck. While tumors located in this area are of concern, not all of them behave similarly; some of these tumors will be bad actors and have substantial disease progression, including metastasis and death, and others will not. Incorporation of an objective finding to separate risk is needed and welcomed by clinicians treating this type of cancer. In a study published by Ibrahim and others in Future Oncology, a novel genetic profile was used to separate risk in cutaneous squamous cell carcinoma patients. Out of the 420 patients included in the study, 278 of them had tumors on the head and neck, and therefore, could be staged by AJCC staging, as was published in a head and neck subset analysis by Arron and colleagues. This genetic profile was able to separate patients into 3 risk categories based on the tumor biology obtained from the primary tissue biopsy. A Class 2A/2B results demonstrated more than twice the sensitivity for prediction of metastasis compared to AJCC T3/T4 (79.6% versus 37.0%) while maintaining similar positive predictive value (28.3% versus 37.0%). When considering a Class 2B result, which predicts highest risk of metastasis, the PPV is superior to AJCC alone (55.6% versus 37.0%). Importantly, the multivariate analysis in head and neck patients clearly demonstrates that this test adds information to AJCC staging. Both Class 2A and Class 2B demonstrate significant hazard ratios and p values in multivariate analysis in the context of AJCC stage. Thus, the test can be used with AJCC staging to refine metastatic risk assessments provided by staging. It is important to note, and should be noted in the coverage policy, that the missing depth of invasion and/or tumor diameter information in patients was shown to not have an impact on the assessment of the added value of this novel genetic profile. The analysis was repeated with highest level of risk assumed for missing information, and did not change the findings. Thus, the concern for under-staging as extensively shared in the coverage policy is incomplete and missing evidence that refutes and diminishes this concern. Please see and appropriately include the supplemental data in the Wysong manuscript published in Journal of the American Academy for Dermatology for analyses.

While the LCD acknowledges improvement of AJCC staging by this novel genetic profile in the summary of evidence of the proposed coverage policy, this data is not discussed in the analysis of evidence, which is puzzling as this evidence directly provides clarity on how the test is used in the context of staging to improve risk prediction; the test can be combined to refine risk estimated by AJCC T stage. As demonstrated in the Ibrahim publication, the test stratifies risk within AJCC stages to improve precision of prediction and identify patients with clinically impactful differences in metastatic rate within each stage. These patients can have risk-aligned management changes, within NCCN guidelines, as demonstrated by Farberg and colleagues in Current Medical Research and Opinion, which provides a framework to integrate AJCC stage with test results to risk-align management. By virtue of the fact that these management decisions are recommended or considered for patients today by guidelines, and these decisions are driven by risk stratification, it is the rational conclusion that improving risk stratification will improve patient outcomes.

Given that more than half of the patients included in the expanded validation study had tumors on the head and neck, with 54 events occurring in the 278 patients with tumors in this location, this test can aid in determining which head and neck tumors are more biologically concerning. This additional information will allow the clinical staff to direct more resources to the riskier patients while allowing for more routine follow up in the patients with lower risk biology.

Please consider addition of this data to the analysis of evidence and updating the conclusion to more appropriately represent the body of literature supporting this test.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

Given the ongoing advancements in molecular tools for objectively predicting risk for poor outcomes in various types of cancers (prostate, breast, melanoma), personalized medicine using molecular testing is becoming standard of care across the field of oncology. When these tests have been well validated and have demonstrated to improve management recommendations known to improve outcomes, it is in the best interest of the patient to apply these tools to help guide decisions for the ideal management pathway for each patient to result in the best net health outcomes for that patient. A molecular test for patients with cutaneous squamous cell carcinoma with one or more risk factors has been developed and clinically validated (the 40-gene expression profile test or 40-GEP by Castle Biosciences). This test can help guide decisions for high-risk cSCC patient management.

The LCD inappropriately discounts Class 2B results as “rare” and states that the impactful analysis is net health improvement between Class 1 and Class 2A. This argument is false and should be revised in the final LCD.

First, the Class 2B result is not rare; even if 5% of the overall high risk SCC population receives a Class 2B, this is 1 in 20 patients identified with a risk of metastasis around 50%, without accounting for additional risk from presence of high-risk factors. Consideration of the utility of the test in the context of various stages/risk groups of disease is requested, though little effort is made to actually conduct such analysis despite availability of data. Therefore in parallel, the distribution of the Class 2B result and its relative impact within each stage of disease should be considered. While it is relatively infrequent in patients with lower risk disease, this is expected considering the overall metastatic rate is low and poorly predicted by other factors. Therefore, to generate a test with a reasonably actionable PPV, the result will not be common because of low prevalence of progression. The test cannot be penalized for detecting a less comment event in the population, unless the intention of the LCD is to state that the metastatic rate in this disease state is acceptable, and availability of tools to identify patients with high biological risk who may metastasize and die from their disease is not worthy of a non-invasive molecular test. Of note, in some disease states, such as cutaneous melanoma, a 5% risk of a positive test result is worthy of an invasive surgical procedure to find patients with high risk of metastasis (see the NCCN guideline recommendations for SLNB in Cutaneous Melanoma).

Thank you for your consideration of these comments. I look forward to an updated LCD that appropriately acknowledges the value of a Class 2B result.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

Thank you for the opportunity to comment on the policy MolDX: Molecular Biomarker Testing for Risk Stratification of Cutaneous Squamous Cell Carcinoma.

There are misstatements in the LCD that preclude accurate interpretation of the value of the gene expression profile {GEP) test for patients with cSCC with high risk factors. Despite the reference section being fairly complete, there is data in the cited literature that is not discussed or reviewed, despite identification of this specific data as an evidence gap that led to the negative coverage decision. The number of errors in this MolDX policy warrant a complete revision to the analysis of evidence.

For one example, this MolDX policy states, "It would be prudent to further explore the patients with Class 2B results who were not identified as very high risk according to all available staging criteria and risk factor analysis, but for whom a Class 2B result would enable them to benefit from more intense management. This data is not available aside from a handful of clinical anecdotes."

This data is available, included in publications cited in the LCD¹, and not reviewed or included in evidentiary analysis, which exhibits a significant lack of understanding the published data available supporting the GEP test. All BWH risk factors are classified as very high risk factors according to NCCN, and staging factors that identify AJCC T3 or higher are also classified very high risk factors by NCCN. Thus, analysis of Class 2B performance within NCCN high risk patient subset accounts for patients who are not identified as very high risk according to all available staging criteria. Considering that BWH and AJCC are the only formalized staging systems used, analyses of Class 2B performance in these Tl staged patients are also valuable, although they may contain other NCCN very high risk factors not incorporated into staging. This data is also published, but not discussed in the MolDX policy.

When considering NCCN division of risk between high risk and very high risk subsets in the expanded validation study, 8/23 of Class 2B results (35%) were found in NCCN high risk patients versus 15/23 of Class 2B results (65%) in NCCN very high risk patients.¹ One-third of Class 2B results occur in patients missed by all risk assessment methods as very high risk is a meaningful proportion of Class 2B results, and is counter to insinuation throughout the MolDX policy that align Class 2B results as present primarily in very high risk patients. In the context of NCCN risk groups, the GEP test result provides significant risk information, as seen in multivariate analysis, meaning the test's validity and stratification can be considered additive to NCCN risk assessment. When looking only at patients in NCCN high risk group, patients with a Class 2B result have a 37.5% metastasis rate, nearly four times higher than the overall cohort risk, and nearly 10 times the risk of NCCN high risk patients with" Class 1 result. Moreover, this rate of metastasis is similar to the risk observed in NCCN very high risk group, which makes the metastatic risk predicted by Class 2B results appropriate for more intensive management options outlined in guidelines, including adjuvant radiation therapy, use of surveillance imaging, SLNB, and more frequent follow-up.

As with NCCN risk stratification, multivariate analysis demonstrates the test adds significant risk stratification information in addition to BWH staging.¹ Patients with BWH T1 tumors are only very high risk by NCCN criteria in the setting of desmoplasia or lymphovascular invasion, both of which are very rare findings. In the validation study, patients with T1 tumors and a Class 2B result have a 43-44% risk of metastasis, similar to the risk observed in the overall T3 population. T3 is the highest risk stage by BWH stage, and these patients are routinely offered more intensive treatment, including adjuvant radiation therapy, surveillance imaging, SLNB, and increased frequency of follow-up. Again, this shows that the risk found by a Class 2B in patients that are not identified as very high risk by any risk assessment system have a level of risk eligible for more intensive therapy. Contrary to assertions in the LCD, 9/23 (40%) of Class 2B results occur in patients with BWH T1 disease, again finding a substantial portion of Class 2B results occurring in patients at lower risk by staging systems.

When considering AJCC division of risk, in patients with T1 tumors who are at the lowest end of metastatic risk in this cohort, a Class 2B result is associated with a 50% risk of metastasis, similar to the overall metastatic risk in AJCC T4 patients which represent the highest risk assessed by AJCC. This level of risk is dearly eligible for more intensive treatment, including adjuvant radiation therapy, surveillance imaging, and SLNB.¹ Again, contrary to assertions in the LCD, 12/23 (52%) of Class 2B results occurred in AJCC T1 tumors.

The impact of these results on patient care have been demonstrated in the real-world clinical utility study assessing management plans from 34 physicians who routinely use the GEP test clinically in 6 real-world patients.² Patient cases 4, 5 and 6 in this study are clinically tested patients who all meet the criteria of "not very high risk by any staging or risk assessment system." These patients are all BWH T1, AJCC T1, and NCCN high risk. In these patients, a Class 2B result significantly increased management plans to the higher intensity, compared to low to moderate intensity management plans at baseline (Hooper et al, Figure 3). When assessing the specific management changes, a Class 2B result led to 70-80% of physicians increasing frequency of follow-up for the first three years, 80-90% of physicians assessing nodes using imaging, ~70% increase in SLNB recommendations, 70-90% increasing surveillance imaging for first 2 years, and ~70% increasing recommendation for adjuvant radiation therapy. These results demonstrate consistency in utility of the test in patients with similar baseline clinicopathologic risk and a high risk Class 2B result.

This single evidentiary analysis is one example of the lack of detail and thoroughness exhibited in many places throughout the evidentiary review for this test. This is one statement that had several analyses, across multiple publications, that could have been reviewed and cited to provide a more accurate depiction of the level of evidence supporting the test. I urge MolDX to take a close look at the many errors and inaccuracies in this policy, which defy a factual, evidentiary review, and revise the analysis of evidence. This GEP test should be covered for the benefit of Medicare patients.

References were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

I am writing to express my disagreement with the policy MolDX: Molecular Biomarker Testing for Risk Stratification of Cutaneous Squamous Cell Carcinoma (cSCC). My primary concern with the policy is that it does not consider the accuracy and clinical utility of DecisionDx-SCC test results as an adjunct to current staging or risk-factor-based assessment. This is a critical flaw as physicians don’t simply stop their use of other risk assessment methodologies when they incorporate novel and independent genetic information. Instead, physicians work to layer information that they know already from their patient’s clinical presentation (eg: clinical size of the lesion) and the diagnostic pathology report (eg: presence or absence of PNI) with genomic information from DecisionDx-SCC (eg: Class 2A result). It is in this way that the DecisionDx-SCC test is actually used by the doctors who are ordering this test for their patients. The policy review of DecisionDx-SCC should take this into perspective and evaluate the clinical validity and clinical utility of the test when used in conjunction with comprehensive risk assessment from clinical and pathological information.

Doctors are trained to order tests that they know – before they order – will be useful to guide their decision making and help them take the best care of their patients. I have reviewed the literature supporting clinical validity of DecisionDx-SCC and have implemented use of this test to inform my management decisions for patients with high-risk cSCC. I use the test as part of shared decision making with my patients to decide on aspects of their care that follow definitive surgical removal of their cSCC (eg: clinic follow-up intervals, the need for surveillance imaging and adjuvant radiation therapy referrals). I believe that the DecisionDx-SCC test, when interpreted in the context of the patient’s known other risk factors, allows me to present my patients with a more informed picture of their regional and distant metastatic risk to then guide personalized, and risk-aligned management plans.

This non coverage policy is at odds with the published evidence and current clinical adoption by physicians, such as myself, who use this test to improve accuracy of cSCC metastatic risk prediction and guide subsequent risk-aligned management decisions for their patients. I encourage MolDX to review the published clinical validation data and fully evaluate the use of the test in combination with other risk-factor-based assessment systems to give a more clinically grounded perspective. This data is available in the Future Oncology publication by Ibrahim and colleagues that is cited in the MolDX LCD, but incompletely reviewed. In doing so, I believe the evidence answers many questions raised in the MolDX LCD regarding clarity on interpretation of clinical, pathological and genetic information to guide management decisions, and will support the conclusion that this test meets the criteria of being medically reasonable and necessary for Medicare and should be considered as a covered test in the final policy.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

As a dermatologist and Mohs surgeon, managing patients with cutaneous squamous cell carcinoma is a daily challenge as patients present with significant uncertainty in risk of progression and tools and staging methods to determine risk are inadequate. The 40-GEP test available from Castle Biosciences represents a significant advance in the accurate risk stratification of these patients. When compared to staging methods, the 40-GEP has been published to demonstrate improvement in accuracy metrics.

The draft LCD implies that the comparable PPV of Class 2 (Class 2A and Class 2B combined) to existing staging criteria, the comparable HR of a Class 2A result comparable to that of BWH T stage T2b/T3 or AJC T3/T4, and the comparable predictive value of Class 1 and Class 2A results to existing staging diminish the value of the test. This is not supported by published evidence.

First, while the PPV may be comparable, a Class 2A/2B result has more than 25% improved sensitivity, making the claim that Class 1 and Class 2A results have comparable total predictive value to staging false. Though the same proportion of patients with Class 2A or Class 2B results experience progression as BWH or AJCC high-stage patients, the Class 2 result captures more than 25% more patients with poor outcomes (see Table 1 in Summary of Evidence in the draft LCD). Even so, the GEP result is used with staging to improve accuracy, not in parallel to staging, making the comparable PPV less relevant to interpretation of test results.

Second, it is critical to understand that a hazard ratio represents the magnitude of risk predicted by the independent factor. As such, the similar hazard ratio of Class 2A to staging or other risk factor only strengthens the assertion that the test identifies meaningful levels of risk known to impact patient management as the same level of risk is already used to make management decisions. Moreover, the Class 2A result is used with all other risk information to predict risk, and the hazard ratio is combined with hazard ratios of other risk factors to form a complete assessment of risk. The draft LCD fails to represent that the test results work with staging and other risk factor assessment to improve accuracy. The draft LCD requires an update in which the statement that Class 1 and Class 2A results have similar predictive value to staging is removed, the additive value of hazard ratios from multiple risk factors, including GEP results is acknowledged, and comparison of PPV between stage and Class results is supplemented by discussion of improved sensitivity.

This objective tool represents a substantial advancement in risk stratification of patients with high-risk SCC with one or more high risk factors over the current standard of care staging systems and is supported by a well-developed body of evidence. Medicare coverage of the 40-GEP test is important to ensure its continued availability to patients.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

I am writing to address the significant value of the DecisionDx-SCC test in making decisions in the management of patients with cutaneous squamous cell carcinoma (cSCC). Specifically, the importance of the test in determining follow up intensity. Frequency of follow-up decisions play a crucial role in optimizing patient outcomes and resource utilization. It is essential to recognize that not all patients with cSCC require continued surveillance from specialized providers. Dermatologists, with their specialized expertise in skin conditions, are well-positioned to provide appropriate and effective long-term follow-up for low-risk cSCC patients. De-escalation and escalation strategies have been shown to be effective and safe for appropriately selected patients and are outlined in the NCCN guidelines for management of cSCC. Follow up frequency recommendations are presented in a spectrum and physicians are responsible for determining the appropriate intensity based on each patient’s risk for metastasis. The DecisionDx-SCC is a great tool to have in my “tool belt” to determine that risk. I encourage you to consider the SCC clinical data as a valuable resource in your review process. Its evidence-based recommendations and documented consistent changes will help inform decision-making regarding the de-escalation of care to dermatology follow-up for low-risk cSCC patients. As we strive to provide patient-centered care and optimize healthcare delivery, it is crucial to align our practices with the evolving evidence and expert consensus. I appreciate your time and consideration, and I eagerly await any insights or actions that may result from our communication.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

Thank you for the opportunity to express my support for the SCC GEP test. Cutaneous squamous cell carcinoma guidelines from the NCCN include many factors that can be considered as high risk/very high risk and leave the responsibility to interpret risk information into the best management plan for each patient with little guidance. In publication by Ibrahim and colleagues, a 40-GEP prognostic assay for cSCC was examined in the context patients with at least one high risk feature by NCCN guidelines in a retrospective cohort and found that the SCC GEP test was able to risk stratify patients into low, moderate and high-risk tumor biology within both the NCCN high and very high-risk categories. By improving the accuracy of risk prediction within these classifications of risk by NCCN, practicing clinicians such as myself are given more objective information to determine which patients are at risk for developing metastatic events and should undergo additional follow up, referral, or adjuvant therapy.

Additionally, as the number of high-risk factors increased, so did the patient’s risk for metastasis, and the SCC GEP test was able to maintain risk separation (Ibrahim Future Oncology 2022). The current draft LCD seems to suggest that this is a limitation of test performance by highlighting that the increased risk in NCCN very high risk Class 2B patients over NCCN high risk Class 2B patients demonstrates the importance of clinical and pathological factors. This interpretation assumes that only one risk prediction modality can be accurate in a population subset, which is of course, not true. Multivariate analysis demonstrates that some risk factors maintain independence of risk prediction while the SCC GEP test also provides independent risk prediction. In fact, the use of multiple risk factors to predict risk of metastasis in patients with SCC is already the current clinical paradigm; the list enumerating NCCN high risk and very high risk factors demonstrates that paradigm. Clinicians are not intended to only use depth of invasion, or any other single factor in their assessment of patient risk. In fact, use of the SCC GEP test can be considered akin to another risk factor, and can be added to the comprehensive clinical and pathological assessment of the patient to drive management decisions. I respectfully request that this interpretation of the data and current management paradigm for patients be included in the LCD and that the SCC GEP test be covered under this policy.

The improvement of risk stratification beyond the classification present in NCCN guidelines with improved accuracy fills a critical unmet need in the treatment of patients with cSCC. When coupled with clinical utility evidence that demonstrates the test changes specific management modalities for patients (Hooper Cancer Investigation 2022) and in the context that these modalities are already management options included in NCCN guidelines (Farberg CMRO 2020), demonstrate the SCC GEP test medically reasonable and necessary for the appropriate care of these patients.

I fully endorse Medicare coverage of these tests.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

I am submitting this letter to you regarding the proposed LCD on behalf of the National Society for Cutaneous Medicine in support of coverage for the DecisionDx-SCC 40-GEP test for use in patients diagnosed with cutaneous squamous cell carcinoma and one or more clinicopathologic risk factors.

The National Society for Cutaneous Medicine is an evidence-based, non-profit organization comprised of over 500 dermatologists and experts in cutaneous oncology. I can attest to the need for a more precise prognostic tool to predict metastatic risk for cutaneous squamous cell carcinoma (SCC). I have followed updates to risk assessment strategies for patients with SCC over the years, including revisions to the American Joint Committee on Cancer Staging Manual and the newer Brigham and Women’s Hospital tumor staging system for SCC. All of this, in conjunction with the continual updates to the National Comprehensive Cancer Network’s treatment recommendations for SCC, represent a continuously shifting risk assessment and management guidance framework that are a direct result of inadequacy of current risk assessment strategies. These models contain only clinical pathological information and do not include the critical information obtained by genomic analysis that can lead to a more precise prognostic assessment.

The DecisionDx-SCC test available from Castle Biosciences meets a critical need for an unbiased and objective, molecular test to better assess prognosis and inform recommendations for individual patients with uncertain metastatic potential. Moreover, it meets or exceeds the level of evidence that has been required of other similar tests that have been made available over the years, and yet, this proposed Medicare policy sets a new, and arguably unattainable, standard for new tests not previously required in the field of dermatology, nor oncology as a whole.

First, most risk factors that are currently included in NCCN guidelines to guide risk prediction of SCC are guided only by retrospective studies. And in fact, none of these risk factors or staging systems have demonstrated that their use improves outcomes (see meta-analysis published by Zakhem et al. Cancers 2023). And yet, these are the tools currently used every day to guide patient management decisions for SCC patients.

There are many examples of other tests that have not demonstrated improvement in outcomes that have had Medicare coverage for years. In fact, the recently posted and soon to be finalized policy MolDX: Prognostic and Predictive Molecular Classifiers for Bladder Cancer establishes a test as covered with no mention of improvement in outcomes. In fact, as with previously posted policies for Breast Cancer and Prostate Cancer prognostic tests, this LCD cites the hierarchical framework of Fryback and Thornbury and the ACCE Model Process for Evaluating Genetic Tests and cites:

“When a proven, well established association of pathway is available, intermediate health outcomes may also be considered. For example, if a particular diagnostic test result can be shown to change patient management and other evidence has demonstrated that those patient management and other evidence has demonstrated that those patient management changes improve health outcomes, then those separate sources of evidence may be sufficient to demonstrate positive health outcomes from the diagnostic test. For patients with bladder cancer, an array of treatment possibilities exists in all stages of disease. Clinicians must consider not only the potential treatment options but must also make an individualized risk to benefit assessment to determine how to treat a specific patient. Diagnostic tests that aid in this assessment are expected to change physician management in a way that improves patient outcomes.”

This rationale and subsequent coverage of the prognostic bladder cancer test is in clear opposition of this presently discussed policy, which is requiring health outcomes and fails to acknowledge the impact of DecisionDx-SCC in guiding management decisions in multiple treatment options outlined in guidelines that have been shown to or are expected to improve patient outcomes.

In light of this policy and coverage criteria, it is the responsibility of MolDX to hold consistent level of evidence requirements for similar tests between disease states to avoid incongruence in availability of effective, advanced tests between disease states. In fact, the only observable difference between these tests appears to be NCCN guideline inclusion. However, relying on NCCN guideline inclusion to grant coverage to a test or service with comparable levels of evidence further reinforces disease state incongruity as the guideline panels are distinct from each other and each set their own level of evidence requirements to determine inclusion.

Moreover, guideline panels are consensus-based and not evidence based; they do not publish evidentiary reviews and do not disclose rationale behind determinations. Thus, the level of evidence deemed sufficient by one panel does not reflect the level of evidence required by another, with little transparency as to differences between the panels. Therefore, because Medicare coverage is anchored in evidentiary review, regardless of NCCN guideline inclusion or exclusion, the level of evidence supporting one test for coverage must also be sufficient for another, similar test to receive coverage. DecisionDx-SCC has a level of evidence at least on par with, if not exceeding, the prognostic tests included in the final bladder cancer LCD.

It is my strong belief that denying the use of the tool will be detrimental to Medicare beneficiaries, and will disproportionately burdens patients with SCC relative to other cancers, such as bladder cancer.

Medicare policy should require clear and consistent, evidence-based criteria for coverage from test to test in the best interest of the patients that Medicare serves.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

High risk cutaneous squamous cell carcinoma (cSCC) is a difficult disease to manage. While the majority of patients have a low risk of metastasis, it is difficult to accurately identify patients with the highest risk for which more intensive management strategies should be considered. Clinical and pathological factors can be used to estimate risk; however, the accuracy of individual factors is poor and as such, lack consistent recommendations for use in clinical decision making. AJCCv8 (only for head and neck) and BWH staging systems improve upon accuracy of individual factors alone and attempt to unify factors, however, these staging systems still have substantial room for improvement and as a result are not routinely used in clinical practice. It is important to acknowledge that not all high stage patients are treated with the most aggressive therapy available as indicated by NCCN guidelines. NCCN guidelines do not incorporate staging to guide management, and instead leave interpretation of the constellation of risk factors and the comprehensive risk of the patient to progress to inform decisions across multiple treatment modalities.

The 40-gene expression profile test by Castle Biosciences can add valuable prognostic information to help clinicians determine the most appropriate patient management modalities for patients with cSCC, including those with multiple high-risk features or high tumor stage (T3 by AJCC or T2b by BWH system).¹ Importantly, the expanded validation data published in Ibrahim, et al. demonstrates that the test stratifies risk within NCCN guideline risk classification (high risk and very high risk factors). In patients classified as having NCCN high risk or very high risk, the test provides significant stratification of risk of metastasis. While the proposed LCD does summarize this data, statements appear to misunderstand the value of an objective test that can be combined with NCCN risk assessment. The proposed LCD voices that, “Interestingly, when cSCC cases are stratified by NCCN criteria into high-risk and very high-risk populations, the metastatic risk associated with a Class 2B result was 37.5% in the NCCN high risk cohort versus 60% in the NCCN very high risk cohort, perhaps reflective of the overall difference in prevalence of metastasis in both cohorts and underscoring the importance of additional clinicopathologic risk factors.” This analysis masks the robust consistency of the FOLD CHANGE associated with the DecisionDx-SCC class results across cohorts and fails to reflect the published clinical validity and clinical utility data that emphasizes the importance of interpreting the GEP test result in the context of the clinical and pathologic risk factors for each patients SCC. It is critical to note that it has never been asserted in published literature that clinicopathologic risk factors become unimportant or no longer predictive with use of the test, or that the test should be used outside the context of other risk information. In fact, published literature makes the use of the test in the context of other risk information abundantly clear.²

When treating a patient with a high risk cSCC, the treating clinician will know the complete range of clinical and pathologic factors that have been identified for the patient. What this means is that the treating clinician will know if the patient’s tumor would be considered “high risk” or “very high risk” by NCCN (or alternatively their AJCCv8 or BWH stage). Therefore, the data comparison of predicted risk by Class 2B results between NCCN high risk and very high risk patients does not answer any clinically meaningful questions as this analysis from the LCD is arbitrary and irrelevant when considering predictive accuracy or clinical utility of the test. Instead, a comparison of Class results within NCCN high risk and very high risk answers a clinically meaningful question – when clinicians use the DecisionDx-SCC test result, the treating clinician is interested in obtaining more accurate risk stratification by interpreting the DecisionDx-SCC result in the context of the known risk factors for their patient, not comparing the relative impact of the Class 2B for tumors with different types of risk factors present.

When considering this data, NCCN high risk patients have an overall metastatic risk of 9.8%, and a Class 1 result is associated with a 4.1% risk of metastasis, whereas a Class 2B result is associated with a 37.5% risk of metastasis. In NCCN very high-risk patients with an overall metastatic risk of 23.0%, a Class 1 result is associated with 11.9% risk of metastasis, whereas a Class 2B result is associated with a 60.0% risk of metastasis. While it is true that Class 2B patients in the very high risk cohort have higher metastatic risk than Class 2B patients in the high risk cohort, this is to be expected. The test does not need to invalidate other risk stratification systems to demonstrate clinical value. The consistency of findings that within each risk grouping a Class 1 is consistently associated with half the metastatic risk of the overall cohort, while a Class 2B is associated with ~3 times the risk of the overall cohort demonstrates a clinically useful and actionable tool. This is the most relevant description of the data, which is absent from the proposed LCD and should be included. Of note, this information is also available in the context of BWH staging and AJCC staging. Though the proposed LCD faults the test for lack of definitive value above the combination of available clinical, pathological, and staging criteria, this evidence is actually published in the supplement of the Ibrahim, et al manuscript published in 2022 and should be included (see figure 2 and table 2 of supplemental material).

Evidentiary review supports, repeatedly, that using the test with other risk assessment tools gives a more accurate prediction of metastatic risk, and when used to guide management decisions that are already being made today on less risk information, provides a clinically useful and actionable result. All statements suggesting or implying otherwise should be removed from the proposed LCD as they are not grounded in review of published evidence, and appear to be the conjecture of atypical, and often irrelevant, comparisons of data.

References were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

Cutaneous squamous cell carcinoma (SCC) is the commonest form of skin cancer in the U.S.^1,2 While the absolute rates of metastasis and death from this disease are low (~2-6% and ~2%, respectively),^3–8 given the high incidence of this disease, this translates into a large number of individuals who develop recurrent disease. Given the median of diagnosis (70 years), this makes cSCC and particularly recurrent cSCC a critical issue for Medicare and Medicare beneficiaries. While most patients with SCC are cured after definitive surgery, others have more aggressive tumors that can metastasize. To this end, adjuvant radiation therapy (RT) is offered to patients with high-risk resected disease, and several phase III trials are evaluating the role of anti-PD-1 immunotherapy following RT in this patient population. However, despite RT, a substantial proportion of patients remain at high-risk of developing recurrence or metastatic disease,^2,7,9–12 and it is critical to identify patients having tumors at high risk for metastasis.

The American Joint Committee on Cancer Staging, Eighth Edition (AJCC8) AJCCv8 staging was established for patients with squamous cell carcinoma. However, in its current iteration, AJCC8 only applies to patients with cSCC tumors located on the head or neck which limits its broad applicability. The Brigham and Women’s Hospital Tumor Classification System (BWH) is an alternative system that assigns a greater weight to high-risk features including perineural invasion. Although both AJCC 8 and BWH appropriately classify high-risk tumors, AJCC 8 upstages twice the number of tumors compared to BWH, and the use of the BWH tumor classification enables more accurate staging of low-risk patients. However, there is substantial heterogeneity in both AJCC8/BWH high-risk and low-risk patients, and it is clear that there remains a pressing need for prognostic scores that are capable of identifying patients at risk for recurrence, metastasis, and death.

The gene expression profile test by Castle Biosciences, Inc. (DecisionDx-SCC) has been clinically validated for stratifying high-risk SCC patients as being either low, moderate, or high risk for metastasis based on gene expression in the primary tumor.¹⁹ In SCC patients considered to be at high risk for poor outcomes by tumor staging, such as AJCC8 stage T3, the test by Castle can further refine risk assessment to determine if an individual’s tumor is likely to develop metastasis within three years of diagnosis and, therefore, if the patient’s management should be elevated in a risk-appropriate manner. As shown in Ibrahim, et al. published in 2022 (supplemental figure 2 and table 2), all patients with an AJCC8 T3 tumor and a Class 2B result experienced metastasis, whereas a Class 1 result (achieved in 44% of patients) had an overall metastasis rate less than half of the overall T3 population, and similar to the overall metastasis rate in T1 tumors.¹⁹ This is a clearly impactful result in some of the highest risk patients identified by AJCC8 staging, and identifies not only the needed improvement for risk stratification within staging systems, but also the clear clinical actionability of the test result.

Importantly, the proposed LCD cites comparisons between BWH and AJCC staging, but fails to acknowledge that all studies conducted comparing accuracy of risk prediction only includes head and neck tumors and confounds attempts to compare published BWH and AJCC accuracy metrics to that of the 40- GEP test. For that reason, the most appropriate comparison of other staging systems to the 40-GEP test should be in the same cohort; in essence, the analysis conducted in the validation is the most appropriate comparator of accuracy. This should be acknowledged in the final LCD, and the correct conclusion that the 40-GEP is the most accurate predictor of risk when compared to existing staging should be stated. This data was published in Ibrahim, et al., available at the time the LCD was drafted, and yet, the LCD seems to fail to understand that this demonstrates precisely the added value of the test to staging and how to use the test in the context of staging. This molecular test can be used as an ancillary tool to further clarify the metastatic risk in high-stage patients to improve decision making. This data was not discussed in the proposed LCD and should be included in lieu of statements regarding unclear validity and utility of the test in comparison to staging methods.

As a clinician and researcher in the field, I think it is important to realize that over 30% of SCC tumors staged as low-risk progress to metastasis^13–15 and more than 60% of those staged as high-risk do not develop metastasis.^16–18 This represents is a substantial proportion of patients with SCC who are potentially being sub-optimally managed. More accurate prognostication to identify which high risk SCC patients are truly at high risk for metastasis, and which are not, is in the best interest of the patient and will allow for more appropriate allocation of our healthcare resources. I consider DecisionDx-SCC to be a useful adjunct in a subset of patients with high-risk SCC. Thought should be given to potentially covering this test in this patient population, particularly given the unmet clinical need and robust data supporting the use of this test.

References were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

I was surprised to see the non-coverage policy from MolDX given the strong body of evidence that supports this test. I submit this letter to support the positive coverage of DecisionDx-SCC as a critically important clinical tool to stratify risk in high-risk patients with SCC, and to clarify questions raised in the policy regarding the UTILISE study. UTILISE is an ongoing prospective clinical utility study that is collecting management plans pre- and post-test and subsequent outcomes in patients with high-risk SCC who are clinically tested with DecisionDx-SCC. A publication from December 2022 analyzes data from this study to determine management plan changes in Medicare patients and demonstrates 24% of patients had a recommended management change as a result of the test results. Of note, this study is comparable in design and demonstrates a rate of utility similar to other broadly used and covered tests in breast cancer and prostate cancer. Additional, unpublished details on specific management plan changes, as opposed to generalized management changes that are described by management plan intensity were requested. First, it is important to note that the change in relative management plan intensity (low, moderate, high) was reported by the clinicians in study and are not a post hoc interpretation of management plan changes. This is the assessment of the management plan intensity from the clinicians themselves before and after reviewing GEP results. Specific management plan changes are available in the data, however, were not included in the publication as an overall assessment of test impact and change in management plans due to diverse baseline management strategies employed pre-test from patient to patient, even when accounting for risk factor assessment. This diversity in pre-test management plans is reflective of the latitude given in guidelines for patient management and insufficiency of staging to accurately predict metastatic risk. Clinical impact studies, such as those published by Hooper in 2022, suggest that incorporation of test results into clinical and pathological risk factor assessment can unify management strategies, but also show wide range of pre-test management plans. In the patients with management changes, 9 patients received Class 1 results and 5 patients received Class 2A results. In patients with Class 1 results, the primary management change was a decrease in follow-up intensity, either by decreasing frequency or by referral back to dermatology for follow-up (instead of with Mohs practice). In patients with Class 2A results, the primary management changes were increased follow-up intensity, referral to medical oncology, and increase in nodal assessment, including the use of imaging to surveil the nodal basin. Other management changes observed included referral for surveillance imaging, and referral for adjuvant radiation therapy. The majority of these patients were BWH staged T1 and T2a; thus, management plans for these patients demonstrate consistency with published management algorithms. Table 4 in the publication exhibits perception of metastatic risk and shows that after a Class 1 result, the majority of clinicians decreased their perception of risk from 5-10% to <5%. Questions were raised regarding congruence of this perception with the published metastatic risk predicted in patients with Class 1 results and more than 1 risk factor, which is 9% in the Ibrahim study. Total risk factor count is one way to assess clinical and pathological risk that is interpretable regardless of the risk assessment system considered. However, it does not account for the accepted doctrine that certain risk factors predict a higher level of risk than others; this is demonstrated by the tiering of risk factors in NCCN into high risk and very high risk, and the selection of a subset of these very high-risk factors for incorporation into BWH and AJCC staging. It is possible to have a patient with 3 risk factors that is a T1 by BWH criteria (if all risk factors are NCCN high risk and not one of the BWH staging factors) and to have a patient with 3 risk factors that is a T2b by BWH criteria (if all risk factors are BWH risk factors). As such, the pre-test risk assessment in this cohort compared to total risk factor count does not account for the full clinical picture. In fact, as mentioned above, the majority of these patients were BWH stage T1 or T2a, both of which are classified as low risk by BWH staging. I use the test regularly in clinical practice myself to manage patients with high risk SCC. One recent case that came to my practice was a Medicare patient with 1 risk factor located in the head/neck region staged as a BWH T1 and AJCC T1 with a Class 2B result. Before receiving the test result, my plan was to follow the patient every 6 months for 2 years, however, after receiving the Class 2B result, associated with a metastatic risk of nearly 50%, I referred this patient to ART. This management decision is in line with published utility studies, management algorithms, and importantly, NCCN guidelines as the patient had a very high risk of metastasis. This is a clear demonstrated utility with specific management change, in line with NCCN guidelines, that interprets risk by considering clinical, pathological, and Class results. The commitment the organization has shown for continued evidence supporting their tests is obvious. The DecisionDx-Melanoma test offered by Castle Biosciences has more than 45 peer reviewed publications that support the use of the test, with more than half of those publications produced in the last 5 years. The same should be expected for evidence supporting DecisionDx-SCC. Additional data from the UTILISE study will be published in a follow-up manuscript with increased patient enrollment and increased follow-up within the next year. Requirement for proven improvement in outcomes in clinical utility studies has not been required for molecular diagnostic or prognostic tests to receive Medicare coverage in the Medicare Program Integrity Manual definition of medical reasonableness and necessity. Moreover, other recently published MolDX policies have granted coverage to molecular prognostic tests without specific management change information by citing that impact to management plans in a risk-aligned manner that are demonstrated management strategies in guidelines can be assumed to have an improve management and subsequent impact on patient outcomes (refer to MolDX: Prognostic and Predictive Molecular Classifiers for Bladder Cancer). Even so, given the long lead time required to generate clinical utility data, it is noteworthy that many prior MolDX policies have been finalized with positive coverage for tests pending anticipation of additional data (refer to MolDX: Melanoma Risk Stratification Molecular Testing). This can appropriately allow test availability for innovative tests that improve patient care while additional data is developed. Given the strong data that has been published supporting the DecisionDx-SCC test, this approach is one viable option to issue Medicare coverage for an important test. This test represents a clinically necessary tool that improves risk-aligned patient management plans in a prospective study and should be a covered service for Medicare beneficiaries.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

As a board-certified AP/CP Pathologist, and dermatopathologist, I am writing to provide a critical assessment of the evidence review and non-coverage determination from MolDX and to advocate for the positive coverage of the DecisionDx-SCC test in the final Local Coverage Determination. Below, I will provide some perspective on the criticisms presented in this Determination regarding histopathological factors used in risk assessment and reporting of these factors by dermatopathologists that impact the discussion in the Determination.

Unlike other malignancies such as colorectal cancer or breast cancer – or even other cutaneous malignancies such as cutaneous melanoma – the College of American Pathologists (CAP) does not have a current synoptic reporting standard for diagnostic biopsies of primary cutaneous squamous cell carcinoma (cSCC). The absence of standardized reporting criteria for cSCC poses a challenge in implementing staging derived from diagnostic pathology reports because of the dramatic variation in the reporting practices observed across the country. While it is generally true that interpreting pathologists will comment on the presence of particularly impactful risk factors in this disease state (such as the presence of perineural invasion) interpreting pathologists are not mandated to comment on these risk factors and the structure of the comment is not standardized (such as documenting the mm measurement of the nerve involved with cSCC). This is clearly problematic when assessing the Brigham and Women’s Hospital stage of a tumor, which requires PNI ≥ 0.1 mm to be considered a risk factor, and which stands in contrast to the broad clinical use of PNI of “any” caliber nerve to inform clinical management decisions. Thus, comparisons to existing staging systems are imperfect; in clinical reality, physicians may not have enough information to fully stage their patients and may use slightly different histologic thresholds than those articulated in guidelines based again on the heterogeneity of reporting practices.

Surprisingly, the draft LCD cites the German literature from the 1990s to describe the importance of desmoplasia in the risk stratification of cSCC (Reference #76: Breuninger H, Schaumburg-Lever G, Holzschuh J, Horny HP. Desmoplastic squamous cell carcinoma of skin and vermilion surface: a highly malignant subtype of skin cancer. Cancer. 1997;79(5):915-919). Desmoplasia is certainly considered as an NCCN very high risk factor in current NCCN guidelines for cSCC (v1.2023), however it is not incorporated in AJCCv8 or BWH staging. As mentioned above, there is no CAP synoptic report for primary cSCC, and the current CAP synoptic report for node positive cSCC of the Head and Neck (which is only for “N+, and pT3 and pT4 cutaneous squamous cell carcinoma including squamous cell carcinomas of dry vermillion lip and commissure”) does not even include “desmoplasia” as a reportable field, further highlighting that this is an uncommonly reported finding in the US (https://documents.cap.org/protocols/HN.SCC_1.0.0.1.REL_CAPCP.pdf [documents.cap.org]).

What this type of evidence review highlights is that the Determination has gone to great lengths to accentuate the level of evidence underpinning the NCCN included high risk and very high risk factors - many of them histologic in nature - which has the effect of obscuring the fact that the level of evidence supporting the histologic factors included in the NCCN is not as strong as readers would be led to believe. Perhaps an even better example of this fact is from a high impact, yet uncited, manuscript that discusses the different approaches used in clinical practice to measure depth of invasion in primary cutaneous SCC. The study from Yildiz et al.[i] highlights the open debate in dermatopathology regarding the different ways pathologists can measure depth of invasion for SCC. The existence of this debate highlights that the data supporting depth of invasion in SCC is less robust than in other cutaneous malignancies such as melanoma (where Breslow depth is standard) and that the appropriate approach to depth measurement is not clarified in published CAP synoptic reporting guidelines.

While the cSCC histologic factors certainly have a role in prognostication when present or reported, they do not achieve a level of reliability to mandate comparison in order to validate an additional, independent prognostic tool (eg: in a nomogram).

The inconsistent and varied reporting practices across the US for the histologic risk factors incorporated in the NCCN and AJCC/BWH leads to potential discrepancies in patient risk stratification, management and outcomes. These inconsistencies are compounded when we look at the weak to moderate concordance of histologic grading between dermatopathologists and between MOHs surgeons.[ii] This underscores the clear unmet clinical need to improve the accuracy of risk stratification in cutaneous SCC. What is more, the evidence level underpinning the currently included NCCN high risk and very high risk factors is not as strong as is suggested by the Determination, and the DecisionDx-SCC test has met or exceeds the level of evidence behind the clinical and pathologic risk factors incorporated into current NCCN risk stratification.

I would like to directly address a few other criticisms mentioned in the Determination. First, the determination states “The relative rarity of the Class 2B result is depicted in the Ibrahim et al validation cohort of 420 patients wherein only 23 patients received a Class 2B result (12 metastatic), 18 of whom had 2 or more risk factors.” This is true, however, cSCC that metastasizes is itself relatively uncommon.[iii] Therefore, if one wants a molecular diagnostic test with an actionable and high PPV (such as Decisiondx-SCC) the highest risk result will logically be relatively uncommon. Second, when considering current staging systems we also must remember the AJCC-8 apply only to tumors from the head and neck while cutaneous squamous cell carcinomas of the skin outside of the head and neck do not currently have an AJCC staging system. Lastly, I would like to point out that “there is no standardized follow-up schedule for patients with cSCC.[iv]

To this last point, I would like to share my practical knowledge since I have had clinical experience with this test. I have discussed clinical management with several providers who use the Decisiondx-SCC test to help arrive at appropriate management decisions. As the AJCC/BWH criteria are imperfect themselves, several clinicians use the DecisionDx-SCC test when additional information can aid in choosing appropriate follow up management. Depending on the results of the test and the clinical scenario, a clinician will be less aggressive in their treatment if a class 1 is resulted. However, a class 2a and definitely 2b will push a clinician to be more aggressive in their treatment even if other factors made them reluctant.

As such, I wholeheartedly believe that the evidence review in the proposed LCD should be re-evaluated – supplemented by omitted articles and level-set to balance the well-known limitations of histologic risk factor reporting for this disease state. Finally, after this re-review, DecisionDx-SCC should be reclassified as a covered test as it has met the standards of clinical validity and utility for Medicare.

References were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

The 40 gene assay for cSCC is an advancement in personalized medicine that is transforming the field by delivering the ability to tailor management plans more specifically to a patient’s specific risk of metastasis. The test has been well validated to provide independent risk information that can be combined with other clinically relevant information to make the best, risk-aligned management plan for patients. Although the test informs many management strategies, the absence of discussion of the utility of the test for follow-up decisions is notable. In the publication, “Impact of a prognostic 40-gene expression profiling test on clinical management decisions for high-risk cutaneous squamous cell carcinoma” published by Dr. Graham Litchman in the journal Current Medical Research and Opinion in 2020, 162 physicians included in the study evaluated 2 patient vignettes (both BWH stage T2a) and formed management plans before and after receipt of test results. In this study, follow-up intervals were substantially impacted by test results. At baseline, with no test result, most physicians agreed on a follow-up plan of 2-4 times a year for both patients. However, a Class 1 result led to ~20% more physicians preferring a reduced follow-up time of 1-2x per year, while a Class 2B result led to ~35% more physicians preferring an increased follow-up time of 4-12x per year. Of note, this aligns with the management algorithm published by Aaron Farberg and colleagues in the article, “Integrating gene expression profiling into NCCN high-risk cutaneous squamous cell carcinoma management recommendations: impact on patient management” in the journal Current Medical Research and Opinion in 2020. In this manuscript, Farberg and colleagues establish a management algorithm for patients based on incorporating test results with staging information and applying to the framework in NCCN guidelines. It is noteworthy that although NCCN guidelines were updated after this manuscript was published to divide risk factors into high risk and very high risk, the proposed management framework did not change, and this algorithm still applies. In this algorithm, a patient with BWH T2a with a Class 1 result was recommended follow-up 1-2x per year for the first 2 years, while the same patient with Class 2B result is recommended follow-up 4x per year for the first 2 years. Statements made about the lack of demonstrated utility in the context of risk factors or the consistency of use of the test should be re-evaluated as they seem based more on conjecture than review of evidence.

As a Mohs surgeon with a busy practice, it is critical to identify the lower risk patients that can be followed less frequently to better manage my patient flow. In the past any patient that met high risk criteria was seen in my office more frequently than patients without high-risk features. With the introduction of the 40-GEP, I am able to risk stratify my patients and only see those with biologically high-risk SCC as frequently. In addition, patients with low-risk Class 1 results can also be de-escalated to follow-up back to dermatology clinics.

Misconceptions about the utility of this test and how it fits into the NCCN management framework create major flaws in the conclusions of this policy that warrant scrutiny and revision. This test is an important medical advancement that is clinically useful and should be covered by Medicare.

I would like to take the example of a gentleman. He had a very large stage 3 tumor that I cleared with Mohs surgery. Historically, based on his tumor, he would have had radiation, imaging with CT scans, and possibly immunotherapy to help prevent metastasis. The expenses for these interventions would easily be over $100,000. Based on his Castle Decision Dx result, which indicated a low risk tumor, I was able to "step down" his treatment and avoid these costly interventions. He is doing well, and we were able to avoid these unnecessary procedures. I hope these points are helpful to you.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

I am writing in support of coverage of the gene expression test developed and validated at Castle Biosciences. Staging systems are helpful during risk assessment of patients with high risk cSCC, but often fall short of accurately predicting risk of metastasis. Therefore, additional prognostic tools like this gene expression test have a critical role to fill this gap. The test has been clinically validated and published data have shown that the test can supplement formal staging and risk assessment methods to enhance prediction of risk of poor outcomes in patients with high-risk cSCC.

The proposed local coverage determination issued by MolDX mistakenly tries to assess the value of the test in the context of risk factors by looking at the association between risk factors and Class call results, and appears to question the independence of the test results with vague statements such as, “The majority of Class 1 results were identified in samples with 1-2 risk factors, confirmative of a low-risk of metastasis. Class 2A and Class 2B were more likely to be found in samples from patients with a greater number of risk factors compared with Class 1”, and “The Class 2B result is rare and occurs more often (though not exclusively) in patients with two or more risk factors, who would already be classified as higher risk by existing tools” followed by conclusion statement that, “current evaluated tests at the time of drafting this policy have not yet demonstrated definitive value above the combination of available clinical, pathological, and staging criteria in accurate risk stratification.”

There are many assumptions and false statements in this misunderstanding of the data that must be corrected in the LCD.

First, no patients in the intended use population of the test are at low risk for metastasis. Moreover, the Class 1 result shows a reduced risk of metastasis, even in the lowest risk subsets by staging or risk factor assessment, when compared to each subset’s overall rate of metastasis. Thus, there is zero evidentiary basis for the statement that a “Class 1 result is confirmative of low risk of metastasis”. Because of this misconception of the value of a Class 1 result, the LCD does not appropriately include the value of a high NPV of a Class 1 result across subsets, that effectively downstages each risk group by one stage or risk group across risk systems. This data is evaluated in Ibrahim Future Oncology 2022 and the NPV can be calculated for Class 1 results in all subsets of NCCN groups, BWH stages and AJCC stages from the data in the main text and in the supplement. This data was available at the time the LCD was drafted, but review of the clinical validity in the context of staging is not included in the proposed determination, despite specific criticism aimed at precisely these questions.

The value of a high NPV Class 1 result in de-escalating care is confirmed in clinical utility studies (see Hooper Cancer Investigation 2022), in which BWH T1 patients with at least one other high-risk factor that makes them eligible for testing, have de-escalated management.

Second, it is reasonable and rational that the Class 1 result is associated with a lower number of risk factors, and that a Class 2A and Class 2B result is associated with a higher number of risk factors. The implications of statements correlated test results with risk factor assessment seem to be that the test duplicates staging and doesn’t add new risk information. Indeed, this is implied again by suggesting Class 2B is more common in patients who would already be identified as high risk by existing tools. Review of prognostic tests across disease states, including those with existing Medicare coverage, will show this is a common finding in molecular prognostic tools; if the Class call was completely discordant with risk factor assessment, the accuracy of the tool would be called into question since it conflicts with published validations of other risk stratification methods. To assess if the Class call duplicates staging, the proper analysis to cite is the multivariate analysis, which demonstrates that the test accurately predicts risk of metastasis after accounting for risk prediction contributed from other risk factors.

Finally, the Class 2B result does not occur most often in patients with highest risk disease, and relying on data that counts risk factors without considering what the risk factors are gives low resolution to actual metastatic risk (for example, it is possible to have a BWH T2b tumor with only 2 risk factors, and T1 disease with 3+ risk factors that are not included in BWH staging). Review of the Supplement Table 2 in Ibrahim Future Oncology 2021 shows that of the 23 Class 2B results, 9/23 (39%) were in T1 patients, 7/23 (30%) were in T2a patients, 5/23 (22%) were in T2b patients, and 2/23 (9%) were in T3 patients. Thus, the majority of Class 2B results are actually found in patients at the low end of metastatic risk by staging systems. Moreover, even in patients at the highest end of risk by staging, clinical utility studies demonstrate that a Class 2B result results in escalation of management (see Hooper Cancer Investigation 2022 and Singh Clinical Cosmetic and Investigational Dermatology 2023), demonstrating that not all T2b tumors are de facto managed at the absolute highest intensity, and that the Class 2B result is actionable in this subset. Of note, in the validation study ~10% of T2b patients find a Class 2B result, which is not a rare finding.

With all things considered, the DecisionDx-SCC test should be an option that is available for use by clinicians and affordable for their patients when assessing each patient’s individual risk for poor outcomes. Use of this test along with traditional risk- assessment methods and expertise while following national guidelines for cSCC could positively impact patient management for better outcomes. Proper analysis and interpretation of the data should be included in the LCD, and based on strong clinical validity and utility data, the test should be a covered service for Medicare beneficiaries.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

I am writing to bring to your attention the DecisionDx-SCC test's significant clinical value and potential benefits in managing high-risk cutaneous squamous cell carcinoma of the head and neck (cSCC-HN) patients. As a practicing physician who is deeply committed to providing the best possible care to my patients, I strongly support the revision and subsequent inclusion of this innovative test in Medicare coverage.

The 40-Gene Expression Profiling (40-GEP) test demonstrated statistically significant independent metastatic risk stratification for high-risk cSCC-HN patients in a recent large multi-center cohort study. Similar findings were published in a smaller cohort (n=278), and are now validated in a larger cohort (n=622).¹ This study provides strong evidence that a 40-GEP test is an indispensable tool for predicting metastatic risk in this patient population, allowing for better treatment decisions.

What sets the 40-GEP test apart is its reliance on tumor biology, rather than risk factors only. Risk factors co-occur with one another, complicating the association of risk factors with outcomes and predictive value of each.² Furthermore, the assessment of these risk factors is subjective, including the degree of differentiation, the ever-evolving definition of PNI, and incomplete biopsy information (eg depth of tumor invasion) at the time of staging, which can have a direct impact on risk assessment, thus patient management.^3-6 40-GEP is independent of clinicopathological factors and staging and is a reliable objective tool. By considering the molecular profile of the tumor, this test offers unique and complementary information to conventional staging methods. As a result, it provides independent prognostic value in predicting metastatic risk compared to staging alone in high-risk cSCC-HN patients. This additional prognostic information allows healthcare providers to tailor treatment plans for each patient’s risk profile.

Implementing the 40-GEP test in clinical practice enables healthcare providers to take proactive measures to mitigate the risk of metastasis and improve patient outcomes. For instance, with the knowledge gained from the test results, healthcare providers can implement closer monitoring protocols, conduct more frequent imaging scans, and initiate adjuvant treatments in a timely manner. These interventions can potentially detect and manage metastasis at earlier stages, thereby improving the chances of successful treatment and long-term survival for high-risk cSCC-HN patients. These treatment decisions are already being made for patients presently diagnosed with cSCC-HN with less precise information, thus, any perceived lack of clarity suggested by Medicare policy on clear and consistent use of the test is missing the point; the improvement in accuracy over risk factor methods alone that has been robustly established means that use of the test is better than care offered by risk factor guided management alone.

By revising and covering the DecisionDx-SCC test under Medicare, CMS can empower physicians and patients with access to a valuable tool that enhances risk stratification and improves clinical decision-making. Integrating this test into the reimbursement framework would ensure equitable access to cutting-edge precision medicine for eligible patients, reducing health disparities and fostering advancements in personalized care.

I request CMS to reevaluate the scientific evidence supporting the clinical utility of the DecisionDx-SCC test and consider the importance of its coverage for high-risk cSCC-HN patients. This revision would significantly contribute to the advancement of precision medicine in the field of oncology and align with CMS's mission of promoting quality care and improved patient outcomes.

References were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

We have recently learned that MolDX has published a non-coverage determination for the 40-GEP DecisionDx-SCC for high-risk squamous cell carcinoma patients (SCC). As recognized leaders in the field of Mohs surgery, we would like to assert our position that this should be a Medicare covered service. This gene expression profile test has provided our community with enhanced risk stratification and has been used to inform clinical management decision making in our practice for the past 2-3 years. Below, we highlight two issues with the non-coverage determination that need to be addressed.

First, the draft LCD does not appropriately evaluate the value of the 40-GEP test with other risk information, possibly due to lack of review of the supplemental data in the validation manuscript published by Ibrahim and colleagues which substantially impacts the conclusion drawn regarding how the test is clinically used. One example of this impact is an unnecessary concern raised over the potential for inappropriate de-escalation of patients with a Class 1 result. This is appropriate as indicated by the metastatic rates associated with Class 1 results published in validation studies. When considering risk assessment by BWH T-stage as published by Ibrahim and colleagues, the Class 1 result is consistently associated with the same level of metastatic risk as one BWH T-stage lower. This is also true in the setting of NCCN risk assessment and AJCC staging. Tangential to this point, but important to raise, is that this reduction of risk also applies to Class 1 results in patients with the lowest level of risk that are eligible for testing; the assertion in the policy that Class 1 results confirm low risk in these patients is a false statement. If a tumor is classified by the 40-GEP test as a Class 1, the patient could be managed less intensely than would be indicated by clinical and pathological factors alone, recognizing that treatment decisions are not necessarily binary. Reducing intensity of management in these tumors does not mean treatment of all patients with a Class 1 result as lowest risk, but rather, as one tier reduced from assessment of clinical and pathological risk alone.

Use of the test to make de-escalation decisions is important to spare the patients with reduced biological risk unnecessary treatment and allocating healthcare resources more appropriately to higher risk patients and is supported by published clinical utility data. In addition, the concern raised for inappropriate de-escalation does not reflect current clinical management of these patients; decisions to tier management approaches based on individual patient risk across multiple factors is the central dogma of clinical practice today and is not prescriptively guided by any available algorithm. Including a test in that practice that has demonstrated accurate, consistent, and objective information can only serve to improve our ability to stratify risk and appropriately manage patients.

Finally, the misconceptions present in the draft LCD regarding treatment pathways prevent an accurate placement of the test in context of patient care. The non-coverage determination does not include adequate review of the NCCN guidelines for Squamous Cell Skin Cancer with and does not detail the multiple management pathways outlined or the considerable latitude given to the treating physician to select the most appropriate treatment course based on the whole assessment of the patient. In addition, MolDX should refer to the recently finalized LCD for bladder cancer to see what appears to be a similar clinical scenario, but for which the test was granted coverage, even in absence of specific articulated utility for a single management modality as is requested for coverage of this test.

In MolDX: Prognostic and Predictive Molecular Classifiers for Bladder Cancer, it is stated that “For patients with bladder cancer, an array of treatment possibilities exists at all stages of disease. Clinicians must consider not only the potential treatment options but must also make an individualized risk to benefit assessment to determine how to treat a specific patient. Diagnostic tests that aid in this assessment are expected to change physician management in a way that improves patient outcomes.”

This appropriately values the clinician’s role in assembling a myriad of information regarding a particular patient, their individual risk factors, and assessing the best treatment plan, without data supporting a clear, specific pathway. Yet, this is the standard applied in the MolDX: Molecular Biomarker Testing for Risk Stratification of Cutaneous Squamous Cell Carcinoma.

We have been using DecisionDx-SCC since November 2020, soon after it was made commercially available and feel strongly that it should be reimbursed for appropriate, high-risk SCC patients. It is important for Medicare beneficiaries that this test remains available. A more thorough evaluation of the validation study by Ibrahim, et al, of the clinical utility studies by Litchman et al, Hooper, et al, Teplitz, et al, and Saleeby et al, all cited in the current non-coverage determination, is needed and will cover the points above. In addition, review of the coverage criteria established for tests with nearly identical intended use in different disease states should be used to ensure that MolDX is issuing consistent policy over time that does not inappropriately disadvantage specific disease states.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

I write this letter in support for coverage of Molecular Biomarker Testing for Risk Stratification of Cutaneous Squamous Cell Carcinoma. This test provides the clinician valuable information to better care for their cancer patients. The rising mortality rate for cutaneous SCC and subsequent increasing cost to care for these patients is directly related to delayed diagnosis of metastatic disease in high risk SCC cases.(1,2). Identifying these high risk patients and providing timely evaluation and management is paramount to both saving lives and preventing costly late stage therapies that often provide “too little, too late” for these patients. Personally, I saw a direct benefit to my own recent patient from information obtained through this testing. Given my patient’s higher risk SCC (larger than 2cm and facial location) I ordered the gene expression profile (GEP) test to gain more information regarding the tumor’s biologic behavior. I performed Mohs micrographic surgery on the tumor and obtained clear margins. Given the high risk stratification found on the GEP testing, I decided to send the tissue block for further evaluation of potential nerve involvement on permanent sections. Despite not seeing any further tumor at margins or perineural involvement in my frozen sections, the further testing revealed significant nerve involvement requiring further radiation therapy to prevent what would have likely progressed to metastatic disease. Frankly, without the knowledge of the high risk GEP my patient was found to have utilizing this test, I would have likely not pursued further than my “clear margins” and missed the noncontiguous neural involvement. My patient would have likely had a poor outcome and become one of the growing number of metastatic SCC cases we are seeing today. Giving clinicians another tool to prevent such outcomes will allow us to better serve our patients with SCC and prevent the growing demand advanced/metastatic SCC has imparted on our health care system. I hope you find this testing worthy of coverage so we may better serve our patients.

References were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

I am writing to provide a perspective to the statement made questioning the superiority of DecisionDx-SCC to current available staging tools and clinical and pathological (CP) factors in the intended use population. I strongly believe that the test adds significant information and provides valuable insights in the management of cutaneous squamous cell carcinoma (cSCC) that has been clearly demonstrated in combination with current CP risk assessment.

The clinical evidence supporting the test's utility in improving risk stratification and guiding treatment decisions for patients with cSCC demonstrates that the test offers additional information beyond the current staging tools and CP factors commonly used in clinical practice. The test provides a more nuanced understanding of individual patient risk, allowing for personalized and targeted management strategies. The clinical studies, highlight the test's ability to accurately identify high-risk patients who may benefit from more intensive management approaches, such as adjuvant radiation therapy, sentinel lymph node biopsy, and increased surveillance. (Arron et al. JDD 2021)

It is important to note that the test does not aim to replace existing staging tools or CP factors but rather complements them by offering additional insights that may not be captured by these conventional methods alone. Of note, the expert consensus paper published (Arron et al.) provides recommendations for patient management based on DecisionDx-SCC test results that are clear and given in the context of other risk assessment systems. Moreover, as acknowledged, this article is clear that the test should be used as an additional data point when determining individualized management plans. In the presence of data answering these queries, any specific analyses needed by MolDx to inform how test results are incorporated with clinical and pathological factors should be articulated instead of vague statements suggesting insufficiency with no guidance as is included in the draft LCD.

I kindly request that you reconsider the statement questioning the test's superiority to current available staging tools and CP factors. The robust clinical evidence supports the significant information that the test provides, enhancing the management and treatment decisions for patients with cSCC.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

More accurate prediction of metastatic risk is needed in high-risk cutaneous squamous cell carcinoma (cSCC), as this disease can be aggressive with subsequent metastasis and potentially disease-specific death, and current risk assessment systems are insufficient to identify which patients are at highest risk. A substantial number of cases categorized as low risk by current staging systems actually progress to metastatic disease, and only approximately one-third of those predicted as highest risk by staging systems have disease progression and poor outcomes.^1–3 Current formal staging systems are not uniformly used, and do not incorporate several high-risk factors that are generally agreed upon by healthcare providers to be impactful (e.g., tumor located on the head, neck, hands, feet, pretibial, anogenital, immunosuppression).^2,4

Formal staging and specified treatment pathways are not included in national guidelines for these reasons. Rather, NCCN guidelines categorize patients as low risk (no risk factors), high risk or very high risk based presence of specific factors, and provide a rather broad range of management recommendations with little distinction in management between high risk and very high risk disease.⁴ Thus, providers have a variety of ways to characterize risk and there is no consensus on a single approach, and no prescribed treatment pathways; a wide net has been cast for indications of which patients qualify for consideration of more intensive therapy. Castle’s 40-gene expression profile test (DecisionDx-SCC) has significant prognostic value for metastatic risk assessment and the test can complement current tumor staging systems or individual risk factor assessment and national guidelines for determining the best management approach for each individual patient.^5,6

DecisionDx-SCC is intended for patients with cSCC and one or more risk factors and the test was clinically validated in this patient population.⁵ Of note, the test is not intended for use in patients with no risk factors. The assertion in the proposed LCD that a Class 1 result confirms a low risk of metastasis is a misinterpretation of the intended use population. All of these patients that are eligible for testing are at an elevated risk and patients with a Class 1 result have reproducibly demonstrated to have half the metastatic risk compared to the overall population, even in subsets of patients predicted to be on the lowest end of the risk spectrum by other risk assessment approaches.^5,7

The test is intended to complement current risk assessment methods to add needed refinement to a patient’s risk of metastasis to guide patient management decisions. It is critically important to stress that, contrary to the suggestions made in the proposed LCD, physicians do not use this test, or any risk factor assessment, in a vacuum. It is the principal tenant of high-quality cancer care that treatment plans should include a comprehensive view of the patient, with all factors weighed and considered. The GEP test only improves the clinician’s ability to do just that, by providing additional risk information that cannot be obtained from other sources. This is demonstrated by multivariate analyses that show independence of the test from other risk factors and staging systems. While this analysis is cited in the LCD, the interpretation of the data is flawed, as there is no question that the test adds statistically independent and clinically significant information to risk assessment for a patient above and beyond clinical and pathological factors alone. These statements in the LCD are not supported by the evidentiary review and should be removed from the LCD.

Because DecisionDx-SCC determines a patient’s risk for metastasis based on the biology of the tumor, the test can be used to better stratify patients and refine management. This test reduces the under-classification of patients who have tumors at high risk for metastasis as well as over-classification of those who are at low metastatic risk. The proposed LCD should be edited to remove opinion-based statements that do not reflect the published evidence or clinical practice in use of this test, and the LCD should be finalized with DecisionDx-SCC as a covered service for Medicare beneficiaries.

References were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

As a practicing radiation oncologist, I am writing to provide my perspective on the DecisionDx-Squamous Cell Carcinoma (DecisionDx-SCC) test and the value of gene expression profile (GEP) testing for patients diagnosed with high-risk cutaneous squamous cell carcinoma (cSCC). I currently utilize this test for high-risk cSCC patients to help assess the likely benefit of adjuvant radiation therapy (ART). The current proposed LCD asserts there is no specific utility for which the test has been demonstrated to guide management which is inconsistent with how the test is currently used clinically. It is important to note that there is no risk-factor guided management framework that defines a treatment pathways for patients, in current clinical practice nor in NCCN guidelines.

Current NCCN guidelines indicate that ART can considered for all patients with NCCN high risk or very high risk disease with poor prognostic features, and patients with PNI or deep invasion or those staged as BWH T2b/T3 or AJCC T3/T4 are routinely referred to radiation oncology for ART based on an elevated risk for poor outcomes. However, these risk factors alone are inadequate predictors of metastatic risk. Importantly, studies show that the DecisionDx-SCC test Class 2A or Class 2B result predicts a risk of metastasis consistent with higher staged disease, even in the absence of these specific clinical and pathological risk factors, and in patients with lower T staged disease (Ibrahim). Published management algorithms and expert consensus consistently supports that patients with Class 2A and Class 2B disease can benefit from the use of ART (Singh, Farberg, Arron). On the other hand, patients with Class 1 results, even in tumors on the higher end of the clinical and pathological risk assessment, can be considered for de-escalation. It is noteworthy that not all patients with high risk disease by clinical and pathological risk factors are treated with ART today (NCCN guidelines); therefore, concern about inappropriate de-escalation of these patients is unfounded and not reflective of clinical practice. Moreover, in patients with higher staged disease, Class 1 results have demonstrated a risk of metastasis on par with lower staged disease, which have low benefit from ART and are not routinely considered for ART (Ibrahim). Therefore, published evidence supports de-escalation of these patients. The draft LCD should be updated to include discussion of use of DecisionDx-SCC results to guide ART decisions, cite the benefit afforded to these patients receiving ART, and removed suppositions of patient harm that are not supported by evidence.

The use of GEP results in the context of risk factors is consistently suggested across multiple publications (Farberg, Singh, Arron). All publications contain nearly identical recommendations for incorporation of test results with risk factor assessment to drive ART decision-making. That is, ART is recommended for patients with high T-stage tumors or a Class 2B result, considered for patients with high T-stage tumors and/or a Class 2A result, and can be avoided in patients with Class 1 result, pending evaluation of other clinical, pathological, and patient-specific factors. Again, the LCD summary of clinical utility of DecisionDx-SCC should be updated to reflect these consistent recommendations for ART use across multiple publications and remove statements suggesting inconsistency of use to guide management decisions.

Utilizing the DecisionDx-SCC result along with other high risk clinical pathologic factors provides an additional level of objective risk information useful in making ART decisions. Combining DecisionDx-SCC with other risk information is useful to clinicians, as evidenced by numerous clinical utility publications showing management changes with the clinical use algorithms referenced above. This test is critically important in sparing some patients from ART, while providing treatment to patients with the potential for greatest benefit. I urge the contractors to finalize the LCD with coverage for DecisionDx-SCC.

References were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

I am writing to inform the MolDX program of an important manuscript that has been accepted for publication in the Journal for Drugs in Dermatology as of April 2023 that will be published in October 2023 that details a consensus panel of 8 dermatologists who evaluated evidence supporting the 40-GEP test for prognostication of metastasis in patients with cSCC with one or more risk features. This panel performed a systematic literature review, evaluated the level of evidence supporting the test by SORT criteria and used a modified-Delphi approach to evaluate 7 specific statements. The literature review found 21 relevant articles, with 2 articles evaluated as Level 1 evidence, 8 articles as Level 2 evidence, and 11 articles as Level 3 evidence. The seven statements evaluated received consensus (8/8) recommendations, with 6 statements receiving SORT Level A and 1 statement receiving SORT Level C.

The statements are listed below:

Statement 1: There is data to support that specific genes influence cSCC clinical behavior. (SORT Level A)

Statement 2: The data supports the 40-GEP test’s ability to identify a subset of cSCCs that are at a high risk for metastasis. (SORT Level A)

Statement 3: The 40-GEP test provides clinically useful data for cSCC prognosis independent of the AJCC8 and BWH staging systems. (SORT Level A)

Statement 4: Adding 40-GEP data to the AJCC8 and BWH staging systems enhances the prognostic assessment of cSCC. (SORT Level A)

Statement 5: The 40-GEP test results can increase the precision and confidence in cSCC management decisions. (SORT Level A)

Statement 6: The 40-GEP test should be considered for use on cSCC tumors with at least 1 high risk feature per AJCC8 and/or BWH and/or NCCN guidelines. (SORT Level A)

Statement 7: The 40-GEP test is not recommended to be used on cSCC in situ or invasive cSCC without high-risk features, or for patients that are not candidates for additional procedures or therapies. (SORT Level C)

Further, the conclusion of this paper states, “The 40-GEP provides accurate and prognostic information beyond standard staging systems that only incorporate clinicopathological data. Incorporation of genomic testing into national guidelines can help further stratify patients based on risk of metastasis and thus improve morbidity and mortality.” This conclusion, based on SORT criteria and modified-Delphi methodology, contradicts the findings of MolDX: Molecular Biomarker Testing for Risk Stratification of Cutaneous Squamous Cell Carcinoma. This statement should be reversed, with data presented in this letter and publication included in the final LCD, and the conclusion revised to acknowledge and be consistent with the conclusion of this important article.

Thank you for your consideration in this important matter, impactful to Medicare beneficiaries.

Reference was provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

Gene expression profiling is becoming standard of care for many cancers, and I am writing to improve the accuracy of metastatic risk prediction for patients with high-risk cSCC.

The need for such a test for high-risk cSCC is critical, due to the various definitions of the "riskiness” for local recurrence of high-risk cSCC and/or progression of cSCC to regional or distant metastasis. cSCC is a historically understudied cancer, with clinical guidelines based on historical practice and/or medium-quality evidence. Given the lack of consensus on which risk factor(s) to use and to what degree the factor(s) indicate poor outcomes, the 40-GEP test has emerged as a well-validated, objective molecular tool that improves the accuracy of risk prediction in combination with existing clinical-pathologic risk assessment systems. In a disease state with so many complicating variables, it is critical that a molecular tool like 40-GEP show added predictive value to the risk factors commonly used in risk assessment, and each staging system; the 40-GEP has strong data demonstrating this feature.

The current policy that has been proposed by the MolDX program has some critical missteps in understanding the statistical analyses that underpin the conclusions regarding risk assessment value of the test, and that impair the ability of the policy to correctly place the 40-GEP test in the current patient journey.

First, this policy misinterprets multivariate analyses. When properly conducted, a multivariate analysis demonstrates the independent value each factor adds to risk prediction, when considering the weight and power of all other factors. Thus, multivariate analyses provide instruction on how to combine risk information to comprehensively evalaute a patient and make management decisions. If a risk factor demonstrates a hazard ratio greater than 1 with a significant p value, the test is adding information, equal in weight to the hazard ratio, that is unique information provided by that factor. In other words, including a new risk factor with statistical significance in a multivariate analysis improves our risk assessment. In this light, the multivariate analysis presented in validation manuscripts represent evidence that the test adds information to our traditional clinical-pathologic risk factors and staging, and does in fact demonstrate, “definitive value above the combination of available clinical, pathological, and staging criteria in accurate risk stratification”, in contrast to the conclusion drawn by the policy. (Wysong, JAAD 2020; Ibrahim, Future Oncology 2022).

The incomplete understanding of multivariate analysis then impacts other sections of the policy, in which comparisons of metastasis rates between different patient subsets attempt to understand the added value of the test when such comparisons are not statistically capable of drawing those conclusions; only a multivariate analysis or statistical comparison of metastatic rate WITHIN a subpopulation (not between subpopulations) are appropriate analytical tools to draw such conclusions. Because the test has already demonstrated independence by multivariate analysis, the test result adds information to other risk information in the context of NCCN risk groupings, BWH stages, AJCC stages, and within individual risk factor assessment. At a practical level, this means that the fact that a Class 2B occurs in a patient with multiple risk factors does not make the result useless or confirmatory of risk factors; it adds risk information to those factors and indicates an even higher level of risk. This is clearly demonstrated in published literature. In patients with BWH T2b or AJCC T3 staged disease, a Class 2B result indicates a substantially higher metastatic risk than in the BWH T2b or AJCC T3 subset alone (Ibrahim Future Oncology 2022). This comparison is more the appropriate evaluations to consider to determine value added from test results to clinical and pathological information and are yet completely absent from evidentiary review and discussion.

Similarly, the statement in the LCD that, “the majority of Class 1 results were identified in samples with 1-2 risk factors, confirmative of low-risk of metastasis” is false as all patients included in the study are at elevated risk of metastasis, and a Class 1 result demonstrates a statistically significant reduction in metastatic events of approximately 50%, consistent across all subset analyses. In addition, patients with 1-2 risk factors, depending on what those factors are, could be staged as high as a BWH T2b, again highlighting the importance of appropriate evaluation of test results across all risk assessment methods. The instances in which Class call results are taken out of context of stage, despite the information being available, are too numerous to count or to comprehensively cite here, and as such, the evidentiary evaluation of this test should be completely revised to correctly reflect the test performance.

I hope to be able to also witness its acceptance for Medicare coverage, as this will allow for the test to benefit such a greater populace.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

I am a dermatologist. I manage patients with skin cancer on a daily basis and have found the DecisionDx SCC test very helpful. I am providing commentary on some of my concerns with the MolDX policy regarding DecisionDx SCC.

A flaw present in the MolDX policy open for comment regarding prognostic testing for patients with SCC with risk factors is the dismissal of the value of a Castle DecisionDxClass 2B result as “rare” and therefore lacking value, which is the opposite of clinical impression. The true problem in this disease state is a complete lack of current risk stratifying methods to identify the relatively low proportion of patients who will experience poor outcomes - we only have low PPV tools available to identify the 2-6% of patients with poor outcomes. The clinical reality is that clinicians are considering or recommending many unnecessary treatments for more patients that don’t need additional aggressive therapy because of inability to predict who does need and will benefit from additional therapy. In this setting, physicians are more likely to err on the side of overtreatment in the face of uncertainty. In order to get an actionable PPV, the test result must, by necessity, be rare, because it accurately identifies patients with the highest risk of metastasis that do warrant all aggressive treatment modalities available.

What’s more, the LCD cites that the Class 2B result is more common in higher stage tumors as a dismissal of its value in these tumors that are “already high risk”, completely missing the added 30-40% PPV that a Class 2B result represents in the highest risk tumor, which is unquestionably clinically actionable. Moreover, it is common and absolutely expected with two well-validated risk prediction systems that there be correlation in “high risk” results. If there is no correlation, then one system is completely wrong in their identification of metastatic risk. This is why the independence of the multivariate analysis is crucial – the test identifies risk that cannot be predicted by staging and risk factors, therefore, the correlation does not and cannot represent a complete overlap in identification of patients.

Finally, after acknowledging that a Class 2B result is often in higher stage tumors, it is disingenuous to characterize the result as rare and, therefore, useless in determining clinical utility at a population level. According to published data, ~10% of patients with a BWH T2b tumors generates a Class 2B result and given the drastic improvement in clinical actionability of a more than 30% increase in PPV, this is a clinically meaningful, and clearly interpretable result.

The LCD demonstrates a lack of thoroughness in the analysis of evidence and no attempt to evaluate and understand the published data showing impact of GEP results in the context of staging nor the appropriate interpretation of added clinical value in each context, despite the primary critique that it is unclear how GEP adds value to staging. These data should be updated to consider the performance of the test in each clinical subset evaluated and compare to the current standard of care for interpretation of the value of the results. It is disingenuous to critique the test for failing to demonstrate these attributes when the data was available in published literature, but not analyzed and evaluated.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

I am submitting these written comments of the proposed LCD DL39583 – Molecular Biomarker Testing for Risk Stratification of Cutaneous Squamous Cell Carcinoma. I disagree with the conclusion that this test has not met Medicare criteria for medical reasonableness and necessity.

Further, I use the DecisionDx-SCC test for my patients when making decisions about adjuvant radiation therapy (ART) in a model of shared decision making. In order to provide high quality cancer care to patients based on retrospective cohorts, genomic data doesn’t replace staging but rather it augments current risk stratification approaches to inform ART decisions.

For this disease state – high risk cutaneous squamous cell carcinoma (cSCC) - the level of evidence supporting our clinical decision making each and every day in the clinic is not at the same stage in evolution as the shape of the data for other cancers such as breast or colon cancer. I mean this as a way to provide context for the field overall – not related to molecular testing – but the shape of data supporting decisions around surgery, immunotherapy, radiation therapy and even staging systems. For example, we are just beginning, in our 50 years of research in cSCC, to start thinking about starting randomized trials in this space; it's never been done before.

High risk cSCC is not tracked in the Surveillance, Epidemiology, and End Results (SEER) cancer registry or National Cancer Database (NCDB) and high-quality evidence lags behind other more common disease states. The draft LCD also mischaracterizes the level of evidence in support of existing clinicopathologic risk stratification approaches (NCCN/BWH/AJCCv8). Current staging systems and patient management decisions are all based on retrospective data and the level of evidence supporting the 40-GEP test falls in line with all other data we use to make our current clinical decisions. While various clinical and pathological factors are used to stratify risk, we can now augment that process by adding genomic data with the 40-GEP. Physicians treating these patients recognize the data limitations inherent to this field and make evidence-based choices for our patients based on the highest quality data available. Patients want physicians to utilize the best data available – especially in the field of skin cancer where current management strategies are not based on “perfect” evidence.

I read the recent draft LCD and was struck by several aspects of the review of evidence and conclusions; I was shocked that the draft LCD ignores the established role and efficacy of ART in treatment of cSCC. Several recently published studies support the efficacy of ART in cSCC, and I think that they warrant review and inclusion in the discussion prior to the finalization of the LCD. While it is true there are currently no prospective studies demonstrating improvements in patient outcomes, ART is included in national guidelines based on retrospective data demonstrating improvement in outcomes, and it would be unethical and, quite honestly, a liability to withhold consideration of ART from appropriate patients with high risk cSCC on that basis alone.

ART triage paradigms that rely on clinicopathologic factors alone can be improved by incorporating additional genomic data such as DecisionDx-SCC; it is likely that we are overtreating cSCC patients with ART and missing some patients that could benefit due to low PPV of staging. The two papers, both published in JAAD in 2022, contain the most up to date data on the utility of ART in cSCC and are important to consider when framing the state of evidence around ART.^1,2 These studies show that use of ART in patients with very high risk cSCC can reduce the risk of disease progression by 50%, which is consistent with the benefit of ART in breast, rectal, cervical, lung and many other epithelial-derived tumor.¹

Ruiz, et al in 2022 retrospectively evaluating over 500 patients from two large academic practices with very high-quality data. The study results showed that the risk of locoregional recurrence for patients who were treated surgically and achieved negative margins was 15.3% without adjuvant radiation and 7.5% with radiation for an overall relative risk reduction of roughly 50%, which is exactly the expected result based on studies of ART in other epithelial tumors. The greatest total benefit was seen in patients with the highest metastatic risk where the metastatic risk was reduced from 31.0% to 17.2% by ART (see Figure 1 below).¹ This study is now cited in the NCCN guidelines for cSCC (v1.2023) and represents the highest quality evidence to date that the greatest benefit from ART was observed in patients with the highest risk of disease progression.

Another study focusing on BWH T2a tumors which are another substantial gray area for treatment recommendations. To level set, approximately 30% of poor outcomes that occur in Brigham and Women’s Hospital (BWH) involve T2a cutaneous squamous cell carcinomas. What this means is that patients with high stage disease are not the only patients at risk for developing poor outcomes and improved risk stratification can inform risk aligned decision making for BWH T2a patients.²

Clinicopathologic factors crossed with genomic test results can help identify patients in this cohort who benefit from ART as well.

All of the major guidelines including the societies of the American College of Radiology, American Academy of Dermatology, ASTRO, endorse the use of ART in certain categories of high risk cutaneous, and ART for these patients is reimbursed by Medicare already. As a physician who has a career focused on trying to enhance risk stratification, I think we overtreat many patients based on inadequate risk stratification. This is not just an opinion; this is informed by assessment of the positive predictive value of staging systems and individual risk factors. For example, patients with BWH T2b tumors have a risk of metastasis around 30%, and based on multiple consensus statements and publications, should be referred for consideration of ART.³ Without any additional risk information, I consider ART for all of these patients when referred to me. However the use of ART in all patients with BWH T2b tumors results in 60-75% of patients being overtreated.⁴ There is significant morbidity and cost associated with ART, and this presents complex clinical discussions about the risk/benefit of ART, particularly for older, Medicare-aged patients who may not have easy access to treatment centers.

On the other hand, many patients are not referred for consideration of ART until the disease has progressed to a point where intervention is too late to be effective. The delay in these referrals is often based on an initial clinical and pathological risk-assessment and reflects the inability of these factors alone to identify the adverse biology of the tumor. Moreover, there are some patients where radiation is not enough. These refractory cases may actually benefit from immunotherapy or may need to be included in currently ongoing clinical trials. The unfortunate reality is that it is unclear how to identify these patients using clinical and pathological risk factors alone. There remains a clear need to identify patients at lower BWH stages with a higher biological risk of disease progression who will therefore also benefit from ART, and identify patients at higher BWH stages with a lower biological risk who can safely forgo ART.

Genomic classifiers are not novel anymore; over the last 5 to 10 years in prostate cancer and breast cancer, they've become indispensable to making the most informed decisions for therapy. This scenario of over and undertreatment in cSCC is exactly the kind of scenario where genomic classifiers are best suited. In fact, DecisionDx-SCC is filling that role, and due to its clinical availability for over 3 years and widespread use, is no longer novel and is regularly used in clinical practice.

After surgery of the primary tumor and initial assessment of risk factors, patients are often referred to radiation oncology, where we historically put them into three buckets for management decisions prior to the availability of a genomic classifier. Patients with BWH T3 we believe are high enough risk to consistently recommend and offer ART, but we do not currently offer adjuvant systemic immunotherapy off of a clinical trial. For BWH T2b patients, we estimate the risk of recurrence is between 10-20% and have a personalized discussion and shared decision making on the best approach for each patient. For BWH T1-T2a patients, we typically observe them because their risk of recurrence – at a population level- tends to be less than 10%. (Figure 2)

Now that DecisionDx-SCC is available, we have biomarker informed ART decision making that is happening in the clinic. Specifically, we have a data-informed discussion with the patient regarding DecisionDx-SCC test result and evaluate the clinical, pathologic and genomic information together with patients and their family members. Unlike the critiques raised in the proposed LCD, never has the fact that the test was not validated with direct outcomes or outcomes validated in a randomized Phase III study of a treatment that we use in practice today been an issue.

This may be due to the fact that formally none of what I have to offer these patients has been validated in that manner for cSCC. Overwhelmingly, patients want this test and the additional information it provides, and then we incorporate the test results with BWH staging to make risk aligned treatment recommendations in this model of shared decision making.

For BWH T3 patients, we are still typically treating them, regardless of DecisionDx-SCC test results, which is in alignment with the proposed intended use population proposed by Castle. However, if I have BWH T2b patients whose tumor returns a Castle Class 2A, we are typically offering them ART and are more likely to offer elective nodal radiation. A BWH T1/T2a, with Castle 2B are being referred for consideration of ART based solely on their genomic classification. Patients with BWH T1/T2a and Castle Class 2A or BWH T2b with Castle Class 1 result require a more nuanced discussion, but again, this conversation is powered by clinical, pathological and now genomic information. In these patients, we have an open and honest conversation with more accurate risk information on the pros and cons for ART. In this setting, many patients are electing to not get treated, which is perfectly reasonable based on the current data. On the other hand, we have patients who are characterized as “low risk” by clinical and pathologic factor-based risk stratification, but due to a Class 2A result now have sufficient concern about an 15-18% risk of regional/distant metastasis and for these patients ART and comprehensive nodal radiation are warranted for discussion.⁵

Based on 420 patients in a large study cohort from Ibrahim et al. 2021, patients know what their Castle score means for the risk of metastasis in the clinical context of their BWH stage and are engaging with clinicians during shared decision making. I feel ethically obligated to tell them about the data related to each risk stratification method (Decision-Dx-SCC, BWH, etc.), even though the validation of each is based on retrospective cohorts. I feel ethically obligated to take them through it, and I have yet to have a patient not feel compelled to get the test and additional genomic information. The level of evidence that supports incorporating DecisionDx-SCC into the shared decision making is as good as, if not better than, many of the other risk features that are incorporated into staging or NCCN risk stratification and go into decision making. Whether it’s tumor diameter, differentiation, perineural invasion, nerve size (which I don’t think is very well evidenced) invasion beyond the fat, or whether it is NCCN high risk or very high risk designation it is clear that the 40-GEP is highly validated and on par with these other factors that the proposed LCD does not appear to critique in the slightest.

In conclusion, cutting edge medicine for cSCC should mirror what has been shown and adopted in NCCN, and by Medicare payers in prostate cancer and breast cancer. What this means is that by combining clinical, pathologic, and genomic risk indicators treating clinicians can enhance our ability to appropriately categorize patients, better estimate risk, and to make adjuvant therapy decisions for management decisions such as ART in that context. Combining BWH staging and NCCN risk classification with DecisionDx-SCC allows our best opportunity to make the best physician and patient shared decision making in this space.

It is critically important to ask a simple question when considering my comments related to the proposed LCD and the DecisionDx-SCC test: Why combine clinical and genomic risk stratification? It's because DecisionDx-SCC has a comparable, if not superior level of evidence to anything else for cSCC risk stratification. Therefore, I feel that DecisionDx-SCC has met the requirements for medical reasonable and necessity as it enables me to make much more informed, nuanced decisions with my patients about the use of adjuvant radiation therapy. Currently, I have 3-4 patients per week consenting to receive the test and use the results to guide management in my practice alone. It is a high enough level of evidence that in the complex, murky disease state of cSCC; it is as strongly evidenced as anything else I use in the clinic.

In this letter, I have a detailed how I approach the data-driven discussions with my patients as we determine their appropriate risk aligned treatment plans. DecisionDx-SCC has a comparable level of evidence to the clinicopathologic factors that we use to inform ART decision making in clinic today. The DecisionDx-SCC test shares similar level of evidence to existing staging systems, the validation cohorts are well balanced for the population of patients I see with cSCC and include data from more centers than were involved with staging systems.^5–9 The 40-GEP metastatic risk information is independent of other staging systems and therefore provides additional risk stratification in the context of NCCN, BWH, AJCC, individual risk factors. A combined clinical-genomic risk stratification scheme enhances our ability to appropriately categorize risk and offer adjuvant therapy with enhanced selection. To me this is an essential aspect of providing high quality cancer care – my patients want to know their Castle 40-GEP score during shared decision making and I am obligated to provide patient with information from best data available. Combining BWH staging WITH DecisionDx- SCC score allows this hybrid risk stratification and more nuanced physician/patient consent discussions and shared decision making.

Decision DX has met the requirements for medical reasonableness and necessity, enables an evidence- based decision regarding use or avoidance of ART and should be available and covered for Medicare beneficiaries. On behalf of myself and my colleagues, the DecisionDx-SCC test is very helpful to us in making decisions and our patients want to know their most precise risk information, and we are using it to make decisions.

Figures and references were provided for review.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.

The following comment was submitted to Palmetto GBA:

We are writing on behalf of the Skin Cancer Prevention Working Group (SCPWG), a dedicated collaboration of experts focused on the prevention of skin cancer. Our group consists of clinical and research specialists who have spent years investigating and comprehending the diagnosis and management of melanoma and non-melanoma skin cancer. The mission of the SCPWG is to foster and analyze evidence-based research, enabling us to gain a deeper understanding of skin cancer pathophysiology, treatment, and prevention. We strive to become leaders in skin health education by utilizing this knowledge effectively.

SCPWG recently aligned on an expert consensus regarding the utilization of DecisionDx-SCC for patients diagnosed with cutaneous squamous cell carcinoma (cSCC). This consensus is in preparation for publication.

It is our unanimous agreement that, similar to all disease states, treatment decisions should align with the patient's risk of disease progression. Although current guidelines (NCCN) and staging systems (BWH, AJCCv8) rely on historical, clinicopathologic, and phenotypic risk factors to determine cSCC prognosis, we unanimously acknowledge that these clinical or pathologic- based risk systems do not optimize prognostic assessment for directing risk-aligned treatment decisions.

All clinicians within the SCPWG concur that literature demonstrates the DecisionDx-SCC result is a significant risk factor in identifying cSCC patients at greater risk than the overall low-risk population for regional and/or distant metastasis. In fact, compared to traditional risk factors recognized by current guidelines, the DecisionDx-SCC result ranks as significant as perineural involvement (PNI) and surpasses all other factors in terms of degree of risk identified by high risk results.

Furthermore, we unanimously agree that the DecisionDx-SCC test is clinically validated, independent of other risk factors, reproducible, and consistent across studies. Integrating the DecisionDx-SCC Class result into NCCN/BWH/AJCCv8 guidelines and staging can enhance prognostic accuracy by providing personalized risk assessment to inform risk-aligned treatment decisions.

Based on the clinical validation and performance characteristics of the test, we unanimously assert that DecisionDx-SCC is appropriate for use, in conjunction with clinical and pathologic factors, to guide clinical follow-up frequency, surveillance modality and frequency, nodal surveillance intensity, and clinical management/treatment decisions including adjuvant radiation therapy.

In addition, all clinicians agree that the DecisionDx-SCC Class 2 results provide a common platform for multi-disciplinary tumor board discussions. The DecisionDx-SCC test can inform clinical decision making, thereby guiding risk-aligned treatment plan decisions and positively impacting patient outcomes.

The consensus is in direct contrast to conclusions in the LCD that the test has not demonstrated definitive value above the combination of available clinical, pathological, and staging criteria in accurate risk stratification. We urge you to consider our findings and change your non-coverage determination so your Medicare patients can maintain access to a crucial prognostic tool utilized by their clinicians to improve their care.

We remain committed to advancing skin cancer prevention and improving patient outcomes, and positive coverage of this tool is critical to this mission.

Thank you for your thoughtful comment. These concerns have been addressed in Response to Comment 1.