Response to Comments: MolDX: Melanoma Risk Stratification Molecular Testing

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

In response to the request for comments on MolDX: Melanoma Risk Stratification Molecular Testing, we wish to present the updated evidence supporting DecisionDx-Melanoma as a powerful predictor of metastatic risk in patients with cutaneous melanoma that influences sentinel lymph node biopsy (SLNB) decision making, frequency of follow-up, use of surveillance imaging and referrals.

We also wish to provide suggestions on the drafted language added to the draft LCD as the most recent updates to this LCD include two articles with methodologic limitations and potential bias while overlooking eleven studies, including two prospective, independent studies and two systematic reviews and meta-analyses demonstrating the consistent, accurate prediction of metastatic risk by DecisionDx-Melanoma.^1–11 We believe that the LCD coverage policy will benefit from improved contextualization of the published literature that has evaluated the clinical validity and clinical utility of DecisionDx-Melanoma (31-GEP) testing in melanoma, and that the strength of this evidence overwhelmingly supports continued coverage for DecisionDx-Melanoma testing for the benefit of patient care. While we are submitting suggested changes to the draft LCD that is currently posted in a separate attachment to this letter. We do not have proposed changes to the Article Billing and Coding: MolDX: Melanoma Risk Stratification Molecular Testing that pertains to DecisionDx-Melanoma specifically.

DecisionDx-Melanoma is a non-invasive prognostic test that accurately and independently predicts the risk of metastasis to the sentinel lymph node as well as regional and distant metastasis within five years as low risk (Class 1A lowest risk, Class 1B lower risk) and high risk (Class 2A higher risk, Class 2B highest risk).^2,12 With the release of two publications in April 2021, there are now 30 peer-reviewed publications demonstrating consistent clinical and analytical validity and clinical utility. All melanoma management decisions are risk-based in that guideline recommendations for sentinel lymph node procedures, follow-up and active surveillance decisions, and referral for specialty care are based upon an individual patient’s risk of metastasis. Thus, the utility of DecisionDx-Melanoma is aligned to improve or influence decision making of these risk-based patient management decisions within the framework of current guidelines and it is these uses that are covered under the LCD noted above.

DecisionDx-Melanoma meets medical necessity and reasonableness standards and continues to have meaningful, positive impact on patient care. In 2020, the test was ordered by 4,300 U.S. clinicians to guide patient management decisions for 16,730 patients (total tests ordered since clinical availability = 78,225). Importantly, 288 commercial insurance companies approved and paid for DecisionDx-Melanoma in 2020. Continued LCD coverage for DecisionDx-Melanoma is clinically impactful as the median age of patients diagnosed with cutaneous melanoma is 65, and since the original LCD effective date of December 3, 2018, more than 4,700 clinicians have ordered the test for 21,120 patients ages 65 years and older to direct management (clinical data on file).¹³ Clinical research data published on over 5,700 patients have consistently demonstrated the prognostic accuracy and added value of DecisionDx-Melanoma test results in prognostication of metastatic risk beyond clinical and pathological features alone. Since the positive decision to expand coverage of DecisionDx-Melanoma in LCD, additional evidence continues to demonstrate the robust accuracy and independent prognostic value of the test and the use of test results by clinicians in the management plans of patients with cutaneous melanoma.^1–11

The data supporting the clinical validity and clinical utility of the test is outlined in the following pages to provide a counterpoint to the proposed additions to draft LCD. This data further strengthens the evidentiary foundation for continuing the LCD coverage determination that offers clinicians access to DecisionDx-Melanoma that has been shown to improve the accuracy of clinical management decisions for patients with Stage I-III melanoma including SLNB decision making, frequency of follow-up, use of surveillance imaging and referrals.

Clinical Validity: DecisionDx-Melanoma has achieved Level 1A Evidence

Two systematic reviews and meta-analyses that utilized the appropriate Cox regression analyses generating hazard ratios or odds ratios were published in mid-2020. Together, these studies confirm the clinical validity of the test, demonstrating that DecisionDx-Melanoma provides prognostic information independent from staging and SLNB.^3,4 Both meta-analyses demonstrate consistent prediction of increased metastatic risk by a high-risk DecisionDx-Melanoma test result.

Importantly, consistency of results demonstrated by a meta-analysis constitutes Level 1A evidence for a prognostic biomarker under the Strength of Recommendation Taxonomy (SORT) used by the American Academy of Dermatology and other groups.¹⁴

Meta-analysis #1 summary (Greenhaw, et. al. JAAD 2020)³:

DecisionDx-Melanoma was shown to be a significant, independent predictor of recurrence-free survival (p<0.0001, HR: 2.90) and distant metastasis-free survival (p<0.0001, HR: 2.75) according to multivariate Cox regression analysis. DecisionDx-Melanoma significantly stratified metastatic risk of patients within AJCC stages, demonstrating added prognostic value to clinical factors. The relative risk associated with a Class 2 result was similar to that of a positive sentinel lymph node biopsy result, a prognostic variable accepted as standard of care.

Patients with low stage (I-IIA) primary cutaneous melanoma but with a high-risk Class 2B GEP result from a cohort included in the Greenhaw meta-analysis (Gastman JAAD 2019)² had a 5-year MSS rate (87.4%) comparable to the reported 5-year MSS of patients with Stage IIB tumors (87%) by AJCC. Patients with a Class 1A GEP result had a 5-year MSS rate comparable to that reported by AJCC in patients with Stage IA tumors (99%). Of note, patients with Stage IIB disease have an AJCC-reported 5-year MSS rate of 87% and these patients are eligible for consideration of increased management intensity, including more frequent follow-ups and use of surveillance imaging based on increased risk of metastasis.^2,3

DecisionDx-Melanoma had higher sensitivity and negative predictive value for prediction of recurrence and distant metastasis than that of a positive SLN status and similar specificity (Table 1). DecisionDx-Melanoma combined with SLNB improved sensitivity over either result alone while maintaining a high NPV. As SLNB remains appropriate for a subset of patients, improvement of prognostic accuracy of SLNB with DecisionDx-Melanoma for prediction of metastasis is a second benefit of the DecisionDx-Melanoma test.

The consistency of 31-GEP results across multiple patient cohorts, led to the conclusion that the test “significantly augmented the ability to identify high-risk patients for heightened clinical surveillance.”

Meta-analysis #2 summary (Litchman, et. al. SKIN: The Journal of Cutaneous Medicine, 2020)⁴:

Additional prospective studies have recently been published:

In addition to the published meta-analyses noted above, evidence from large, independent prospective cohorts has also been recently published.^5,6 One study performed independently of Castle Biosciences included 383 melanoma patients from the Oregon Health & Science University with 32 months mean follow-up time. The study findings confirmed previous retrospective and prospective studies’ findings that patients with Class 2 results have significantly reduced recurrence-free and distant metastasis-free survival in patients with stage I-III disease.⁵ Importantly, after adjusting for predicted AJCC staging factors, DecisionDx-Melanoma was significantly associated with survival outcomes and was a significant, independent predictor of recurrence in the overall cohort (stage I-III) as well as the AJCC low-risk Stage I-IIA patients, demonstrating prognostic value over AJCC staging alone. Importantly, in all patients with T1-T4 melanoma, Class 1A and Class 2B results were associated with significantly different rates of SLNB positivity (5% and 28%, respectively) confirming previously published data on the use of DecisionDx-Melanoma to guide SLNB decisions.

A second prospective, multi-center, study was also recently published demonstrating long-term outcomes from a previously published interim analysis.^6,17 Similar to the study noted above the results from this large prospective cohort of 323 melanoma patients (38 months median follow up time) demonstrated that the test is a significant, independent predictor of recurrence, distant metastasis, and death.^6,17 Patients in this cohort with Stage I-IIA disease and Class 2 results had similar outcomes to all patients with Stage IIB-III disease in this cohort. This finding demonstrates that the metastatic risk of a subset of patients with AJCC low-stage disease (Stage I-IIA) who also obtain a Class 2 GEP result is similar to the metastatic risk of patients with high-stage disease (Stage IIB-III) for whom guidelines recommend increased follow-up and surveillance intensity as well as referral to specialty care. This study also demonstrated that the sensitivity of DecisionDx-Melanoma was higher than AJCC staging alone, with similar specificity, negative predictive value, and positive predictive value. Moreover, when DecisionDx-Melanoma test results were incorporated in AJCC staging, the sensitivity for detection of recurrence increased by 19% and sensitivity for distant metastasis increased by 26%. This evidence strongly supports incorporation of DecisionDx-Melanoma test results with AJCC staging for improved prognostic accuracy.

Clinical Utility: DecisionDx-Melanoma significantly influences patient management, including guiding SLNB decisions

It is important to provide meaningful context for evaluating the clinical utility of GEP testing that aims to improve the accuracy of risk stratification as this is key to deciding on optimal risk-appropriate patient management plans. Traditionally, the standard of care for prognostication in cutaneous melanoma relies substantially on the risk stratification provided by AJCCv8 staging information. At the time of diagnosis, prior to SLNB, the primary prognostic factors are those pathologic factors (Breslow thickness and ulceration status) that are used to calculate the melanoma tumor stage (T stage). Current guidelines recommend “discussion and consideration” of SLNB for patients when the population-based likelihood of a positive SLN is >5% (this is estimated to occur in patients with T1a with high-risk features or T1b melanoma), to “discuss and offer” SLNB for patients when the population-based likelihood of a positive SLNB is >10% (this is estimated to occur with T2-T4 melanomas), but to not perform an SLNB if the likelihood of a positive SLNB is <5%.18 Overall, only ~12% of all patients will have a positive SLN; that is, ~88% of patients who undergo the SLNB procedure have a negative SLN and do not benefit from an improvement in risk stratification.¹⁹ Moreover, rates of SLNB positivity are known to be inversely related to age; that is, older patients with the same Breslow thickness and ulceration status have a lower likelihood of a positive SLN than a younger patient^.20,21 This point is particularly important for the Medicare population. As such, there is a clinical need to identify patients with a low risk of sentinel node metastasis who could safely forgo SLNB.^{20, 22}

The first clinical utility timepoint for GEP testing occurs before SLNB is performed, and that clinical utility is to guide SLNB decision making for patients with T1-T2 melanoma. Prior studies have determined that DecisionDx-Melanoma Class 1A results do correlate with <5% predicted risk of nodal metastasis and that these patients have extremely good outcomes with >99% melanoma-specific survival (MSS) in the T1-T2, ≥65 year-old population, suggesting these patients could forego SLNB.²² Conversely, the test increases the yield of the SLNB procedure by identifying patients with high-risk tumors (Class 2B), who are more likely to have a positive node. It is important to note that the landmark MSLT-I trial definitively demonstrated that the SLNB procedure itself does not confer a specific survival benefit and that the majority of melanoma deaths in the study cohort were sustained by patients with negative SLNB results; meaning that the SLNB procedure is only a prognostic variable.²³

Traditionally, following completion of the initial melanoma workup, patients are then risk stratified according to their AJCCv8 stage (stage I-IV) which incorporates the Breslow thickness, ulceration status and additional information, such as nodal status, when applicable. After the decision to either obtain or forgo the SLNB has been made, the second clinical utility of DecisionDx-Melanoma is to provide improved accuracy of risk stratification to inform decisions regarding frequency of patient follow-up, surveillance imaging and referrals. As noted above, recent prospective studies demonstrate that patients with AJCCv8 early-stage disease (stage I-IIA) and high-risk Class 2B results have a recurrence risk similar to patients with later stage disease (stage IIB-III) who are currently recommended to receive more intensive follow-up, surveillance imaging, and referrals.^5,6 In this way, DecisionDx-Melanoma testing identifies a level of risk in early-stage patients that is determined to be clinically actionable by national guidelines. Finally, it remains important to direct patients who are at high risk for recurrence and metastasis to enhanced surveillance regimens as new data strengthens the evidence that detection of asymptomatic recurrence through enhanced surveillance imaging can lead to improved patient survival.²⁴

The current LCD references three clinical utility studies demonstrating the use of DecisionDx-Melanoma in informing decisions related to sentinel lymph node biopsy surgical procedure recommendations, follow-up, referral, and use of advanced imaging.^25–27 These clinical utility studies highlight that there are not only multiple clinical decisions that GEP results impact, but that GEP results can substantially inform decisions that are made at different time points both early in the patient journey after diagnosis with invasive cutaneous melanoma and after the initial melanoma workup is complete. Specifically, GEP results can inform SLNB decision making to both avoid unnecessary SLNB procedures in low-risk Class 1A patients and advocate for SLNB procedures in high-risk Class 2B patients most likely to benefit from the SLNB surgical intervention. Following the impact on surgical decisions, the same test result can inform the intensity of follow-up visits, referrals and surveillance for patients after final staging is complete. Additionally, more recent publications confirm and expand on these findings.

A recent study that describes test experience at a single institution cohort, included both clinical validity and clinical utility data that is relevant for the Medicare eligible population.⁸ Importantly, during the course of the study the surgeons adopted the use of the test to guide SLNB decisions in patients ≥65 years of age with T1-T2 melanomas and described avoidance of an SLNB in 50% of patients meeting this criteria. Recent correspondence with the senior surgeon of this surgical oncology group confirmed that their standard workflow, today, is to use DecisionDx-Melanoma to determine which patients ≥65 years of age with T1-T2 melanomas can avoid the SLNB surgical procedure due to a Class 1A result (personal communication). The study also reported that "to date, no cases of locoregional recurrence or distant metastasis have been identified in this cohort” suggesting that the individuals over age 65 who were triaged away from obtaining a SLNB in this cohort sustained excellent 3-year health outcomes. In addition, post-SLNB management decisions were also impacted for patients who received high risk GEP results (Class 2A/2B) as these patients were offered augmented surveillance measures and increased follow-up visit frequency. The article ends with the statement, “We conclude that the 31-GEP has added to the assessment of patients within our practice, and given the available literature suggesting its utility, we are beginning to see changes to our clinical practice.”

During the recent Open Comment meeting (6/28/21), a request was made for quantitative data showing the clinical application of 31-GEP test results for reducing unnecessary SLNB procedures. Following the finalization of the initial LCD (2019) that covers DecisionDx-Melanoma, Castle Biosciences initiated the DECIDE trial to determine the long-term outcomes of SLNB-eligible patients managed by clinicians (dermatologists and surgical oncologists) who are using the test to inform decisions about the SLNB procedure. We performed a data extraction from this ongoing study following the Open Comment meeting and the results showed that 45% of the Medicare eligible patients enrolled in that study who had a Class 1A DecisionDx-Melanoma test result did not have SLNB performed. These results show that the 31-GEP is being utilized to inform decisions about the SLNB, reflecting the awareness of results from published studies.^8,22

In 2020, a panel of expert dermatologists evaluated the full body of published evidence on DecisionDx-Melanoma to provide a clinical management algorithm incorporating test results into patient management decisions^.9 Importantly, the expert panel concluded that, although helpful, current AJCC staging has room for improvement in accuracy of risk prediction, and determined that DecisionDx-Melanoma is useful and actionable for patient care when used in conjunction with AJCC staging and applied to inform clinical management decisions in the context of current NCCN guidelines. The resulting clinical management algorithm provides rationale for GEP testing in patients with AJCC stage I-III melanoma and characterizes patient management decision points informed by integrating GEP results with melanoma staging. The management decisions incorporated a patient-centered approach and included data-based patient reassurance, informing SLNB decision making, adjustments in the frequency of follow-up visits, consideration of baseline imaging, adjusting thresholds for symptomatic imaging, and referrals.

Separately, a 2021 study of 589 dermatologists attending two distinct national virtual conferences demonstrated that the majority of physicians report benefit of DecisionDx-Melanoma in understanding a patient’s true risk of metastasis (72.5% of respondents) and in personalizing treatment options (58.8% of respondents).⁷ Thirty-six percent (36%) of clinicians surveyed indicated that they use DecisionDx-Melanoma as part of the decision to recommend or not recommend a sentinel lymph node biopsy to patients. Escalation of management of patients with Class 2B results was indicated by 59% of respondents when considering patients with stage I disease, and 61% of respondents when considering patients with T1 melanoma (Breslow depth <1.0 mm). Additionally, 60% of respondents acknowledged the value of a favorable Class 1A test result in easing patient uncertainty about their future. In summary, the combined clinical utility evidence continues to demonstrate 1 out of 2 patients have clinical management decisions altered by their DecisionDx-Melanoma test results.^25–28 This evidence further demonstrates that DecisionDx-Melanoma delivers meaningful and actionable results to physicians managing patients with cutaneous melanoma.

The two single center cohort experiences^8,10, expert opinion clinical workflows⁹, and clinical impact survey responses⁷ reviewed above as well as the three additional clinical utility studies that are cited in the current LCD, demonstrate that clinicians are currently using the GEP test to guide management decisions - both before SLNB decisions are made and after SLNB has either been obtained or avoided. Table 4 demonstrates the consistency of management change percentages across previously cited clinical utility studies and three recent studies that specifically address how management decisions to avoid SLNB procedures are impacted by GEP results. These data show that clinicians are using the GEP test to improve or influence decision making of risk-based patient management decisions in a manner that is consistent with the current LCD and accompanying billing article.

Economic considerations for avoiding unnecessary SLNB surgical procedures:

While we recognize that cost savings is not a part of medical necessity and reasonableness determination, it is critical to understand the costs and inequity of access of SLNB, for which DecisionDx-Melanoma provides guidance. While the cost of SLNB varies between regions, published literature from the San Jose, California Medicare region notes the price for SLNB as $19,000.²⁹ This reported price-point aligns with the $18,183 reimbursement calculated by Castle Biosciences in early 2019 (Data on file). Another way to look at the cost of an SLNB procedure is to look at the cost to identify a positive SLN. As 88% of patients who undergo the procedure receive a negative result,19,30,31 the cost to identify a positive sentinel node exceeds $100,000.

Importantly, access to the SLNB procedure is not equal in the U.S. This inequality is listed as one of the major challenges in melanoma care in 2021.³² The DecisionDx-Melanoma test is available for all patients with stage I-III melanoma, without a need for additional travel to surgical centers or access to multi-disciplinary teams. As such, this test provides a critical service to guide decision-making where real management and surgical barriers exist, particularly for older patients and those in rural areas. Thus, DecisionDx-Melanoma improves the allocation of resources to appropriate patients resulting in better patient care and substantial cost savings.

Response to inclusion of new articles in Draft LCD

The current proposed draft LCD has included summaries of two articles critical of GEP testing without inclusion of the above summarized publications that strengthen the body of evidence supporting the use of DecisionDx-Melanoma to improve accuracy of risk prediction for patients with cutaneous melanoma. While we agree with the language in the draft LCD that “none of the data directly contradicts the initial published data demonstrating concordance with SLNB status”, the discussion of these two articles in the draft LCD is unbalanced, presenting these two articles as equal weight to the more than 30 peer-reviewed articles to date that support the use of the test to inform a range of management decisions and without any discussion of their substantial limitations. Importantly, neither of these articles contribute additional GEP tested cases to the published literature.

The article by Grossman, et al. provides opinion statements informed by survey data on test usage from a limited subset of physicians.³³ These authors requested more evidence prior to routine use of the DecisionDx-Melanoma test, yet they failed to cite or thoroughly review many published manuscripts that support the current clinical validity and clinical utility of DecisionDx-Melanoma that address questions raised. The reported survey in this publication had a low response rate (only 28 [14%] respondents completed both surveys within the study analysis), without characterization of survey non-responders. These two points are significant as JAMA Dermatology generally requires a ≥60% response rate and appropriate characterization of non-responders to ensure that nonresponse bias does not threaten the validity of the findings. Other limitations to this article include disproportionate representation from a single academic center, and inclusion of medical students, research scientists, and internationally based participants with limited clinical experience or no clinical experience with DecisionDx-Melanoma, all of which substantially restrict the applicability of the study findings. Without surveying diverse opinions from around the country and without including numerous centers that have clinical familiarity with DecisionDx-Melanoma, the conclusions regarding the utility of GEP testing does not reflect a balanced consensus opinion. Specifically, the manuscript does not reflect the opinions of clinicians who have incorporated DecisionDx-Melanoma testing into their practice for the benefit of patient care, including the 4,300 clinicians who used DecisionDx-Melanoma for clinical management decisions in 2020 (clinical data on file). Furthermore, the Grossman et al. manuscript concludes that, “Decisions on performing GEP testing and patient management based on these results should only be made in the context of discussion of testing limitations with the patient or within a multidisciplinary group.”

For these reasons, instead of characterizing the Grossman et al manuscript in the current Draft LCD as “casting doubt on the utility of gene expression profiles (GEPs) for risk stratification” we believe that this article could be more fairly described as bringing awareness to the ongoing debate in academia around how to incorporate GEP testing in the management of patients with cutaneous melanoma. Finally, reference to the Grossman et al manuscript should highlight the existence of other expert opinion papers and clinical impact studies that arrive at substantially different conclusions about the strength of current evidence supporting the use of DecisionDx-Melanoma to improve the accuracy of risk prediction and inform a range of risk-based patient management decisions.^7,9,10,25-28

The second article from Marchetti, et al. reported analyses characterized as a meta-analysis of GEP tests for melanoma.³⁴ However the bulk of the article focused on DecisionDx-Melanoma, concluding that the test has prognostic value for patients with stage II melanoma, but limited value for those with stage I disease.³⁴ This study was marked by significant limitations because it did not evaluate the test in the context of current clinical factors or use standard analytic methodology common to meta-analysis. For instance, the authors noted that the prognostic outcomes are best analyzed using hazard ratios or odds ratio in meta-analyses; however, this standard analysis was not performed. We analyzed the same source articles that were cited in Marchetti et al and were able to calculate the positive likelihood ratio from the same published data. We found that the DecisionDx-Melanoma Class 2 result was associated with a 2.8 times greater risk of recurrence than a Class 1 result in stage I patients. Further, we evaluated a subset of the articles to compare those with the highest (Class 2B) and lowest (Class 1A) risk DecisionDx-Melanoma results and found that a Class 2B result conferred a 6.5 times greater risk of recurrence than a Class 1A result (positive-likelihood ratio for Class 2B = 6.45). Thus, if Marchetti et. all had performed this analysis, they would have shown that the DecisionDx-Melanoma test did, in fact, add prognostic value. The results of Castle’s analysis have recently been accepted for publication.

The Marchetti study also failed to directly compare the accuracy metrics of DecisionDx-Melanoma to the accuracy metrics of pathologic staging by AJCCv8, or show the added value when combining all prognostic factors, as has been performed in other studies (meta-analysis by Greenhaw³, multi-center, prospective study by Hseuh⁶, single-center, prospective study by Arnot).⁵ In addition, as alluded to in the draft LCD, the manuscript does not evaluate the demonstrated clinical utility of the test to influence sentinel lymph node biopsy decisions in patients with T1-T2 primary cutaneous melanoma, though the seminal study describing this utility of DecisionDx-Melanoma was cited by the authors.²² Specifically, as shown in Vetto et al, a DecisionDx-Melanoma Class 1A test result was seen in 65% of patients ≥65 years of age with a T1-T2 melanoma – this is a significant clinical impact in patients with thin melanoma. Importantly, the Marchetti et al. article cites the potential for bias or reverse bias in the study limitations. These represent significant flaws, both methodological and interpretive, that are not only misleading to readers but are in direct conflict with the preponderance of published evidence, including recent high-impact publications, that clearly articulate the consistent clinical validity and clinical utility of DecisionDx-Melanoma.^{1-12,15-17,22,25-28,35-38 2–6,12,15}

Improved context for LCD:

A primary theme of the Castle Biosciences presentation made during the open meeting for MolDX: Melanoma Risk Stratification Molecular Testing on 6/28/21 was that the isolated citation and discussion of two articles^33,34 critical of GEP testing in melanoma does not provide the meaningful context of the complete publication landscape since the last reconsideration request. As such, the evidence cited and discussed in the draft LCD would benefit by inclusion of a balanced review of additional published manuscripts that draw different conclusions and contain responses to viewpoints of these two articles. As mentioned in the open meeting, there are multiple peer-reviewed published manuscripts that warrant consideration by MolDX medical directors as they affirm and strengthen the clinical validity and clinical utility evidence that supports continued LCD coverage for GEP testing in cutaneous melanoma, and DecisionDx-Melanoma specifically.

At the Open Comment meeting with Noridian medical directors on 6/24/21 a request was made to submit the full text of any article that falls under this umbrella of relevant articles that Castle Biosciences believes warrant review by MolDX medical directors. In order to facilitate review of these published manuscripts highlighted in the open meeting and/or discussed in this letter, we have compiled a table and file of the full text articles that should be included for a balanced review of recently published literature. The table and pdf article files have been submitted as an addendum to this letter. For transparency and completeness, two additional articles not discussed in the main text of this letter are briefly reviewed in the Appendix below. Suggested modifications to the text of the draft LCD have also been submitted as a separate addendum to this letter, and are informed by this expanded set of manuscripts to present a more complete context of the evidence that has been published regarding DecisionDx-Melanoma.

Summary:

In summary, substantial additional evidence has been published in eleven additional publications, including two large, independent, prospective datasets with increased follow-up time, two systematic reviews and meta-analyses, an expert consensus on appropriate use of the test within AJCC staging, and a clinical impact survey study confirming the value and actionability of DecisionDx-Melanoma test results. The two most recently published prospective studies both demonstrate that DecisionDx-Melanoma produces significant risk stratification for patients with stage I-IIA cutaneous melanoma.^5,6 Taken together with previously published studies, this new evidence provides additional support to continue the existing LCD and coverage criteria that enables coverage of DecisionDx-Melanoma for Medicare beneficiaries, and reinforces the positive impact this coverage decision has had on physician management decisions and patient care.

Given the weight of supportive evidence and the limitations of the critical articles cited above, we believe the two paragraphs added to the draft LCD should be modified substantially. The suggested language submitted separately as an addendum to this letter achieves a balanced perspective and avoids the implication that two articles without novel data bear equal weight or “cast doubt” on the clinical validity or utility that is demonstrated by over 30 peer-reviewed articles that support the current LCD. It is important to accurately characterize the publication landscape, and the draft LCD should be updated to include summaries of the articles described in this letter and included as a separate addendum. These publications further bolster the evidence for the improved accuracy of risk prediction provided by DecisionDx-Melanoma that guides SLNB decision making and augments risk stratification for recurrence and metastasis that informs patient follow-up, surveillance imaging and referrals. As asserted by Castle Biosciences during the 2019 reconsideration request, DecisionDx-Melanoma meets medical necessity and reasonableness standards, and recent evidence reviewed in this letter and open comment meeting further supports this claim.

As noted at the beginning of this submission, we do not have proposed changes to the Billing and Coding: MolDX: Melanoma Risk Stratification Molecular Testing that pertains to DecisionDx-Melanoma specifically.

Please refer to the three addenda to this letter submitted as separate attachments for Proposed Specific LCD Language Modification, article PDFs of publications with summary table, and PDF of the slide presentation given during the 6/28/21 Open Meeting.

Thank you for your consideration to this important LCD. We look forward to the final published version.

Appendix: Response to additional critical articles not included in draft LCD:

To achieve full transparency, we wanted to briefly summarize two recently published articles not cited in the LCD or discussed in the Open Comment meeting that we believe could be inaccurately positioned by others as critical of GEP testing in melanoma. It is important to emphasize that we are informing MolDX of these two articles despite the belief that neither article warrants inclusion in the draft LCD given the substantial limitations to the reported results and conclusions in each and discussions to retract or correct the articles. As with other critical articles referenced in the draft LCD,^33,34 it is important to consider these additional papers in light of the intended use of these tests as covered in this policy and highlight that none of the data directly contradicts the initial published data demonstrating concordance of DecisionDx-Melanoma GEP results with SLNB status.²²

1) Garg, et al. Nat Comm. 202139: This manuscript reports a discovery study by a group of international authors who used an RNAseq approach to develop a GEP signature to predict progression-free survival and overall survival for patients with cutaneous melanoma. This manuscript published data stating that they compared their discovery effort to the DecisionDx-Melanoma test in the same samples. This data and statement around this data is false. Specifically, this paper did not include any patients tested with DecisionDx-Melanoma and the article authors did not use the DecisionDx-Melanoma test in their work. Instead, they performed RNAseq to quantify the expression of individual genes that are included in the DecisionDx-Melanoma GEP test. However, they did not have access to and did not evaluate these genes using the proprietary DecisionDx-Melanoma algorithm. Additionally, as you know, the DecisionDx-Melanoma test is performed on qRT-PCR and there was no evidence confirming gene expression concordance across these two platforms. The article’s reference to DecisionDx-Melanoma testing is therefore false and misleading. Given that this article did not evaluate DecisionDx-Melanoma, this article should not be cited in the LCD.

2) Kangas-Dick, et. al. Ann Surg Oncol. 202140: This manuscript reports experience from a single center retrospective cohort. The article has substantial errors, conflicting statements and methodological flaws that cause concern regarding the validity of the analysis and, as such, the conclusions. For example, the study shows in tabular form that GEP is the strongest predictor of RFS under univariate analysis, with a strong p-value of <0.0001, yet it becomes non-significant (without showing the data) upon multivariate analysis. SLN status was also significant upon univariate analysis but insignificant upon multivariate analysis. However, in the text of the article as well as in the underlying poster that was presented 5 weeks prior to article submission, GEP results were shown and stated to be an independent predictor under multivariate analysis. Similarly, GEP was the most significant predictor of DMFS under univariate analysis but becomes non-significant (as does SLN status) upon multivariate analysis. Another example of inconsistency is the fact that the article states that they evaluated “all patients” who underwent surgery for cutaneous melanoma and for whom DecisionDx-Melanoma test results were received. They state this number to be 361. However, a check of Castle’s clinical database shows 401 patients (Data on file). There is no description of the inclusion or exclusion of clinically tested patients, and the possibility of a 10% potential change in cohort size raises concerns regarding the integrity of the data and the analysis. Finally, the data presented in this paper conflict with the results of other single-center studies using DecisionDx-Melanoma ^{5, 16} and the MSLT-I trial.²³ Based upon these limitations, Castle Biosciences submitted a request for retraction, and this article should not be cited in the LCD until a corrected analysis of this cohort is completed and published.

References, Appendix, Tables and Redline LCD, were received and reviewed

Thank you for the in-depth review of the draft revision of this policy, as well as for the numerous publications. These have been reviewed and have been incorporated into the policy. Of note, we do not believe any data presented contradicts the intent of the policy as written in the initial policy or reconsideration draft; namely that the GEP tests have the ability to better stratify patients based on SLN positivity and can likely drive better patient management in regards to the performance of a SLNB.

The following comment was submitted to CGS and Noridian:

I am writing to provide my support for the DecisionDx-Melanoma gene expression profile test and request that CMS continue to support of the test in its local coverage determination. I have contributed and/or led several studies showing the accuracy of the test, both independently and in combination with clinical and pathologic prognostic factors, for melanoma prognosis. We previously reported the accuracy of DecisionDx-Melanoma in three low-risk populations of melanoma patients, including sentinel lymph node-negative patients and those with stage I-IIA melanoma. The accuracy of the test in T1 (<1 mm) tumors was also assessed. T1 tumors account for most newly diagnosed melanoma and because they are frequently diagnosed, many melanoma-related deaths each year. As such, my practice carefully considers how to identify which thin tumors are likely to recur or metastasize to determine appropriate treatment strategies. The test from Castle Biosciences has been a valuable tool in this regard, and our clinical use is based on data showing that those patients with thin tumors and a high-risk test result had significantly lower 5-year recurrence-free and distant metastasis-free survival rates than those with a low-risk result. The results were consistent across all low-risk populations of patients studied, and provided confidence that the test adds important prognostic information to clinical factors. Although thin tumors (AJCC stage I if SLN negative) have good survival, it is apparent from these data that some tumors are riskier than others. Therefore, identifying these high-risk tumors with Castle’s test in conjunction with AJCC helps me alter patient management decisions in a risk-appropriate way, which would be hindered by removing patient access to the test.

Other studies to which I’ve contributed have demonstrated the value of the test for patients with a tumor on the head or neck region. This is a sensitive area for surgical intervention, and the sentinel lymph node biopsy procedure has important clinical limitations that can be alleviated by use of gene expression profile testing. In our study, combining the DecisionDx-Melanoma test with SLNB results improved prognosis compared to SLNB alone. Independently, the test more accurately stratified patient risk for recurrence, distant metastasis, overall survival and melanoma specific survival.

With a wide breadth of experience with both the validation process and clinical implementation of the DecisionDx-Melanoma test, I support its use in combination with clinical factors to both inform sentinel lymph node biopsy decisions and guide decisions about patient management. The data continues to reflect consistent results from study to study, which has led to its incorporation to our management workflows for patients diagnosed with melanoma at Cleveland Clinic or referred to our clinic.

Thank you for your letter of support.

The following comment was submitted to CGS:

I have been in dermatology practice for the last 35+ years and cared for many cutaneous melanoma patients. Today, I would like to express my support for the use of Decision DX Melanoma, a 31-genetic expression prolile (GEP). This assay helps me determine each of my individual patient's risk for experiencing a recurrence or metastases. I frequently utilize this assay and the Class score it provides to assist me in developing treatment plans, specifically, the frequency of follow up and whether to send the patient for routine imaging studies in order to pick up an early metastasis. There are many published papers that explore utility of this test in clinical practice, and I want to highlight a recent publication that is in line with my utility of this test. Hsueh and colleagues recently showed that in stage I or II melanoma, patients with a GEP results of Class 2 were more likely to have a recurrence (27% vs. 4%) or metastatic (23% vs. 1 %) event compared to patients with a Class 1 result (Hsueh et al. JCO Precis One 5:589-601. 2021 ). Additionally, for the entire study population, a Class 2 GEP score was a significant predictor of recurrence (JIR=4.34 p<0.001), metastatic (HR=5.45 p<0.001) and overall survival (HR=3.13 p=0.016). I strongly believe that Decision DX testing is an important and useful tool in early staging and ultimate survival of melanoma patients. Thank you for your time and consideration.

Thank you for your letter of support of our policy.

First, I want to disclose that I am a paid speaker for Castle Biosciences, the company who owns DecisonDx Melanoma.

DecisionDx-Melanoma has become a valuable tool for our dermatology practice. We want to gather as much information as possible when making tough decisions about management strategies for patients diagnosed with melanoma, and we understand the limitations of the clinical factors that are either used for staging or not included as staging criteria but available for consideration (regression, lymphovascular invasion, etc.). We have reviewed results from numerous studies (Greenhaw, Dermatologic Surgery; Hsueh, Journal of Hematology & Oncology; Keller, Cancer Medicine, etc.) and the data clearly supports that the 31-GEP test is one of the most accurate tools available for risk assessment. This has been demonstrated in single and multi-center studies, in retrospective and prospective studies, and in studies sponsored or unsponsored by Castle Biosciences. As a dermatologist who has seen many patients with thin, low-risk melanoma undergo the sentinel lymph node biopsy procedure, I have appreciated the information provided by the 31-GEP. We are comfortable managing patients with melanoma and following those patients if they choose to forego the sentinel lymph node procedure. DecisionDx-Melanoma allows us to identify those patients more accurately than any other prognostic factor. For this reason, we support Medicare coverage of the DecisionDx-Melanoma test and the current LCD.

Please do not hesitate to contact me with any questions.

Thank you for your time and consideration.

Thank you for your letter of support.

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

Following my presentation during your Open Meeting, I am pleased to submit the attached comments for your consideration related to the above noted MolDX: MELANOMA Risk Stratification Molecular Testing.

The recently proposed update to the local coverage determination for the 31-GEP has called into question the test's utility for risk stratification, particularly in the stage I population of patients diagnosed with melanoma. This is based on two recent publications by Grossman et al. and Marchetti et al. in JAMA Dermatology.^1,2 Notably, it was disappointing that factual responses to these articles from clinicians independent of Castle were rejected without review by JAMA Dermatology. While the LCD claims that the Consensus statement and meta-analysis evaluate a “host of clinical studies,” it should be noted that the two studies report on just a combined nine studies on the 31-GEP. More than 30 peer-reviewed studies support the 31-GEP test for use in patients with stage I-III melanoma, suggesting that the two reports are based on less than 30% of the available material.

Independent of the lack of completeness of the Grossman article, since the last LCD update there have been 11 publications supporting the clinical validity and utility of the 31-GEP.^3–12 That these publications are not included in the updated evidence report while the Grossman and Marchetti articles are, is concerning. Included in the 11 recently published manuscripts are two meta-analyses that provide primary evidence (in contrast to the Marchetti article) by supporting the use of the 31-GEP in patients with stage I-III melanoma. Similarly, new publications by Arnot et al. and Hsueh et al. provide compelling evidence that the 31-GEP accurately stratifies risk in patients with stage I-III melanoma.

In addition to these data, I am a senior author on a study under consideration for publication at JAMA Dermatology supporting the use of the 31-GEP for risk stratification of stage I-III tumors. For each main AJCC stage (I, II, and III), the 31-GEP identified patients with higher (Class 1A) and lower (Class 2B) melanoma specific survival (MSS) than is predicted by AJCC alone. For example, we found that the 31-GEP identified patients with stage I melanoma with MSS rates similar to Stage II patients. Conversely, the 31-GEP test identified patients with stage II melanoma with MSS rates similar to patients with stage I disease.

Of note, there is continued debate about the extent of sentinel lymph node biopsy use for patients with thin melanoma tumors. The melanoma field is one of the few areas of medicine in which patients with a risk of metastasis to the sentinel lymph node as low as 5% are considered for surgical intervention. Patients diagnosed with T1b melanoma are recommended to “discuss and consider” the SLNB procedure even though 90-95% of those patients will have a negative outcome following surgery. Similarly, national guidelines suggest that patients with T1a disease who have accompanying lymphovascular invasion, a mitotic rate above 2 per mm2, or uncertain microstaging also fall into that category of “discuss and consider” with a risk of metastasis equal to only 5-10%.¹³ The DecisionDx-Melanoma test has allowed us to further separate those patients who have low risk according to the molecular biology of their tumor and can safely avoid surgical intervention.¹⁴ The Grossman article recognizes this clinical hurdle and more than 70% of respondents to the survey included in that study called for a comparison of GEP profiling to SLNB. Our own studies have shown that dermatologists use DecisionDx-Melanoma test results to inform SLNB decisions. In a study of physicians who used the test, 36% said that they use the results “as part of [their] decision to recommend or not recommend a patient having an SLNB.” This lines up well to the 50% avoidance of SLNB as reported by Scott, et al.

The Consensus statement and meta-analysis by Marchetti and colleagues does not consider the entire body of published literature for the 31-GEP but use only selected manuscripts with questionable methodology to advance their ideas. More importantly, when analyzing the totality of data, it is clear that support of the 31-GEP for use in stage I, stage II, and stage III melanoma is clear and consistent; the 31-GEP test can accurately stratify patients by risk of recurrence and adds prognostic value to current staging guidelines.

References were received and reviewed.

Thank you for your comments regarding the Grossman et al. and the Marchetti et al publications.

The following comment was submitted to CGS and Noridian:

Re: Proposed LCD MolDX: Melanoma Risk Stratification Molecular Testing

We are contacting you on behalf of the Dermatology Contractor Advisory Committee (DermCAC), which represents the board-certified dermatologists in your carrier region. We appreciate the opportunity to share the dermatology perspective on the proposed Local Coverage Determination (LCD) MolDX: Melanoma Risk Stratification Molecular Testing.

When appropriately used, melanoma risk stratification molecular testing will help to identify a specific group of patients that might have otherwise been subjected to sentinel lymph node (SLN) mapping and biopsy. Melanoma risk stratification molecular test can provide the information that we need to spare them of this procedure. Additionally, SLN biopsy is costly and can carry with it a significant amount of morbidity which is difficult for patients.

An additional benefit of this tool is that it will also provide useful information when making decisions about adjuvant therapies that might alter a patient’s chance of survival from melanoma. The ability to use this test—when needed—to spare our patients from unnecessary procedures and suffering in addition to its use with helping to guide therapy, adds a new level of critical information needed to guide dermatologists in our treatment of melanoma.

As such, the DermCAC agrees with and supports the proposed LCD MolDX: Melanoma Risk Stratification Molecular Testing.

Thank you for your group’s comments.

The following comment was submitted to CGS:

I am writing during the open comment period for MolDx: DecisionDx-Melanoma as I strongly believe that the DecisionDx-Melanoma gene expression profile (31-GEP) test from Castle Biosciences can deliver more clinical information for our melanoma patients and help us reduce unnecessary sentinel lymph node biopsy (SLNB) procedures while finding those most in need of the SLNB procedure. From my perspective as a surgical oncologist, an important part of cancer patient care is providing quality care while reducing cost. I have been utilizing SLNB to stage my patients and direct their care since my Surgical Oncology Fellowship in 1995. While SLNBs are an important tool used for the risk stratification of melanoma patients studies have shown that in those patients who undergo SLNB with negative lymph nodes can also go on to develop local recurrence and distant metastasis (Morton 2014).

I have reviewed the published evidence in support of DecisionDx-Melanoma and believe that the evidence is strong that the test can both risk stratify patients for their risk of recurrence and metastasis as well as identify a subset of patients with T1-T2 tumors, >65 years of age with Class 1A 31-GEP results who have a very low-likelihood of SLNB positivity (<5%) and who can therefore safely forgo SLNB surgical procedure (Greenhaw 2020; Vetto 2019).

I am aware of a question posed by a Palmetto medical director during the June 28^th 2021, open meeting covering the LCD and I believe that my clinical experience can directly address questions about how clinicians are currently integrating DecisionDx-Melanoma 31-GEP scores to guide clinical decision making for SLNB procedures. My surgical oncology group and I have integrated results from published data into our clinical practice and have published a case series that describes the clinical performance of patients treated using our clinical workflow that incorporates 31-GEP testing (Scott 2020). As noted in this manuscript, our clinical decision-making was altered and the SLNB procedure was omitted for patients >65 years of age with a low risk (Class 1A) 31-GEP result and this resulted in a clinical management change for 50% of the Medicare-age population in our single center study cohort. Further, there have been no cases of locoregional recurrence or distant metastasis identified in our study cohort – including in patients who did not obtain SLNB based on their 31-GEP result and clinicopathologic risk profile. Thus, confirming that patients with a low-risk Class 1A 31-GEP result are both at (a) low risk of having a positive SLN and (b) low risk of having a subsequent metastatic event.

It is worth noting that our experience with 31-GEP testing to rule out SLNB for 50% of our Medicare age patients is in line with other data. Specifically, a recently published clinical utility study that highlights the fact that other physicians are also using DecisionDx-Melanoma 31-GEP results to impact SLNB management decisions (Marson 2021). Castle Biosciences also relayed interval data from the ongoing DECIDE trial that shows alterations in SLNB management decisions for 45% of the Medicare-aged patients enrolled in the study to date (personal communication with Castle Biosciences); indicating that groups enrolling in the DECIDE trial are using the DecisionDx-Melanoma 31-GEP test to guide SLNB management decisions in a similar manner and similar frequency. [See Table below for overview of these three study findings.]

Patient selection is an important part of evidence-based medicine and there is strong published evidence supporting the two clinical utilities for DecisionDx-Melanoma. From my perspective, the non-invasive 31-GEP test that can identify a subset of patients who can safely avoid the SLNB surgical procedure is beneficial for patient care and appropriate utilization of healthcare resources. Again, as noted in our recent single center study, we have incorporated DecisionDx-Melanoma testing into the clinical workflow for SLNB decision making and follow-up surveillance algorithms for clinical follow-up and imaging. Recently published prospective studies further support the clinical validity and utility of this test in patients who are being evaluated by surgical oncologists both for guiding SLNB decision making and, independently, providing valuable risk stratification for risk of recurrence in patients after they have decided to obtain or forgo the SLNB procedure (Arnot 2021 and Hsueh 2021).

Having a test that can spare my patients unneeded surgery, or conversely find those that are more likely to have a positive sentinel node, is an important tool to facilitate clinical management decisions in my clinical practice and I strongly support the continued Medicare coverage for DecisionDx-Melanoma as put forth in the existing LCD.

Thank you for your consideration

References and table were received and reviewed.

Thank you for your letter and sources.

The following comment was submitted to Noridian:

As a board-certified dermatologist and dermatopathologist who sees many elderly melanoma patients with thin tumors, I need to have additional tools to assess which patients are at higher risk of metastasis, and which patients are likely to do well. I use the DecisionDx-Melanoma test to identify patients with a higher risk of metastasis, so that I can determine which patients need to be referred to surgical oncology for sentinel node biopsy and who may need more frequent clinic visits or surveillance imaging.

I’ve found that the additional prognostic value added by the DecisionDx-Melanoma test is particularly useful in the Medicare population who have a lower risk of having a positive sentinel node, but who may be poorer candidates for surgery, and who tend to have worse outcomes than their younger counterparts. Therefore, I fully support the continued coverage of DecisionDx-Melanoma for Medicare patients.

Thank you for your comments.

The following comment was submitted to Noridian:

I am writing to provide my perspective on the DecisionDx-Melanoma test and the value of gene expression profile testing for patients diagnosed with melanoma. I have been associated with the validation of DecisionDx-Melanoma as a contributor to clinical studies and as a co-author of publications describing its accuracy. Most recently I am an author on an article submitted to, and currently under review at JAMA Dermatology that reports the clinical validity of the 31-GEP for determining the risk of death from melanoma for patients with stage I, II, or III cutaneous melanoma.

Notably, the 31-GEP is an independent predictor of MSS after accounting for AJCC staging factors and age. Further, the 31-GEP had high sensitivity and NPV in patients with stage I cutaneous melanoma. Because all patients with stage I CM are considered low risk, and NCCN guidelines recommends low-intensity management strategies accordingly, the 31-GEP could identify patients with stage I CM with a higher likelihood of experiencing an adverse event earlier than would otherwise be detected. While risk assessment can be further improved, the information provided by the 31-GEP improves upon current prognosis based solely on AJCC staging. Any patient with stage I CM correctly identified by the 31-GEP as high risk (experiences an event) is a patient that current guidelines would likely have missed. Thus, the 31-GEP helps identify which patients with stage I CM should be followed more closely for early recurrence detection, thus promoting earlier treatment and better long-term outcomes.

An important advance for the 31-GEP was the proven association of the 31-GEP test result with sentinel lymph node positivity. This finding provides dermatologists an opportunity to counsel early-stage patients considered for surgical interrogation of their lymph nodes. The vast majority of T1 patients (~95%) will have a negative sentinel lymph node biopsy result, as demonstrated by numerous studies, so the incorporation of information about the molecular biology of the tumor for guiding this important decision is critical. Based on the validity of the test for informing decisions about both sentinel lymph node status and 5-year recurrence risk, I have incorporated the test into the workflow for my patients and recommend that Medicare continues to support reimbursement for the test for the patients in my dermatology practice.

Thanks for your comments.

The following comment was submitted to Noridian:

I am a surgical oncologist and I specialize in the treatment of melanoma here in Arizona.

I’ve found that knowing molecular data about a patient’s primary melanoma is beneficial when trying to determine a strategy for patient care. There are some patients whose true risk is missed by conventional staging, so it is helpful to have molecular data that the 31 GEP provides. In a recent study published in Journal of Clinical Oncology Precision Oncology, patients with stage I-IIA melanoma and a class 2 result had risk similar to that of patients in the high risk (stage IIB-III) population overall indicating that patients with a Class 2 result may require additional surveillance, like more frequent clinic visits or routine imaging to catch distant metastases while tumor burden remains low. This extra level of information provided by the 31-GEP helps me to develop a more precise plan for patients with localized disease. Moreover, a Class 1 result from a tumor indicates that patients are at a much lower risk of recurrence and de-escalation of their surveillance can save time and resources for the patient and for me as the physician.

Please consider incorporating the i31-GEP Decision Dx Melanoma test into your coverage. This allows for closer surveillance for those patients that need it and for less intense surveillance for others.

Thank you for your comments.

The following comment was submitted to Noridian:

DecisionDx-Melanoma has become a valuable tool for our dermatology practice. We want to gather as much information as possible when making tough decisions about management strategies for patients diagnosed with melanoma, and we understand the limitations of the clinical factors that are either used for staging or not included as staging criteria but available for consideration (regression, lymphovascular invasion, etc.). We have reviewed results from numerous studies (Greenhaw, Dermatologic Surgery; Hsueh, Journal of Hematology & Oncology; Keller, Cancer Medicine, etc.) and the data clearly supports that the 31-GEP test is one of the most accurate tools available for risk assessment. This has been demonstrated in single and multi-center studies, in retrospective and prospective studies, and in studies sponsored or unsponsored by Castle Biosciences. As a dermatologist who has seen many patients with thin, low-risk melanoma undergo the sentinel lymph node biopsy procedure, I have appreciated the information provided by the 31-GEP. We are comfortable managing patients with melanoma and following those patients if they choose to forego the sentinel lymph node procedure. DecisionDx-Melanoma allows us to identify those patients more accurately than any other prognostic factor. For this reason, we support Medicare coverage of the DecisionDx-Melanoma test and the current LCD.

Thank you for your comments.

The following comment was submitted to Noridian:

As a community dermatologist who sees many melanoma patients, I would like to offer my support for the current Local Coverage Determination for Castle’s 31-GEP test. In my high-risk melanoma clinic, I use the Castle test to better understand the prognosis for patients diagnosed with invasive melanoma. The test can identify those who are at higher risk for metastasis (class 2A or 2B). Thus, I can adjust their follow-up protocol in line with their increased risk of recurrence. Conversely, a lower risk result enables me to establish a less aggressive follow-up schedule appropriate for their risk. After receiving these results, I find that my patients benefit from increased understanding and relief from uncertainty, consistent with recently published results from Marson et al. (SKIN 2021) showing that 70% of physicians agreed that the test improved patient understanding of their situation and over 80% would use or recommend the use of the test.

For these reasons, this test is part of my overall melanoma work-up for most Stage I and II patients along with American Joint Committee on Cancer (AJCC) staging. As I see in my practice every day, if identification of metastatic disease can be done at an early stage, patients have a higher chance of improved outcomes.

Castle’s 31-GEP test is important to melanoma care in my practice. I strongly recommend that the existing LCD remain in place to support coverage and ensure that there are no interruptions in my ability to provide the best patient care.

Reference was received and reviewed.

Thank you for your letter of support.

The following comment was submitted to Noridian:

Gene expression profiling is an important part of melanoma patient care in my practice. I want to do what is best for my patients while ensuring I don’t overburden the health care system. So, I participated in a study with Castle to define the appropriate use criteria for using the 31-GEP test for risk of recurrence. As outlined in Marks et al, SKIN 2019, we found that the 31-GEP stratifies risk of recurrence for patients with tumors with Breslow thickness of 0.3mm and above. Patients with thin tumors below 0.3 mm in thickness had a 5-year recurrence-free survival rate of 100%, while the 31-GEP significantly stratified risk for patients with tumors 0.3 - 1.0 mm thick (P < 0.0001). Patients with a Castle Class 2B result had a 75.5% 5-year recurrence free survival while those with a Class 1A result had a 96.7% recurrence-free survival. This dramatic reduction in Class 2B survival rates led me to change how I would manage patients with a melanoma diagnosis, reducing the frequency of office visits for low-risk patients and increasing the number of visits for Class 2B patients who need more frequent skin checks and, potentially, consideration of regular imaging in order to find asymptomatic metastases early. While I still use the GEP test for sentinel node biopsy guidance in patients with any Breslow thickness—even in thin tumors with concerning features— I use the test in patients with tumors of 0.3mm and above for determining who is more likely to have a recurrence.

Thank you very much for your consideration!

Thank you for your comment.

The following comment was submitted to Noridian:

As a board-certified dermatologist and fellowship-trained Mohs micrographic surgeons with over 10 years of experience, I’ve worked closely without hospital’s surgical oncologist to care for the Medicare population presenting with malignant melanoma. Patient management with surgery is often complicated by the location of the tumor, co-morbidities, or simply the anxiety of the patient about the procedure. This decision making has been particularly challenging amid the COVID-19 pandemic. It continues to be a challenge for my at-risk patients in California. Some patients have anxiety and fear to come into the office even for biopsies and routine surveillance. Castle Bioscience’s Decision-Dx (31-GEP) test provides information about the risk of a particular melanoma. In this manner, it has assisted me in discussions with my patients about medical and surgical management, including how long to wait, risk of recurrence or metastasis and even eliminating unnecessary procedures.

As a physician, I am primary responsible for the patient in front of me, but also the healthcare system at large. Using tumor biology supported by strong, published evidence to make decisions that both help my patient and the system at large, and providing patients with the right procedure while avoiding waste, align with the values as a physician. The 31-GEP test has had a substantial impact on our ability to accurately assess risk for our patients and I appreciate the continued coverage of this test.

Thank you for your comment.

The following comment was submitted to Noridian and Palmetto GBA:

I am the Executive Director of IMPACT Melanoma. I am a MD, board certified in dermatology and dermatopathology, am President and CEO of StrataDx and serve as a Director of IMPACT Melanoma. We were recently made aware of your decision to re-open the LCD and billing article for the DecisionDx-Melanoma (31-GEP) test. IMPACT Melanoma is a nationwide patient advocacy group focused on improving melanoma prevention, through awareness, care and teaching. Much of our work at IMPACT is focused on skin cancer prevention, education, and education patients diagnosed with early-stage melanoma and the treatment options available to them. As a practicing clinician and dermatopathologist, I, Lisa Cohen, can also speak to the impact of the availability of DecisionDx-Melanoma test to Medicare beneficiaries.

While there have been significant advances in the management of metastatic melanoma, we have not seen much progression in the management of newly diagnosed early-stage melanoma, namely Stages I and II. Specifically, it has been over thirty years since the surgical community began evaluating the sentinel lymph node biopsy surgical procedure (SLNBx) for identifying patients at higher risk of progression (metastasis) than those who are SLN negative. Additionally, it was fiercely believed that a SLNBx would be therapeutic- not only in identifying aggressive disease but also improving melanoma specific survival.

Unfortunately, from a patient care perspective, the multi-center MSLT-1 study demonstrated that there was no survival benefit in patients who received SLNBx with their wide local excision vs those who received wide local excision alone. As you may know, prior to the availability of DecisionDx-Melanoma, the only way to identify patients who should consider undergoing SLNBx was by pathologic features such as tumor depth (Breslow’s thickness), ulceration status and the presence of other risk features (mitotic rate, etc.) Using these features, and a recommended threshold of 5% SLN positivity for recommending the SLNBx procedure, we as a medical community have performed far too many unnecessary SLNBx. In fact, of all patients referred for SLNBx, only 12% will be SLN positive.

SLNBx has an 11% complication rate, and 18% false negative rate for regional metastasis to the SLN, and an estimated fully allocated cost of more than $18,000. Two-thirds of patients who die from melanoma are SLN negative. There is significant room for improvement in how we select patients for SLNBx. As you know from the current LCD, the DecisionDx-Melanoma test enables a significant number of patients to forgo SLNBx, since they would be projected to have a SLN positivity rate of <5%. Additionally, data recently presented by Whitman et al at the EACO April 2021 conference from a prospective multi-center study demonstrated that patients predicted to have SLN positivity rate <5% by incorporating the DecisionDx-Melanoma test score with clinical and pathologic features had 3 year recurrence-free survival rates of 98.6% and distant metastasis-free survival rates also of 98.6%.

This data brings us to our second use of the DecisionDx-Melanoma test: to guide management decisions more accurately than staging alone. As you undoubtedly know, melanoma management plans are based upon the estimated risk of recurrence or metastasis. This use of DecisionDx-Melanoma is also clinically valuable as two out of three deaths from melanoma occur in patients who were misclassified as low-risk with traditional clinical and pathologic features alone.

In the referrals that I see, clinicians use the DecisionDx-Melanoma test for its important clinical impact. Including the Medicare population with the LCD policy is imperative as many Medicare patients undergo unnecessary SLNBx, and given their age and co-morbidities are at greater risk of complications. We request that you maintain coverage for the DecisionDx-Melanoma test so that clinicians and patients who want to use the molecular test have access to the valuable information it provides.

Thank you for your letter of support to the current policy.

The following comment was submitted to Palmetto GBA and Noridian:

The 31-gene expression profile (31-GEP) prognostic test for cutaneous melanoma has been a critical tool that I use in my dermatology clinic to augment traditional AJCC staging. AJCC staging, a population-based risk estimator, misses THE MAJORITY OF MELANOMA DEATHS each year in the United States by assigning them a “low risk,” status. However, numerous outliers cause the majority of melanoma deaths. Surprisingly, more people die from thin melanoma’s (low risk) than from thick melanomas both in Australia and in the United States. The 31-GEP test allows me to identify biologically high-risk patients so I may individualize their care and essentially, better allocate the health care resources available to best treat them. Additionally, the 31-GEP objectively reinforces those who are at low risk and are unlikely to experience a recurrence. This is strongly represented across various studies with consistently significant negative predictive values (NPV). My conversations with truly low risk patients have dramatically improved as a result of implementing the 31-GEP into my melanoma patient care.

I have found there to be two primary utilities of the test in my decision-making process. First, in overall risk of recurrence. The test further stratifies risk beyond AJCC staging to offer a more comprehensive risk profile for patients with Stage I-III melanoma. This gives me an additional safety net to find the deadly outliers missed by AJCC staging. This utility is supported consistently throughout the robust body of evidence behind the development and validation of the test going back to 2015. Secondly, the test helps to better identify patients who may or may not benefit from sentinel lymph node biopsy (SLNB). There exists a nationally recognized threshold of eligibility for SLNB (5%) in melanoma patients that results in a very low yield. In addition the only randomized data on SLNB and survival outcome (the MSLT-1 trial) conclusively demonstrates that while 16% of SLNB patients are dead at year 12, over twice as many of these deaths came from the node-negative patients than from the node-positive patients. Also, only 16% of SLNB procedures found a positive node. This is a tremendous amount of surgery for poor information and NO SURVIVAL BENEFIT. In using the 31-GEP, my multidisciplinary team and I are able to most appropriately select patients who undergo the SLNB procedure, as well as better tailor their care when they do not benefit from surgical intervention.

I recognize the inclusion of recent negative articles for review and urge the committee to consider the most up to date evidence provided to support the test, most notably the meta-analysis by Greenhaw et al, and an expert clinical workflow by Kwatra et al. These studies have fortified the years of science behind the validity of the test and how it continues to change my practice and those of my colleagues across the United States.

The current standard of care with AJCC staging and it’s 50+ year old tool, the SLNB, is missing the majority of melanoma-related deaths. I did a quick calculation using my own practice and the 30 invasive melanomas per year I find. I was stunned to learn that if I simply follow the current AJCC/SLNB standard of care, during my 30-year career, I will incorrectly reassure between 30 and 60 of my OWN DEAR PATENTS that their melanomas are low risk. These patients will then go on die of their melanomas. I refuse to accept that this will be my future as a dermatologist.

References were received and reviewed.

Thank you for your comments.

The following comment was submitted to WPS:

I am a dermatologist and dermatopathologist with extensive experience with pigmented lesions and melanoma. I now practice in Jefferson City, Missouri.

I appreciate the ability to use the Castle Decision Dx test for my Medicare patients with melanoma. Using this test with patients with thin tumors helps me identify which patients I can follow up with myself and which ones I need to send to a high-risk clinic. For my Medicare patients in particular, who may be managing multiple medical appointments, it helps to have refined risk so my patients with higher risk results can 1nake sure their follow up is not delayed.

I follow the algorithm in the Kwatra, 2020 article which has been very useful in my practice. This is the standard of care for melanoma patients.

I cannot imagine not having this prognostic tool at my disposal. Please ensure this testing is a covered service for our patients. Thank you.

References were received and reviewed.

Thank you for your comments.

The following comment was submitted to WPS:

In the last several years, there has been dramatic advancement in new therapies for cutaneous melanoma.^1,2

This includes advances in adjuvant interventions for patients having asymptomatic disease following definitive treatment. Effective adjuvant therapy for melanoma has been focused mainly on patients with Stage IIIB and higher disease, and results from clinical trials investigating adjuvant therapy in early stage melanoma (i.e., Stage I – IIC)³ are pending study completion.

With respect to recent suggestions by Grossman et al. (2020) for demonstration of validity and utility of gene expression profile (GEP) testing in melanoma,⁴ following processes previously used in breast cancer GEP test development is not plausible for the 31-GEP test for melanoma as the indications for breast cancer GEP testing and the 31-GEP test are different. Nevertheless, the prognostic 31-GEP test⁵ can be used within current national guidelines⁶ to further stratify risk of recurrence and metastasis in Stage I-III melanoma, and help guide patient management decisions with regards to adjuvant therapy.

The results of the test can and should be discussed between clinicians and their patients along with clinicopathologic risk assessment. This would allow for improved clinician-patient decision-making for risk-appropriate patient management.

References were received and reviewed.

Thank you for your comments.

The following comment was submitted to WPS:

As a practicing Dermatologic surgeon and J5CAC representative for Iowa, I am in support of the proposed LCD regarding DecisionDX-Melanoma. The proposed criteria for use of this test are in concordance with my personal clinical practice and are also in line with that of my colleagues who use this test in their practice.

Thank you for your letter of support to the current policy.

The following comment was submitted to WPS and Palmetto GBA:

My name is Dr. XXXXXX, a surgical oncologist at Saint Louis University School of Medicine in St. Louis, MO. Our practice performs hundreds of sentinel lymph node biopsies each year for patients with melanoma. While SLNBx is a critical part of melanoma staging, I have found that the Castle DecisionDx-Melanoma test provides additional information that I incorporate into my clinical judgement as a surgeon. Typical tumor features, like thickness, are essential for decision making, but for those patients with thinner tumors, the choice to perform surgery, especially in my older patients, can become more complicated. Having additional information from the Castle test to bring to discussions with my patients often will help us both settle on the

right decision for the particular situation. Often for patients with thinner tumors who may only be slightly at risk for a positive node, the Castle test can help identify who really doesn’t need the procedure because of the low risk that the surgery will identify nodal disease. Conversely, in patients with thicker disease, this test helps find those patients likely to have nodal burden and thus are the right people to undergo SLNBx.

I also support the DecisionDx-Melanoma test because I have been integrally involved with studies demonstrating the prognostic accuracy of the test.^1,2 Most recently I was the primary author on a publication in the journal JCO Precision Oncology that reported its accuracy in over 300 melanoma patients with longer-term follow-up.³ The study showed that the test significantly separated high- and low-risk patients throughout the cohort and, importantly, also in those patients with early-stage I-IIA disease. Results of the study aligned with results from our single center analysis of the DecisionDx-Melanoma test, and to those from other groups, emphasizing the consistency of DecisionDx-Melanoma prognosis.^4,5 Based on that consistency and first-hand experience in my own practice I support the CMS decision to cover this test according to the specifications of the current, previously approved LCD.

References were received and reviewed.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

I am submitting a comment for the MolDx: Melanoma Risk Stratification molecular testing as I am a co-author in several articles cited in the draft LCD. I am an Assistant Professor in the section of Dermatology at Baylor University Medical Center in Dallas, Texas and am a board certified dermatologist and Mohs Surgeon. As a researcher and expert in the field of cutaneous melanoma, I hope that my perspective can be considered.

It is my strong belief that the coverage criteria for MolDx: Melanoma Risk Stratification molecular testing should remain unchanged and should continue to provide coverage for DecisionDx-Melanoma testing as this GEP test in particular has demonstrated clinical validity to improve the accuracy of risk prediction for melanoma and has demonstrated clinical utility to inform a range of risk appropriate management decisions for my patients with melanoma.

As much of my published research has focused on the clinical actionability of genetic testing in dermatologic practice, I wanted to focus my comments on the current clinical utility of DecisionDx-Melanoma testing. In our practice at Baylor University Medical Center, DecisionDx-Melanoma has provided information to guide important clinical decisions both for the use of SLNB surgical procedure as well as management decisions such as follow-up intervals and surveillance intensity for patients with melanoma. We have been able to both avoid unnecessary SLNB procedures while directing those patients to surgery who are most likely to benefit from the procedure. With improved risk stratification, we can also more appropriately triage patient follow-up and surveillance by combining GEP test results with AJCC staging information. A study I led (Farberg et al, J Drug Dermatol, 2017), provided survey data with a 99% response rate (an appropriate response rate for publication, in contrast to published survey response rates in a recent JAMA Dermatology article from Grossman et. al. with a combined survey response rate of 14%). This study (and many others) supports our current real-life clinical utility of the test, demonstrating that the 31-GEP results are clinically utilized in a risk-appropriate manner to guide established clinical workflows.

An objective review of all the published data demonstrates how the 31-GEP may be utilized in the management of cutaneous melanoma. Thousands of clinicians currently have, like myself, incorporated the 31-GEP into their clinical practice for the benefit of their patients. I have patients scheduled this week who have benefitted from the 31-GEP. The coverage criteria for this LCD should remain unchanged and should continue to provide coverage for DecisionDx-Melanoma.

Thank you for the letter of support to the current policy.

The following comment was submitted to Palmetto GBA:

I am a board-certified dermatologist and consultant in the Department of Dermatology at Mayo Clinic in Minnesota with a joint appointment in Biochemistry and Molecular Biology. I have been practicing clinical dermatology for more than ten years. My clinical interests are in the care of melanoma patients. In clinical practice, I see melanoma patients of all stages. My research focuses on the role of secreted proteins for cellular transformation. The secretory phenotype of cancer cells is often associated with changes in gene expression, which can be reliably quantified by gene expression profiling (GEP). I have published extensively on GEP based testing in malignant melanoma.^1–4 For example, my research has led to the development of the Merlin assay, which is a qPCRbased test that uses clinicopathologic and gene expression variables to predict if a patient with a thin or intermediate thickness primary melanoma can forgo a sentinel lymph node biopsy (SLNb) due to the low risk of nodal metastasis.

Primary cutaneous melanoma accounts for only 1-2% of all skin cancer diagnoses but 75% of skin cancer deaths. This is due to the extraordinary ability of melanoma cells to metastasize. Global melanoma incidence rates are rising, and this trend is projected to continue for several decades in the United States and other nations with fair-skinned populations.⁵ With an estimated combined incidence of greater than 250,000 cases in the United States, Europe, and Australia in 2020, melanoma incidences approximate those of colon cancer.

The risk assessment of primary melanoma is now more critical than ever because for the first time in history, we have therapies available that reduce the likelihood of disease progression.⁶ However, these therapies cannot be indiscriminately provided to all melanoma patients: most patients with primary melanoma do well without adjuvant interventions;⁷ adjuvant therapies can induce serious adverse reactions,⁸ and they are expensive.⁹ In current practice, patients are selected for adjuvant interventions based on the presence of metastasis.⁷ The most sensitive test for detecting metastasis is the sentinel lymph node biopsy (SLNb), a complicated invasive procedure that requires extensive outpatient, inpatient, and pathology resources, including many subspecialists (dermatologists, surgeons, radiologists, anesthesiologists, and pathologists).¹⁰ Not all melanoma patients are eligible for this staging procedure. Only patients with a pretest probability of metastasis of greater than 5 to 10% are referred based on guidelines, equivalent to patients with a ≥0.8 mm Breslow thickness melanoma.^11,12 If the SLNb shows metastasis, patients are referred to medical oncology for adjuvant interventions (such as PD-1 inhibitors). Using this approach, approximately 85% of SLNb procedures are negative and without a therapeutic benefit.^1,13 Moreover, despite the large number of SLNb, some patients with positive SLNb remain unidentified, either because the SLNb is falsenegative¹⁴ or because patients with thinner primaries forgo the procedure.¹⁵ These patients miss an opportunity for adjuvant interventions. There is an unmet clinical need for improved primary melanoma risk assessment to better select patients for SLNb and avoid unnecessary procedures.

SLNb is a more complex and time-consuming procedure than is often appreciated. This can be illustrated by a representative case of a Mayo Clinic patient who enrolled in a feasibility study of the CP-GEP/Merlin assay, which identifies patients who are at sufficiently low risk for metastasis so they can safely forgo SLNb (Fig. 1). Following a melanoma diagnosis, the patient met with a surgeon, underwent preoperative evaluation with customized testing to ensure a safe recovery, and completed a technetium-99m scintigraphy (single-photon emission computerized tomography [SPECT]) in nuclear medicine combined with computed tomography (CT) to produce SPECT-CT scans. This was followed by SLNb and wide local excision of the primary tumor in the operating room under general anesthesia. Tissue was paraffin-embedded, sectioned, and stained by hematoxylin and eosin, and melanocyte lineage-specific immunohistochemistry was determined and interpreted by anatomic pathologists. In this case, weeks passed from melanoma diagnosis to finalized SLN biopsy results. Surgery was followed by postoperative care.

SLNb also carries a significant complication rate. For example, a review of 577 patients who underwent SLNb without completion lymph node dissection at Mayo Clinic between 2004 and 2018 revealed an overall complication rate of 17.4%. Patients developed seromas (9.3%), lymphedema (4.3%), infection/cellulitis (4.8%), hematomas (3.3%), and wound dehiscence (2.5%). 2.7% of patients had to visit the emergency room or were readmitted to the hospital (unpublished data).

Using the CP-GEP/Merlin test, up to 80% of patients with T1b, 48% of patients with T2a, and 24% of patients with T2b melanoma can forgo the SLNb procedure as they are identified as low risk for nodal metastasis. The error rate is less than 5%. Test results are available within a few days (Fig. 1) and do not require patient interaction with healthcare providers. Hospital resources can be prioritized to other tasks.

At Mayo Clinic, we are fortunate to have access to the Merlin assay. Several of my patients were tested as part of routine clinical care. For example, we tested:

• A patient with a stage T1b melanoma and multiple comorbidities, including congestive heart failure, diabetes mellitus, lower extremity edema, and frailty age-related physical debility; the melanoma was on the thigh, a location which increases the risk of SLNb associated complications (seroma, infection,lymphedema); the patient wanted to have the primary melanoma tested by CP-GEP as they very much wished to avoid SLNb; unfortunately, the test came back high-risk and the patient was therefore referred to general surgery to discuss SLNb.
• A 48-year old male with a stage T1b melanoma of the left cheek and a history of metastasized cancer and cancer-associated pain, the primary melanoma was categorized as low risk by Merlin; SLNb was not performed.

While the molecular staging of melanoma will undoubtedly become part of mainstream medical care, GEP-based testing must be appropriately validated, e.g., as outlined in a recent consensus paper.¹⁶ The CP-GEP/Merlin assay has by been clinically validated in several independent cohorts.^1,17,18 Crucially, the assay has been:

• Benchmarked against standard established clinicopathologic variables, AJCC 8th edition risk stratification models, and nomograms from the Melanoma Institute of Australia (MIA) and the Memorial Sloan Kettering Cancer Center (MSKCC).^1,19
• Developed from primary diagnostic biopsy tissue (as opposed to re-excision material)
• Evaluated across the spectrum of intended use, not convenience samples
• Developed based on a defined endpoint, i.e., SLNb status at the time of melanoma diagnosis
• Evaluated in large unselected and independent validation sets.

Of note, CP-GEP achieves excellent SLNb reduction rates in patients 65 years and older (46.5% for patients diagnosed with T1-T2 melanoma).²⁰

In conclusion, I fully endorse the coverage of molecular testing for melanoma patients to improve risk assessment and individualized care. Harvard economist Michael Porter defined value in health care as "the health outcome achieved for patients relative to the cost of achieving them".²¹ Reducing the need for expensive surgery will facilitate clinical care delivery in lower resource settings, in underserved populations, and during times of healthcare crises such as the current Covid-19 pandemic.

I believe that CP-GEP will positively impact patient care.

Please feel free to contact me if you need any further information.

References and Tables were received and reviewed.

Thank you for your support of the coverage policy.

The following comment was submitted to Palmetto GBA:

I am emailing to indicate my thoughts regarding the role of DecisionDx gene expression profile testing in melanoma patients. Generally, surgeons take a limited role in patient follow-up for patients with disease that has not spread to the lymph nodes. However, Vetto et al at OHSU wanted to assess whether gene expression profiling had an impact on surgeon recommendations for patients with pathologic stage I and II melanoma. In the Journal of Drugs in Dermatology, 2018, they published a study assessing Class 1 (low risk) and Class 2 high risk in patients with localized disease. They found that the majority of Class 2 patients were followed by a combination of dermatology, surgical oncology, and medical oncology, while most patients with Class 1 GEP scores were followed up by dermatology alone due to the high NPV (94%) seen in patients with stage I -II disease and a Class 1 result. The most striking example of this dichotomy was seen in stage I patients with 82% of GEP Class 1 patients following up with dermatology alone, and 100% of Class 2 following up with surgical oncology. The majority of patients in the study that were referred to medical oncology had a Class 2 result. Since catching tumor metastases early improves outcomes, increasing disease surveillance for patients with a high molecular risk according to the 31-GEP is a part of our strategy to improve patient care.

I utilize Decision Dx testing in many of my higher risk stage I and II patients with the idea that a class IA result assists in reassuring that this is likely to be a non-aggressive lesion, whereas a class II result warrants a more intense approach therapeutically, diagnostically, and from a follow up perspective. Thank you for your attention to this matter.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

In the management of cutaneous melanoma, high-risk patients have options for consideration of treatment in the adjuvant and metastatic settings, as effective therapies have been expanding rapidly over the last decade. With regards to adjuvant therapy, it is imperative to stratify risk early in the course of disease for better-informed decision-making with respect to the most appropriate therapy for optimal outcomes. Tumor burden is an important factor associated with response to treatment, and lower tumor burden at the time of intervention corresponds with improved responses and survival outcomes. Therefore, improving risk stratification earlier in disease while tumor burden is low (asymptomatic relapse identified by surveillance imaging) relative to later stages of disease when tumor burden is often higher and clinically symptomatic is key to risk-appropriate treatment and more favorable outcomes.

The gene expression profile test offered by Castle Biosciences (DecisionDx-Melanoma) has been prospectively and independently validated for risk stratification of Stage I-III melanoma patients. The test can identify patients at highest risk of recurrence and improves upon stratification within AJCC Stage II and III melanoma for which decisions regarding therapeutic intervention are necessary. This test has clinical application for guiding those decisions and loss of access to the test would negatively impact patient management. I routinely identify patients with early stage melanoma with asymptomatic relapse of their cancer because I intensify surveillance for high GEP risk patients- as they behave clinically like higher stage patients. If we lose access to this test, patients with high risk early stage melanoma will lose access to personalized surveillance plans, treatments, and clinical trials.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

I am a dermatologist in a busy private practice, and I have routinely cared for cutaneous melanoma patients for the past 18 years. I am writing to discuss my utility of a 31-Genetic Expression Profile (GEP), namely, Decision-Dx Melanoma. This test fills a knowledge gap we have in managing our melanoma patients. This elusive cancer can spread though the bloodstream and bypass the lymphatic system. A negative sentinel lymph node biopsy (SLNB) does not mean a patient is out of the woods. I have had several patients go on to have metastatic disease who had a negative SLNB or thin melanomas that did not qualify for SLNB. However, the GEP helps me to identify patients that are in a higher risk group, despite their classification by NCCN. These patients will get more intensive follow-up as well as referrals to medical oncology for consideration of imaging and immunotherapy. Conversely, the GEP has a very high negative predictive value, and I am very confident that patients with a 1A score will not need any referrals or imaging. It is nice to spare these patients unnecessary worry and procedures. I find the GEP especially helpful in patients over the age of 65. In this group it has been shown that it can inform which patients will need a SLNB. I have been able to confidently counsel patients that they do not need the expensive SLNB procedure, and they are thrilled to be spared the morbidity and potential complications. Therefore, this particular assay has changed who I refer to surgical oncology groups to be considered for the SLNB procedure. I have referred patients that would not otherwise qualify for SLNB, and they were found to have positive nodes which may have saved or extended their life. There are several studies that support the use of GEP in the care of melanoma patients. I will highlight a recent one for you: In a meta analysis by Greehaw and colleagues, (Greenhaw et al. JAAD 2020), a hazard ratio of 2.75 was calculated for a GEP Class 2B result. That was similar to the calculated hazard ratio of 2.8 for SLNB. When you examine the accuracy metrics of both the GEP and SLNB alone in this meta analysis, the GEP outperforms SLNB; however, when you combine GEP with SLNB, the sensitivity for DMFS increase to 88% while the NOV remains high at 91%. This shows that the combination of SLNB and GEP results is superior to either modality on its own and provides a much clearer prognostic accuracy. I routinely order this test on all of the melanomas I diagnose in my practice that are 0.3mm and deeper. I feel this is an indispensable piece of information that unlocks a patient’s personal risk based upon the specific biologic behavior of their tumor. I have been using this test routinely for the past 5 years, and I would welcome any questions that you may have related to my utility of the GEP.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

In my clinic, we utilize DecisionDx-Melanoma (31-GEP) to identify patients at low risk for sentinel lymph node metastasis, and stage I and II patients who are high risk according to the 31-GEP in order to follow them more closely than would be recommended by current guidelines, in the hopes of detecting recurrence or metastasis early. Recently, I participated in a study to understand how DecisionDx-Melanoma has performed over multiple studies in patients with stage I-III melanoma.¹ Our meta-analysis of the 31-GEP showed that the test stratified risk in all main AJCC stages, and it was striking that those patients with stage I melanoma classified as being at the lowest risk by the 31-GEP (Class 1A) had significantly higher 5-year RFS (97.6%) and DMFS (98.4%) than patients with the highest risk result (Class 2B, RFS: 76.1%; DMFS: 86.0%, P<.001 for RFS and DMFS). Similar results were seen in both stage II and stage III melanoma. As the primary author of this meta-analysis study that stands in stark contrast to the meta-analysis being considered as part of this LCD review, I find it curious that our study was not included in the evidence supporting the 31-GEP. This study provides compelling evidence for the accuracy of the 31-GEP not only in stage I, but also stage II and stage III. It also highlights the enhanced prognosis achieved when the test is added to current clinical factors and sentinel lymph node biopsy results.

The 31-GEP is critical in my practice because 1) a Class 1A result can increase confidence in reducing the intensity of a patient management plan, and 2) a Class 2B result provides a reason to communicate with the patient that we should consider an increase in the frequency of clinical visits and patient self checks, or imaging and referral to surgical oncology.

We previously published that the 31-GEP can also be clinically impactful for busy dermatology practices by simply identifying Class 1 patients who can “undergo clinical skin and nodal examinations twice yearly for 2 years, then yearly thereafter” compared to Class 2 patients who may be “examined every 3 months for 2 years, then every 6 months for 3 years, then yearly thereafter.”² This workflow methodology fits within the wide recommendations of the NCCN panel, and substantially reduces the burden on dermatology practices that are seeing more and more patients each year. Because many patients may not be able to afford the test without Medicare coverage, removing coverage for DecisionDx-Melanoma would be problematic for providing the most current, up-to-date clinical care.

References were received and reviewed.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

I am writing to encourage directors of the MolDx program to continue coverage for the 31-GEP test in the interest of improved care for my patients diagnosed with melanoma. As a practicing dermatologist and skin cancer specialist who works with a multidisciplinary skin cancer team, I have first-hand experience of the value of molecular profiling when used in conjunction with other staging factors. In my practice, the evidence supporting the accuracy of the test and the information provided by the test have an important impact on the management of my patients, guiding who can be seen less often at our office and who should be watched more closely.

Although cutaneous melanoma treatments have greatly improved in recent years, many patients who later die from melanoma were initially assessed with the sentinel lymph node procedure and continued with an early-stage diagnosis following a negative result. To help address this clinical gap, I use the DecisionDx-Melanoma test as part of my assessment of patients who return to me following a negative sentinel lymph node biopsy. Additionally, I have confidence that patients who have thinner tumors and a Class 1A result can avoid referral to surgical oncology because they have a very low risk of SLN positivity.

Published studies support this use of the test to identify those patients (Vetto et al. Future Oncology, 2019) and they have been shown to have excellent outcomes in long-term studies regardless of lymph node status (Morton et al. NEJM, 2014), enabling some patients to forgo SLN surgical biopsies and eliminate the risks and costs of that procedure. I also use the test to risk stratify patients and allow me to plan their follow-up in a way that is aligned with their risk of melanoma recurrence/metastasis. Published evidence supports routine use of the 31-GEP test, including two recent prospective studies (Hsueh et al. JCO Precision Oncology, 2121 and Arnot et al. Am Journal of Surgery, 2021) that show the test accurately and independently predicts risk of recurrence and distant metastasis for melanoma patients.

In my practice I now use the test result to establish follow-up protocols, including decisions on referrals and imaging, that are appropriate for each patient’s individual risk.

I am confident that the DecisionDx-Melanoma test offers information that can improve patient care. I urge you to maintain the existing LCD and continue to provide coverage for the DecisionDx-Melanoma.

References were received and reviewed.

Thank you for your comments and citations

The following comment was submitted to Palmetto GBA:

I am writing on behalf of the Melanoma Research Foundation (MRF) in response to the request for comment on the draft LCD MolDX: Melanoma Risk Stratification Molecular Testing.

The MRF is the largest independent organization devoted to melanoma. Committed to the support of medical research in finding effective treatments and eventually a cure for melanoma, the MRF also educates patients and physicians about prevention, diagnosis, and the treatment of melanoma. The MRF is an active advocate for the melanoma community, helping to raise awareness of this disease and the need for a cure.

We appreciate the opportunity to comment on the proposed Local Coverage Determination (LCD) published by the Centers for Medicare and Medicaid Services (CMS) concerning revisions to the policy based on recent publications concerning the use of DecisionDX as a standard of care prognostic for patients diagnosed with cutaneous melanoma (CM).

Prior to 2013, treatment plans for patients with CM relied solely on clinical and pathologic prognostics factors, such as sentinel lymph node biopsies (SNLBx), a surgical procedure to identify if cancer cells have metastasized regionally within a patient’s lymph nodes. SNLBx surgeries are both invasive to the patient and many times cause debilitating side effects such as lymphedema, which can raise a patient’s risk of infection, slow the healing of wounds and can a lifetime of unnecessary pain, rehabilitation and costs to the patient and their insurance plans.

The use of noninvasive prognostics to influence SLNBx decisions as well as subsequent management decisions enables the Medicare beneficiary and their clinician to make a more informed, more accurate decision to undergo an SLNBx or to avoid it. The MSLT-1 trial proved that the SLNBx procedure in melanoma has no therapeutic benefit – meaning that patients who undergo SLNBx with their wide local excision compared to wide local excision alone die from melanoma at the same rate. So SLNBx is a tool to fully stage patients but not therapeutic.

Furthermore, relying upon just clinical and pathologic factors to select candidates for SLNBx results in only 12% of patients shown to be SLN positive meaning that 88% of patients could have forgone this invasive procedure that requires general anesthesia. Finally, the SLNBx cost to Medicare is not insignificant. Kanzler, et al⁴ states that the 2007 Medicare reimbursement rate was $19,000 for the San Jose, California region. This correlates well to the $18,183 cost for an SLNBx as reported by Erickson, et al in 2014. Thus, being able to potentially avoid upwards of 70% of patients from having to undergo SLNBx not only reduces unnecessary complications but saves the Medicare Trust Fund significant money.

The risk of CM is increasing at dramatic rates in the United States According to the American Academy of Dermatology (AAD), invasive CM is projected to be the fifth most commonly diagnosed cancer for both men and women in 2021.^1-2 In fact, in 2021, melanoma is expected to take the lives of approximately 7,180 Americans.^1-2 For CM, the risk that a patient will develop invasive or metastatic disease is important to know in regards to surveillance and medical intervention.

With the incidence of invasive melanoma is projected to increase to approximately 112,000 by 2030, with a 250% increase in the cost of treating new cases, we cannot afford to change the existing LCD coverage⁵.

Therefore, we are requesting the existing LCD coverage to remain the same.

References were received and reviewed.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

Thank you for the opportunity to comment on the draft LCD regarding the 31-GEP test for melanoma patients. I would like to offer support for continuing with the current LCD.

I am a dermatologist who has experience using the test to guide discussions with my patients on their sentinel lymph node biopsy and surveillance plan decisions. Because it is based on tumor biology, the test offers an additional, independent assessment of patient risk that I cannot get from other tests or AJCC staging methods. Based on my experience I recognize that AJCC staging for melanoma is an imperfect risk assessment?tool. In fact, more patients who are diagnosed with early stage melanoma (Stage I or II) die of melanoma then those that are Stage III at diagnosis. Thus, there is a significant clinical need to improve prognostic accuracy and the 31-GEP test is indeed an improvement to other risk assessment means available to me.

This test is extremely useful to me in patients with T1-T2 tumors in discussing sentinel lymph node biopsy, imaging, and clinical follow-up decisions. I also find it helpful in patients who have thinner tumors <0.3mm Breslow’s thickness for whom a sentinel lymph node biopsy is being considered. This would include those whose tumor is transacted at the base, leaving the true depth unknown, and/or with other high-risk tumor features.

I agree with Castle’s recommendation that GEP results should be considered in a multidisciplinary setting in the context of other clinicopathologic factors, and that patient management decisions need to align with the overall risk of the patient. An expert panel recently published an algorithm outlining test criteria and management changes for Stage I-III patients (Kwatra et al. JCAD 2020), and I believe that this provides an excellent overview of integration of the test in dermatology practices such as mine. Importantly, a recent prospective study provides utility for more frequent follow-up for Class 2 patients compared to Class 1 patients (Hyams et al. Future Onc 2021). I recommend that newer publications such as these be included in the LCD to provide a more complete view of the data supporting the 31-GEP test.

I fully support the current LCD policy so that I can continue to use this information with my patients to understand their individual risk and make more informed care decisions.

Thank you for taking into account the clinical experience of dermatologists and the most recent evidence.

References were received and reviewed.

Thank you for your support of the policy.

The following comment was submitted to Palmetto GBA:

I am a board certified dermatologist, dermatopathologist and prior president of the American Academy of Dermatology with decades of clinical experience and I am writing to encourage the continued coverage of DecisionDx-Melanoma by the current Medicare LCD. My dermatology practice has utilized the test for many years as an addition to currently recommended factors that aid in directing patient care according to risk of recurrence. More recently, the test has been instrumental in helping us identify patients diagnosed with melanoma who are not likely to have positive results from the sentinel lymph node procedure.¹ This is an important utility of the test because we have a substantial number of patients who are low risk, with thin tumors less than 1mm in thickness, but who still qualify for surgical assessment of their lymph nodes because they have high risk T1a or T1b disease.² Current guidelines specify use of the sentinel lymph node biopsy because these patients have a 5-10% risk of metastasis to the lymph node, and Castle Biosciences has reported convincing studies that support the use of the 31-GEP in conjunction with AJCC staging to better identify patients likely to have a negative SLN who can safely forego the procedure.^1,3-6 Importantly, I was the senior author on a recently published manuscript that established a floor for GEP testing at 0.3mm Breslow thickness.⁷ The data support the 0.3mm cut point for risk stratification of patients who may be at elevated biological risk of recurrence and metastasis and whose clinicians will change management following their receipt of GEP test results. The management decisions that were impacted include the frequency of clinical visits, imaging, referrals, laboratory tests and sentinel lymph node biopsy decision making. Taken together, this paper highlights both the current clinical validity and clinical utility of DecisionDx-Melanoma in the thin (T1) cutaneous melanoma population.

Finally, the potential for reducing patient anxiety, saving money for patients (and the healthcare system), and protecting patients from potential surgical harms such as lymphedema from unnecessary SLNB procedures, is why I have incorporated DecisionDx-Melanoma into my patient assessments. Continued coverage of DecisionDx-Melanoma will allow me access to the most current methodologies for risk determination, promoting the best treatment decisions for my patients.

References were received and reviewed.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

I am a surgical oncologist at Desert Surgical Oncology in Rancho Mirage, CA. I direct a high volume surgical program that provides care to patients with malignant melanoma, among other diseases. I would like to provide written comments to the above-referenced MolDx local coverage de-termination, which concerns the DecisionDx-Melanoma test developed and distributed by Castle Biosciences, Inc. (Friendswood, TX).

In particular, I would like to comment about the value of the DecisionDX gene expression profile (GEP) in avoiding unnecessary lymph node assessment surgery in an important population of patients. This comment is based on empiric investigative data I will provide you, and an extensive personal experience that has saved my patients pain, morbidity, and cost.

I would first reference a study led and published by Dr. John Vetto at Oregon Health & Science University demonstrating a low likelihood of sentinel lymph node positivity in older patients with a favorable prognostic risk classification by the DecisionDX GEP.

The study of 1065 patients showed that patients 65 and older, having T1 or T2 melanoma, and a lowest-risk DecisionDx-Melanoma Class 1A result (448 patients) had only a 1.6% (95%CI= 0.5-3.6%) rate of detectable nodal disease following sentinel lymph node biopsy (SLNB).¹ Current national standards, recommend consideration of SLNB in individuals with a minimum 5% likelihood of a positive SLNB result.² This evidence strongly suggests that DecisionDx GEP can identify patients whose combination of age, stage, and biologic risk does not warrant nodal biopsy.

I have served as a co-investigator in a separate study, with a novel cohort of patients, that confirms and expands the results published by Vetto and colleagues. This study, initially presented at the Society of Melanoma Research with 840 patients, and currently being prepared for publication with a larger population of 1253 identically eligible patients, evaluated individuals using the same age and stage cut-points. For the population 65 and older, having T1 or T2 melanoma, and a lowest-risk DecisionDx- Melanoma Class 1A result (n= 604 patients), 2.0% (95%CI= 0.8-4.0%) had a positive sentinel lymph node biopsy.

The data I refer to from the two individual studies, consisting of the Novel Cohort, as well as that of Vetto et al., is reproduced in Table 1.

Table 1. SLN positivity in T1-T2 melanoma patients ≥ 65 years stratified by 31-GEP Class (Hyams, Identification of melanoma patient subsets with a low likelihood of sentinel lymph node positivity and favorable prognosis using a 31-gene expression profile (GEP) test. Manuscript in preparation.)

Combining patients from the original Vetto paper with the original 840 patients of the Novel Cohort originally presented at the Society of Melanoma Research, allowed for a combined cohort of 1905 patients with melanoma tumors of T1 or T2 thickness (<0.2mm). Within the group aged 65 and above who underwent a SLNB procedure (n=866), patients with DecisionDx-Melanoma Class 1A test results had a 2.7% (95%CI= 1.3-5.0%) SLNB positivity rate. This result falls below the >5% likelihood threshold for SLNB used as a minimum for consideration in national guidelines, and well below the 10% likelihood threshold at which national guidelines actively recommend SLNB.

For details of positive SLNB likelihood, as a function of age and stage, please see Table 2.

Table 2. SLN positivity in T1-T2 melanoma patients ≥ 65 years stratified by 31-GEP Class (Cook, Identification of melanoma patients with low risk of sentinel lymph node positivity and favorable prognosis using a 31-gene expression profile (GEP) test. Society of Melanoma Research, 2019).

My practice is largely skewed toward a Medicare-age population, for whom SLNB surgery provides additional expense and morbidity. Avoiding such interventions in these individuals, when the likelihood of a positive result is lower than the false negative rate (5%), should be a goal for all of us.

Having the molecular information provided by DecisionDx-Melanoma is critical to the decision-making process for my melanoma patients and provides valuable detail that is absent with clinicopathological assessment alone. There are numerous reports of the limitations of prognostic clinical features, including Breslow thickness and ulceration which are the two best established factors, and the only factors aside from SLNB that contribute to current staging of patients diagnosed with melanoma. CMS must do its due diligence in reviewing the evidence. No new data has been presented to counter the utility proposed by Dr. Vetto and our Novel Cohort. It is my hope you will continue to support this adjunct test which is both helpful to patients and cost-effective.

References and Tables were received and reviewed.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

As a matter of introduction, I am a practicing dermatopathologist with more than 20 years of experience in the field; I was also one of the pathologists in charge of several projects on ECOG (Eastern Cooperative Oncology Group) dealing with prognosis and survival of melanoma. I was also a long-time director of laboratory proficiency testing for American Society of Dermatopathology. I am writing to request that the current LCD for Decision Dx Melanoma continues to support existing coverage. Cutaneous melanoma treatments have markedly progressed in the recent years; however, most patients diagnosed without metastatic disease at onset, who later did die from melanoma, were initially assessed as low risk. To help address this clinical gap, both my colleagues in the practice and myself (Georgia Dermatology & Skin Cancer Center and Georgia Dermatopathology) use the decision Dx Melanoma test as part of our assessment of risk for sentinel lymphoid positivity. Published studies support the use of the test to identify patients with low, but definitive risk for sentinel lymphoid positivity. These patients have had excellent outcomes in long term studies, thus enabling patient to forgo sentinel lymphoid procedure and eliminate all the risks associated with it. Both my colleagues and I use this test to risk stratify patients and allow us to plan their follow up with further treatment in a way that is aligned with their risk of melanoma recurrence/metastasis. Growing body of published evidence supports routine use of the 31-GEP decision Dx Melanoma test, including two recent prospective studies (see below). This study shows that the test accurately and independently predicts risk of recurrence and distant metastasis for melanoma patients studied. My colleagues and I use the test result to establish follow up protocols including decisions on referrals and imaging that are appropriate and aligned for each patient’s individual risk. We are all confident that the Decision Dx Melanoma test offers information, which is not available otherwise or elsewhere, that can and does improve patient care. I strongly urge you to maintain the existing LCD and continue to provide coverage for Decision Dx Melanoma test.

References were received and reviewed.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

I am a board-certified pathologist with subspeciality board certification in dermatopathology and am involved in the diagnosis of patients with invasive cutaneous melanoma in my daily clinical practice. I wanted to submit a comment during the open comment period for MolDx: Melanoma Risk Stratification molecular testing. I believe that the coverage criteria for this LCD should remain unchanged and should continue to provide coverage for DecisionDx-Melanoma testing given that this gene expression profile (GEP) test has demonstrated clinical validity and clinical utility for patients with cutaneous melanoma.

GEP testing is an important part of melanoma patient care in my practice. I strive to provide the highest level of care when I interpret biopsies from my patients. In combination with other factors, I have found that GEP can be of great clinical use to the dermatologist in managing our shared patients. In addition to my belief that the test has clinical validity and clinical utility I have also been significantly involved in research to evaluate the performance and utility of GEP testing for cutaneous melanoma. Specifically, I participated in a study with Castle Biosciences to define the appropriate use criteria for using the 31-GEP test for risk of recurrence and metastasis.¹ As outlined in Marks et al, SKIN 2019, we found that the 31-GEP stratifies risk of recurrence for patients with tumors with Breslow thickness of 0.3mm and above. Patients with thin tumors below 0.3mm in thickness had a 5-year recurrence-free survival rate of 100%, while the 31-GEP significantly stratified risk of recurrence and distant metastasis for patients with tumors 0.3 - 1.0 mm thick (P < 0.0001). Patients with a Castle Class 2B result had a 75.5% 5-year recurrence free survival while those with a Class 1A result had a 96.7% recurrence-free survival. Furthermore, the paper described how clinicians have adjusted their management based on the results of the 31-GEP test for patients with T1 melanomas. Additionally, multiple clinical impact studies have shown that 25% of thin melanoma lesions had patient management changes based on the GEP test results. These changes included the frequency of clinical visits, imaging, referrals, laboratory tests and sentinel lymph node biopsy (SLNB) decision.

The dramatic reduction in Class 2B survival rates and the demonstration that the 31-GEP results impact patient management decisions has led me to recommend this test be used for cutaneous melanomas with certain criteria.^1-6 I believe that GEP test results can help improve the accuracy of risk prediction for recurrence and distant metastasis and that this information impacts clinician decision making around SLNB biopsy procedures and follow-up surveillance approaches.^7-9 For example, GEP testing can help guide clinicians with regard to the need for a SLNB for patients diagnosed with a T1a invasive melanoma with high risk features or a T1b melanoma. Subsequently, and independent from SLNB decision making, after clinical staging is complete, the GEP test results can inform the necessary frequency of office visits for low-risk Class 1A patients (decreased) or high-risk Class 2B patients (increased) with the possible addition of surveillance imaging in the latter case in order to find asymptomatic metastases at earlier time points.

Taken together, I recommend that my dermatology colleagues consider the use of DecisionDx-Melanoma for SLNB guidance in patients with T1-T2 melanomas over the age of 55 – including thin tumors with concerning features (T1a with high risk features) – and, independent of SLNB decision making, I advise the use of this test in patients with tumors of 0.3mm and above for determining who is more likely to have a recurrence or distant metastasis to help guide important patient management decisions. I believe that my dermatology colleagues would benefit from thoughtful conversations around the appropriate use of GEP testing for patients diagnosed with invasive cutaneous melanoma. Additionally, I discuss and facilitate ordering GEP testing for my colleagues when I believe the results will benefit the patients we care for as a team. After reviewing the published and peer reviewed data regarding test performance, clinicians can establish an appropriate use of GEP testing in their practices in order to improve the accuracy of risk prediction and inform a range of management decisions for patients with invasive cutaneous melanoma.¹⁰

References were received and reviewed.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

DecisionDx-Melanoma has become a valuable tool for my dermatology practice. We want to gather as much information as possible when making tough decisions about management strategies for patients diagnosed with melanoma, and we understand the limitations of the clinical factors that are either used for staging or not included as staging criteria but available for consideration (regression, lymphovascular invasion, etc). We have reviewed results from numerous studies (Greenhaw, Dermatologic Surgery; Hsueh, Journal of Hematology & Oncology; Keller, Cancer Medicine, etc) and the data clearly supports that the 31-GEP test is one of the most accurate tools available for risk assessment. This has been demonstrated in single and multi-center studies, in retrospective and prospective studies, and in studies sponsored or independent of Castle Biosciences. As a dermatologist who has seen many patients with thin, low-risk melanoma undergo the sentinel lymph node biopsy procedure, I have appreciated the information provided by the 31-GEP. We are comfortable managing patients with melanoma and following those patients if they choose to forego the sentinel lymph node procedure. DecisionDx-Melanoma allows us to identify those patients more accurately than any other prognostic factor. We support Medicare coverage of the DecisionDx-Melanoma test and the current LCD.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

As Professor of Surgery in the Division of Surgical Oncology at Oregon Health & Science University (OHSU) and the clinical Lead for the OHSU Multidisciplinary Melanoma Program I have extensive experience with the management of patients diagnosed with melanoma of all stages. Our multidisciplinary approach to the decisions about surgical intervention, frequency of follow-up and intensity of surveillance has been positively impacted by the addition of molecular information derived from the 31-GEP prognostic test^.1,2 As I described in recent presentations to Noridian, Palmetto and WPS (see attached presentation), the 31-GEP informs management decisions before and after sentinel lymph node biopsy (SNB) for i) patients with T1-T2 melanoma with a very low risk of a positive SNB, and ii) patients who were considered for SNB, but who either did not obtain the procedure, had a negative result (stage I-II) or had a positive result (stage III).^3,4

There is a significant clinical challenge for patients diagnosed with early-staged melanoma who are considering SNB. Current application of national guidelines to identify patients who should undergo the procedure results in SNB positivity rates of only 12%, with lower rates for patients with T1-T2 tumors less than 2mm in thickness, especially in patients of Medicare age ^{3, 5,6} As a result, there is ongoing debate in the surgical oncology community about the appropriateness of SNB for all patients with T1-T2 thickness tumors.^7,8 As shown in a study I led and published in Future Oncology, the 31-GEP further refines risk prediction before performance of the SNB, and can identify those patients with a risk of metastasis to the sentinel node that is below the threshold established by the NCCN (<5%) for considering the procedure.^3,9 We showed that patients who are over 65 years old and have tumors below 2mm in thickness and a Class 1A result have only a 1.6% risk of sentinel node metastasis, compared to Class 2B patients who have a risk of 11.9%.³ These results have led to a shift at OHSU away from considering or offering SNB to all eligible patients with T1-T4 melanoma. Instead, for patients who are >65 years of age with T1-T2 thickness tumors, we have begun making decisions using all available information, including the 31-GEP test result. We have recently opened an IRB-approved registry study (DECIDE) to track outcomes from this approach.

Expanding on the utility of the 31-GEP test, our study recently published in the American Journal of Surgery found that patients with stage I- II melanoma who underwent sentinel node biopsy and had Class 2 (high risk) 31-gene expression profile results had increased risk of recurrence compared to those with a Class 1 (low risk) result.² While the NCCN does not recommend routine surveillance imaging for patients with stage I-IIA disease, it is well known that patients with these stages comprise the majority of recurrences and deaths from melanoma. Based on the results of our study our group here at OHSU believes that patients with Class 2 tumor biology have a compelling additional risk of metastasis, and that routine surveillance imaging can be important for these patients. Further, in patients with a positive sentinel node, those that had a Class 2 result had the highest risk. As demonstrated by our data, patients with Class 1A/SNB negative profiles had only a 2% risk of recurrence, compared to 11%, 17% and 38% for patients with Class 1A/SNB positive, Class 2B/SNB negative, or Class 2B/SNB positive results, respectively. The resolution of risk provided by the 31-GEP for patients with a node-negative SNB is critical to improved patient care.

In summary, I have worked with my colleagues at OHSU to exhaustively evaluate the data supporting clinical use of the 31-GEP before consistently including it in our patient management workflows. Ultimately, we found it useful for identifying the risk of sentinel node metastasis, and for guiding the management of “double negative” (Class 1/SNB negative) patients who could be followed by the dermatology department alone, without need for follow-up surveillance by surgical oncologists.

The potential impact of this management shift at our institution has been published as a decision-tree analysis in the Journal of Drugs in Dermatology ¹ and is marked by a dramatic reduction of patient visits to my office since that time (results presented in Arnot et al. support that decision, showing that the 172 Class 1A patients included in our study had only a 3% risk of recurrence.² The test also provides important information when considering management of patients who have AJCC “low risk” disease who harbor a high risk according to their tumor biology, and a subset of stage IIIA patients who have a lower 5-year survival rate (and should therefore be more strongly considered for adjuvant therapy) than the more common good prognosis IIIA patient.¹⁰ Taken together, these results suggest that the 31-GEP adds prognostic accuracy to the evaluation of cutaneous melanoma from stage I to stage III, and should be included in the decision-making process for patients diagnosed with melanoma.

References were received and reviewed.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

I am a surgical oncologist in the Cincinnati area. I specialize in melanoma and breast cancer care and have incorporated tumor genomics for nearly a decade. Melanoma has been a disease without predictability until the development of a predictive assay. I use this assay to identify patients who are at high risk of nodal metastasis as well as recurrent disease.

Conversely, most patients are low risk and can have the discussion of avoiding sentinel node biopsy and who are at very low risk of recurrence. This is precious information and is greatly valued by patients and clinicians alike.

This assay saves money and lives; I highly recommend it to my patients.

I am writing to voice my support for the approved LCD detailing use of the DecisionDx-Melanoma to inform sentinel lymph node biopsy use and decisions about surveillance opportunities. My practice manages hundreds of patients each year who are diagnosed with melanoma and are subsequently put in a position where they must consider what the appropriate management decisions are. The information provided by the test has substantial utility for guiding those decisions and it would be a strike against patient care to lose access to it. The data is strong and consistent,^1–3 and studies that do not align with that consistency have been shown to be marked by poor methodology.^4,5 If these inconsistent studies are being considered in the evaluation of gene expression profiling tests, then I would ask the committee to also consider all supporting studies.^6–10 I look forward to continued use of the DecisionDx-Melanoma test for my patients with melanoma.

References were received and reviewed.

Thank you for your support for the current policy.

The following comment was submitted to Palmetto GBA:

I am a dermatologist and clinical researcher practicing in New York, New York. I appreciate the ability to use the Castle DecisionDx-Melanoma test for my Medicare patients with melanoma.

Using this test with patients with thin tumors helps me identify which patients I can follow up with myself and which ones I need to send to a high-risk clinic. For my Medicare patients in particular, who may be managing multiple medical appointments, it helps to have refined risk so my patients with higher risk results can make sure their follow up is not delayed.

I follow the algorithm in the Kwatra, 2020 article which has been very useful in my practice, and I support the current LCD.

Thank you for your support for the current policy.

The following comment was submitted to Palmetto GBA:

I am a clinical diagnostics consultant based in San Francisco, who has worked with >30 companies in the field over 20 years guiding them through evidence and product development choices, market entry strategies, funding and a diverse range of business development transactions. In my experience, clinical diagnostics should be developed in a way that their findings have a positive impact on physician and patient decision-making.

A company that develops diagnostics should engage with clinical and payer stakeholders to define the evidentiary requirements for their indication. Better clinical diagnostic data leads to the efficient delivery of health care. The path to market introduction requires diligent investment in the development of appropriate evidence for the intended use population and a quantification of the clinical and economic impact.

The sentinel lymph node biopsy procedure supports clinical staging. However, the procedure itself can be evaluated for clinical utility and performance. A higher age (≥65 years) is associated with high false negative biopsy rate (20%).¹ In other words, a nodal recurrence following negative sentinel lymph node biopsy for melanoma is possible to occur in 20% of this group of patients. Molecular diagnostics of the primary tumor provides an opportunity to reduce the false negative rate of this surgical procedure.

A diagnostic test with the right performance metrics and clinical data would allow a better triage of patients with melanoma and reduce the over prescription of the sentinel lymph node surgery.

I have reviewed the data and publications^2-6 for the Merlin/CP-GEP assay which has been designed to identify patients with primary cutaneous melanoma stages T1b, T2a & (T2b) who are at low risk for nodal metastasis to enable them to safely forgo sentinel lymph node biopsy.

Specifically, the large, combined data set for the CP-GEP test (multiple independent validation sets) is benchmarked against standard established clinicopathologic variables (AJCC 8th edition risk stratification model). Furthermore, there has been direct comparison between the discovery cohort and nomograms from the Melanoma Institute of Australia (MIA) and the Memorial Sloan Kettering Cancer Center (MSKCC) as there is clear variability among prediction outcomes of clinical risk prediction tools.^4,7 The CP-GEP data reflects the complete spectrum of the intended use population (rather than convenience samples). Importantly the signature was developed on primary diagnostic biopsy tissue (rather than reexcision material) and uses SLNb status at the time of the melanoma diagnosis as “clinical truth”, making the application of this development to clinical practice straightforward.

Of note, CP-GEP also achieves excellent SLNb reduction rates in patients 65 years and older (46.5% for patients diagnosed with T1-T2 melanoma) with the opportunity to reduce the use of an invasive nontherapeutic procedure, decrease the cost of care and focus healthcare resources to those patients that need them most.

References were received and reviewed.

Thank you for your comments and citations for review.

The following comment was submitted to Palmetto GBA:

I am writing in support of the DecisionDx-Melanoma prognostic gene expression profile test for cutaneous melanoma. I am a board certified dermatologist in a large practice diagnosing approximately 75 melanomas a year. I started using the test approximately 2 years ago and now incorporate it routinely as an integral part of further decision making and patient management. The DecsionDx-Melanoma has helped guide my decisions regarding oncologic referrals, sentinel lymph node biopsy, and disease surveillance. The test has helped guide the most appropriate use of health care dollars. For instance, patients older than 65 years old with a low-risk Decision-Dx-Melanoma result have a very low rate of sentinel lymph node positivity. This reduces the number of sentinel lymph node biopsies my patients need. I also learn from the low-risk Decision-Dx Melanoma test that these patients develop low rates of metastases allowing me to direct high cost imaging to appropriate patents.

Thank you for your comment.

The following comment was submitted to Palmetto GBA:

I am submitting a public comment during the open comment period for MolDx: Melanoma Risk Stratification molecular testing as I am a co-author on previous publications that examine the clinical validity of GEP testing in cutaneous melanoma. As a national expert in the diagnosis and management of cutaneous melanoma, active researcher in the clinical management of melanocytic neoplasms, Professor at the University of Pittsburgh Department of Dermatology and Director of the Pigmented lesions at the UPMC Melanoma program, I believe that my perspective is valuable to provide support for gene expression profiling (GEP) testing for cutaneous melanoma. It is important to submit my comments as I believe that the coverage criteria for MolDx: Melanoma Risk Stratification molecular testing should remain unchanged and should specifically continue to provide coverage for DecisionDx-Melanoma testing as this GEP test has demonstrated clinical validity to improve the accuracy of risk prediction for invasive cutaneous melanoma. (Gerami, CCR, 2015; Gastman, JAAD, 2019; Greenhaw, JAAD, 2020; Hsueh, JCOPO, 2021; Arnot, AJS, 2021). What is more, DecisionDx-Melanoma has currently applicable clinical utility at our UPMC melanoma clinic to both guide SLNB decision making and, independent from surgical decision making, the test helps triage patients to low vs. high intensity follow-up and surveillance regimens after clinical staging has been completed by combining GEP results with AJCC staging data. (Vetto, Future Oncology, 2019; Schuitevoerder, JDD, 2018)

Patients with stage I-II cutaneous melanoma have good overall outcomes, but many will experience a recurrence of disease. In our multidisciplinary practice at the University of Pittsburgh Medical Center, DecisionDx-Melanoma has provided valuable information to better identify patient risk for recurrence so that I can offer the appropriate management plan for each patient. A study I led (Ferris, JAAD, 2017) showed that 85% of the patients in the AJCC low-risk population (stage I-IIA) who died following a diagnosis of melanoma were correctly identified as high risk by DecisionDx-Melanoma. That test had higher sensitivity than AJCC for the endpoints of recurrence-free, distant metastasis-free and overall survival, and sensitivity increased further when combining GEP results with AJCC-determined risk. In addition, when we combined low-risk test results with low-risk AJCC and high-risk test results with high-risk AJCC, there was better stratification of risk than observed with each tool independently. The totality of evidence supporting the DecisionDx-Melanoma test supports that gene expression profile testing can add value in combination with AJCC staging factors to further improve risk prognosis (Gastman, JAAD, 2019; Greenhaw, JAAD, 2020).

Having worked on clinical studies with Castle in the past, I am excited by what they have done to provide patients with access to methods for accurate prognosis, and I have confidence in the data supporting the DecisionDx-Melanoma test. Improved identification of patients who will not benefit from sentinel lymph node biopsy procedures is critical. We have worked across disciplines at UPMC to optimize this utility of the test and, as a group, we have reviewed the data supporting the ability of the test to accurately identify patients at low risk for SLN positivity (Vetto, Future Oncology, 2019). Our colleagues in surgical oncology agree that those patients who are eligible for the SLN procedure can benefit from the test results. Patients, particularly those who are older than 55 years of age and have thinner tumors (T1-2), find the information they receive from DecisionDx-Melanoma to be helpful in making a decision to forgo SLNB in many cases, and we feel comfortable following a low intensity surveillance regimen, provided by me and my colleagues in dermatology, that does not include imaging or bloodwork. By comparison, a worse prognosis of Class 1B or higher results in a discussion with patients and consideration of the SLNB according to the guidelines provided by the National Comprehensive Cancer Network. Class 2B patients, in particular, are strongly considered for SLNB, given the high rate of positivity in that group. This information is particularly helpful in older patients for whom comorbid conditions may make surgery particularly risky.

It is important to highlight that we are currently utilizing the DecisionDx-Melanoma test at UPMC to guide important clinical decisions both immediately after a melanoma diagnosis in the dermatology clinic (eg: before SLNB has been performed) and then after complete clinical staging (eg: after clinical staging is completed with or without SLNB performed). The DecisionDx-Melanoma test is, therefore, used for guiding both the use of SLNB and longer-term surveillance for patients with melanoma. In so doing, I believe that we can work with our multidisciplinary team to both avoid unnecessary SLNB procedures while directing those patients to surgery who are most likely to benefit from the procedure, and, independent from SLNB decision making, more appropriately triage patient follow-up and surveillance through improved risk stratification by combining GEP test results with AJCC staging information.

As an author on the recently published Grossman et al manuscript (Grossman, JAMA Derm, 2020) it is important to emphasize that while the article can be accurately described as raising awareness of an ongoing academic debate, the article does not contain new data from tested patients and therefore I do not believe that it casts “doubt on the utility of GEP profiles for risk stratification” as stated in the draft LCD. Further, while the article advises against the routine use of GEP in melanoma it also clearly articulates a current appropriate clinical use for GEP testing. In the conclusion of the Grossman et al manuscript, the article states, “The MPWG recommends further research to assess the validity and clinical applicability of existing and emerging GEP tests. Decisions on performing GEP testing and patient management based on these results should only be made in the context of discussion of testing limitations with the patient or within a multidisciplinary group.” I believe that this statement both describes the value of continued ongoing and future research while also describing a clinical setting where GEP testing can be employed in current clinical practice outside of a clinical trial or study. In medicine, we should always evaluate our standard of care and research patient outcomes. I believe that after physicians have read and interpreted the published and peer reviewed data regarding the DecisionDx-Melanoma test performance and clinical utility, they are able to establish thoughtful workflows within their current practice patterns that combine GEP results with AJCC staging information to augment risk prediction, improve patient counseling conversations and inform a range of management decisions for their patients.

Again, I believe that the coverage criteria for this LCD should remain unchanged and should continue to provide coverage for DecisionDx-Melanoma testing given that this GEP test has demonstrated clinical validity and clinical utility for patients with cutaneous melanoma. Removing access to DecisionDx-Melanoma from providers will take away a tool that is used at UPMC to counsel patients on SLNB and follow-up/surveillance decisions.

Thank you for your comments and support of the current policy.

The following comment was submitted to Palmetto GBA:

We are contacting you on behalf of the Dermatology Contractor Advisory Committee (DermCAC), which represents the board-certified dermatologists in your carrier region. We appreciate the opportunity to share the dermatology perspective on the proposed Local Coverage Determination (LCD MolDX: Melanoma Risk Stratification Molecular Testing.

When appropriately used, melanoma risk stratification molecular testing will help to identify a specific group of patients that might have otherwise been subjected to sentinel lymph node (SLN) mapping and biopsy. Melanoma risk stratification molecular test can provide the information that we need to spare them of this procedure. Additionally, SLN biopsy is costly and can carry with it a significant amount of morbidity which is difficult for patients.

An additional benefit of this tool is that it will also provide useful information when making decisions about adjuvant therapies that might alter a patient’s chance of survival from melanoma. The ability to use this test—when needed—to spare our patients from unnecessary procedures and suffering in addition to its use with helping to guide therapy, adds a new level of critical information needed to guide dermatologists in our treatment of melanoma.

As such, the DermCAC agrees with and supports the proposed LCD MolDX: Melanoma Risk Stratification Molecular Testing.

Thank you for your comments.

I am submitting a public comment during the open comment period for MolDx: Melanoma Risk Stratification molecular testing as I am a board-certified Dermatologist, prior chairman of the Icahn School of Medicine at Mount Sinai Department of Dermatology and past president of the American Academy of Dermatology with clinical experience using gene expression profile testing for my patients with invasive melanoma.The 31-gene expression profile (31-GEP) prognostic test for cutaneous melanoma has been a critical tool that I use in my dermatology clinic to augment traditional AJCC staging. The test allows me to identify biologically high-risk patients so I may individualize their care and better allocate the health care resources available to best treat them. Additionally, the 31-GEP objectively reinforces those who are at low risk and are unlikely to experience a recurrence or metastasis. This is strongly represented across various studies with consistently significant negative predictive values. My conversations with truly low risk patients have dramatically improved as a result of implementing the 31-GEP into my melanoma patient care.

I have found there to be two primary utilities of the test in my decision-making process. First, the test helps me understand a patient’s overall risk of recurrence and metastasis. The test further stratifies risk beyond AJCC staging to offer a more comprehensive risk profile for patients with Stage I-III melanoma. This utility is supported consistently throughout the robust body of evidence behind the development and validation of the test going back to 2015. Secondly, is the ability of the test to better identify patients who may or may not benefit from sentinel lymph node biopsy (SLNB). There exists a nationally recognized threshold of eligibility for SLNB (5%) in melanoma patients that results in a very low yield. In using the 31-GEP, my multidisciplinary team and I are able to most appropriately select patients who undergo the SLNB procedure, as well as better tailor their care when they do not benefit from surgical intervention.

I recognize the inclusion of recent negative articles for review and urge the committee to consider the most up to date evidence provided to support the test, most notably the meta-analysis by Greenhaw et al, two recently published prospective studies by Arnot et al and Hsueh et al, and an expert clinical workflow by Kwatra et al.^1-4 These studies have fortified the years of science behind the validity of the test and how it continues to change my practice and those of my colleagues across the United States.

References were received and reviewed.

Thank you for your support of the policy and comments.

The following comment was submitted to Palmetto GBA:

I am writing in response to the request for comment on the draft LCD MolDX: Melanoma Risk Stratification Molecular Testing.

I am President of Colorado Melanoma Foundation. Our organization owns and operates The Sun Bus, a free community skin cancer education and screening program. As part of our program, we interact with melanoma and other skin cancer patients on a regular basis.

I would like to emphasize some of our findings as we fulfill our community mission. First, patients are in general very grateful to have testing options that tell them something about their likelihood of having a positive sentinal lymph node biopsy. In particular, those who are in the Castle Biosciences DecisionDX melanoma low risk group greatly benefit from not having to have a sentinal node biopsy and the possible sequelae that can result. For example, lymphedema can be a lifelong battle that can emerge from this frequently unnecessary procedure. In addition, we believe that knowledge of this test result gives patients a sense of relief and helps put their minds more at ease as they come to terms with their recent cancer diagnosis. Even those identified at high risk for sentinal node positivity using this test are able to benefit, since they become super diligent with their follow up surveillance visits and with their own at-home sun safety and surveillance activities once they learn of their result.

Second, as a melanoma clinical researcher, I strongly urge you to keep the existing LCD coverage as it is and unchanged. It is clear from gene expression profiles used in other cancer types that these are the way of the future and are increasingly important in predicting care levels for patients. Ultimately the proposed changes to LCD coverage will result in decreased patient care.

I am requesting the existing LCD coverage to remain the same.

Thank you for your comments and support of the current policy.

The following comment was submitted to Palmetto GBA:

In response to the open comment period posted for (MolDX: Melanoma Risk Stratification Molecular Testing), I would like to provide my perspective as a board-certified General Surgeon who specializes in melanoma with 6 years in practice on the clinical utility of this test beyond guidance of SLNB decisions alone.

The 31-GEP is a prognostic test that predicts a patient’s likelihood of recurrence within 5 years, including the likelihood of SLN positivity. Numerous articles have supported the accuracy of the test for predicting outcomes, adding value to both AJCC staging and SLNB to detect early-stage patients with a higher likelihood of recurrence. Patients with T1-T2 melanoma ages 55 years or older with Class 1A results have demonstrated a low likelihood of SLN positivity, under the 5% threshold established by national guidelines for consideration of the procedure. Thus, use of the test prior to SLNB identifies a substantial proportion of patients with low likelihood of positive results that have good survival outcomes.

Moreover, addition of the high-risk 31-GEP result add prognostic value to those who do undergo the SLNB procedure. When used with the SLNB procedure, the 31-GEP test detects an additional ~30% of patients who will experience distant metastasis than SLNB alone without sacrificing the positive predictive value (PPV) of a high-risk result. Notably, critics of the 31-GEP fail to appreciate the fact that this test adds information to current staging methods, demonstrating a fundamental misunderstanding of the utility of this highly valuable prognostic tool.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

The DecisionDx-Melanoma test has provided valuable information to our center for guiding patient management. In our academic surgical oncology practice at the University of Rochester Medical Center, we have reviewed supporting and critical data extensively^1–4 and, understanding that no diagnostic test achieves perfection, have included DecisionDx-Melanoma as a part of our patient workflow to be used in combination with clinical and pathologic features of the tumor. The value added when including test results with those features, including improved sensitivity for identifying patients likely to have or develop metastatic disease, is substantial. As guidance for the committee when reviewing the evidence, it is critical that the test results be evaluated in the context of staging factors rather than in comparison to staging factors. Sensitivity and negative predictive value are consistently improved across the majority of studies. Those that don’t report an increase in sensitivity and NPV are outliers that either did not measure those endpoints or had poor study design to achieve those results.^3,5

It is also important that the LCD review committee understand the value of experience with the DecisionDx-Melanoma test. Articles critical of GEP testing are most often led by authors who do not have that experience, which greatly impacts their ability to fairly evaluate the test. We have been encouraged by the recently published studies from centers like ours who have utilized the test in large cohorts of completely staged melanoma patients.^6,7 Again, the data is marked by consistency and improved identification of high-risk patients. If you evaluate the data from those papers closely, you can also see that they support the use of the test to identify patients who were at low risk for having positive SLNB results. Arnot et al., for example, reported that patients with DecisionDx-Melanoma Class 1A results, regardless of age and tumor thickness, had a 5.2% (9/172) risk of metastasis to the lymph node, while Class 2B patients had a nodal positivity rate of 28% (21/75). Our group sees this as further support of the article published by Dr. Vetto, and an opportunity for better targeted use of the SLNB. Given the results of the Multicenter Selective Lymphadenectomy Trial-1 that indicated no survival advantage for SLNB positive patients compared to negative patients,⁸ we expect that Class 1A patients who have a low risk of SLNB metastasis would not be adversely effected by a watch-and-wait management plan, and have implemented that process for our patients.

Thus, based on current data we recommend that Medicare continue to support the LCD for gene expression profiling for patients with melanoma.

References were received and reviewed.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

I am a Mohs surgeon and regularly treat patients with melanoma. The DecisionDx-Melanoma (31-GEP) test has been useful for determining risk of metastasis for my patients.

While use of the DecisionDx- Melanoma prognostic test in the context of AJCC staging is intuitive for me, I’ve read recent articles from others within the dermatology community (Grossman, Marchetti) who have asked for more informaton on the use of GEP tests in the context of other clinicopathologic features. Therefore, my colleagues and I wanted to scientifically demonstrate how to calculate the risk of metastatic disease recurrence using the 31-GEP test. So, we developed a study aimed at developing a metastasis risk calculator based on patients we have tested with the 31-GEP. Currently, we have an article under review at the Journal of the American Academy of Dermatology that describes this study in which we developed a nomogram to integrate the 31-GEP and T stage for personalized patient risk of metastatic recurrence.

This study demonstrates that the 31-GEP adds prognostic value to the current staging system and that the 31-GEP can be used in the context of tother factors that clinicians commonly assess. This study has strengthened my view that the DecisionDx-Melanoma GEP test is an important part of melanoma assessment for my practice.

Thank you for your comments.

The following comment was submitted to Palmetto GBA:

I wish to submit this public comment during the open comment period for MolDx: Melanoma Risk Stratification molecular testing. I believe this is of particular relevance to this issue because I am a co-author of the article recently added to the Draft LCD by Grossman et al.¹ I am also generally considered to be a national expert in the diagnosis and management of cutaneous melanoma, and am active in the field of cutaneous melanoma research, Professor and Chair of Oregon Health & Science University Department of Dermatology and Director of the Melanoma research program at the Knight Cancer Institute. I believe that my perspective is valuable in providing meaningful context to this recently published manuscript. I feel qualified to make an informed and meaningful contribution to your discussion.

Any change in the coverage criteria for MolDx: Melanoma Risk Stratification molecular testing should be carefully considered in light of these comments and the context in which the Grossman et al manuscript was prepared. Importantly, as has already been concluded by prior evaluation of the test, this GEP test has demonstrated clinical validity and has been shown to improve the accuracy of risk prediction for invasive cutaneous melanoma.^2-6 What is more, DecisionDx-Melanoma has applicable clinical utility and is currently used in our OHSU melanoma program to guide SLNB decision in equivocal or difficult cases, and helps triage patients to low vs. high intensity follow-up and surveillance regimens after clinical staging has been completed. We have found the combination of GEP and AJCC staging data to be of clinical value to our patients.^7,8

While the current Draft LCD includes a thoughtful discussion of the Grossman et al manuscript, it is essential to be aware of the context of the paper and understand potential unstated biases in the article. I urge you to carefully review the methodologic limitations, and I wish to provide you with an alternative and perhaps more balanced perspective on the impact of the article conclusions for the community of physicians who currently manage patients with cutaneous melanoma.

Substantial biases exist on BOTH sides – those critical and those supportive of GEP testing. These types of biases infiltrate and compromise the conclusions that experts draw regarding the clinical validity and utility of these technologies, including opinions published in the manuscript. As a “consensus” manuscript, co-authors are required to compromise in the interpretation of opinions (in both directions), but these compromises are never completely reflected. For example, during preparation of the manuscript, I recommended inclusion of an acknowledgement that “surgical oncologists may be biased against a test that could limit the use of SLNB” and that there was “a relatively low combined response rate to both surveys,” but these acknowledgements were not incorporated into the final version of the paper. I raised other potential biases that were not as clearly articulated as I would have liked, including the fact that non-academic experts with alternative opinions were not included in the open discussion and Delphi process. Certainly, economic influences can lead to conscious and unconscious biases, and these must be mitigated. However, academic biases also exist, particularly with respect to the need to appear “objective and appropriately critical” of new technologies. Academic experts also frequently have competing technologies of interest that may not yet require disclosure. The purpose of a Delphi process is to include all aspects of expert opinion, from every side and angle being considered, and to utilize anonymous surveys interposed with discourse to minimize undue influence from dominant “voices in the room.” The modified Delphi process used in this paper was not, in my opinion, completely successful in achieving this goal, despite our best intentions.

I have strong concerns that you are allowing the Grossman paper to be an overly influential consideration in your decision regarding the GEP test. Unlike the numerous studies that contribute real, peer-reviewed data for consideration, the Grossman paper is just expert opinion. Further, the experts included in the modified Delphi process did not include all stakeholders and excluded the members of the pharmaceutical and non-academic communities that might have had valuable and differing opinions. The group meeting that was held to discuss the results of the initial Delphi survey also excluded those voices. Further, the discourse was arguably dominated by a group of individuals with strong personalities, potential bias against a technology that would reduce sentinel lymph node biopsy, and a “power gradient” (since several of these individuals occupy leadership positions in national organizations). In addition, some of the subsequent “consensus” opinions of the authors were drafted and edited in identifiable email correspondence where participating authors did not have anonymous voice; other controversial conversations were had “behind the scenes” and not made available in their entirety to all authors until the near-final manuscript was circulated. Finally, even in this group of relatively like-minded experts, the “consensus” was defined as 50-60%, and only barely exceeded 50% in several of the contentious areas (i.e., at least 40% of the participants in the conversation felt differently to the published “consensus” statements included in the article). Thus, our expert opinion might have been better characterized as “majority” rather than “consensus” opinion without support of GEP testing as a rhetorical counterbalance.

I would advocate that you use this expert manuscript as one important, but also limited, factor in deciding the future of the GEP test. It should not be stressed above the peer-reviewed data published in numerous prestigious journals. It should be regarded for what it is, a summary of the opinions of a limited number of experts about a newly available test. Instead of describing the article as “casting doubt on the utility of gene expression profiles (GEPs) for risk stratification” as is currently written in the Draft LCD, it would be more accurate to state that this article highlights the ongoing debate about the role for GEP testing in the management of patients with cutaneous melanoma. It is important to emphasize that the Grossman et al manuscript argues against the routine use of GEP in melanoma and in fact presents a clear path forward for the current appropriate clinical use for GEP testing. Castle Biosciences does not endorse the routine testing of every invasive melanoma and has set limits about the thickness of lesions that are acceptable for testing and demonstrates the clinical utility of the test results. In the conclusion of the Grossman et al manuscript, the article states, “The MPWG recommends further research to assess the validity and clinical applicability of existing and emerging GEP tests. Decisions on performing GEP testing and patient management based on these results should only be made in the context of discussion of testing limitations with the patient or within a multidisciplinary group.” I believe that this statement describes the value of continued research that is already ongoing, and that can be undertaken in the future, and also describes a clinical setting where GEP testing can be employed in clinical practice now outside of a clinical trial or study.

While OHSU is currently participating in multiple enrolling prospective trials, there are also established clinical workflows for GEP testing that align with the concluding statement from the Grossman et al article.⁸ As with all clinical testing, we advocate for a careful discussion between physician and patient regarding the benefits and limitations of GEP testing and stress the importance of interpreting results in the context of clinicopathologic factors. Furthermore, GEP testing takes place in our department within a multidisciplinary group setting to inform a range of clinical management decisions. These clinical utilities include: SLNB decision making to avoid unnecessary SLNB procedures and advocate for SLNB procedures for those most likely to benefit from the surgical intervention and, separately from SLNB decision making, informing the intensity of follow-up visits and surveillance for patients after clinical staging is complete. This use of the multidisciplinary team is vital to providing patients with the highest standard of care; it is not uncommon for prospective clinical trials to lag behind implementation of useful, novel tests and technologies. For example, while the therapy prediction value of Oncotype DX was ultimately confirmed in the TAILORx prospective study, the test became standard of care based on earlier archival studies.^9-10 As a result, patients benefited from GEP testing in breast cancer for almost 15 years while awaiting the prospective study results.

As an important adjunct to the Grossman et al manuscript, I was also recently involved in an expert opinion paper, written by clinicians with experience using GEP tests in the management of their patients, that contains a clinical workflow for GEP testing that can be harmonized with current AJCC staging and risk appropriate management decisions.¹¹ The management decisions include data-based patient reassurance, informing SLNB decision making, adjustments in the frequency of follow-up visits, consideration of baseline imaging, adjusting thresholds for symptomatic imaging, and referrals for multidisciplinary follow-up. This manuscript is influenced by the authors’ belief that Patient-Centered outcomes (eg: those supported by the Patient-Centered Outcomes Research Institute - PCORI) are also valuable health outcomes of interest for our patients with melanoma. The clinical workflows described therefore incorporate management decisions that can improve risk stratification in combination with AJCC staging and that also focus on improving patient quality of life metrics by utilizing GEP testing in appropriate patient groups that benefit from more accurate risk prediction than what is provided by an AJCC staging approach alone.

Fundamentally, I use GEP testing when I believe that the test result will benefit my patients. I believe that physicians equipped with published and peer reviewed data regarding test performance are able to establish careful approaches within their current practice to incorporate DecisionDx-Melanoma to inform important clinical decisions for patients with cutaneous melanoma. It is a dangerous precedent to allow a consensus paper of limited experts to obstruct the forward progress of a useful test – the statement that the Grossman “consensus” casts doubt on the utility of GEP testing is an overstatement. While it is important to acknowledge the active academic debate around GEP testing in melanoma, the LCD should also reference alternative expert opinion papers that reach different conclusions and provide clear clinically actionable workflows for how GEP can be currently incorporated into clinical practice for the benefit of patients.

References were received and reviewed.

Thank you for your comments. Of note, the Grossman et al paper was taken as face value, we recognize it as expert opinion and have identified the liabilities in the survey. That said, we feel it is important when considering evidence (that in the case of coverage policy often includes expert opinion) to ensure that dissenting opinion is presented in controversial topics, even if the weight or totality of that evidence is not equal. In this case, we could not ignore a publication that was published as a consensus paper, especially since the findings were critical of the tests covered in this policy (although the language was nuanced). That stated, the draft and subsequent coverage policy did not alter coverage for these tests.

The following comment was submitted to Palmetto GBA:

As a practicing board-certified surgical oncologist with 20 years of clinical experience, I can attest to the need to improve the identification of patients with an increased risk of metastasis across the spectrum of AJCC staging. While a sentinel node biopsy (SNB) is useful in identifying some of these patients, the clinical scenario remains that the majority of SNBs yield negative results and within this population subsist two-thirds of those diagnosed with regional disease that will ultimately die from melanoma. DecisionDx-Melanoma is used for two purposes in our clinic: first, to identify patients at low risk for a positive node that can avoid SNB; and second, to further stratify metastatic risk to guide risk-appropriate follow-up and surveillance in concordance with NCCN guidelines.

Patients with Stage I melanoma (T1a-T2a with no regional disease detected) represent an important clinical challenge for appropriate management. The nodal positivity rate of these patients as a population hover around the 5% threshold set forth by national guidelines for consideration of the procedure, demonstrating the need to more accurately identify the risk of SLN positivity for an individual patient. DecisionDx-Melanoma test results in combination with T stage and age improve the prediction of a patient’s risk of positive node such that patients 55 years and older with a Class 1A result have a risk of positivity less than 5% (Vetto 2019). In combination with a high NPV (~98% for MSS) and good survival outcomes, these patients can safely avoid SNB. As demonstrated in a recent meta-analysis, patients with Stage I disease and a Class 2 result have significantly higher rates of recurrence and distant metastasis, indicating the test results are also useful in this population for guiding management decisions for increased surveillance and follow-up (Greenhaw 2020). Clinical utility studies published to date have demonstrated that experts understand how to incorporate the test results with AJCC staging and NCCN guidelines to guide management decisions (Kwatra 2020), that physicians find clinical value in both low- and high-risk results (Marson 2021), and that the test result influences patient management decisions in 1 out of 2 patients tested (Berger 2016; Schuitevoerder 2018; Dillon 2018). Importantly, even in patients with thin melanomas less than 1 mm, 25% of patients had clinical management changes based on the DecisionDx-Melanoma test result (Marks 2019). As such, claims of reduced utility in this population of patients is unfounded, and most likely are a result of inexperience with the clinical use of the test. Data supporting the accuracy and utility of the test far outweighs articles of a critical nature, and the current LCD should not be altered.

Thank you for your comment.