Response to Comments: MolDX: Prostate Cancer Genomic Classifier Assay for Men with Localized Disease

The following comment was submitted to Palmetto GBA and Noridian:

We are in support of the policy and respectfully submit the following requested changes for the purpose of clarification:

• Change the title from “MolDX: Decipher^® Biopsy Prostate Cancer Classifier Assay for Men with Favorable Intermediate Risk Disease” to “MolDX: Decipher^® Prostate Genomic Classifier for Men with Prostate Cancer”
• In the section titled “Coverage Indications, Limitations, and/or Medical Necessity”:
  • In the first paragraph, replace “Decipher^® Biopsy Prostate Cancer Classifier Assay” with “Decipher^® Prostate Genomic Classifier for Men with Prostate Cancer (Herein referred to as “Decipher Prostate”)”
    • In the remainder of the document, update references to the test as “Decipher Prostate”
  • In the first paragraph, replace “identify men with localized Favorable Intermediate Risk Disease Prostate Cancer and a life expectancy of at least 10 years who are good candidates for active surveillance” with “guide treatment decisions for men who meet the coverage criteria below”
  • In the second paragraph, replace “indication” with “indications”
  • In the third paragraph, replace “a man” with “a man diagnosed”
  • In the third paragraph, replace “who have a life expectancy” with “who has a life expectancy”
  • In the fourth to last bullet, replace “radical prostatectomy specimen” with “resected specimens”
  • In the penultimate bullet, remove “or radical prostatectomy”
  • In the final paragraph, replace “Additionally, a similar transcriptome-based test” with “Additionally, a similar whole-transcriptome-based test”
• In the section titled “Background”, replace “Prostate cancer is a heterogeneous disease, which to better risk stratify this patient cohort was the creation of favorable and unfavorable intermediate risk disease groups developed by Zumsteg and Spratt at Memorial Sloan Kettering, now adopted by NCCN guidelines” with “Prostate cancer is a heterogeneous disease, and physicians have historically utilized clinical and pathological features to classify patients into risk categories”
• In the section titled “Test Description”, replace “Decipher^® Biopsy Prostate Cancer Classifier Assay” with “Decipher^® Prostate”
• In the section titled “Test Description”, replace “Decipher Biopsy^®” with “Decipher Prostate”
• In the section titled “Test Description”, replace “, measuring the expression of over 1.4 million RNAs (from coding and non-coding genes). The assay is performed on formalin-fixed paraffin embedded FFPE) prostate cancer tumor tissue from diagnostic biopsy needle cores” with “that is performed on formalin-fixed paraffin embedded (FFPE) prostate tissue from biopsy or resected specimens”
• In the section titled “Test Performance”, replace the entire body of the section with:

Decipher has been examined in over 20,000 patients including more than 40 completed clinical validity and utility trials, 2,900 unique patients with long-term follow up, and treatment and outcomes data from a variety of patient populations and practice settings.

In the initial treatment setting, Decipher Prostate reports an overall GC score and risk estimates of metastasis, prostate-cancer specific mortality (PCSM), and high-grade disease at RP. It provides information to help guide decision-making for men considering active surveillance, and definitive therapy.

In the post-RP setting, Decipher Prostate reports an overall GC score and risk estimates of metastasis 28 and PCSM. The results help guide decision-making regarding choice and timing of treatment after RP, and treatment intensification.

As assessed by the area under the ROC curve (“AUC” or c-index), the prognostic accuracy of Decipher at biopsy for each of the three endpoints, and at RP for both endpoints, has been shown to be consistently higher than that of clinical and pathological risk factors alone. Decipher risk is objectively derived from tumor biology alone, has strong association with prostate cancer-specific outcomes, and its use leads to significant re-classification of clinically determined patient risk. Decipher has also demonstrated clinical utility in its ability to change patient management and improve patient outcomes.

Tables 2 and 3 below demonstrate the analytical and clinical performance of Decipher Prostate:

Table 2: Decipher Prostate Analytical Performance

Table 3: Decipher Prostate Clinical Performance

MVA-multivariable analysis; CI-confidence interval; HR-hazard ratio; OR-odds ratio; ref-reference group; BCF-biochemical failure; PCSM-prostate cancer specific mortality; NR-not reported; CGRG-Clinical Genomic Risk Groups.

a-Hazard ratios or odds ratios of continuous Decipher score were reported per 10% increase in score

b- Primary Gleason pattern 4 or 5, pT3b or higher

c- lymph node involvement (PSMA-avid pelvic nodal)

d- Grade group 3-5, pT3b or higher, or LNI

e- Primary pattern 4 or 5, pT3b or greater, or LNI

• There are a number of studies relevant to the clinical validity and utility of Decipher Prostate that we believe should be included in the LCD. In the section titled “Bibliography”, replace the entire body of the section with the references submitted with this comment.

Thank you for your comments and support of the policy. We will not be able to rewrite the entire policy as you have suggested as this would delay bringing the policy to fruition. The policy has been modified in areas which resolves the issues that you have brought to the attention of MolDX in our communications.

The following comment was submitted to Palmetto GBA:

I am submitting this letter regarding MolDX: Decipher Biopsy Prostate Cancer Classifier Assay for Men with Favorable Intermediate Risk Disease (DL38292).

I am the founder and medical director of TruCore Pathology, PA. We work with approximately 90 urologists from around the country in the diagnosis of prostate and bladder cancer specifically. I personally review approximately 5000 unique prostate biopsies a year or nearly 60,000 individual prostate cores. In helping to ensure that I and my colleagues provide the absolute best information possible, we have adopted the routine use of genomic classifiers or genomic biomarkers. The biomarker space has become an increasingly confusing one to our urologist brethren. Many of these tests have different tumor requirements that are only well known to the pathologist. We routinely attempt to guide the urologist in a biomarker direction based on tumor composition and have thereby attained the nation’s lowest cancellation rate while providing drastically reduced turnaround times (critically important for decision making). While we retain our neutrality amongst the tests, we are uniquely positioned to help provide the most accurate information. That being said, we pride ourselves on staying up to date with the latest research. This is why I’m concerned over the recent LCD submission regarding the Decipher biopsy test. By definition, the Decipher test looks at 22 unique genes and is an excellent predictor of necessity of early adjuvant radiation versus salvage radiation following treatment with radical prostatectomy. The Decipher biopsy test referenced in the submitted LCD is not however, as suggested, a whole transcriptome assay, but only a measure of the 22 gene panel. The whole transcriptome assay, to which they refer, reflects what they call the “GRID” which is intended for research purposes only (as their clinical report mentions). Within the GRID is a Cuzick score which is intended to give a proliferation index on how fast the tumor cells are programmed to divide, attempting to predict tumor behavior. I have personally compared their GRID Cuzick scores to other validated rtPCR (gold standard for Cell Cycle Proliferation scores (CCP)) and found it to be lacking. While the information portrayed on the GRID is potentially exciting, it remains, to their clear delineation, for research purposes only and is not validated for clinical use. The actual Decipher biopsy test does offer some significant clinical data that can be useful. These nuances are however often lost on the community urologist who is seeing a clinic full of patients, running off to surgery and then being handed a report with what appears to have everything provided on the GRID.

The risk stratification conundrum is inherently tricky. Most urologists weigh the extremely subjective Gleason grading heaviest when deciding whether AS is an appropriate option. That’s based on years of correlating Gleason grading with tumor behavior and until recently was the best we had. There are several other biomarkers with equivalent or superior validation studies that have been published in peer-reviewed journals. When reviewing Deciphers appeal, I urge the committee to focus purely on the 22 gene test and its validations instead of being distracted by a whole genome transcriptome that has no validation. Additionally, I would strongly support the use of any biomarker that provides equivalent or better clinical information to the Decipher biopsy test and would recommend their inclusion for approved utilization as a class.

Thank you for your comments and support of the policy. We have edited the test description to reflect your suggestion also identifying the 22 gene panel.

The following comment was submitted to Palmetto GBA:

Myriad Genetics, Inc. appreciates the opportunity to comment on the proposed LCD titled “MolDX: Decipher^® Biopsy Prostate Cancer Classifier Assay for Men with Favorable Intermediate Risk Disease (DL38292)”.

Given the uncertainty regarding treatment choices within current guidelines for men with prostate cancer, use of a test that provides more accurate prognostic information compared to existing clinical and pathological features allows for improved treatment decisions. We recognize the significant efforts of the MolDX team in reviewing the supporting evidence for several prostate cancer prognostic assays and providing Medicare beneficiaries with access to these tests. Based upon our review of DL38292, we share the following comments and suggested revisions for your consideration.

1) Description of the Decipher test

The draft LCD describes the Decipher biopsy test as “a clinical-grade whole-transcriptome assay, measuring the expression of over 1.4 million RNAs (from coding and non-coding genes)”. This description differs from the description in each of the validation studies referenced in the draft LCD, in which the Decipher prostate test is consistently described as a 22-gene genomic classifier. While the development of the Decipher prostate test used a whole transcriptome approach through which 22-genes were eventually selected, it is inaccurate to describe the commercially available test that is used in clinical practice as a “whole-transcriptome assay”. In fact, the Decipher prostate biopsy sample report available on the company’s website describes the test as using “oligonucleotide microarrays to measure 22 RNA expression biomarkers, extracted from formalin fixed paraffin embedded (FFPE) prostate biopsy specimens”.

The “1.4 million RNAs” refers to the Decipher Genomic Resource Information Database (GRID) Profile, which is not interchangeable with the Decipher prostate biopsy test. The Decipher Prostate requisition form includes a checkbox stating “Check here if the patient has authorized you to receive their GRID Profile, which is Research Use Only and not to be used for treatment decisions or medical management”. Additional information on the reverse side of the requisition form states “The Decipher assay collects up to 1.4 million data points for each patient when the Decipher test is performed, and such data is used to create the GRID Tumor RNA Expression Profile (GRID Profile). The GRID Profile is a summation of genomic signatures and biomarkers for Research Use Only.”

In summary, the current language describing the nature of the Decipher prostate biopsy test is confusing and not consistent with the published clinical validation studies cited in the draft LCD. If the intent of the draft LCD is to provide coverage of a clinically validated test used for medical management and not to cover a ‘research use only’ test, then the Decipher test should be described as a 22-gene genomic classifier and not as a whole transcriptome assay.

2) Coverage of similar tests

In the “Coverage Indications, Limitations, and/or Medical Necessity” section, the draft LCD states that “a similar transcriptome-based test with analytical and clinical validity at least as good as Decipher will be considered reasonable and necessary for the same indications”. As described above, the clinically available Decipher Prostate test is a 22-gene genomic classifier. Therefore, coverage of “similar” tests under this framework should not be limited to transcriptome-based tests. Rather, coverage should be considered for any genomic-based prognostic test for prostate cancer with comparable AV and CV data, e.g. Prolaris.

Prolaris is a prognostic test that measures the expression of 31 cell-cycle progression (CCP) genes in prostate cancer tissue. The 31 genes were selected through a process in which the Gene Expression Omnibus database was queried, and cancer outcomes were associated with transcriptome-wide gene expression patterns. From this analysis a set of 126 CCP genes were identified that could be robustly associated with adverse outcomes. Finally, 31 genes were ultimately chosen based on their correlation with the mean expression of the entire set. A validated combination of the CCP score and clinical features generates a combined clinical risk (CCR) estimate of prostate cancer mortality and metastatic disease within 10-years of diagnosis.

The accompanying table provides a summary of key validation studies utilizing the Prolaris test across a variety of different clinical scenarios, in a format analogous to the table included in DL38292. This compilation of studies demonstrates unequivocally that the Prolaris test augments the prognostic ability of clinicopathologic features alone for important oncologic endpoints across the entirety of the prostate cancer risk strata. Unlike other prostate cancer genomic classifiers, the Prolaris test does not focus on prediction of adverse pathology as a solitary oncologic endpoint. Prolaris is also the only prognostic biomarker that has been extensively validated in cohorts of conservatively managed men who did not receive initial treatment with progression of disease specific mortality as the ultimate endpoint.

In all these studies, multivariate analyses show that the Prolaris test provides significant prognostic information about disease outcome that cannot be captured by clinicopathologic features. Prolaris has also been validated to predict progression to metastatic disease (and BCR) after curative intervention in men of all disease risk groups. These studies include men who were treated with radical prostatectomy and primary radiation. As with conservatively managed men, the test provided highly significant prognostic information even after accounting for clinical variables. The power of the Prolaris test is additionally enhanced in that it is the only prognostic biomarker that has been validated in tissue from diagnostic biopsy, Transurethral Resections of the Prostate (TURP), prostate biopsy specimens, and tissue from radical prostatectomy (RP) samples. Finally, it should be noted that the most sensitive way to test for an improvement in prediction is a standard hypothesis test for the risk factor, which is how Prolaris was consistently evaluated (as presented by the HR and p-values in the attached table). AUC is widely considered to be less sensitive, because it does not adequately adjust for variability in estimated regression coefficients.

In summary, based on the evidentiary comparison presented in the attached table, the analytical and clinical validation data for Prolaris are equivalent or superior to the test for which coverage is proposed under DL38292. Prolaris provides more accurate information about clinically relevant oncologic endpoints in comparison to clinicopathologic features, which allows for better informed treatment decisions that impact patient outcomes. Therefore, we request that the coverage described in the draft LCD for use of a validated prognostic test across all localized prostate cancer risk groups be extended to Prolaris.

3) Clarification of LCD Title

Although the title of draft LCD and the first paragraph under “Coverage Indications, Limitations, and/or Medical Necessity” refer to “Men with Favorable Intermediate Risk Disease”, the remainder of the document and the body of evidence delivered appears to address all localized prostate cancer risk categories. This approach would be consistent with the MolDX team’s public comments and recent efforts to consolidate LCDs by disease states, which led to our current request for revisions to this draft LCD to include Prolaris (as opposed to requesting changes to existing Prolaris LCDs that are specific to certain risk categories). Finally, the discrepancy between the LCD title and content will likely raise questions among physicians and laboratory providers who rely on the LCD to determine which patients are expected to have coverage for specified services. Therefore, we recommend revising the title of the LCD to “Prostate Cancer Genomic Classifier Assays for Men with Localized Disease”.

Thank you for your comments and support of the policy. We have modified the test description based on your suggestion. There is language covering the Myriad Prolaris test in the current policy as we are trying to write policies to be less provider specific.