Summary Of Evidence

Cutaneous melanoma (CM) is increasing in incidence in the U.S., with an estimated 96,480 cases expected to be diagnosed in 2019 with 7,230 deaths.¹

In the treatment of CM, the risk that a patient has or will develop metastatic disease is central to many of the decision management choices in cutaneous melanoma, with more aggressive management or treatment strategies recommended for patients who are at a higher risk.² Per current national guidelines, a SLNB procedure is considered for all patients with melanoma pathologic Stage T1b and above, as well as those patients with T1a tumors in whom there is significant uncertainty about the adequacy of microstaging.^2,3 Patients with a positive SLN are at substantially increased risk for distant metastatic disease and death;⁴ however, the procedure only provides prognostic information, and the MSLT-II study showed no survival benefit associated with completion lymphadenectomy in SLN positive patients.⁵ Currently, the identification of SLN positive patients helps identify Stage III patients who can potentially benefit from targeted and immunotherapeutic agents in the adjuvant setting.^6-8 The procedure can be associated with complications in a substantial proportion of patients such as pain, seromas, nerve damage and edema, and requires a large team of dedicated personnel, including nuclear medicine physicians, surgeons, and pathologists.^9-11 It has been estimated that the cost of a SLNB can be 10 times that of a wide excision alone, and the cost per life saved in a patient population with low prevalence of positive SLN can approach 1 million dollars.¹² Overall, the likelihood of a positive SLN after the SLNB procedure is 16%,^4,13 but this is variable for specific populations.^13-16 Elderly patients account for a substantial proportion of CM patients, and 60% of melanoma-related deaths occur in patients ≥65 years-old. While older age is associated with a poor prognosis, fewer elderly patients are SLN positive,^14-18 which indicates that the prognostic value of SLNB is limited in this population.^19,20 In general, a 5% likelihood for a positive SLN is recommended as a threshold for performing this procedure in a patient population.¹³

Molecular diagnostic tests have been proposed to help managing clinicians risk stratify patients for selecting their most appropriate management based on their probability of developing metastatic disease; these tests may score patients’ probabilities of resultant metastatic disease by measuring tumor biomarkers such as relevant gene expression.^21-24 One gene expression profile (GEP) test (DecisionDx Melanoma, Castle Biosciences) was evaluated in a retrospective cohort (n=782) to evaluate its ability to predict metastasis and ability to predict SLNB status with tumors with a Breslow thickness <2.0 mm (American Joint Committee on Cancer [AJCC] T1 T2).²⁵

The ability of the test to identify a low-risk group was initially assessed and compared to SLNB in 2 contemporary, multi-center, prospective study cohorts: a 584 patient cohort from 2 published prospective studies (overall 14% SLN positive rate)^23,26 and a 837 patient cohort from prospectively tested patients at 5 large academic institutions (overall 12% SLN positive rate).²⁵ The rate of SLN positivity in both prospective study cohorts aligns with the SLN positivity rate in the general population of melanoma patients who have undergone SLNB. The results show that in patients from the Medicare-eligible population (65 years old and over) who were determined to be low-risk by this test, the concordance of a negative SLNB was 98.4%. These studies showed improved performance in other patient groups as well.

SLNB positivity rates for T3 tumors with low-risk score for this test is 8.7%. Importantly, the 5-year melanoma specific survival (MSS) rate for T1/T2 low-risk group remains favorable; with 99% MSS, comparable to that observed in T1a tumors and for which current guidelines do not recommend SLNB.^2,27 Furthermore, T1/T2 low-risk patients show 5-year overall survival (OS) of 97% and distant metastasis free survival (DMFS) of 93%.²⁸ The MSLT-II study demonstrated that a delay in lymph node dissection does not adversely affect survival, thus clinical follow up of low-risk patients and lymphadenectomy for those few who develop clinically detectable nodal disease should achieve similar outcomes to those who currently undergo a planned SLNB.⁵ Thus, the test identifies a patient population with <5% likelihood of a positive SLN and high survival rates and therefore, has utility in guiding SLNB decisions in patients 65 years-old and over with T1-T2 CM tumors. In this population, the test could potentially reduce the rate of SLNB by up to 78% while still maintaining an MSS survival rate of 99% in those patients with low-risk tumor biology who can safely avoid the procedure.

Clinical validation of this same test as a prognostic test for CM patients was performed in 3 multicenter, prospectively designed archival tissue studies including 782 patients.^21,22,24 These studies have shown that the test accurately predicts risk for local/regional recurrence, distant metastasis, melanoma-related mortality, and all-cause mortality independent of clinicopathologic factors used in staging and that the test shows improved sensitivity and negative predictive value (NPV) for recurrence-free (RFS), distant metastasis-free (DMFS), melanoma-specific (MSS) and overall survival (OS) individually or in conjunction with established clinicopathologic factors. A study focused on patients with melanoma of the head and neck (H&N) demonstrated that patients with H&N melanoma have poorer outcomes and lower rates of SLN positivity which makes the prognostic value of the SLNB procedure limited and thus additional prognostic information provided by this test is important in this group of patients.²⁸ Four prospective, independent studies (n=510) and an interim analysis of 2 prospective registries (n=322) have confirmed the prognostic accuracy of the assay.^23,29-32 This suggests that the test may be useful in stratifying risk and patient treatment decisions.

A retrospective study directly compared the ability of DecisionDX Melanoma with the ability of AJCC melanoma staging to predict longer term outcomes.³³ This study looked at 205 archived formalin-fixed, paraffin-embedded primary melanoma tissue blocks from 6 centers. This included 109 Stage I and 96 Stage II cancers, and median time to follow-up was 6.9 years. The median time to recurrence was 1.7 years, and the median time to distant metastasis was 1.6 years. In general, the test scoring alone had greater sensitivity to recurrence, distant metastasis, and death than AJCC staging. Alternatively, AJCC staging had greater specificity for these outcomes. The use of this assay with AJCC staging had a higher sensitivity to these adverse outcomes than either prognostic measurement alone. Studies have also suggested that clinicians value and use this test in their medical decision making regarding aggressiveness of melanoma management.^34-36

Recent publications have provided further evaluation of these types of genetic risk stratification tests. Some studies and meta-analyses evaluating a host of clinical studies have cast doubt on the utility of gene expression profiles (GEPs) for risk stratification.^37-38 Grossman et al³⁷ described the findings of the National Melanoma Prevention Working Group, which evaluated 3 prognostic GEP tests. This included a review of online surveys, journal articles from 2015-2019, and relevant abstracts. One of the findings from the group was that it was difficult, due to a lack of clinical trial data reviewed, to clearly separate known clinical pathological factors that are associated with risk of recurrence/metastases from the GEP findings, stating that while some such factors were considered, not all such factors were considered in clinical studies that support GEP use. Additionally, it was noted that some of the staging used in the favorable studies were not using current AJCC8 staging criteria. Additionally, Marchetti et al³⁸ performed a systematic review and meta-analysis of 7 GEP studies, including 5 that evaluated DecisionDx-Melanoma and 2 studies evaluating MelaGenix. They evaluated patients in these studies, who were characterized as high-risk or low-risk, dependent on staging, with recurrence (primary endpoint) or metastases (secondary endpoint). They determined that among patients with recurrence, DecisionDx-Melanoma correctly classified 29% of stage I patients and 82% of stage II as high-risk. Among patients without recurrence, this test identified 90% of stage I patients and 44% of stage II correctly as low-risk. The Melagenix test classified 32% of stage I and 76% of stage 2 patients with recurrence correctly as high-risk; and 77% of stage I and 43% of stage II patients without recurrence correctly as low-risk. The data evaluating metastasis as an endpoint was considered poor and not significantly evaluated.

Other studies, including meta-analyses, maintained and strengthened the evidence for the use of GEP tests. One example is Greenhaw et al⁴² that evaluated three studies that met criteria and analyzed raw data for a systematic review of 1479 patients. This study demonstrated that the Melanoma DecisionDX test was an independent predictor of both recurrence and distant metastasis-free survival based on multivariate analysis (including Breslow thickness, ulceration, age, and SLNB status; with the GEP having the highest Hazard Ratio of 2.9). The NPV of the GEP test was 92% (90-94% CI) for recurrence-free survival and 93% (91-95% CI) for distant metastasis-free survival, both outperforming SLNB. Although sensitivity for RFS and DMFS were 76%, this was superior to SLNB, although these data were not broken down by patient stage. Specificity for RFS and DMFS was similar to SLNB. Litchman et al⁴⁶ evaluated 6 studies and noted high-risk GEP results correlated with poorer RFS and DMFS, as well as correlated with OS. A recent prospective study by Hsueh et al⁴³ evaluated the use of DecisionDX Melanoma in 323 patients in two prospective trials with three years of follow-up (INTEGRATE and EXPAND registries), and demonstrated that a GEP high risk result was a significant predictor of RFS, DMFS, and OS in univariate analysis, and was demonstrated to be an independent predictor of RFS, DMFS, and OS in multivariate analysis, including AJCC 8^th ed. staging (staging was also an independent predictor). Other studies had similar findings.⁴⁷ Combining AJCC staging or other clinicopathological factors and GEP was also found to improve risk stratification.^44,48-49 Other prospective and retrospective studies have subsequently been performed for GEP testing and demonstrate similar findings. Of importance, some studies have shown that the use of GEP has reduced some unnecessary SLNB, although these data are limited.^45. It is understood that comprehensive data are currently being compiled.

Analysis of Evidence

For cutaneous melanoma, a well-established set of clinical pathways is available for patient management, which are heavily based on risk stratification of patients with higher risk patients being recommended to have a more intense diagnostic workup or more intensive treatment. While the consensus guidelines also give specific clinical, pathologic, and imagining findings that can help to risk stratify a patient, additional evidence suggests that molecular diagnostic tests can be used to further improve risk stratification, thereby improving the accuracy of the risk stratification that is key to deciding on the optimal patient management plan in these clinical pathways. This is particularly the case for identifying the patients at a higher risk of adverse outcomes as a result of the melanoma.

Although some recent publications and evidence has cast doubt in the clinical utility of these tests, it is important to consider these in light of the intended use of these tests as covered in this policy.^37,38 The negative studies demonstrate disagreement in expert consensus on test use or question the predictive ability of the tests to accurately identify RFS in a stage dependent manner; however, other prospective trials and meta-analyses support the use of GEP testing and evaluate the evidence in univariate and multivariant analyses including AJCC 8^th ed. staging criteria. Despite this, it is important to note that from a clinical utility perspective, a major of these GEP tests as described in this policy is to modify management of patients who have >5% likelihood of a positive SLN to preclude unnecessary SLNB. The data reviewed suggest not only that GEP tests can better predict SLN positivity than staging and clinicopathological criteria alone; they may be more predictive of DFS and DMFS than SLNB and may be even more effective when combined with clinicopathological factors. The Marchetti et al³⁸ paper did not evaluate DMFS or SLB positivity because of limited data, and thus did not directly address the value this payor sees in these tests as a tool to help reduce the need for unnecessary SLNB in low-risk patients. Based on the evidence published and reviewed to date, this payor sees evidence that GEP testing may be used to better risk stratify and manage melanoma patients as defined in the coverage section of this policy. Specifically, this test has shown to be able to predict SLN status better than other methodologies, and thus likely reduce unnecessary SLNB in a low-risk population who would otherwise have received a SLNB as well as advocate for SLNB in those patients who may otherwise forgo the procedure based on staging criteria and clinical factors alone. Admittedly, the data supporting this test is limited and imperfect; however, the current alternative tools for assessing patient risk stratification in this population in accepted practice are even more so; with this policy we leave it to clinicians to use these services to mitigate a challenging clinical situation and hope that further evidence is developed to clearly guide melanoma patient management. We await more comprehensive published and peer-reviewed prospective data that support or refute changes in patient management based on the use of these or other similar tests.

Since the clinical utility of testing is dependent upon consensus-based management recommendations based on risk, this coverage decision is subject to change pending changes in the consensus recommended use of risk strata for patient management and the accuracy of alternative risk-stratification tools. This coverage decision will be periodically re-evaluated.

Molecular diagnostic tests used to assist in risk stratification of melanoma patients are covered when ALL of the following are true:

1. The patient has a personal history of melanoma AND:
   1. Either:
      1. Has Stage T1b and above OR
      2. Has T1a with documented concern about adequacy of microstaging
   2. Is undergoing workup or being evaluated for treatment, AND
   3. Does not have metastatic disease AND
   4. Presumed risk for a positive Sentinel Lymph Node Biopsy (SLNB) based on clinical, histological, or other information is >5% AND
   5. Has a disease stage, grade, and Breslow thickness (or other qualifying conditions) within the intended use of the test
2. The TEST has demonstrated, as part of a Technical Assessment:
   1. Clinical validity of analytes tested in predicting metastatic disease (or the absence of metastatic disease) in peer-reviewed scientific literature
   2. Utility beyond clinical, histological, and radiographical factors in the ability to accurately stratify patients into risk groups to manage patient care, such by precluding unnecessary sentinel lymph node biopsies
   3. Appropriate analytical validity
   4. Performance characteristics equivalent or superior to other covered, similar tests