Response to Comments: MolDX: EndoPredict® Breast Cancer Gene Expression Test

The following comment was submitted to Palmetto GBA and Noridian:

We appreciate the opportunity to submit comments on draft LCDs for MolDX: EndoPredict Breast Cancer Gene Expression Test which propose non-coverage of the Endopredict test when used to aid in decision-making for extended endocrine therapy (EET) in patients with breast cancer.

We request your review of the data presented below and consideration of the attached suggested revisions to the draft LCD, to allow coverage of EndoPredict in women with node-negative, hormone-receptor positive, HER2-negative breast cancer when considering extension of endocrine therapy beyond five years.

Executive Summary:

1. Societal guidelines, i.e. ASCO, clearly state that recurrence risk, specifically, should be considered when deciding about EET for women with node-negative, hormone-receptor positive breast cancer and that women with low-risk node-negative tumors should not uniformly be offered extended therapy (Recommendation 1).¹ Furthermore, the ASCO guideline states that genomic signatures, including EndoPredict, are sufficiently robust to reasonably inform this clinical decision.
2. In published analyses of three randomized trial cohorts, ABCSG-6&8 and TransATAC, EndoPredict was shown to identify a group of node-negative patients who could safely forgo EET based on sufficiently low risk of distant recurrence (DR) by 10 years or 15 years.^2,3
3. In a head-to-head comparison, EndoPredict compared similarly to Breast Cancer Index (BCI) for predicting DR.³ The BCI test currently has MolDX coverage for selecting patients with hormone receptor-positive, HER2-negative, node-negative disease who may safely forgo extended endocrine therapy.

Guidelines

The draft LCD appropriately acknowledges that current professional society guidelines suggest consideration of EET for certain subsets of patients with breast cancer.^1,4 The ASCO Clinical Practice Guideline on Adjuvant Endocrine Therapy for Women with Hormone Receptor-Positive Breast Cancer is specifically mentioned, which is understandable given the recent publication of a focused update about the optimal duration of aromatase inhibitor (AI) treatment .¹ However, we believe the statement, “ASCO Guidelines suggest using clinical criteria, including assessing tolerability, for determining which patients should be given EET” lacks important details from the guideline about incorporating the risk of recurrence. An accurate predictor of recurrence risk provides a key piece of information which patients can use in their decision-making, since some women will prefer to avoid EET (and its associated chance of side effects and complications) if their DR risk is predicted to be low. The ASCO focused update addresses postmenopausal women who have received aromatase inhibitor-(AI) based therapy as part of their initial 5 years of adjuvant treatment. Following an evidence review of the efficacy of additional ongoing AI therapy, ASCO describes the 2-4% improvement in disease-free survival (based on a relative reduction in recurrence risk of 15-20%) as “measurable but modest”. They go on to conclude that lower-risk patients will garner less benefit on average. When detailing Recommendation 1 in their guideline [page 434], ASCO states, “Many women with node-negative breast cancer are potential candidates for and may be offered extended AI therapy for up to a total of 10 years of adjuvant endocrine treatment based on considerations of recurrence risk using established prognostic factors. The use of the word ‘many’ underscores that not all women with node-negative breast cancer will benefit from extended AI therapy. In women with node negative breast cancers, T stage, grade, and genomic signature [emphasis added] are all known to serve as prognostic markers for risk of recurrence after 5 years of endocrine therapy. The Expert Panel favored extended AI treatment among higher-risk, node-negative patients, including women with stage T2/T3 tumors and T1c tumors with higher-risk prognostic factors based on the assumption that a higher residual risk of recurrence will translate into a larger clinical benefit with extended therapy.” Finally, ASCO’s expert panel “believed that retrospective findings on the importance of established prognostic factors, including stage, grade, and genomic signatures (Table 4), for both early and late recurrence have become sufficiently robust that a clinical risk stratification that reflected these prognostic factors could reasonably be used to inform the clinical decision about extended therapy with AI treatment” [page 435]. Note that Table 4 in the guideline lists prognostic factors for recurrence after 5 years of adjuvant endocrine therapy, including EndoPredict. EndoPredict’s performance in the late (5-10 and 5-15 year) time periods among women with hormone-receptor positive, HER2-negative, node-negative breast cancer.

The draft LCD Analysis of Evidence (Rationale for Determination) questions the ability of EndoPredict to delineate risk of late recurrence and contribute to decision-making about the use of EET in node-negative patients. In the combined ABCSG-6&8 cohorts, Filipits et al. showed a very low 5-10-year DR rate of 2.1% (CI 0.9-3.3%) for node -negative patients determined to be low risk by EndoPredict and who were recurrence-free at 5 years (Supplementary Table S5).² The corresponding 5-10 year DR rate for the high-risk group was almost 3-fold higher than that for the low risk group; yet still rather modest– 5.9% (95% CI 2.2-9.5%) – because some of the high-risk patients had not yet experienced a recurrence. The difference in 5-10 year DR rates between the node-negative EPclin Low vs EPclin High groups at 5-10 years was statistically significant (HR=3.13, p=0.0077, not shown in publication), even though both groups had DR rates of <10% during this time period. In a cohort with a lower baseline risk (as in the ABCSG-6&8 cohorts, in which 67% had T1 tumors and 89% had Grade 1/2 disease), the difference in DR rates between the low- and high-risk groups may become more apparent 10-15 years after diagnosis.

Although based on fewer patients, analysis of the data from the 5-15 time period showed larger and statistically significant differences in DR rates between EPclin Low vs High groups (Figure 2B and Table S5).² Evaluation of this time period is clinically important, as the risk of recurrence is known to persist out to 20 years post-diagnosis. Although some patients experienced DR in years 5-10, a similar number experienced DR in years 10-15 as well, primarily in the high-risk group. For the 5-15 year time period, node-negative patients determined to be low risk by EndoPredict continued to experience a low DR rate of 3.1% (95% CI 1.5-4.8%), while patients determined to be high risk by EndoPredict had a significantly higher DR rate of 15.1% (95% CI 4.0-24.9%) (HR=3.77, p<0.0001). These data clearly demonstrate that patients with EPclin Low scores continue to have a very low risk of late DR at 15 years and cannot be discounted on the basis that the absolute difference from the EPclin High risk group at 10 years was not as large.

Furthermore, Sestak et al. (2018) also provides data that is relevant to the use of EndoPredict in making decisions about EET.³ The performance of 6 different breast cancer gene expression profile tests was compared in a head-to-head analysis using archived tumors from patients in the TransATAC randomized controlled trial. Importantly, Sestak et al. included BCI, which has MolDX coverage for use in selecting patients with N0 disease who may safely forgo EET. In a subgroup analysis of patients that had not recurred in the first 5 years, both EndoPredict and BCI similarly were able to discriminate 5-10 year DR risk among node-negative patients with low vs. high scores. As shown in the table below, the EPclin 5-10-year DR rate was <5% in the low risk group and >10% in the high risk group. (Compared to ABCSG-6&8, the TransATAC cohort was generally higher risk, with 18% of patients having grade III tumors.)

Question raised during Noridian Open Meeting on June 2

During the Noridian Open Meeting on June 3, Dr. Clark posed a question about the following sentence (first paragraph of the Summary of Evidence) and asked for input from stakeholders: “Approximately 54% of early-stage breast cancers are estrogen receptor (ER)-positive and HER2-negative, leading to treatment with adjuvant endocrine therapy (e.g., tamoxifen or aromatase inhibitors) that significantly improves prognosis.^2,3” He questioned the discrepancy with a similar phrase that appears in the LCD for the Breast Cancer Index test, which states that ~80% of breast cancers are hormone-receptor positive. We reviewed the reference associated with the sentence above (Howlader et al. 2014), which clearly states that 72.7% of breast cancers are hormone-receptor-positive, HER2-negative. The 54% figure in the draft LCD documents appears to be an error; we have noted this in the suggested revisions in our attached document. We cannot comment on the origin of the 80% figure, as the associated reference (Pritchard et al 2017) does not appear to provide such information.

Summary
EndoPredict has been shown to reliably identify a group of hormone-receptor positive, HER2-negative, node-negative patients who are recurrence-free at five years and who have sufficiently low risk of DR by 10 years or 15 years that they could safely forgo EET. Societal guidelines support the use of genomic signatures to provide individualized recurrence risk information that aids in decision-making about EET, especially in node-negative patients who should not routinely be offered extended therapy if low-risk.
We appreciate your review of the data presented herein and request that you provide coverage for EndoPredict in women with node-negative, hormone-receptor positive, HER2-negative breast cancer when considering extension of endocrine therapy beyond five years.

Thank you for your comments. We believe the clinical utility of this test for determining which patients would benefit from EET is not clear based on the data. This test is indicated in node-negative and 1-3 node-positive patients. First, the data presented in Filipits et al. for 1-3 node-positive patients was not significant for the period studied (5-15 years), and the difference in DRFR was 84% vs 87% between the EPclin high and low group. While the difference between EPclin high and low groups are significant in node–negative patients for the duration of the study (96% vs 85% DRFR), a subset analysis of the best supported data shows that 5-10 years after ET therapy there is little difference in the performance of these patients; neither the EPclin high or low group would be defined as “high risk” (>10% risk of recurrence), and the difference in DRFR between these two groups is 97.9% vs. 94.1%. Based on this study, this contractor does not believe there is sufficient evidence that this test can clearly delineate patients who can benefit from EET (whose benefits are already known to be ‘modest’).

The following comment was submitted to Palmetto GBA and Noridian:

I am a private practice medical oncologist in Idaho Falls ID where I have practiced breast oncology for the past seventeen years. I have seen the advent of extended adjuvant hormonal therapy and the difficulty in decision making that can ensue given the modest benefit and sometimes significant toxicity. The best example would be a woman who wants to be as “aggressive as possible” in treating her cancer but is having significant side effects from five years of an aromatase inhibitor.

It has been very useful to have a biomarker in these instances to guide decision making. On the basis of the data presented by Sestak and Buss in JAMA Oncol. April 1, 2018, I have selected the EndoPredict test to use in my practice. As you know, EndoPredict performed similarly to BCI in this setting, a test for which a positive coverage determination has already been made. I believe EndoPredict to be the superior test insofar as it can extend to 15 year risk of recurrence rather than only 10 years.

I hope that you can reconsider coverage of EndoPredict for this indication in the Medicare population.

Thank you for your comments. We have addressed them in the policy.