Response to Comments: 4Kscore Assay

Thank you for the opportunity to respond to the Proposed/Draft LCD for the 4Kscore Assay. We appreciate the discussion at the Open Meeting, and are now providing the supporting documents as requested. I have attached the 4Kscore dossier, which describes the validity and clinical utility of the assay, along with a PowerPoint summary of the information.

In addition, below is information specifically addressing the points brought up in the discussion.

1) The 4Kscore is indicated in men aged 45-75 years of age, with a PSA between 1.5-10 ng/ ml . A sub-analysis of this group in the US validation study shows excellent discrimination, with a higher AUC than total PSA, % free PSA, and age adjusted PSA (See table below, sub-analysis of Parekh et al.).

[Table available upon request, would not format into this article] Cut Point

2.) The value of the 4Kscore algorithm to improve prostate cancer evaluation is demonstrated in the US validation study with an improvement vs comparative tests in AUC, decision curve analysis, and a very close calibration to prostate biopsy results (Parekh, et al.). The value of intact PSA and hk2, which have been associated with a higher risk of aggressive prostate cancer, is also indicated in this study (Table 2). The AUC of the ful l model is 0.821 vs 0.751 when these biomarkers are removed (p < 0.0001).

3.) The study reporting long term risk of developing distant metastasis is a nested case control study of 12,542 men (Stattin et al.). It examines the scenario of a single 4Kscore reading at 50 or 60 years of age without any follow up for 15-20 years, which would be a worse case than any expected in clinical practice. Even without follow-up, men with an elevated PSA (<". 2.0 ng/ml in men aged 50 years and<". 3.0 ng/ml in men aged 60 years) and a low 4Kscore (< 7.5%) had a very low risk of developing distant metastasis over a period of 20 years.

[Table available upon request, would not format into this article]

4.) The clinical utility study is a retrospective decision impact study (Konety et al.). It demonstrates the significant reduction in prostate biopsies in men who have been referred to urologists with an abnormal PSA and/or digital rectal examination. The 64% reduction in prostate biopsies, with a 94% reduction in men with a low risk 4Kscore (<7.5%), demonstrates the significant impact on reducing prostate biopsies. The accuracy demonstrated in the US validation study (AUC and calibration to prostate biopsy results), along with the numerous clinical studies validating the accuracy of the 4Kscore for predicting risk of aggressive prostate cancer, demonstrate the ability of 4Kscore to reduce prostate biopsies in low risk men, while accurately identifying higher risk men for further evaluation.

5.) The dossier and PowerPoint detail the analytic and clinical validation of the 4Kscore along with the clinical utility of the test, along with the clinical need for a more accurate and specific test to follow an abnormal PSA and/or DRE.

We appreciate the comments. Collectively, these comments seem to imply that a 4K score should be used as a binary test with a threshold score of 7.5 used as a cutoff to predict which men would have Gleason 7 or greater score if they had a prostate biopsy. However, guidelines for clinical decision making in the management of prostate cancer require additional information obtained on biopsy, such as number of positive cores. As such, it is not clear that a test that predicts Gleason score alone serves to adequately allow risk stratification without a tissue sample so as to improve patient outcomes.

I would like to weigh in on the 4K score LCD. I have been using it for 6 months and it is an excellent tool to determine which patients need biopsies and which can avoid the discomfort, cost and risk of bleeding and sepsis. Since the PSA has been determined to be a D rating by the USPSTF, it is disappointing that something that would be more accurate would not be covered by Medicare. Since over 26,000 men die of prostate cancer annually, it is still a disease with significant mortality and a better screening test would be helpful.

The commenter’s conclusions form a viable hypothesis for how the test could be used. However, no prospective studies presently substantiate this use.

This comment compared several aspects of testing for prostate cancer and the information is appreciated. That said this policy focuses on the 4K Score therein the other references are not included as they have not all been validated by Noridian nor the MolDX contractor.

Newer PCa Biomarkers

Improved diagnostic accuracy, but modest especially compared to the costs

Adjunct to PSA – not instead of.

CMS vs FDA approval

? Change outcome

? Person vs Population

? Change plans

? Cost effective

? Paid for

Trade off of increasing specificity at the expense of sensitivity, i.e. decreasing the biopsy burden at the risk of missing high-grade cancer.

Potential usefulness is more clear in men willing to forego biopsy if the risk of progressive cancer is deemed low and in select men in whom the risks of prostate biopsy are particularly high, such as those with prior sepsis after prostate biopsy and in those who cannot discontinue anticoagulation to undergo biopsy.

4Kscore (OPKO Lab) 4 kallikrein panel

www.4Kscoretest.com, 1.877.922.8364,

Approved 2014 - Before 1st bx or after negative biopsy - does not yet have a validated algorithm for AS.

Based on research from Memorial Sloan-Kettering Cancer Center

Measures total PSA, free PSA, intact PSA, and hK2 (last 2 are proprietary)

Uses a proprietary algorithm, which also incorporates age, DRE, and prior biopsy status along with the 4 kallikrein markers to estimate the number necessary to biopsy to find one HGPCa and risk of HGPCa

Assigns a probability score from less than 1% to greater than 95% of having biopsy-detectable, clinically significant PCa (Gleason ≥7)

100% minus the score provides the NPV/probability that a patient will not have Gleason ≥7

Annual US savings of $19 million - $1 billion, depending on the test adoption and actual cost of test.

94% accurate in detecting aggressive PCa at 6% cutoff.

30-71% of bx avoided but miss 12% HGPCa

PSA >3 but low risk 4Kscore <1% diagnosed w/ advanced PCa w/in 5 yrs

Test score          Fewer % Bx       Missed % HGPCa

>6%                      30                       1.3

>7.5%                   38 - compared to PSA >3.0

>9%                      43                       20

>10%                    29                       10

>15%                    58-71                    5

Summarizing a chart that was included “At a 4K score test result of 27%, about 1 in 4 men biopsied would have high grade prostate cancer.

We appreciate the concerns voiced in this comment and duly note that the need for a well designed and validated test for prostate cancer would greatly improve the decision making as to when active surveillance or other intervention is better suited for the individual.

• 7500 men with PSA ≥3.0 ng/mL who had a sextant biopsy with over 1900 ERSPC patients having evidence of any grade prostate cancer at biopsy – allowed researchers opportunity to determine whether prediction of risk of PC at biopsy using a statistical model based on the 4K-panel was accurate and could be independently replicated across multiple clinical settings using large, highly representative population-based cohorts of men aged 50-74 who were either previously unscreened, subject to previous PSA-testing, having a previous negative biopsy, or had been subject to clinical work-up prior to biopsy.
• First report on 4K-panel was based on cryopreserved serum and data from 740 previously unscreened participants in the prospective Goteborg (Swedish arm of ERSPC) who had PSA ≥3.0 ng/mL and were followed up with DRE, TRU, and sextant biopsy.
• The next report was an evaluation of 4K-panel in 1241 recently screened participants in the Goteborg trial. These studies involved the technical evaluation of monoclonal antibodies vs recombinant Fab-capture to measure intact fPSA and hK2.
• The first change in assay protocols for intact fPSA and hK2 was in Vickers et al. (J Clin Oncol., 2010) based on 2914 previously unscreened men in ERSPC-Rotterdam with PSA ≥3.0 ng/mL who were followed up by DRE, TRUS, and a sextant biopsy. The change in assay protocols for intact fPSA and hK2 led us to build a new statistical model to predict prostate biopsy-outcome based on the 4k-panel testing in cryopreserved serum from the ERSPC-Rotterdam participants. The Rotterdam cohort was randomly divided 1:3 into a training and validation set. The validation set utilized a locked down, statistical model based on levels of the 4K-panel, age and DRE with model-based prediction compared with empirical biopsy outcome.
• Clinical practice changes during the early 2000s, e.g. use of ≥10-core prostate biopsy, changes in Gleason grading criteria for Gleason score ≥7 cancer (high grade), and cryopreservation of anti-coagulated plasma rather than serum, led us to develop a new statistical model to predict risk of Gleason score ≥7 prostate cancer at ≥10-core prostate biopsy based on 4K-panel testing from 4765 biopsied 50-70 year old participants in the large, prospective randomized ProtecT-trail in UK who had a PSA ≥3.0 ng/mL. The 4K-panel enhanced cancer detection compared with PSA and age alone. AUC for the 4K-panel was 0.719 (95% CI: 0.704; 0.734) vs. 0.634 (95% CI: 0.617; 0.651) for PSA and age alone for any-grade cancer, and 0.820 (95% CI: 0.802; 0.838) vs 0.738 (95% CI: 0.716; 0.761) for high-grade cancer. Using a 6% risk of high-grade cancer as illustrative cutoff, the 4K-panel model reduced the need for biopsy in 43% of mean with PSA≥3.0 ng/mL, detected 90% high-grade cancers, and delayed diagnosis of 10% high-grade cancers.

The European studies prior to publication of the Bryant study (JNCI, 2015) use a different model/algorithm because both biopsy and grading practice changed in the late 1990's and early 2000’s. Subsequent to the Bryant study, the algorithm was “locked down”. Nevertheless, results of the pre- and post-Bryant model/algorithm change are not directly comparable.

• It is demonstrably false that the model has never been “locked down”. In the study on the French ERSPC cohort (Benchikh et al, BMC, 2010) the abstract reads: “We applied a previously published predictive model”. Similarly, in the Rotterdam ERPCS study (Gupta et al, BJC, 2010) the abstract states: “A previously published statistical model was applied”. In both cases the model applied was the one built on a “training” set of the Rotterdam ERPSC cohort.
• The “LCD uses the Bryant study (JNCI, 2015) to “highlight” that the model has never been locked down. He states “But there was a very specific reason why the Bryant study was conducted, why a new model was generated and why that model was used in subsequent studies. The reason is that both biopsy and grading practice changed in the late 1990’s and early 2000’s. The commenter states that the model developed in the Rotterdam ERSPC was of value for demonstrating “proof of principle” and it should not be applied to contemporary patients. The decision to create a new model in the Bryant study using the ProtecT cohort- which involved at least 10 core biopsy and predominately modern grading practices – was not, as implied in the LCD, an arbitrary one, but a conscious choice to ensure contemporary clinical applicability. The model created using the Bryant data was then applied in all subsequent cohorts, including the Swedish community study, the Statin long-term follow-up study and the Parekh validation study.”

• The PSA units and reference to the assay not being “locked down” will be corrected in the policy.
• The validation studies appear adequate to show the predictive accuracy of 4Kscore.
• However, there is no evidence of clinical utility. As stated above, the intended use population, the intended user and PSA cutoff for testing have never been defined. Current company marketing appears directed towards the PCP for automatic reflex testing without regard to PSA cutoff. Prospective studies must demonstrate that 4Kscore testing improves patient outcomes, or changes physician management that results in improved patient outcomes.
• Despite the absence of clinical utility to meet Medicare’s reasonable and necessary criteria for coverage, this test is marketed as a “screening” test and would not be covered under NCD 210.1
• The NCCN has not endorsed 4Kscore. The NCCN says “… 4Kscore are potentially informative in patients who have never undergone biopsy or after a negative biopsy, and “consider…, 4Kscore…” and “Tests that improve specificity in the post-biopsy state ….4Kscore…should be considered in patients thought to be higher risk despite a negative prostate biopsy.” Medicare has no requirement to cover biomarkers listed in NCCN.