Ed Weber, M.D., Editor
March – April, 2015

GENOMIC PROFILERS for RISK STRATIFICATION: A Primer: What Are They? How Can They Help in Clinical Management?

Genomic profilers are gene expression panels that assess the activity of multiple genes associated with tumor aggressiveness. The assay’s results provide an individual with an estimate of his prognoses for biochemical recurrence, metastatic disease and prostate cancer-specific mortality based on his individual pattern of gene expression. Two classifiers are commercially available and are in common use, “Prolaris” and “Oncotype Dx,” and a third one is less common, “Decipher”.

The current research into improving risk stratification in all phases of prostate cancer is overwhelmingly extensive. But the bar for success is set quite high. The venerable classifiers of risk — by D’Amico and the NCCN (i.e. groupings of PSA, tumor stage, and Gleason score, often supplemented by extent of cancer in a single positive core and the number of cores positive), provide an accuracy of prediction of risk in the 80+% range. This accuracy may be further augmented by integration of the CAPRA and CAPRA-S scores, and with various nomograms which also may include elements such as percentage of cores positive for cancer, PSA density and velocity, PCA3 status, and many others. These classifiers attempt to address the vexing requirement of accounting for tumor heterogeneity, which increases as the risk groupings bundle cancers of greater aggressiveness, i.e. intermediate and high risk cancer. It is unlikely that any system of classification can account for 100% of this heterogeneity. Therefore the goal for the new genomic risk classifiers is to add independent information to improve upon the current systems by another 10% or more.

The task for the physician is to utilize genomic classifiers in areas of clinical decision making where the results of these assays are likely to make a significant contribution.


First, … the Oncotype Dx from Genomic Health, Inc, introduced in May 2013 yielding what is termed the genomic prostate score (GPS). The development work combined the efforts of the researchers at UCSF, notably Drs. Peter Carroll and Matthew Cooperberg, and at the Cleveland Clinic, Dr. Eric Klein. Their product relied heavily on data from the 15,000 men followed by CaPSURE (Cancer of the Prostate Strategic Urologic Research).

The assay provides a risk assessment — independent of the Gleason score, for the likelihood of high-grade cancer, extra capsular extension and clinical recurrence. The GPS is especially applicable for men initially judged to be low-risk who are considering active surveillance.

The initial research compared the GPS assessed on micro dissected tissue from the surgical specimen and compared the GPS estimate to the final surgical pathology. “The 17-gene GPS [was] specifically developed to overcome tumor heterogeneity and biopsy under-sampling” (AUA 2013 Abstract #2131). “In the validation study (395 pts), GPS, was assessed in biopsies with as little as 1mm tumor length from patients suitable for AS and strongly predicted (p=<0.005) high grade and/or pT3 disease after adjusting for CAPRA (see below) or other standard pretreatment factors,” (ibid)

The initial 288 gene candidates eventually were compressed to the 17 that compose the panel. They signal for cellular organization, androgen signaling, proliferation, and stromal response. An individual’s patten of expression of these 17 genes determines the GPS by which the aggressiveness of his cancer can be robustly predicted, p=0.002.

The genomic prostate score ranges from 0 to 100 and each 20 points increase doubles the risk of adverse pathology — 2.5X for predicting high-grade disease and 2.2X for predicting pT3 stage.

OncotypeDx is in the process of preparing a panel to predict outcome after prostatectomy.


The 31 gene Prolaris assay, Myriad Genetics, Inc, assesses the extent to which these selected genes are expressed or not expressed, and yields the cell cycle progression (CCP) score, designed to predict cancer aggressiveness. The score can reinforce the selection of candidates appropriate for active surveillance (AS), and conversely, can suggest the presence of significant cancer that renders men less suitable for AS. The foundation study for Prolaris was reported by Cuzick et al., “Prognostic value of the cell cycle progression signature for prostate cancer death in a conservatively managed needle biopsy cohort” (Br J Cancer. 2012 March). The CCP score is reported on a scale of -2 to +5, with each one-unit increase conferring 2X the risk for prostate cancer death. In the Cruzik study, the median score for the 349 men who did not receive immediate treatment was 1.03. “The 10-year death rate … for those with a score less than 0, was 19.3%, and increased to 19.8%, 21.1%, 48.2% and 74.9% for CCP score groups 0-1, 1-2, 2-3, >3, respectively.” Cruz concludes: “The most obvious clinical use of the CCP score is to help identify low-risk patients who can be safely managed by surveillance.”

The Prognostic Utility of the Cell Cycle Progression Score Generated from Biopsy in Men Treated with Prostatectomy” (J Urol. August 2014) by Bishop et al. reported their findings that in 582 men the CCP based on the biopsy specimen “was associated with biochemical progression and metastatic disease” when stratified into CCP scores of -2 to 0, 0 – 1, 1 – 2, and 2 – 3. They concluded: “In the current study the score identified men in whom surgery was likely to fail.”

In the poster session at the 2014 AUA meeting, Dr. Cooperberg and others presented their work on: “Predicting Radical Prostatectomy Outcome Among Men Who Upgrade from Clinical Gleason 6 to Pathological Gleason 7.” The 10-year Kaplan Meier projection for progression-free survival (PFS) showed important differences for different CCP scores: -2 to 0, about 80%; 0 – 1, ~50%; and ~25% PFS for CCP 1-3.


Integrating OncotypeDx and Prolaris with CAPRA and CAPRA-S:

The predictions of the OncotypeDx and Prolaris can be strengthened by integration with the CAPRA score (Cancer of the Prostate Risk Assessment) developed at UCLA notably by Drs. Cooperberg and Carroll. (J Urol. June 2005).

CAPRA is based on the biopsy Gleason score and readily available clinical parameters. It estimates the risk of recurrence in patients with localized cancer. It incorporates 5 parameters: PSA, Gleason score, clinical T-stage, biopsy results and age. The score spans 0 – 10. The risk of recurrence doubles for each 2 point increase in score. The CAPRA score is useful in selecting men for active surveillance.

CAPRA-S incorporates data from the prostatectomy specimen and predicts for biochemical recurrence and mortality.(Euro Urol. 2013 Apr). It can be useful in determining the need for adjuvant therapy.

Both CAPRA and CAPRA-S are available on line as free apps.

Both tools have been validated and are strongly predictive of outcome (possibly with as great as 80% accuracy).To improve upon the CAPRA estimates is the goal of both OncotypeDx and Prolaris.

NCCN Guidelines Version 1.2015

The most recent guideline describe and discuss both the OncotypeDx and Prolaris assay in the section on Molecular Testing. However, no recommendation for their use was made — “Their clinical utility awaits evaluation by prospective randomized clinical trials, which are unlikely to be done.

References were cited in association with summary comments such as: “Prolaris has changed treatment recommendations in 32% to 65% of cases and may enhance adherence to the treatment recommended.” “Oncotype DX GPS improved NCCN risk group assignments, which may enhance rate of compliance with recommended active surveillance or diminish the number of surveillance prostate biopsies.”


While these tools — CAPRA(S), OncotypeDx, and Prolaris — have been validated to improve risk stratification over and above standard metrics, the clinician remains faced with the question of whom to test and how to integrate the results in management decisions. In each case a middling score might not be persuasive as a tie-breaker. An unanticipated low or high score might be convincing, but the likelihood may be that unexpected results at extremes of the scales would be uncommon. Sound clinical judgement is always the necessary and final ingredient.