A genomic approach to active surveillance: A step toward precision medicine

In the last 25 years, PSA screening has resulted in a large gap between the likelihood of being diagnosed with, and of dying of, prostate cancer, leading to the clinical problems of over-diagnosis and over-treatment. Although there is agreement in urologic circles that early detection and aggressive treatment of higher-grade cancers reduces prostate cancer-specific mortality, the widespread overtreatment of low-grade, non-aggressive disease led the U.S. Preventive Services Task Force to recommend against routine screening. There is accumulating data from several institutions on an alternative management strategy called active surveillance. Active surveillance is defined as expectant management with curative intervention delayed until signs of tumour progression. The main advantage of this approach is that it avoids overtreatment of indolent disease, thereby restricting the cost and morbidity of curative-intent treatments only to those who have potentially life-threatening cancers. Despite the favourable outcomes reported for surveillance, its clinical use is limited, with >90% of men in the U.S. diagnosed with potentially indolent disease undergoing immediate treatment with radiation or surgery.

There are several reasons that surveillance has not been more widely adopted — legal (fear of being sued if the window of curability is missed), economic (doctors get paid to intervene, not watch), and emotional (patient and family anxiety over not treating a known cancer). However, the major limitation to wider use of surveillance is the lack of a tool that can distinguish indolent from aggressive prostate cancer at the time of diagnosis, and one that can be used on subsequent biopsies to determine if someone on surveillance has true biologic progression.

Working with Genomic Health, Inc., we have designed a novel strategy built upon the approach used to successfully develop molecular profiling of breast and colon cancers designed to address the challenge of tumour heterogeneity inherent in prostate cancer. Results from the first two studies in this strategy were recently presented at the 2012 American Society of Clinical Oncology (ASCO) annual meeting (Klein EA, Maddala T, Millward C, et al., abstract #4560). We first conducted a gene discovery study in fixed, paraffin-embedded (FPE) tumour tissue from patients treated by radical prostatectomy, where the relationship between gene expression and clinical tumour recurrence in two separate tumour foci selected to represent the primary and highest Gleason patterns were examined. We found a group of 288 genes in six biological pathways that predict for clinical recurrence expressed in common by both the primary and highest Gleason patterns. We then conducted a second study to demonstrate that the majority of the most highly predictive genes identified in radical prostatectomy specimens in the first study, when assayed in tumour from prostate needle biopsies, could also predict adverse pathology at the time of prostatectomy.

METHODS

For the gene identification study, we sampled 441 tumour specimens from a large pool of approximately 2,600 men treated by RP between 1987 and 2004 at the Cleveland Clinic Glickman Urological & Kidney Institute. For the biopsy study, FPE prostate needle biopsy specimens were selected from an additional 167 patients (92 low-risk and 75 intermediate-risk) who had both a prostate biopsy and RP at our institution. All specimens were re-reviewed and assigned Gleason pattern and score using the 2005 International Society of Urological Pathology Consensus guidelines. In the RP study, we sampled two spatially distinct tumour specimens which represented the primary Gleason and highest or secondary Gleason patterns. For the biopsy study, representative tissue blocks were selected for each patient.

For the gene discovery study, 727 candidate genes selected from a meta-analysis of publicly available DNA microarray datasets were analysed. Candidate genes were assayed for expression by quantitative RT-PCR assays. Eighty-one candidate genes identified in the gene discovery study were assayed using the same methods in the needle biopsy study.

STATISTICAL METHODS

For the gene discovery study, the primary objective was to identify genes associated with time to clinical recurrence (local recurrence or distant metastases), and for the needle biopsy study, the presence of adverse pathology (high-grade or non-organ confined disease) in the RP specimen. Cox proportional hazards regression and logistic regression models were used to evaluate associations between genes and outcome variables. The false discovery rate (FDR) was controlled at 10%.

RESULTS

Figure 1
In data presented at the 2012 ASCO Annual Meeting, we identified 288 genes that were similarly predictive of clinical recurrence (as assessed by standardised hazard ratios) in both primary and highest Gleason pattern. This result demonstrated that certain genes could predict tumour aggressiveness regardless of the Gleason pattern tumour in which they were assessed (see figure 1). These genes span multiple pathways that are differentially associated with aggressiveness, for example, higher expression of the stromal response and proliferation genes is associated with higher risk of clinical recurrence, while for other groups (cellular organisation, basal epithelial, androgen and stress), higher expression is associated with lower recurrence risk. After adjustment for AUA risk group (based on pretreatment PSA, T stage and PSA), 198 genes, including representative genes from the six groups identified in univariable analyses, remained strongly associated with clinical recurrence in tumour taken from either the primary or highest Gleason pattern. Importantly, in the second study, expression patterns in these groups were also predictive of adverse pathology on RP in tumour samples taken from needle biopsy specimens (see figure 2). Overall, 58 of 81 (72%) tested genes predicted high grade and/or non-organ-confined disease (FDR<10%).

Figure 2
DISCUSSION

In 2011, the Institute of Medicine issued a call for the development of a new system of disease classification that would link molecular data to health outcomes in order to allow more precise clinical decision making that is tailored to individual patients, a concept termed ‘precision medicine’. The overdiagnosis of nonlethal prostate cancer by PSA screening, coupled with recent advances in genomic profiling of prostate and other human cancers, represents a significant opportunity to apply the concept of precision medicine to the management of prostate cancer.

The major question most newly diagnosed men today face is no longer “What is the best treatment for my cancer”, but rather “Does my cancer need to be treated at all?”. Current clinical predictors, including nomograms, lack the discriminative ability to answer this question for most newly diagnosed tumours. We have attempted to address this limitation by studying the biology of prostate cancer as revealed by gene expression profiling from both radical prostatectomy specimens and prostate biopsies. The capacity to predict clinically meaningful outcomes from biopsies is essential for those considering active surveillance, since it is the only material available on which to make a judgment.

Our studies, as presented at the 2012 ASCO Annual Meeting, have revealed that sampling the expression of genes contained in multiple biological families has the ability to predict outcomes in ways that can be used in inform clinical decision making. The discovery study identified 288 genes that can predict for the development of metastasis or prostate cancer death whether they are assayed in the primary or highest Gleason pattern present in prostatectomy specimens; a subset of these genes assayed on biopsy samples also predicted for adverse pathology at RP. Together these observations suggest that meaningful information on outcomes is contained in the small amounts of tissue obtained at biopsy. The fact that this information can be obtained from either the primary or highest Gleason pattern tumour suggests that the sampling error inherent with needle biopsy consequent to tumour multifocality and heterogeneity may be overcome with this approach (although this conclusion requires a great deal more study). A multigene assay developed from this work is now undergoing validation using biopsy specimens from an independent cohort of patients from another centre. If our initial findings are confirmed, gene expression profiling of biopsy samples at the time of diagnosis and subsequently in those initially managed by surveillance could have major clinical impact, bringing precision medicine to the prostate cancer clinic in the near future (see figure 3).


Figure 3
ABOUT THE CLEVELAND CLINIC GLICKMAN UROLOGICAL & KIDNEY INSTITUTE

The Glickman Urological & Kidney Institute is one of 26 institutes at Cleveland Clinic. In 2012, U.S. News & World Report ranked Cleveland Clinic #1 in Urology and #1 in Nephrology care in the United States. The Glickman Urological & Kidney Institute is a world leader in treating complex urologic and kidney conditions in adults and children. Cleveland Clinic physicians have pioneered medical advances including partial nephrectomy, laparoscopic, robotic and single-port urologic surgery, while serving tens of thousands of patients annually. Last year alone, people from all 50 states and 34 countries traveled to Glickman Urological & Kidney Institute for cures, treatment and long-term kidney care.

Cleveland Clinic is an integrated healthcare delivery system with local, national and international reach. At Cleveland Clinic, 2,800 physicians represent 120 medical specialties and subspecialties. The Cleveland Clinic health system includes a main campus, 18 family health centers, eight community hospitals, Cleveland Clinic Florida, the Cleveland Clinic Lou Ruvo Center for Brain Health in Las Vegas, Cleveland Clinic Canada, Sheikh Khalifa Medical City, and Cleveland Clinic Abu Dhabi.

In 2012, Cleveland Clinic was ranked one of America’s top 4 hospitals in U.S. News & World Report’s annual “America’s Best Hospitals” survey. The survey ranks Cleveland Clinic among the nation’s top 10 hospitals in 14 specialty areas, and the top hospital in three of those areas.

ACKNOWLEDGEMENTS

Genomic Health, Inc. provided research funding to Cleveland Clinic for performance of this study. A patent application for the development of a genomic assay based on the genes identified in this study has been submitted by Genomic Health, Inc.