BACKGROUND: In prostate cancer management, differentiating relatively indolent from more aggressive disease is a key clinical issue. The Gleason grading system, based on microscopic features, is one of the strongest predictors of many clinical endpoints, including PSA recurrence following radical prostatectomy or radiotherapy. However, the Gleason grading system is subject to inter-observer variation. Further, among low or intermediate grade cancers (Gleason scores 6 or 7), the ability of Gleason score to predict the clinical course is limited. Due to the sampling effect and tumor heterogeneity, the Gleason grade in biology specimen may not reflect the true grade in the prostate gland. It is accepted that due to the limitations of current prognostic factors, many patients are over treated and new prognostic factors are needed to assist clinical decision making.
METHODS: We studied 191 patients with localized prostate cancer treated with radical prostatectomy at Massachusetts General Hospital between 1993 and 1995. The clinical endpoint of progression was PSA recurrence and the median follow-up was 7.2 years. Formalin-fixed paraffin-embedded tumor sections were analyzed for the expression of 1536 specifically selected prostate cancer related genes using the DASL platform. The samples were divided randomly into a training set of 124 tumor samples (53 recurrences) and a validation set of the other 67 samples (24 recurrences).
RESULTS: A 62-gene predictor was developed and strongly prognostic for PSA recurrence in both the training set (hazard ratio=8.3, 95% CI 3.9-17.7, p < 0.0001) and the validation set (hazard ratio=6.3, 95% CI 2.45-16.0, p=0.00013). After adjusting for Gleason score, the 62-gene predictor remained a powerful predictor with hazard ratios of 6.3 (95% CI 2.8-14.1, p < 0.0001) and 4.6 (95% CI 1.61-13.1, p=0.0045) in the training set and validation set, respectively. Importantly, the 62-gene predictor separated patients with Gleason scores 6 and 7 cancers into good and poor prognosis groups in the training set (p <0.0001) and the validation set (p=0.0015).
CONCLUSIONS: A set of 62 predictive genes has been identified that strongly separates men with Gleason score 6 and 7 prostate cancers into good and poor prognosis groups for PSA recurrence following prostatectomy. If further validated, this gene expression-based signature may provide strong prognostic information beyond the Gleason grading system.