In this issue of ONCOLOGY, Dr. Anthony D’Amico outlines a prognostic model for grouping localized prostate cancer patients into high-risk (1992 American Joint Committee on Cancer stage T3b, Gleason score > 7, prostate-specific antigen [PSA] > 20 ng/mL), intermediate-risk (T2b-T3a, Gleason score = 7, PSA = 10-20 ng/mL), or low-risk (T1c-T2a, Gleason score < 7, PSA < 10 ng/mL) categories. As noted by Dr. D’Amico, the large intermediate-risk category is heterogeneous, and further refinements in prognostication within this group may help guide therapeutic decision-making.
Controversies remain about appropriate local treatment as well as the benefits of adjuvant systemic therapy for intermediate-risk patients. Dr. D’Amico proposes a combined-modality staging systemadding the percentage of positive biopsies to clinical stage, Gleason score, and pretreatment PSAto stratify intermediate-risk patients with more precision.
Examining surgical series from the University of Pennsylvania and Brigham and Women’s Hospital as well as a radiation series from the Joint Center for Radiation Therapy, Dr. D’Amico stratifies intermediate-risk prostate cancer patients by percentage of positive biopsies and demonstrates three distinct biochemical freedom-from-failure curves. This system may improve our ability to separate patients with a high risk of failure and disease progression from more favorable patients, and may improve treatment individualization and patient selection for future clinical trials.
Bulkier disease is more likely to extend beyond the prostate capsule, involve the seminal vesicles or pelvic lymph nodes, and result in positive surgical marginsall factors that increase the likelihood of failure after local therapy. Staging procedures incorporating some estimate of disease volume might be useful, and percent positive biopsies would seem an appropriate surrogate for tumor bulk. We suggest, however, that for core biopsy-based prognostic criteria such as percent positive biopsies, standardization of staging methods will be important for ultimate general applicability.
In the studies cited, the definition of percent positive biopsies (number of positive biopsies divided by total number of biopsies) was based on traditional sextant biopsies. Additional investigations will be needed to determine how to apply this definition to alternative sampling methods, including biopsy techniques in which several additional cores of tissue are obtained from palpable or transrectal ultrasound-detected focal abnormalities, or extended biopsies with eight or more cores (which are now commonly employed).
Other parameters, including measurements of how much disease is present in each core, may yield additional information regarding disease volume. How should one compare the prognostic implications of a sextant biopsy showing 4 of 6 biopsies positive (67%), each containing < 5% adenocarcinoma, with another showing 2 of 6 biopsies positive (33%) and 80% cancer in each positive core? Some method incorporating both the number (or percentage) of positive biopsies and quantity of disease within each core might further enhance prediction of local tumor burden.
Prognostic Model End Points
Most prognostic protocols for prostate cancer have been developed with a focus on biochemical (PSA) failure, with the implicit assumption that biochemical failure will predict prostate cancer-specific mortality. Unfortunately, given the long natural history of prostate cancer and the fact that PSA measurement has only been available for about 15 years, this assumption has not yet been proven. We agree with Dr. D’Amico that longer follow-up and additional research are needed to validate this assumption as well as the utility of prognostic models based on biochemical failure as an end point.
Roach et al recently reported that four distinct prognostic groups could be identified based on clinical stage and tumor grade, using patient data from four large, multi-institutional phase III randomized trials conducted by the Radiation Therapy Oncology Group (RTOG). The four identified subgroups were shown to predict the 5-, 10-, and 15-year disease-specific survival of patients treated with radiation therapy alone. The majority of patients in these RTOG trials were treated prior to the availability of the PSA blood test, but our preliminary findings suggest that the RTOG risk groups do, in fact, predict PSA failure, progression-free survival, and overall survival in a modern, PSA-era patient cohort. Furthermore, pretreatment PSA levels appear to yield additional prognostic information regarding disease-specific survival.
Survival end points are needed in prostate cancer studies, and PSA-era data are only beginning to become available. Prognostic models developed with survival end points may be superior to those developed with surrogate end points such as biochemical failure. The ultimate goal is to enhance the practitioner’s ability to counsel patients on the risks and benefits of PSA screening and the relative efficacy of various treatment options for the patient’s individual disease. The model should also help with patient selection for clinical trials by targeting higher-risk patients for adjuvant therapy (as in the ongoing RTOG 9902 phase III trial, which is combining androgen suppression and chemotherapy with definitive external-beam radiation therapy).
We agree that the ideal prognostic model will eventually incorporate measures of not only a patient’s disease, but also his overall health status and long-term impact of his treatment selection. Only with complete information can our patients truly make informed decisions about the influence of prostate cancer on their overall health.