© 2 0 0 7 B J U I N T E R N A T I O N A L | 1 0 0 , 1 2 5 9 – 1 2 6 3 | doi:10.1111/j.1464-410X.2007.07136.x 1259 NO CLAIM TO ORIGINAL US GOVERNMENT WORKS Original Article PROGNOSIS IN PROSTATE CANCER CONCATO et al. Molecular markers and mortality in prostate cancer John Concato*†, Dhanpat Jain‡, William W. Li§, Harvey A. Risch¶, Edward M. Uchio∞ and Carolyn K. Wells*† *Clinical Epidemiology Research Center, VA Connecticut Healthcare System, West Haven, Departments of †Medicine (General Medicine), ‡Pathology, ¶Epidemiology and Public Health, and °Surgery (Urology), Yale University School of Medicine, New Haven, CT, and the §Angiogenesis Foundation, Cambridge, MA, USA Accepted for publication 8 June 2007 ambulatory care during 1989–90 at nine Veterans Affairs (VA) medical centres in New England, 1274 had incident prostate cancer during 1991–95. We obtained the medical records and diagnostic tissue for these men, and then extracted demographic data and clinical information, and conducted immunohistochemical assays of molecular markers in biopsy tissue, as potential prognostic factors. In this interim analysis, data on 250 patients were analysed; the main outcome was overall mortality to 31 December 2003, providing 8–13 years of follow-up. RESULTS In 228 (91%) patients with available medical record and laboratory data, the median age was 72 years and the median prostate- specific antigen level was 10.4 ng/mL. In adjusted (multivariate) analyses that included traditional prognostic factors, bcl-2 staining (hazard ratio 2.14, 95% confidence interval 1.27–3.58, P = 0.004) and high microvessel density (1.76, 1.19–2.60; P = 0.005) had an independent effect on the outcome. CONCLUSIONS Bcl-2 and microvessel density are independent predictors of subsequent death among men with prostate cancer and might have a clinical role in assisting in deciding on treatment. KEYWORDS prostatic neoplasms, prognosis, angiogenesis, bcl-2 proto-oncogene Study Type – Prognosis (retrospective cohort study) Level of Evidence 2b OBJECTIVE To evaluate prognosis in prostate cancer by assessing the independent effect of selected molecular factors (e.g. markers of cell-cycle regulation), in addition to the effect of traditional clinical factors (e.g. anatomical stage, histological grade), in predicting long- term mortality among men newly diagnosed with prostate cancer. PATIENTS AND METHODS In a community-based population of 64 545 USA veterans aged ≥ 50 years and receiving INTRODUCTION Published reports have examined various aspects of prognosis in prostate cancer. Patient-orientated research has focused on clinically relevant issues, e.g. predicting the pathological stage at surgery [1], assessing clinical or biochemical progression of disease after radical prostatectomy [2], examining the effect of preoperative PSA velocity on mortality after surgery [3] or radiation therapy [4], and describing long-term survival after conservatively treating localized prostate cancer [5]. Concurrently, laboratory- based and translational research has focused on molecular markers [6,7], but usually in studies that assess the correlation of such markers with tumour stage or grade, or as risk factors for resistance to androgen ablation or radiotherapy. In a more comprehensive approach, a combination of clinical and laboratory-based factors would be evaluated, at the time of diagnosis of prostate cancer (rather than in subgroups of patients who have received a particular therapy), to determine their effect on long-term survival. In this context, the objective of the present study was to determine whether factors related to angiogenesis and cell-cycle regulation, when assessed simultaneously with traditional clinical factors, provide independent predictive information about mortality among men with incident prostate cancer; in this report we describe an interim analysis from an ongoing study of prognosis in prostate cancer. PATIENTS AND METHODS The present sample was selected from a community-based population of 64 545 men receiving ambulatory care during 1989–90 at nine Veterans Affairs (VA) medical centres in New England, USA. Medical records and pathology registries identified 1274 men with incident prostate cancer during 1991–1995. Candidate prognostic factors, assessed up to the time of initial treatment, include baseline demographic, clinical and molecular characteristics. These factors are being evaluated for their effect on overall mortality to 31 December 2003, providing a median (range) of 11 (8–13) years of follow-up. The primary analysis adjusted for treatment(s) received by patients; the secondary analyses included a treatment-stratified approach, and cause-specific (prostate cancer) death as the outcome. This interim analysis includes over-sampling of mortality to provide adequate outcome events; 250 patients were identified based on their vital status as of 31 December 1999. To take advantage of additional follow-up, the current report considers mortality to 31 December 2003. Among the 250 patients, 228 (91%) had available medical records and biopsy tissue. Data were collected from three sources: first, a review of the medical records to obtain clinical data, based on a