1. ¹Cancer Research UK Centre for Epidemiology, Mathematics and Statistics, Wolfson Institute of Preventive Medicine, St Bartholomew's Medical School, Queen Mary, University of London, Charterhouse Square, London EC1M 6BQ, UK
2. ²Department of Quantitative Health Sciences, The Cleveland Clinic Foundation, Cleveland, OH 44195, USA
3. ³Department of Histopathology, St Bartholomew's Hospital, London E1 2ES, UK
4. ⁴Department of Medical Oncology, St Bartholomew's Hospital, London EC1A 7BE, UK
5. ⁵Department of Cellular Pathology and Molecular Genetics, Liverpool University Hospital, Liverpool L1 3GA, UK
6. ⁶King's College, Thames Cancer Registry, London SE1 3QD, UK
7. ⁷Department of Pathology, Memorial Sloan Kettering Cancer Center, New York, NY 10021, USA
8. ⁸Department of Urology, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 10021, USA

Correspondence: Professor J Cuzick, E-mail: jack.cuzick@cancer.org.uk

⁹Principal investigators.

Received 15 May 2006; Revised 10 August 2006; Accepted 10 August 2006.

Abstract

Optimal management of clinically localised prostate cancer presents unique challenges, because of its highly variable and often indolent natural history. There is an urgent need to predict more accurately its natural history, in order to avoid unnecessary treatment. Medical records of men diagnosed with clinically localised prostate cancer, in the UK, between 1990 and 1996 were reviewed to identify those who were conservatively treated, under age 76 years at the time of pathological diagnosis and had a baseline prostate-specific antigen (PSA) measurement. Diagnostic biopsy specimens were centrally reviewed to assign primary and secondary Gleason grades. The primary end point was death from prostate cancer and multivariate models were constructed to determine its best predictors. A total of 2333 eligible patients were identified. The most important prognostic factors were Gleason score and baseline PSA level. These factors were largely independent and together, contributed substantially more predictive power than either one alone. Clinical stage and extent of disease determined, either from needle biopsy or transurethral resection of the prostate (TURP) chips, provided some additional prognostic information. In conclusion, a model using Gleason score and PSA level identified three subgroups comprising 17, 50, and 33% of the cohort with a 10-year prostate cancer specific mortality of <10, 10–30, and >30%, respectively. This classification is a substantial improvement on previous ones using only Gleason score, but better markers are needed to predict survival more accurately in the intermediate group of patients.