Prostate Cancer Staging Tables—A Predictive Model for the UK

Abstract

Introduction: The use of accurate risk stratification is a prerequisite to informed decision-making when considering potentially curative treatments for prostate cancer. Most models are derived from cases managed in the United States. The validity of these methods may be compromised when used on a population other than that used for generating the predicted outcomes. We present predictive tables derived from the observed outcomes of men treated by radical prostatectomy in the United Kingdom. Methods: Using logistic regression a pilot study identified the best predictors of pathological stage from eight pre-operative variables. All full BAUS members were asked to submit their consecutive RP patients' age, biopsy Gleason score, pre-operative PSA, number of biopsy cores, number of biopsy cores containing cancer (% positive cores) and pathological stage. Predictive tables were constructed using this data to predict pT2, pT3a or pT3b/4/N1 disease at radical prostatectomy and assessed using internal cross-validation methods. Results: 1912 patients undergoing radical prostatectomy by 39 consultant urologists in 19 centres were included. The impact of age was equivocal but a robust model was developed to predict outcomes based on Gleason sum score, pre-operative PSA and positive biopsy cores. A series of tables have been constructed to allow for use in practice. Conclusions: In this study we have generated a validated prostate cancer predictive table derived entirely from a UK surgical cohort and which is simple to use.