Objective: To show the impact of learning curve and patient selection on complication rate and biochemical recurrence-free survival of a UK radical prostatectomy series for localised prostate cancer and to model the influence of common preoperative variables on biochemical recurrence after controlling for learning curve. Patients and Methods: From 1989 to 1999, 280 of 350 patients who underwent anatomical radical retropubic prostatectomy (RRP) at our institution had complete records and follow-up of at least 1y. After exclusions of preoperative staging, factors reflecting the learning curve, early complications and prostate-specific antigen (PSA) outcome were recorded on 217 patients. Procedures before 1995 were compared with procedures after 1995. Results: Comparison of the two groups showed a significant decrease in operating time (mean 152 vs 130min), blood loss (mean 1500 vs 1000ml), transfusion rate (83 vs 42%) and hospital stay (mean 7 vs 6days). Median preoperative PSA changed significantly from 13.2 to 11.5ng/ml. Only 17% were diagnosed by rectal examination compared to 27% in the early years. The number of clinical T1 tumours increased from 33 to 47%. This did lead to an increase of organ-confined tumours on pathological staging by 25% Biochemical recurrence-free survival improved significantly after 1995. After controlling for the learning curve PSA and clinical stage were significant predictors of PSA recurrence. Conclusion: Time trends of case selection, stage migration and a steep learning curve are shown over a 10-y period. Factors associated with the learning curve as well as case selection have a significant impact on outcome. There may be other as yet not specified factors over time, which have a significant impact on PSA recurrence-free survival. Patients with a PSA of 20ng/ml and above have a poor outcome and do not appear to be suitable candidates for RRP.