A risk index for prostate cancer

BACKGROUND: The aim of the present study was to create a simple numerical index predicting the presence of prostate cancer in a group of high risk patients, for the purpose of selecting those most likely to need prostate biopsy. METHODS: 100 consecutive patients at high risk of having prostate cancer seen at Ramathibodi Hospital, Thailand between November 2000 and February 2002 were prospectively studied. All patients underwent transrectal prostate biopsies. The following predictor variables were obtained: age, digital rectal examination (DRE) findings, prostate specific antigen level, transrectal ultrasonography (TRUS) findings, and prostate volume determined by TRUS. The outcome was the presence of prostate cancer on histological examination of the biopsy specimens. A risk index for prostate cancer based on the linear predictor of a multiple logistic regression model was created. RESULTS: Almost all predictor variables were significantly related to the presence of prostate cancer. The final multiple logistic regression model with four categorized predictors (excluding DRE) was shown to have good discrimination, calibration, and cross-validity. For a cutoff risk index of 10, corresponding to a 10% probability of having prostate cancer, the sensitivity for detecting prostate cancer was 96.2%, with a specificity of 73.0%. Based on this cutoff, 55% of patients in this series might not require prostate biopsy. CONCLUSION: A risk index for prostate cancer was developed. If this index can be externally validated, the potential savings from avoiding unnecessary prostate biopsies, on the basis of selection using the index, could be significant.