Predictive modeling for the presence of prostate carcinoma using clinical, laboratory, and ultrasound parameters in patients with prostate specific antigen levels ≤ 10 ng/ml

BACKGROUND. The objective of the current study was to develop a model for predicting the presence of prostate carcinoma using clinical, laboratory, and transrectal ultrasound (TRUS) data. METHODS. Data were collected on 1237 referred men with serum prostate specific antigen (PSA) levels ≤ 10 ng/mL who underwent an initial prostate biopsy. Variables analyzed included age, race, family history, referral indication(s), prior vasectomy, digital rectal examination (DRE), PSA level, PSA density (PSAD), and TRUS findings. Twenty percent of the data were reserved randomly for study validation. Logistic regression analysis was performed to estimate the relative risk, 95% confidence interval, and P values. RESULTS. Independent predictors of a positive biopsy result included elevated PSAD, abnormal DRE, hypoechoic TRUS finding, and age 75 years or older. Based on these variables, a predictive nomogram was developed. The sensitivity and specificity of the model were 92% and 24%, respectively, in the validation study for which the predictive probability ≥ 10% was used to indicate the presence of prostate carcinoma. The area under the receiver operating characteristic curve (AUC) for the model was 73%, which was significantly higher compared with the prediction based on PSA alone (AUC, 62%). If it was validated externally, then application of this model to the biopsy decision could result in a 24% reduction in unnecessary biopsy procedures, with an overall reduction of 20%. CONCLUSIONS. Incorporation of clinical, laboratory, and TRUS data into a prebiopsy nomogram significantly improved the prediction of prostate carcinoma over the use of individual factors alone. Predictive nomograms may serve as an aid to patient counseling regarding prostate biopsy outcome and to reduce the number of unnecessary biopsy procedures.