Predicting Prostate Biopsy Results

Abstract

We studied whether age and race have a significant role in predicting the results of prostate biopsy after consideration of prostate-specific antigen (PSA) alone. We evaluated the results of 884 prostate biopsies performed on 695 men, including 188 black men. We used logistic regression analysis to evaluate relationships between presence of cancer in the biopsy specimen and the key predictor variables of PSA, age and black race. We also evaluated the importance of a Bayes-derived positive predictive value (PPV), which used age- and race-specific values of specificity and sensitivity of PSA and age- and race-specific Surveillance, Epidemiology, and End Results incidences of prostate cancer. In univariate analysis, the Bayes PPV was associated significantly with presence of cancer (chi2 = 216; P approximately .000); however, serum PSA provided more information (chi2 = 248; P approximately .000). With serum PSA in the logistic model, the Bayes PPV provided no further information (P > .08), and as additional variables, neither age nor black race contributed further information (P > .1). Serum PSA provides the most predictive information about the results of biopsy of the prostate, probably because it naturally correlates with age and race. Decisions about whether to biopsy should rest on values of serum PSA and need not consider age and race.