Background: The distribution of prostate-specific antigen (PSA) values for men with or without prostate carcinoma are confounded because of verification bias. Correcting for verification bias, the means and variances of PSA values were estimated in specific clinical scenarios.
Methods: Existing receiver operating characteristic (ROC) curves, adjusted for the presence of verification bias in a screening population, were used to estimate the mean and variance of PSA values for men with or without prostate carcinoma, stratified by age and the presence or absence of benign prostatic hyperplasia. Men with a suspicious digital rectal exam (nodular) were excluded from analysis.
Results: Among men with cancer and the absence of benign prostatic hyperplasia, mean PSA values were 2.05 ng/mL and 2.66 ng/mL for younger (<60 yr) and older (> or =60 yrs) men, respectively. These estimates were 2.56 ng/mL and 3.90 ng/mL in the presence of benign prostatic hyperplasia for younger and older men, respectively. For men without prostate carcinoma, these values were 0.78 ng/mL and 1.23 ng/mL for younger and older men, respectively, among those without benign prostatic hyperplasia, and 0.97 ng/mL and 1.75 ng/mL for younger and older men, respectively, among those with benign prostatic hyperplasia.
Conclusions: Accurate estimates of the mean and variance of PSA values for men with or without cancer may provide PSA thresholds for biopsy that are specific for age and prostate size as assessed by digital rectal exam. Therefore, the current threshold of 4.0 ng/mL should not be applied indiscriminately.
Copyright 2006 American Cancer Society.