What is the 'normal range' for prostate-specific antigen? Use of a receiver operating characteristic curve to...

To compare the relative sensitivity and specificity of prostate-specific antigen (PSA) as a test for prostate cancer over a range of PSA values in a variety of patient groups, and to compare the sensitivity and specificity of PSA and prostatic acid phosphatase (PAP).

Receiver operating characteristic (ROC) curves (sensitivity plotted against 1-specificity) were constructed to compare the ability of PSA to discriminate men with prostate cancer (n = 257) from those with benign prostatic hyperplasia (BPH) (n = 220) or control patients (n = 164). Receiver operating characteristic curves were also constructed to compare PSA and PAP in 173 men with either BPH or prostate cancer.

When patients with symptomatic BPH and those with advanced prostate cancer are excluded, a PSA of 8 ng/mL has a sensitivity of 94% and a specificity of 98% for prostate cancer. In patients presenting with symptoms suggestive of bladder outflow obstruction, PSA remains a sensitive marker for prostate cancer (93% sensitivity at 10 ng/mL) but its specificity (65%) is poor. PSA is a sensitive test for skeletal metastases but levels of 60-80 ng/mL are required to achieve a specificity of 70% or more. The sensitivity of PSA is far superior to that of PAP.

Serum PSA provides good discrimination between patients with and without prostate cancer. The sensitivity and specificity of PSA can be improved by excluding men with symptomatic BPH. The specificity of PSA as a diagnostic test for prostate cancer is reduced in men with symptoms of bladder outflow obstruction. For reasonable sensitivity and specificity, a PSA of 60-80 ng/mL is required for differentiating non-metastatic from metastatic prostate cancer. The ROC curve comparing PSA and PAP provides a graphical demonstration of the superiority of PSA as a tumour marker. The ability of PSA to identify prostate cancer can be improved by selecting out groups of patients and by adjusting the cut-off level of PSA to the population under study.