Intraindividual variation of prostate specific antigen measurement and implications for early detection of prostate carcinoma

Abstract

BACKGROUND. Many researchers have described intraindividual prostate specific antigen (PSA) variation, but its implications for a prostate carcinoma screening program have not been well appreciated. In this study, the author explored misclassification problems associated with intraindividual PSA variation. METHODS. The previous research work on intraindividual PSA variation was reviewed and summarized. Then, on the basis of the results from the previous work and a simple measurement error model, the author estimated misclassification probabilities for a single PSA measurement. RESULTS. With a variance-based approach of meta-analysis, the author derived a summary coefficient of variation (13.1%) of the intraindividual PSA variation from the previous relevant studies. They found that there were substantial risks for misclassification, especially among those with PSA values near the recommended cutoff point (4.0 ng/mL). For an individual with a true PSA level at 3.9 ng/mL, the risk for misclassification as PSA higher than 4.0 ng/mL was approximately 42%; for an individual with a true PSA value at 3.0 ng/mL, the risk was less than 1%. CONCLUSIONS. Intraindividual PSA variation should become a part of interpreting PSA test results, especially for men with a PSA value near the cutoff point. Both sensitivity and specificity of PSA test reported in the literature have been underestimated, and the risk of prostate carcinoma associated with PSA test results has been biased down to the null (no effect) because of the intraindividual PSA variation. © 2001 American Cancer Society.