A prospective evaluation of plasma prostate-specific antigen for detection of prostatic cancer

Abstract

Objective: To evaluate the validity of prostate-specific antigen (PSA) in identifying men who subsequently were or were not clinically diagnosed with prostate cancer, assess optimal test cutoff, measure lead time, and estimate relative risks (RRs) associated with discrete PSA levels.

Designs: Nested case-control study of men providing plasma samples before a 10-year follow-up.

Setting: The Physicians' Health Study, an ongoing randomized trial that enrolled 22,071 men aged 40 to 84 years in 1982.

Participants: A total of 366 men (cases) diagnosed with prostate cancer and 1098 men (three controls per case), matched by age, randomly selected from all cohort members at risk at the time of case diagnosis.

Main outcome measures: Sensitivity and specificity for each year of follow-up and for aggressive and nonaggressive cancers separately.

Results: At a cutoff of 4.0 ng/mL, sensitivity for the entire 10-year follow-up was 46% for total cases. Sensitivities for detection of total, aggressive, and nonaggressive cancers occurring in the first 4 years were 73%, 87%, and 53%. Overall, specificity was 91% and changed little by year of follow-up. Optimal validity was achieved at a cutoff of 3.3 ng/mL. Estimated mean lead time for all cancers was 5.5 years. Only 40% of cancers detected more than 5 years from baseline were nonaggressive. Compared with men with PSA levels less than 1.0 ng/mL, those with PSA levels between 2.0 and 3.0 ng/mL had an RR of 5.5 (95% confidence interval, 3.7 to 9.2).

Conclusions: A single PSA measurement had a relatively high sensitivity and specificity for detection of prostate cancers that arose within 4 years. Prostate-specific antigen values less than the usual cutoff were associated with substantial increases in risk compared with the lowest levels. Final evaluation of PSA screening must also consider cost and the ability of current treatments to improve the prognosis of screen-detected cases.