PSA level had moderate sensitivity for prostate cancer presenting within 4 years
ACP J Club. 1995 July-Aug;123:19. doi:10.7326/ACPJC-1995-123-1-019
Gann PH, Hennekens CH, Stampfer MJ. A prospective evaluation of plasma prostate-specific antigen for detection of prostatic cancer. JAMA. 1995 Jan 25;273:289-94.
To determine the diagnostic accuracy of plasma prostate-specific antigen (PSA) in men who subsequently develop prostate cancer.
Blinded comparison of PSA levels (from blood drawn at the beginning of the Physicians' Health Study) in men with a subsequent diagnosis of prostate cancer with men who did not develop prostate cancer (nested case-control study).
366 male physicians (cases) with stored plasma at baseline and a diagnosis of prostate cancer during follow-up and 1098 (3 controls per case) male physicians, matched by age, randomly selected from the entire cohort of 22 071 male physicians who had stored plasma and were disease-free at the time of case ascertainment. The Physicians' Health Study excluded those with a history of myocardial infarction, stroke, transient ischemic attacks, unstable angina, or cancer; current renal or liver disease; peptic ulcer or gout; contraindication to aspirin; or current use of aspirin, other platelet-active agents, or vitamin A.
Description of test and diagnostic standard
Blinded plasma samples were analyzed for PSA. Incident cases of prostate cancer were identified during 10 years of follow-up. Medical records were reviewed by a study physician unaware of the baseline PSA results. Disease stage at diagnosis, tumor grade, Gleason score, type of presentation, PSA level immediately before treatment, and types of treatment were recorded.
Main outcome measures
Sensitivity, specificity, and lead-time to diagnosis.
Of 366 prostate cancers identified, 183 were classified as aggressive, 160 as nonaggressive, and 23 as indeterminate. At a PSA abnormal cutoff level of 4.0 ng/mL, sensitivity for the entire 10-year follow-up period was 46% (95% CI 41% to 52%) (Table) and decreased with duration of follow-up. The sensitivities for detection of total, aggressive, and nonaggressive cancers that developed in the first 4 years of follow-up were 73%, 87%, and 53%, respectively. Overall specificity was 91% (CI 89% to 93%) (Table) and changed little by year of follow-up. Maximum accuracy over 7 years of follow-up was obtained at a PSA cutoff level of 3.3 ng/mL. The baseline PSA level could have moved the diagnosis of prostate cancer forward by an average of 5.5 years.
A single prostate-specific antigen measurement had moderate sensitivity and higher specificity for detecting prostate cancers.
Source of funding: National Institutes of Health.
For article reprint: Dr. P.H. Gann, Department of Preventive Medicine, Northwestern University Medical School, 680 North Lake Shore Drive, Suite 1102, Chicago, IL 60611, USA. FAX 312-908-9588.
Table. Test characteristics of prostate-specific antigen with abnormal cutoff point of 4.0 ng/ml for detecting prostate cancer*
|Outcome||Sensitivity (95% CI)||Specificity (CI)||+LR||-LR|
|Prostate cancer||46% (41 to 52)||91% (89 to 92)||5.1||0.59|
*Abbreviations defined in Glossary; LRs calculated from data in article.
The true sensitivity and specificity of PSA are unknown because no large study has used a gold standard, such as biopsy, to define which patients tested with PSA have or do not have cancer. Using their novel case-control approach, Gann and colleagues document that 87% of men with aggressive cancers appearing within 4 years had an abnormal PSA result at baseline compared with only 53% of those with nonaggressive cancers.
The aggressive cancers in this study were unlikely to be cured at the actual time of diagnosis with aggressive therapy. If a PSA had been done and responded to several years earlier, how many cancers would have been curable? The answer is unknown. Because the time of diagnosis would have been advanced, however, the prognosis would have seemed to be better (lead-time bias). This study documented a much lower sensitivity for nonaggressive cancers, which are believed by advocates of early detection to be the potentially curable precursors of future aggressive cancers and the true targets of screening.
Gann and colleagues also document a false-positive rate (FPR) of 9%; however, the mean age of participants at baseline was only 63 years. The FPR is even higher among older men with more concurrent benign prostatic hyperplasia. In a recent large screening study of older men, about 12% in their 60s and 18% in their 70s had negative systematic biopsy findings after an elevated PSA result (1). These FPRs are higher than those for other screening tests that have better evidence of efficacy, such as stool guaiacs (FPR, 1% to 4% without rehydration; 10% with rehydration) (2) or mammography (FPR, 5% to 7%) (3).
It is unknown whether PSA will decrease mortality in the trials that have just started. What should clinicians do in the meantime? Patients should be informed about the downstream consequences of PSA testing and the uncertainty about whether it does more good than harm (4). Some will want the test in these circumstances; many will not.
Michael J. Barry, MD
Massachusetts General HospitalBoston, Massachusetts, USA