The diagnostic performance of prostate-specific antigen (PSA) in early detection of prostate cancer : considerations of sensitivity, specificity, lead-time and survival

Background: Prostate cancer (PC) is the most common form of cancer, as well as the single major cause of cancer-related mortality among Swedish males. Recent reports indicate that screening of the serum levels of prostate-specific antigen (PSA) in asymptomatic men may result in decreases in both the incidence of metastatic disease and PC-specific mortality. However, the diagnostic performance of PSA-based screening is limited by rather low sensitivity and specificity. An improved understanding of the natural course of PC in relationship to PSA levels at the time of diagnosis is required in order to calculate how much earlier diagnosis can be achieved by screening (the lead-time) and when the beneficial effect in the form of improved survival can be expected to become apparent. Patients and methods: This study is based on two well-defined cohorts of men both selected randomly from the general population. One of these cohorts (die Stockholm South PC study) included 1782 men who participated in a single intervention study for evaluation of early diagnostic methods in PC detection in 1988-89, whereas the other cohort (n=657) was involved in a screening of general health in men, not including PC, in 1980 (The Study of Men Born in 1913). The diagnostic performance of percentage free PSA (%fPSA) was evaluated in the Stockholm cohort. For estimations of lead-time and PC-specific survival, men from the two cohorts, with a median age of 67 years, were monitored for 12 and 20 years, respectively, following PSA testing. Individuals diagnosed as having PC as well as causes of mortality were ascertained utilizing the national Cancer Registry together with the Cause of Death Registry. Results and conclusions: Retrospective application of a %fPSA "cut-off 'value of < 18% improved the diagnostic sensitivity among men with normal PSA levels (i.e., < 3.0 ng/ml), whereas a "cut-off" of 22% increased the specificity at slightly elevated PSA levels (3.0-9.9 ng/mL). In men without PC the %fPSA ratios were influenced by age and prostate volume, which at PSA levels in the interval of 4.0-9.9 ng/mL together accounted for much of the variation in this ratio. Men with PSA levels < 1.0 ng/mL were at a very low risk of being diagnosed as having, or of dying from PC during 20 years of follow-up. In the PSA interval of 3.0-9.9 ng/mL the various estimates of median lead-time obtained ranged from 4.5-11.2 years. Successful screening for PC (i.e., decreased mortality) should be discernible