Prostate-specific antigen utilization in Ontario: extent of testing in patients with and without cancer

To ascertain the extent of prostate-specific antigen (PSA) testing in patients with prostate cancer (PC), with other cancers (OC), and with no cancer (NC) in two clinical laboratory databases.

PSA test records were obtained from a tertiary care hospital, Sunnybrook Health Science Centre (SHSC) and from a private laboratory, Gamma-Dynacare Medical Laboratories (GDL), during the period 1988 to 1995. These records were linked with the Ontario Cancer Registry (OCR) to establish a diagnosis of PC, OC, or NC. Trends in PSA testing according to diagnostic category, testing laboratory, patient age (by decade), and PSA value (in microgram/L) were determined.

Major cancer sites identified in the patients tested for PSA were prostate (60%), bladder and colon (7% each), lung (5%), kidney (3%), and rectum (3%). There were 11,867 patients (8.5%) with PC, 8,002 (5.9%) with OC, and 118,954 (86%) with NC. The total number of PSA tests performed on these patients was 230,756, of which 21% were on PC, 5% on OC, and 74% on NC; of these tests, 64% were performed through GDL and 36% through SHSC. The mean (median) number of tests per patient was: PC, 4.0 (2); OC, 1.4 (1); and NC, 1.5 (1). For PC 89% and for OC 72% of all tests occurred after diagnosis. Between 1990 and 1995 the number of PSA tests increased two-fold in PC and OC, and 20-fold in NC. We estimate that about one-half of the PSA tests in the NC group were for screening purposes. The proportion of PSA tests occurring in PC, OC, and NC for patients 50 to 70 years of age was 41%, 50%, and 63%, respectively; for patients over 70 years of age, this proportion was 58%, 46%, and 22% respectively; and for patients under 50 it was 1%, 4%, and 15%, respectively. Between 1990 and 1995, the largest increase in testing frequency was in the NC group, particularly in patients 50 to 70 years of age, which was accompanied by a decrease in patients over 70. Less than 10% of testing occurred in patients under 50 in all diagnostic groups. We estimate that about 26% of PSA screening tests in NC occurred outside the guidelines for patient age. Between 1988 and 1995, the proportion of PSA results below our detection limit (< 0.2 micrograms/L) showed a steady rise in the PC group, as did the proportion between 0.2 and 3.9 micrograms/L; these were accompanied by a fall in the proportion > 20.0 micrograms/L. However, the proportion of PSA results within these ranges did not change much during the same time period for the OC and NC groups. At cutoffs of PSA = 4.0 micrograms/L (or PSA = 10.0 micrograms/L), estimates of clinical specificity were 84.0% (or 96.3%), and of clinical sensitivity were 83.4% (or 47.1%).

Most (86%) PSA testing occurred in men with NC, consistent with diagnosis or screening. There were more PSA tests per patient in PC than in OC, and most testing occurred after diagnosis. PSA testing in the NC group continues to increase rapidly. The proportion of PSA tests in patients over age 70 decreased in the order of PC > NC > OC. Between 1990 and 1995, there was an increase in the proportion of patients tested who were between 50 and 70 in the NC group, which may suggest more screening in this group. Over this same time period, there was an increase in the proportion of undetectable PSA values, possibly suggesting increased use of radical therapy; there was also a decrease in the proportion of PSA > 20 micrograms/L, possibly suggesting a decrease in the prevalence of advanced stage PC.