Variability of repeated serum prostate-specific antigen (PSA) measurements within less than 90 days in a well-defined patient population.

OBJECTIVES

To determine the variability of repeated serum prostate-specific antigen (PSA) measurements within less than 90 days in a well-defined patient population.

METHODS

A retrospective review of the PSA database at the Dallas Veterans Affairs Medical Center in Dallas was performed to identify patients who had two serum PSA measurements within less than 90 days, with the first PSA being less than 10 ng/mL (monoclonal assay, Abbott IMx). Patients' age and the dates and results of the PSA 1 and 2 measurements were captured in a database. Charts were reviewed on all patients, and those who had undergone a manipulation likely to alter the PSA value either before PSA 1 or between the two PSA measurements were excluded. The results of digital rectal examination (DRE) were classified as follows: no pathologic condition, benign prostatic hyperplasia, or suspected carcinoma of the prostate. The data were stratified in a variety of ways and analyzed to determine the variability of repeated PSA measurements under varying conditions.

RESULTS

A total of 295 men were identified who fulfilled the conditions. Mean age was 66.3 +/- 8.3 (standard deviation [SD]) years, and the mean PSA 1 was 2.7 +/- 2.4 (SD) ng/mL. When stratified by whether PSA 1 was 4.0 or less or from 4.1 to 10.0 ng/mL, the mean values (1.41 versus 1.43 and 6.00 versus 5.89 for PSA 1 and 2, respectively) were not significantly different. Similarly, when stratified by whether PSA 2 was obtained within 30 days, 30 to 60 days, or 60 to 90 days, there was no significant difference between the mean values for PSA 1 and 2. When stratified by decade of life, there were no differences between PSA values for any decade, although a clear relationship was seen between mean PSA and age (less than 50 years: 1.39 versus 1.06; 50 to 60 years: 1.89 versus 1.70; 60 to 70 years: 2.47 versus 2.48; 70 to 80 years: 3.65 versus 3.70; more than 80 years: 3.45 versus 3.56). A stratification by results of the DRE (1 = normal, 2 = benign prostatic hyperplasia, 3 = prostate cancer suspected) yielded the following values: DRE 1: 2.40 versus 2.47; DRE 2: 2.99 versus 2.75; DRE 3: 3.64 versus 3.81; the difference was not significant for all three groups. Forty-six percent of patients either had an identical PSA or an increase in the PSA, and 54% had a decrease. One third of the patients had a difference of greater than +/- 1.0 ng/mL. The largest differences noted were -5.3 and +7.5 ng/mL. Cumulative distributions of differences were calculated, and the patients were stratified by time intervals, age, DRE findings, and PSA 1 values. With the exception of the latter, there were no significant differences noted in the other three stratifications.

CONCLUSIONS

There is a significant variability between two serum PSA measurements obtained within a short-time interval, which is due to chance alone. These results highlight the problem when relying on a single PSA measurement and using either a single cutoff, age-specific reference ranges, or a rate of change to trigger further diagnostic tests, such as a biopsy. Furthermore, they raise the question whether two measurements and in case of discrepancy a third measurement should precede any further recommendation or invasive testing. Decision aids are offered to physicians to select ranges of PSA within which they may wish to repeat a PSA test, depending on calculated probabilities for the second PSA to cross the predefined cutoff value.