Interpreting results of prostate-specific antigen testing for early detection of prostate cancer.J Gen Intern Med 1996; 11(9):505-12JG
Calculation of likelihood ratios for serum prostate-specific antigen (PSA) levels to discriminate potentially curable prostate cancer in men selected for having benign prostatic hyperplasia (BPH) or in randomly selected men.
Retrospective analysis of prospectively measured PSA levels.
A tertiary referral center, a multicenter randomized controlled trial, and a community-based study, all providing PSA data.
We used PSA measurements from four groups of men aged 50 to 79 years: 276 men with organ-confined prostate cancer treated with radical prostatectomy, 305 randomly selected men without clinical evidence of prostate cancer or a history of surgery for BPH recruited for a community study, 173 men without cancer but with BPH coming to prostatectomy, and 770 men without cancer and with symptoms of BPH enrolled in the North American finasteride clinical trial.
MEASUREMENTS AND MAIN RESULTS
Age-standardized, stratum-specific likelihood ratios for organ-confined prostate cancer were calculated separately for unselected men in the community sample and for selected men with BPH (pooling both BPH populations). Likelihood ratios ranged from 0.2 for PSA between 0.0 and 2.0 ng/mL to 54.8 for a PSA level of 10.1 ng/mL or higher in unselected men, but rose to only 2.9 for PSA values of 10.1 ng/mL or higher in men with BPH. Forty percent of the men in the community study had moderate to severe lower urinary tract symptoms. In these men, likelihood ratios ranged from 0.2 for PSA values between 0.0 and 2.0 ng/mL to 17.2 for PSA values of 6.1 ng/mL or higher, while in men with no or mild symptoms, likelihood ratios rose to 26.9 for PSA values of 6.1 ng/mL or higher.
Likelihood ratios for PSA test results allow stratification of men along a continuum of risk for prostate cancer. Likelihood ratios demonstrate that the ability of the PSA test to discriminate potentially curable prostate cancer from BPH is dramatically lower in men selected with lower urinary tract symptoms than in randomly selected men.