Purpose: Free and total serum prostate specific antigen (PSA) levels are frequently used for early prostate cancer detection in men with a total PSA in the 2.6 to 10 ng/ml range. Results from a longitudinal study suggested that the free-to-total PSA ratio (F/T PSA), declining the decade before prostate cancer diagnosis, is the earliest serum marker predicting a subsequent diagnosis of prostate cancer. Other than this study, there are limited data on the predictive value of longitudinal F/T PSA measurements for subsequent cancer. To evaluate the clinical importance of a decreasing F/T PSA, we compared longitudinal F/T PSA in volunteers in a prostate cancer screening study.
Materials and methods: We evaluated 657 volunteers in a screening study with a total PSA of 2.6 to 10 ng/ml but whose initial biopsies were either negative for cancer or were waived. To obtain meaningful F/T PSA slopes, the patients have been followed with at least 3 F/T PSA levels measured for 3 or greater years (mean 3.5, median 3.5, range 3 to 5.3). Of these men 22 subsequently were diagnosed with prostate cancer. We calculated the F/T PSA slope as the last F/T PSA minus the first F/T PSA divided by the interval between the measurements. The F/T PSA slopes of the patients with and without cancer were compared using chi-square analysis.
Results: Mean F/T PSA slope for the patients with cancer was -0.06 (median -0.03, range -0.63 to 0.22) compared to mean slope of -0.02 (median -0.01, range -0.57 to 0.69) for patients without cancer (p = 0.21). Of the patients with cancer 61% had a negative slope compared with 54% of patients without cancer (p = 0.59). The positive predictive value of a negative F/T PSA slope for subsequent cancer detection to date is only 4%.
Conclusions: A decreasing F/T PSA during a 3 to 5-year interval has little value in predicting subsequent diagnosis of prostate cancer. Further followup of a larger number of patients will be required to determine the value of longitudinal F/T PSA measurements in early prostate cancer detection.