Abstract
BACKGROUND AND OBJECTIVE: Although the effects of 5-α-reductase inhibitors (5-ARIs) on prostate-specific antigen (PSA) have been well-studied, the effects on other kallikrein markers have not been firmly established. Accordingly, a statistical model based on blood measurements of four kallikreins-known commercially as the "4Kscore"-cannot be used for men taking 5-ARIs. We investigated how finasteride affects kallikrein markers and whether suitably adjusted marker levels could be used in the four-kallikrein model to accurately predict the prostate biopsy results. METHODS: We analyzed 500 participants from the Prostate Cancer Prevention Trial (PCPT) with PSA ≤3 ng/ml before finasteride and who had marker measurements after 1 yr on finasteride. Conversion factors were generated from this cohort to estimate prefinasteride marker levels in a separate PCPT cohort of 459 men on finasteride biopsied for cause. Adjusted marker levels were entered into the prespecified 4K model and performance characteristics assessed for predicting high-grade prostate cancer on biopsy. KEY FINDINGS AND LIMITATIONS: Finasteride use halved total PSA (β 0.51, 95% confidence interval 0.49, 0.54). human kallikrein 2 was also halved (β 0.50); free PSA and intact PSA were slightly more than halved (β 0.44 for both). Predictions from the 4K model using adjusted markers improved discrimination over adjusted total PSA alone (AUC 0.734 vs 0.595; p < 0.001). While this cohort underwent only sextant biopsy, the ProtecT model, created on 10-12 core biopsies was still well-calibrated at clinically important thresholds. In decision-curve analysis, the 4K model had the highest net benefit for risk thresholds of ≥7%. CONCLUSIONS AND CLINICAL IMPLICATIONS: The 4K model can be used to inform prostate biopsy decision-making in men taking 5-ARIs for at least 3 mo by incorporating adjusted kallikrein levels into the scoring algorithm.