Comparative Performance of Prostate Cancer Risk Prediction Tools in Gay, Bisexual, and Other Men Who Have Sex with Men Versus Heterosexual Men.
[OBJECTIVE] To compare the accuracy of clinical models and MRI-based prostate cancer (PCa) risk calculators in men who have sex with men (MSM) versus heterosexual men (HSM).
- p-value p=0.02
- Sensitivity 94%
APA
Fink KE, Liu Y, et al. (2026). Comparative Performance of Prostate Cancer Risk Prediction Tools in Gay, Bisexual, and Other Men Who Have Sex with Men Versus Heterosexual Men.. Urology. https://doi.org/10.1016/j.urology.2026.04.015
MLA
Fink KE, et al.. "Comparative Performance of Prostate Cancer Risk Prediction Tools in Gay, Bisexual, and Other Men Who Have Sex with Men Versus Heterosexual Men.." Urology, 2026.
PMID
41985632
Abstract
[OBJECTIVE] To compare the accuracy of clinical models and MRI-based prostate cancer (PCa) risk calculators in men who have sex with men (MSM) versus heterosexual men (HSM). MSM comprise 3-6% of U.S. men yet remain underrepresented in validation studies of PCa risk tools. Behavioral and anatomic factors may influence the performance of PSA, digital rectal examination (DRE), and prostate MRI.
[METHODS] We analyzed 486 men (243 MSM, 243 HSM) undergoing prostate biopsy at a single academic center, matched by age, PSA, race/ethnicity, and MRI status. Sexual orientation, demographics, imaging, laboratory data, DRE findings, and biopsy results were extracted from electronic health records and validated by chart review. MRI-derived prostate volume, PSA density, and PI-RADS scores were included when available. Logistic regression models using PSA, DRE, PSA+DRE, PI-RADS, and PI-RADS+DRE generated predicted probabilities of clinically significant PCa (csPCa; Grade Group ≥2). Predictions from PLUM, UCLA, and My nMRIsk were evaluated using area under the curve (AUC), calibration, decision-curve analysis, and biopsy outcomes at 10% and 30% thresholds.
[RESULTS] MSM and HSM were similar in age, PSA, prostate volume, and PI-RADS distribution. Abnormal DRE was more frequent in MSM (12.8% vs 6.6%, p=0.02) but was not associated with higher csPCa risk. PI-RADS ≥3 demonstrated higher sensitivity in MSM (94% vs 87%). MRI-based calculators showed higher AUCs in MSM, with best performance for PLUM and UCLA.
[CONCLUSIONS] MRI-based risk calculators demonstrated higher diagnostic accuracy in MSM, supporting use of current PCa risk tools without recalibration.
[METHODS] We analyzed 486 men (243 MSM, 243 HSM) undergoing prostate biopsy at a single academic center, matched by age, PSA, race/ethnicity, and MRI status. Sexual orientation, demographics, imaging, laboratory data, DRE findings, and biopsy results were extracted from electronic health records and validated by chart review. MRI-derived prostate volume, PSA density, and PI-RADS scores were included when available. Logistic regression models using PSA, DRE, PSA+DRE, PI-RADS, and PI-RADS+DRE generated predicted probabilities of clinically significant PCa (csPCa; Grade Group ≥2). Predictions from PLUM, UCLA, and My nMRIsk were evaluated using area under the curve (AUC), calibration, decision-curve analysis, and biopsy outcomes at 10% and 30% thresholds.
[RESULTS] MSM and HSM were similar in age, PSA, prostate volume, and PI-RADS distribution. Abnormal DRE was more frequent in MSM (12.8% vs 6.6%, p=0.02) but was not associated with higher csPCa risk. PI-RADS ≥3 demonstrated higher sensitivity in MSM (94% vs 87%). MRI-based calculators showed higher AUCs in MSM, with best performance for PLUM and UCLA.
[CONCLUSIONS] MRI-based risk calculators demonstrated higher diagnostic accuracy in MSM, supporting use of current PCa risk tools without recalibration.