Table 3.

Full-cohort internal validation of the final logistic regression model.

Validation analysis	n/CSPCa events	AUROC (95% CI)	Brier score	Calibration intercept	Calibration slope	Sensitivity (95% CI)	Specificity (95% CI)
5-fold cross-validated out-of-fold predictions	193/41	0.840 (0.754-0.916)	0.105	-0.122	0.917	0.512 (0.365-0.657)	0.967 (0.925-0.986)
Bootstrap optimism-corrected model	193/41	0.862	0.097	0.019	1.013	NA	NA
Apparent full-cohort model	193/41	0.867 (0.790-0.936)	0.092	0.077	1.064	0.512 (0.365-0.657)	0.980 (0.944-0.993)

The final logistic regression model was presented using age, PHI, and PHID. Five-fold cross-validated estimates were derived from out-of-fold predicted probabilities in the full cohort, with LASSO feature selection repeated within each outer training fold. Bootstrap optimism-corrected estimates were obtained from 1,000 bootstrap resamples. The Brier score measures overall probabilistic prediction error rather than calibration alone. Calibration intercept and slope were estimated by regressing the observed outcome on the logit of the predicted probability. Classification metrics were calculated using a probability threshold of 0.50. AUROC, area under the receiver operating characteristic curve; CI, confidence interval; CSPCa, clinically significant prostate cancer; PHI, Prostate Health Index; PHID, PHI density.