Variable(s) OR 95% CI P value Sensitivity Specificity Hosmer-Lemeshow test (Pr>c2) |β| AUC Univariate analysis KLK3 1.575 [1.229-2.020] <0.0001 73.53% 60.87% 0.398 0.851 0.837 KLK11 1.579 [1.187-2.101] 0.0003 67.65% 62.86% 0.913 0.563 0.739 miR-9-3p 0.719 [0.588-0.881] 0.0002 81.08% 53.33% 0.359 0.580 0.769 miR-19a-3p 0.820 [0.718-0.935] 0.001 76.47% 64.29% 0.091 0.494 0.723 Bivariate analysis KLK3+KLK11 KLK3 1.500 [1.167-1.928] < 0.0001 85.29% 65.22% 0.692 0.758 0.860 KLK11 1.331 [0.920-1.926] 0.307 KLK3+miR-9-3p KLK3 1.631 [1.200-2.216] < 0.0001 87.10% 72.22% 0.106 0.929 0.877 miR-9-3p 0.827 [0.612-1.118] 0.293 KLK3+miR-19a-3p KLK3 1.833 [1.289-2.607] < 0.0001 85.71% 72.22% 0.312 1.127 0.880 miR-19a-3p 0.932 [0.764-1.137] 0.177 KLK11+miR-9-3p KLK11 1.420 [1.016-1.985] 0.0001 86.67% 71.43% 0.546 0.404 0.828 miR-9-3p 0.729 [0.576-0.924] 0.567 KLK11+miR-19a-3p KLK11 1.435 [1.006-2.049] 0.0004 78.57% 76.92% 0.500 0.395 0.811 miR-19a-3p 0.803 [0.688-0.937] 0.529

Table 3: Logistic regression analysis of variables predicting prostate cancer. Differentially expressed between PCa and BPH group (KLK3, KLK11, miR-9-3p and miR-19a-3p) were included. Prediction models are based on mRNA expression of each gene. |β|, absolute value of regression coefficient; OR, odds ratio; CI, confidence interval; AUC, Area Under Curve.