Table 5.
Logistic regression analyses predicting probability of prostate cancer in subjects with tPSA 4–10 ng/mL
| Predictors | AUC of individual predictor variable (95% CI); P value* |
Bivariate analysis OR(95%CI); P value |
Multivariate analysis | ||
|---|---|---|---|---|---|
| Base model ** OR (95%CI); P value |
Base model + %p2PSA OR (95% CI); P value |
Base model + PHI OR(95%CI); P value |
|||
| Age |
0.63 (0.51–0.74) 0.235 |
1.069 (1.01–1.13); 0.031 |
1.081 (1.02–1.15);0.013 |
1.084 (1.01–1.16);0.018 |
1.084 (1.01–1.16);0.018 |
| tPSA, ng/mL |
0.50 (0.38–0.61) – |
1.024 (0.76–1.38); 0.875 |
1.615 (0.66–3.86); 0.281 |
1.384 (0.49 − 3.88); 0.536 |
1.243 (0.44–3.48); 0.679 |
| fPSA |
0.59 (0.47–0.70) 0.418 |
0.529 (0.18–1.53); 0.240 |
0.038 (0.00–11.2); 0.261 |
0.070 (0.00–53.9); 0.433 |
0.071 (0.00–56.5); 0.437 |
| %fPSA |
0.63 (0.52–0.74) 0.090 |
0.961 (0.89–1.03); 0.284 |
1.155 (0.80–1.66); 0.448 |
1.108 (0.73–1.69); 0.631 |
1.108 (0.73–1.69); 0.632 |
| p2PSA, pg/mL |
0.59 (0.47–0.70) 0.286 |
1.023 (1.00–1.05); 0.089 |
– | – | – |
| %p2PSA |
0.70 (0.59–0.81)0.038 |
1.020 (1.00–1.04); 0.007 |
– |
1.022 (1.01–1.04);0.007 |
– |
| PHI |
0.70 (0.59–0.81)0.020 |
1.743 (1.19–2.56); 0.005 |
– | – |
1.760 (1.17–2.64);0.006 |
| AUC of multivariate models (95% CI); | – | – |
0.682 (0.57–0.78) 0.003 |
0.784 (0.68–0.87) < 0.001 |
0.787 (0.68–0.87) < 0.001 |
|
Gain in predictive accuracy (95% CI); P value |
– | – | – |
0.102 (− 0.00–0.21); 0.054 |
0.104 (0.00–0.21); 0.048 |
AUC, area under the receiver operating characteristic curve; PSA, prostate-specific antigen; fPSA, free PSA; p2PSA, [-2]proPSA; PHI, prostate health index; tPSA, total PSA, OR, odds ratio; CI, confidence interval
*P value: Comparison of AUC using tPSA as standard
**Base model includes age, tPSA, fPSA and %fPSA
P value to be statiscally significant with bolditalics