Table 3 Diagnostic performance.
Indicator | Cutoff | Sensitivity (95% CI) | Specificity (95% CI) | PPV | NPV | AUC (95% CI) | P for AUC |
---|---|---|---|---|---|---|---|
Training cohort | |||||||
PSAD | 0.162 | 73.3% (63.0–82.4%) | 62.1% (55.8–68.5%) | 41.8% (33.9–50.1%) | 86.5% (80.4–91.2%) | 0.712 (0.648–0.775) | P < 0.001a |
PI-RADS | ≥ 3 | 95.4% (88.8–98.2%) | 55.9% (49.6–62.1%) | 44.4% (37.1–51.8%) | 97.1% (92.6–99.2%) | 0.875 (0.834–0.916) | 0.01a |
≥ 4 | 86.2% (77.4–91.9%) | 84.7% (79.6–88.8%) | 67.6% (58.0–76.1%) | 94.3% (90.3–97.0%) | |||
PSAD + PI-RADS | – | 84.7% (75.8–91.8%) | 87.4% (82.4–91.3%) | 71.2% (61.4–79.6%) | 94.1% (90.1–96.8%) | 0.893 (0.853–0.933) | – |
Validation cohort | |||||||
PSAD | 0.183 | 50.0% (31.9–68.1%) | 75.5% (65.8–83.6%) | 40% (24.9–56.7%) | 82.2% (72.7–89.5%) | 0.677 (0.569–0.785) | P < 0.001a |
PI-RADS | ≥ 3 | 96.9% (84.3–994%) | 56.1% (46.3–65.5%) | 41.9% (30.5–53.9%) | 98.2% (90.4–100%) | 0.829 (0.759–0.899) | 0.02a |
≥ 4 | 71.9% (54.6–84.4%) | 81.6% (72.9–88.1%) | 56.1% (39.7–71.5%) | 89.9% (81.7–95.3%) | |||
PSAD + PI-RADS | – | 83.7% (62.7–94.7%) | 84.4% (76.0–91.2%) | 64.3% (48.0–78.4%) | 94.3% (87.2–98.1%) | 0.871 (0.807–0.934) | – |