Table 5.
Comparison of the diagnostic performance in biopsy-naive patients with published studies
| Current study | Published study | |||||
|---|---|---|---|---|---|---|
| PI-RADS threshold | Facility I | Facility II | Facility III | van der Leest | Rouviere | |
| Sensitivity | ≥ 3 | 80.4% [77.8–83.2%] | 79.5% [75.4–83.7%] | 81.2% [78.3–84.3%] | 93.2% (136/146) | 93.6% (88/94) |
| ≥ 4 | 69.0% [66.8–71.3%] | 73.5% [69.6–77.6%] | 76.8% [74.1–79.6%] | 89.0% (130/146) | 83.0% (78/94) | |
| 5 | 35.6% [34.5–36.8%] | 39.5% [37.4–41.7%] | 35.1% [33.9–36.4%] | 65.8% (96/146) | 58.5% (55/94) | |
| Specificity | ≥ 3 | 73.6% [72.9–74.3%] | 68.3% [67.2–69.2%] | 77.0% [76.3–77.6%] | 62.3% (299/480) | 29.9% (47/157) |
| ≥ 4 | 88.0% [87.6–88.3%] | 78.4% [77.6–79.0%] | 83.9% [83.5–84.3%] | 69.4% (333/480) | 61.8% (97/157) | |
| 5 | 97.9% [97.8–98.0%] | 93.8% [93.6–94.0%] | 96.3% [96.1–96.4%] | 90.6% (435/480) | 93.0% (146/157) | |
| PPV | ≥ 3 | 70.9% [70.6–71.2%] | 58.8% [58.2–59.4%] | 71.4% [71.2–71.7%] | 42.9% (136/317) | 44.4% (88/198) |
| ≥ 4 | 82.1% [81.9–82.3%] | 65.9% [65.7–66.3%] | 77.2% [76.8–77.4%] | 46.9% (130/277) | 77.8% (98/126) | |
| 5 | 93.2% [93.1–93.4%] | 78.4% [77.7–79.0%] | 87.0% [86.6–87.2%] | 68.1% (96/141) | 83.3% (55/66) | |
| NPV | 82.5% [79.5–85.4%] | 85.4% [81.5–89.0%] | 85.3% [82.4–88.1%] | 96.8% (299/309) | 88.7% (47/53) | |
| CDR | 35.7% [35.6–35.9%] | 28.9% [28.6–29.1%] | 33.6% [33.5–33.7%] | 21.7% (136/626) | 39.8% (137/344) | |
| AIR | 50.4% (1073/2130) | 49.1% (421/858) | 47.1% (527/1119) | 50.6% (317/626) | 78.9% (198/251) | |
| Prevalence | 44.4% [43.0–45.9%] | 36.3% [34.4–38.4%] | 41.4% [39.9–42.9%] | 23.3% (146/626) | 37.5% (94/251) | |
| AUC | 0.83 [0.81–0.84] | 0.79 [0.76–0.82] | 0.83 [0.81–0.85] | 0.86 [0.83–0.89] | 0.81 [0.76–0.86] | |
Performance metrics were calculated for the present study and two multi-center prospective studies that evaluated the diagnostic performance of MRI-guided biopsy (van der Leest: [28], Rouviere: [29])
The current study shows the estimated statistics calculated from all biopsy-naive patients with clinical suspicion of clinically significant prostate cancer (csPCa) in the three facilities. In contrast, the published studies show the observed statistics calculated from all enrolled biopsy-naive patients
When calculating the estimated statistics, the estimated number of examinations with csPCa was used, assuming a 100% pathological confirmation rate. The 95% confidence intervals are shown in square brackets
AIR abnormal interpretation rate, CDR cancer detection rate, NPV negative predictive value, PI-RADS Prostate Imaging-Reporting and Data System, PPV positive predictive value