Table 3.
Discrimination of the IT-3 model and other non-invasive models.
| AUC (95%CI) | p-valuea | NRI (95%CI)b | p-value | IDI (95%CI)c | p-value | |
|---|---|---|---|---|---|---|
| Training cohort | ||||||
| IT-3 | 0.888 (0.813–0.962) | – | – | – | – | – |
| GPR | 0.731 (0.641–0.821) | 0.007 | 1.27 (0.938–1.610) | <0.001 | 0.21 (0.124–0.302) | <0.001 |
| APRI | 0.740 (0.646–0.834) | 0.001 | 1.36 (1.023–1.691) | <0.001 | 0.22 (0.134–0.302) | <0.001 |
| FIB-4 | 0.645 (0.546–0.743) | <0.001 | 1.40 (1.086–1.704) | <0.001 | 0.23 (0.142–0.317) | <0.001 |
| Validation cohort | ||||||
| IT-3 | 0.833 (0.695–0.970) | – | – | – | – | – |
| GPR | 0.669 (0.522–0.815) | 0.147 | 1.21 (0.750–1.672) | <0.001 | 0.29 (0.145–0.433) | <0.001 |
| APRI | 0.616 (0.453–0.779) | 0.009 | 0.97 (0.466–1.480) | <0.001 | 0.30 (0.153–0.441) | <0.001 |
| FIB-4 | 0.631 (0.484–0.777) | 0.050 | 1.00 (0.496–1.507) | <0.001 | 0.29 (0.149–0.426) | <0.001 |
APRI, aspartate aminotransferase-to-platelet ratio index; AUC, the area under curve; CI, confidence interval; FIB-4, fibrosis index based on the four factors; GPR, gamma-glutamyl transpeptidase to platelet ratio; IDI, integrated discrimination improvement; NRI, net reclassification improvement.
a Compared with the IT-3 model.
b,c NRI or IDI > 0 indicated the new model (IT-3) had better prediction performance than reference model (GPR, APRI or FIB-4). Cut-off of NRI: 0.2, 0.4.