Table 2.
Correlation, calibration, and agreement metrics between the LiverRisk score and LSM, as well as metrics on the diagnostic accuracy of LiverRisk score for the diagnosis of cACLD in cohorts I and II.
| Metric | Cohort I (n = 5,897) | Cohort II (n = 1,558) |
|---|---|---|
| LSM (kPa, linear) | ||
| Pearson’s r | 0.325 | 0.422 |
| R2 | 0.106 | 0.178 |
| Intercept (SE) | 4.502 (0.260) | 1.788 (0.457) |
| Slope (SE) | 0.791 (0.030) | 1.002 (0.055) |
| Mean difference (95% CI)∗ ≙ calibration in the large (mean calibration) | 2.93 (2.67–3.18) | 1.80 (1.36–2.25) |
| Lower limit of agreement (95% CI)∗ | -16.75 (-17.19 to [-16.31]) | -15.82 (-16.58 to [-15.06]) |
| Upper limit of agreement (95% CI)∗ | 22.61 (22.17–23.04) | 19.42 (18.66–20.19) |
| Interval of agreement (kPa/points)∗ |
39.36 |
35.24 |
| cACLD (≥10 kPa) | ||
| Prevalence, n (%) | 1,690 (28.7) | 322 (20.7) |
| AUROC (95% CI) | 0.757 (0.743–0.770) | 0.790 (0.762–0.819) |
| Sensitivity† (95% CI) (%) | 30.2 (28.0–32.4) | 34.5 (29.3–39.9) |
| Specificity† (95% CI) (%) | 91.9 (91.1–92.7) | 93.4 (91.8–94.7) |
| Positive predictive value† (95% CI) (%) | 60.7 (57.0–63.0) | 57.6 (51.2–63.7) |
| Negative predictive value† (95% CI) (%) | 76.6 (76.0–77.2) | 84.5 (83.4–85.5) |
| Accuracy† (95% CI) (%) | 74.2 (73.1–75.3) | 81.2 (79.1–83.1) |
These include Pearson’s r, linear regression, Bland–Altmann analysis, ROC analysis.
AUROC, area under the receiver operator characteristics curve; cACLD, compensated advanced chronic liver disease; LSM, liver stiffness measurement.
∗
Based on Bland–Altmann analysis (LSM - LiverRisk score).
†
Applying at a cut-off of 10 points, corresponding to 10 kPa.