Table 4. Model Performance Parameters.

	Model Performance
Cross-validated AUROC (95% CI)	0.83 (0.77–0.89); p<0.0001
PPV	65.5%
NPV	86.2%
Sensitivity	79.2%
Specificity	75.8%
AIC	190
BIC	213
Correct Prediction of Histologic Response	77%
Probability Cutoff of Histologic Response	0.39
Clinical Model for Probability of Response to OCA
$log\frac{p}{1-p}$=−3.8520+1.1941*Treatment(OCA=1, PBO=0) −1.8836* baseline NAS(1 when<=5; 0 when >5) + 1.3949*Baseline AST (1 when <=49 U/L; 0 when >49 U/L) + 2.3717*ALT change at week 24 (1 when <=−17 U/L, 0 when >−17 U/L) +0.9124* Baseline Triglycerides (1 when <=154 mg/dL, 0 when >154 mg/dL) + 1.2376* Baseline INR (1 when <=1, 0 when >1)

P-value of the AUROC determined by a chi-square test.

For use of the clinical model, the selected parameters are input into the model and if the resultant probability is above 0.39 then the patient is predicted to achieve histologic response at Week 72.

AUROC: area under the receiver operator characteristic curve; PPV: positive predictive value; NPV: negative predictive value; AIC: Akaike information criterion; BIC: Bayesian information criterion; OCA: obeticholic acid; NAS: nonalcoholic fatty liver disease activity score; AST: aspartate aminotransferase; ALT: alanine aminotransferase; INR: international normalized ratio