Table 4.
Biomarkers accurately predict peak INR of more than 1·5 in patients who had a normal ALT and INR at hospital presentation
|
Derivation cohort (MAPP) n/N=26/875 |
Validation cohort (BIOPAR) n/N=4/176 |
|||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ROC–AUC | p value | Specificity | Sensitivity | PPV | NPV | ROC–AUC | p value | Specificity | Sensitivity | PPV | NPV | |
| ALT | 0·55 (0·39–0·72) | 0·473 | 0·95 | 0·23 (0·09–0·44) | 70·0 | 52·5 | 0·57 (0·31–0·80) | 0·369 | 0·95 | 0·25 (0·12–0·45) | 0·0 | 63·6 |
| Paracetamol concentration | 0·53 (0·36–0·70) | 0·699 | 0·95 | 0·00 (0·00–0·15) | 0·0 | 45·8 | 0·55 (0·40–0·79) | 0·773 | 0·95 | 0·00 (0·00–0·13) | 0·0 | 66·7 |
| miR-122 | 0·73 (0·59–0·88) | 0·0043 | 0·95 | 0·46 (0·27–0·66) | 92·3 | 62·2 | 0·75 (0·41–1·00) | 0·016 | 0·95 | 0·50 (0·31–0·67) | 66·7 | 80·0 |
| HMGB1 | 0·94 (0·88–1·00) | <0·0001 | 0·95 | 0·88 (0·70–0·98) | 92·0 | 88·0 | 0·90 (0·73–1·00) | 0·025 | 0·95 | 0·65 (0·50–0·88) | 75·0 | 88·9 |
| Full-length K18 | 0·81 (0·69–0·93) | 0·0001 | 0·95 | 0·27 (0·11–0·48) | 87·5 | 54·8 | 0·86 (0·65–1·00) | 0·045 | 0·95 | 0·25 (0·10–0·43) | 50·0 | 72·3 |
| Caspase-cleaved K18 | 0·82 (0·70–0·94) | <0·0001 | 0·95 | 0·27 (0·11–0·48) | 77·8 | 53·7 | 0·81 (0·55–1·00) | 0·008 | 0·95 | 0·25 (0·10–0·43) | 50·0 | 72·3 |
| GLDH | 0·66 (0·51–0·82) | 0·042 | 0·95 | 0·42 (0·23–0·63) | 73·3 | 57·1 | 0·58 (0·26–0·88) | 0·643 | 0·95 | 0·25 (0·10–0·44) | 0·0 | 66·7 |
ROC–AUC (95% CI), sensitivity at 95% specificity (95% CI), PPV, and NPV were calculated to identify the potential of novel and established stratification biomarkers to predict the development of hepatic dysfunction. INR=international normalised ratio. ROC=receiver operator characteristic. AUC=area under the curve. PPV=positive predictive value. NPV=negative predictive value. ALT=alanine aminotransferase. HMGB1=high mobility group box-1. GLDH=glutamate dehydrogenase.