Potential role of predictive models in assessment of liver inflammation in patients with hepatocellular carcinoma: a two-center cohort study

Abstract

Background Hepatic inflammation in patients with hepatocellular carcinoma (HCC) remains unclear. This study aimed to construct a clinically expedient predictive model to grade hepatic inflammation in HCC patients. Methods This is a two-center retrospective cohort study of HCC patients comprising Derivation cohort and External Validation cohort of 1201 and 505 patients, respectively. Variables of liver inflammation identified through uni- and multi-variate logistic regression analyses were incorporated into predictive nomograms and applied to Derivation cohort, subject to internal and external validation. Results Liver fibrosis severity score, portal hypertension severity, and model for end-stage liver disease-sodium independently predicted hepatic inflammation grade. Performance for distinguishing G1 and non-G1 (≥ G2) patients was good with C-index of 0.810 and 0.817 in Derivation and External Validation cohort, respectively. The nomogram performed poorly to predict grade G2, G3 and G2 + G3, but performed well to predict G4. Conclusions Our nomogram exhibited good performance for scaling hepatic inflammation (G1 and G4) in HCC, and could be employed as adjunctive diagnostic tools to guide HCC management strategy.