Construction and validation of a nomogram prediction model for antiviral efficacy based on clinical characteristics and intestinal microflora distribution in patients with chronic hepatitis B

TL;DR

This study creates a prediction model to assess antiviral treatment effectiveness in chronic hepatitis B patients using clinical data and gut microbiome features.

Contribution

A novel nomogram model combining clinical and microbiome data to predict antiviral therapy outcomes in chronic hepatitis B.

Findings

The nomogram model achieved AUCs of 0.869 in training and 0.829 in verification sets, showing strong predictive power.

Key predictors included AST, HBV DNA levels, and gut microbiome diversity indices like Shannon-Wiener and Simpson.

The model demonstrated good calibration and fitting in both training and verification datasets.

Abstract

To construct and validate a nomogram prediction model based on clinical characteristics and intestinal flora distribution in patients with chronic hepatitis B.

Patients with chronic hepatitis B were divided into training set (n = 175) and verification set (n = 75) according to the ratio of 7:3 by complete random method. In the training set, multivariate logistic regression was used to analyze the risk factors for the failure of antiviral therapy and the nomogram prediction model was constructed. The ROC curve and calibration curve were drawn to evaluate the prediction efficiency of the nomogram model and were verified in the verification set.

There was no significant difference in the incidence, clinical characteristics and distribution parameters of intestinal flora between the training set and the verification set (p > 0.05). Univariate analysis showed that the training set treatment ineffective group and the effective group had statistical differences in ALT, AST, hepatitis B virus DNA quantification, Shannon-Wiener index, Simpson index, Chao1 index, ACE index, relative abundance of Sclerotinia sclerotiorum, relative abundance of Bacteroides immitis, and PCA clustering separation (p < 0.05). Multivariate logistic regression analysis identified AST, hepatitis B virus DNA quantification, Shannon-Wiener index, Simpson index, and the relative abundance of Firmicutes and Bacteroides as independent risk factors for antiviral therapy failure (p < 0.05). Further, the nomogram prediction model was constructed, and the nomogram model had good calibration and fitting between prediction and reality in the training set and the verification set (ROC curves were shown in the training set and the verification set); AUC of the nomogram model for predicting the antiviral treatment effect was 0.869 and 0.829.

The nomogram model shows good discriminative ability for predicting suboptimal antiviral response, requiring multicenter validation. It should complement, not replace, clinical judgment and virological monitoring, aiding early risk identification and targeted interventions.

Figures