融合中医体质分型与多因素信息的抗结核药物性肝损伤风险预测模型构建及评估OA

Objective To integrate traditional Chinese medicine(TCM)constitution types with conventional clinical factors,develop and evaluate a risk prediction model for anti-tuberculosis drug-induced hepatotoxicity(ATDH),and to provide a quantitative tool for risk prevention and control within an integrative medicine framework during tuberculosis treatment.Methods A single-center retrospective cohort study was conducted.A total of 1 396 patients diagnosed with pulmonary tuberculosis and receiving standard anti-tuberculosis therapy at Rugao People's Hospital from January 2019 to October 2022 were enrolled.Based on the occurrence of ATDH during treatment,patients were divided into a hepatotoxicity group(180 cases)and a non-hepatotoxicity group(1 216 cases).General data were compared between the two groups.Multivariate logistic regression analysis was used to identify influencing factors for ATDH and to establish a prediction model.All patients were randomly divided into a training set(977 cases)and a validation set(419 cases)at a ratio of 7:3.The model's discrimination,calibration,and clinical application value were comprehensively evaluated using receiver operating characteristic(ROC)curves,calibration curves,and decision curve analysis.Results The proportions of extra-pulmonary tuberculosis,cardiac insufficiency,underlying liver disease,qi-deficiency constitution positivity,and damp-heat constitution positivity in the hepatotoxicity group were higher than those in the non-hepatotoxicity group,while the age,proportion of initial treatment status,electrolyte disturbance,hypoproteinemia,balanced constitution positivity,and phlegm-dampness constitution positivity were lower than those in the non-hepatotoxicity group(P<0.05).Multivariate logistic regression analysis showed that age,initial treatment status,underlying liver disease,electrolyte disturbance,hypoproteinemia,damp-heat constitution positivity,and phlegm-dampness constitution positivity were independent risk factors for ATDH(OR=1.036,95%CI:1.031-1.053,P<0.001;OR=2.111,95%CI:1.171-4.208,P=0.021;OR=4.476,95%CI:2.764-7.153,P<0.001;OR=3.372,95%CI:2.434-4.678,P<0.001;OR=2.601,95%CI:1.842-3.642,P<0.001;OR=1.690,95%CI:1.170-2.406,P=0.004;OR=2.369,95%CI:1.645-3.372,P<0.001).ROC curve analysis showed that the area under the curve(AUC)of the prediction model in the training set was 0.754(95%CI:0.711-0.797),with sensitivity 0.812 and specificity 0.571 at the optimal cutoff point.In the validation set,the AUC was 0.730(95%CI:0.658-0.803),with sensitivity 0.726 and specificity 0.648.The Hosmer-Lemeshow goodness-of-fit test(χ²=10.882,P=0.209 for the training set;χ²=1.712,P=0.989 for the validation set)indicated good model calibration.Calibration curve analysis showed good consistency between the predicted probability and the actual probability in both the training and validation sets.Decision curve analysis demonstrated that the prediction model provided stable and additional clinical net benefit across both datasets.Conclusion The ATDH risk prediction model integrating TCM constitution types and clinical factors shows moderately good discriminatory ability and satisfactory calibration for identifying high-risk patients,enabling individualized risk assessment at the initial stage of anti-tuberculosis treatment.Damp-heat and phlegm-dampness constitutions significantly increase ATDH risk,suggesting the potential value of TCM constitution in assessing susceptibility to drug-induced liver injury.This model contributes to the early identification and prevention of ATDH within an integrative medicine framework but requires further external validation and optimization in multicenter prospective cohorts.