The Development and Validation of a Predictive Model for Voriconazole-Related Liver Injury in Hospitalized Patients in China

Abstract

Voriconazole is widely used in the treatment and prevention of invasive fungal diseases. Common drug-induced liver injuries increase the economic burdens and the risks of premature drug withdrawal and disease recurrence. This study estimated the disposal cost of voriconazole-related liver injury, explored the risk factors of voriconazole-related liver injury in hospitalized patients, and established a predictive model of liver injury to assist clinicians and pharmacists in estimating the probability or risk of liver injury after voriconazole administration to allow for early identification and intervention in patients at high risk of liver injury. A retrospective study was conducted on the selected inpatients whose blood concentration of voriconazole was measured in the West China Hospital of Sichuan University from September 2016 to June 2020. The incidence and disposal cost of voriconazole-related liver injuries were calculated. The incidence of voriconazole-related liver injury was 15.82% (217/1372). The disposal cost has been converted to 2023 at a discount rate of 5%. The median (P25, P75) disposal cost of severe liver injury (n = 42), general liver injury (n = 175), and non-liver injury (n = 1155) was 993.59 (361.70, 1451.76) Chinese yuan, 0.00 (0.00, 410.48) yuan, and 0.00 (0.00, 0.00) yuan, respectively, with a statistically significant difference (p < 0.001). Single factor analysis and multiple factor logistic regression were used to analyze the risk factors of voriconazole-related liver injury. The voriconazole-related liver injury was related to the trough concentration (Cmin, OR 1.099, 95% CI 1.058–1.140), hypoproteinemia (OR 1.723, 95% CI 1.126–2.636), and transplantation status (OR 0.555, 95% CI 0.325–0.948). The prediction model of liver injury was Logit (P)= −2.219 + 0.094 × Cmin + 0.544 × Hydroproteinemia − 0.589 × Transplantation, and the prediction model nomogram was established. The model validation results showed that the C-index of the derivation set and validation set was 0.706 and 0.733, respectively. The area under the curve (AUC) of the receiver operating characteristic (ROC) curve was 0.705 and 0.733, respectively, indicating that the model had good prediction ability. The prediction model will be helpful to develop clinical individualized medication of voriconazole and to identify and intervene in the cases of patients at high risk of voriconazole-related liver injury early on, in order to reduce the incidence of voriconazole-related liver injuries and the cost of treatment.