Voriconazole-associated liver injury: clinical risk factor identification and predictive nomogram construction.

OBJECTIVE

Investigate the high risk factors contributed to voriconazole-induced liver injury and develop a predictive nomogram for voriconazole-induced liver injury risk, thereby optimizing the clinical medication safety.

METHODS

This observational study retrospectively analyzed the electronic health records of hospitalized patients who received voriconazole treatment or prophylaxis at a tertiary hospital. Patients who met the inclusion and exclusion criteria were enrolled between June 2020 and June 2024, and their general biological data were collected. The diagnosis and severity grading of voriconazole-associated liver injury were determined according to the diagnostic and treatment guidelines for drug-induced liver injury. Patients were categorized into liver injury and non-injury groups based on the results of post-treatment liver function tests. Associations between baseline characteristics and liver injury were analyzed using non-parametric and chi-squared tests. To avoid omitting potentially significant factors, variables demonstrating P ≤ 0.1 in univariate screening and retained through least absolute shrinkage and selection operator (LASSO) regression underwent multivariate logistic regression to identify independent predictors. The nomogram was developed and internally validated through receiver operating characteristic (ROC) analysis, calibration curves, and decision curve analysis (DCA).

RESULTS

Among the 1,964 patients screened, 1,196 were excluded, leaving 768 patients included in the statistical analysis. Liver injury occurred in 95 patients, resulting in an incidence rate of 12.4%. Multivariable logistic regression analysis demonstrated that total cholesterol (TC) (OR = 1.893, P < 0.01), concomitant use of glucocorticoids (OR = 1.861, P = 0.041), ezetimibe (OR = 7.453, P = 0.047), and caspofungin (OR = 2.485, P = 0.032). Patients with TC levels exceeding 4.485 mmol/L exhibited a significantly elevated risk of liver injury. ROC analysis revealed an area under the curve (AUC) of 0.728 (95% CI: 0.66-0.797), with a sensitivity of 0.661 and specificity of 0.722. Internal validation indicated good discrimination and calibration, with predicted probabilities closely aligned with actual outcomes. The decision curve analysis suggested a substantial net clinical benefit. These findings were subsequently validated in a test cohort.

CONCLUSION

Concomitant use of ezetimibe, caspofungin, glucocorticoids, and TC levels exceeding 4.485 mmol/L are independent risk factors for voriconazole-induced liver injury. The developed nomogram model offers clinically meaningful predictions for drug-induced liver injury risk, facilitating the optimization of voriconazole therapy safety.