Development of a nomogram risk prediction model for subtherapeutic valproic acid plasma concentration: A retrospective cohort study.

OBJECTIVE

Limited research has explored the factors contributing to subtherapeutic concentrations (< 50 μg/mL) of Valproic acid (VPA) in patients with mental disorders. To address this gap, the present study aims to identify key factors associated with subtherapeutic VPA levels and to develop a visual nomogram prediction model.

METHODS

This retrospective single-center study collected the therapeutic drug monitoring (TDM) results of patients treated with valproate at the Xi'an Mental Health Center from January to June 2024. Single-factor analysis was performed to identify candidate variables for multivariate logistic regression, which was then used to determine the final risk factors.

RESULTS

Among 686 patients receiving VPA therapy, 138 (20.12 %) had plasma concentrations below the therapeutic threshold (<50 μg/mL). Multivariate logistic regression identified age (OR = 0.449, 95 % CI: 0.259-0.780), body mass index (OR = 3.266, 95 % CI: 1.907-5.592), and daily dose (OR = 0.006, 95 % CI: 0.002-0.020) as independent predictors of subtherapeutic VPA levels (p < 0.01). The nomogram model demonstrated strong predictive performance with an area under the curve (AUC) of 0.766 through 10-fold cross-validation. The calibration curve confirmed a concordance index (C-index) of 0.788 using 1000 bootstrap resamples, and the Hosmer-Lemeshow test showed good model fit (χ2 = 4.778, p = 0.311). Decision curve analysis (DCA) indicated favorable clinical utility, with a wide range of threshold probabilities (13 %-78 %).

CONCLUSIONS

A visual nomogram model for predicting VPA concentrations was developed for Chinese patients with mental disorders. This tool offers a personalized approach to optimizing VPA dosing in clinical psychiatric practice.