A Clinical Nomogram for Predicting Substandard Serum Valproic Acid Concentrations in Chinese Patients With Epilepsy

Abstract

Background

It is well-known that substandard serum valproic acid (VPA) concentrations may lead to treatment failure of epilepsy. However, there is still a lack of a quick method to predict whether a patient's serum VPA concentration will reach the standard.

Objective

The aims of this study were to investigate the factors leading to substandard serum VPA concentrations in Chinese patients with epilepsy and develop a related nomogram for risk prediction.

Methods

From January 2019 to March 2022, a total of 1143 serum VPA concentrations were collected from 630 hospitalized Chinese patients with epilepsy who were monitored by the Department of Pharmacy of Lu'an People's Hospital, and complete clinical data were collected from the corresponding patients for retrospective analysis. All monitored serum VPA concentrations were further divided into a training cohort and a validation cohort. For the training cohort, serum VPA concentrations below 50 µg/mL and between 50 and 100 µg/mL were classified into the subtherapeutic group and therapeutic group, respectively. The variables were selected from the clinical data, and differences between the variables of the subtherapeutic and therapeutic groups were analyzed. The influencing factors leading to substandard serum VPA concentrations were screened via logistic regression analysis, and the screened influencing factors were used to establish the nomogram prediction model.

Results

Multivariate logistic regression analysis revealed that the daily dose per unit of body weight (mg/kg/d), route of administration, presence of hepatic lesions, hypoalbuminemia, and combination with carbapenems or barbiturates were independent factors influencing the occurrence of substandard serum VPA concentrations. On the basis of the results of the multivariate logistic regression analysis, a nomogram risk prediction model for substandard serum VPA concentration was established. The values of the C-index and internal verification results indicated that the nomogram model had good accuracy and discrimination. The decision curve revealed that the nomogram that predicted the risk of substandard serum VPA concentration had a greater net benefit value (ranging from 12% to 94%), indicating that the model had a wide prediction interval.

Conclusions

Our study established a nomogram risk prediction model for substandard serum VPA concentrations in Chinese patients with epilepsy, which can help doctors or patients control the serum VPA concentration within the target concentration range as soon as possible.