Abstract Valproic acid (VPA) is highly protein‐bound, thereby impacting its free fraction and clearance in hypoalbuminemia. There is limited data on VPA use in such patients. Thus, this study evaluates the impact of adjusted VPA concentration (aVPAc) in predicting effectiveness and adverse effects compared to total VPA (tVPA) levels in hypoalbuminemic patients. A retrospective cohort study involved adult patients with seizures or epilepsy between January 1, 2016, and December 31, 2022. The levels of tVPA and aVPA (adjusted for albumin) were compared using receiver operating characteristic curves, and AUC differences were assessed using the DeLong method. Safety endpoints included hepatotoxicity, hyperammonemia, hyponatremia, and thrombocytopenia, while effectiveness endpoints were seizure occurrence, status epilepticus, and the use of additional antiepileptic medications during hospitalization. Of the 1621 screened patients, 71 with hypoalbuminemia received VPA. An aVPAc threshold of 154.19 mg/dL demonstrated higher sensitivity (86%) but lower specificity (47%) for predicting hepatotoxicity compared to a tVPA threshold of 67.53 mg/dL (sensitivity: 71%, specificity: 72%). Although aVPAc yielded a comparable negative predictive value (96% vs 95%), tVPA showed superior positive predictive value (25% vs 18%) and a higher Youden index (0.43 vs 0.33), indicating better overall discriminatory performance; however, these findings did not achieve statistical significance. In contrast, an aVPAc threshold of 188 mg/dL showed a sensitivity of 100% and a specificity of 82% for predicting hyperammonemia, which is superior to the tVPA threshold of 74.32 mg/dL that has a sensitivity of 40% and a specificity of 88%. The aVPAc also achieved a higher Youden index of 0.82 compared to 0.28 for tVPA. Adjusted VPA concentrations showed greater sensitivity than tVPA in predicting hepatotoxicity and hyperammonemia, suggesting potential utility for ruling out these adverse effects in hypoalbuminemic patients. Further research with a larger sample size is needed to validate these findings.
Naheet et al. (Thu,) studied this question.