RESUMO
STUDY OBJECTIVE: Because the pharmacokinetic evaluation of valproic acid (VPA) based on total drug concentration may be misleading in patients with hypoalbuminemia as a result of saturable protein binding and saturable metabolism, we sought to investigate the usefulness of therapeutic drug monitoring of unbound VPA concentration in a real-world clinical context, with a focus on clinically significant neurologic adverse outcomes. DESIGN: Retrospective analysis. SETTING: Large academic tertiary care hospital in Montreal, Canada. PATIENTS: Forty-one adults, hospitalized or followed as outpatients, for whom unbound VPA concentration testing was performed between January 1, 2008, and April 30, 2015. Patients were retrospectively identified by using the hospital's central laboratory database. MEASUREMENTS AND MAIN RESULTS: In the multiple linear regression analysis, the two variables that significantly predicted unbound VPA concentration were total VPA concentration (p<0.001) and albumin concentration (p<0.001). The correlation between total VPA concentration and the number of neurologic adverse symptoms was 0.187 (p=0.241), whereas the correlation between unbound VPA concentration and the number of neurologic adverse symptoms was 0.384 (p=0.013). The performance of total and unbound VPA concentrations in predicting the presence of at least one neurologic adverse symptom, as determined by the receiver operating characteristic curve, was 0.642 (95% confidence interval [CI] 0.449-0.836, p=0.167) and 0.776 (95% CI 0.629-0.923, p=0.007), respectively. CONCLUSION: This study showed that in the presence of hypoalbuminemia, high unbound VPA concentrations can be observed despite normal or low total VPA concentrations. It also demonstrated that high unbound VPA concentrations are associated with clinically significant neurologic adverse symptoms. Clinicians should be aware that unbound VPA concentration monitoring may be required in the presence of hypoalbuminemia.
