(982) External Validation of the Fraser Equation to Estimate Free Valproate Concentration in the ICU

Introduction: Valproate (VPA) has complex protein binding and free VPA concentrations (fVPA) are often occultly toxic in the critically ill even when total VPA concentrations (tVPA) are within the reference range. Measurement of fVPA is rarely available at medical center laboratories; the Fraser equation was derived to estimate fVPA in ICU patients. We sought to externally validate the Fraser equation in a separate cohort of critically ill patients at two academic medical centers.

Methods: Adult patients ≥18 years of age who were admitted to an ICU and received VPA with concurrently measured tVPA and fVPA were screened. Patients with incomplete data required to execute the Fraser equation (tVPA, albumin, blood urea nitrogen [BUN], and propofol or aspirin use) were excluded. The primary outcome was the mean bias between the Fraser estimated and measured fVPA concentrations using Bland-Altman statistics. Secondary outcomes included the performance an alternative predictive equation published by Dore and colleagues and the therapeutic concordance between estimated and measured fVPA (therapeutic fVPA defined as 5-15mg/L).

Results: 315 unique patients with 556 fVPA concentrations were included; 210 (67%) were female, mean (±SD) age was 58 (±17) years, and 141 (45%) were admitted to a neurosciences ICU. Median (IQR) tVPA and fVPA were 46 mg/L (34-63) and 17 mg/L (11-24), respectively, and the median free fraction was 36% (25-54), well above the expected 5-10% range. The mean (95% CI) bias between Fraser estimated and measured fVPA was -2.8 mg/L (-18.9-13.5), well below that for the Dore equation (-4.7 mg/ml [-21.1-11.8], p< 0.001). Underestimation of fVPA tended to increase as measured fVPA increased. Fraser estimates led to a therapeutically concordant interpretation for 71% of all fVPA concentrations versus 64.9% for the Dore equation (p=0.03). The most common discrepancy was an estimated therapeutic fVPA but measured supratherapeutic concentration (n=76, 13.7%).

Conclusions: The Fraser equation-estimated fVPA generally underestimated measured fVPA in an external cohort but led to appropriate interpretation for 71% of all concentrations. Further refinements are necessary before broad application of the Fraser equation in clinical practice.