RESUMO
INTRODUCTION: The debate about observing total versus free serum valproate levels for therapeutic drug monitoring (TDM) has been unresolved for decades. This study was planned to assess the agreement between the total versus free valproate levels and the advantage of one method over the other in TDM. METHODS: The present cross-sectional study was done on 93 patients with bipolar disorder. The intraclass correlation coefficient, Bland Altman analysis, and Lin's concordance analysis were done to assess the agreement between the total and free valproate concentrations. Linear and polynomial models were constructed to evaluate the relation between the two measurements. Receiver operating characteristics analysis was done to compare the accuracy for differentiating remission from non-remission on Young's mania rating scale (YMRS). RESULTS: The intraclass correlation coefficient and Lin's concordance correlation coefficient were 0.491 (p = .002) and 0.055 (95% CI:0.037, 0.073), respectively. Bland Altman's analysis showed proportional bias. A polynomial model of second order was found to be the best fit for the prediction of free valproate from the data for total valproate, and 81.4% of the variability in free valproate could be explained when adjusted for albumin levels. The area under the curve for total valproate was 0.60 when compared to free valproate 0.56 for differentiating between remission and non-remission, but the comparison between the two ROC analyses was not statistically significant. CONCLUSION: Free valproate does not provide any added advantage over the total valproate levels; hence, total valproate levels may continue to be used as the marker for drug monitoring.
