Prediction of unbound propofol concentrations in a diabetic population

Propofol is a short-acting general intravenous anesthetic characterized by a wide interindividual variability in the response after the same dose. Its binding to serum proteins exceeds 98%, so small changes in protein concentrations can be amplified in the unbound fraction of the drug and hence possibly in the effect. It is then likely that part of the variability in the response could be attributed to differences in protein levels among individuals and particularly among those with pathologies such as diabetes. The aim of this study was to establish predictive regression models in a diabetes mellitus (DM) population between unbound:bound propofol ratios and demographic and biochemical indices. Unbound:bound propofol ratios can be routinely obtained in the clinic as opposed to the free fraction of the drug. In DM patients (30 women and 37 men aged between 17 and 78 y) with mellitus type 1 (n = 37) and type 2 (n = 30) diabetes, the authors measured the lipoproteins (HDL, LDL, and VLDL), cholesterol, triglycerides, albumin, alpha1-acid glycoprotein (AAG), free fatty acids (FFA), glycosylated hemoglobin, and the unbound fraction (Fu) and the bound/free ratio (B/F) of propofol. A linearized regression model between the above variables--as well as age, sex, and type of diabetes--and Fu was then developed. Patients had blood drawn and sera separated by centrifugation and spiked with propofol to a concentration of 10 microg/mL. The Fu was determined via ultrafiltration. Multiple linear regression analysis was used to identify significant predictor variables of Fu in this population and two models were originated: one with lipoprotein serum concentrations as explanatory variables (Model A) and another that depended on cholesterol and triglycerides (Model B). Both models presented high correlation coefficients (r2 = 0.71 and 0.68, respectively; P < 0.0001), and each was used to predict Fu in an independent group of 15 DM patients of similar characteristics and biochemical indices as the model development group. Bias and precision were for Model A, 0.9% and 7.8%, and for Model B, 3.0% and 8.7%, respectively. Both models were compared with each other and to a naive predictor (the mean) and each was better than the naive model in predicting the unbound fraction of propofol. Model A and model B could be used in estimating Fu of propofol in DM patients based on the more routine clinical measures of lipoprotein serum concentrations or cholesterol and triglyceride levels.