High-performance liquid chromatographic determination of propofol in human plasma: comparison of different heteroscedastic calibration curve models.

PURPOSE The aim of this study was to select the best calibration model for determination of propofol plasma concentration by high-performance liquid chromatography method. METHODS Determination of propofol in plasma after deproteinization with acetonitrile containing thymol (as internal standard) was carried out on a C18 column with a mixture of acetonitrile and trifluoroacetic acid 0.1% (60:40) as mobile phase which delivered at the flow rate of 1.2 mL/minute . Fluorescence detection was done at the excitation and emission wavelengths of 276 and 310 nm, respectively. After fitting different equations to the calibration data using weighted regression, the adequacy of models were assessed by lack-of-fit test, significance of all model parameters, adjusted coefficient of determination (R(2) adjusted) and by measuring the predictive performance with median relative prediction error and median absolute relative prediction error of the validation data set. RESULTS The best model was a linear equation without intercept with median relative prediction error and median absolute relative prediction error of 4.0 and 9.4%, respectively in the range of 10-5000 ng/mL. The method showed good accuracy and precision. CONCLUSION The presented statistical framework could be used to choose the best model for heteroscedastic calibration data for analytes like propofol with wide range of expected concentration.