Cross-validation of methods for the quantitative determination of phenazepam and its active metabolite in human blood plasma at various extractions

Abstract

Introduction. In conducting of therapeutic drug monitoring (TDM), often such situation arises where the drug concentration has measured by different methods or in different laboratories. To combine and analyze the data obtained with different methods, it is necessary to perform cross-validation procedure. Insufficient attention is paid to the statistical approaches used for this purpose.Aim. Performing cross-validation of different analytical methods for the quantitative determination of phenazepam (PHEN) and 3-hydroxyphenazepam (3-OH-PHEN) using the Bland – Altman analysis.Materials and methods. PHEN and 3-OH-PHEN concentrations in the blood plasma of patients (n = 100) with alcohol withdrawal syndrome were measured using high-performance liquid chromatography with tandem mass spectrometry (HPLC-MS/MS). The quantification of both analytes in each sample was measure twice by two different methods: solid phase extraction (SPE) and supported liquid extraction (SLE). Both methods have been fully validated before the experiment began. Cross-validation was performed at the end of the experiment using data from study samples. The Bland – Altman analysis was used to evaluate accuracy and precision. Deming regression was also used to identify a systematic error between measurement results.Results and discussion. The regression equations have been obtained between concentrations both analytes measured by different sample preparation methods. 95 % confidence intervals (CI) of the regression coefficients of both equations included one, and 95 % CI of the intercepts included zero. 95 % CI of the geometric mean of the individual SLE/SPE ratios was within the acceptable range (0.87; 1.15). These results confirm the absence of the influence of quantitative methods on the measurement of both analytes concentration. 66.7 % CI of the percent difference between two measurements was within acceptable limits (–0.2; 0.2), not exceeding 20 % of the range of their mean value. This confirms the acceptable precision between the methods. The estimated CIs were displayed in the Bland – Altman plots.Conclusion. The statistical approaches used in the work have confirmed the reproducibility of the results of different sample preparation methods. In addition to cross-validation, the statistical algorithm from this paper using Bland – Altman analysis can be successfully employed to assess accuracy and precision during bioanalytical method validation and evaluation of the acceptance of analytical runs, as well as to determine the level of reproducibility of incurred samples.