PRECISION, ACCURACY, AND DATA ACCEPTANCE CRITERIA IN BIOPHARMACEUTICAL ANALYSIS

Accuracy and precision are the most important criteria in the assessment of an analytical method, and monitoring quality control during sample analysis is essential to ensure the validity of reported results. Various approaches to testing accuracy, precision, and quality control were applied to 10 analytes from seven chromatographic bioanalytical methods. These methods include fixed interval bias and significance testing for accuracy; fixed interval percentage relative standard deviation (%RSD) and analysis of variance (ANOVA) approaches for precision; +/-20% fixed range, 99% confidence interval, multiple rules. and range chart for individuals approaches for quality control acceptance criteria. Quality control approaches were also applied to the entire run and to a bracketed approach whereby results are considered valid only if bracketed by acceptable quality control. Accuracy and precision were assessed for six runs of each analyte at three concentrations established to represent the calibration range of the analytical method. Quality control acceptance criteria were evaluated using all data sets from each of the analytical methods collected during the course of running various numbers of real samples. The data suggest that the fixed interval bias criteria for accuracy was a more liberal method of accuracy assessment because three of the seven methods would have been rejected according to the significance testing criteria whereas all were acceptable by the fixed internal bias criteria. Precision can be effectively assessed for between- and within-run data by criteria set on unconfounded %RSD values or by separation of the sources of variation using an ANOVA approach applied to confounded data. The percentage of samples rejected for the 99% confidence interval applied to brackets, the multiple rules approach applied to the entire run, and the individual range chart approach applied to brackets were comparable and were found to be 7.0, 6.2 and 8.3 percent respectively. The +/-20% fixed range criteria applied to the two-thirds of the run resulted in just 2.9% sample rejection and was not considered comparable to the other methods.