From experimental design to uncertainty estimation for the European Pharmacopoeia HPLC analysis of human insulin

In this paper the process from experimental design (e.g. ruggedness test) to uncertainty estimation is described. The uncertainty estimate was calculated for the peak area of insulin plus A21 desamido insulin resulting from an HPLC analysis of a sample of an injectable human insulin preparation, Actrapid HM 100 IU ml-1 (Novo Nordisk A/S). The analytical method used was the European Pharmacopoeia assay.4 An expanded uncertainty1 of 1.8% (of the area from the HPLC analysis) at an approximately 95% confidence level was found and confirmed by a validation study. This uncertainty refers to the peak area of the analyte from a single injection of the sample. The input parameters to the uncertainty estimate were found from a factorial experimental design (e.g. ruggedness test) consisting of 9 factors applied to the HPLC analysis. The input parameters were chosen to cover the probable contributors to the variability of the measured area of the HPLC analysis, including the sample preparation, but excluding uncertainty deriving from the reference material(s).