Application of prediction intervals to the interpretation of the robustness study of a UHPLC method for the separation of cannabinoids

Abstract

Design of Experiments (DoE) is a well-established tool used for analytical methods robustness studies, because of its ability to assess the effect of a great number of factors in a minimal number of experiments. However, when assessing the robustness of an analytical method the analysis of the individual effect of each factor is not sufficient on its own. Some factors may not influence the robustness of the method, but their effect combined with the effects of other factors may have a significant contribution on the robustness of the method, which is not given by conventional analysis of DoE results. The aim of this work is to propose, in addition to the analysis of the individual effects of the factors, to estimate the joint effect of the factors by means of the matrix experimental results prediction interval. This prediction interval is the interval in which, with a given probability, should fall the next results, therefore it is an interesting tool to estimate the variation limits of the method results during routine use. We also propose the use of two other prediction intervals which can help to analyze the DoE results and give a conclusion on the method robustness. The first one is based on the DoE experimental error information, and it gives an estimation of the experimental error component impact on the factors joint effect. The second one is based on the factors non-significance limits, and it provides the information regarding the factors impact on the responses in the case where the conditions are, by definition, robust. We applied these proposals to the robustness study of a UHPLC method for the separation of phytocannabinoids and we could demonstrate that, in addition to the calculated effects values and robustness information, the use of the prediction intervals information provided additional information that allowed a better interpretation of the method performance parameters.

Keywords: Analytical target profile; Cannabinoids; Design of experiments; Method operable design region; Prediction interval; Robustness; UHPLC.