Multi-Level Robustness Evaluation Approach: From Robustness Criterion to Adapted Algorithm of Dong

Abstract

This paper proposes an innovative concept for robustness evaluation guided by two crucial aims: indubitable identification of the factors that significantly affect the LC method and avoidance of unnecessary time and money wasting. The first phase of the proposed strategy includes robustness screening during the method optimization. Initial assumptions of the method robustness can be tracked as the rate of the response change while the factors deviate within the expected range. Therefore, partial and total robustness criteria are calculated. If the results obtained are not satisfactory, re-optimization of the method should be considered. Otherwise, extensive robustness testing defined by experimental design and multi-level factors estimation should be performed to confirm the method robustness. Firstly, the important factors are investigated by the standard graphical (normal probability plots) and statistical (algorithm of Dong and error estimation based on a priori declared negligible effects) procedures. Since these approaches have several drawbacks, they can result in the appearance of false negative or false positive results. Thus, the modification of the statistical tests is advised in order to make the final conclusions. Special attention was dedicated to the advantages of the adapted algorithm of Dong (so-called 75 % approach) in the absence of the effect sparsity. The new approach is presented on the optimization and robustness testing of LC method for determination of ramipril and its five impurities. It is proved that the proposed strategy can perform an overall robustness estimation and successfully reveal all important factors.