Abstract :
[en] As defined by ICH [1] and FDA, Quality by Design (QbD) stands for “a systematic approach to development that begins with predefined objectives and emphasizes product and process understanding and process control, based on sound science and quality risk management”. A risk–based QbD–compliant approach is proposed for the robust development of analytical methods. This methodology based on Design of Experiments (DoE) to study the experimental domain models the retention times at the beginning, the apex and the end of each peak corresponding to the compounds of a mixture and uses the separation criterion (S) rather than the resolution (RS) as a Critical Quality Attribute. Stepwise multiple linear regressions are used to create the models. The estimated error is propagated from the modelled responses to the separation criterion (S) using Monte Carlo simulations in order to estimate the predictive distribution of the separation criterion (S) over the whole experimental domain. This allows finding ranges of operating conditions that will guarantee a satisfactory quality of the method in its future use. These ranges define the Design Space (DS) of the method. In chromatographic terms, the chromatograms processed at operating conditions within the DS will assuredly show high quality, with well separated peaks and short run time, for instance. This Design Space can thus be defined as the subspace, necessarily encompassed in the experimental domain (i.e. the knowledge space), within which the probability for the criterion to be higher than an advisedly selected threshold is higher than a minimum quality level. Precisely, the DS is defined as “the multidimensional combination and interaction of input variables (e.g., material attributes) and process parameters that have been demonstrated to provide assurance of quality” [1]. Therefore, this DS defines a region of operating conditions that provide prediction of assurance of quality rather than only quality as obtained with traditional mean response surface optimisation strategies. For instance, in the liquid chromatography there is a great difference in e.g. predicting a resolution (RS) higher than 1.5 vs. predicting that the probability for RS to be higher than 1.5 (i.e. P(RS> 1.5)) is high. The presentation of this global methodology will be illustrated for the robust optimisation and DS definition of several liquid chromatographic methods dedicated to the separation of different mixtures: pharmaceutical formulations, API and impurities/degradation products, plant extracts, separation of enantiomers, …
