I. CURRENT STAGE OF KNOWLEDGE
Liquid chromatography is the analytical technique of choice in the pharmaceutical field. It is mainly used to separate, identify and quantify the compounds of interest and related substances. Whether for quality control of drugs, the study of their stability or to provide qualitative and quantitative informations to researchers in disciplines as varied as synthetic chemistry, toxicology, pharmaceutics, liquid chromatography has become a necessity. Research seeking innovative strategies for the development of chromatographic methods is therefore ever more important in the analytical field.
In the current environment guided by the concepts of quality by design (QbD) and design space (DS) recommended by the International Conference on Harmonization (ICH), the methodologies for optimization of chromatographic methods should provide analysts along with more robust chromatographic conditions, a good understanding of process optimization and increased credibility of the predictions.
The first part of this thesis is structured in three chapters that are a succinct presentation of liquid chromatography concepts, regulatory support for biopharmaceutical development and chemometrics applied to analytical method development.
In the personal contribution part of this work, the first chapter describes a methodology combining both chromatographic behavior modeling compounds and optimization of analytical conditions that was tested. This methodology, based on design of experiments, the inclusion of the prediction error and the propagation of this error, from the modeled responses to the separation criterion, has allowed the formalization and the identification of design spaces. The second chapter is an application of this new strategy for the separation of tertiary alkaloids extracted from Strychnos usambarensis leaves, while the third one presents the implementation of this methodology for enatiomeric separations in polar organic solvent chromatography.
II. PERSONAL CONTRIBUTIONS
1. The design of experiment – design space methodology
The optimization of LC operating conditions for the separation of several compounds in complex samples is often intricate. Indeed, it can be tricky to separate compounds due to their similar chromatographic behaviors or fussy to elute all of them well separated when some have widely distinct physico-chemical properties (e.g. polarities, pKa, log P). Since past decades and mainly those last years, a lot of improvements in the fields of method development have been done and it is thus possible to foresee different strategies that could be applied to find optimal separations in a rather automated way.
The main difficulty is that these strategies need to give accurate and robust predictions prior to the validation and transfer of these analytical methods. ICH Q8 (R2) guideline provides a harmonized guidance to improve the robustness and reliability of pharmaceutical development. In this guideline 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“.
To decrypt this definition, it can be assumed that, the “multidimensional combination and interaction of input variables and process parameters” is a multidimensional space (or subspace) whose dimensions are the factors used during the method development.
Afterwards, the guideline mentions “that have been demonstrated to provide assurance of quality“ which means that the size of this space is defined by the set of combinations of factors ranges wherein a process provides quality results. In the LC field, and more precisely when the separation optimization is the main aim, the DS predicts a space wherein the separation is achieved taking into account the measurements, process and models uncertainties.
The separation quality can be evaluated by a chromatographic criterion such as the resolution (RS = 2∙(tR,2 – tR,1)/(wb,1 + wb,2); with tR,2 > tR,1 and wb,1, wb,2 are the peaks widths at baseline; tR,1 and tR,2 being the retention times of the critical pair peaks) and the method uncertainty is estimated by the probability to reach a given criterion threshold (e.g. the probability for RS to be higher than 1.5) in future uses of the method. It also states that “working within the design space is not considered as a change”.
Therefore, in the framework of separation methods development, DS can be clearly considered as a zone of theoretical robustness since modifications of the method parameters will not significantly affect the separation quality. Thus, no decrease of quality in the separation should be observed while working in DS. Consequently, to optimize HPLC separations and advisedly compute DS, DoE is one of the most appropriate strategies.
Thus, based on DoE strategy, a recent methodology was used to model and predict the retention times according to selected chromatographic factors (e.g. mobile phase composition, gradient time, mobile phase pH) and subsequently optimize the separation. Afterwards, the prediction error was also estimated in order to allow the DS computation.
However, DoE involves the recording of chromatograms at very different operating conditions. By this way, very distinct selectivities are obtained. Peak detection and identification in each chromatogram can thus become tedious and time-consuming. Hence, an additional methodology based on independent component analysis (ICA) has been developed to detect and match peaks among the chromatograms resulting from DoE.
Two test samples were selected to evaluate these complementary methodologies. The first sample (sample 1) was sent from Eli Lilly and Company and was considered as an unknown sample. A similar test mixture was already used by Biswas et al. to conduct a method screening study. However, in this work, the amount of compounds and their nature were deliberately occulted to test out both methodologies.
The second sample (sample 2) is a common-cold pharmaceutical formulation. Studies involving some of these compounds were previously published: the separation of the three active drugs using a poly(ethyleneglycol) column, the separation of the three active drugs using cyano columns, the comparison between five HPLC columns, the comparison between electrophoresis and liquid chromatography separation and the validation of a HPLC method for the quantification of the active drugs. Furthermore, a review of analytical methods published for the separation of some of these substances can be found in the works of Marín et al.. These abundant publications highlight the interest that still remains in separating and quantifying these compounds. Nevertheless, as the sweeteners of a new sugar-free formulation were not taken into account in the aforementioned methods, a new method development was necessary.
The automated optimization of chromatographic separation is the first critical step in the framework of the automated development of chromatographic method. In this paper, DoE, ICA, multiple linear regression, error propagation and DS methodologies were successfully applied to separate nine compounds of an unknown sample mixture in less than 40 min and the seven compounds of a pharmaceutical formulation. Furthermore, for this latter sample, the analysis time was shortened to less than 14 min.
This global methodology is also very flexible as the choice of each criterion and their respective acceptance limit are made by the analyst. In addition, an evaluation of DS robustness was carried out during the present study. The separation criterion S clearly demonstrated its robustness capability within the identified DS. It strengthens the fact that DS defines a space wherein the separations are complete and the method is robust and even more robust if the DS is large. Nevertheless, as the size of DS depends on the value of π, further works are still required to define possible adjustment strategy for this parameter. Eventually, if one compound is discarded from one of the tested mixtures, such as an active ingredient of the pharmaceutical formulation, the present DS would obviously have defined optimal and robust space for the separation of the subsequent mixture without any additional experiments.
2. Application of a new optimization strategy for the separation of tertiary alkaloids extracted from Strychnos usambarensis leaves
The HPLC separation of six alkaloids extracted from Strychnos usambarensis leaves has been developed and optimized by means of a powerful methodology for modelling chromatographic responses, based on three steps, i.e. DoE, ICA and DS. This study was the first application of a new optimization strategy to a complex natural matrix. The compounds separated are the isomers isostrychnopentamine and strychnopentamine, 10-hydroxyusambarine and 11-hydroxyusambarine, also strychnophylline and strychnofoline.
Three LC parameters have been optimized using a multifactorial design comprising 29 experiments that includes 2 center point replicates. The parameters were the percentage of organic modifiers used at the beginning of a gradient profile which consisted in different proportions of MeOH and MeCN, the gradient time to reach 70 % of organic modifiers starting from the initial percentage and the percentage of MeCN found in the mobile phase. Subsequent to the experimental design application, predictive multilinear models were developed and used in order to provide optimal analytical conditions.
The optimum assay conditions were: methanol/acetonitrile-sodium pentane sulfonate (pH 2.2; 7.5 mM) (33.4:66.6, v/v) at a mobile phase flow rate of 1mL/min during a 40.6 minutes gradient time. The initial organic phase contained 3.7 % MeCN and 96.3 % MeOH.
The method showed good agreement between the experimental data and predictive value throughout the studied parameters space. Improvement of the analysis time and optimized separation for the compounds of interest was possible due to the original and powerful tools applied.
The partial replacement of MeCN by MeOH was successful, allowing reducing costs and maintaining a high quality of results in respect with the ICH Q8 guidelines and the DS approach.
An optimal separation was achieved especially for strychnopentamine, isostrychnopentamine, 10–hydroxyusambarine and 11-hydroxyusambarine. A good separation was also obtained for another pair of isomers that is represented by strychnophylline and strychnofoline, succeeding so to acquire for the first time, to our knowledge, a LC-UV profile for the compounds of interest.
In addition, the method resulting from the strategy of simultaneous multifactorial optimization reduced overall assay development time and provided information regarding separation and sensitivity due to the detection of new compounds in the analyzed mixture of plant origin, compounds whose separation and identification are still delicate.
Finally, this study permitted the acquisition of isomers profiles allowing the identification of the optimal collecting period of Strychnos usambarensis.
3. Implementation of a Design Space Approach for Enatiomeric Separations in Polar Organic Solvent Chromatography
This paper focuses on implementing a Design Space approach and on the CPPs to consider when applying the QbD concepts outlined in ICH Q8(R2), Q9 and Q10 to analytical method development and optimization for three chiral compounds. In this sense, an HPLC method using a polysaccharide-based stationary phase containing a cellulose tris (4-chloro-3-methylphenylcarbamate) selector in polar organic solvent chromatography mode was considered. The effects of trifluoroacetic acid (TFA) and n-hexane concentration in a MeCN mobile phase were investigated under a wide range of column temperatures.
Good adequacies were found between the observed data obtained after using a central composite design and the expected chromatographic behaviours predicted by applying the DoE-DS methodology. The critical quality attribute represented here by the separation criterion (Scrit) allowed assessing the quality of the enantioseparation. Baseline separation for the compounds of interest in an analysis time of less than 20 minutes was possible due to the original and powerful tools applied which facilitated an enhanced method comprehension.
Finally, the advantage of the DoE-DS approach resides in granting the possibility to concurrently assess robustness and indentify the optimal conditions which are compound dependent.
The set objective of reaching baseline enatioseparation was achieved given that the three compounds studied were enantiomerically separated using a QbD compliant DoE-DS methodology in the polar organic solvent chromatography mode. Preliminary results with regards to the nature of the compounds were obtained using the HPLC method combined with circular dichroism and UV-VIS spectrometry.
A good understanding of the CPPs effects on a polysaccharide-based stationary phase containing a cellulose tris (4-chloro-3-methylphenylcarbamate) selector represents a step forward to apprehend the chromatographic behaviour of the three chiral compounds. In addition, the CQA chosen provided acceptable separation in an overall short analysis time.
Moreover, this promising methodology allowed for the optimal condition and robustness to be concurrently determined which along with its effectiveness strongly indicates a powerful strategy that can be applied in the analysis of pharmaceutical compounds.
The results presented here demonstrated the applicability of a DoE-DS methodology to identify the DS on an enantioselective case.