7.1 Overview of Multivariate Analysis as a Part of Pharmaceutical Product Design
The classical approach to the development of pharmaceutical products is all too often based on univariate thinking. There might be a huge arsenal of state-of-the-art analytical tools and a broad expert team, but still at the end of the day, the critical decision-making will be performed using experience-based and intuitive processes. Documentation of this type of decision-making can turn out to be a lengthy and difficult-to-follow process. An alternative approach would be the utilization of well-defined multivariate methods as supportive decision-making tools for combining the huge amount of data and precisely documenting the extraction of information from the huge amount of development data. The pharmaceutical business area is often a late adaptor of new thinking, but there are more and more published examples of the use of multivariate analysis (MVA) as a part of the drug development process. This type of holistic approach is a crucial element of the Quality by Design (QbD)-based development process. This chapter will introduce a broad range of examples of using MVA as a part of innovative product design.
Different MVA methods have a huge potential for diverse use throughout the whole drug development process. Examples of the MVA-based approaches are often related to exploring the chemical space and range from early phases of development (physicochemical parameters) to commercial manufacturing (process analytics). Physicochemical parameters, such as aqueous solubility, partitioning between different phases (e.g., logP), and biological barrier permeation are examples of critical parameters that can be explored and predicted using the multivariate approach. Classical examples are based on the use of molecular descriptors for cluster analysis and predictive models. Quantitative structureâactivity relationship (QSAR) models can be used for predicting the relationship between a high number of molecular descriptors and the selected response (e.g., aqueous solubility). Bergström et al. (2003) utilized this approach for exploring the chemical space of the orally administered drugs selected from the World Health Organization (WHO)âs list of essential drugs. In this work (Bergström et al., 2003), the prediction of both aqueous drug solubility and drug permeability based on multivariate tools is suggested as an approach supporting the early phase decision-making related to the oral drug absorption. Multivariate methods will be increasingly important when aiming for personalized therapies. A central element of precision medicine will be the incorporation of individual characteristics, such as genomic and metabolic capacity-related data. This type of big data can be explored using basic multivariate methods (Trygg, Holmes, & Lundstedt, 2007). Similar relationships can be used as a part of the risk management strategy (e.g., evaluation of toxicity), reducing the number of experiments (e.g., evaluation of solvent similarity as a part of experimental polymorph screening), and identifying the relationship between material properties and processing behavior (quantitative structureâproperty relationship, QSPR).
MVA is a well-established part of many analytical methods and chemical imaging is an example of an analytical method, where the amount of data can rapidly become enormous and the logical approach to extract relevant information is the use of multivariate methods (Ravn, Skibsted, & Bro, 2008). Another trend in the drug development process is the increasing use of different high-throughput screening (HTS) methods. The HTS approach is based on maximizing the number of experiments by performing them in a smaller and smaller scale, e.g., using well-plate-based technologies (Aaltonen et al., 2009). This approach coupled with faster and faster analytical methods results again in a huge amount of data, which is underpinning the need for multivariate methods. It can be a very efficient approach to start by clustering a large experimental data set with MVA and identifying trends and main factors affecting the system behavior. At the other end of the broader drug development work, process analytics is an area where multivariate methods are becoming a standard solution. Near-infrared (NIR) spectroscopy is a well-established method for understan...