Formulation Tools for Pharmaceutical Development

J E Aguilar

• Editorial: Woodhead Publishing
• Año: 2013
• ISBN:
  9781908818508
• Categoría: Medicina
• Categoría: Farmacología

A range of new and innovative tools used for preformulation and formulation of medicines help optimize pharmaceutical development projects. Such tools also assist with the performance evaluation of the pharmaceutical process, allowing any potential gaps to be identified. These tools can be applied in both basic research and industrial environment. Formulation tools for pharmaceutical development considers these key research and industrial tools.Nine chapters by leading contributors cover: Artificial neural networks technology to model, understand, and optimize drug formulations; ME_expert 2.0: a heuristic decision support system for microemulsions formulation development; Expert system for the development and formulation of push-pull osmotic pump tablets containing poorly water-soluble drugs; SeDeM Diagram: an expert system for preformulation, characterization and optimization of tables obtained by direct compression; New SeDeM-ODT expert system: an expert system for formulation of orodispersible tablets obtained by direct compression; and 3D-cellular automata in computer-aided design of pharmaceutical formulations: mathematical concept and F-CAD software.

• Coverage of artificial intelligence tools, new expert systems, understanding of pharmaceutical processes, robust development of medicines, and new ways to develop medicines
• Development of drugs and medicines using mathematical tools
• Compilation of expert system developed around the world

1

Introduction

Johnny Edward Aguilar

The way in which medicines are developed is changing and the regulatory environment is also changing. Consequently, formulators require full understanding of a product and its process of development. In addition, it is desirable for formulators to detect any gaps in drug formulas which, if not addressed, could be linked to inadequate quality or problems with the product. Different methodologies have been implemented to try to improve the existing pharmaceutical process in the industrial environment, such as lean and six sigma. High variability and continuous problems during manufacturing could be avoided by ensuring that a good product design is used in the initial stages when developing new medicine. This is not an easy step because of complex non-linear relationships between the formulation composition, process conditions, and product properties. In most cases, a formulation consists of a drug, a number of formulation ingredients, and process conditions, interactions between which affect the quality of the final product. Thus, formulation design is based on a multi-dimensional space that is difficult to conceptualize for scientists working in this field (Rowe and Roberts, 1998; Shao et al., 2007).

A good understanding of processes and interactions between different components of formulations is key to understanding the complex relationships in product formulations. This can be attained using appropriate tools that avoid unnecessary trials in the laboratory and optimize this goal in an efficient manner. These kinds of tools also provide information which can be used in the optimization of the formulation, so that the final formulation is obtained by fixing any gaps previously detected by these formulation tools. The tools also assist formulators in avoiding problems related to quality which can occur in the subsequent development phase or during commercial manufacturing.

Understanding of these processes and implementation of continuous improvements are becoming ever more important. Therefore, such tools and their derivate software are highly appreciated for a better understanding of our processes by formulators, scientists, and similar professionals in the pharmaceutical industry or research centers. It is noted that the development of such tools has increased recently, for example they are being used in design or development of formulations such as expert systems, artificial intelligence technologies and tools such as artificial neural networks, etc.

The methodologies termed lean and six sigma are commonly used in routine manufacturing. These apply basic statistics to evaluate the behavior of a process, permitting identification of an advantageous change or detection of a possible trend beforehand. However, there are alternatives that can be used to reach this goal, such as preformulation and formulation tools. In contrast to the traditional statistical approach, these tools allow analysis of complex and non-linear relations and provision of additional information that can be used during the analysis phase. They can help to propose assertive solutions during optimization. For example, SeDeM methodology, detailed in this book, can provide information on differences in rheology properties in a powdered formulation for tablets, which can be used when comparing suppliers used for raw materials. This tool uses routine tests of pharmacopeia to allow identification of variances between two different suppliers of the same component, excipient, or drug substance, and provides information on any gaps that must be corrected before executing the pilot and commercial batches. Analysis using this tool ensures a successful formula and a robust validation.

Factors related to productivity and reduction of cost are also taken into account when developing medicines. The tools described in the subsequent chapters can assist with cost reduction by providing information to lead to a better understanding of formulations under development, and by decreasing the lead time in development and avoiding unnecessary trials because the old (expensive) methodology trial error is not applied. The use of these tools is highly appreciated by pharmaceutical companies and research centers as good product design leads to lean processes and cost improvements.

During the lab phase, the physical and chemical properties of a drug are determined and then the desired dosage form and critical attributes are designed. The design of experiment is performed in the pilot scale, which helps to obtain a detailed understanding of the different steps implemented in the process. The data are generated and used in the scale-up and the subsequent phase corresponding to commercial manufacturing. Preliminary design space and the criteria of final specifications are determined during this phase.

Review of the design space is then initiated. This information related to the manufacturing process is used in improvement studies and for future troubleshooting, which can be necessary in routine manufacturing of commercial batches.

All these phases are strictly linked and require exchange of information in trying to understand the complex non-linear relationships between the formulation composition, process conditions, and product properties. This information is not only useful at the development stage, but also subsequently for identifying root causes and supporting implementation of effective corrective and preventive actions.

The pharmaceutical development phase provides information critical to form the basis of process understanding. This can be used for various new technologies; it facilitates scientists to reach a better understanding of the chemical and physical phenomena of the drug. There are some cases wherein this learning is compiled on paper, in electronic data, books or in the personal experience of pharmacists or professionals working in development of medicines; however, there are also unpublished experiences and knowledge, which are therefore unknown to the scientific community. If that information were treated and compiled using appropriate software or managed with an adequate methodology, it could provide a high probability of a good and effective solution in case of problems with the formulation. The use of an expert system or other artificial intelligence tools is recommended to achieve this. ‘Expert system’ (ES) is a versatile term, as ES occur in many disciplines such as economics, mathematics, etc; however there are some common definitions:

– ‘Computer program that draws upon the knowledge of human experts captured in a knowledge base to solve problems that normally require human expertise’ (Partridge and Hussain, 1994).

– ‘The label “expert system” is broadly speaking, given to a computer program intended to make reasoned judgements or give assistance on a complex area in which human skills are fallible or scarce’ (Lauritzen and Spiegelhalter, 1988).

There is a need to introduce newer methods in mathematical modeling of stochastic phenomena, such as power behavior which could be of a single component or a mixture in a final formulation. However, it is important to have an overview of the main directions of past modeling trends. One of the main objectives in the second half of the twentieth century was to develop artificial intelligence-based modeling methods for aiding design of pharmaceutical dosage forms. Artificial intelligence can capture the knowledge of a formulation expert, document it, and make it available and user-friendly. Turban compared artificial intelligence (AI) with—as he called it—natural intelligence (NI) of experts as follows (Turban, 1995):

NI depends on persons, which results in a dependency on personnel changes.

NI is difficult to transfer, whereas AI can be moved from one computer to another.

AI can reduce costs.

AI is consistent, decisions are traceable and can easily be documented.

AI is not creative.

NI uses a wider context of experience to solve problems.

These methods are restricted to sequential processing of knowledge; however, a different approach is to use neural networks. As the name implies, artificial neural networks are inspired by the functionality of the human brain. The artificial neuron takes one or more inputs, each multiplied with a weight factor, and potentially creates an output which is forwarded to another neuron. Whether an output is generated or not depends on the inputs, which must exceed a defined threshold. The threshold activation is computed by transformation functions, which can be linear or non-linear. Compared with expert systems, neural networks need short development time, but need to be trained. The training consists of linking inputs and outputs and adapting weight-values until inputs give a result that is close to the experimentally determined result.

A classic algorithmic overview of pharmaceutical development indicates that it requires a recompilation of knowledge with a foundation in many disciplines that could assist with understanding drug substances and the different interactions with excipients. It is important to consider the variables used during the process which could potentially impact the quality of the medicines, and to avoid those considered unnecessary. However, as previously mentioned, this is not an easy task because they are not universal theories or principles. Mechanisms can be identified by those with professional experience; however, innovative preformulation and formulation tools are under development which could help reach better understanding of these complex relations. These tools could suggest a model for use to define the final formulation and the appropriate process to apply, therefore having a high impact on the final formulation.

Finally it is concluded that the life sciences industry is changing rapidly and the historical rules, regulations, and government oversight are under pressure to modernize. The recent introduction of Quality Systems and Quality by Design (QbD) concepts has challenged the traditional view that simple compliance with the basic Good Management Practices (GxP) rules is enough to satisfy stakeholders, regulators, and patients. A better understanding of processes is required. The strategies used for development of new medicines are also changing and they are being carried ...