Predictive Modeling of Pharmaceutical Unit Operations

Preetanshu Pandey, Rahul Bharadwaj

• Publisher: Woodhead Publishing
• Year: 2016
• Print ISBN: 9780081001547
• eBook ISBN: 9780081001806
• Topic: Medicine
• Subtopic: Pharmaceutical, Biotechnology & Healthcare Industry

The use of modeling and simulation tools is rapidly gaining prominence in the pharmaceutical industry covering a wide range of applications. This book focuses on modeling and simulation tools as they pertain to drug product manufacturing processes, although similar principles and tools may apply to many other areas. Modeling tools can improve fundamental process understanding and provide valuable insights into the manufacturing processes, which can result in significant process improvements and cost savings. With FDA mandating the use of Qualityby Design (QbD) principles during manufacturing, reliable modeling techniques can help to alleviate the costs associated with such efforts, and be used to create in silico formulation and process design space. This book is geared toward detailing modeling techniques that are utilized for the various unit operations during drug product manufacturing. By way of examples that include case studies, various modeling principles are explained for the nonexpert end users. A discussion on the role of modeling in quality risk management for manufacturing and application of modeling for continuous manufacturing and biologics is also included.- Explains the commonly used modeling and simulation tools- Details the modeling of various unit operations commonly utilized in solid dosage drug product manufacturing- Practical examples of the application of modeling tools through case studies- Discussion of modeling techniques used for a risk-based approach to regulatory filings- Explores the usage of modeling in upcoming areas such as continuous manufacturing and biologics manufacturingBullet points

1

Modeling of drug product manufacturing processes in the pharmaceutical industry

P. Pandey¹, R. Bharadwaj² and X. Chen¹, ¹Drug Product Science and Technology, Bristol-Myers Squibb, New Brunswick, NJ, United States, ²Rocky-DEM, Houston, TX, United States

Abstract

This chapter introduces the various modeling techniques that are typically used in the pharmaceutical industry. The different modeling approaches are classified into two main categories, namely, physics-based models and empirical models, and examples of each are provided. The different stages of model development, including calibration, validation, verification, and maintenance, are discussed. An introduction to modeling techniques associated with each unit operation in a typical process workflow of a solid dosage form development is also included. In addition, the importance of understanding model limitations and desired outcomes before model selection is highlighted.

Keywords

Modeling techniques; unit operations; model development; drug product; oral solids

1.1 Introduction

The use of modeling techniques is becoming more common in the pharmaceutical industry, covering a wide range of application areas, such as drug discovery, pharmacokinetics and pharmacodynamics, biopharmaceutical (effect of pH, food, etc.), drug substance (chemistry and manufacturing), and drug product. This book primarily focuses on modeling as it pertains to the drug product manufacturing processes, although similar modeling principles and techniques apply to other areas as well. The drug product manufacturing processes spans the entire range that includes the preparation of a formulation to the processing and packaging of the final drug product.

The use of modeling tools enable fundamental process understanding and provide insights into the unit operations, which can lead to significant process improvements and cost savings. With FDA mandating the use of Quality by Design (QbD) principles during manufacturing, the establishment of a formulation and process design space during drug product development is imperative (U.S. Department of Health and Human Services FDA, 2006, 2009a,b; Pandey and Badawy, 2016; Pandey et al., 2006d; Yu et al., 2014) This means that on the formulation end, the effects of variations in input material properties (API and excipients), referred to as potential critical material properties (CMAs), on drug product critical quality attributes (CQAs) have to be studied. A similar effort is needed on the manufacturing process where the effects of potential critical process parameters (CPPs) from each unit operation on drug product CQAs are to be established. Additionally, there is significant interaction between formulation and process components that needs to be characterized.

Given that a typical pharmaceutical formulation has 5–6 different components (API and excipients—filler, disintegrant, binder, flow aid, lubricant), and 5–6 different unit operations (such as blending, granulation, milling, compression, coating, packaging) are involved, studying the effects of variations of each and their interactions can be a tedious and expensive experimental study. Reliable modeling and simulation tools can help tremendously to alleviate such an elaborate effort. Appropriate modeling tools can be used to create both in silico formulation and process design space by varying potential CMAs and CPPs and studying their impact on drug product CQAs. (U.S. Department of Health and Human Services FDA, 2006, 2009a,b) The models can be used to conduct a sensitivity analysis that will in turn reduce the number of variables that need to be studied experimentally. Models can also be used to identify edge of process failure, which can be expensive experimentally. A good process and scale-up model will also allow for more small-scale experiments (cost savings) with limited amount of work required at the larger scale. In an era of “speed to patient,” there is often limited availability of API during early development, and certain formulation and manufacturing process selection decisions that could have a long-term impact on the product are required. For example, one may have to choose between wet granulation and dry granulation process with only a limited amount of information on API powder properties. In order to gauge the risk levels appropriately at an early stage, material-sparing tools such as minipiloting and modeling are becoming increasingly important (LaMarche et al., 2014) Therefore, the use of predictive tools is not only important during the later stage of drug product development but also during early stages of development.

In a QbD paradigm, regulators are looking for a risk-based approach towards drug product development. A risk assessment identifies the formulation, process, and any other risks that can affect a drug product CQA and identifies the failure modes, their probability of detection and occurrence, and their severity levels. Once the knowledge gaps are appropriately identified and ranked, a mitigation plan is put in place and executed. A process control strategy is put in place when a certain risk can’t be eliminated totally, in which case a residual risk level is identified and monitored. From a regulatory perspective, the use of predictive tools such as modeling and simulation can enable a quantitative risk assessment, facilitating the quality assessment of manufacturing processes. It can also support the evaluation of control strategies by demonstrating system capabilities to handle multiple sources of variability (formulation and process). FDA has provided guidelines that in part discuss the role and usage of models in QbD (U.S. Department of Health and Human Services FDA, 2006, 2009a,b) ICH recommends categorization of models based on their impact to drug product quality. The three categories on the types of models based on impact include high impact, medium impact, and low impact models. A high impact model is defined as one where the model is the sole predictor of the drug product quality, a medium impact is defined as ones which are important for assuring quality of the product but are not the sole indicators of the drug product quality, and low impact models are defined as ones that are used to support formulation and process development type of activity. The high impact models (e.g., chemometric model for drug product assay) are not encountered routinely but when they are they undergo the most scrutiny (given their impact). High impact models require development of appropriate calibration and validation procedures with the ability to discern OOS (out of specification) batches, development of a model monitoring system (model maintenance and updates during the life cycle), and tracking and trending of the process within the Quality system. In order to categorize models based on their mode of implementation, they can be classified into models for supporting analytical methods (e.g., NIR methods), models for supporting process design (e.g., design space and scale-up models such as DEM, CFD, etc.), and models for process monitoring and control (e.g., multivariate statistical models).

The following sections explain some of the modeling methods used in process design/control and also the resources required for implementing them.

1.2 Modeling techniques

The various kinds of modeling approaches currently being used can be broadly classified into two categories, namely physics-based models and empirical models. The physics-based models utilize fundamental first principles (conservation of mass, momentum, and energy) to predict the behavior of the fluid or powder in the process. The more commonly used modeling approach for granular materials is the discrete element method (DEM). Continuum methods for modeling include finite element method (FEM), computational fluid dynamics (CFD), and, more recently, hybrid methods that utilize a combination of one or two other methods.

1.2.1 First principle predictive models

1.2.1.1 Discrete element method

DEM is a powerful predictive tool commonly used in applications that include powder mixing (blending), powder conveying (e.g., hopper discharge, die filling, twin screw extruder feeding, and conveying), milling (by including breakage kernels), granulation, and tablet movement during film coating. A good review of application of process modeling in the pharmaceutical industry using DEM was provided by Ketterhagen et al. (2009).

The discrete element method takes into account the forces acting on each individual particle and then integrates them over time (Newton’s laws of motion) to get its velocity and subsequent position. The forces in the model can include gravity, contact, cohesive, and fluid forces. Since pharmaceutical powder size distributions are typically in the micron meter ranges, this approach becomes computationally expensive if the actual size of the powder is used in the model due to large number of particles in the simulation. Hence, it is common practice to scale the particle sizes in the DEM model to study mixing and flow patterns. In addition, particles are generally assumed to be spheres due to the additional complexity that arises by using different non-spherical shape representations in the model.

Particle shapes in DEM have traditionally been represented by a glued-sphere approach where spheres are rigidly glued together to form close to the desired shape (Favier et al., 1999) Although this representation of the shape has various limitations, it is relatively easy to program and implement and hence is available in various commercial software (EDEM, Star CCM, PFC 3D) (Nakamura et al., 2013) Limitations with this method include restrictive computational times, poor representation of shapes with large aspect ratios, and large bulk friction produced due to particle interlocking (due to the bumpy surface) (Suzzi et al., 2012) Another, more recently evolving technique, such as that used in commercial software Rocky DEM, applies a new method using a polyhedral shape representation to overcome the limitations associated with the glued sphere approach. The ability to scan the exact particle shape into Rocky DEM, such as a tablet, enables a virtual design environment for studying scale-up of processes such as tablet coating and blending. Although the contact detection using the polyhedral shape is computationally intensive, the ability to use the graphics processing unit (GPU) with thousands of cores within the Rocky DEM solver, enable practical run times. In addition, most commercial DEM software allow the import of process equipment in native CAD drawing formats which make for efficient comparison between different process equipment performance.

Other challenging aspects for DEM are the ability to incorporate the physics of a liquid being present in the system (e.g., water in granulation), and cohesive forces between particles, although there are existing models (or kernels) in literature that have been used before (Asmar et al., 2002, 2003; Matuttis and Schinner, 2001). Although there are certain limitations with the use of DEM, advances in computer hardware power and technology are making this a practical tool and a well-calibrated DEM model can provide valuable insights into the process (Bharadwaj, 2012; Pandey et al., 2006c; Toschkoff et al., 2013, 2015; Hilton et al., 2013; Kulju et al., 2016).

1.2.1.2 Computational fluid dynamics

CFD is widely used to model fluid flow in a process and finds its pharmaceutical applications in modeling mixing in agitated vessels (API drying, mixing in bioreactors, lyophilization, suspension preparation, etc.), spray drying, and air flow in systems such as fluidized bed, film coaters etc. (van Wachem et al., 2001; Petitti et al., 2013; Al...

Table of contents

• Cover image
• Title page
• Table of Contents
• Copyright
• List of contributors
• Predictive modeling of pharmaceutical unit operations
• Preface
• 1. Modeling of drug product manufacturing processes in the pharmaceutical industry
• 2. Quality risk management for pharmaceutical manufacturing: The role of process modeling and simulations
• 3. Powder flow and blending
• 4. Dry granulation process modeling
• 5. Mechanistic modeling of high-shear and twin screw mixer granulation processes
• 6. Fluid bed granulation and drying
• 7. Modeling of milling processes via DEM, PBM, and microhydrodynamics
• 8. Modeling of powder compaction with the drucker–prager cap model
• 9. Modeling approaches to multilayer tableting
• 10. Computational modeling of pharmaceutical die filling processes
• 11. Modeling tablet film-coating processes
• 12. Modeling in pharmaceutical packaging
• 13. Continuous secondary process selection and the modeling of batch and continuous wet granulation
• 14. Process modeling in the biopharmaceutical industry
• Index