Developing a pharmaceutical Fluid bed granulation & drying process via design of experiments based on a multivariate model

Fluid bed granulation and drying (FBG&D) is an important process within a drug product manufacturing framework since it sets the properties of the intermediate product (granules) that make the final drug product (e.g., tablet). In a typical FBG&D process for drug product manufacturing, an aqueous binder solution of a given concentration is sprayed into the fluidized bed of particles through spray nozzle(s). Once the appropriate amount of binder is delivered, the spray is stopped, and the granules are dried to remove the excess moisture. There are several important process variables that influence the parallel mechanisms of fluidization, granulation and drying (e.g., material properties of the powder bed and operational control variables (Emenike et al., 2021)). Hence, FBG&D is a complex process to develop due to interactions among multiple physical phenomena. Despite its complexity, this process is regularly used in industries because of several benefits, including ease of automation and enhanced manufacturing efficiency by reducing the process footprint and cost. Hence, a large volume of work already exists on the scientific understanding of FBG&D (Nascimento et al., 2022). Different aspects of FBG&D in terms of scale-up strategies (Gavi and Dischinger, 2021), process analytical techniques for monitoring (Burggraeve et al., 2013, Rimpiläinen et al., 2012) and physical understanding of the granulation mechanism (Rajniak et al., 2009) have been discussed in the literature.

Due to the complex interactions of several process variables, many studies in the literature (Rambali et al., 2001, Zhao et al., 2019) propose comprehensive statistically designed experiments to understand the relationship between the process inputs (material properties and operating control variable settings) and outputs (granule and tablet properties). For example, Rambali et al (Rambali et al., 2001). had as many as 30 experiments in their design. Executing extensive DOEs like this at scale is not sustainable in the current business environment that favors lean and fast development. Moreover, in many cases it is not necessary to resolve and understand every variable interaction for the development of an industrial process. The identified process risks should determine which parameters are evaluated. For example, supersaturated statistical DOEs are often used in industrial studies to significantly reduce the number of experiments while providing relevant information (Gilmour, 2006).

We present a workflow for defining the operating conditions of a pharmaceutical FBG&D unit. Pharmaceutical process development is an interesting problem since it must satisfy regulatory requirements, meet the business objectives of reducing the time to market and being cost effective, and ensure the quality of the product. This work addresses an important gap in the literature as publications rarely document commercial FBG&D process development with explicit consideration of regulatory perspectives.

Traditionally, the operating conditions in a drug product manufacturing process are registered as targets and PARs (proven acceptable ranges) (FDA, 2017) of the process variables. The PAR of a process variable is determined by perturbing it while keeping the other variables constant and demonstrating that the drug product CQAs (critical quality attributes) (FDA, 2017) are met. It is typically a univariate experimental exercise that does not contain information regarding interactions with other relevant process variables. Operating within a PAR will ensure that the product CQAs are met if the other variables are still at target. If multiple variables deviate from their targets at the same time, the univariate nature of the PARs is no longer sufficient to ensure the process will produce acceptable quality material. However, this situation can be resolved with a multivariate design space where parameter interactions are understood. A design space is defined as “the multi-dimensional (or multivariate) combination and interaction of input variables and process parameters that have been demonstrated to provide assurance of quality” (FDA, 2017). Identifying an operating design space can result in less process deviations and product quality issues during routine commercial manufacturing.

This work highlights the identification of the operating design space of a FBG&D process. Since properties of the granules are key to get acceptable tablets, the granulation operating design space must result in granules that can provide robust and acceptable tablet CQAs. Granulation is a size enlargement (growth) operation, and growth rate is a key factor for granulation. Therefore, in a granulation operating design space, the two extreme corners should be the ones resulting in the lowest and highest growth rate (Parikh, 2021), referred to as under-granulation and over-granulation corners, respectively. Characteristics of under-granulation are small sized granules, often with presence of ungranulated fines. Characteristics of over-granulation are large granules, often with a wide size span and high density. If these two extreme corners of an operating design space meet the tablet CQAs, then all other operating conditions bound by these two extremes should also pass under the assumption of linearity and continuity.

A PLS (partial least squares) model was used to identify the dominant process variables and demonstrate the parameter settings at the two extreme granulation corners. These extreme corners were then confirmed experimentally and shown to meet the drug product CQAs. This workflow demonstrates a combination of experiments and data driven model (like PLS) for proposing an operating design space for the FBG&D operation. The understanding from the experiments and PLS model was used to adjust the processing condition to arrive at a robust operating design space.

Many forms of mechanistic models for FBG&D are available in the literature (Ochsenbein et al., 2019, Sen et al., 2014, Hayashi et al., 2020). One of the most relevant and recent industrial applications of a mechanistic FBG&D model is documented by Ochsenbein et al (Ochsenbein et al., 2019). Their dynamic model can predict the %LOD (loss on drying) of the granules throughout the granulation and drying process while accounting for parameter uncertainties. They used the model to test for process robustness within an operating space and discussed its application for developing PARs. Their model can be effectively implemented in a FBG&D process development workflow. However, we followed a simpler PLS model-based approach to analyze the experimental data and determine the specific experiments to run to confirm a robust operating design space. There are many commercial packages available for PLS analysis; therefore, custom coding by a process engineer is unnecessary to adapt our approach.

Read more

Maitraye Sen, Sydney Butikofer, Chad N. Wolfe, Shashwat Gupta, Adam S. Butterbaugh, Developing a pharmaceutical Fluid bed granulation & drying process via design of experiments based on a multivariate model, Chemical Engineering Research and Design, 2024, ISSN 0263-8762,