Defense Date

5-20-2021

Graduation Date

Summer 8-7-2021

Availability

Immediate Access

Submission Type

dissertation

Degree Name

PhD

Department

Pharmaceutics

School

School of Pharmacy

Committee Chair

Carl A. Anderson

Committee Member

James K. Drennen, III

Committee Member

Peter L.D. Wildfong

Committee Member

Ira S. Buckner

Committee Member

Zhenqi Shi

Keywords

Feedforward Control, Quality by Design, Process Analytical Technology

Abstract

Objectives

For a fluid bed film coating process to consistently deliver quality products, its control system needs to be robust against the variability of input materials and environmental disturbances. Presently, limited studies have been reported to understand the effects and interactions of the material attributes, environmental variables, and process parameters on the product in vitro drug dissolution. A control system can be developed with a proper understanding of the coating process, by adjusting the process parameters in feedback and feedforward manners to compensate for the undesired effect caused by disturbances, and ensure consistent product quality.

Methods

The control system was developed and evaluated using a quality by design approach. The formulation variables, material attributes, and process parameters of the coating process were systematically assessed using Ishikawa and failure mode and effect analysis. The risk assessment was followed by a fractional factorial design to screen the criticality of four process variables: product temperature, airflow volume, atomization air pressure, and inlet air relative humidity. The size distribution of the input granules was constrained to a narrow range in the factorial design. The information gained from the screening study was used to guide the response surface design for process modeling, in which granule size distribution, relative humidity, inlet air volume, and target coating weight gain were investigated, and the studied response was in vitro dissolution. Using two regression methods (partial least squares and Gaussian process regression) and two curve-fitting methods (Weibull function and principal component analysis) in conjunction, four modeling approaches were applied to analyze the experimental data and establish the process models. A control system was subsequently developed. The feedback loops relied on the real-time measurements of near-infrared spectroscopy (NIR) to stabilize the in-process moisture level and determine the process endpoint. The feedforward components were built upon the process models. The controllers modified the target weight gain and airflow volume to accommodate the undesired size distribution of input granules and relative humidity. The combined feedforward-feedback control system was evaluated by comparing the control performance with and without the feedforward elements, using Monte Carlo simulation and 12 additional test runs.

Results

The initial risk assessment and the statistical designs of experiments identified the critical material attributes and process parameters and elucidated their impacts on the coating process and final product quality. The in-process moisture level was found to play an essential role in preventing batch collapse and improving coating efficiency. The hydration of the active pharmaceutical ingredient (API), theophylline, was identified as a high-risk failure mode that requested proper control. The in-line NIR models had 0.3% and 0.5% errors in predicting the moisture level and coating weight gain. The process models were established using different modeling algorithms including partial least squares regression, Gaussian process regression, Weibull function fitting, and principal component analysis. The partial least squares regression model coupled with the Weibull function as a dissolution curve-fitting method outperformed the other models as it had the lowest error profile and great simplicity for control application. The feedforward controllers were established by mathematically transforming the process model into an optimization problem, which searched for the best solution of process parameters given the initial condition of material attributes and environmental variables. The tolerance space of the coating process supervised by the feedforward-feedback control system was established.

Conclusion

The combined feedforward-feedback control system reduced batch failures and improved product quality consistency in both Monte Carlo simulations and test batches. The combined control system also showed robustness against the variability of incoming material attributes, which would grant pharmaceutical companies tremendous flexibility in choosing the sources of raw materials.

Language

English

Download

Share

COinS