Tablet Quality-Prediction Model Using Quality Material Attributes: Toward Flexible Switching Between Batch and Continuous Granulation

Purpose The purpose of the study was to develop a model to predict the critical quality attribute (CQA) of tablets during continuous and batch manufacturing using only critical material attributes (CMAs). Methods Experiments were performed using ethenzamide as the active pharmaceutical ingredient processed with batch and continuous high-shear granulators. The disintegration time of tablets was defined as the CQA, and the particle-size distribution of granules and tablet hardness were defined as the CMAs. We first investigated the influence of granulation conditions on particle-size distribution during batch and continuous granulation. We then proceeded to construct the CQA estimation model by producing tables using batch and continuous granulation. Results The results indicated the similarity of the granulation mechanisms, as observed by the bimodality of the distributions and the significant causal factors. Principal component analysis revealed that the CQA was influenced strongly by the particle-size distribution and that the CMA-CQA correlations were similar for both processes. Finally, a model based on partial least-squares regression could be developed that could reasonably estimate the CQA using CMAs without involving any process parameters. Conclusion This approach of using process-independent CQA prediction could enable flexible switching between batch and continuous manufacturing during a product life cycle, thus offering new possibilities for efficient life cycle management.

Automated summary

The study aimed to develop a model to predict the critical quality attribute (CQA) of tablets during continuous and batch manufacturing using only critical material attributes (CMAs). By investigating the influence of granulation conditions on particle-size distribution and constructing a CQA estimation model, it was found that the CQA was strongly influenced by the particle-size distribution and that the CMA-CQA correlations were similar for both processes. A model based on partial least-squares regression could reasonably estimate the CQA using CMAs without involving any process parameters.