Department of chemical and biotechnological engineering, Université de Sherbrooke, Sherbrooke, Quebec, Canada.
Pfizer Canada, Kirkland, Quebec, Canada.
Pharm Dev Technol. 2023 Jan;28(1):40-50. doi: 10.1080/10837450.2022.2162081. Epub 2023 Jan 12.
Tablet sticking is a continuous accumulation of pharmaceutical powder onto tooling surfaces during compression. Its occurrence greatly impacts tablet productivity, quality attributes, and tooling age. In a previous study, the authors proposed a multivariate data analysis approach to gain insights into tablet sticking directly on the industrial stage. The objective was to determine the combination of factors that could help distinguish between batches affected and unaffected by sticking. The present study aims to generalize this approach by extending it to quantitative predictions of punch sticking intensity. A total of 345 variables was gathered on 28 industrial batches of an ibuprofen and methocarbamol-based formulation.
Using PLS regression models, it was shown that the association of granulation duration and compression force allows to significantly explain ∼60% of sticking variations of studied formulation. In addition, unlike the classification models developed in the earlier work, the validation residues in the present study were found to be normally distributed (Shapiro-Wilks value = 0.96) and independent from the target variable (R = 9.5%).
片剂粘连是在压片过程中,药物粉末不断积聚在压模表面上。它的发生极大地影响了片剂的生产效率、质量属性和压模寿命。在之前的研究中,作者提出了一种多元数据分析方法,可以直接在工业阶段深入了解片剂粘连问题。目的是确定哪些因素的组合可以帮助区分受粘连影响和不受粘连影响的批次。本研究旨在通过将其扩展到定量预测冲头粘连强度来推广这种方法。共收集了 28 批布洛芬和甲灭酸基配方的 345 个变量。
研究表明,使用 PLS 回归模型,颗粒化时间和压缩力的关联可以显著解释研究配方中约 60%的粘连变化。此外,与早期工作中开发的分类模型不同,本研究中的验证残差被发现呈正态分布(Shapiro-Wilks 值=0.96),并且与目标变量独立(R=9.5%)。