Usefulness of Applying Partial Least Squares Regression to T2 Relaxation Curves for Predicting the Solid form Content in Binary Physical Mixtures

• Published in: Journal of pharmaceutical sciences 2023-04, Vol.112 (4), p.1041-1051
• Main Authors: Chiba, Yuya, Okada, Kotaro, Hayashi, Yoshihiro, Kumada, Shungo, Onuki, Yoshinori
• Format: Article
• Language: English
• Subjects: Active pharmaceutical ingredient; NMR relaxometry; Partial least squares regression; Process analytical technology; Solid form; T2 relaxation
• ISSN: 0022-3549; 1520-6017
• DOI: 10.1016/j.xphs.2022.11.028

This study applied partial least squares (PLS) regression to nuclear magnetic resonance (NMR) relaxation curves to quantify the free base of an active pharmaceutical ingredient powder. We measured the T2 relaxation of intact and moisture-absorbed physical mixtures of tetracaine free base (TC) and its hydrochloride salt (TC·HCl). The obtained T2 relaxation curves were analyzed by two methods, one using a previously reported T2 relaxation time (T2), and the other using PLS regression. The accuracy of estimating TC was inadequate when using previous T2 values because the moisture-absorbed physical mixtures showed biphasic T2 relaxation curves. By contrast, the entire measured whole of the T2 relaxation curves was used as input variables and analyzed by PLS regression to quantify the content of TC in the moisture-absorbed TC/TC·HCl. Based on scatterplots of theoretical versus predicted TC, the obtained PLS model exhibited acceptable coefficients of determination and relatively low root mean squared error values for calibration and validation data. The statistical values confirmed that an accurate and reliable PLS model was created to quantify TC in even moisture-absorbed TC/TC·HCl. The bench-top low-field NMR instrument used to apply PLS regression to the T2 relaxation curve may be a promising tool in process analytical technology.