Sustainable quantification of glycopyrronium, indacaterol, and mometasone along with two genotoxic impurities in a recently approved fixed-dose breezhaler formulations and biological fluids: A machine learning-augmented UV-spectroscopic approach

• Published in: Microchemical journal 2024-11, Vol.206, p.111586, Article 111586
• Main Authors: Abbas, Ahmed Emad F., Gamal, Mohammed, Naguib, Ibrahim A., Halim, Michael K., Said, Basmat Amal M., Ghoneim, Mohammed M., Mansour, Mohmeed M.A., Salem, Yomna A.
• Format: Article
• Language: English
• Subjects: Chemometric models; Genotoxic impurities; Glycopyrronium, Indacaterol, and Mometasone; Sustainable pharmaceutical analysis; UV-spectroscopy
• ISSN: 0026-265X
• DOI: 10.1016/j.microc.2024.111586

[Display omitted] •A novel Machine Learning-chemometrics approach exploiting the Kennard Stone Clustering Algorithm.•Simultaneous quantification of recently approved combinations and genotoxic impurities.•Hazardous solvents eliminated by GSST and SDAGI screening of greener alternatives.•Comprehensive greenness assessments by NEMI, ComplexGAPI, AGREE, and carbon footprint.•Apply new blueness and whiteness evaluations using BAGI and RGB12. This study presents an innovative, sustainable approach for the simultaneous quantification of glycopyrronium (2–14 μg/mL), indacaterol (6–18 μg/mL), and mometasone (4–20 μg/mL) in a recently approved fixed-dose breezhaler formulations and biological fluids, along with two genotoxic impurities: methyl para-toluene sulfonate (2–10 μg/mL) and 4-dimethylamino pyridine (2–10 μg/mL). We developed robust UV spectrophotometric machine-learning chemometric models to address the limitations of existing chromatographic methods. The calibration set was carefully selected at five concentration levels using the multilevel-multifactor experimental design, resulting in 25 calibration mixtures. The Kennard-Stone Clustering Algorithm was employed to construct a representative 13-mixture validation set, overcoming biases associated with random data splitting. Five chemometric models (CLS, PCR, PLS, GA-PLS, and MCR-ALS) were rigorously evaluated, with MCR-ALS demonstrating superior performance. This model achieved 98–102 % recovery percentages for all analytes, with low root mean square error of calibration and prediction of (RMSEC: 0.0225 to 0.5246) and (RMSEP: 0.0039 to 0.4226). The method exhibited excellent relative root mean square error of prediction (RRMSEP: 0.1306 to 0.8517 %), a negligible bias-corrected mean square error of prediction (BCMSEP: −0.0073 to 0.0025), and good sensitivity (LOD: 0.022 to 0.893 μg/mL) across all analytes. Green solvents were selected using the Green Solvents Selection Tool and Greenness Index Spider Charts. The method’s sustainability was comprehensively evaluated using seven state-of-the-art assessment tools. This approach not only offers a green alternative to traditional chromatographic methods but also ensures high accuracy in quantifying both active ingredients and genotoxic impurities, thereby enhancing pharmaceutical quality control and patient safety.