A Comparative Approach to Screen the Capability of Raman and Infrared (Mid- and Near-) Spectroscopy for Quantification of Low-Active Pharmaceutical Ingredient Content Solid Dosage Forms: The Case of Alprazolam. Issue 6 (June 2020)

• Record Type: Journal Article
• Authors: Makraduli, Liljana; Makreski, Petre; Goracinova, Katerina; Stefov, Stefan; Anevska, Maja; Geskovski, Nikola
• Is Part Of: Applied spectroscopy. Volume 74:Issue 6(2020)
• Page Start: 661
• Page End: 673
• Publication Date: 2020-06
• Subjects: Alprazolam -- low-content active pharmaceutical ingredient -- API -- quantification -- partial least squares -- PLS -- Raman spectroscopy -- mid-infrared -- near-infrared; Spectrum analysis -- Periodicals; 543.505
• DOI: 10.1177/0003702820905367
• Languages: English
• ISSNs: 0003-7028

Abstract

Content uniformity is a critical attribute for potent and low-dosage formulations of active pharmaceutical ingredient (API) that, in addition to the formulation parameters, plays pivotal role during pharmaceutical development and production. However, when API content is low, implementing a vibrational spectroscopic analytical tool to monitor the content and blend uniformity remains a challenging task. The aim of this study was to showcase the potentials of mid-infrared (MIR), near-infrared (NIR), and Raman spectroscopy for quantitative analysis of alprazolam (ALZ) in a low-content powder blends with lactose, which is used as a common diluent for tablets produced by direct compression. The offered approach might be further scaled up and exploited for potential application in the process analytical technology (PAT). Partial least square and orthogonal PLS (OPLS) methodologies were employed to build the calibration models from raw and processed spectral data (standard normal variate, first and second derivatives). The models were further compared regarding their main statistical indicators: correlation coefficients, predictivity, root mean square error of estimation (RMSEE), and root mean square error of cross-validation (RMSEEcv). All statistical models presented high regression and predictivity coefficients. The RMSEEcv for the optimal models was 1.118, 0.08, and 0.059% for MIR, NIR, and Raman spectroscopy, respectively. The scarce information content extracted from the ALZ NIR spectra and the major band overlapping with those from lactose monohydrate was the main culprit of poor accuracy in the NIR model, whereas the subsampling instrumental setup (resulting in a non-representative spectral acquisition of the sample) was regarded as a main limitation for the MIR-based calibration model. The OPLS models of the Raman spectra of the powder blends manifested favorable statistical indicators for the accuracy of the calibration model, probably due to the distinctive ALZ Raman pattern resulting in the largest number of predictive spectral points that were used for the mathematical modeling. Furthermore, the Raman scattering calibration model was optimized in narrower scanning range (1700–700 cm −1 ) and its prediction power was evaluated (root mean square error of prediction, RMSEP = 0.03%). Thus, the Raman spectroscopy presented the most favorable statistical indicators in this comparative study and therefore should be further considered as a PAT for the quantitative determination of ALZ in low-content powder blends.