A comparative approach of MIR, NIR and Raman based chemometric strategies for quantification of Form I of Meloxicam in commercial bulk drug

Abstract

Meloxicam (MLX), a widespread non-steroidal anti-inflammatory drug, exhibits five reported solid forms. However, only crystalline structures of Form I and Form IV, a monohydrate (MH), have been fully elucidated. Form I is the polymorph of choice for pharmaceutical development, nonetheless, the concomitant crystallization of Form I and MH was found in commercial bulk drugs, and the dehydration of MH may lead to the Form V. Such a thermodynamic relationship between Form I/Form V/MH may cause products with inadequate polymorphic purity compromising their final product bioavailability. Therefore, the objective of this work was to evaluate the feasibility of middle- and near-infrared spectroscopies (MIR and NIR, respectively) as well as Raman spectroscopy, coupled to multivariate calibration to quantify MLX-Form I in commercial raw material of this active pharmaceutical ingredient (API). For this purpose, calibration (n = 30) and validation (n = 9) sets of ternary composition (Form I/Form V/MH) were prepared and measured. Each partial least squares (PLS) model was first optimized by interval-PLS; and then, multiplicative scattering correction (MSC), standard normal variate (SNV), detrend, first derivative (D’) and their combinations were evaluated for spectra pre-treatment and exploring the figures of merit of PLS models using a smart approach based on radial diagrams. Three latent variables (LV) and the full spectral ranges were found as optimal conditions for MIR (4000–600 cm−1, MSC) and NIR (12500–4000 cm−1, SNV) while the Raman-PLS method required a reduced spectral range (1650–1030 cm−1, D’) and four LV. The performance of all three optimal PLS models was compared during the validation step using the radial diagrams and the best results were achieved by NIR-PLS. The suitability of the methods was finally determined through the analysis of commercial batches of the API from different suppliers (raw and spiked samples). NIR-PLS model presented the best predictive capacity (recovery 98.4 %, RSD 0.41–2.6 %), as expected. On the other hand, MIR- and Raman-PLS recoveries (99.7 and 100.7 %, respectively) presented higher dispersions (RSD ranges 0.70–4.4 and 0.72–9.3 %, respectively). NIR-PLS emerged as the best technique to evaluate the polymorphic purity in MLX bulk drugs and the radial diagram optimization approach shows to be a powerful tool to increase the performance of PLS models.