Use of chemometrics to compare NIR and HPLC for the simultaneous determination of drug levels in fixed-dose combination tablets employed in tuberculosis treatment
Kelly Sivocy Sampaio Teixeira,Said Gonçalves da Cruz Fonseca,Luís Carlos Brigido de Moura,Mario Luís Ribeiro de Moura,Márcia Herminia Pinheiro Borges,Euzébio Guimaraes Barbosa,Túlio Flávio Accioly De Lima E Moura
DOI: https://doi.org/10.1016/j.jpba.2017.11.001
2018-02-05
Abstract:The World Health Organization recommends that TB treatment be administered using combination therapy. The methodologies for quantifying simultaneously associated drugs are highly complex, being costly, extremely time consuming and producing chemical residues harmful to the environment. The need to seek alternative techniques that minimize these drawbacks is widely discussed in the pharmaceutical industry. Therefore, the objective of this study was to develop and validate a multivariate calibration model in association with the near infrared spectroscopy technique (NIR) for the simultaneous determination of rifampicin, isoniazid, pyrazinamide and ethambutol. These models allow the quality control of these medicines to be optimized using simple, fast, low-cost techniques that produce no chemical waste. In the NIR - PLS method, spectra readings were acquired in the 10,000-4000cm-1 range using an infrared spectrophotometer (IRPrestige - 21 - Shimadzu) with a resolution of 4cm-1, 20 sweeps, under controlled temperature and humidity. For construction of the model, the central composite experimental design was employed on the program Statistica 13 (StatSoft Inc.). All spectra were treated by computational tools for multivariate analysis using partial least squares regression (PLS) on the software program Pirouette 3.11 (Infometrix, Inc.). Variable selections were performed by the QSAR modeling program. The models developed by NIR in association with multivariate analysis provided good prediction of the APIs for the external samples and were therefore validated. For the tablets, however, the slightly different quantitative compositions of excipients compared to the mixtures prepared for building the models led to results that were not statistically similar, despite having prediction errors considered acceptable in the literature.