Development and validation of a chemometric method for direct determination of hydrochlorothiazide in pharmaceutical samples by diffuse reflectance near infrared spectroscopy Marcus H. Ferreira a, b , Jez W.B. Braga c, e , Marcelo M. Sena a, d, e, ⁎ a Mestrado em Ciências Moleculares, UnUCET, Universidade Estadual de Goiás, P.O. Box 459, 75001–970, Anápolis, GO, Brazil b Indústria Química do Estado de Goiás SA, IQUEGO, 74450–0101, Goiânia, GO, Brazil c Instituto de Química, Universidade de Brasília, 70904–970, Brasília, DF, Brazil d Departamento de Química, ICEx, Universidade Federal de Minas Gerais, 31270–901, Belo Horizonte, MG, Brazil e Instituto Nacional de Ciência e Tecnologia em Bioanalítica, 13083–970, Campinas, SP, Brazil abstract article info Article history: Received 30 November 2011 Received in revised form 26 February 2012 Accepted 16 March 2012 Available online 24 March 2012 Keywords: Multivariate calibration Analytical validation Net analyte signal NIR Quality control Active pharmaceutical ingredient This work developed and validated a new multivariate diffuse reflectance near infrared method for direct determi- nation of hydrochlorothiazide in powder pharmaceutical samples. The best partial least squares (PLS) model was obtained in the spectral region from 1640 to 1780 nm, with mean centered data preprocessed by first derivative and Savitzky–Golay smoothing followed by vector normalization. This model was built with 4 latent variables and provided a root mean square error of prediction of 1.7%. The method was validated according to the appropriate reg- ulations in the range from 21.25 to 29.00 mg of hydrochlorothiazide per 150 mg of powder (average mass tablet), by the estimate of figures of merit, such as accuracy, precision, linearity, analytical sensitivity, capability of detection, bias and residual prediction deviation (RPD). The concept of net analyte signal (NAS) was used to estimate some figures of merit and to plot a pseudo-univariate calibration curve. The results for determinations in pow- dered manufactured tablets were in agreement with those of the official high performance liquid chromato- graphic method (HPLC). Finally, the method was extrapolated for determinations in intact tablets, providing prediction errors smaller than ±9%. The developed method presented the advantage of being about fifteen times faster than the reference HPLC method. © 2012 Elsevier B.V. All rights reserved. 1. Introduction The combination of near infrared spectroscopy (NIRS) and multivar- iate calibration has emerged in the last decade as a promising alterna- tive for the quality control of active pharmaceutical ingredients (API) [1–8], providing methods that are simple, rapid, non-destructive and of low cost. In addition, methods based on NIRS are environmentally friendly and solvent free, generate no chemical waste, may not require any sample pre-treatment, and provide sufficient accuracy and sensitiv- ity with less human intervention. The development of these methods requires a robust calibration design that incorporates all the possible sources of variation, thus improving the quality control of the final product. However, in practice few quantitative NIRS methods are used for the determination of APIs in the quality control laboratories of the pharmaceutical industry. The main challenge is developing of NIRS methods that meet the stringent requirements of this highly regulated industry [3]. In the last years, Brazilian and US Pharmacopoeias [9,10] have published general monographs about NIRS methods, but they have no monographs about multivariate methods for the quantification of specific APIs in pharmaceutical formulations. The most of the present regulation has been established based on chromatographic methods in a univariate way, such as the guidelines of ANVISA (National Health Surveillance Agency) [11], in Brazil, and ICH (International Conference on Harmonisation) [12,13]. Thus, these regulations should be harmo- nised in order to encompass the specific aspects of NIRS technology re- lated to multivariate methods. The main aspect of this harmonisation is the incorporation of the multivariate thinking in the traditional analytical validation. The re- quirement of total selectivity/specificity [11,12] should be eliminated for multivariate methods, since they are only useful when a selective variable/wavelength does not exist. The traditional calibration curves (signal as a function of analyte concentration) should also not be employed with multivariate methods. Some of these aspects have been discussed since about ten years ago [14], but the main concept that emerged in the last years is the net analyte signal (NAS). The NAS concept is an advance in the multivariate calibration theory that allows separating the information specific of the analyte from the whole signal, and can be used for estimating important figures of merit (FOM) in pharmaceutical applications. In addition, NAS Microchemical Journal 109 (2013) 158–164 ⁎ Corresponding author at: Departamento de Química, ICEx, Universidade Federal de Minas Gerais, 31270–901, Belo Horizonte, MG, Brazil. Tel.: + 55 31 34096389; fax: +55 31 34095700. E-mail address: marcsen@ufmg.br (M.M. Sena). 0026-265X/$ – see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.microc.2012.03.008 Contents lists available at SciVerse ScienceDirect Microchemical Journal journal homepage: www.elsevier.com/locate/microc