© 2014, Springer Science+Business Media New York. Introduction: In recent years, the simplicity and expeditiousness of near infrared spectroscopy (NIRS) have substantially fostered its use for the pharmaceutical analysis. Developing appropriate NIRS calibration models requires careful selection of calibration sets containing all potential sources of variability in the production samples to be analysed. Once the whole variability is included, the predictive capacity of the calculated models is enhanced and those can be considered useful in the quality control. In this work, we assessed three different calibration strategies for the quantification of the active pharmaceutical ingredient (API) in a pharmaceutical granulate in low concentration (10 mg g−1). Such strategies were used to construct calibration models, allowing all potential variability in new, unknown samples to be considered.Method: The models were constructed by PLS using samples of variable origin including laboratory-made powder mixtures and industrial samples, and variability in production samples was incorporated via a mathematical algorithm.Results: All three strategies provide effective multivariate models that were validated according to the normative on various production batches of industrial granulates manufactured at different times.Conclusions: The quantification of an API in such a low concentration represented a new challenge in the field and the development of the proposed methodologies contributed for further studies in this regard. Although all three are suitable for the strategy involving calculation and addition of the process spectrum has some advantages over the other two including robustness, a good predictive ability (RMSEP = 0.225 mg g−1) and the need for no reference method, so it is the preferred choice.
- Analytical quality control
- Calibration sets
- Near infrared spectroscopy
- Partial least-squares regression
- Solid dosage forms