A robust calibration model was created using Near Infrared Spectroscopy (NIR) and by minimizing the correlation between the active pharmaceutical ingredient (API) and components in a pharmaceutical process. An experimental model was used to minimize correlation and create a calibration model to detect API in a mixture. A design of experiment (DoE) was prepared using six components. A total of nine placebos were prepared using different concentrations of each component.

A series of mixtures were prepared using placebo, granulation and API or using placebo and API. The correlation between components when using granulation, placebo and API was of 99% because concentrations followed a pattern: if API concentration increased, components concentrations decreased. To break that pattern and to decrease the high correlation, only placebo and API were used and the concentrations of the components were modified.

With the new approach, if the concentration of API increases, not all of the concentrations of the components will decrease. A more robust model was then created because of the change in correlation.

The predictions obtained from the calibration model show a very robust model which was achieved by reducing the high correlation presented originally.