Multivariate Calibration: A Science-Based Method

A new method for multivariate calibration [1 – 3] eliminates many of the problems associated with traditional methods, making the process easier and more transparent to the user. It uses physical methods to estimate the spectral signal and statistical methods to estimate noise. It is also less expensive to apply than partial least squares (PLS) or principal component regression (PCR) calibration methods and allows specificity of response to be proven directly from first principles, without a reference method. Early adopters in the process analytical technologies (PAT) field refer to this as a “science-based” method. This article will discuss this new approach and how it can be applied effectively to pharmaceutical manufacturing and quality assurance.