Application of partial least-squares (PLS) modeling in quantifying drug crystallinity in amorphous solid dispersions

Among the different experimental methods that can be used to quantify the evolution of drug crystallinity in polymer-containing amorphous solid dispersions, powder X-ray diffractometry (PXRD) is commonly considered as a frontline method. In order to achieve accurate quantification of the percent drug crystallinity in the system, calibration curves have to be constructed using appropriate calibration samples and calculation methods. This can be non-trivial in the case of partially crystalline solid dispersions where the calibration samples must capture the multiphase nature of the systems and the mathematical model must be robust enough to accommodate subtle and not so subtle changes in the diffractograms. The purpose of this study was to compare two different calculation and model-building methods to quantify the proportion of crystalline drug in amorphous solid dispersions containing different ratios of drug and amorphous polymer. The first method involves predicting the % drug crystallinity from the ratio of the area underneath the Bragg peaks to total area of the diffractogram. The second method is multivariate analysis using a Partial Least-Squares (PLS) multivariate regression method. It was found that PLS analysis provided far better accuracy and prediction of % drug crystallinity in the sample. Through the application of PLS, root-mean-squared error of estimation (RMSEE) values of 2.2%, 1.9%, and 4.7% drug crystallinity was achieved for samples containing 25%, 50%, and 75% polymer, respectively, compared to values of 11.2%, 17.0%, and 23.6% for the area model. In addition, construction of a PLS model enables further analysis of the data, including identification of outliers and non-linearity in the data, as well as insight into which factors are most important to correlate PXRD diffractograms with % crystallinity of the drug through analysis of the loadings.