Statistical analysis of differences in the Raman spectra of polymorphs.

Raman spectroscopy is a useful tool for identifying polymorphs of pharmaceutical compounds. One limitation of the technique is that the small differences in Raman spectra require confirmation of polymorphs by other methods. Fourteen compounds, both commercial and proprietary pharmaceutical compounds and their polymorphs, were analyzed by Raman microscopy. By using descriptive statistics and analysis of variance (ANOVA), several methods are proposed that provide an approach for comparing the spectra of suspected polymorphs. Because it is difficult to determine the exact amount that a peak may shift before two forms should be considered different; a guideline of a shift greater than 1.6/cm(-1) is proposed. A standard ANOVA analysis is used to compare individual peaks both within and between polymorphs, as well as an alternative method that proposes the use of a total ANOVA table that considers the entire spectrum. Both methods have their advantages and disadvantages, but they provide a starting point for the comparison of a large number of spectra, and effectively differentiate between polymorphs of a given compound. The method accurately identified true polymorphs in all cases, but showed a bias towards misidentifying some samples as polymorphs when they were in fact the same form. This bias was not significant and even in these situations, the magnitude of the calculated F values was a useful indicator of whether the result was a false positive or not. (c) 2005 Wiley-Liss, Inc.