Median-based robust regression methods in prediction of drug stability

The classical isothermal approach for the prediction of drug stability exploits least squares regression. In this paper the use of some robust regression techniques to estimate the rate constants at different temperatures has been evaluated. These techniques are able to give accurate estimates when data are contaminated by the presence of outliers. The successful application of two robust methods, single median and repeated median, to real stability data from the literature is shown. Moreover, the authors have modified the original methods in order to apply them to data sets with replicates, typical of stability studies. The performances of the modified techniques have been investigated with simulated data sets containing outliers and with real data. They appear suitable for preliminary stability studies, especially on solid dosage forms. For a quick implementation of these methods, macroprograms written for a widely used spreadsheet are reported.