High Throughput Prediction of the Long-Term Stability of Pharmaceutical Macromolecules from Short-Term Multi-Instrument Spectroscopic Data

Abstract

Changes in the measurements of a macromolecular biopharmaceutical's physical form are often used to predict changes in the drug's long-term stability. These can in turn be used as important markers of changes to a drug's efficacy and safety. Such stability estimates traditionally require human judgment and are frequently tentative. We introduce methods for developing mathematical models that predict a drug's long-term storage stability profile from measurements of short-term physical form and behavior. We measured the long-term (2 year) chemical and colloidal stability of Granulocyte Colony Stimulating Factor (GCSF) in 16 different liquid formulations. Shortly after formulations were placed on stability, we also employed various spectroscopic techniques to characterize the short-term thermal unfolding response of GCSF in the 16 formulations. The short-term data were processed using several data reduction methods, including reduction to spectra at low temperature, to melt curves, and to transition temperatures. Least squares fitting was used to predict the long-term stability measurements from the reduced short-term spectroscopic measurements. On the basis of the cross-validation and a permutation test, many of the long-term stability predictions have less than 1% probability of occurring by chance.