Multivariate data analysis of variables influencing the dissolution rate of prednimustine : a case of disconformity with the Noyes-Whitney equation

Prednimustine is a drug consisting of crystallites - small plates of approximately 1-20 μm - aggregated and agglomerated into larger composite structures; i.e. it has a wide, multi-modal size distribution. Prednimustine is practically insoluble in water. The relationship between particle size and dissolution rate was investigated. Seven small-scale batches and four production campaigns, giving a total of 35 batches, were examined. 66 variables - including, inter alia, three dissolution-rate variables, 18 particle-size parameters (laser-diffraction technique), 12 gas-adsorption (BET) parameters, impurity and process variables - were included in the data set. Principal component analysis (PCA) indicated that the dissolution rate was influenced by particle-size variables as well as by impurity, process and BET variables. Consequently, there was no general relationship between dissolution rate and particle size alone. Partial least squares projections to latent structures modelling (PLS) indicated the existence of relationships between dissolution rate and particle-size, impurity, process and BET variables. The dissolution rate was reasonably well predicted, the correlation coefficient being ≥0.90. Due to the type of investigation, the data set comprised a lot of extra tests apart from quality control data and recorded process information, which means that 66 variables are not routinely obtained during production. This would not be feasible. PLS modelling did not indicate any general relationships between the impurity profile - impurities, solvent residue and loss on drying (LOD) - and process data, but two of the impurities and LOD were reasonably well predicted. PLS modelling indicated the existence of relationships between discrete process variables and other variables