3B). It should be noted that all Tg-values except one (bezafibrate) used in stability modelling have been experimentally determined. Since no test set was available for validation, the stability model developed was evaluated using the calculated fraction MG-132 manufacturer of the amorphous phase transformed during storage (α).

A plot of α as a function of the prediction values generated by the model is displayed in Fig. 4. This shows the model is not only able to separate the two classes stable and non-stable with 78% certainty, but also able to assign the lowest values (<−0.5) for all the compounds that was fully crystallized upon storage, and highest values (>0) for all the compound that did not crystallize during storage (the only exception being griseofulvin having high prediction value but low stability). There

seem to be a sigmoidal relation between the predicted values and α which further support the validity of the model. The rational for why a model based on the parameters Tg and Mw is able to predict glass stability can be deducted in a similar way as for glass-forming ability, i.e. it is the balance between the molecular mobility (the rate of molecular motion) and the configurational space (how many configurations that can be probed) that governs crystallization tendency of a compound. It has been shown that molecular mobility determines the rate of crystallization of an amorphous phase when analysing the temperature dependency of single amorphous compounds ( Aso et al., 2001, Bhattacharya and Suryanarayanan,

2009 and Bhugra et al., 2008). However, when it comes to comparing crystallization XL184 solubility dmso tendencies for a number of structurally diverse compounds other factors has to be considered to predict physical stability ( Van Eerdenbrugh et al., 2010) and one factor identified to be important is the configurational entropy ( Graeser et al., 2009 and Zhou et al., 2002). Based on this we hypothesize that Tg and Mw is describing molecular mobility and configurational entropy well enough to, when combined, be able to predict glass stability. It is interesting to note that the compound being poorest predicted by the Mw–Tg-storage model, griseofulvin, has been extensively studied as to find out the reason for its sensitivity to production conditions, since its stability is Oxymatrine inhibitors higher when amorphisized by melt-cooling as compared to milling (34–36). A glass heated above its Tg may crystallize before it reach the thermodynamic melting temperature. The onset of this crystallization is dependent on the nucleation tendency and crystal growth rate of the heated amorphous system ( Bhugra and Pikal, 2008 and Hancock and Zograf, 1997). At a well-defined heating rate and sample size, the onset temperature of crystallization (Tcr) can be regarded as an indicator of the crystallization tendency of the amorphous compound.