No useful correlations were obtained with AUC0±1 (r
ranging from approximately 0.5±0.8). For the other pharmacokinetic
characteristics the D value or 1/D value, giving the
highest correlation coef®cient and lowest unexplained
variance, was established for each IVIVC. An example of
establishing the optimum D value for Cmax is shown in Table
3 and Fig. 3. In Fig. 4a&a±c the relationships between D20,
D40 and D50 in SLS and Cmax are presented graphically
(program, Microsoft Excel 5.0). In Table 4, the best correla-