Developing an In Vitro to In Vivo Extrapolation (IVIVE) Model to Predict Human Milk-to-Plasma Drug Concentration Ratios.

Determining the amount of drug transferred into human milk is critical for benefit-risk analysis of taking medication while breastfeeding. In this study, we developed an in vitro and in vivo extrapolation (IVIVE) model to predict human milk/plasma ( M / P ) drug concentration ratios. Drug unionized fractions at pH 7.0 ( F ni,7.0 ) and 7.4 ( F ni,7.4 ), drug fractions unbound in human plasma ( f up ) and milk ( f um ), and in vitro cell permeability in both directions (efflux ratio, ER) were incorporated into the IVIVE model. A multiple regression E max model was chosen to predict f um from f up and polar surface area (PSA). A total of 97 drugs with experimental ER from Caco-2 cells were used to test the IVIVE model. The M / P ratios predicted by the IVIVE model had a 1.93-fold geometric mean fold error (GMFE) and 72% of predictions were within two-fold error (Pw2FE), which were superior to the performance of previously reported five models. The IVIVE model showed a reasonable prediction accuracy for passive diffusion drugs (GMFE = 1.71-fold, Pw2FE = 82%, N = 50), BCRP substrates (BCRP: GMFE = 1.91-fold, Pw2FE = 60%, N = 5), and substrates of P-gp and BCRP (GMFE = 1.74-fold, Pw2FE = 75%, N = 8) and a lower prediction performance for P-gp substrates (GMFE = 2.51-fold, Pw2FE = 55%, N = 22). By fitting the observed M / P ratios of 39 P-gp substrates, an optimized ER (1.61) was generated to predict the M / P ratio of P-gp substrates using the developed IVIVE model. Compared with currently available in vitro models, the developed IVIVE model provides a more accurate prediction of the drug M / P ratio, especially for passive diffusion drugs. The model performance is expected to be further improved when more experimental f um and ER data are available.