Evaluation of mathematical models for predicting medicine distribution into breastmilk - considering biological...

A significant proportion of mothers take medication during the breastfeeding period, however knowledge of infant safety during continued breastfeeding is often limited. Breastmilk exhibits significant physiological heterogeneity, with a range of milk fat (creamatocrit), protein and pH values available within the literature. Mathematical models for the prediction of infant exposure are available and these predict that variable milk physiology will significantly affect accumulation of drugs within the breastmilk. These models are typically validated against limited datasets only, and to the best of our knowledge no widescale review has been conducted which accounts for the heterogeneity of breastmilk.

Observed area under the curve milk-to-plasma (M/P) ratios and physicochemical properties were collected for a diverse range of drugs. The reliability of previously published mathematical models was assessed by varying milk pH and creamatocrit across the physiological range. Subsequently, alternative methods for predicting lipid and protein binding within the milk, and the effect of ionisation and physicochemical properties were investigated.

Existing models mis-predicted >40% of medications (Phase Distribution model), exhibited extreme sensitivity to milk pH (Log-Transformed model) or exhibited limited sensitivity to changes in creamatocrit (LogP_o:w model). Alternative methods of predicting distribution into milk lipids moderately improved predictions, however altering the way in which milk protein binding was predicted and the effect of ionisation on this demonstrated little effect. Many drugs were predicted to have a significant range of M/P ratios.

These data show that consideration of the biological heterogeneity of breastmilk is important for model development and highlight that increased understanding of the physiological mechanisms underlying distribution within the milk may be essential to continue improving in silico methodologies to support infant and maternal health.