Prediction of drug distribution into human milk from physicochemical characteristics.

Decisions about the safety of breast feeding during maternal ingestion of drugs require knowledge of the amount of drug which might be present in the milk. For many drugs this has not been studied, and mothers are usually advised against breast feeding. In many cases this is undoubtedly unnecessary, as the total dose to which the baby is exposed is often negligible. It would be very helpful, therefore, to be able to predict the approximate amount of drug which might be present in milk. Existing theory of pH partitioning enables estimation of the distribution of unbound drug, i.e. milk: plasma unbound ratios. However, these ratios are poor estimates of the concentration ratios for whole milk, because whole milk contains proteins and lipid in which drugs will distribute in amounts which depend on their particular physicochemical properties. To predict the milk: plasma concentration ratios for whole milk the amount of drug present in the protein and lipid phases must be considered along with the unbound drug distribution. A 'phase distribution model' has therefore been developed which permits estimation of whole milk: plasma concentration ratios. The model requires a knowledge of the unbound drug concentration ratio, the plasma and milk unbound fractions and the milk lipid: ultrafiltrate partition coefficient. Evaluation of the model by comparison of predicted whole milk ratio values with literature milk: plasma area under the curve (AUC) ratios indicated a trend to overprediction for acidic and neutral drugs and underprediction for basic drugs. Transformation of the phase distribution equation by taking logarithms results in a relationship which can be analysed by multiple linear regression to derive predictive equations for acidic and basic drugs which take into account the relative contributions of each component of the model. Regression of the logarithms of the literature milk: plasma AUC values against the independent variables resulted in good correlations for acidic and basic drugs. The independent variables explained 93.1% and 82.9% of the variance in the values for acidic and basic drugs, respectively, with random scatter of residuals. The equations, together with those to predict unbound fractions of drug in milk and milk lipid: ultrafiltrate partition coefficients, enable the ratio of the milk: plasma AUCs to be estimated for any acidic or basic drug for which the distribution into human milk is not known, using the pKa, octanol: water partition coefficient and plasma protein binding values of the drug. The data set for neutral drugs (n = 3) was too small to develop a correlation equation.(ABSTRACT TRUNCATED AT 400 WORDS)