Purpose: The aim of this investigation was to demonstrate that nonlinear mixed-effects population pharmacokinetic (PK) modeling can be used to evaluate data from studies of drug transport/excretion into human milk and hence to estimate infant exposure.

Methods: A sparse dataset from a previously published study of the use of oral tramadol for post-cesarean pain management in 75 lactating women was used. Milk and plasma samples were collected during days 2-4 of lactation, and tramadol and O-desmethyltramadol (ODT) concentration measurements in these samples were available. Absolute infant dose was obtained from the concentration measurements and estimated milk volume ingested, and expressed in micrograms per kilogram per day. Relative infant dose was calculated as a percentage of the absolute infant dose divided by the maternal dose (μg/kg/day). Nonlinear mixed-effects modeling was used to fit a population PK model to the data.

Results: The disposition of tramadol and ODT in plasma and the transition of these substances into milk were characterized by a five-compartment population PK mixture model with first-order absorption. The polymorphic ODT formation clearance in the plasma compartment was able to be characterized in both CYP2D6-poor and -extensive metabolizers. Milk creamatocrit was a significant covariate in ODT transfer between the plasma and milk compartments. The estimated relative infant doses in extensive and poor metabolizers, respectively, were 2.16 ± 0.57 and 2.60 ± 0.57% for tramadol, and 0.93 ± .20 and 0.47 ± 0.10% for ODT.

Conclusions: This study demonstrates that a population PK approach with sparse sampling of analytes in milk and plasma can yield quality information about the transfer process and that it also can be used to estimate the extent of infant exposure to maternal drugs via milk.