Quantifying milk yield-dependent aflatoxin B1-to-M1 transfer in dairy cows: a Bayesian consensus toxicokinetic model

Aflatoxin B1 (AFB1) is a hepatocarcinogenic mycotoxin found in animal feed contaminated by mold from the genus Aspergillus. It can be transferred into food of animal origin, such as the milk of dairy cows, as the metabolite aflatoxin M1 (AFM1), which has 10% of the potency of AFB1. A multitude of studies has been conducted over the last decades to quantify the transfer of AFB1 as AFM1 in milk. The transfer rates derived from these studies vary considerably, with one variability source being the different milk yields of the experimental animals. A simple deterministic model was developed in the past to predict the influence of milk yield on the transfer rate in a steady-state setting. Although that previous model could predict transfer, it lacked dynamic predictions and did not quantify uncertainty, as is necessary for risk analysis. In this work, we synthesized transfer data from 19 published controlled dairy cow feeding studies (AFB1 doses: 6.6-2800 µg/d up to a 20,000 µg single dose ; milk yields: 6.4-46.4 L/d) using a hierarchical Bayesian approach with a reparametrized two-compartment toxicokinetic model designed to overcome parameter correlations, capture between-study heterogeneity and quantify prediction uncertainty. The model communicates prediction uncertainty using milk yield-dependent credible intervals (CI) for concentration profiles in milk across studies. It predicts a median transfer rate of 0.41% (95% CI: 0.13-1.40%) at 5 L/d, increasing to a median of 3.05% (95% CI: 1.07-8.19%) at 50 L/d. The model can be used in risk assessment and risk management to quantitatively predict transfer including credible intervals.