Fig. 2
Influence of model design on reverse toxicokinetic predictions for EDCs with estrogenic activity (a) The predicted environmental concentration of EEDCs from the 7C (black dots), 6C (red dots) and 1C (blue dots) models vs. the experimentally determined LOEC. For the 1C model, only 21 of the 36 predicted concentrations are less than the experimentally-determined LOEC (p = 0.405, binomial test). The 6C and 7C model predictions are often lower than the LOECs (30/36, p = 6.96 × 10−5 and 32/36, p = 1.942 × 10−6, respectively). b The predicted environmental concentration of EEDCs from the 7C (black dots) and 6C (red dots) models vs. the experimentally determined LOEC. The published AC10 concentrations were, in this case, assigned to be the steady state chemical concentrations in the gonads and the environmental concentrations predicted solely from the gonad compartments. 35 of 36 predictions were less than the experimental LOEC in the 7C model (p = 1.077 × 10−9, binomial test) while all 36 predictions were higher than the experimental LOEC in the 6C model (p = 2.911 × 10−11, binomial test). c The kernel density estimates of the ratio of the predicted environmental concentrations from the rTK models to the associated LOEC for the EEDCs, using the published AC10 values as whole body concentrations. The 6C (red) and 7C rTK models (black) produce predicted environmental concentrations whose ratios to the respective LOECs are significantly smaller than the predictions from the 1C rTK models (blue) (p = 4.1 × 10−5 and 5.785 × 10−9, respectively, Student’s t-test). The kernel density estimates were obtained in R [24]. d The kernel density estimates of the ratio of the predicted environmental concentrations from the rTK models to the associated LOEC for the EEDCs, using the published AC10 values as the chemical concentrations in gonads. The 6C (red) and 7C (black) rTK models produce predicted environmental concentrations whose ratios to the respective LOECs are significantly different (p < 2.2 × 10−16, Student’s t-test)