Mixed modeling of the effect of pharmaceutical pollution on the metamorphosis of amphibians from incomplete trajectory data

Abstract

Abstract In this paper, we aim to evaluate the potential of mixed models in assessing the impact of ibuprofen, fluoxetine, and a mix of them, on the Gosner stage development of Bufo bufo tadpoles, using incomplete trajectory data. We consider the mixed effects models as a good choice to deal with missing data, the within-subject correlations over time, and the feasibility to model time as a continuous covariate. We fitted several models with different random and fixed effects structures and compared them using the Likelihood Ratio (LR) test and Akaike Information Criteria (AIC). Influential analysis and multiple imputation are also performed to check the stability of the results. Mixed-effects models successfully modeled the Gosner stage incomplete longitudinal trajectories. Based on all fitted models, a significant difference in the evolution of the Gosner stage is observed between the group treated with a mix of pharmaceuticals and the other three groups (i.e., ibuprofen, fluoxetine, control).