Antibiotic resistance presents a serious and still growing threat to human health. Environmental exposure levels required to select for resistance are unknown for most antibiotics. Here, we evaluated different experimental approaches and ways to interpret effect measures, in order to identify what concentration of trimethoprim that are likely to select for resistance in aquatic environments. When grown in complex biofilms, selection for resistant E. coli increased at 100 µg/L, whereas there was only a non-significant trend with regards to changes in taxonomic composition within the tested range (0-100 µg/L). Planktonic co-culturing of 149 different E. coli strains isolated from sewage again confirmed selection at 100 µg/L. Finally, pairwise competition experiments were performed with engineered E. coli strains carrying different trimethoprim resistance genes (dfr) and their sensitive counterparts. While strains with introduced resistance genes grew slower than the sensitive ones at 0 and 10 µg/L, a significant reduction in cost was found already at 10 µg/L. Defining lowest effect concentrations by comparing proportion of resistant strains to sensitive ones at the same time point, rather than to their initial ratios, will reflect the advantage a resistance factor can bring, while ignoring exposure-independent fitness costs. As costs are likely to be highly dependent on the specific environmental and genetic contexts, the former approach might be more suitable as a basis for defining exposure limits with the intention to prevent selection for resistance. Based on the present and other studies, we propose that 1 µg/L would be a reasonably protective exposure limit for trimethoprim in aquatic environments.