Ecology-based approach to predict no-effect antibiotic concentrations for minimizing environmental selection of resistance.

Selection for antibiotic resistance has been demonstrated at low, environmentally relevant antibiotic concentrations. The concept of minimum selective concentrations (MSCs) has been adopted in environmental regulation to define maximum permissible antibiotic concentrations. Such empirically determined MSC values often fail to reflect the complexity of natural communities, where susceptibility and resistance-associated fitness costs vary widely across species. To address this limitation, computational approaches have been developed to predict no-effect concentrations for selection of antibiotic resistance (PNECres) from routinely collected minimum inhibitory concentration (MIC) data. However, these approaches, using assessment factors to convert MICs to PNECres, often lack a strong ecological basis, undermining confidence in their predictions. Here, we propose a simple but biologically consistent framework to derive PNECres values by integrating MIC data with probabilistic estimates of resistance-related fitness costs. We demonstrate mathematically and empirically that for typical high-level resistances, the MSC/MIC ratio is approximately equal to the resistance cost, allowing for cost-based estimation of MSCs. In experimental validation across 26 strain-antibiotic combinations, 66% of computed MSCs deviated by less than factor two from empirical values. Leveraging these findings, we explored the general distribution of fitness costs of resistance determinants to establish a cost-based probabilistic model for replacing conventional fixed assessment factors. When applied to current MIC databases, our framework suggests that regulatory environmental threshold concentrations should be lowered by at least one order of magnitude to guard against selection for antibiotic resistance. Our approach offers a feasible and biologically transparent alternative for deriving PNECres values in environmental risk assessment.