Background: Data regarding antimicrobial pharmacokinetics (PK) in critically ill dogs are lacking and likely differ from those of healthy dogs. The aim of this work is to describe a population PK model for intravenous (IV) amoxicillin–clavulanic acid (AMC) in both healthy and sick dogs and to simulate a range of clinical dosing scenarios to compute PK/PD cutoffs for both populations.Methods: This study used a prospective clinical trial in normal and critically ill dogs. Twelve client-owned dogs hospitalized in the intensive care unit (ICU) received IV AMC 20 mg/kg every 8 h (0.5-h infusion) during at least 48 h. Eight blood samples were collected at predetermined times, including four trough samples before the next administration. Clinical covariates and outcome were recorded, including survival to discharge and bacteriologic clinical failure. Satellite PK data were obtained de novo from a group of 12 healthy research dogs that were dosed with a single AMC 20 mg/kg IV. Non-linear mixed-effects model was used to estimate the PK parameters (and the effect of health upon them) together with variability within and between subjects. Monte Carlo simulations were performed with seven dosage regimens (standard and increased doses). The correlation between model-derived drug exposure and clinical covariates was tested with Spearman's non-parametric correlation analysis. Outcome was recorded including survival to discharge and bacteriologic clinical failure.Results: A total of 218 amoxicillin concentrations in plasma were available for healthy and sick dogs. A tricompartmental model best described the data. Amoxicillin clearance was reduced by 56% in sick dogs (0.147 L/kg/h) compared with healthy dogs (0.336 L/kg/h); intercompartmental clearance was also decreased (p <0.01). None of the clinical data covariates were significantly correlated with individual exposure. Monte Carlo simulations showed that higher PK/PD cutoff values of 8 mg/L could be reached in sick dogs by extending the infusion to 3 h or doubling the dose.Conclusions: The PK of AMC is profoundly different in critically ill dogs compared with normal dogs, with much higher interindividual variability and a lower systemic clearance. Our study allows to generate hypotheses with regard to higher AMC exposure in clinical dogs and provides supporting data to revise current AMC clinical breakpoint for IV administration.