The purpose of the present study was to establish in the horse the relationship between plasma concentration profiles of phenylbutazone (PBZ) and flunixin meglumine (FM) and their pharmacological effects in order to build a predictive pharmacokinetic/pharmacodynamic (PK/PD) model. In five horses, an experimental arthritis was induced by injecting Freund's adjuvant into a carpal joint. PBZ (4 mg/kg) and FM (1 mg/kg) were injected by the intravenous route as a single intravenous dose in two different trials. Five pharmacodynamic end-points were regularly measured after test article injection using standardized procedures: local skin temperature, stride length, the rest angle flexion and the maximal carpal flexion of the injured leg and circumference of the inflamed joint. Plasma drug concentrations and pharmacodynamic data were analysed according to an integrated PK/PD model: for the stride length, the PBZ ECSOr i.e. the plasma concentration for which half the maximum effect could be obtained, was 3.6 ± 2.2 yglml and the maximum potential effect was 10.7 ± 9.4% above the control value. For FM, the corresponding values were 0.93 ± 0.35 μg/ml and 16.3 ± 4.6%. ECSO values for rest angle flexion and local skin temperature were similar to that obtained for stride length. Maximal carpal flexion was an unreliable end-point, and circumference of the joint did not display significant response to the drugs. Using these experimental parameters, a dose-effect relationship was simulated for both drugs: it was shown for PBZ that the model predicts an absence of effect for a 1 mg/kg dose and a maximum effect at about 2 mg/kg: at higher PBZ doses, the maximum effect was not modified, but its duration was increased from 8 h with a 2 mg/kg dose to about 24 h with an 8 mg/kg dose. For FM the model predicts that a dose of 0.5 mg/kg will be without significant effect, whereas a 1 mg/kg dose allows a nearly maximal effect with a return to the control value after a delay of 16 h. A 2 mg/kg dose allows the effect to be maintained for 24 h. It is concluded that PK/PD is a tool of potential value for the preclinical screening of a dosage regimen.