OBJECTIVE: There is a lack of correlation between cyclooxygenase (COX) inhibition and nonsteroidal anti-inflammatory drug (NSAID)-induced gastrointestinal (GI) damage; it has been suggested that mucosal damage may be initiated by a "topical" action of NSAIDs involving mitochondrial injury. We evaluated the effect of a range of NSAIDs and related compounds on mitochondrial function and assessed the differences between them in relation to their physicochemical properties. METHODS: Stimulation of respiration, as an indicator of mitochondrial uncoupling, was measured in isolated coupled rat liver mitochondrial preparations, using an oxygen electrode. RESULTS: Conventional NSAIDs and acidic prodrugs all had stimulatory effects on mitochondrial respiration at micromolar concentrations (0.02-2.7 microM); higher concentrations were inhibitory. The uncoupling potency was inversely correlated with drug pKa (r = -0.87, P < 0.001; n = 12). Drugs known to have good GI tolerability, including modified flurbiprofen (dimero-flurbiprofen and nitrobutyl-flurbiprofen), nabumetone (a non-acidic prodrug), and non-acidic highly selective COX-2 inhibitors, did not cause uncoupling. CONCLUSION: The ability to uncouple mitochondrial oxidative phosphorylation is a common characteristic of antiinflammatory agents with an ionizable group. Modification or absence of an ionizable moiety reduces the effect on mitochondria and could lead to improved NSAID GI safety.