There are many examples of compounds showing neuroprotective efficacy in animal models of stroke but not in clinical trials. It is possible that some or all of these compounds possess poor therapeutic ratios, which results in the administration of sub-efficacious doses in order to avoid the emergence of side-effects. In order to explore this possibility, this study compared the therapeutic ratios of a number of neuroprotective agents that have undergone clinical trials. Neuroprotective efficacy was established using the mouse permanent (24 h) middle cerebral artery occlusion model. Side-effect liability was determined by assessment of motor coordination using the rotarod test. The therapeutic ratio was calculated as the ratio between the minimum effective dose (MED) for significant impairment in rotarod performance and the MED for significant neuroprotection. Compounds were administered i.p. 30 min prior to rotarod testing or onset of ischemia. Drugs such as Ifenprodil, Cerestat and Selfotel, that have failed in clinical trials, were found to have very low therapeutic ratios of < or = 1, whereas compounds with more tolerable clinical side-effect profiles were found to have higher therapeutic ratios (2, 10 and 10 for Sipatrigine, Remacemide and sPBN, respectively). It is concluded that the lack of efficacy of a number of neuroprotectants in clinical trials may well be a consequence of their poor therapeutic ratios.