Reduced animal use in efficacy testing in disease models with use of sequential experimental designs.

Although the use of animals in efficacy tests has declined substantially, there remains a small number of well-documented disease models which provide essential information about the efficacy of new compounds. Such models are typically used after extensive in vitro testing, to evaluate small numbers of compounds and to select the most promising agents for clinical trial in humans. The aim of this study was to reduce the number of animals required to achieve valid results, without compromising statistical power, via sequential designs, in tests of efficacy of potential neuroprotective agents. In a rat middle cerebral artery occlusion model of acute stroke, infarct volume was measured by magnetic resonance imaging 24 hours post-occlusion. We performed 13 vehicle-controlled studies of neuroprotectives, with the aim of achieving false negative and false-positive rates of 5% for compounds with 40% neuroprotection. The sequential design used "EaSt" software (Cytel), with significance testing at N=6, 9, 12, and 15 rats. Ten compounds were declared active: four at N=6, three at N=9, two at N=12, and one at N=15. Three compounds were declared inactive: one at N=9, one at N=12, and one at N=15. This equates to a reduction of 138 animals, or 35%, over a conventional design with groups of N=15. Sequential studies require more effort to manage, and are not appropriate in all cases, e.g. where drift occurs in control values, or where the model pathology develops over a considerable period of time. In our implementation, the saving in animals can be anything between 0% and 60%, depending on the true efficacy of the compounds tested. We have demonstrated a practical approach that achieves significant savings in animal use.