Neuroprotection: Establishing Proof of Concept in Human Stroke
The failure to translate the positive effects of a variety of neuroprotective strategies from animal models to human trials has perplexed investigators. This contrasts with the success of tissue plasminogen activator, where efficacy in animal models was translated into positive trial results. Postulated reasons for these expensive trial failures include defective trial design (particularly time window and trial size) and the heterogeneity and complexity of human stroke compared with animal infarct models. Further explanations might be the lack of entry of neuroprotective compounds into the ischemic brain (suggesting that these agents might require thrombolysis to facilitate access), their lack of benefit in white matter, and the fairly modest effect of neuroprotection in penumbral salvage, even if reperfusion has occurred.
We agree with Grotta that the animal experiments should be replicated as closely as possible in trials, including standardization of stroke severity and appropriately short times to treatment. Importantly, he also mentions that compounds with multiple sites of action on the neurotoxic cascade may be more effective. Lees has pointed out that improved trial design, tightened entry criteria, and more sophisticated endpoint selection have already occurred and should increase the chances of success.
Despite the plethora of negative neuroprotective trials, an array of compounds has been shown to dramatically reduce infarct volume in animal models. It seems quite implausible to us that these experiments should not be translated to human stroke, given a potent agent, adequate drug levels in ischemic tissue, a short time window, lack of toxicity, and well-designed stroke trials. We suggest a stepwise approach, which includes demonstration of a highly significant infarct volume reduction in small animal models (at least 80% as suggested by Grotta), together with functional improvement. A similar benefit should then be shown in larger animal models, perhaps incorporating neuroimaging endpoints as a further intermediate step. Proof of concept studies in humans, using perfusion/diffusion-weighted MRI as a surrogate endpoint, should more reliably indicate the chance of efficacy in phase III trials.
The views of both our debaters clearly indicate that we need to rethink our approach to translational research. We conclude that however sophisticated our current phase III trial designs, the testing of the hypothesis that neuroprotection is an effective stroke treatment requires a far more rigorous approach than was originally perceived.
The opinions expressed in this editorial are not necessarily those of the editors or of the American Stroke Association.