Abstract WP337: Publication Bias and Low Power Contribute to the Efficacy Decline of Neuroprotective Drugs From Experimental Stroke Studies to Phase Iii Clinical Trials

Background: In spite of several successful experimental studies, neuroprotective drugs have always failed in large clinical trials. We aimed to determine the efficacy decline of neuroprotectants from animal studies to early phase and phase III clinical trials. Moreover, we aimed to identify study characteristics of experimental studies and clinical trials, which contribute to the efficacy decline of neuroprotective drugs. Methods: We identified all phase III clinical trials and corresponding experimental studies and early clinical trials of neuroprotective drugs for ischemic stroke. The overall efficacy of neuroprotective drugs in experimental studies, early phase and phase III clinical trials was determined by means of a meta-analysis. We performed power analyses and estimated the publication bias in experimental studies and clinical trials. Stratified meta-analyses will be performed to identify study characteristics (e.g. randomization, blinded outcome assessment, time to treatment) which influence the results of experimental studies and clinical trials. Results: We identified 47 phase III trials, 190 corresponding experimental studies and 66 corresponding early clinical trials. Seventy-seven percent of experimental studies and 39,4% of early clinical trials reported a positive result in at least one treatment arm. By contrast, there was only one positive phase III trial, which was followed by a larger negative trial. Funnel plot asymmetry revealed a major publication bias in experimental studies. The mean power of experimental studies was 47,3%. Conclusion: Our study illustrates the efficacy decline from experimental studies to phase III clinical trials. Publication bias and low power inflate the efficacy of neuroprotective drugs in experimental stroke studies.