Outcome analysis in clinical trial design for acute stroke: physicians' attitudes and choices

Abstract

Background: Thrombolysis remains the only proven therapy to benefit acute ischemic stroke (AIS) patients. Recent studies have introduced more sensitive outcome measures such as the shift analysis to detect a treatment effect in AIS trials and are also including imaging as a surrogate of injury.

Methods: We conducted a cross-sectional, internet-based survey of academic neurologists regarding their attitudes, choices and understanding of various outcome measures in clinical trial design for AIS. The survey population consisted of neurologists who specialize in the care of stroke patients and are on faculty at university-affiliated hospitals in the USA.

Results: 152 of 300 neurologists completed the survey. There were 79% men and 21% women. Among commonly used outcome scales in acute stroke, the most frequent ones selected for use as trial primary endpoints were the global statistic (59%), modified Rankin scale (mRS) (52%), and NIHSS (30%). When given choices about which outcome on the mRS would justify a therapeutic intervention, 54% chose a shift analysis of change in the distribution of outcomes and 39% chose a dichotomized outcome (mRS <or=2). A majority of respondents favored health transition states of 4-3, 3-2 and 2-1 on the mRS as clinically worthwhile. Only 2% of the respondents thought that a single transition point on the mRS was clinically meaningful. However, 20% of the respondents did not understand the shift analysis. In addition, nearly two thirds of the respondents believed that the presence of a mismatch on brain imaging is relevant to the success of neuroprotective agents.

Conclusion: The majority of respondents accepted an analysis of the entire distribution of the mRS scores as an appropriate endpoint analytic technique in AIS trials and did not require the traditional dichotomized outcome to demonstrate a treatment effect; however, a better understanding of the shift strategy is needed. Our data also support the importance of incorporating mismatch imaging into future neuroprotection trials.