Assessment of clinical outcomes in acute stroke trials.

BACKGROUND AND PURPOSE Adequate outcome assessment is crucial to randomized trials. We wished to assess the types of outcomes used in acute stroke trials and the appropriateness of these outcomes and their analyses. METHODS Acute stroke trials from the Cochrane Stroke Group's database were included from 1955 to 1995 if they were published in full text in English. For each trial we collected year of publication, number of patients randomized, blinding of outcome assessment, the specific outcome instruments used, the statistical methods used for analysis, and the significance of the results. The validity and reliability of each outcome measure were assessed by review of the literature. RESULTS Our study included 174 trials. Outcomes were assessed blindly in 69%. Death was recorded in only 76% of trials, impairment in 76%, disability in 42%, and handicap or quality of life in only 2%. Of the trials that measured impairment, 35% used a measure of established validity or reliability. For disability and handicap, the proportions with valid or reliable measures were 70% and 25%, respectively. Impairment and handicap measures were primarily analyzed as continuous variables, while disability was mainly analyzed as a dichotomous variable. Continuous data were usually analyzed with inappropriate parametric statistics. There was no relationship between the method of analysis, the type of outcome, and the statistical significance of results. CONCLUSIONS Most acute stroke trials up to 1995 have used clinical outcome measures that were inadequate in terms of their content, reliability, validity, blinded assessment, and statistical analysis. This has important implications for future stroke research.