The pragmatic approach to stroke trial design: stroke register, pilot trial, assessment of neurological then functional outcome.

Assessment of the results of medical intervention in acute stroke presents particular difficulties. Firstly the condition is extremely variable both in its clinical presentation and in its underlying pathology. Secondly, extreme care must be taken over the methods used to assess outcome. A reduction in mortality is no indication of success if there is a consequent increase in the number of survivors with extreme disability. Nor is it valid to compare functional or neurological outcome in survivors, since differences in mortality between treatment groups may upset the balancing effect of initial randomisation. Furthermore, overall 'neurological scores' are of dubious validity when the progress of patients with different patterns of disability has to be compared. Despite these difficulties a pragmatic approach can be used in which the design and conduct of a stroke trial closely resemble the normal clinical situation. This requires a knowledge of the range and natural history of stroke cases seen, best obtained from a stroke register. A pilot trial is then needed to assess the extent of variation in outcome and thus to estimate the numbers needed for the main study. The latter should then be conducted in two stages. Initially 'neurological benefit' should be assessed by counting the number of individual neurological signs which show improvement or deterioration in patients in each treatment group. Finally, a simple functional end-point should be compared in large numbers of patients with death being treated as equivalent to the worst possible non-fatal outcome. These principles are now being applied in a large, double-blind, placebo-controlled trial of low-dose beta-blockers in conscious stroke victims.

RCT Entities:   Population Interventions Outcomes