Boosting the chances to improve stroke treatment

Abstract

Background and purpose: There is a lack of agreement regarding measuring the effects of stroke treatment in clinical trials, which often relies on the dichotomized value of 1 outcome scale. Alternative analyses consist mainly of 2 strategies: use all the information from an ordinal scale and combine information from several outcome scales in a single estimate.

Methods: We reanalyzed 3 outcome scales that assessed patient recovery (modified Rankin Scale, National Institutes of Health Stroke Scale, and Barthel Index). With data collected from the 1652 patients in the Citicoline pooling data analysis, we used 2 standard techniques of exploratory multivariate analysis to analyze the distances among ranks and to isolate the common and the unique information provided by each of the 3 scales.

Results: The different scale values correspond to gradually different patient status, confirming that information is lost when a scale is collapsed to just 2 values, whether recovered or not. The scales shared 90.7% (95% CI, 84.5-96.9) of their information, with no individual scale contributing unique information.

Conclusions: Salient stroke outcome information is lost when an ordinal scale is collapsed into fewer categories. In contrast, the full scales provide a comprehensive patient outcome estimate. Furthermore, in the context of stroke clinical trials, those scales are highly correlated, providing the rationale to pool them into a single estimate. These insights may be used to optimize the analysis of stroke trials to increase study power to detect efficacious interventions.