Exploration of time-course combinations of outcome scales for use in a global test of stroke recovery

Goldie, F. C., Fulton, R. L., Dawson, J. , Bluhmki, E. and Lees, K. R. (2014) Exploration of time-course combinations of outcome scales for use in a global test of stroke recovery. International Journal of Stroke, 9(6), pp. 755-758. (doi: 10.1111/ijs.12131) (PMID:24148224)

Full text not currently available from Enlighten.


Background: Clinical trials for acute ischemic stroke treatment require large numbers of participants and are expensive to conduct. Methods that enhance statistical power are therefore desirable.<p></p> Aims: We explored whether this can be achieved by a measure incorporating both early and late measures of outcome (e.g. seven-day NIH Stroke Scale combined with 90-day modified Rankin scale).<p></p> Methods: We analyzed sensitivity to treatment effect, using proportional odds logistic regression for ordinal scales and generalized estimating equation method for global outcomes, with all analyses adjusted for baseline severity and age. We ran simulations to assess relations between sample size and power for ordinal scales and corresponding global outcomes. We used R version 2·12·1 (R Development Core Team. R Foundation for Statistical Computing, Vienna, Austria) for simulations and SAS 9·2 (SAS Institute Inc., Cary, NC, USA) for all other analyses.<p></p> Results: Each scale considered for combination was sensitive to treatment effect in isolation. The mRS90 and NIHSS90 had adjusted odds ratio of 1·56 and 1·62, respectively. Adjusted odds ratio for global outcomes of the combination of mRS90 with NIHSS7 and NIHSS90 with NIHSS7 were 1·69 and 1·73, respectively. The smallest sample sizes required to generate statistical power ≥80% for mRS90, NIHSS7, and global outcomes of mRS90 and NIHSS7 combined and NIHSS90 and NIHSS7 combined were 500, 490, 400, and 380, respectively.<p></p> Discussion: When data concerning both early and late outcomes are combined into a global measure, there is increased sensitivity to treatment effect compared with solitary ordinal scales. This delivers a 20% reduction in required sample size at 80% power. Combining early with late outcomes merits further consideration.

Item Type:Articles
Glasgow Author(s) Enlighten ID:Macisaac, Dr Rachael and Lees, Professor Kennedy and Goldie, Mr Fraser and Dawson, Professor Jesse
Authors: Goldie, F. C., Fulton, R. L., Dawson, J., Bluhmki, E., and Lees, K. R.
College/School:College of Medical Veterinary and Life Sciences > School of Cardiovascular & Metabolic Health
Journal Name:International Journal of Stroke
ISSN (Online):1747-4949
Published Online:22 October 2013

University Staff: Request a correction | Enlighten Editors: Update this record