Letter by Bath Regarding Article, “A Simple, Assumption-Free, and Clinically Interpretable Approach for Analysis of Modified Rankin Outcomes”
To the Editor:
Howard and colleagues1 describe a variant of the Mann-Whitney U test for analyzing the modified Rankin Scale (mRS).1 Unfortunately, the authors do not reference the international Optimizing the Analysis of Stroke Trials (OAST) collaboration that has already assessed Mann-Whitney U test, and other statistical approaches, in the analysis of mRS.2–5 OAST showed that methods that take account of all 7 levels of the mRS, for example, ordinal (including Mann-Whitney U test) and continuous (such as Student t test) are statistically more efficient than those that dichotomize mRS into “good” and “bad” outcomes (eg, mRS 0, 1 versus 2–6 or mRS 0–2 versus 3–6).2 This enhanced efficiency can be expressed as an increase in the proportion of trials that would have been significant (positive or negative, not neutral),2 maintaining statistical power for a smaller sample size or increasing power for the same size.3 Furthermore, it is advantageous to incorporate baseline covariates (including minimization factors if adaptive randomization is used) in the primary analysis4; this allows a further reduction in sample size (approximately 20%) due to an increase in the precision of the estimated treatment effect and deals with minor imbalances of baseline factors.4
In an accompanying editorial, Lyden discusses the issue of how to report trials that use optimized analyses that, unfortunately, generate less-than-intuitive numbers (eg, OR) than absolute risk reduction. However, treatment effects may be described as “number-needed-to-treat” and number-needed-to-treat are lower with alteplase if the whole mRS is examined, that is, number-needed-to-treat 4 versus 8 for ordinal versus dichotomous assessment.5 Hence, number-needed-to-treat could be used to describe treatment effects in trial reports, irrespective of the primary method of analysis, with the additional advantage that they allow the health economics of an intervention to be estimated.
Although these approaches are now being used in academic trials, regulators such as the European Medicines Agency and Food and Drugs Administration should support them as well so commercial trials may also benefit. To this end, the European Stroke Organisation held a workshop at the 2011 International Stroke Conference and has recently published its recommendations in this journal; first, mRS should usually be the primary outcome measure, and second, ordinal, continuous, or sliding dichotomous methods should be used to analyze mRS, ideally with adjustment for baseline covariates.
Finally, as a caveat, ordinal (or continuous) analysis approaches may not be appropriate for interventions that have asymmetrical effects on benefit and hazard,3 as with thrombolysis, which tends to shift all participants in a beneficial direction but may adversely affect a subgroup (eg, those with severe stroke who are more likely to have a catastrophic intracerebral bleed).
Philip M.W. Bath, FRCP, FESO
Stroke Trials Unit
University of Nottingham
P. Bath is Chief Investigator of OAST and chairs the European Stroke Organisation (ESO) Industry Roundtable that is promoting the use of optimized methods for design and analysis of stroke trials. He is The Stroke Association Professor of Stroke Medicine.
Stroke welcomes Letters to the Editor and will publish them, if suitable, as space permits. Letters must reference a Stroke published-ahead-of-print article or an article printed within the past 3 weeks. The maximum length is 750 words including no more than 5 references and 3 authors. Please submit letters typed double-spaced. Letters may be shortened or edited. Include a completed copyright transfer agreement form (available online at http://stroke.ahajournals.org and http://submit-stroke.ahajournals.org).
- © 2012 American Heart Association, Inc.
- Howard G,
- Waller JL,
- Voeks JH,
- Howard VJ,
- Jauch EC,
- Lees KR,
- et al
The Optimising Analysis of Stroke Trials (OAST) Collaboration. Can we improve the statistical analysis of stroke trials? Statistical re-analysis of functional outcomes in stroke trials. Stroke. 2007;38:1911–1915.
The Optimising Analysis of Stroke Trials (OAST) Collaboration. Should stroke trials adjust functional outcome for baseline prognostic factors? Stroke. 2009;40:888–894.