- Poster presentation
- Open Access
Identification and evaluation of novel methods for the analysis of ordinal data in acute stroke trials in comparison to ordinal logistic regression
© Nunn and Gray 2015
Published: 16 November 2015
Ordinal outcomes are common; however, selecting an appropriate analysis is often problematic. Acute stroke trials routinely measure dependency using an ordinal scale as the primary outcome. Historically trials have dichotomised ordinal scales to compare the proportion dependent across groups, which limits statistical power to detect an effect. The OAST Collaboration (2007) showed ordinal logistic regression increased statistical power. Alternative methodologies for ordinal data proposed since 2007 were compared to ordinal logistic regression (OLR).
PubMed and conference proceedings were searched for novel methods for analysis of ordinal outcome scales in stroke. Identified statistical methods were applied to data from the International Stroke Trial which randomised 19,435 patients to receive no treatment, aspirin, heparin or aspirin and heparin. All identified methods were compared to OLR in detection of the effect of aspirin versus no aspirin.
Two regression based techniques were described for the analysis of ordinal scales: partial proportional odds and adjacent categories models. Two methods were proposed using prognosis based cut points on ordinal scales: sliding dichotomy and trichotomy. Three absolute measures of treatment effect were identified: permutation test, number needed to treat (NNT) for ordinal data and win ratio. None of the methods assessed were observed to increase the power to find a treatment effect compared to OLR.
None of the recently proposed methods appear to offer statistical advantage over OLR. However, reporting an absolute measure of effect such as the NNT alongside the results from OLR aids interpretability of the common odds ratio.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.