Statistical meta-analysis for ordinal categorical data

Abstract

Traditionally the odds ratio (OR) is used for measuring the extent of association between exposure and its binary outcomes in randomised controlled trials (RCTs) and similar studies. It is inapplicable if the outcomes are on an ordinal scale with more than two categories. In those studies, the generalised odds ratio (GOR) is used for summarising the difference between two stochastically ordered distributions of an ordinal categorical variable. Meta-analysis combines data from various independent trials in estimating the overall effect measure to make the sample size larger so that the inference based on the combined data is more reliable. In this paper we developed a method of meta-analysis using the GOR under independent multinomial sampling scheme for ordinal categorical data.