Modelling Risk Ratios from Matched Cohort Data: An Estimating Equation Approach

In certain epidemiological studies, subjects come from an underlying cohort of matched pairs, but only those pairs in which an outcome event occurs are ascertained. Such data can be analysed by using odds ratio modelling methods, but these methods present interpretational problems and do not use information from matched pairs in which both subjects experience the outcome. This paper presents a method which models the risk ratio and makes use of double-outcome pairs. The method is compared with conditional logistic regression in the analysis of the effectiveness of motorcycle helmets in preventing deaths in motorcycle accidents.