A simulation-based comparison of techniques to correct for measurement error in matched case-control studies

The presence of measurement errors affecting the covariates in regression models is a relevant topic in many scientific areas, as, for example, in epidemiology. Among the techniques proposed in literature to reduce the effects of measurement errors on inferential results, likelihood-based methods have received less attention than alternatives. The main reason relies on the computational complexity and on the difficulties related to selection and specification of the relationships between the involved variables. Despite this, likelihood methods have the advantage of providing inferential results which can be more accurate than those from other approaches. In this paper, we focus on the application of likelihood methods to correct for measurement errors affecting a single covariate, within a matched case-control setting. The approach is compared to commonly used alternatives as regression calibration and SIMEX. The comparison is performed by simulation, under a broad range of measurement errors structures. We base the analysis on matched case-control data generated according to a simulation scheme. This aims to emulate the data collected by an epidemiological population-based study on the aetiology of childhood malignancies, which is currently under completion in Italy.