Printed in Great Britain Combined association and aggregation analysis of data from case-control family studies

Genetic epidemiologists are increasingly interested in family studies using the case-control family study design, and this has motivated several recent developments in related statistical methodology. By summarising and extending some of these developments, this paper proposes an estimating-equation-based framework for analysing family data col-lected from case-control studies, motivated by the likelihood approach proposed by Whittemore (1995). The proposed estimating equation approach requires modelling assumptions about means, correlation coefficients and, at most, trivariate correlation coefficients, and it can be used for analysing family data with arbitrary pedigree structures, which would be prohibitive with the likelihood-based approach. Further, this new approach utilises all available data from case and control probands and from their relatives in assessing association between disease outcome and covariates, and data from relatives in evaluating aggregation of disease phenotypes within families, thus enhancing its efficiency. The approach is illustrated by analysing data collected from case-control studies on ovarian and colorectal cancers