modest increase in power and precision would probably be sacrificed for an increase in computational efficiency and for practical reasons (such as the availability of suitable software). A more recent study investigated the differences in power between these methods (using simulations) for genetic associations of coronary heart disease (CHD) in a cohort of familial hypercholesterolaemia patients.9 The Cox model had significantly greater power compared with logistic regression for SNP–CHD associations in this very high-risk population. However, this study is not generalisable to cohorts with lower disease incidences. Furthermore, practitioners have thus far assumed findings for cohort studies would extrapolate to case-cohort studies, but this has not been formally investigated.
