Studies that assess the association of one or a small number of genetic variants with a trait. "Candidate" genes in such studies are commonly selected based on hypothesised functional relevance for the trait being assessed. Candidate genes might then be used as instrumental variables (IVs) in MR studies testing the effect of that trait on an outcome.
Potential limitations include: 1) There is an assumed effect/robust association with the trait (exposure of interest in MR analysis) but this is may not be the case if replication for the candidate gene has not been demonstrated in independent studies; 2) single or a few genetic IVs might increase the likelihood of weak instrument bias; 3) it may not be feasible to use some IV methods, such as MR-Egger and weighted median-based methods, that provide different means of exploring IV validity, with single or few genetic IVs. Potential strengths include: 1) Having knowledge of how genetic IVs function to influence the exposure of interest can be valuable in terms of considering MR assumptions and the extent to which these are likely to be violated.
References
- Colhoun HM, McKeigue PM, Davey Smith G. Problems of reporting genetic associations with complex outcomes: can we avoid being swamped by spurious findings? Lancet 2003; 361: 865-872.
- Lawlor DA, Windmeijer F, Davey Smith G. Is Mendelian randomization 'lost in translation?': comments on 'Mendelian randomization equals instrumental variable analysis with genetic instruments' by Wehby et al. Stat Med 2008; 27: 2750-5.