We suggest that a potentially useful approach is to adopt a two-stage study design: initial genome-wide analysis of case-control cohorts for GSV discovery, although likely to be underpowered, will nevertheless yield a set of candidate GSVs; unselected population cohorts can then be screened for individuals carrying these GSVs. The key advantage of this approach is that carriers identified from population cohorts are not biased (either qualitatively or quantitatively) by pre-existing ascertainment criteria, so that the impact of the GSV on phenotype can be directly analysed using more powerful quantitative methods.
