GWA studies are prone to detection of false positives; to ensure that the most promising signals were taken through to the replication stage we applied a ‘clumping’ strategy [22]. Significant SNPs were clumped together as a single association signal if in linkage disequilibrium over a 250 kilobase pair distance. Prioritization for replication was as follows; SNPs with multiple supporting independent clumps in the same region; support from quantitative trait results, as this is not dependent on the WTCCC2 controls; significant SNPs with a p-value ≤10−6 with some support in the surrounding region; finally, where clumped results located within a gene, relevant functional data was also taken into consideration.
