We concluded that the substantial effort required to recall and reanalyze all of the affected plates in the T2D studies (14/29 in HPFS and 15/41 in NHS) was very unlikely to make a significant improvement in the quality of the data, so these data sets were released to dbGaP without doing so. However, we advise GWAS analysts to consider re-calling a plate whenever one or more of the following occur: (a) a significant portion of the plate consists of low quality samples; (b) the plate is an outlier with respect to median missing call rate and/or the allelic frequency-plate association test; (c) the study has high genomic inflation and low quality cluster plots for association test hits; (d) case-control association with affected plates.
