There are a number of potential problems in the application of our approach to expanding the control cohort. First, there may be overlapping aetiologies between cases and the disease cohorts used to expand the control cohort, for example, cases of T1D and additional samples from other autoimmune disease cohorts. Thus, by including diseases of overlapping aetiologies in the case and control cohorts, we reduce our power to detect association with pleiotropic genes. A second problem arises as a result of strong signals of association among cohorts utilized to expand the control group. False-positive signals of association for the disease of interest will be identified unless sufficiently large numbers of samples from other reference cohorts are included as controls. Both of these problems highlight the need for careful selection of cohorts used to expand the control group and care in the interpretation of the results of such analyses. In addition, increases in the false-positive error rates of association tests applied to expanded control cohorts can arise as a result of systematic genotyping error differences between studies. It is crucial, therefore, to follow up any positive signals of association with genotyping in independent samples for replication.
