There have been impressive advances in the genetics of type 1 diabetes (T1D) that are increasingly being translated into a greater understanding of the disease mechanisms, as reviewed in this journal by Polychronakos and Li (Understanding type 1 diabetes through genetics: advances and prospects. Nature Reviews Genetics 12, 781–792)1. Each susceptibility gene confers a familial relative risk (λ), and multiple susceptibility genes can be modelled to give a joint λ. The derived λ can be compared to the empirical familial clustering of the disease as measured in epidemiological family studies that consider different family members: for example, parents and offspring or siblings. Sibling familial risk (λs) is often used in genetic studies. Polychronakos and Li gave λs values for 34 loci that are reliably associated with T1D (Figure 1 of Polychronakos and Li1), and the authors concluded that the joint contribution of the identified loci was ~60% of the empirical λs of 15. We would like to raise concerns about the use of an empirical λs of 15 for T1D and to highlight the wider need for empirical λs values to be regularly re-evaluated for other complex diseases.

The λs of 15 for T1D has been used in other prominent studies and reviews without discussion of where it has come from1,2,3. In 2003, Hirschhorn4 cited a paper by Spielman and co-workers5 from 1980 as the origin of this value; this paper based λs on a literature review. As the incidence of T1D has increased rapidly, it is likely that a pre-1980 value for λs is no longer accurate. In 2009, we published a family study of T1D and other autoimmune diseases based on nationwide hospitalizations in Sweden between the years 1964 and 2007 (Ref. 6). Practically all childhood patients with T1D are hospitalized in Sweden7. These data suggested that λs for T1D is about 12 for Swedes who were diagnosed in the mid-1990s. Using a λs of 12 instead of 15 suggests that the joint gene effects actually account for a larger proportion of the familial clustering than was previously believed.

In summary, we recommend for T1D and other diseases that the effects of temporal changes in the empirical λs need to be monitored synchronously with genetic λs.