To the Editor: In a recent Editorial, Edwin Gale reviewed the congenital rubella syndrome (CRS) and the postulated role of viruses in type 1 diabetes, pulling together most of the literature that links congenital rubella and diabetes [1]. He concluded that the case appears sound for type 2 diabetes, but is at most only suggestive for type 1 diabetes. However, he did not reveal the whole story, in particular, the evidence for HLA associations with diabetes in CRS.

Gale quotes Menser et al. [2] on the incidence of diabetes following CRS, but fails to mention that these investigators also reported that the HLA A1-B8 haplotype was present in 44% (eight out of 18) of their CRS participants with diabetes. HLA A1-B8 was first documented as being increased in frequency in insulin-dependent diabetes just 2 months earlier in the same journal in 1974 [3]. In the CRS participants, this haplotype was present in 75% (three out of four) who were insulin-dependent, with ages at onset of 1.5, 12, 12 and 24 years, and in 36% (five out of 14) who were on oral agents or diet alone, with ages of onset of 28–34 years. The frequency of HLA A1-B8 in the background white population is 10%. Gale states that these studies were carried out ‘by authors with no special interest in diabetes’, but in fact co-author J.A. Burgess was, and still is, a diabetologist.

HLA A1-B8 (-DR3-DQ2) is the classic autoimmune haplotype—associated not only with susceptibility to type 1 diabetes in the non-CRS population but also with other autoimmune diseases such as myasthenia gravis, systemic lupus erythematosus, chronic active hepatitis and coeliac disease—and has been associated with the epidemiology of rubella [4, 5]. The CRS–diabetes–HLA association was confirmed in the New York cohort of 274 patients [6]. In this case, Gale acknowledges the high rate of insulin-requiring diabetes following the rubella epidemic of 1964–1966 in New York, but judges that ‘given the interest of the investigators, the sample may have included a disproportionate number of children with diabetes’, and, ‘ascertainment bias may have contributed to this very high rate, which is otherwise unexplained’. In the context of a critical evaluation of the literature, these statements appear distinctly subjective. The New York data strengthened the association of CRS with type 1 diabetes, because not only was the frequency of HLA DR3 higher in the CRS patients, but also the frequency of HLA DR2, known to be protective against type 1 diabetes in non-CRS populations [7], was decreased.

Further evidence for an autoimmune mechanism in CRS-associated diabetes comes from other studies not included in the review. These demonstrated that T cell clones from both CRS and non-CRS type 1 diabetic individuals, raised in response to epitopes in the islet autoantigen glutamic acid decarboxylase, reacted to peptides from rubella virus [8]. There are several cogent reasons why CRS could have set the stage for type 2 diabetes. Babies with CRS were ‘small for dates’ and had a reduced beta cell mass, and may have had persisting rubella virus infection with reduced postnatal beta cell proliferation. Catch-up growth, resulting in a large child, or puberty with increasing body size and insulin resistance leading to a demand for insulin that exceeds supply, may then lead to type 2 diabetes. Moreover, insulin resistance is now a recognised risk factor for type 1 diabetes [9]. Therefore, in the presence of HLA susceptibility genes for type 1 diabetes, the mechanisms leading to type 2 diabetes after CRS could also promote development of type 1 diabetes. In our view, the evidence, although incomplete, indicates that CRS was associated with diabetes across the spectrum of clinical stereotypes.