There are important differences between the prospective studies from birth and the case-control studies of type 1 diabetes. First, the prospective studies have been performed on subjects who have relatives with type 1 diabetes and, therefore, have the so-called familial type 1 diabetes, a minority of all type 1 diabetes cases. None of the case-control studies have examined risk factors in individuals having a relative with the disease. A second difference is that prospective studies examine the relationship between environmental factors and the development of islet autoimmunity. This is very important because 1) most of the environmental factors examined(breast-feeding, introduction of cow's milk, vaccinations, etc.) are likely to have causative effects early in life and 2) as we have previously shown, islet autoantibodies appear early in life; type 1 diabetes, on the other hand, can manifest at any age(1). Case-control studies can only relate environmental factors to end-point disease; therefore, unlike these prospective studies, case-control studies cannot distinguish between environmental factors that could be primary or secondary in the disease process. Hence, the accumulation of data from these comparatively few labor-intensive prospective studies from birth is fundamental in the identification of factors that may trigger autoimmunity.
As pointed out by Dahlquist(2), the question of power calculations remains crucial to determining associations between environmental factors and disease. Although the power calculations presented by Dahlquist are empirically correct, they fail to take into consideration the actual data from the BABYDIAB cohort. The motivation for our conclusion, that there was no major effect of breast-feeding duration and islet autoimmunity in the offspring of mothers with type 1 diabetes, was that breast-feeding duration in children developing islet autoimmunity tended to be longer, rather than shorter, than those remaining autoantibody-negative (median exclusive breast-feeding duration 8 vs. 4 weeks). Indeed, the odds ratio for developing antibodies in children not breast-fed versus those breast-fed was 0.5 (i.e.,autoantibodies were twice as frequent in breast-fed children) with an upper CI of 1.4, suggesting that even if considerably more children were studied,finding an association similar to that reported by Gerstein(3) [odds ratio of 1.63 (95% CI 1.22-2.17)] would be highly unlikely. Nevertheless, this can only be verified with more data. In addition, the prospective studies should be encouraged to perform the type of meta-analyses reported for case-control studies. Moreover,because the validity of meta-analyses is dependent on an unbiased representation, reporting the data is necessary even if no associations are found (4). Undoubtedly, a study the size of the German BABYDIAB study, which found an association between one of the environmental factors and the appearance of islet autoimmunity, would be published in a high-impact journal. This is true for many of the published case-control studies in which the study populations were often small.
The second issue raised by Dahlquist relates to the definition of what constitutes a major effect. Again, we acknowledge that an association reported by Gerstein for cow's milk may be considerable in certain circumstances. However, we would like to stress that the issue of these environmental factors listed as causative agents has gone beyond the types of associations observed by Gerstein et al. Several studies on anti-bodies to cow's milk proteins report a markedly increased prevalence of such antibodies in patients compared with control subjects; for example, bovine serum albumin antibodies were reported in 100% of patients vs. 0% of control subjects(5). Even when allowing for agents that are more selective with particular HLA genotypes or more persistent in subjects who develop type 1 diabetes, such major differences are not supported by our study, Diabetes Autoimmunity Study in the Young(6), or the Australian BABYDIAB Study (7). We cannot exclude effects from each of the environmental factors presented; however, through the examination of autoimmunity from birth, we cannot attribute them major causative roles in the pathogenesis of type 1 diabetes.
