Infant Diets and Type 1 Diabetes: Title and subTitle BreakToo Early, Too Late, or Just Too Complicated?

Important questions about type 1 diabetes mellitus (DM) remain unanswered, but few have been more perplexing, more exhausting, and perhaps even more unproductive than the issue of what causes the disease.^{1 - 4} While geneticists and immunologists have made measurable progress, the work of epidemiologists at times recalls Veblen's comment that "the outcome of any serious research can only be to make two questions grow where only one question grew before."⁵ Despite this confusion, the notion that environmental influences, or the loss of such influences, may in some way promote the development of type 1 DM appears well-founded, and the changing incidence of the disease is indeed difficult to explain on any other basis.⁶

Earlier explorations of disease causation looked for simple one-to-one correlations between the environment and disease, an approach that worked well for infectious agents, nutritional deficiency, and toxins. The quest for simple explanations should not be discarded too readily given that causative agents have been assigned to a number of conditions that previously were assumed to have a multifactorial etiology, such as the role of Helicobacter pylori in peptic ulcer disease.

For diseases that still resist straightforward explanation, the trend has been for 2-dimensional disease models based on genes and environment to be superseded by dynamic models that incorporate the element of growth and development.^{7 - 8} In type 1 DM this approach assumes interaction between the developing immune system, genetically hard-wired patterns of response, and the timing, duration, or combination of environmental exposures.^{9 - 10} Since it is now clear that autoimmune processes associated with eventual onset of type 1 DM can begin very early in life,^{11 - 12} this strategy mandates prospective studies from birth. In this issue of THE JOURNAL, the reports by Ziegler and colleagues¹³ from the BABYDIAB study and by Norris and colleagues¹⁴ from the DAISY (Diabetes Autoimmunity Study in the Young) outline a link between early infant diet and the development of autoantibodies directed against the pancreatic islet cells.

The BABYDIAB study, established in 1989, is a prospective study of a high-risk population defined as offspring of mothers, fathers, or both with type 1 DM. DAISY was established in 1994 to investigate newborns with and without a family history of type 1 DM. The 2 studies are generally similar in design; both evaluate risk for diabetes in terms of parental diabetes, genetic susceptibility conferred by HLA molecules, the formation of autoantibodies associated with type 1 DM, and candidate environmental influences. Both studies now suggest that the age at which an infant is fed cereal is important in determining his or her risk of type 1 DM. However, both studies fail to provide support for the hypothesis that early exposure to cow's milk might be responsible, and BABYDIAB even suggests that this might be protective. Breastfeeding might still play a role, but had less of an impact than cereal consumption.

These 2 studies point back toward a single trigger hypothesis of disease causation, although several issues remain unresolved in more detailed analysis of the 2 investigations. BABYDIAB implies a specific role for early introduction of gluten, a beguiling possibility because of its role in celiac disease, but it cannot exclude a role for gluten-free cereals whereas DAISY shows similar risk from introduction of rice-based–containing (nongluten) diets. DAISY also suggested a more complex dietary interaction, such that cereal introduction before 3 months and after 7 months of age creates a "window" of relative safety for the intervening period.

How much confidence should researchers, clinicians, and parents place in these findings? At this stage, cautious interest might seem the appropriate response. The advantage of the prospective study from birth is that it has the potential to allow dynamic processes to be viewed over time as in a movie; the practical limitation is that the script of the movie must be inferred from a series of intermittently performed snapshots. Furthermore, autoantibody positivity is used as the outcome measure, and, while this is a powerful predictive measure, the most critical outcome must be actual development of diabetes. A final concern with observational studies like these is the low number of end points—in this case islet autoimmunity with or without progression to diabetes. For example, DAISY identified 34 children with islet autoimmunity from 1149 recruited, and BABYDIAB, while deriving its conclusions from 85 high-risk cases, placed a heavy weight of inference based on 4 children with early exposure to gluten who developed multiple autoantibodies. The potential for overinterpretation of such data is clear.

A major problem with type 1 DM is that too many signposts point to an uncertain future: too much information from animal studies, too many models of immune causation, too many environmental leads, and too many potential interventions. Previous studies of early dietary influences on risk of type 1 DM have, despite their limitations,¹⁵ placed the major emphasis on cow's milk. As a result, an international consortium of researchers led by Finnish and Canadian researchers are now committed to a major intervention trial to test this hypothesis.¹⁶ Rather than plan another dietary intervention, perhaps these new observations first should be reconciled with existing information, seeking to determine what they might reveal about mechanisms underlying disease development. Environmental agents (eg, dietary or enteroviral) implicated in type 1 diabetes have the common element of mediation by gut immune mechanisms.¹⁷ Investigation of this common pathway may prove more productive than advocacy of one potential environmental agent at the expense of another. As argued previously,¹⁵ further intervention trials should not be undertaken until more insight into underlying disease mechanisms has been acquired.

For these reasons, it is clear that the reports from the BABYDIAB study and the DAISY do not present sufficient evidence to suggest that "infant cereal causes diabetes," and hopefully will not be misinterpreted as such by parents and the public. Accordingly, given current knowledge, the recommendation put forth by Norris et al¹⁴ is most appropriate—current infant feeding guidelines should not be changed. In addition, further prospective cohorts must be examined to confirm and extend these findings. The infrastructure for doing so has recently been established with a consortium of 6 international centers designed to identify environmental factors that contribute to the risk of type 1 DM (Triggers and Environmental Determinants in Diabetes of the Young [TEDDY]).¹⁸ Large-scale collaborative studies are needed to determine whether cereal introduction should be included among them. Only then will it be possible to adequately address the question of whether as a risk factor for type 1 DM, infant diets are strongly linked to the issues of too much, too little, too early, or too late.