Elementos didácticos 78

the association of vitamin e with clinical type 1 diabetes was assessed using frozen serum samples from the dime study. despite the small number of cases, inverse associations of borderline significance were observed both in the nested case-control design including cases with diabetes and their matched, seronegative and non-diabetic controls, and in the subcohort of children with preclinical type 1 diabetes. the associations were of similar magnitude to those in an earlier report by Knekt et al. (1999) of a significant inverse association between serum vitamin e and type 1 diabetes among adult finnish men, and lent some support to the hypothesis of a protective effect of vitamin e against developing clinical type 1 diabetes.

the associations of vitamin e with preclinical type 1 diabetes, based on the presence of type 1 diabetes-associated autoantibodies in serum, were analyzed using both serum samples and maternal food frequency data from the DIPP Study. Serum α- and γ-tocopherol concentrations of cases with pre-type 1 diabetes and matched seronegative, non-diabetic controls were compared in a nested case-control design. Neither α- or γ-tocopherol

6 dIscussIoN

concentration in the serum sample collected at one year of age, nor the overall α- or γ-tocopherol concentration up to the age of seroconversion of the case were associated with the risk of preclinical type 1 diabetes. The only significant finding was an interaction between high vs. intermediate concentration of γ-tocopherol at the age of 1 year and the time of seroconversion, which could indicate short-term protection of γ-tocopherol against developing advanced β-cell autoimmunity.

correspondingly, maternal dietary intake of vitamin e during pregnancy was not associated with the risk of preclinical type 1 diabetes in the child. neither was maternal vitamin e intake associated with early endpoints appearing before the age of 2.5 years or with clinical type 1 diabetes. because of the poor validity of the ffQ for vitamin e (erkkola et al. 2001), the lack of association could result from attenuation caused by unreliable intake estimates. the main food sources of vitamin e – dietary fats, cereal products and vegetables – were not associated with pre-type 1 diabetes in the child either, except for low-fat margarines, which showed an inverse association. the validity of the ffQ was acceptable for cereal products and vegetables, but poor for vegetable oils, the most important source of vitamin e. therefore, a protective effect of maternal vitamin e intake during pregnancy against pre- type 1 diabetes cannot be ruled out despite the finding of no association in the present study.

The finding of no association between serum vitamin E concentrations and pre-type 1 diabetes in the DIPP cohort was unexpected, since significant (Knekt et al. 1999) and borderline (study i) inverse associations were found in the two earlier analyses on the mobile clinic study (knekt et al. 1999) and the dime study (study i). the discrepancy of the results could be explained by false inverse association in the mobile clinic and DiMe Studies, or false null association in the DIPP Study. The inverse results in the first two studies may be due to chance, as the number of cases was small and accordingly the confidence intervals of the measure of association were wide in both analyses. Also, the serum samples from both the mobile clinic study and the dime study had been stored frozen at –20°c for several years, and their quality is thus likely to be inferior to the quality of the dipp samples. indeed, vitamin e concentrations were low especially in the dime samples. this may have caused inaccuracy in the results, even if one would expect that the association was attenuated towards null. in consideration of this, confounding by some factor in serum which slows down the degradation of vitamin e upon storage seems possible.

Even if no significant association was observed in the DIPP analysis, confidence intervals did not exclude a moderate inverse association of serum vitamin e concentrations with the risk of pre-type 1 diabetes. the results could also be attenuated by the considerable intraindividual variation in serum vitamin e concentrations (tangney et al. 1987). We had only one serum sample from each individual per year, and for over 40% of cases there was

only one serum sample available overall. furthermore, the validity of serum vitamin e concentrations as an indicator of dietary intake is uncertain, especially in children (byers et al. 1993, hercberg et al. 1994, ortega et al. 2005, drewel et al. 2006, kim et al. 2006).

However, it can also be argued that the findings may reflect a real phenomenon of vitamin e protecting against overt type 1 diabetes, but not against the preclinical stage of the disease. the development of type 1 diabetes begins with genetic disposition and environmental triggers, which are followed by autoimmunity, loss of β-cell mass, and finally overt diabetes (Babaya et al. 2005). In many subjects, the insulitis present in the prediabetic stage will spontaneously recover and not progress to clinical disease. the environmental risk factors may be different for insulitis and for clinical disease (Ludvigsson 2006). In NOD mice, for example, small modifications of the immune system can prevent progression to diabetes, even if they usually do not prevent insulitis (babaya et al. 2005). accordingly, vitamin e may not prevent the autoimmune process which causes the insulitis and initiates β-cell damage, but may protect the β-cells against the cytotoxic effects associated with the autoimmune attack (beales et al. 1994, hayward et al. 1992, Hyppönen 2004). The findings on vitamin E in the present and previous work (Knekt et al. 1999) – borderline or significant inverse association of serum concentration with clinical diabetes, borderline inverse association of serum concentration for advancing from preclinical to clinical diabetes, no proven association of serum concentration or maternal intake with pre-type 1 diabetes – fit well into such a scenario.