SUBJECTS AND METHODS

A population based type 1 diabetes register was established in Lazio region in 1989 and was validated by the standards of EURODIAB ACE which defines type 1 diabetes in accordance with the World Health Organisation (WHO) classification. The type 1 diabetic patients selected for this study (330 cases) were diagnosed within the EURODIAB ACE study, and were born between 1977 and 1989. Population based control subjects born in the same period were chosen from primary and secondary school enrolment records in the region, and the selection process was designed in collaboration with the EURODIAB Substudy 2 Group. It is well known that children within the same school may tend to have more similar characteristics compared to children from different schools. (There is a social class correlation among schools.) Therefore, to minimise the effect of this correlation and ensure an accurate sampling of the general population, we selected our controls from 301 different schools while maintaining the same distribution by type (for example, scientific, classic, technical) and grade (elementary, middle, senior high) as exist in Rome and its province.

Questionnaires were distributed to all parents with affected children and those of control subjects. The parents received questionnaires either from their child’s physician for cases or from the child’s teacher for controls. The teachers participating in the survey were trained to motivate the parents of the control subjects to fill out the forms accurately and thereby minimise the bias of inaccurate recall frequent in retrospective studies⁹ because of a lack of interest from the control respondents. The teachers were given detailed written information packets prepared according to the same protocol used in the EURODIAB Substudy 2 study, which provided them with instructions on gathering participant information and obtaining consent, preparing interview schedules, and completing the record sheet.¹⁰ Permission was obtained from the ethical committee at the coordinating centre.

Of 330 cases of type 1 diabetes born between 1977 and 1989, we were able to contact 150 patients (45% of the eligible cases) who maintained the same phone number or address or frequented the same diabetes clinic as recorded in the register. The remaining 55% of the patients were not able to be contacted because of change of residency. All 150 contacted patients replied by filling out the questionnaire. Of 1200 questionnaires distributed among controls, 1100 replied (91%)—a surprisingly high percentage, as a result of the motivational preparation of the parents. Based on the replies obtained from cases (n = 150), 750 were randomly taken from controls—that is, matching one case with five controls by age, which is appropriate for this type of analysis.

In the questionnaire, we presented questions related to exposure to risk factors occurring during the mother’s pregnancy, the neonatal period, and early years of life. Among such questions were: mother’s pregnancy history including drugs taken; occurrence of autoimmune, infectious, and allergic diseases during the neonatal period; birth characteristics and diseases which occurred during the neonatal period; vaccinations received by the subjects; and eating habits during early life, including vitamin D supplementation.

Core variables were transferred from the questionnaire to a standardised record sheet, and the data computerised centrally.

Statistical analysis

For each variable analysed, the odds ratio (OR) and 95% confidence intervals (CI) were calculated (Stat-View). Conditional logistic regression analysis was also performed to adjust the possible confounders. The odds ratio after adjustment is similar to the pooled odds ratio.

RESULTS

Of those who sent back questionnaires, 90–100% responded to all the questions; there was no significant difference between cases and controls in the number of questions answered. Table 1 shows the frequency distribution of age among cases and controls. Overall data analysis showed an expected association with family history for type 1 diabetes significantly increasing the risk of developing the disease (logistic regression analysis: OR 1.17, 95% CI 1.05 to 1.76). Two per cent of patients had a diabetic father, 1.4% had a diabetic mother, and 5.2% had at least one sibling with type 1 diabetes. No patients had both parents or more than one sibling affected by the disease.

Environmental factors occurring during pregnancy

For environmental factors occurring during the mother’s pregnancy (table 2), no significant differences between patients and controls were observed on the use of coffee, tea, and smoking habits. Among maternal diseases occurring during pregnancy, infectious diseases were more frequent in cases than in controls (OR 1.73, 95% CI 1.13 to 2.64) (the questionnaire did not specify between bacterial or viral infections). However, the occurrence rates of gestational diabetes and autoimmune diseases in the mothers did not differ between the two groups, nor did maternal consumption of drugs.

Environmental factors during the neonatal period

Birth characteristics (table 3) did not vary between patients and control subjects: frequency of birth in the hospital (95.0% cases v 95.0% controls), full term birth (87.3% cases v 84.0% controls), average time of gestation (38.2 weeks for cases v 38.2 weeks for controls), average birth weight (3336 g for cases v 3356 g for controls), and postpartum treatment (13.3% cases v 14.0% controls). Furthermore, caesarean section (OR 1.30, 95% CI 0.87 to 1.94), infections (OR 1.20, 95% CI 0.33 to 4.30), respiratory diseases (OR 0.86, 95% CI 0.29 to 2.52), and jaundice (OR 1.18, 95% CI 0.79 to 1.75) were not significantly different between cases and controls.

Environmental factors during early life

During early life (table 4) thyroid gland disorders (OR 18.45, 95% CI 3.79 to 89.82) were more frequent in cases than in controls; this did not reach statistical significance because of the wide confidence interval. In contrast, eczema (OR 1.96, 95% CI 1.16 to 3.32) was significantly more frequent in cases than in controls. No significant differences were observed between the two groups for receiving mandatory and non-mandatory vaccinations or in vitamin D supplementation.

Regarding feeding habits during the first year of life, it was discovered that less than three months of breast feeding, early commencement of weaning, and suspension of breast feeding in favour of cows’ milk formula before 3 months of life were all higher for type 1 diabetic cases than for controls (table 5). However, because of the confidence levels, the only factor reaching statistical significance as a risk factor for type 1 diabetics was short duration (<3 months) of breast feeding.

Logistic regression analysis for overall risk factor significance

Table 6 presents the logistic regression analysis, adjusted for sex and age, for those risk factors which overall have been found to be statistically associated with type 1 diabetes: family history of type 1 diabetes, presence of infectious diseases during mother’s pregnancy, breast feeding <3 months, and occurrence of eczema.

DISCUSSION

This population based study in Rome and its province is the first one describing environmental risk factors for type 1 diabetes in a well defined geographical area of Italy and in a continental Italian population where incidence data for the disease are available.⁵ Overall, our results confirm what has been reported in other Caucasian populations, but differ from Austria where intake of cows’ milk was not associated with type 1 diabetes.⁸

In conducting our study, every effort was made to trace cases, but we lost contact with 55% of them because the families had moved to different regions. It is possible that moving may have led to potential bias in case responses because of different characteristics of the “movers” such as parental income, lifestyle, and social class. Nevertheless, the number of questionnaires obtained from cases (n = 150) comes from one single region, and is still higher than the 102 cases (from eight different European centres) reported by the EURODIAB Substudy 2 Group.¹¹

In this case-control study we analysed several environmental factors that may have occurred during gestation and early life to influence type 1 diabetes development, with the aim of identifying non-genetic risk factors associated with the disease. As expected, family history of type 1 diabetes and infectious diseases during the mother’s pregnancy (without distinction between viral or bacterial infections) were associated with type 1 diabetes.² These findings are consistent with the concept that autoimmune genes are clustered, and families susceptible to one immunologically mediated disease may also be susceptible to others. Our data regarding breast feeding habits indicate that as in other parts of Europe,² in the Lazio region breast feeding for less than three months is a risk factor for type 1 diabetes. Therefore, even in a southern European region such as Italy, breast feeding and avoidance of cows’ milk in early life is an important preventative factor for type 1 diabetes.

An observation of interest, especially as compared to other studies of this kind, was the slightly higher incidence of caesarean section births among cases compared to controls. Similar to the recent report of the EURODIAB Substudy 2 Group (not including continental Italy),¹⁰ we found that among prenatal factors, caesarean section delivery was slightly, but not significantly, associated with type 1 diabetes in our population. However, in both the UK and Sweden, caesarean section birth was found to be a significant risk factor for type 1 diabetes.^12,13 They suggest that exposure to anaesthesia given during a caesarean section may have a deleterious effect on β cells in individuals genetically susceptible to type 1 diabetes.^14–16

Recently great attention has been paid to the role of vitamin D in type 1 diabetes. Vitamin D supplementation has been reported to have a protective effect against type 1 diabetes in northeast Europe,¹⁷ and cod liver oil supplementation during pregnancy was reported to be associated with a reduced risk of type 1 diabetes in offspring.¹⁸ These findings suggest that vitamin D or the n-3 fatty acids, eicosapentaenoic acid and docosahexapentoic acid in the cod liver oil, or both, have a protective effect against type 1 diabetes. However, in our study, no significant correlation was found between vitamin D and risk for type 1 diabetes. This result may be partly explained by the fact that the Lazio region is located in a Mediterranean area where a sunny climate (2500 hours of sunlight annually) predominates throughout the year. Ultraviolet irradiation of the skin is a major source of this vitamin, and skin synthesis of vitamin D can obtain levels up to 10 000 IU (250 μg) when the whole body is exposed to the sun.¹⁹ Thus, even without supplementation, the entire population, including those at risk for type 1 diabetes, may have sufficient levels of vitamin D in their system to prevent disease onset.

By our logistic regression analysis, presence of eczema in early life was found to be associated with type 1 diabetes, in contrast to the Eurodiab Substudy 2¹¹ and Meerwaldt and colleagues,²⁰ where atopic disease and asthma were associated with significant reduction in risk for type 1 diabetes. Allergic contact dermatitis or eczema is a type IV T cell mediated reaction occurring in sensitive individuals with lesions characterised by a mononuclear infiltrate consisting mainly of T cells.²¹ Therefore, we can consider eczema to be an immunologically mediated disease, and it is not surprising that it is associated with type 1 diabetes. Finally, although it did not reach statistical significance, thyroid disorders were more frequent in cases than in controls. Among such disorders, hypothyroxinemia, which is common in the preterm infant, is associated with a number of other autoimmune conditions as recently reported.²²

Our data for the association of eczema with type 1 diabetes and the lack of association of coeliac disease with type 1 diabetes differ from that of other studies in the literature. Such variance reinforces the idea that the same environmental risk factor may have a different impact on different Caucasian populations and further shows the complicated “puzzle” of the role of the environment in the pathogenesis of this disease.