Seasonality of birth in children with diabetes in Europe: multicentre cohort study

There is increasing evidence that environmental factors in early life, particularly viral infections, influence the risk of developing type 1 (insulin dependent) diabetes.¹ The high incidence of diabetes in the congenital rubella syndrome suggests that intrauterine infection may be important,¹ and the high incidence of enteroviral infection during pregnancy in the mothers of children who subsequently develop type 1 diabetes suggests that other viruses may also be involved.² Since most common viral infections are seasonal, if a significant proportion of cases of childhood diabetes were caused by intrauterine infection it might be expected that the pattern of dates of birth of affected individuals would be abnormal. We previously reported abnormal seasonality of birth in three large independent populations of children with type 1 diabetes in the United Kingdom,³ and a similar pattern has since been reported in the Netherlands.⁴ To determine whether regional variation in seasonal patterns of birth of children with diabetes might provide clues to the aetiology of the disease we studied 20 cohorts of children with diabetes from 16 European countries.

Subjects, methods, and results

We obtained data from population based incidence cohorts of children (0-14 years) who had type 1 diabetes diagnosed after 1989 and were registered with the European diabetes (EURODIAB) project. The dates of birth spanned 20 years (1974-94), and national monthly birth rates were obtained for this period. Data for this extended period were obtained for the previously reported British cohorts and duplicate cases removed. The pattern of births in each cohort was analysed separately. We adjusted for the seasonality of live births in the general population by constructing pseudocohorts of births based on the number of births by month in the period under study. Significance of seasonal trends was tested by the method of Walter and Elwood.⁵

Only the cohorts from Great Britain showed significant differences in the seasonality of birth between children with diabetes and the general population (table). There were no convincing seasonal trends elsewhere in Europe. Combining cohorts from Great Britain revealed a significant sinusoidal pattern of births with a peak in early summer, a trough in winter, and an amplitude of 20% (P=0.006).

Comment

Although many of the cohorts from continental Europe were relatively small and had little power to exclude a modest degree of seasonality of birth, none of the cohorts showed any suggestion of abnormal seasonality. Moreover, aggregation of cases into three larger cohorts relating to areas of low, medium, and high incidence of diabetes (data not shown) also showed no abnormal patterns. Only the five cohorts in Great Britain showed any signs of a seasonal trend. The consistency of the finding in these studies and the significance of the overall seasonality suggest that the abnormal seasonal pattern of births is not due to chance, and it is difficult to conceive of any bias that might account for the observation.

If the abnormal seasonality of birth of children with diabetes in Great Britain, and that reported in the Netherlands,⁴ is a consequence of infection in utero or in early infancy, then the infectious agent(s) responsible would be predicted to be less prevalent, to exhibit a less seasonal pattern of infection, or to be different, in most other parts of Europe. Further studies are required to determine whether abnormal seasonality of birth exists in childhood diabetes in other parts of the world and to identify likely causes. These should be multicentre collaborative studies to avoid bias due to the selective reporting of positive results by single centres.

Table.

Seasonal pattern of dates of birth in cohorts of children with diabetes compared with that expected from national birth rates

Region	No of cases	χ2 for seasonality of birth	P value
Great Britain:
Scotland*	2408	8.9	0.01
Yorkshire*	1232	8.3	0.01
England and Wales*	1270	4.9	0.09
Oxford	758	5.2	0.07
Leicester	329	7.2	0.03
Total	5997	10.2	0.006
Austria	895	0.4	0.8
Czech Republic	1512	3.0	0.2
Denmark	296	3.3	0.2
Germany (Baden-Württemberg)	1303	1.4	0.5
Italy (Lazio)	471	1.9	0.4
Lithuania	486	0.8	0.6
Luxembourg	72	4.7	0.1
Malta	98	5.6	0.06
Northern Ireland	462	1.8	0.4
Poland	448	0.4	0.8
Romania	185	2.4	0.3
Italy (Sardinia)	886	1.7	0.4
Slovakia	889	3.0	0.2
Spain	914	1.6	0.5
Sweden (Stockholm)	520	0.3	0.8

^*Cohorts contain patients reported previously.³