OBJECTIVE To investigate whether there was a temporal and geographical clustering of time of birth for infants with childhood-onset diabetes.

RESEARCH DESIGN AND METHODS The nationwide Swedish Childhood Diabetes Registry, which ascertains 99% of children with recent-onset diabetes (0-14 years), was linked with the Swedish Medical Birth Registry. Clustering of 3,725 patients as to place and time of birth was studied compared with the general population. For each municipality (and in the three large cities of Sweden for each parish), the observed number of patients was compared with the expected number calculated from the average total rate and the number of births in that municipality. Clustering in time of birth within municipality was analyzed using a modification of a set technique by Chen (14).

RESULTS There was no consistent variability in diabetes risk by calendar birth month, but for specific years, the risk varied during the year. When geographic localization for place of birth was studied on a municipality level, four municipalities snowed a statistically significant case excess while one would have been expected by chance. When we looked for clusters in both time and space for date of birth, clearly more clusters than expected were identified (P < 0.01). Of the total of 198 primary clusters, 42 included three or more patients being born in the same municipality within an unlikely short period always <2 years.

CONCLUSIONS This is the first study indicating a clustering according to place and time of birth for later risk to develop type I diabetes. Such a phenomenon would agree with the hypothesis that infections in early life, including fetal infections, can increase the risk for diabetes.