Background: The incidence rates of childhood onset type 1 diabetes are almost universally increasing across the globe but the aetiology of the disease remains largely unknown. We investigated whether birth order is associated with the risk of childhood diabetes by performing a pooled analysis of previous studies.

Methods: Relevant studies published before January 2010 were identified from MEDLINE, Web of Science and EMBASE. Authors of studies provided individual patient data or conducted pre-specified analyses. Meta-analysis techniques were used to derive combined odds ratios (ORs), before and after adjustment for confounders, and investigate heterogeneity.

Results: Data were available for 6 cohort and 25 case–control studies, including 11 955 cases of type 1 diabetes. Overall, there was no evidence of an association prior to adjustment for confounders. After adjustment for maternal age at birth and other confounders, a reduction in the risk of diabetes in second- or later born children became apparent [fully adjusted OR = 0.90 95% confidence interval (CI) 0.83–0.98; P = 0.02] but this association varied markedly between studies (I2 = 67%). An a priori subgroup analysis showed that the association was stronger and more consistent in children <5 years of age (n = 25 studies, maternal age adjusted OR = 0.84 95% CI 0.75, 0.93; I2 = 23%).

Conclusion: Although the association varied between studies, there was some evidence of a lower risk of childhood onset type 1 diabetes with increasing birth order, particularly in children aged <5 years. This finding could reflect increased exposure to infections in early life in later born children.