Abstract Background Studies have suggested increased risks of childhood leukaemia after prenatal exposure to antibiotics, particularly nitrofurantoin. However, these findings may be related to the underlying maternal infection. This multinational study aimed to investigate the association between prenatal nitrofurantoin exposure and childhood leukaemia while accounting for maternal infection. Methods In a population-based cohort study of children born in Denmark, Finland, Norway or Sweden from 1997 to 2013, prenatal exposure to nitrofurantoin or pivmecillinam (active comparator) was ascertained from national Prescription Registries. Childhood leukaemia was identified by linkage to national Cancer Registries. Poisson regression was used to estimate incidence rate ratios (IRRs) and incidence rate differences (IRDs) with inverse probability of treatment weights applied to account for confounding. Results We included 44 091 children prenatally exposed to nitrofurantoin and 247 306 children prenatally exposed to pivmecillinam. The children were followed for 9.3 years on average (standard deviation 4.1). There were 161 cases of childhood leukaemia. The weighted IRR for prenatal nitrofurantoin exposure when compared with pivmecillinam was 1.34 (95% confidence interval 0.88, 2.06), corresponding to an IRD of 15 per million person-years. Higher point estimates were seen for first- and third-trimester exposure. There was no evidence of a dose–response relationship. Conclusions Prenatal exposure to nitrofurantoin was not substantially associated with childhood leukaemia, although a slightly elevated IRR with confidence intervals including the null was observed, corresponding to a small absolute risk. The lack of a dose–response relationship and a clear biological mechanism to explain the findings suggests against a causal association.