Background: Substantial evidence links exposure to moderate or high doses of ionising radiation, particularly in childhood, with increased risk of leukaemia. The association of leukaemia with exposure to low-dose (<100 mSv) radiation is less certain, although this is the dose range most relevant to the general population. We aimed to estimate the risk of leukaemia associated with low-dose radiation exposure in childhood (age <21 years). Methods: In this analysis of historical cohort studies, we pooled eligible cohorts reported up to June 30, 2014. We evaluated leukaemia and myeloid malignancy outcomes in these cohorts with the relevant International Classification of Diseases and International Classification of Diseases for Oncology definitions. The cohorts included had not been treated for malignant disease, had reported at least five cases of the relevant haematopoietic neoplasms, and estimated individual active bone marrow (ABM) doses. We restricted analysis to individuals who were younger than 21 years at first irradiation who had mean cumulative ABM doses of less than 100 mSv. Dose-response models were fitted by use of Poisson regression. The data were received in fully anonymised form by the statistical analyst. Findings: We identified nine eligible cohorts from Canada, France, Japan, Sweden, the UK, and the USA, including 262 573 people who had been exposed to less than 100 mSv enrolled between June 4, 1915, and Dec 31, 2004. Mean follow-up was 19·63 years (SD 17·75) and mean cumulative ABM dose was 19·6 mSv (SD 22·7). 154 myeloid malignancies were identified (which included 79 acute myeloid leukaemias, eight myelodysplastic syndromes, and 36 chronic myeloid leukaemias, in addition to other unspecified myeloid malignancies) and 40 acute lymphoblastic leukaemias, with 221 leukaemias (including otherwise unclassified leukaemias but excluding chronic lymphocytic leukaemia) identified overall. The fitted relative risks at 100 mSv were 3·09 (95% CI 1·41–5·92; ptrend=0·008) for acute myeloid leukaemia and myelodysplastic syndromes combined, 2·56 (1·09–5·06; ptrend=0·033) for acute myeloid leukaemia, and 5·66 (1·35–19·71; ptrend=0·023) for acute lymphoblastic leukaemia. There was no clear dose-response for chronic myeloid leukaemia, which had a relative risk at 100 mSv of 0·36 (0·00–2·36; ptrend=0·394). There were few indications of between-cohort heterogeneity or departure from linearity. For acute myeloid leukaemia and myelodysplastic syndromes combined and for acute lymphoblastic leukaemia, the dose-responses remained significant for doses of less than 50 mSv. Excess absolute risks at 100 mSv were in the range of 0·1–0·4 cases or deaths per 10 000 person-years. Interpretation: The risks of acute myeloid leukaemia and acute lymphoblastic leukaemia were significantly increased after cumulative doses of ionising radiation of less than 100 mSv in childhood or adolescence, with an excess risk also apparent for cumulative radiation doses of less than 50 mSv for some endpoints. These findings support an increased risk of leukaemia associated with low-dose exposure to radiation and imply that the current system of radiological protection is prudent and not overly protective. Funding: National Cancer Institute Intramural Research Program, National Cancer Institute, and US National Institutes for Health.