Background: A previous investigation of the occurrence of childhood acute leukemia around the Belgian nuclear sites has shown positive associations around one nuclear site (Mol-Dessel). In the following years, the Belgian Cancer Registry has made data available at the smallest administrative unit for which demographic information exists in Belgium, i.e. the statistical sector. This offers the advantage to reduce the potential misclassification due to large geographical scales.Methods: The current study performed for the period 2006-2016 uses Poisson models to investigate (i) the incidence of childhood acute leukemia within 20 km around the four Belgian nuclear sites, (ii) exposure-response relationships between cancer incidence and surrogate exposures from the nuclear sites (distance, wind direction frequency and exposure by hypothetical radioactive discharges taking into account historical meteorological conditions). All analyses are carried out at statistical sector level.Results: Higher incidence rate ratios were found for children <15 years (7 cases, RR = 3.01, 95% CI: 1.43;6.35) and children <5 years (< 5 cases, RR = 3.62, 95% CI: 1.35;9.74) living less than 5 km from the site of Mol-Dessel. In addition, there was an indication for positive exposure-response relationships with the different types of surrogate exposures.Conclusion: Results confirm an increased incidence of acute childhood leukemia around Mol-Dessel, but the number of cases remains very small. Random variation cannot be excluded and the ecological design does not allow concluding on causality. These findings emphasize the need for more in-depth research into the risk factors of childhood leukemia, for a better understanding of the etiology of this disease. ; This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. The authors gratefully acknowledge Prof. Elisabeth Cardis (Institute for Global Health, Spain), Dr Enora Cléro (Institute for Radiological Protection and ...