BACKGROUND: It has been hypothesized that risk factors of childhood cancers may already operate during the prenatal and neonatal period. Results of previous epidemiological studies have been inconsistent. METHODS: During 1992-1997 a large case-control study on childhood cancers and a variety of potential risk factors was conducted in Germany. Cases were ascertained by the German Childhood Cancer Registry. Each case was matched to a population-based control of the same age and gender, sampled from the district where the case lived at the date of diagnosis. For the analyses, 2358 cases and 2588 controls were available. RESULTS: Risk of childhood acute leukaemia increased with maternal age < or =20 years at time of delivery (odds ratio [OR] = 1.9, 95% CI: 1.1-3.2), lower (<2500 g: OR = 1.7, 95% CI: 1.1-2.8) and higher birthweight (>4000 g: OR = 1.4, 95% CI: 1.0-1.8, P < 0.05), and hormonal treatment because of infertility (OR = 1.6, 95% CI: 1.0-2.5, P < 0.05). No associations were seen for parental smoking habits, maternal alcohol consumption during pregnancy and fetal losses. Parity was associated only with subgroups of acute leukaemias. Regarding non-Hodgkin's lymphoma we observed an elevated OR for lower birthweight and heavy maternal smoking during pregnancy (>20 cigarettes/day) and a decreased OR for children with one or two siblings. Only a few significant findings were seen for the different groups of solid tumours. CONCLUSIONS: Overall, only weak associations were identified and the evaluated risk factors operating during the neonatal and prenatal period account at most for only a small proportion of childhood cancers.