The aim of this cohort study was to assess the risk of developing cancer, specifically leukaemia, tumours of the central nervous system and lymphoma, before the age of 15 years in children previously exposed to computed tomography (CT) in Germany. Data for children with at least one CT between 1980 and 2010 were abstracted from 20 hospitals. Cancer cases occurring between 1980 and 2010 were identified by stochastic linkage with the German Childhood Cancer Registry (GCCR). For all cases and a sample of non-cases, radiology reports were reviewed to assess the underlying medical conditions at time of the CT. Cases were only included if diagnosis occurred at least 2 years after the first CT and no signs of cancer were recorded in the radiology reports. Standardised incidence ratios (SIR) using incidence rates from the general population were estimated. The cohort included information on 71,073 CT examinations in 44,584 children contributing 161,407 person-years at risk with 46 cases initially identified through linkage with the GCCR. Seven cases had to be excluded due to signs possibly suggestive of cancer at the time of first CT. Overall, more cancer cases were observed (O) than expected (E), but this was mainly driven by unexpected and possibly biased results for lymphomas. For leukaemia, the SIR (SIR = O/E) was 1.72 (95 % CI 0.89–3.01, O = 12), and for CNS tumours, the SIR was 1.35 (95 % CI 0.54–2.78, O = 7). Despite careful examination of the medical information, confounding by indication or reverse causation cannot be ruled out completely and may explain parts of the excess. Furthermore, the CT exposure may have been underestimated as only data from the participating clinics were available. This should be taken into account when interpreting risk estimates.