In this retrospective cohort study, we show significant associations between the estimated radiation doses provided by CT scans to red bone marrow and brain and subsequent incidence of leukaemia and brain tumours. Assuming typical doses for scans done after 2001 in children aged younger than 15 years, cumulative ionising radiation doses from 2–3 head CTs (ie, ~60 mGy) could almost triple the risk of brain tumours and 5–10 head CTs (~50 mGy) could triple the risk of leukaemia.
Although no previous cohort studies have assessed the risk of cancer after CT, several studies have reported significantly increased cancer risks after radiation exposure in the range received from multiple CT scans (100 mGy).19
Such studies include those of survivors of the atomic bombs in Japan,20
and patients who received tens of diagnostic radiographs.22
A few case-control studies have also assessed cancer risks from CT scans on the basis of self-reported history of diagnostic radiograph exposures.23,24
These studies might be subject to recall bias whereby patients are more likely to recall previous medical radiation exposures than are unaffected controls, and also high levels of reporting error. We avoided such bias by taking a cohort approach and assessing more accurate exposure histories from medical records (panel
Panel. Research in context
We searched PubMed and Medline databases without date or language restriction for articles with the search terms “computed tomography”, “ionizing radiation”, “cancer”, “radiation-induced neoplasms”, “case-control”, and “prospective”. We reviewed reports from scientific committees such as the International Commission on Radiological Protection (ICRP), United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR), and Biological Effects of Ionizing Radiations (BEIR), and also a broader range of publications and reports covering medical imaging and radiation exposure. We checked references from selected publications for relevance to this study including comments, correspondence, and editorials. Exposure to ionising radiation is an established risk factor for leukaemia and brain tumours.10,16
Although CT has important clinical uses, concerns exist about the potential cancer risks from the associated ionising radiation, particularly for children. Rates of CT use have been rising rapidly in the developed world.
Increases that we noted in incidence rates of leukaemia and brain tumours after childhood exposure to CT scans are unlikely to be due to confounding factors. The evaluated risks per unit dose were consistent with those derived from recent analyses of cohorts exposed to higher average radiation doses and dose rates. The current study supports the extrapolation of such risk models to doses from CT scans.
In terms of the quantitative estimates of the risk, our primary comparison for leukaemias and brain tumours is with the Life Span Study20
of Japanese atomic bomb survivors, which is the most comprehensive study of cancer after radiation exposure currently available.10,16
The dose-response for leukaemia following childhood exposure and similar follow-up time (<15 years after exposure) in the Life Span Study was 0·045 per mSv (95% CI 0·016–0·188; appendix
) which was much the same as our estimate (ERR of 0·036 per mGy [0·005–0·120]; 1 mSv=1 mGy). For brain tumours, our result (ERR 0·023 per mGy [0·010–0·049]) was about four times higher than was the Life Span Study estimate (0·0061 per mSv [0·0001–0·0639] <20 years after exposure; appendix
), but the CIs are wide and overlapped. We had reduced power to examine risks by subtype of neoplasm, age, or time since exposure compared with the Life Span Study, partly because of the more restricted ranges of length of follow-up and age at exposure. The increased risks noted in our study compared with the Life Span Study might be because existing tumours in some patients were not detected at the time of their first CT. The relatively low-energy x-radiation from CT scans might also be about twice as biologically effective per unit dose as the mainly high-energy γ-rays that were the predominant exposure source from the atomic bombings in Hiroshima and Nagasaki.16
Our large study sample was collected from a wide range of hospitals in Great Britain. Because most medical attendances at hospitals in Great Britain, particularly for the age group in this study, are in public, free-to-access, NHS hospitals, the sample is probably representative of the childhood and young adult population in the country as a whole who undergo CT. Ascertainment of cancer diagnoses by NHSCR is estimated to be 97%25
and therefore there is a low likelihood of losses to follow-up. Patients who were excluded because linkage to their records was not possible had similar characteristics to those who were linked and thus should not have biased conclusions. Because we assessed children and young adults, our results are directly applicable to a highly radiosensitive section of the population,10
although whether the results can be generalised to adulthood CT scans has not been established. Moreover, because most (>80%) of the population assessed was white, whether the results are generalisable to other ethnic groups is unknown.
CT is often used as a diagnostic technique when a solid cancer is suspected. However, information about the reasons for CTs and other clinical variables were not available for this study. Instead, we excluded all scans undertaken in the 2 years before a leukaemia diagnosis and 5 years before a brain tumour diagnosis. Young patients with leukaemia are unlikely to have a CT because of their disease,26
but we still used a cautious approach of applying an exclusion period. By contrast, patients with brain tumours will probably have a number of CT examinations during the diagnostic period, hence the longer exclusion period. Nevertheless, we noted much the same results in sensitivity analyses in which all scans in the 10 years before a brain tumour diagnosis were excluded. The absence of data for other exposures, such as radiographs, is unlikely to have introduced a major bias because the doses from these scans are typically ten-times smaller than those for CT scans. However, we cannot rule out this bias and the increased dose response noted for brain tumours compared with the survivors of the atomic bombs in Japan is also a possible indication of some residual bias despite the long exclusion period.
Previous dose estimates for CT typically provided effective dose rather than organ doses and were restricted in terms of the ages covered. In this study, a series of phantoms with a higher age resolution from newborn to adult was used for both males and females. We also used more realistic anatomy and bone marrow dosimetry models compared with previous computational phantoms. These advanced features allow more accurate and valid estimates of organ-specific doses. Despite these advanced methods, uncertainties exist for our dose estimates. However, such uncertainties are likely to be mainly Berksonian (resulting from applying group-averaged estimates), and thus would not be expected to bias the dose response.27
Collection of detailed scan parameter data for individual patients was not possible. Instead, we used average CT machine settings from two national surveys and assumed that no technical adjustment was made for paediatric patients before 2001.5
Absolute excess risk estimates are necessary to put the risks into perspective with the benefits of the scans. Good evidence from the long-term study of the atomic bomb survivors in Japan suggests that cancer risk persists indefinitely after radiation exposure and most cancer types are inducible by radiation.10,16
At present, we only have sufficient case numbers to assess brain tumours and leukaemia, and the maximum age of patients at the end of follow-up is 45 years, with a minimum age of 6 years and maximum follow-up time of 23 years. Provisional estimates of excess absolute risk for the end of follow-up at about 10 years after exposure suggest that, of 10 000 people between the ages of 0–20 years receiving 10 mGy from a CT scan, there would be about 0·83 (95% CI 0·12–2·77) excess leukaemia cases and 0·32 (0·14–0·69) excess brain tumours (appendix
). Applying the dose estimates for one head CT scan before the age of 10 years () this estimate would translate into approximately one excess case of leukaemia and one excess brain tumour per 10 000 patients. Increased follow-up and analysis of other cancer types is needed to identify the lifetime excess cancer risk associated with CT scans. Some evidence28
suggests that doses in the range delivered by several CT scans might increase the risk of cardiovascular disease. Investigating this feature would require not only the same long-term follow-up required for adulthood cancer outcomes, but also a new approach to obtain cardiovascular incidence data, which is not currently recorded on a registry rather than reliance on mortality data.
Various studies have estimated the potential lifetime excess cancer risks from CT scans from risk projection models, which are largely based on risk models from studies of survivors of the atomic bombs in Japan. Because our relative risk estimates are broadly consistent with the results from the Life Span Study, this study provides additional direct support for the existing lifetime absolute cancer risk projections for paediatric patients.3,7,8,29
The most recent risk projections8
suggest that, for children with normal life expectancy, the lifetime excess risk of any incident cancer for a head CT scan (with typical dose levels used in the USA) is about one cancer per 1000 head CT scans for young children (<5 years), decreasing to about one cancer per 2000 scans for exposure at age 15 years. For an abdominal or pelvic CT scan, the lifetime risks for children are one cancer per 500 scans irrespective of age at exposure. These absolute excess lifetime cancer risks (to age 100 years) are very small compared with the lifetime risk of developing cancer in the general population, which is about one in three, and are also likely to be small compared with the benefits of the scan, providing it is clinically justified.1
We estimated doses for each scan that every patient received, obtained outcome data for the patients, and provided direct evidence that doses at the level children and young adults can receive from CT are associated with increased risks of leukaemia and brain tumours. The dose-response relation that we noted and relative risks of more than 2 for an exposure that is an established carcinogen at higher dose-levels10,16
is evidence that this relation is unlikely to be entirely due to confounding factors. With the increasing use of CT worldwide, particularly within this young population,8
knowledge of the risks based on empirical data will be crucial to assess safety in relation to the benefits that CT provides. Frequent calls have been made to decrease doses, following the as low as reasonably achievable (ALARA) principle, and only scan when justified as in the current image gently campaign.30
In the UK, the Ionising Radiation (Medical Exposure) Regulations mean that a CT scan should only be done when clinically justified, which might explain the low levels of CT use in the UK compared with other countries that do not have such regulations. The immediate benefits of CT outweigh the long-term risks in many settings31
and because of CT's diagnostic accuracy and speed of scanning, notably removing the need for anaesthesia and sedation in young patients, it will remain in widespread practice for the foreseeable future. Further refinements to allow reduction in CT doses should be a priority, not only for the radiology community but also for manufacturers. Alternative diagnostic procedures that do not involve ionising radiation exposure, such as ultrasound and MRI might be appropriate in some clinical settings.