Abstract
20 years ago, 3 manuscripts describing doses and potential cancer risks from CT scans in children raised awareness of a growing public health problem. We reviewed the epidemiological studies that were initiated in response to these concerns that assessed cancer risks from CT scans using medical record linkage. We evaluated the study methodology and findings and provide recommendations for optimal study design for new efforts. We identified 17 eligible studies; 13 with published risk estimates, and 4 in progress. There was wide variability in the study methodology, however, which made comparison of findings challenging. Key differences included whether the study focused on childhood or adulthood exposure, radiosensitive outcomes (e.g. leukemia, brain tumors) or all cancers, the exposure metrics (e.g. organ doses, effective dose or number of CTs) and control for biases (e.g. latency and exclusion periods and confounding by indication). We were able to compare results for the subset of studies that evaluated leukemia or brain tumors. There were eight studies of leukemia risk in relation to red bone marrow (RBM) dose, effective dose or number of CTs; seven reported a positive dose–response, which was statistically significant (p < 0.05) in four studies. Six of the seven studies of brain tumors also found a positive dose–response and in five, this was statistically significant. Mean RBM dose ranged from 6 to 12 mGy and mean brain dose from 18 to 43 mGy. In a meta-analysis of the studies of childhood exposure the summary ERR/100 mGy was 1.05 (95%CI: -0.58, 2.69) for leukemia/myelodisplastic syndrome (n = 5 studies) and 0.80 (95%CI: 0.48–1.12) for brain tumors (n = 4 studies) (p-heterogeneity >0.1). Confounding by cancer pre-disposing conditions was unlikely in these five studies of leukemia. The summary risk estimate for brain tumors could be over estimated, however, due to reverse causation. In conclusion, there is growing evidence from epidemiological data that CT scans can cause cancer. The absolute risks to individual patients are, however, likely to be small. Ongoing large multicenter cohorts and future pooling efforts will provide more precise risk quantification.
Introduction
20 years ago a trio of manuscripts describing the dose levels and potential cancer risks from CT scans in children raised awareness of a growing public health problem.1–3 Within months of these publications, plans were underway for the first epidemiological studies to investigate the potential risks. Although the levels of CT scan use were highest in Japan and the US at that time,4 feasibility work suggested that the disjointed healthcare systems were an impediment to the proposed studies that required large-scale medical record linkage. Researchers turned then to countries with universal healthcare systems and national cancer registries, which led quickly to the initiation of the UK and Australian pediatric CT scan cohorts, and later the multicenter European effort EPI-CT. There was close co-ordination of these efforts to ensure, where possible, similar protocols and study designs that would eventually enable the studies to be combined. More recently, several studies have been conducted using data from national insurance databases that are well suited to medical record linkage studies, in countries such as Taiwan and Korea. A large-scale North American study is now also in progress nested within the US health maintenance organizations and the Ontario Health Insurance Plan.
Here, we review what we have learnt about the cancer risks from CT scans from the last two decades of this research. We focused on the studies that are based on medical record linkage, as it is well established that self-reported medical radiation exposures can be subject to biased and unreliable recall.5 As well as reviewing the study findings, we discuss the issues that have been raised regarding study methodology and make recommendations for optimal study design for new efforts.
Literature review
We included cohort or case–control studies designed to estimate cancer risk after exposure to CT scans at any age. Studies in which exposure information was not based on medical records (i.e. self-reported) were excluded. Only manuscripts written in English were included and no restriction was applied regarding year of publication.
Studies were identified through a search conducted in Pub-Med using the following terms: “(((Children OR adults OR population OR participants OR patients) AND ((CT-scan OR “CT scan” OR “computed tomography” OR “diagnostic” OR “medical radiation”) AND (“ionizing radiation” OR “ionising radiation”))) AND ((Cancer OR malignancy) AND (risk OR ERR OR SIR OR “standardized incidence ratio”))) AND ((Cohort OR case-control OR epidemiological))”. The search was performed on January 25, 2021. Pub-Med records were considered for inclusion by screening, in a first step, the title and abstract, and subsequently the full text. Additional records were searched screening the references list of included articles. For each included study, details on the design, population characteristic and risk estimates were extracted in parallel by EP and LV.
Study summaries
We identified 441 records in our Pub-Med search. As detailed in the Prisma flowchart (Figure 1), 60 were selected with the first title-abstract screening, and, of these, 21 records were included after the text screening. An additional 11 records were included from a manual review of cited references. Finally, 32 manuscripts were included, corresponding to 17 studies in 13 countries and 1 multicenter European study (EPICT)6–37
Figure 1.
PRISMA flowchart for article selection.
Of the 17 studies, 13 have already published results on risk estimation and 4 are still in progress (Table 1). Most of the studies focused on CT scans in childhood and young adulthood; only two assessed adulthood exposure. All the studies were retrospective passive record linkage studies using either hospital records (n = 9), insurance databases (n = 7) or both (n = 1) to ascertain CT scan exposure with outcome data from cancer registries or the same insurance databases (Table 2). Most studies were cohort design but there were three case–control studies that focused on radiosensitive outcomes such as leukemia and thyroid cancer.
Table 1.
Study design and characteristics for the eligible record linkage studies of CT scans and cancer risk
| Study type and countrya (ref) | Age at exposure (yrs) | Calendar period | Sample size (approximate) | Exclusion period (yrs) | Average follow-up (yrs) | Results available |
|---|---|---|---|---|---|---|
| Australia (cohort)6,7 | 0–19 | 1985–2005 | 11 000 000 | 1 | Mean exposed=9.5; unexposed=17 | Yes |
| Canadab (cohort)8 | 16–50 | 1995–2014 | 1 300 000 | 1 | Median exposed=5.9; unexposed=11.1 | Yes |
| Europe (cohort) (EPI-CT)9,10 | 0–22c | 1980–2013c | 950 000 | na | na | No |
| Finland (case-control)11 | 0–15 | 1978–2011 | 911 cases; 2730 controls | 2 | nk | Yes |
| France (cohort)12–14 | 0–10 | 2000–2010 | 60 000 | 2 | Mean=4 | Yes |
| Germany (cohort)15–17 | 0–15 | 1980–2010 | 40 000 | 2 | Mean=4.1 | Yes |
| Israel (cohort)18,19 | 0–18 | 1985–2005 | 30 000 | na | na | No |
| South Korea (cohort) [20] | 0–19 | 2006–2015 | 12 000 000 | 2 | Mean=8 | Yes |
| Spain (cohort)21 | 0–21 | 1991–2013 | 100 000 | na | na | No |
| Sweden (matched cohort)22 | 0–99 | 1973–1995 | 80 000 | 5 | Mean exposed=20; unexposed=19 | Yes |
| Taiwan (adult case–control)d23 | > 25 | 2000–2013 | Cases: thyroid 22,853; leukemia 13,040; NHL 20,157 (ca/co= 1:10) |
3 | Mean=9-10 | Yes |
| Taiwan (childhood case–control)d24 | 0–16 | 1997–2013 | Cases: Leukemia 1423; lymphoma 272; intracranial 838 (ca/co= 1:10) |
1 | nk | Yes |
| Taiwan (matched cohort)d25 | 0–18 | 1996–2008 | 1 20 000 | 2 | Mean=4 | Yes |
| The Netherlands (cohort)26,27 | 0–18 | 1979-2012 | 1 40 000 | Leukemia= 2, solid tumors=5 | Mean=8.5 | Yes |
| USA/Canada (cohort) (RIC Study)e28–30 | 0–20 | 1995–2016 | 2,200,000 pregnant women (FC) & 3,500,000 children (CC) | na | na | No |
| USA (cohort)31 | 0–6 | 1991–2001 | 104 | 0 | Mean=15 | Yes |
| UK (cohort)32–37 | 0–21 | 1985–2002 | 1 80 000 | Leukemia= 2, solid tumors=5 | Mean leukemia=9.6 years; brain tumors=6.7 | Yes |
CC, Childhood cohort; FC, Fetal cohort; NHL, Non hodgkin lymphoma; ca/co, Case:Control ratio; na, Not applicable; nk, Not known.
Study name if applicable.
Ontario Health Insurance Plan.
Variation between participating countries.
Based on data from Taiwan National Health Insurance Research Database (NHIRD).
Six U.S. healthcare systems (Kaiser Permanente Northern California, Northwest, Washington, and Hawaii, Harvard Pilgrim Health Care, and Marshfield Clinic Health System and Ontario Health Insurance Plan).
Table 2.
Exposure information, outcome and medical history data in studies of CT scans and cancer risk
| Country (design) | Exposure | Outcome | Medical history data | ||||
|---|---|---|---|---|---|---|---|
| CT exposure data (main source) | Organ dose estimation | % exposed | Cancer outcomes evaluated | Cancer ascertainment | CT indication (source) | Cancer predisposing conditions | |
| Australia (cohort) | Insurance claims | Noa | 6.2% | All cancers & multiple individual sites | Cancer registries | No | No |
| Canada (cohort) | Insurance claims | No | 0.4% | Breast cancer | Cancer registries | No | No |
| Europe (cohort) EPI-CT | RIS | Yes | 100% | na | Cancer registries | Yes (subset) | Yes (subset) |
| Finland (case–control) | RIS | Yes | <1% | Leukemia | Cancer registries | No | Yes |
| France (cohort) | RIS | Yes | 100% | Leukemia, brain tumors | Cancer registry | No | Yes |
| Germany (cohort) | RIS | Yes | 100% | All cancers, leukemia, lymphoma, brain tumors | Cancer registry | Yes (RIS/radiologist reports) | Yes |
| Israel (cohort) | RIS & Insurance claims | Yes | 100% | na | Cancer registries | Yes (Radiologist reports) | No |
| South Korea (cohort) | Insurance claims | No | 9.8% | All cancers & multiple individual sites | Insurance claims | No | No |
| Spain (cohort) | RIS | Yes | 100% | na | Cancer registries | No | No |
| Sweden (matched cohort) | RIS | Yes | 22% | Meningioma | Cancer registries | Yes (Radiologist reports) | No |
| Taiwan (adult case–control) | Insurance claims | No | 4-5% | Leukemia, NHL, thyroid cancer | Insurance claims | No | Yes |
| Taiwan (childhood case–control) | Insurance claims | No | 2-5% | Leukemia, brain tumors, lymphoma | Insurance claims | No | No |
| Taiwan (matched cohort) | Insurance claims | No | 25% | All cancers, leukemia, brain tumors | Insurance claims | No | Yes |
| The Netherlands (cohort) | RIS | Yes | 100% | Leukemia, brain tumors | Cancer registries | No | Yes (subset) |
| USA/Canada (cohort) | RIS | Yes | <6% | na | Cancer registries | nk | Yes |
| USA (cohort) | RIS | Yes | 100% | All cancers | Medical records | No | No |
| UK (cohort) | RIS | Yes | 100% | Leukemia, brain tumors, Hodgkin lymphoma | Cancer registries | Yes (RIS subset) | Yes (subset) |
NHL, Non hodgkin lymphoma; RIS, Radiologists information systems; na, Not applicable; nk, Not known.
Individual organ dosimetry recently completed6.
Organ doses were estimated for 11 of the studies (Table 2); studies based on insurance databases generally did not attempt to reconstruct organ doses. The approach used to estimate organ doses ranged from using published doses according to scan region and age (least individualized) to doses calculated directly from abstracted sets of CT films from the study population (most individualized). No study was able to abstract all the CT films for their entire study population. Another key design difference was whether the studies only included exposed individuals (n = 8) vs those that were based on general population data. Whilst many of these studies were very large, the proportion of the population that was exposed was generally quite low (<10%). Analytic approaches were driven by the type of exposure data and ranged from analysis of total cancer risk in relation to number of CT scans (regardless of scan region) to analysis of specific cancer subtypes in relation to estimated organ dose or effective dose. Other important sources of variability in the analytic approach include how latency and exclusion periods were applied, with several studies using exclusion periods of <2 years (n = 4). More than half the studies had information on cancer predisposing conditions (n = 9) but few had information on the indication for the CT scan (n = 5).
There were eight studies that evaluated leukemia risk according to radiation dose (n = 5 red bone marrow dose and n = 1 effective dose) or number of CTs (n = 2), and of these seven reported a positive dose–response, which was statistically significant in four of the studies (Table 3). The number of cases in some studies was quite small (e.g. <20 in the French and German cohorts). Mean red bone marrow dose in the studies with individual estimates ranged from 6 to 12 mGy. Although the estimated excess relative risk per 100 mGy (ERR/100 mGy) varied from 0.04 to 13 across the five studies with available data the confidence intervals were wide. A meta-analysis using a random effects model performed using STATA38 could not reject homogeneity (p = 0.14) and the summary ERR/100 mGy was 1.05 (95%CI: −0.58, 2.69) (Supplementary Material 1). Most of these studies of leukemia (n = 6) controlled for confounding by cancer pre-disposing conditions, generally by exclusion. The South Korean and Australian studies evaluated leukemia (and brain tumor risk), but only compared exposed to unexposed subjects.
Table 3.
Results for the studies of leukemia/MDS risk in relation to estimated radiation dose or number of CT scans
| Country (design) | Outcome (cases) | Exposure metric | Dose (average, max) | Dose response | Risk estimate (95% CI) | Control for pre- disposing conditionsa | Adjustment/matching factors |
|---|---|---|---|---|---|---|---|
| Finland (case–control) | Leukemia (n = 1093) | Red bone marrow dose | 6 mGy, 30 + mGy | ↗* | EOR/100 mGy = 13 (2-26) | Yes | Sex, year of birth |
| France (cohort) | Leukemia (n = 17) | Red bone marrow dose | 7 mGy, 100 + mGy | ↗ | ERR/100mGy = 1.6 (-2.3 to 2.7) | Yes | Sex, age |
| Germany (cohort) | Leukemia (n = 12) | Red bone marrow dose | 12 mGy | ↗ | ERR/100mGy = 0.9 (-2 to 3.7) | Yes | Sex, age |
| Taiwan (adulthood case–control) | Leukemia (n = 13,040) | Effective dose | nk | ↗* | OR (>30 mSv vs 0 mSv)=9 (3–31) p-trend < 0.001 for age < 45 years | Yes | Sex, age |
| Taiwan (childhood case–control) | Leukemia (n = 1423) | Number of CTs | na | ↗ | OR/CT = 1.07 (0.78 to 1.45) p-trend = 0.69 | No | Sex, year of birth, urbanization |
| Taiwan (matched cohort) | Leukemia (n = 25) | Number of head CTs | na | ↗* | HR (3 + head CTs vs 0)=17.1 (2.32–131) p-trend = 0.045 | Yes | Sex, age |
| The Netherlands (cohort) | Leukemia/MDS (n = 63) | Red bone marrow dose | 10mGy, 20 + mGy | ↔ | ERR/100mGy = 0.04 (-0.12 to 1.6) | No | Sex, year & age |
| UK (cohort) | Leukemia/MDS (n = 74) | Red bone marrow dose | 12mGy, 50 + mGy | ↗* | ERR/100mGy = 3 (0.3 to 10.9) | Yes | Sex, age & socioeconomic status |
MDS, Myelodisplastic syndrome; Nk, Not known; na, Not applicable.
↗Increased risk with increasing dose/CT scans; ↗* Statistically significant increased risk with increasing dose/CT scans; ↔ No evidence of a dose-response relationship.
EOR, Excess Odds Ratio; ERR, Excess Relative Risk; OR, Odds Ratio; HR, Hazard Ratio
Analyses excluded patients with leukemia-predisposing conditions, including: organ transplant, human immunodeficiency virus infection, primary immune deficiencies, Fanconi anemia, ataxia telangiectasia, xeroderma pigmentosum, Bloom, Noonan and Down genetic syndromes.
Six of the seven studies that evaluated brain tumors found a positive dose–response and in five, this was statistically significant (Table 4). The estimated ERR/100 mGy varied from 0.7 to 1.2 across the four studies that provided organ-specific dose–response estimates. We could not reject homogeneity of risks across studies (p = 0.9) and the summary ERR/100 mGy was 0.79 (95%CI: 0.47–1.11) (Supplementary Material 1). The cumulative mean brain dose was generally higher than the red bone marrow dose; ranging from 18 to 43 mGy. Most of these studies evaluated the impact of longer exclusion periods as an indirect method for controlling for reverse causation (where the CT scan was performed to detect a potential cancer), and generally found that there were still increased risks with a 10-year exclusion period, but that the risk estimates were reduced. Only two of these studies directly evaluated reverse causation with data on CT indication. Whilst these findings all point in the direction of supporting increased cancer risks, there are several issues that need to be considered in the interpretation, which we summarize below.
Table 4.
Results for the studies of brain tumor risk in relation to estimated brain dose or number of CT scans
| Country (design) | Outcome (cases) | Exposure metric | Dose (average, max) | Dose–response | Risk estimate (95% CI) | Control for reverse causationa | Adjustment/matching factors |
|---|---|---|---|---|---|---|---|
| France (cohort) | CNS tumors (n = 22) | Brain dose | 18mGy, 100 + mGy | ↗ | ERR/100mGy = 0.7 (-0.1 to 1.0) | No | Sex, age at first CT scan |
| Germany (cohort) | CNS tumours (n = 7) | Brain dose | 34mGy | ↗* | ERR/100mGy = 0.8 (0.4 to 1.3) | Yes | Sex, attained age |
| Sweden (matched cohort) | Meningioma (n = 96) | Brain dose | 100 + mGy | ↔ | OR 100 + mGy vs 0 mGy = 2.13 (0.52 to 8.82) p-trend = 0.07 | Yes | Sex, age, residence |
| Taiwan (childhood case-control) | Intracranial malignancy (n = 838) | Number of CTs | na | ↗* | OR/CT = 1.55 (1.17 to 2.04) p-trend = 0.002 | No | Sex, year of birth, urbanization |
| Taiwan (matched cohort) | Malignant/benign brain tumors (n = 49) | Number of head CTs | na | ↗* | HR 3 + head CTs vs 0 = 10.4 (1.41 to 76) p-trend < 0.0001 | No | Sex, age |
| The Netherlands (cohort) | Malignant/benign brain tumors (n = 84) | Brain dose | 39mGy, 120 + mGy | ↗* | ERR/100mGy = 0.86 (0.20 to 2.22) | No | Sex, year, age |
| UK (cohort) | Malignant/benign brain tumors (n = 135) | Brain dose | 43mGy, 350 + mGy | ↗* | ERR/100mGy = 1.2 (0.4 to 3.1) | Yesa | Sex, age & socioeconomic status |
CNS, Central nervous system tumors; na, Not applicable.
↗ Increased risk with increasing dose/CT scans; ↗* Statistically significant increased risk with increasing dose/CT scans; ↔ No evidence of a dose–response relationship.
ERR, Excess Relative Risk; OR, Odds Ratio; HR, Hazard Ratio
Using CT indication data, analysis excluded patients with previous/possible undiagnosed cancer.
Study strengths and weaknesses
In reviewing these studies, we have highlighted several challenges in the study design. It is important to distinguish issues that can introduce bias into the central risk estimate (and potentially its standard error) from those that reduce the power and precision of the risk estimate. Figure 2 describes the key components for an optimal epidemiological study design including patient characteristics, exposure data and outcome data.
Figure 2.
Key components for optimal record-linkage study design.
Patient characteristics: information on cancer pre-disposing conditions should be collected if possible to evaluate confounding by indication. For example, leukemia risk could be confounded by Down syndrome if this condition is related to the probabilityof undergoing CT scans. Medical history can be ascertained from a variety of sources such as disease registries (e.g. transplant registries) if full electronic medical records are not available. If the condition that predispose to cancer is not related to the frequency of CT scanning, then this will not be a confounder. There are few other established risk factors for brain tumors and leukemia,39 which are the outcomes of primary interest in most of these studies. Therefore, apart from age, sex and the cancer predisposing conditions the only other potential confounder that was routinely considered was high socioeconomic status/income, which has been associated with an increased risk of childhood leukemia in many studies.39 This will only confound the relationship with CT scans if there is also an association with the number of CTs or the estimated organ dose. Some studies have reported a higher CT frequency among children with low socioeconomic status, possibly due to a higher rate of injuries.40 This would therefore act as a negative confounder, biasing the risk coefficient towards the null. It has also been shown that the relationship could depend on the particular socioeconomic status indicator.21 These examples highlight the importance of considering the potential direction of the bias. In addition, the strength of the confounding should be considered and assessment of whether the observed effect could be entirely explained by the confounder. Several methodological studies have assessed this question for confounding by indication and shown that it is unlikely to be a strong confounder in CT scan studies.41–43
Exposure data: ideally, we would estimate individualized organ doses using CT films for each individual scan and accounting for the age, sex and anthropometry of the patient. In practice, the films have proven difficult to obtain, especially for the studies that predated electronic film storage. The approach used by the majority of studies was the use of radiology information system (RIS) data for the scan region combined with other data sources such as national surveys, scanner or hospital protocols to reconstruct typical doses by year, age and sex.6,35 Whilst this approach will introduce dose error, this shared error is unlikely to bias the dose–response estimate.44 Results from analyses of number of CT scans or effective doses, rather than organ dose, are difficult to interpret because of the heterogeneity in organ doses from different scan regions.
Outcome data: linkage to a high-quality population cancer registry that covers the entire study period is the optimal method for ascertaining cancers. In addition, reasons for exiting the study including death or emigration should be collected, e.g. by linkage to national vital status data systems. Pathology reports are a valuable additional source of information on disease subtype, which is critical for classifying hematopoietic malignancies.33 It is still rare for cancer registries to capture non-malignant cancers such as benign brain tumors and myelodysplastic syndrome (MDS), and there remain questions about the completeness of reporting in those countries, such as the UK that do register these outcomes. Misclassification of outcomes will most likely bias the risk estimate towards the null, unless it is related to exposure levels.45
Power/sample size: the red bone marrow and brain are the most radiosensitive organs in childhood and as head CTs are the most common pediatric CT scan and leukemia and brain tumors are the most common childhood cancers, these will be the outcomes with the highest power.46 Lack of statistical significance in some of the smaller studies, therefore, might be explained by lack of power.44
Analytic approach: radiation-related solid cancers generally take at least 5 years to develop after exposure and leukemia at least 2 years and so an appropriate latency period should be used by lagging the doses appropriately.46 Indication for the CT is important for assessing reverse causation, whereby the CT scan is part of the cancer diagnostic work-up rather than the cause of the cancer. As noted above, indication data were not available for most of the studies conducted to date. The UK and German studies used the radiologist’s notes from the RIS database to assess whether the CT scan was performed because of cancer symptoms for a subset of the population15,33 and in the Swedish study referral notes were used.22 An indirect approach to managing reverse causation is to use an exclusion period to only evaluate cancers that are diagnosed a certain period after the first CT scan. This is a particular concern for brain tumors, and at least 2 or more years exclusion is probably necessary,33 or longer for benign brain tumors.22 As CT scans are rarely used to diagnose leukemia, this is not a major concern for this outcome.
Discussion
In the last decade, 13 studies of CT scans and subsequent cancer risk have published findings and at least 4 more studies are in progress. Most of the studies focused on pediatric CT scans and to date, they include more than 3.5 million exposed children. Even within the electronic medical record linkage studies considered here, there was variation in methods. The most informative are those large studies with individualized organ dose estimates for leukemia or brain tumors, which are the most radiosensitive outcomes in children. Seven of eight studies reported a positive dose–response relationship for leukemia/MDS and six of seven studies found increasing risk of brain tumors with increasing exposure levels. Only a few studies published dose–response estimates based on organ doses, but the risk coefficients for childhood exposure were statistically compatible with a summary ERR/100 mGy of 1.05 (95%CI: −0.58, 2.69) for leukemia/MDS and 0.80 (95%CI: 0.48–1.12) for brain tumors. Confounding by cancer pre-disposing conditions was unlikely in most of these studies for leukemia. It remains challenging to assess the role of reverse causation in the relationship between head CT scans and brain tumors, and indirect evidence from longer exclusion periods suggests that the brain tumor risk could be over estimated.
There have been several previous studies of repeated diagnostic X-rays and cancer risk that established the principle that multiple small doses of ionizing radiation can cause cancer.47–49 The cumulative doses in these historical studies of fluoroscopy for monitoring tuberculosis and spine X-rays for monitoring scoliosis were generally much higher than in the CT cohorts considered here. Nevertheless, the positive findings we summarized for leukemia and brain tumors within this low-dose range are also consistent with the growing body of epidemiological data from occupational and environmental exposures that support cancer risks from low-dose (<100 mGy) protracted radiation exposures.50
There are several strengths and limitations of these studies and we reviewed these in detail. Key strengths include the use of record linkage of radiology and cancer registry databases, which rules out the risk of recall bias and outcome ascertainment bias. Evidence of a dose–response relationship is also less likely to be due to confounding than comparison of exposed to unexposed individuals. Many studies collected information on cancer predisposing conditions, which also showed that these were unlikely to be strong confounders, at least in the specific study setting. By evaluating childhood leukemia and brain tumors statistical power was maximized as the red bone marrow and the brain are highly radiosensitive organs commonly exposed during pediatric CT scans. It is challenging to reconstruct highly individualized organ doses and five of the studies had not done so. Few were able to collect the indication for the CT scan, which prohibited direct evaluation of reverse causation for the relationship between head CT scans and brain tumor risk. However, as CT scans are not typically involved in the diagnosis of leukemia, this should not be a source of bias in that relationship. Future studies should prioritize collection of indication data to enable evaluation of reverse causation.
Use of CT scans continues to increase in the UK51 and the US.52 In 2019, there were 5.7 million CT scans performed in England, a 70% increase compared to 2012 (3.3 million). The rate remains much lower than in the US, however; currently about 90 CT scans per 1000 population in England vs 250 in the USA.51,52 Image Gently and other campaigns have reduced CT doses in children,53,54 but clinically unjustified exposures likely continue. Continued efforts to collect data on trends in pediatric CT scans is important as children are at much greater risk of radiation-related cancer than adults.46 There is also limited data on exposure levels in lower- and middle-income countries, but there are growing concerns about dose levels and appropriateness of use.55,56 More research is clearly warranted as well as continuing monitoring of exposure levels in higher-income countries.
In conclusion, most of the epidemiological studies we reviewed reported positive relationships between CT scans and cancer risk, particularly for childhood leukemia and brain tumors. Organ-specific dose–response estimates, where available, were statistically compatible across studies and the summary ERR was significantly increased for brain tumors. The absolute risks are likely to be small, however, because the outcomes such as childhood leukemia are rare; the estimate from the UK CT study was about 1 excess cancer per 10,000 CT scans in the 10 years after exposure.40 Hence, the benefits from the CT scan will outweigh the risks for individual patients if the procedure is clinically justified, and the dose is optimized. There are several large efforts ongoing including the multicenter North America study, the combined European analysis (EPI-CT), and an updated analysis of the Australian study with individual doses and extended follow-up, which are all due to report findings in the next 1–2 years. These large multicenter cohorts and future pooling efforts of these studies will provide more precise quantification of the potential cancer risks.
Contributor Information
Amy Berrington de Gonzalez, Email: berringtona@mail.nih.gov.
Elisa Pasqual, Email: elisa.pasqual@nih.gov.
Lene Veiga, Email: lene.veiga@nih.gov.
REFERENCES
- 1.Brenner D, Elliston C, Hall E, Berdon W. Estimated risks of radiation-induced fatal cancer from pediatric CT. AJR Am J Roentgenol 2001; 176: 289–96. doi: 10.2214/ajr.176.2.1760289 [DOI] [PubMed] [Google Scholar]
- 2.Donnelly LF, Emery KH, Brody AS, Laor T, Gylys-Morin VM, Anton CG, et al. Minimizing radiation dose for pediatric body applications of single-detector helical CT: strategies at a large children's Hospital. AJR Am J Roentgenol 2001; 176: 303–6. doi: 10.2214/ajr.176.2.1760303 [DOI] [PubMed] [Google Scholar]
- 3.Paterson A, Frush DP, Donnelly LF. Helical CT of the body: are settings adjusted for pediatric patients? AJR Am J Roentgenol 2001; 176: 297–301. doi: 10.2214/ajr.176.2.1760297 [DOI] [PubMed] [Google Scholar]
- 4.Berrington de González A, Darby S. Risk of cancer from diagnostic X-rays: estimates for the UK and 14 other countries. The Lancet 2004; 363: 345–51. doi: 10.1016/S0140-6736(04)15433-0 [DOI] [PubMed] [Google Scholar]
- 5.Berrington de Gonzalez A, et al. Comparison of documented and recalled histories of exposure to diagnostic x-rays in case-control studies of thyroid cancer. Am J Epidemiol 2003; 157: 652–63. doi: 10.1093/aje/kwg026 [DOI] [PubMed] [Google Scholar]
- 6.Brady Z, Forsythe A, McBain-Miller J, Scurrah KJ, Smoll N, Lin Y, et al. CT dosimetry for the Australian cohort data linkage study. Radiat Prot Dosimetry 2020; 191: 423–38. doi: 10.1093/rpd/ncaa175 [DOI] [PubMed] [Google Scholar]
- 7.Mathews JD, Forsythe AV, Brady Z, Butler MW, Goergen SK, Byrnes GB, et al. Cancer risk in 680 000 people exposed to computed tomography scans in childhood or adolescence: data linkage study of 11 million Australians. BMJ 2013; 346(may21 1): f2360. doi: 10.1136/bmj.f2360 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Burton KR, Park AL, Fralick M, Ray JG. Risk of early‐onset breast cancer among women exposed to thoracic computed tomography in pregnancy or early postpartum. J Thromb Haemost 2018; 16: 876–85. doi: 10.1111/jth.13980 [DOI] [PubMed] [Google Scholar]
- 9.Bernier M-O, Baysson H, Pearce MS, Moissonnier M, Cardis E, Hauptmann M, et al. Cohort profile: the EPI-CT study: a European pooled epidemiological study to quantify the risk of radiation-induced cancer from paediatric CT. Int J Epidemiol 2019; 48: 379–81. doi: 10.1093/ije/dyy231 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Bosch de Basea M, Pearce MS, Kesminiene A, Bernier M-O, Dabin J, Engels H, et al. EPI-CT: design, challenges and epidemiological methods of an international study on cancer risk after paediatric and young adult CT. J. Radiol. Prot. 2015; 35: 611–28. doi: 10.1088/0952-4746/35/3/611 [DOI] [PubMed] [Google Scholar]
- 11.Nikkilä A, Raitanen J, Lohi O, Auvinen A. Radiation exposure from computerized tomography and risk of childhood leukemia: Finnish register-based case-control study of childhood leukemia (FRECCLE. Haematologica 2018; 103: 1873–80. doi: 10.3324/haematol.2018.187716 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Bernier M-O, Rehel J-L, Brisse HJ, Wu-Zhou X, Caer-Lorho S, Jacob S, et al. Radiation exposure from CT in early childhood: a French large-scale multicentre study. Br J Radiol 2012; 85: 53–60. doi: 10.1259/bjr/90758403 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Journy N, Rehel J-L, Ducou Le Pointe H, Lee C, Brisse H, Chateil J-F, et al. Are the studies on cancer risk from CT scans biased by indication? elements of answer from a large-scale cohort study in France. Br J Cancer 2015; 112: 185–93. doi: 10.1038/bjc.2014.526 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Journy N, Roué T, Cardis E, Le Pointe HD, Brisse H, Chateil J-F, et al. Childhood CT scans and cancer risk: impact of predisposing factors for cancer on the risk estimates. J Radiol Prot 2016; 36: N1–7. doi: 10.1088/0952-4746/36/1/N1 [DOI] [PubMed] [Google Scholar]
- 15.Krille L, Dreger S, Schindel R, Albrecht T, Asmussen M, Barkhausen J, et al. Risk of cancer incidence before the age of 15 years after exposure to ionising radiation from computed tomography: results from a German cohort study. Radiat Environ Biophys 2015; 54: 1–12. doi: 10.1007/s00411-014-0580-3 [DOI] [PubMed] [Google Scholar]
- 16.Krille L, Jahnen A, Mildenberger P, Schneider K, Weisser G, Zeeb H, et al. Computed tomography in children: multicenter cohort study design for the evaluation of cancer risk. Eur J Epidemiol 2011; 26: 249–50. doi: 10.1007/s10654-011-9549-6 [DOI] [PubMed] [Google Scholar]
- 17.Krille L, Zeeb H, Jahnen A, Mildenberger P, Seidenbusch M, Schneider K, et al. Computed tomographies and cancer risk in children: a literature overview of CT practices, risk estimations and an epidemiologic cohort study proposal. Radiat Environ Biophys 2012; 51: 103–11. doi: 10.1007/s00411-012-0405-1 [DOI] [PubMed] [Google Scholar]
- 18.Chodick G, Levin M, Kleinerman RA, Shwarz M, Shalev V, Ashkenazi S, et al. Differences in characteristics of pediatric patients undergoing computed tomography between hospitals and primary care settings: implications for assessing cancer follow-up studies. Isr J Health Policy Res 2015; 4: 33. doi: 10.1186/s13584-015-0031-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Chodick G, Ronckers CM, Shalev V, Ron E. Excess lifetime cancer mortality risk attributable to radiation exposure from computed tomography examinations in children. Isr Med Assoc J 2007; 9: 584–7. [PubMed] [Google Scholar]
- 20.Hong J-Y, Han K, Jung J-H, Kim JS. Association of exposure to diagnostic low-dose ionizing radiation with risk of cancer among Youths in South Korea. JAMA Netw Open 2019; 2: e1910584. doi: 10.1001/jamanetworkopen.2019.10584 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Bosch de Basea M, Espinosa A, Gil M, Figuerola J, Pardina M, Vilar J, et al. CT scan exposure in Spanish children and young adults by socioeconomic status: cross-sectional analysis of cohort data. PLoS One 2018; 13: e0196449. doi: 10.1371/journal.pone.0196449 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Nordenskjöld AC, Bujila R, Aspelin P, Flodmark O, Kaijser M. Risk of meningioma after CT of the head. Radiology 2017; 285: 568–75. doi: 10.1148/radiol.2017161433 [DOI] [PubMed] [Google Scholar]
- 23.Shao Y-H, Tsai K, Kim S, Wu Y-J, Demissie K. Exposure to tomographic scans and cancer risks. JNCI Cancer Spectr 2020; 4: pkz072. doi: 10.1093/jncics/pkz072 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Li I-G, Yang Y-H, Li Y-T, Tsai Y-H, . Paediatric computed tomography and subsequent risk of leukaemia, intracranial malignancy and lymphoma: a nationwide population-based cohort study. Sci Rep 2020; 10: 7759. doi: 10.1038/s41598-020-64805-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Huang W-Y, Muo C-H, Lin C-Y, Jen Y-M, Yang M-H, Lin J-C, et al. Paediatric head CT scan and subsequent risk of malignancy and benign brain tumour: a nation-wide population-based cohort study. Br J Cancer 2014; 110: 2354–60. doi: 10.1038/bjc.2014.103 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Meulepas JM, Ronckers CM, Smets AMJB, Nievelstein RAJ, Gradowska P, Lee C, et al. Radiation exposure from pediatric CT scans and subsequent cancer risk in the Netherlands. J Natl Cancer Inst 2019; 111: 256–63. doi: 10.1093/jnci/djy104 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Meulepas JM, Ronckers CM, Smets AMJB, Nievelstein RAJ, Jahnen A, Lee C, et al. Leukemia and brain tumors among children after radiation exposure from CT scans: design and methodological opportunities of the Dutch pediatric CT study. Eur J Epidemiol 2014; 29: 293–301. doi: 10.1007/s10654-014-9900-9 [DOI] [PubMed] [Google Scholar]
- 28.Kwan ML, Miglioretti DL, Marlow EC, Aiello Bowles EJ, Weinmann S, Cheng SY, et al. Trends in medical imaging during pregnancy in the United States and Ontario, Canada, 1996 to 2016. JAMA Netw Open 2019; 2: e197249: e197249. doi: 10.1001/jamanetworkopen.2019.7249 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Miglioretti DL, Johnson E, Williams A, Greenlee RT, Weinmann S, Solberg LI, et al. The use of computed tomography in pediatrics and the associated radiation exposure and estimated cancer risk. JAMA Pediatr 2013; 167: 700–7. doi: 10.1001/jamapediatrics.2013.311 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Smith-Bindman R, Kwan ML, Marlow EC, Theis MK, Bolch W, Cheng SY, et al. Trends in use of medical imaging in US health care systems and in Ontario, Canada, 2000-2016. JAMA 2019; 322: 843–56. doi: 10.1001/jama.2019.11456 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.White IK, Shaikh KA, Moore RJ, Bullis CL, Sami MT, Gianaris TJ, et al. Risk of radiation-induced malignancies from CT scanning in children who underwent shunt treatment before 6 years of age: a retrospective cohort study with a minimum 10-year follow-up. J Neurosurg 2014; 13: 514–9. doi: 10.3171/2014.2.PEDS12508 [DOI] [PubMed] [Google Scholar]
- 32.Berrington de Gonzalez A, Journy N, Lee C, Morton LM, Harbron RW, Stewart DR, et al. No association between radiation dose from pediatric CT scans and risk of subsequent Hodgkin lymphoma. Cancer Epidemiology Biomarkers & Prevention 2017; 26: 804–6. doi: 10.1158/1055-9965.EPI-16-1011 [DOI] [PubMed] [Google Scholar]
- 33.Berrington de Gonzalez A, Salotti JA, McHugh K, Little MP, Harbron RW, Lee C, et al. Relationship between paediatric CT scans and subsequent risk of leukaemia and brain tumours: assessment of the impact of underlying conditions. Br J Cancer 2016; 114: 388–94. doi: 10.1038/bjc.2015.415 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Journy NM, McHugh K, Harbron RW, Pearce MS, Berrington De Gonzalez A. Medical conditions associated with the use of CT in children and young adults, great Britain, 1995–2008. Br J Radiol 2016; 89: 20160532. doi: 10.1259/bjr.20160532 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Kim KP, Berrington de González A, Pearce MS, Salotti JA, Parker L, McHugh K, et al. Development of a database of organ doses for paediatric and young adult CT scans in the United Kingdom. Radiat Prot Dosimetry 2012; 150: 415–26. doi: 10.1093/rpd/ncr429 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Lee C, Morton LM, Berrington de Gonzalez A. A novel method to estimate lymphocyte dose and application to pediatric and young adult CT patients in the United Kingdom. Radiat Prot Dosimetry 2018; 178: 116–21. doi: 10.1093/rpd/ncx084 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Pearce MS, Salotti JA, Little MP, McHugh K, Lee C, Kim KP, et al. Radiation exposure from CT scans in childhood and subsequent risk of leukaemia and brain tumours: a retrospective cohort study. The Lancet 2012; 380: 499–505. doi: 10.1016/S0140-6736(12)60815-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.StataCorp .Stata statistical software. 16 ed. College Station, TX: StataCorp; 2019. [Google Scholar]
- 39.Roman E, Lightfoot T, Picton S, Kinsey S. Chapter 59: Childhood Cancers. Cancer Epidemiology & Prevention, 4th ed. New York, NY: OUP; 2017. [Google Scholar]
- 40.Pearce MS, Salotti JA, McHugh K, Pyo Kim K, Craft AW, Lubin J, et al. Socio-economic variation in CT scanning in northern England, 1990-2002. BMC Health Serv Res 2012; 12: 24. doi: 10.1186/1472-6963-12-24 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Meulepas JM, Hauptmann M, Lubin JH, Shuryak I, Brenner DJ. Is there unmeasured indication bias in radiation-related cancer risk estimates from studies of computed tomography? Radiat Res 2018; 189: 128–35. doi: 10.1667/RR14807.1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Meulepas JM, Ronckers CM, Merks J, Weijerman ME, Lubin JH, Hauptmann M. Confounding of the association between radiation exposure from CT scans and risk of leukemia and brain tumors by cancer susceptibility syndromes. J Radiol Prot 2016; 36: 953–74. doi: 10.1088/0952-4746/36/4/953 [DOI] [PubMed] [Google Scholar]
- 43.Schubauer-Berigan MK, Berrington de Gonzalez A, Cardis E, Laurier D, Lubin JH, Hauptmann M, et al. Evaluation of confounding and selection bias in epidemiological studies of populations exposed to low-dose, high-energy photon radiation. JNCI Monographs 2020; 2020: 133–53. doi: 10.1093/jncimonographs/lgaa008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Gilbert ES, Little MP, Preston DL, Stram DO. Issues in interpreting epidemiologic studies of populations exposed to low-dose, high-energy photon radiation. JNCI Monographs 2020; 2020: 176–87. doi: 10.1093/jncimonographs/lgaa004 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Linet MS, Schubauer-Berigan MK, Berrington de González A. Outcome assessment in epidemiological studies of low-dose radiation exposure and cancer risks: sources, level of ascertainment, and misclassification. JNCI Monographs 2020; 2020: 154–75. doi: 10.1093/jncimonographs/lgaa007 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Berrington de Gonzalez A, Rajaraman, P A, Schubaeur-Berigan M. Chapter 13: ionizing radiation. Cancer epidemiology & prevention. 4th ed. New York, NY: OUP; 2017. [Google Scholar]
- 47.Boice JD, Preston D, Davis FG, Monson RR. Frequent chest X-ray fluoroscopy and breast cancer incidence among tuberculosis patients in Massachusetts. Radiat Res 1991; 125: 214–22. doi: 10.2307/3577890 [DOI] [PubMed] [Google Scholar]
- 48.Miller AB, Howe GR, Sherman GJ, Lindsay JP, Yaffe MJ, Dinner PJ, et al. Mortality from breast cancer after irradiation during fluoroscopic examinations in patients being treated for tuberculosis. N Engl J Med Overseas Ed 1989; 321: 1285–9. doi: 10.1056/NEJM198911093211902 [DOI] [PubMed] [Google Scholar]
- 49.Ronckers CM, Doody MM, Lonstein JE, Stovall M, Land CE. Multiple diagnostic x-rays for spine deformities and risk of breast cancer. Cancer Epidemiology Biomarkers & Prevention 2008; 17: 605–13. doi: 10.1158/1055-9965.EPI-07-2628 [DOI] [PubMed] [Google Scholar]
- 50.Hauptmann M, Daniels RD, Cardis E, Cullings HM, Kendall G, Laurier D, et al. Epidemiological studies of low-dose ionizing radiation and cancer: summary bias assessment and meta-analysis. JNCI Monographs 2020; 2020: 188–200. doi: 10.1093/jncimonographs/lgaa010 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.NHS England. Diagnostic imaging dataset annual statistical release. 2021. Available from: https://www.england.nhs.uk/statistics/statistical-work-areas/diagnostic-imaging-dataset/.
- 52.Mettler FA, Mahesh M, Bhargavan-Chatfield M, Chambers CE, Elee JG, Frush DP, et al. Patient exposure from radiologic and nuclear medicine procedures in the United States: procedure volume and effective dose for the period 2006-2016. Radiology 2020; 295: 418–27. doi: 10.1148/radiol.2020192256 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Arch ME, Frush DP. Pediatric body MDCT: a 5-year follow-up survey of scanning parameters used by pediatric radiologists. AJR Am J Roentgenol 2008; 191: 611–7. doi: 10.2214/AJR.07.2989 [DOI] [PubMed] [Google Scholar]
- 54.Lee C, Pearce MS, Salotti JA, Harbron RW, Little MP, McHugh K, et al. Reduction in radiation doses from paediatric CT scans in Great Britain. Br J Radiol 2016; 89: 20150305. doi: 10.1259/bjr.20150305 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Dovales ACM, da Rosa LAR, Kesminiene A, Pearce MS, Veiga LHS. Patterns and trends of computed tomography usage in outpatients of the Brazilian public healthcare system, 2001–2011. J. Radiol. Prot. 2016; 36: 547–60. doi: 10.1088/0952-4746/36/3/547 [DOI] [PubMed] [Google Scholar]
- 56.Vassileva J, Rehani MM, Al-Dhuhli H, Al-Naemi HM, Al-Suwaidi JS, Appelgate K, et al. IAEA survey of pediatric CT practice in 40 countries in Asia, Europe, Latin America, and Africa: part 1, frequency and appropriateness. AJR Am J Roentgenol 2012; 198: 1021–31. doi: 10.2214/AJR.11.7273 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.