Hammer, G. P., Seidenbusch, M. C., Schneider, K., Regulla, D. F., Zeeb, H., Spix, C. and Blettner, M. A Cohort Study of Childhood Cancer Incidence after Postnatal Diagnostic X-Ray Exposure. Radiat. Res. 171, 504–512 (2009).
Ionizing radiation is an established cause of cancer, yet little is known about the health effects of doses from diagnostic examinations in children. The risk of childhood cancer was studied in a cohort of 92.957 children who had been examined with diagnostic X rays in a large German hospital during 1976–2003. Radiation doses were reconstructed using the individual dose area product and other exposure parameters, together with conversion coefficients developed specifically for the medical devices and standards used at the radiology department. Newly diagnosed cancers occurring between 1980 and 2006 were determined through record linkage to the German Childhood Cancer Registry. The median radiation dose was 7 μSv. Eight-seven incident cases were found in the cohort: 33 leukemia, 13 lymphoma, 10 central nervous system tumors, and 31 other tumors. The standardized incidence ratio (SIR) for all cancers was 0.99 (95% CI: 0.79–1.22). No trend in the incidence of total cancer, leukemia or solid tumors with increasing radiation dose was observed in the SIR analysis or in the multivariate Poisson regression. Risk did not differ significantly in girls and boys. Overall, while no increase in cancer risk with diagnostic radiation was observed, the results are compatible with a broad range of risk estimates.
Little is known about the causes of childhood cancers. Recent reviews mention genetic disorders, infections, environmental factors and ionizing radiation as risk factors (1, 2). Current epidemiological knowledge on radiation risks in children, especially those exposed to low doses of ionizing radiation, is not as good as for that adults, since fewer and smaller study populations could be investigated (3, 4). The risk of cancer after childhood exposure to high radiation doses has been investigated in the survivors of the atomic bombs (5, 6) and in patients treated with ionizing radiation for either benign diseases (7–9) or cancer (10– 14).
Public awareness about the potential risks of diagnostic uses of ionizing radiation in children was first raised by a publication in 1956 on the risk in children after diagnostic X-ray exposure in utero (the Oxford Survey of Childhood Cancers) (15), which resulted in major changes in medical practice (16, 17). Doll and Wakeford found support for the association reported by Stewart et al. in many case-control studies conducted in different countries and estimated the excess absolute risk (EAR) per gray for all childhood cancers to be about 6% (16). Similarly, Wakeford and Little compared the risk estimates from studies on persons with in utero exposure to the results of the Life-Span Study of atomic bomb survivors and found them to be compatible after accounting for known sources of uncertainty (18, 19). Controversy continues regarding the causal interpretation of the association (20).
The risk of childhood cancer after postnatal diagnostic irradiation has been studied less extensively (21–23). Studies of persons exposed in early childhood have uncertainties similar to those in persons exposed in utero. These are related mainly to problems of recall bias and radiation dose reconstruction (24).
This publication reports the risk of childhood cancer observed in a cohort of about 100,000 children who had been examined using diagnostic X rays.
MATERIAL AND METHODS
The study was performed by linking a large cohort of children exposed to X rays for diagnostic reasons at the Dr. von Hauner Children's Hospital, University of Munich (DvHCH) to the nationwide German Childhood Cancer Registry (GCCR).
The present study includes patients who had been examined in the DvHCH's radiology department between 1976 and 2003 and who were less than 14.5 years of age at the time of first examination. The patients had to be German residents and without any previous diagnosis of cancer. Those who were diagnosed with cancer at or up to 6 months after the first documented X-ray examination in the department were considered as prevalent cases and thus were excluded from the cohort. The referral criteria and diagnoses of examinations were coded according to the International Classification of Diseases, 10th revision (ICD-10), by trained staff.
Source of Data
An electronic documentation system for all diagnostic radiology procedures carried out in the radiology department of the Dr. von Hauner Children's Hospital, University of Munich, Germany, was set up in 1976. In addition to basic information on the individual patients, the documentation also includes information on the body part that was examined, the radiographic views done, and the exposure parameters. Until the introduction of a new Radiological Information System (RIS) in 1998, the referral criteria for the examination and the radiological diagnosis were also recorded. The RIS system does not record these latter items as readily available database fields but instead contains the complete discharge letter.
Individual radiation doses were estimated by using the PAEDOS algorithm developed in the DvHCH's department of radiology. Based on the known exposure parameters for each individual examination, organ doses and whole-body doses were reconstructed for 96% of all examinations. Details have been described elsewhere2 (25, 26).
Briefly, PAEDOS uses conversion coefficients to derive organ doses and whole-body doses from the entrance dose or the dose area product for the given examination. These conversion coefficients were determined for all combinations of examination type, target organ and patient age occurring at the DvHCH using the Monte Carlo software PCXMC developed by the Finnish Radiation and Nuclear Safety Authority (27). This software virtually irradiates hermaphrodite mathematical phantoms of different sizes based on the following parameters: individual records comprising age at exposure, type of examination, target organ, as well as the individual dose area product and other known exposure parameters such as the direction of the beam projection, the patient's position, the distance from source to detector, field dimensions, tube voltage and filtering. These parameters were taken from the study database, and missing parameters were reconstructed from other available data and the documentation of the radiological procedures. When available information was insufficient, no dose was estimated.
About 3.7% of the patients were diagnosed using computed tomography (CT) or contrast medium examinations. These examinations were recorded, but corresponding doses were not reconstructed. This task is much more complex and is planned for the future. As described in the Data Analysis section below, these patients were assigned a special tag. The first CT scanner was installed in the department in 1997.
In the present study, cumulative effective dose (in μSv) was used. Missing doses were replaced with the median dose of subjects of the same age and sex examined in the same year.
Cohort follow-up for newly diagnosed cancers occurring between 1980 and 2006 was done through the GCCR. The GCCR records all childhood cancers (under 15 years old at diagnosis) in Germany since 1980 with a high degree of completeness for most childhood cancer types; only brain tumors are somewhat under-reported, especially in Upper Bavaria. The GCCR has been used previously as a study base for several large-scale childhood cancer investigations at the national and international level (28–30). Pseudonymized cohort data were linked to the pseudonymized data of the GCCR using the software Merge Toolbox for stochastic record linkage (31). An experienced medical documentalist reviewed all matches.
Estimation of Person-Years
All persons in the cohort contributed person-years of observation starting 6 months after the date of initial attendance in the radiology department until whichever of the following came first: date of cancer diagnosis, their 15th birthday, or December 31, 2006. This method of computing person-years overestimates the true figure slightly since mortality is not taken into account. Childhood mortality rates are very low; therefore, a mortality follow-up was not done. To compensate for this, (1) patients with a high a priori mortality risk were tagged, and (2) person-years accumulated after the individual date of last X-ray examination were discounted for non-cancer mortality by multiplying person-years by the appropriate survival rates. For this, childhood mortality rates for West Germany (excluding Berlin)3 were used.
The analyses were restricted to the following cancer groups: all cancers, all leukemia, lymphocytic leukemia, acute myeloid leukemia, tumors of the central nervous system (CNS), lymphoma, and all other cancers.
Risks were quantified by calculating standardized incidence ratios (SIR) and incidence rate ratios (IRR) with 95% confidence intervals (CI). For the SIR calculations, the cancer incidence rates in West Germany (excluding Berlin) provided by the GCCR were used as the reference. The cumulative effective dose (in μSv) was used as exposure measure for all cancers, using a latent period of 6 months. In the SIR analyses, dose was categorized as <1, 1–4.99, 5–9.99, 10–24.99, 25–49.99, 50– 99.99, 100–249.99, 250–499.99 and 500+ μSv. Additionally, the number of examinations (categorized as 1, 2, 3+) was used as a dose proxy to achieve comparability with previous studies. In the IRR analyses, Poisson regression was used to describe incidence rate ratios for all cancers, leukemia, and lymphoma and solid tumors by dose categorized as 0–9.99, 10–49.99 and 50+ μSv, adjusted by age and sex. To allow for a non-linear effect of age at diagnosis, it was categorized as 0–4, 5–9 and 10– 14 years in the regression analysis.
The cohort included some heterogeneous subgroups of patients. Individual children were labeled accordingly, based on the available diagnostic or exposure information. Each individual could be in more than one group.
Children were labeled as “high mortality risk” if their diagnoses were included in a predefined list of ICD-10 codes for serious diseases such as AIDS, complex congenital heart defects, cystic fibrosis or hydrocephalus (Table 1).
Likewise, children with syndromes known to be associated with elevated cancer incidence (e.g. Down's syndrome) included on a second list of ICD-10 codes were marked as “elevated incidence risk”.
Premature children were tagged separately as “premature” when this was noted as an indication or diagnosis. These children carry an increased mortality risk and repeatedly undergo X-ray examinations to check the development of the lungs.
Patients with recorded CT or contrast media examinations were labeled as “highly exposed”. As noted above, exposures from these examinations were not reconstructed. Person time for such patients counted as “high exposure” from the date of first CT or contrast media examination onward.
Children for whom the indication for examination and the diagnosis were completely missing because the RIS system does not record this information could not be tagged in the same manner as other children. These children were marked as “RIS” patients.
The initial cohort consisted of 105,847 patients, of whom 12,890 were excluded: 9757 were 14.5 years or older at their first X-ray examination, 1547 were 15 years or older at the beginning of follow-up, six were examined after 2003, 993 had a prevalent cancer at first examination, 395 had a cancer diagnosis within 6 months after first examination, 16 were diagnosed with cancer before 1980, and 176 had inconsistent dates of birth and examinations that could not be corrected. The net cohort consists of 92,957 children: 50,005 boys, 41,432 girls, and 1520 of unknown sex (Table 2).
On average, 3423 patients per year newly entered the cohort. The mean age of patients entering the cohort was 5.6 years, with 20,546 (22.1%) entering the cohort below 1 year of age and 9096 (9.8%) at 1 year of age. Nearly one quarter of all patients (n = 21,319, 23%) entered the cohort in 1998 or later, the “RIS” patients (Table 2). Among the other 71,638 patients, 14,174 (19.8%) were tagged as “high mortality risk”, 398 (0.6%) as “elevated incidence risk” and 279 (0.4%) as “premature”. A total of 3428 patients were tagged as “highly exposed”, 442 (2.1%) among “RIS” patients and 2986 (4.2%) among the others.
The patients contributed a total of 726,200 person-years of observation time, of which 107,612 person-years (14.8%) were from “RIS” patients. The mean follow-up time was 7.8 years (Table 2).
In total, 87 incident cancer cases (52 in boys and 35 in girls) were found in the GCCR's files (Table 3). Seventeen of the incident cancer cases were in “RIS” patients (nine boys and eight girls). No cases were found among patients marked as “elevated incidence risk” or “premature”. Nine additional suspected cases were identified from hospital records. Because they would not be included in the denominator for rate comparisons, they were excluded from the main analysis.
Most patients (n = 54,605, 59%) had only one recorded examination. A further 17,818 (19%) patients had two, 7515 (8%) had three, and 13019 (14%) had four or more examinations, the maximum being 120 for one patient. The median radiation dose per individual examination declined from 18 μSv [interquartile range (IR): 8–66 μSv] in 1976 to 3 μSv (IR: 0–8 μSv) in the year 2002, with a marked drop in 1982 due to changes in procedures. The median exposure per examination was highest (25 μSv, IR: 10–90 μSv) in 0-year-olds and lowest (<1 μSv) in 12- to 14-year-olds.
Individual cumulative dose (Fig. 1) ranged from 0 to 343.4 mSv, with a median of 7 μSv (IR: 0–48 μSv) and arithmetic mean of 135 μSv, and did not differ in boys and girls. A total of 1984, 77 and seven patients had doses ≥1, 10 and 100 mSv, respectively. In cancer cases, the median and mean cumulative dose were 22 μSv (IR: 6–111 μSv) and 116 μSv, respectively (Table 2). Two cases had cumulative doses above 1 mSv: 1.04 and 3.13 mSv.
Medical procedures involving “high exposures” (CT scans and or contrast media examinations) were mostly used among children with higher mortality risk: Among the 2986 non-“RIS” cohort members classified as “highly exposed”, 2223 (74.4%) were classified as having “high mortality risk”, 89 as having “elevated incidence risk”, and 260 as “premature”.
The overall SIR was 0.99 (95% CI: 0.79–1.22), based on 87 cases (Table 4). SIRs generally did not differ between sexes. The SIR for all cancers was 0.99 (95% CI: 0.74– 1.29) for boys and 1.00 (95% CI: 0.69–1.38) for girls. SIRs were not significantly different from 1.0 for any cancer entity except CNS tumors (all of which were brain tumors), for which it was 0.52 (95% CI: 0.25–0.95). An SIR below unity was observed for the lowest dose category (<1 μSv) and above unity for most other dose categories, although none was significant and no trend was observed (Fig. 1). The same pattern is observed for the number of examinations.
There were only four cases among the patients labeled as “highly exposed”, so no regression model was fitted for this group. Among the other patients, IRRs adjusted for age and sex did not differ from unadjusted IRR. No significant risk increase and no trend were observed for all cancers combined, for leukemia and lymphoma, or for solid tumors combined (Table 5). Patients with “high mortality risk” had IRR below unity for all cancers combined (15 cases), driven by the low IRR for leukemia and lymphoma (seven cases). In other patients, non-significant IRRs of 1.13 (95% CI: 0.53–2.44) and 1.24 (95% CI: 0.54–2.83) were observed for leukemia and lymphoma in the dose categories 10-49.9 μSv and 50+ μSv compared to <10 μSv (total 37 cases), while no elevation was seen for solid tumors (31 cases).
DISCUSSION AND CONCLUSION
This study investigated the childhood cancer risk associated with postnatal diagnostic ionizing radiation exposure in a large cohort of children who underwent diagnostic radiological procedures and were followed up through the nationwide and comprehensive childhood cancer registry in Germany.
Overall, no increase in the risk of solid tumors or leukemia in childhood associated with radiation exposure was observed in this cohort, nor was an increase in risk observed in patients labeled as having had high exposures or “high mortality risk”. The apparent risk reduction for brain tumors might be explained by the documented under-reporting in Upper Bavaria and fact that, in Munich, they are often treated in hospitals other than the DvHCH, where the chance of reporting to the GCCR is even lower.
The strengths of this cohort study are the large population included in the study, the prospective assessment of detailed exposure data over a long period, the ability to reconstruct absorbed doses, and the independent case ascertainment through a cancer registry with a high degree of completeness. This avoids the problem of recall bias encountered in case-control studies. On the other hand, the use of only one source of data bears the risk of missing exposures for some children. This could be of particular relevance should the participants have had procedures which might involve high radiation doses, such as computed tomography (CT). Doses for CT or contrast media examinations were not reconstructed. However, three quarters of cohort members with such exposures were classified a priori as having “high mortality risk”. The low radiation exposure in the cohort, with a median cumulative dose of 7 μSv, reflects the constant efforts of the DvHCH radiology department to optimize and document instruments and procedures. The DvHCH is the largest children's hospital in Munich; there is a small chance that the children seen here could have received large doses of radiation elsewhere. An underestimation of the radiation dose would most likely have led to an overestimation of risks, which does not seem to be the case here.
A further limitation lies in the practical restrictions imposed by the Radiological Information System introduced in 1998. While in principle some of the necessary information such as the indication for the radiological examination and the resulting diagnosis is available from scanned (often handwritten) medical files, abstracting this information would have meant a huge amount of manual record review. This could not be done in the framework of the current study.
Sensitivity analyses including the nine cases ascertained from hospital records generated essentially unchanged results. For four of these nine cases, there were plausible explanations why they were not reported to the GCCR: Two brain tumors and an ovarian cancer were probably treated in an adult ward, and the fourth was an HIV-positive patient with lymphoma. Intensified manual search in the GCCR did not yield any linkable records for these cases.
To the authors' knowledge, there are no cohort studies directly comparable to the one presented here. Previous case-control studies conducted in Germany involving children exposed to postnatal diagnostic X rays did not show any increased risk for leukemia or solid tumors (23, 32). In recent large case-control studies, Shu et al. found a significant increase in the risk of total cancer with the number of postnatal exposures (34), and both Shu et al. and Infante-Rivard et al. reported increases in the risk of acute leukemia with the number of exposures (21, 33). In another study by Shu et al., no such increase for acute leukemia was observed (22). In a recent report from the life-span study of atomic bomb survivors, Preston et al. found a significant excess relative risk of solid cancer in teens and adults with postnatal and in utero doses. The risks decline with attained age and are 1.70 per Sv (95% CI: 1.1–2.5) at age 50 (6). In contrast to earlier studies, a recent review of studies on prenatal and postnatal diagnostic exposures and the risk of childhood cancer (24) found no evidence for an increased risk of leukemia in studies published since 1990.
The results of the present large cohort study suggest that postnatal exposure to diagnostic X rays is not linked to an appreciable increase in the incidence of solid tumors or leukemia in this cohort. No trends with increasing radiation dose from diagnostic imaging were detected. It should be noted that only 2.1% of all patients, including two cases, had cumulative doses above 1000 μSv.
The relevance of the present findings needs to be assessed in the light of current trends in diagnostic imaging. There has been an increase in the use of computed tomography, including on children, which has led to a significant increase in patient exposure (35). These developments pose a continuous challenge for the quantification of cancer risk from diagnostic radiography and associated issues such as improved radiation protection. Recently, Brenner and Hall and Chodick et al. estimated the number of excess lifetime cancer cases due to CT examinations in the U.S. (35) and Israel (36), respectively, using frequency data for CT scans and cancer risk models. They concluded that CT examinations during childhood present non-negligible lifetime health risks. Their findings were criticized because they applied models used in radiation protection for adults. Clearly, more data obtained directly from observational studies among children exposed to diagnostic radiation are required, and the current large, well-documented cohort study contributes to this. It can also serve as a base for further investigations. An extension of the current study including doses from CT and contrast media is planned for the future, as is a further follow-up of cancer incidence and mortality, also into adult age.
The authors wish to thank the staff of the German Childhood Cancer Registry for the excellent collaboration, in particular Claudia Bremensdorfer, who helped with the clerical review. Prof. Doris Bardehle and Dr. Monika Deml coded diagnoses according to ICD-10, Florence Samkange-Zeeb assisted in revising the manuscript, and Susanne Seuchter expertly coordinated study tasks. Their work is gratefully acknowledged. This study was funded by the German Federal Office for Radiation Protection, contract StrSch 4432.
List of ICD-10 Codes Used to Classify Patients as Having “High Mortality Risk” or “Elevated Incidence Risk” in a Cohort of Children who Underwent Diagnostic X-Ray Procedures at Dr. von Hauner Children's Hospital, Munich, Germany, in the Period 1976–2003
Selected Characteristics of the Cohort of Children who Underwent Diagnostic X-Ray Procedures at Dr. von Hauner Children's Hospital, Munich, Germany, in the Period 1976–2003
Number of Incident Cancer Cases Occurring in the Period 1980–2003, by Cancer Diagnosis, Sex and Age at Diagnosis, in a Cohort of Children who Underwent Diagnostic X-Ray Procedures at Dr. von Hauner Children's Hospital, Munich, Germany, in the Period 1976–2003
Observed and Expected Numbers of Incident Cancer Cases and Standardized Incidence Ratios (SIR) of Cohort Members by Cancer Type, Sex, Assigned Tags, Cumulative Exposure and Number of Examinations
Incidence Rate Ratios (IRR) for Categories of Cumulative Effective Dose (in mSv) Adjusted for Sex and Age at Diagnosis of Cancer Obtained through Several Multilevel Poisson Regression Models in a Cohort of Children who Underwent Diagnostic X-Ray Procedures at Dr. von Hauner Children's Hospital, Munich, Germany, in the Period 1976–2003
 M. Seidenbusch, Rekonstruktion von Organ- und Effektivdosen bei konventionellen Röntgenuntersuchungen am Dr. von Haunerschen Kinderspital der Universität München mit einer Berechnung neuer Konversionsfaktoren für die pädiatrische Radiologie (Dissertation). LMU München: Medizinische Fakultät, 2006.