Article Text
PDF
Abstract
Objectives To evaluate cumulative occupational radiation dose response and haematopoietic malignancy mortality risks in the US radiologic technologist cohort.
Methods Among 110 297 radiologic technologists (83 655 women, 26 642 men) who completed a baseline questionnaire sometime during 1983–1998, a retrospective cohort study was undertaken to assess cumulative, low-to-moderate occupational radiation dose and haematopoietic malignancy mortality risks during 1983–2012. Cumulative bone marrow dose (mean 8.5 mGy, range 0–430 mGy) was estimated based on 921 134 badge monitoring measurements during 1960–1997, work histories and historical data; 35.4% of estimated doses were based on badge measurements. Poisson regression was used to estimate excess relative risk of haematopoietic cancers per 100 milligray (ERR/100 mGy) bone-marrow absorbed dose, adjusting for attained age, sex and birth year.
Results Deaths from baseline questionnaire completion through 2012 included 133 myeloid neoplasms, 381 lymphoid neoplasms and 155 leukaemias excluding chronic lymphocytic leukaemia (CLL). Based on a linear dose-response, no significant ERR/100 mGy occurred for acute myeloid leukaemia (ERR=0.0002, 95% CI <−0.02 to 0.24, p-trend>0.5, 85 cases) or leukaemia excluding CLL (ERR=0.05, 95% CI <−0.09 to 0.24, p-trend=0.21, 155 cases). No significant dose-response trends were observed overall for CLL (ERR<−0.023, 95% CI <−0.025 to 0.18, p-trend=0.45, 32 cases), non-Hodgkin lymphoma (ERR=0.03, 95% CI <−0.2 to 0.18, p-trend=0.4, 201 cases) or multiple myeloma (ERR=0.003, 95% CI −0.02 to 0.16, p-trend>0.5, 112 cases). Findings did not differ significantly by demographic factors, smoking or specific radiological procedures performed.
Conclusion After follow-up averaging 22 years, there was little evidence of a relationship between occupational radiation exposure and myeloid or lymphoid haematopoietic neoplasms.
- epidemiology
- occupational health practice
- leukaemia
- longitudinal studies
- Ionising radiation
Statistics from Altmetric.com
Key messages
What is already known about this subject?
Despite a well-established link between radiation exposure and leukaemia, few studies have evaluated whether low dose, low-dose rate cumulative occupational radiation exposure influences risks of haematopoietic malignancies in medical radiation workers and there is a lack of data on cumulative occupational radiation dose-response.
What are the new findings?
In a follow-up (average 22 years) of a US nationwide cohort of 110 297 radiologic technologists for mortality during 1983–2012, there was no evidence of cumulative occupational radiation dose-response and elevated risk of myeloid neoplasms including acute myeloid leukaemia, acute myeloid leukaemia combined with myelodysplastic syndromes or chronic myeloid leukaemia or for all leukaemias excluding chronic lymphocytic leukaemia. There were no significant trends in cumulative occupational radiation dose-response and excess risk of lymphoid neoplasms overall.
The findings from the study of US radiologic technologists should be interpreted in light of low average cumulative bone marrow doses and uncertainty of estimated doses prior to the 1970s.
How might this impact on policy or clinical practice in the foreseeable future?
No impact on policy or clinical practice.
Introduction
There were an estimated 7.3 million medical and 0.4 million dental workers monitored for radiation exposure in the last published worldwide survey during 2000–2002.1 Despite a well-established link between radiation exposure and leukaemia, few studies have evaluated whether cumulative occupational radiation exposure influences risks of haematopoietic neoplasms in medical radiation workers.1 2
Understanding of radiation and risk of haematopoietic neoplasms has been based mostly on the Japanese atomic bomb survivors’ mortality3 4 and incidence data.5 The 1945 bombings resulted in acute radiation total body exposures >0–4+ Sv; follow-up of the survivors did not start until 1950.3 Radiation from the bombings has been estimated to account for half of the subsequent incident leukaemias5 vs 4.8% of the incident solid cancers6 in the survivors. Elevated mortality4 and incidence risks5 of acute myeloid (AML), chronic myeloid (CML), acute lymphocytic leukaemias (ALL) and incidence of myelodysplastic syndromes (MDS)7 were observed. Dose-response increases of the three leukaemia subtypes were apparent within a few years of the bombings but declined rapidly over time except for AML.4 Among survivors, there was weak evidence of radiation-related non-Hodgkin lymphoma (NHL) incidence and mortality particularly in males3 5 and multiple myeloma (MM) mortality mostly in females,3 but no increase of Hodgkin lymphoma incidence.5 The reported significant linear dose response for incident chronic lymphocytic leukaemia (CLL)5 is difficult to interpret because the results are based on only 12 cases that incorrectly include 2 hairy cell leukaemias which should have been combined with NHL.
Dose-response studies of protracted external radiation have shown increased mortality and incidence risk of leukaemia excluding CLL following low-dose (mean<100 mSv)8–12 and moderate-dose (100 to <1000 mSv)13–18 cumulative radiation exposures. Results for radiation dose-response and risk of specific myeloid and lymphoid subtypes is more limited and generally less consistent than for leukaemia excluding CLL.8–21
The 146 021 US radiologic technologists (USRT), one of the largest medical radiation worker cohorts studied, have been followed up for cancer and other radiation-related diseases during 1983–2012.22–24 The primary occupational exposures of cohort members are to intermittent X-rays with mean energies of 35–50 keV.25 A comprehensive exposure assessment for the USRT covers the period 1916–1997. Previously, the USRT study team found increased risks of leukaemia excluding CLL among technologists who worked five or more years before the 1950s or reported high-frequency of holding patients for radiological examinations or long duration work history (eg, 30+ years),26 27 but these reports lacked individualised dose information. This report provides the first occupational radiation dose-response estimates for mortality risks from haematopoietic malignancies in medical radiation workers (the USRT) based on reconstructed individual doses28 corroborated with analysis of chromosomal translocations.29
Materials and methods
The study protocol has been approved annually by the institutional review boards of the National Cancer Institute and the University of Minnesota. Completion of the baseline questionnaire was taken as implied consent; the requirement for informed consent was subsequently waived by the review boards because of minimal risk, inability to carry out the research without the waiver, lack of adverse effect of the waiver on subjects’ rights and provision of research findings to subjects from the data collection.
Study population and data sources
The study population and survey methods have been previously described22–24 and is available at http://radtechstudy.nci.nih.gov/. In brief, the USRT cohort was identified in the early 1980s from the American Registry of Radiologic Technologists (ARRT) which was established in 1926. Eligible for inclusion were 146 021 technologists who were certified for two or more years during 1926–1982. The exposure period began in 1916 (earliest year worked) and continued through 1997 (see Dose estimation section below). The ARRT provided updated residential addresses based on annual renewals; inactive members were traced using many sources as described in detail in Ref. 30. Follow-up for incident cancers and other outcomes started in 1983 using mailed questionnaires. Deaths were identified through linkages with state and national databases (including the Social Security Administration death master file and many other sources) and causes of death were obtained through linkage with the National Death index in 1997, 2008 and 2014.
Four questionnaires (baseline administered during 1983–1989 or 1994–1998 and subsequent questionnaires during 2003–2005 and 2012–2014) collected history of specific cancers and other health outcomes, work history, demographic and lifestyle characteristics. Response rates were 68%–72% for the first three surveys and 63% for the fourth survey. Among the 110 373 individuals completing one or more questionnaires, the USRT team excluded those with a previous history of cancer other than non-melanoma skin cancer before the date of entry into follow-up. The final analytic sample comprised 110 297 radiologic technologists (83 655 women, 26 642 men).
Follow-up and ascertainment of deaths from hematopoietic neoplasms, 1983–2012
Cohort members were followed from the baseline questionnaire until the date of death, loss to follow-up or 31 December 2012 (last vital status follow-up), whichever occurred first. Mortality was evaluated since it is available for the entire cohort whereas incidence was restricted to those who completed a baseline and at least one additional questionnaire. Overall, cohort members had worked an average of 22 years. ICD-931 and ICD-1032 codes for cause of death are reported in online supplementary table E1.31 32
Supplemental material
Dose estimation
The dose estimation methods are described in detail in Refs. 25, 28 and 29. and in online supplementary file 1. Briefly, doses were estimated for 110 373 USRT technologists who completed one or more self-administered questionnaires. The reconstruction of individual doses25 28 was based on an internationally recognised strategy for estimating organ doses from external photon radiation exposures.33 In short, the International Commission Radiation Protection uses published factors to convert badge dose to air kerma and then to organ dose.25
Because complete badge dose measurements covered only 35.4% of the occupational radiation exposure of the USRT cohort members and badge measurements were limited before the late 1970s, the dose reconstruction used: (1) individual badge measurements (covering 1960–1997), (2) detailed questionnaire-based work history and radiation protection measures and (3) historical literature on diagnostic radiologic exposures to medical radiation workers (covering 1916–1959). We developed probability density functions of badge dose by decade from which each missing dose was sampled.
The estimated bone marrow dose was individualised to account for body mass index using baseline questionnaire data and coefficients derived from phantoms, with thinner individuals receiving higher and larger individuals receiving lower estimated bone marrow doses (see figure 2 in Ref. 34).
Lead aprons are effective at reducing bone marrow doses between 93% and 99% (table 6 in Ref. 25) depending on the X-ray energy and apron thickness (tables 1 and 6 in Ref. 25). The total bone marrow dose was derived as the weighted sum of lead apron-shielded and unshielded doses, weighted by fraction of the body not exposed and exposed.25 35 Questionnaire-based work history information and probabilistic methods were employed together with literature-derived information to estimate bone marrow dose;28 35 36 when lead aprons were worn, 20% of bone marrow was estimated to be exposed to X-rays.25
US radiologic technologists followed up for mortality, 1983–2012 by mean total cumulative red bone marrow dose and percentage of dose from badge dose readings according to year first worked*
The estimated annual population mean badge doses (derived from table 4 in Ref. 28) decreased substantially from about 1500 mSv in the 1920s to about 5.5 mSv in the 1990s, with a similar decline in estimated bone marrow dose (overall mean: 8.5 mGy; median: 4.7 mGy; range: 0–430 mGy).
The estimated bone marrow doses were corroborated by chromosome aberration analysis using fluorescent in situ hybridisation.37 We found: (i) a significant excess translocation rate with increasing estimated bone marrow dose and (ii) a dose-response similar to results from atomic bomb survivors (see table 6 in Ref. 29). These findings conferred confidence in our methods for estimating bone marrow doses.
Demographic, lifestyle and work history characteristics
Demographic, lifestyle and work history characteristics from the baseline questionnaires were evaluated as potential confounders and/or effect modifiers. The variables examined included year of birth, sex, smoking, year first worked and ever worked with fluoroscopically guided or nuclear medicine procedures.
Statistical analysis
The USRT investigators used Poisson regression to estimate the excess relative risk of haematopoietic neoplasms per 100 milligray (ERR/100 mGy) with 2-year lagged cumulative occupational radiation bone marrow dose in the retrospective cohort follow-up. Due to the small number of deaths, parsimonious models, for example, those with the minimal number of explanatory variables that yield acceptable fit,38 were preferentially chosen to facilitate model convergence.
The basic risk model was:
where baseline rates (
) were described as
where 
is 1 for the indicated sex and 0 otherwise and 
represents other risk factors (such as smoking). For specific outcomes, the quadratic effect in log age, the birth cohort effect or other effect modifiers were omitted when they were not statistically significant. The ERR was described either with a simple linear function of dose, 
, with separate effects in different categories, 
, such as sex or year of initial exposure, or using categorical dose-group effects. Effect modification was described using loglinear models, for example, 
. Trend tests were based on the hypothesis that the dose response parameter was equal to 0. In models with different dose response trends for different subgroups, a test of a common dose response over the subgroups was the test for heterogeneity.
The baseline rate model used for the analysis was the same (unstratified) parametric model used for the simple dose-response modelling. The associations were also separately examined in strata defined by year first worked as a radiologic technologist (before 1950, 1950–1959, 1960–1969, 1970+), sex, year of birth (before 1930, 1930–1939, 1940–1949, 1950–1966), smoking (ever versus never) and in subgroups who reported ever working with fluoroscopically guided or nuclear medicine procedures due to their potentially higher doses in recent decades.39 40 Two-sided hypothesis tests and 95% CIs were based on likelihood ratio tests.41 All analyses were conducted using Epicure (V.2.00.02, Risk Sciences International, Ottawa, Canada).
Hodgkin lymphoma deaths were excluded from detailed consideration because of small numbers, potential misclassification with NHL42 and the limitation of mortality due to the high survival and cure rate. Small numbers precluded assessment of CML and acute lymphocytic leukaemia by year first worked. The USRT team evaluated the combined grouping of MDS with AML due to the substantial potential for misclassification43 44 and also provide results for AML only to enable comparison with historical studies.
Results
Cohort participants were mostly women (77%) and white (95%); close to half were born in 1950 or later and one-third during 1940–1949, with a small proportion (6%) born before 1930 (online supplementary table E2). Half first worked in 1970 or later, 30% during 1960–1969 and 6% before 1950. The mean length of follow-up was 22 years and the median was 27 years. During 1983–2012, there were 133 deaths from myeloid neoplasms (85 AML, 31 MDS, 9 CML and 8 other myeloid neoplasms) and 381 deaths from lymphoid neoplasms (201 from NHL, 32 CLL, 112 MM, 18 Hodgkin lymphomas, 15 acute lymphocytic leukaemias and 3 other lymphoid leukaemias). In addition, there were 15 leukaemias, not otherwise specified, and 155 leukaemias excluding CLL. The mean total cumulative bone marrow dose was estimated to be 8.5 mGy, with a notable decline from a mean of 190 mGy for those first working before 1930 to 1.6 mGy for those first working in 1980 or later (table 1). For the entire cohort, the percentage of estimated doses based on badge dose measurements was 35.4%; badge dose data were very limited before 1960.
Myeloid neoplasms and leukaemia excluding CLL
For the overall dose-response for AML (ERR/100 mGy=0.0002 (95% CI <−0.02 to 0.24, p>0.5)) and for the combined grouping of AML/MDS (ERR/100 mGy=0.015 (95% CI <−0.023 to 0.25, p>0.5)) the USRT investigators found no significant ERR/100 mGy or in any of the year first worked categories (table 2). Findings were similar for AML and for AML/MDS based on models that were unadjusted (data not shown) compared with those adjusted for birth cohort. Based on nine deaths from CML, there was no evidence of an excess risk before 1970 or for 1970 and later or overall evidence of a dose-response relationship (data not shown). There was no evidence for effect modification on dose-response and risk by sex, year of birth, ever versus never smoking or ever versus never worked with fluoroscopically guided or with nuclear medicine procedures (table 3).
ERR of radiation and haematopoietic neoplasm mortality by year first worked, US radiologic technologists, 1983–2012
ERRs of radiation and myeloid neoplasms and leukaemia excluding chronic lymphocytic leukaemia mortality overall and by selected risk factors, US radiologic technologists, 1983–2012
Lymphoid neoplasms
CLL findings are based on a small number of cases. There was no evidence of an overall association (ERR/100 mGy=−0.023 (95% CI <−0.025 to 0.18, p-trend=0.45, 32 cases)). An increased occupational radiation dose response is seen for those who first worked during 1960–1969 but not among those who first worked in earlier decades; only one technologist who developed CLL first worked in 1970 or later (table 2). If the baseline for risk of CLL is not adjusted for birth cohort, the fit is less good, but the findings are quite different with a borderline significant increased ERR in overall dose response (ERR/100 mGy=0.78 (95% CI −0.002 to 13, p=0.07)) (detailed data not shown). Risk of NHL was increased for those who first worked during 1950–1959, but no excess was observed among those who first worked in other time periods nor an overall ERR for occupational radiation dose response and NHL (ERR/100 mGy=0.03 (95% CI <−0.2 to 0.18, p=0.4, 201 cases)) (table 2). There was no evidence of a radiation dose response for MM among those first working in any calendar year period or overall (ERR/100 mGy=0.003 (95% CI −0.02 to 0.16, p>0.5, 112 cases)) (table 2). For acute lymphocytic leukaemia, there was no evidence of an overall dose-response (ERR/100 mGy=0.058 (95% CI <−0.02 to 1.03, p=0.5, 15 cases)) (table 2).
Other haematopoietic neoplasms
Dose-response risks were not increased for the year first worked categories or overall for leukaemia excluding CLL (ERR/100 mGy=0.05 (95% CI <−0.09 to 0.24, p=0.21)) (table 2). There was no effect modification by sex, year of birth, smoking or ever having worked with fluoroscopically guided or with nuclear medicine procedures (table 3).
Discussion
In the first dose-response assessment of cumulative occupational radiation exposure and haematopoietic malignancy mortality risks in US medical radiation workers, there was no evidence of elevated risk of AML, AML combined with MDS or CML, the latter based on small numbers, during 1983–2012 in the USRT. These findings did not differ significantly by sex, year of birth, smoking or history of ever working with fluoroscopically guided or nuclear medicine procedures. There were no overall significant dose-response trends for lymphoid neoplasms, although there were small increases for CLL and NHL in certain decades before 1970. There was no evidence of dose-response for cumulative occupational radiation exposure and risk of leukaemia excluding CLL.
Comparison of haematopoietic neoplasm mortality risks for the USRT with other relevant mortality and incidence studies is shown in table 4; more than one report is summarised if different haematopoietic neoplasms are reported, additional years of follow-up are included or different population subsets are described. No dose-response relationship was apparent for leukaemia excluding CLL in the USRT compared with significant mortality increases in INWORKS8 and non-significant mortality increases in the 15-country study11 and mortality and incidence in the UK radiation workers.9 10 Significant increased dose-response for mortality and incidence of leukaemia excluding CLL was reported in populations with higher doses, for example, Mayak nuclear workers,13 14 Techa River residents15 16 and Chernobyl liquidators.17 18
ERRs per 100 mGy* for haematopoietic neoplasms mortality and incidence among radiation-exposed populations mostly exposed in adulthood except Techa River and Japanese atomic bomb survivors at all ages
Neither occupational radiation dose-response for AML or the combined MDS/AML was increased in the USRT compared with modest non-significant increases of AML mortality in the INWORKS nuclear workers,8 the US pooled nuclear workers19 or the UK radiation workers,9 all populations with bone marrow doses similar to the USRT. Substantially higher mean bone marrow doses may have been the reason for a significantly elevated risk of AML incidence in Mayak nuclear production workers14 and for non-significantly elevated dose-response incidence risks for AML in the Chernobyl liquidators from Belarus, the Russian Federation and Baltic countries.18
No increase of CML mortality was seen based on very small numbers for the USRT in contrast with significantly elevated CML mortality in the INWORKS nuclear workers8 and mortality and incidence in the UK radiation workers.9 10 Non-significant increases for dose-response and CML mortality were described in the 15-country study11 and the US pooled nuclear workers19 and non-significant increases in incidence in the Mayak workers.14
In general, comparison with the exposure response for haematopoietic neoplasm mortality from other studies of adult low-dose protracted radiation exposures (mostly nuclear workers) shows lower ERRs for almost all of these outcomes in the USRT, although there was overlap of CIs of the USRT mortality results for leukaemia excluding CLL with those from other radiation worker studies8 9 12 and the Mayak and Techa populations13 15 and for AML.8 9 11 19 The USRT AML mortality findings (adult exposures) do not overlap with those of the atomic bomb survivors (exposures at all ages).4 The notably increased ERRs for CML in the nuclear workers and atomic bomb survivors and to a lesser extent in the Mayak and Techa River cohorts were not evident in the USRT, although small numbers of CML in the latter precluded precise estimates.
The nuclear and radiation worker cohorts listed in table 4 8 9 11 19 show CIs for dose-response estimates for CLL mortality that overlap with those for the USRT. Statistical power is much lower for the USRT than for INWORKS,8 US pooled nuclear worker19 and UK radiation worker9 cohort studies, with the latter three studies including 2–4 times as many CLL cases as the USRT. Except for the pooled study of mortality in US nuclear workers,12 one of the incidence studies of Chernobyl liquidators18 and mortality in sex-specific subsets of atomic bomb survivors,3 none of the cohorts in table 4 showed radiation dose-response associations with NHL or MM, consistent with the lack of association observed in the USRT (table 4).
USRT study strengths included a large cohort size, and comprehensive individualised historically reconstructed doses25 28 corroborated for a sample of USRT subjects by an observed dose-response for chromosomal translocations.29 Other strengths, compared with most occupational radiation investigations, were incorporation of USRT questionnaire data on smoking, demographic and work history factors in the analyses of myeloid neoplasms.
Limitations included low bone marrow doses, lack of widespread badge dose monitoring before the late 1970s, small numbers of haematopoietic malignancy deaths and lack of detailed haematopoietic malignancy subtype characterisation and potential misclassification. Haematopoietic malignancy deaths included in the dose-response assessment were restricted to those occurring in baseline questionnaire respondents, thus excluding technologists dying of haematopoietic malignancies neoplasms before start of follow-up (a particular shortcoming for myeloid neoplasms due to their short latency)45 or not completing a baseline questionnaire. The historical dose reconstruction had greater uncertainty for earlier workers due to lack of badge doses before 1960. Questionnaire-related limitations included recall problems for lead apron use and other early work practices and lack of detailed data collection about fluoroscopically guided or nuclear medicine procedures. These limitations resulted in low statistical power. Also, residual confounding by unmeasured or unknown risk factors could have obscured a true positive association.
The USRT investigators considered evaluating incidence, which would be preferable to mortality for indolent haematopoietic malignancies (eg, CLL and some types of lymphoma) or for those with long survival (eg, more recently diagnosed CML, MM or AML treated with bone marrow transplant) or a high cure rate (eg, Hodgkin lymphoma). However, incidence data would not capture those cohort members who did not complete one or more follow-up questionnaires (administered roughly 10 years apart). Also, complete incidence ascertainment was not feasible in the absence of a long-standing nationwide US population-based cancer registry. The median cohort age of 63 at the end of follow-up in 2012 and the use of mortality for the endpoints may have precluded identification of haematopoietic malignancies with onset in the elderly, namely CLL and MM.46
In conclusion, the USRT investigators found little evidence of a relationship between estimated occupational bone marrow dose and deaths from myeloid, lymphoid and other haematopoietic malignancies after more than two decades of follow-up. The ongoing effort to establish a nationwide registry of state and other population-based cancer registries in the USA (https://www.naaccr.org/about-vpr-cls/) may provide opportunities to improve ascertainment of haematopoietic cancer incidence in the USRT but will be limited by lack of complete state registry participation, coverage before the mid-to-late 1990s and under-reporting of haematopoietic neoplasms diagnosed in physician’s offices (eg, CLL, MDS and others). Combining USRT data with other medical radiation worker cohorts with individual dose data will also be important given the low doses and rarity of haematopoietic malignancies.
Acknowledgments
The authors are indebted to the radiologic technologists who participated in the study. The USRT investigators thank Dr Jerry Reid of the American Registry of Radiologic Technologists for continued support, Allison Iwan and Diane Kampa of the University of Minnesota for data collection, Jeremy Miller of IMS, Inc. for data management and data file preparation and Ka Lai Lou for assistance with manuscript preparation.
References
Footnotes
Contributors Guarantees of integrity of entire study: all authors. Study conception/design: MSL, MPL, CMK, EKC, DLP. Data acquisition: MMD, BHA. Development of dosimetry system: SLS, MMD, DLP, ML. Data analysis: DLP. Data interpretation: all authors. Manuscript drafting: MSL, MPL, SLS, DLP. Manuscript revisions: all authors. Approval of final revised version: all authors.
Funding Intramural Research Program of the National Institutes of Health, National Cancer Institute, and the US Public Health Service of the Department of Health and Human Services, Bethesda, Maryland. No funding was received from industry.
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement Data are available on reasonable request. Deidentified data are available from the principal investigator of the US radiologic technologists cohort study (Dr Cari Kitahara) on reasonable request.