Occupational radiation exposure and cause-specific mortality in a cohort of Canadian medical workers

Abstract

Purpose

The health effects of high and moderate doses of ionizing radiation are recognized, but the impacts at lower doses are less clear. Medical workers represent the largest occupational group exposed to radiation. Herein, we examine mortality patterns in a Canadian cohort of medical workers, sex differences in risk, and trends in exposure over time.

Methods

Annual whole-body effective doses for 301,740 medical workers between 1951 and 2018 were obtained from the Canadian National Dose Registry. Dose trends were characterised by job class and sex. Underlying cause of death was ascertained using linkage to mortality data through 2020. Standardized mortality ratios (SMRs) were estimated to compared the risk of mortality among 124,180 workers with with a lifetime cumulative dose > 0 mSv to the Canadian general population. Internal cohort comparisons incorporated person-time and events among unexposed workers. Poisson regression was used for internal cohort comparisons to explore healthy worker bias.

Results

The mean annual radiation dose among workers declined from 0.56 mSv during 1948–1977 to 0.09 mSv in 2000–2018. Compared to the general Canadian population, all-cause mortality was lower for both men (SMR = 0.54, 95% CI: 0.52–0.55) and women (SMR = 0.57, 95% CI: 0.55–0.59). No elevated cause-specific mortality risks were observed. Internal cohort analyses suggests that the healthy worker effects may be contributing to lower mortality rates in workers exposed to low doses of radiation.

Conclusions

Consistent with earlier studies, radiation-exposed medical workers have reduced mortality relative to the general population, likely reflecting a healthy worker effect.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00420-026-02211-9.

Background

Ionising radiation is a human carcinogen (UNSCEAR 2008) and has been shown to increase the risk of cardiovascular diseases (Little et al. 2012), ocular disorders such as cataracts and glaucoma (Little et al. 2021), dementia, and Parkinson’s disease (Srivastava et al. 2023). Our understanding of the long-term effects of radiation exposure on health is derived from a limited number of international cohorts of radiation-exposed workers (Little et al. 2024), as well as the Japanese atomic bomb survivors who were exposed to high-dose intensity exposure over a short time interval (Ozasa et al. 2012). However, extrapolating health risks from such studies focusing on high-dose acute radiation exposures to the substantially lower doses (< 100 mSv) commonly encountered in occupational and public settings introduces uncertainties. Given these uncertainties, epidemiological studies of low radiation exposures in large occupational cohorts with lengthy follow-up provide opportunities to directly assess health risks associated with prolonged exposure at lower levels (Wakeford 2009).

Within the expanding global healthcare workforce, the number of medical workers occupationally exposed to ionizing radiation has increased substantially, from an estimated 1.3 million in 1975–1979 to nearly 9 million in 2010–2014 (UNSCEAR 2022). While this group represents only a fraction of the total healthcare workforce, which exceeded 65 million by 2020 (Boniol et al. 2022), the growth in exposed workers reflects increases in the diagnostic and therapeutic use of radiation (Boice Jr et al. 2020). Indeed, the average annual collective effective dose (person-Sv) for workers in the medical sector accounts for about 83% of the total average annual collective effective dose (person-Sv) from human-made sources of radiation in 2010–2014 (UNSCEAR 2022). Considering the adverse health effects of radiation, it is therefore important to regularly monitor those medical workers who are exposed to radiation and evaluate changes in exposure over time to ensure that radiation protection measures remain adequate.

Historical data on occupational radiation exposure among Canadian medical workers are limited in scope and relevance to contemporary conditions. Earlier reports of medical workers also identified from the Canadian National Dose Registry that covered the period from 1951 to 1987 presented only mean annual (zero and non-zero annual doses combined) doses by job class (Zielinski et al. 2009; Zielinski et al. 2005). They provided limited data on non-zero annual doses, an important measure for understanding exposure among workers who received detectable radiation. From 1985 to 2006, Health Canada documented mean annual doses for most job categories, but again provided little detail on non-zero annual doses (Health Canada). More recently, a study provided both mean annual and mean non-zero annual doses from 1998 to 2018, though it focused exclusively on nuclear medicine technologists (Chen et al. 2021). The current study expands on these prior efforts by offering a comprehensive trend analysis of both mean annual and mean non-zero annual doses across multiple medical job classes, spanning a broader timeframe (from as early as 1948 to 2018) and using the most up—to-date data available.

Epidemiological studies comparing mortality among radiation-exposed medical workers to that of the general population have been conducted in several countries, including the United States (Boice Jr et al. 2023a; Mohan et al. 2003), the United Kingdom (Berrington et al. 2001), Japan (Aoyama et al. 1998; Yoshinaga et al. 1999), South Korea (Lee et al. 2018b), Canada (Zielinski et al. 2009; Zielinski et al. 2005), Denmark (Andersson et al. 1991), and more recently, France (Lopes et al. 2023). For the most part, these studies have reported lower rates of all-cause and cancer mortality among medical workers relative to the general population, however, cancer excesses have been reported in some studies (Andersson et al. 1991; Aoyama et al. 1983). While this general pattern of risk likely reflects, in part, the healthy worker effect (Choi 1992), among medical workers this effect is influenced by greater access to healthcare, the adoption healthier lifestyles (Lee et al. 2024), and higher socioeconomic status (Olfson et al. 2024). It is well established that sociodemographic disparities in mortality exist, with more affluent individuals tending to have lower death rates (Franks et al. 2003; Tjepkema et al. 2013). Lower standardized mortality ratios reported in epidemiological studies may reflect the under-ascertainment of deaths through record linkage processes or incomplete follow-up.

Unlike other cohorts occupationally exposed to ionizing radiation, such as nuclear power plant workers or underground miners, a larger proportion of medical workers exposed to radiation are female. These workers therefore provide opportunities to look at sex differences in risk. Previous research suggests that females are at a higher risk of ionizing radiation-induced cancers (e.g. breast, thyroid, lung) than males who receive the same dose (Council et al. 2006). This increased vulnerability to the effects of radiation among females may be due to processes related to hormonal regulation and genetic risks (Schmitz-Feuerhake et al. 2016). Notably, there have been five previous studies that have presented mortality risk estimates separately for male and female medical radiation workers (Boice Jr et al. 2023a; b; Lee et al. 2018b; Lopes et al. 2023; Mohan et al. 2003; Zielinski et al. 2009) compared to the general population. In general, SMRs were similarly reduced among both sexes in Canada (Zielinski et al. 2009), the U.S. (Boice Jr et al. 2023a; b; Mohan et al. 2003) and France (Lopes et al. 2023). The study on the U.S. radiation-exposed medical workers found a considerably greater mortality deficit in men than their female counterparts compared to the general population, corresponding to a ratio of SMRs of 0.79 (95% CI: 0.76–0.82) (Boice et al. 2023b, a). However, the reporting of sex-specific risks has been limited in several cohorts, such as those in South Korea (Lee et al. 2018b) and France (Lopes et al. 2023), due to a small number of cause-specific deaths among female workers.

Despite published findings from several cohorts of medical workers exposed to radiation, few have evaluated differences in mortality across different medical specialty categories. Such analyses can be informative, offering insight into differences in mortality patterns across medical job classes, which may reflect variation in exposure levels and work responsibilities. Boice et al. (Boice Jr et al. 2023a; b) analyzed data from the U.S. medical workers study and noted a similar decrease in the overall death rates for specialists in general radiography (Standardized mortality ratio, SMR = 0.64; p < 0.001), interventional fluoroscopy (SMR = 0.51; p < 0.001), nuclear medicine, and radiation oncology (SMR = 0.45; p < 0.001), compared to general U.S population. Similar deficits were observed for overall or cause-specific mortality in separate studies among radiological technologists in the U.S. (Mohan et al. 2003) and Japan (Yoshinaga et al. 1999), British radiologists (Berrington et al. 2001), and South Korean physicians (Lee et al. 2024). We know of no previous Canadian study that has evaluated differences in mortality among different types of medical workers exposed to ionizing radiation.

Currently, active follow-up of medical radiation worker cohorts is limited to a few countries, specifically Canada (Chen et al. 2021), South Korea (Lee et al. 2015b; Lee et al. 2018b; Lee et al. 2019), France (Lopes et al. 2023) and the U.S. (Linet et al. 2010). In Canada, despite the existence of a well-defined cohort through the National Dose Registry (NDR), the data have been underutilized, primarily because mortality follow-up has not been updated in over two decades. Earlier analyses of dental and medical radiation workers in the NDR by Zielinski et al. (Zielinski et al. 2009, 2005) found lower mortality rates for both cancer and non-cancer causes compared to the general population for the period 1951 to 1987 (Zielinski et al. 2009). A slight excess in esophageal cancer mortality was observed among medical radiation workers (SMR = 1.40; 95% CI: 0.38, 3.62), though this result was based on just three deaths and was not statistically significant. Notably, this period also saw a substantial decline in occupational radiation doses among both dental and medical workers (Zielinski et al. 2009, 2005).

Given this context, the objective of our study was to characterize temporal trends in whole-body external radiation doses across medical occupations and by sex in a cohort of Canadian exposed medical workers. This study also sought to evaluate mortality rates among Canadian medical workers with a positive (> 0 mSv) lifetime cumulative dose of ionizing radiation, compared to the general population, stratified by medical job class and sex. Additionally, internal cohort analyses of all-cause mortality were carried out to partly address potential biases associated with the healthy worker effect (Boice Jr et al. 2023a; b; Choi 2000; Chowdhury et al. 2017). These analyses were enabled by the most recent record linkage of the Canadian National Dose Registry (NDR) cohort with national mortality databases, extending follow-up by approximately 30 years relative to earlier evaluations (Zielinski et al. 2009, 2005).

Methods

Data sources

Canadian national dose registry

The Canadian National Dose Registry (NDR), maintained by the Radiation Protection Bureau of Health Canada, is a centralized database established in 1951 to track occupational radiation exposures across Canada (Chen et al. 2021). It was developed to record doses for all workers potentially exposed to ionizing radiation above regulatory thresholds. For example, dosimetry service providers are to submit monitored doses to the ensure workers do not exceed occupational exposure limits of 50 mSv in one year, and 100 mSv over five years (Government of Canada 1997). The Canadian Nuclear Safety Commission (CNSC), under the authority of the Nuclear Safety and Control Act, is responsible for regulating all aspects of nuclear energy, including the development, production, use, and control of nuclear substances, prescribed equipment, and related information. Dosimetry service providers (DSPs), licensed by the CNSC, are required to submit dose records to the NDR for all CNSC-regulated activities. In addition, DSPs routinely provide dose records for workers in provincially and territorially regulated activities, even in cases where reporting is not explicitly mandated. External radiation dose measurements submitted to the Canadian National Dose Registry (NDR) are typically derived from passive dosimeters, which store a signal generated in response to ionizing radiation exposure (Government of Canada 1997). Among medical radiation workers, commonly used devices include thermoluminescent dosimeters (TLDs) and optically stimulated luminescence dosimeters (OSLDs), which detect photon and beta radiation from diagnostic and therapeutic procedures. After a predefined wear period, dosimeters are returned to licensed dosimetry service providers, where the accumulated signal is processed and converted into standardized operational dose quantities. These dose records are then transmitted to both the individual service users and the NDR (Government of Canada 1997). As of 2018, the NDR contains dose histories for approximately one million individuals across 114 occupational categories, encompassing nuclear industry personnel, uranium miners, medical and dental workers, and various industrial workers in Canada (Chen et al. 2021). The registry captures a broad range of exposure types, including external exposures to gamma rays, x-rays, beta particles, and neutrons, as well as internal exposures to radionuclides such as radon and tritium (Chen et al. 2021).

Canadian vital statistics death (CVSD) database

The Canadian Vital Statistics-Death Database (CVSD) is a comprehensive database that collects vital statistics reports from all Canadian provinces and Territories, including some demographic and medical information. The database includes vital details such as age, sex, marital status, place of residence and death, and the underlying cause of death, which is classified using the International Classification of Diseases (ICD) system. For deaths before 2000, the International Classification of Diseases, Ninth Revision (ICD-9) coding was used to classify the underlying cause of death, whereas the 10th revision was used for deaths after 2000. The reporting of deaths in the CVSD database is virtually complete, as the registration of deaths is a legal requirement in all Canadian provinces and territories (Statistics Canada). The accuracy of reporting various identifying information (e.g. sex, age, year of birth, place of residence) in the CVSD is also high, with only 0.8% of deaths requiring adjustment for errors in identifying information (Andreev and Bourbeau 2025).

Record linkage: ascertainment of mortality

The Canadian National Dose Registry (NDR) was linked with national mortality data from the Canadian Vital Statistics—Death Database (CVSD; 1950–2020) (Statistics Canada) through Statistics Canada’s Social Data Linkage Environment (SDLE). The SDLE facilitates innovative use of administrative datasets by enabling linkages between tax records, mortality data, cancer registries, and census information, enhancing research capabilities across domains (Statistics Canada).

The record linkage, performed by Statistics Canada, employed probabilistic matching methods, relying on personal identifiers captured by the Canadian NDR such as Social Insurance Number (SIN), date of birth, first and last names, place of birth (province, territory, or country), and sex. However, we note that the SIN number was only available from its inception in 1964, and it is not captured in the death or cancer registrations, but is available in the income tax records. Through this process, 883,150 of the 957,585 unique individuals in the NDR were successfully linked to mortality data, with approximately 8% remaining unmatched due to incomplete or missing data required for the linkage. These workers were excluded from the analyses. This updated record linkage extended the follow-up from 1987 in the previous studies (Zielinski et al. 2009, 2005) to 2020 since the last linkage in 1999. It was validated at Statistics Canada as part of the Derived Record Depository (DRD). In addition to mortality linkage, the income tax file, namely, the T1 Personal Master File (T1PMF; 1981–2020) (Statistics Canada), was also linked to the NDR and used to confirm the vital status of individual workers. The workers with deaths reported in (1) both the CVSD and T1PMF (85.1% of 13,980 deaths) and (2) the CVSD only but not in the T1PMF (9.5%) were retained for the all-cause and cause-specific analyses. Deaths that were not reported in the CVSD but in T1PMF (5.4%) were retained for all-cause mortality analysis.

Study population and design

A retrospective cohort of medical workers occupationally exposed to radiation was derived from the NDR cohort linked to mortality records of 883,150 workers in Canada as of December 31, 2018. This cohort of exposed medical workers represents only a subset of the national healthcare workforce: approximately 1.7 million individuals are employed in health occupations in Canada (Bernard and Seddiki 2025), yet on average, about 100,000 exposed workers per year are monitored and contribute dose data to the NDR (Health Canada 2019). Workers who had a recorded monitoring dose between 1951 and 2018 in at least one of the job classes within the medical sector were eligible for inclusion in the cohort. The Job Class Code variable in the NDR dose record file, indicating an employee's job category at a given exposure year, was used to identify the medical workers (Health Canada 2019). The 21 job classes to define a medical worker and their corresponding distribution in the cohort were presented in Table 1.

Table 1.

Demographic and occupational characteristics of the cohort of all medical workers exposed to radiation in NDR from 1951 to 2018 by Sex

Characteristic	Male N = 71,270		Female N = 230,470		Total N = 301,740
	N	%	N	%	N	%
Decade of birth
1896–1909	70	0.1	10	0.0	80	0.0
1910–1919	440	0.6	220	0.1	660	0.2
1920–1929	1910	2.7	1060	0.5	2970	1.0
1930–1939	3780	5.3	3430	1.5	7210	2.4
1940–1949	9370	13.1	14,580	6.3	23,950	7.9
1950–1959	15,170	21.3	40,030	17.4	55,200	18.3
1960–1969	15,900	22.3	50,790	22.0	66,690	22.1
1970–1979	13,020	18.3	49,790	21.6	62,810	20.8
1980–1989	9080	12.7	47,050	20.4	56,130	18.6
1990–2002	2530	3.5	23,510	10.2	26,040	8.6
Decade of first monitoring for occupational radiation
1948–1959	380	0.5	240	0.1	620	0.2
1960–1969	2900	4.1	3520	1.5	6420	2.1
1970–1979	11,760	16.5	25,500	11.1	37,260	12.3
1980–1989	14,780	20.7	40,000	17.4	54,780	18.2
1990–1999	13,570	19.0	46,380	20.1	59,950	19.9
2000–2009	15,750	22.1	64,520	28.0	80,270	26.6
2010–2018	12,130	17.0	50,310	21.8	62,440	20.7
Age at first monitoring
25th percentile	31.2		27.2		28.1
75th percentile	31.4		27.3		28.2
Mean age at first monitoring (SD)	31.3 (9.5))		27.3 (8.9)		28.2 (9.2)
Years monitored for radiation
1	4920	6.9	17,580	7.6	22,500	7.5
2–10	35,980	50.5	134,250	58.3	170,230	56.4
11–20	15,130	21.2	48,450	21.0	63,580	21.1
21–30	8830	12.4	19,100	8.3	27,930	9.3
31–40	5350	7.5	9375	4.1	14,725	4.9
41–55	1060	1.5	1715	0.7	2775	0.9
Mean number of years monitored for radiation (SD)	12.2 (10.8)		9.9 (9.1)		10.5 (9.6)
Years of follow-up
0–10	12,680	17.8	51,030	22.1	63,710	21.1
11–20	16,550	23.2	65,280	28.3	81,830	27.1
21–30	14,620	20.5	47,160	20.5	61,780	20.5
31–40	15,100	21.2	39,960	17.3	55,060	18.2
41–50	9970	14.0	23,680	10.3	33,650	11.2
51–69	2350	3.3	3360	1.5	5710	1.9
Mean duration of follow-up (SD), in years	26.3 (13.9)		23.3 (13.1)		24.0 (13.4)
Dose categories of cumulative effective dose at the end of follow-up, in mSv
0	32,950	46.2	144,610	62.7	177,560	58.8
(0, 5)	31,820	44.6	77,240	33.5	109,060	36.1
[5, 10)	2510	3.5	4220	1.8	6730	2.2
[10, 15)	1080	1.5	1360	0.6	2440	0.8
[15, 20)	650	0.9	780	0.3	1430	0.5
[20–25)	440	0.6	500	0.2	940	0.3
[25–50)	980	1.4	1050	0.5	2030	0.7
[50–75)	390	0.5	360	0.2	750	0.2
[75–100)	170	0.2	180	0.1	350	0.1
[100–473]	280	0.4	170	0.1	450	0.1
Mean cumulative effective dose (SD), in mSv	2.85 (13.6)		1.01 (5.9)		1.44 (8.4)
Mean non-zero cumulative effective dose (SD), in mSv	5.30 (18.2)		2.71 (9.4)		3.51 (12.8)
Job classes*
Chiropractor & Chiropractor Assistant	2460	3.5	1200	0.5	3660	1.2
Dentist	14,100	19.8	6930	3.0	21,030	7.0
Dental Hygienist	880	1.2	32,070	13.9	32,950	10.9
Dental Assistant	550	0.8	43,820	19.0	44,370	14.7
Dental Therapist/Nurse	40	0.1	260	0.1	300	0.1
Gynaecologist	70	0.1	40	0.0	110	0.0
Laboratory Technician	6640	9.3	11,920	5.2	18,560	6.2
Medical Physicist	940	1.3	360	0.2	1300	0.4
Nuclear Medicine Technologist	1770	2.5	3390	1.5	5160	1.7
Nurse	2690	3.8	33,340	14.5	36,030	11.9
Physician	9020	12.7	2790	1.2	11,810	3.9
Radiological Technologist	7840	11.0	28,210	12.2	36,050	11.9
Radiation Therapist	840	1.2	2700	1.2	3540	1.2
Radiologist (Diagnostic)	3520	4.9	1570	0.7	5090	1.7
Radiologist (Therapeutic)	290	0.4	190	0.1	480	0.2
Speech Language Pathologist	20	0.0	330	0.1	350	0.1
Veterinarian	4770	6.7	7900	3.4	12,670	4.2
Veterinary Technician	1320	1.9	21,950	9.5	23,270	7.7
Ward aid/Orderly	2600	3.6	5220	2.3	7820	2.6
Other (Medical)	10,910	15.3	26,280	11.4	37,190	12.3

*Workers were classified according to the job class in which they had the most monitored exposure records over their career in the NDR. In case of a tie, the last recorded job class was considered; SD indicates standard deviation; mSv, millisieverts; NDR, National Dose Registry

Since the NDR database did not include the retirement date of a worker, person-years were accrued from the last monitored job class to the earliest of death or the administrative censoring date (December 31, 2020). Some medical workers had dose records from as early as 1948. We included those dose records to calculate the lifetime cumulative and annual effective dose. A total of 318,880 workers were identified as medical workers in the NDR dose history data from 1948 to 2018. Of them, 210 medical workers were excluded due to missing data on sex or birth year. For dose records, individual doses, recorded at varying intervals from biweekly to annually, were aggregated to determine annual doses for each cohort member. However, monitoring certain medical workers, such as those in the dental sector, is not mandatory but is recommended and varies by province (Ashmore et al. 1998; Zielinski et al. 2005). Consequently, some dose records may be incomplete. Therefore, additional exclusion criteria for missing or invalid (e.g. negative) yearly doses (n = 170) were applied. Individuals with missing (n = 14,260) or high (500 mSv or more, n = 10) lifetime cumulative dose were excluded since the analysis focused on low ionizing radiation exposure doses. Workers alive over 110 years with no death link (n = 30), alive with missing birth year (n = 760), or without information on alive status (n = 1620) were excluded. We also excluded medical workers whose age at first exposure was below 14 or above 80 years, as these records likely indicated errors in birth year or linkage inaccuracies (n = 80). After applying these exclusion criteria, a total of 301,740 medical workers occupationally exposed to radiation were included in the study (Fig. 1). This cohort was used to characterize trends in occupational radiation exposure and included all workers, regardless of whether their lifetime cumulative dose was zero or greater than zero. For the exposure trend analysis by medical job class, the job classes were treated as a time-varying variable, assigned annually based on the medical job held in each specific exposure year. Of the 301,740 workers in the full cohort, 124,180 workers (41.1%) had a positive lifetime cumulative effective dose and comprised the analytical cohort for the external and internal comparisons. For the mortality risk analysis by medical job class, each worker was assigned a single job class based on the one with the highest number of monitored exposure records in the NDR over their career. When multiple job classes had equal numbers of records, we used the most recently recorded job class. The participants were followed from January 1 of the first year of exposure recorded as a medical worker in the NDR between 1951 and 2018 until death or December 31, 2020, whichever occurred first.

Fig. 1.

Derivation of the cohort of radiation-exposed medical workers in Canadian National Dose Registry (NDR)

Dosimetry

Dosimetry information on the individual annual estimates of whole-body dose attributable to the effective dose of external penetrating radiation was acquired for all radiation-exposed medical workers in the study population. These external exposures were predominantly from penetrating gamma radiation affecting the entire body and were submitted to the NDR as effective doses expressed in millisieverts (mSv). The NDR does not maintain records of organ-specific absorbed doses (for example, doses to the brain), and reliable reconstruction of such doses was not feasible because key exposure parameters—such as photon energy spectra and irradiation geometry—were unavailable. The reported whole-body dose values also accounted for contributions from tritium and other less frequently encountered radionuclides. These components were integrated into the annual effective dose calculations by dosimetry service providers in accordance with established monitoring procedures and applicable radiation weighting factors. The annual dose was registered as zero when a recorded dose fell below the detection threshold of the dosimeter used (typically ranging from 0.01 to 0.10 mSv) (Health Canada 2019). The cumulative dose for each cohort member was calculated by summing all individual annual doses throughout the working career. The mean annual effective dose was also calculated based on the actual annual dose recorded in the NDR. In contrast, the non-zero mean annual effective dose was calculated using the actual non-zero annual doses (i.e., annual doses > 0 mSv).

Ascertainment of vital status and cause of death

All-cause mortality, as well as cardiovascular, cancer, dementia and Parkinson’s disease-related mortality, were identified using the ICD-9 and ICD-10 codes for the underlying cause of death (Table S1). Specific cardiovascular and cancer mortality types were likewise defined using the relevant ICD-9 and ICD-10 codes.

Statistical analysis

The key demographic and occupational characteristics of the study population, categorized by the sex of the workers, were described using proportions and the mean for categorical and continuous variables, respectively. The mean annual dose (mSv) and mean non-zero annual dose (mSv) were used to demonstrate the dose trend across various medical professions and both sexes.

The mortality of exposed medical workers with positive lifetime cumulative dose was compared with that of the Canadian general population using the indirect method of standardization method and expressed as the standardized mortality ratio (SMR). The SMR refers to the ratio of observed to expected cause-specific deaths, where ‘expected’ is the number expected to occur if medical workers were subject to the same death rate as the general population. The general population death rates, specified by sex, age (5-year intervals), calendar year, and cause of death, were calculated by dividing the number of deaths in the general Canadian population, as in the CVSD death data, by the total population for the same strata (sex, 5 year age groups, calendar year). All analyses were based on person-years at risk. The DATAB module in Epicure (Hirosoft International LLC, California, USA) (Preston et al. 1993) was utilized to create a person-year table categorized by sex, age (5 year intervals), and calendar year. The associated confidence intervals (CI) for SMRs were calculated assuming the observed deaths followed a Poisson distribution. For job class-specific SMRs, person-years were further stratified by medical job class, and SMRs were calculated only for job categories in which the number of observed deaths was at least 10 for broad categories of causes of death. These included the dental worker (dentist, dental hygienist, assistant, and therapist/nurse), laboratory technician, nurse, nuclear medicine technologist, physician, radiological technologist, and radiation therapist/radiologist (diagnostic/therapeutic).

We also performed an internal cohort comparison among medical workers with positive lifetime cumulative dose using Poisson regression. The outcome of interest of this analysis was all-cause mortality. Overall and sex-stratified estimates of dose effects on mortality were calculated. Person-years of follow-up were cross-classified by sex, attained age (< 40, 40-, 45-, 50-, 55-, 60-, 65-, 70-, ≥ 75 years) and calendar year at risk (< 1970, 1970-, 1980-, 1990-, 2000-, ≥ 2010), duration of monitoring (< 5, 5-, 10-,15-, ≥ 20 years), birth cohort (< 1900, 1900-, 1910-, 1920-, 1930-, 1940-, 1950-, 1960-, 1970-, 1980-, 1990, ≥ 2000), age at first exposure (< 20, 20-, 25-, 35-, ≥ 45 years), time since last exposure (< 10, 10-, 20-, ≥ 30 years), and cumulative external whole-body radiation dose. The dose was considered in categorical (0, > 0–2, 2–< 5, 5—50, 50–< 75 and ≥ 75–< 500 mSv) and quasi-continuous forms. In the quasi-continuous analysis, the cumulative dose was categorized into 9 groups to ensure that death counts remained comparable across each group. A minimum count of 10 deaths in each dose category was ensured to enhance statistical power for the regression analyses. A mean dose, weighted by person-years of follow-up, was calculated for each of these 9 dose categories and modelled as a continuous variable. Cumulative doses were lagged by 5 years to account for a possible latent period between radiation exposure and all-cause mortality.

The effect of cumulative radiation dose on mortality rates within the cohort of medical workers was described using the linear-no-threshold (LNT) assumption that provided an estimate of the excess relative risk (ERR) (Kelly-Reif et al. 2023). Mathematically, this model can be expressed as follows (Lubin et al. 1995):

1

where a, s, t, and D represent attained age, sex, calendar time, and cumulative lagged radiation dose, respectively. The coefficient indicates the ERR per mSv. The background death rate was described as and was adjusted for attained age, sex, and calendar time. Parameter estimates and their corresponding 95% confidence intervals were obtained using the maximum likelihood method with the AMFIT module in Epicure (Preston et al. 1993). (Preston et al. 1993). Sensitivity analyses were conducted by (i) including workers with a lifetime cumulative dose of 0 mSv, (ii) censoring the follow-up at age 90, and (iii) censoring follow-up on December 31, 2019, to minimize potential bias from excess mortality during the COVID-19 pandemic. All statistical analyses were performed using SAS software 9.4 (SAS Institute Inc.).

Ethical approval for this study was granted by Carleton University’s Research Ethics Board (Clearance# 119,942). The analysis was conducted at Statistics Canada’s Research Data Centres (RDC) located at Carleton University (CURDC) in Ottawa, Canada, to adhere to data confidentiality and protection requirements. All results in this study underwent confidentiality vetting rules related to the data and were reviewed and approved for release by a Vetting Analyst at CURDC. The minimum cell size requirement for descriptive statistics (such as proportion and mean) was set at 10, and sample size numbers presented in the tables were rounded to the nearest 10 to adhere to the reporting requirements of the RDC. The SMRs and RRs presented in this paper are based on unrounded data.

Results

Among the 301,740 medical workers in the cohort for the dose trend analysis, majority were female (76.4%) (Table 1). Overall, 32.7% of the cohort members were dental workers (dentists, dental hygienist, assistant, and therapist/nurse), followed by 15% radiology workers (radiological technologists, radiation therapist, and radiologists) and nurses (11.9%). The mean age at the start of the monitoring was 31.3 and 27.3 years for male and female workers, respectively. A higher proportion of males were born before 1950 (21.8% vs. 8.4%), whereas females were predominantly born in the 1970s–2000s (34.5% vs. 52.2%).. Males had longer tenure than their female counterparts. Most females (58.9%) and a substantial portion of males (46.2%) received zero mSv of lifetime cumulative dose. The mean cumulative effective dose was 1.4 mSv, with males exhibiting a notably higher average cumulative dose (p value < 0.001) at the end of follow-up (2.9 mSv) than females (1.0 mSv). The mean non-zero cumulative dose was 5.3 mSv and 2.7 mSv for males and females, respectively (p-value < 0.001). Higher exposures were rare, with only 0.2% of workers exposed to over 100 mSv (100–473 mSv).

Over time, the number of medical workers in Canada has steadily increased, while the average annual effective dose has consistently declined (Fig. 2, Panel A). The mean annual effective dose among all workers in the early period between 1951 and 1977 was 0.56 mSv. The corresponding mean non-zero annual dose from 1951 to 1977 was 2.30 mSv. However, during the last few decades (1990–2018), the mean annual dose dropped to 0.08 mSv (non-zero annual dose, 0.63 mSv), indicating a significant reduction in radiation exposure over time. Both male and female workers show similar declining patterns in exposure from 1978 to 2018, with females consistently receiving slightly lower annual doses than males (Fig. 2, Panel B). Over the 40 years from 1978 to 2018, this decrease in mean annual dose is based on 11% for males and 67% for females.

Fig. 2.

Trend in mean annual effective dose (mSv) among medical workers exposed to radiation in the NDR: Panel A, Number of workers and mean annual dose among all workers in the cohort from 1948 to 2018; Panel B, mean annual dose among all workers in the cohort from 1978 to 2018, by sex; Panels C–D: Yearly mean dose for nuclear medicine technologists C, and radiation technologists D from 1978 to 2018. The mean annual dose before 1978 in Panels B–D was not presented due to a small number of workers (< 10) per calendar year

Analysis of the annual dose trends among nuclear medicine technologists over four decades, from 1978 to 2018, revealed a stable mean annual dose at low levels, with a notable decrease in the past two decades (Fig. 2, Panel C). Radiological technologists, a key occupational group in medical imaging, received a comparatively lower annual dose (non-zero dose < 1 mSv) over the observation period (Fig. 2, Panel D). In general, a decline in the mean annual effective dose has been observed (Table 2) between 1948 and 1979 and 2000–2018, with an approximately 55%–88% reduction across most job classes.

Table 2.

Mean annual effective dose (mSv) for medical workers exposed to radiation by job class and calendar period in the NDR cohort, 1951–2018

Job classes	Calendar period of monitoring
	1948–1979				1980–1999				2000–2018
	Number of all workers exposed to radiation	Number of workers with non-zero dose	Mean annual effective dose, mSv		Number of all workers exposed to radiation	Number of workers with non-zero dose	Mean annual effective dose, mSv		Number of all workers exposed to radiation	Number of workers with non-zero dose	Mean annual effective dose, mSv
			All workers exposed to radiation	Workers with non-zero dose			All workers exposed to radiation	Workers with non-zero dose			All workersexposed to radiation	Workers with non-zero dose
Chiropractor & Chiropractor Assistant	670	310	0.21	0.69	2070	740	0.06	0.61	2550	540	0.03	0.53
Dentist	3880	580	0.05	0.47	12,600	2610	0.02	0.61	16,250	2940	0.01	0.33
Dental Hygienist, Assistant & Therapist/Nurse	4200	410	0.03	0.47	32,100	3030	0.02	0.59	71,070	3710	0.01	0.29
Gynaecologist	20	10	0.24	0.63	140	40	0.18	1.15	30	10	0.03	0.15
Laboratory Technician	2090	750	0.22	0.86	12,370	4030	0.10	0.73	10,790	3200	0.10	0.63
Medical Physicist	130	60	0.29	0.83	710	280	0.19	0.99	1060	390	0.06	0.40
Nurse	3380	1150	0.34	1.27	16,100	5200	0.09	0.64	25,340	9480	0.08	0.43
Nuclear Medicine Technologist	450	320	1.75	2.63	3070	2290	1.48	2.18	4300	3520	1.46	1.87
Physician	1300	640	0.27	0.76	5340	2660	0.19	0.88	9830	4200	0.26	1.03
Radiological Technologist	5370	2900	0.24	0.62	19,930	10,920	0.10	0.59	31,270	17,030	0.10	0.45
Radiation Therapist	230	170	1.04	1.62	2020	1270	0.48	1.15	3420	1880	0.08	0.37
Radiologist-Diagnostic	810	530	0.39	0.79	3160	1850	0.19	0.79	5090	2590	0.24	1.06
Radiologist-Therapeutic	80	50	0.82	1.51	420	200	0.26	0.95	590	230	0.13	0.72
Veterinarian & Veterinary Technician	990	480	0.26	0.74	8330	2620	0.07	0.54	31,460	7160	0.03	0.15
Ward aid/Orderly	1230	460	0.17	0.61	6560	1770	0.08	0.66	3620	970	0.06	1.71
Others*	5120	1040	0.12	0.76	24,950	5930	0.09	0.86	27,800	10,570	0.13	0.55

*Includes speech language pathologists and other medical workers; NDR indicates National Dose Registry

Table 3 displays the all-cause and cause-specific SMRs for the restricted cohort of workers with positive lifetime cumulative dose, compared to the Canadian general population. A total of 124,180 workers were followed for 3,542,480 person-years, and 8,280 deaths were observed during the study period. SMR analysis of all-cause mortality revealed a substantial deficit for male (SMR = 0.53; 95% CI: 0.51, 0.54) and female workers (SMR = 0.57; 95% CI: 0.55, 0.59) compared with the general Canadian population. For specific causes, mortality from cardiovascular disease (CVD) was substantially lower in our cohort compared to the general population, with an SMR of 0.48 (95% CI: 0.45, 0.50) for both sexes. We also note that statistically significant reduced risks were observed for both men and women for each specific cardiovascular disease type (Table 3). Similarly, the SMRs for nearly all cancer-related deaths were less than unity, with statistically significant differences noted for nearly all sites, except for soft tissue cancers, brain cancer, and melanoma. Deaths due to dementia were all significantly low in both sexes. In contrast, a comparable mortality risk was observed for deaths due to Parkinson’s disease (SMR = 0.88; 95% CI: 0.70, 1.09) compared to the general population.

Table 3.

Standardized Mortality Ratio (SMR) for medical workers with positive (> 0 mSv) lifetime cumulative effective dose in the NDR cohort, 1951–2018

Causes of death (ICD-10 codes)	Male		Female		Total
	Observed	SMR (95% CI)	Observed	SMR (95% CI)	Observed	SMR (95% CI)
All CVD (I00-I99)	1220	0.48 (0.45, 0.51)	580	0.47 (0.43, 0.51)	1800	0.48 (0.45, 0.50)
Hypertension (I10-I15)	30	0.43 (0.29, 0.62)	40	0.74 (0.52, 1.02)	70	0.56 (0.43, 0.72)
IHD (I20-I25)	690	0.45 (0.42, 0.49)	220	0.40 (0.35, 0.45)	910	0.44 (0.41, 0.47)
Stroke (I60-I69)	190	0.52 (0.45, 0.60)	150	0.52 (0.44, 0.61)	340	0.52 (0.47, 0.58)
Other CVD (I00-I09, I16-I19, I26-I59, I70-I99)	310	0.54 (0.48, 0.60)	180	0.50 (0.43, 0.58)	490	0.52 (0.48, 0.57)
All Cancer (C00-C97)	1560	0.56 (0.54, 0.59)	1750	0.64 (0.61, 0.67)	3310	0.60 (0.58, 0.62)
Head and neck (C00–C14, C30–C32)	40	0.37 (0.26, 0.50)	20	0.38 (0.21, 0.63)	60	0.37 (0.28, 0.49)
Oesophagus (C15)	40	0.43 (0.31, 0.58)	10	0.44 (0.23, 0.75)	50	0.43 (0.33, 0.56)
Stomach (C16)	50	0.56 (0.42, 0.74)	30	0.51 (0.34, 0.74)	80	0.55 (0.43, 0.68)
Colon excluding rectum (C18, C26.0)	160	0.63 (0.53, 0.73)	140	0.67 (0.57, 0.80)	300	0.65 (0.58, 0.73)
Rectum and rectosigmoid (C19, C20)	40	0.46 (0.33, 0.62)	30	0.55 (0.38, 0.77)	70	0.50 (0.39, 0.62)
Intrahepatic bile ducts (C22.1)	20	0.62 (0.36, 0.99)	20	0.76 (0.49, 1.13)	40	0.69 (0.50, 0.94)
Pancreas (C25)	130	0.81 (0.68, 0.97)	100	0.70 (0.57, 0.85)	230	0.76 (0.66, 0.86)
Lung and bronchus (C34)	280	0.35 (0.31, 0.40)	340	0.49 (0.44, 0.54)	620	0.41 (0.38, 0.45)
Connective and other soft tissue, including heart (C38.0, C47, C49)	10	0.73 (0.40, 1.23)	20	0.97 (0.62, 1.44)	30	0.87 (0.61, 1.19)
Melanoma of the skin (C43)	50	1.12 (0.84, 1.48)	30	0.86 (0.59, 1.21)	80	1.01 (0.80, 1.25)
Breast, female (C50)	–	–	350	0.68 (0.61, 0.76)	–	–
Cervix uteri (C53)	–	–	30	0.50 (0.34, 0.72)	–	–
Uterus (body, NOS) (C54–C55)	–	–	50	0.64 (0.48, 0.85)	–	–
Ovary (C56)	–	–	150	0.94 (0.80, 1.11)	–	–
Prostate (C61)	180	0.79 (0.68, 0.92)	–	–	–	–
Urinary bladder (C67)	60	0.72 (0.55, 0.92)	10	0.41 (0.20, 0.73)	70	0.64 (0.50, 0.81)
Kidney and renal pelvis (C64–C65)	40	0.49 (0.34, 0.67)	30	0.62 (0.40, 0.91)	70	0.53 (0.41, 0.68)
Brain (C71)	100	1.13 (0.92, 1.37)	70	0.73 (0.57, 0.94)	170	0.93 (0.80, 1.08)
Non-Hodgkin Lymphoma (C82–C86)	70	0.68 (0.53, 0.85)	70	0.81 (0.62, 1.03)	140	0.73 (0.62, 0.87)
Multiple Myeloma (C90.0, C90.2, C90.3)	30	0.65 (0.45, 0.91)	30	0.84 (0.57, 1.19)	60	0.73 (0.56, 0.93)
Leukemia (C91–C95, C90.1)	80	0.82 (0.65, 1.02)	60	0.75 (0.57, 0.97)	140	0.79 (0.67, 0.93)
All solid cancers (C00-C75)	1300	0.55 (0.52, 0.58)	1520	0.63 (0.59, 0.66)	2820	0.59 (0.57, 0.61)
All haematological cancers (C81-C96)	200	0.73 (0.63, 0.84)	160	0.81 (0.69, 0.94)	360	0.76 (0.69, 0.84)
Other cancer (C76-C80, C97)	60	0.53 (0.41, 0.68)	60	0.56 (0.43, 0.71)	120	0.54 (0.45, 0.65)
Dementia (F00, F01, F03)	170	0.79 (0.67, 0.92)	110	0.70 (0.58, 0.85)	280	0.75 (0.67, 0.84)
Parkinson’s disease (G20-G21)	60	0.90 (0.69, 1.16)	20	0.81 (0.49, 1.25)	80	0.88 (0.70, 1.09)
All-cause (A00-Y89)	4580	0.53 (0.51, 0.54)	3700	0.57 (0.55, 0.59)	8280	0.54 (0.53, 0.56)

CI indicates Confidence Interval; NDR, National Dose Registry; ICD, International Classification of Diseases; CVD, Cardiovascular Diseases; IHD, Ischemic Heart Disease. Estimates were not presented when the observed number of deaths were 0, less then 10 or not applicable and represented with a ‘dash’

Across all job classes, both male and female medical workers with positive lifetime cumulative dose consistently exhibited lower all-cause, cardiovascular, and cancer mortality than the general Canadian population (Table 4). SMRs for all-cause mortality ranged from 0.45 to 0.59, with minimal variation between sexes within the same job class. While female workers tended to show slightly higher SMRs for cancer in some occupations (e.g., lab technicians and dental workers), the differences were modest and within overlapping confidence intervals. Notably, the all-cancer SMR for male nuclear medicine technologists was 0.94 (95% CI: 0.70–1.25), which was higher than that observed in most other male worker groups, though it remained statistically not significant.

Table 4.

Standardized Mortality Ratio (SMR) for medical workers with positive (> 0 mSv) lifetime cumulative effective dose in the NDR cohort, 1951–2018, classified by job class and sex

Job class & Causes of death	Male		Female		Total
	Observed	SMR (95% CI)	Observed	SMR (95% CI)	Observed	SMR (95% CI)
Dental workers*
All-cause (A00-Y89)	1100	0.51 (0.48, 0.54)	440	0.52 (0.47, 0.57)	1540	0.51 (0.49, 0.54)
All CVD (I00-I99)	300	0.45 (0.40, 0.50)	50	0.35 (0.26, 0.47)	350	0.43 (0.39, 0.48)
All cancer (C00-C97)	390	0.57 (0.51, 0.63)	240	0.65 (0.57, 0.73)	630	0.60 (0.55, 0.64)
Dementia (F00, F01, F03)	40	0.70 (0.50, 0.94)	10	0.70 (0.33, 1.28)	50	0.70 (0.52, 0.92)
Parkinson’s Disease (G20-G21)	20	0.80 (0.45, 1.33)	–	–	–	–
Lab technicians
All-cause (A00-Y89)	230	0.46 (0.40, 0.52)	320	0.55 (0.49, 0.62)	540	0.51 (0.47, 0.55)
All CVD (I00-I99)	50	0.38 (0.28, 0.50)	40	0.38 (0.28, 0.52)	90	0.38 (0.31, 0.47)
All cancer (C00-C97)	70	0.45 (0.35, 0.56)	150	0.62 (0.53, 0.73)	220	0.55 (0.48, 0.63)
Dementia (F00, F01, F03)	–	–	20	1.05 (0.59, 1.73)	–	–
Parkinson’s Disease (G20-G21)	–	–	–	–	–	–
Nurse
All-cause (A00-Y89)	60	0.49 (0.38, 0.63)	940	0.56 (0.53, 0.60)	1000	0.56 (0.52, 0.59)
All CVD (I00-I99)	20	0.44 (0.25, 0.73)	170	0.48 (0.41, 0.56)	190	0.48 (0.41, 0.55)
All cancer (C00-C97)	30	0.64 (0.42, 0.93)	430	0.62 (0.56, 0.68)	460	0.62 (0.57, 0.68)
Dementia (F00, F01, F03)	–	–	30	0.61 (0.42, 0.86)	–	–
Parkinson’s Disease (G20-G21)	–	–	–	–	–	–
Nuclear Medicine Technologist
All-cause (A00-Y89)	90	0.54 (0.43, 0.66)	70	0.47 (0.37, 0.60)	160	0.51 (0.43, 0.59)
All CVD (I00-I99)	20	0.46 (0.28, 0.72)	–	–	–	–
All cancer (C00-C97)	50	0.94 (0.70, 1.25)	40	0.63 (0.45, 0.86)	90	0.77 (0.62, 0.95)
Dementia (F00, F01, F03)	–	–	–	–	–	–
Parkinson’s Disease (G20-G21)	–	–	–	–	–	–
Physician
All-cause (A00-Y89)	690	0.46 (0.42, 0.49)	40	0.41 (0.30, 0.56)	730	0.45 (0.42, 0.49)
All CVD (I00-I99)	190	0.41 (0.35, 0.47)	–	–	–	–
All cancer (C00-C97)	230	0.47 (0.41, 0.54)	20	0.47 (0.28, 0.74)	250	0.47 (0.41, 0.53)
Dementia (F00, F01, F03)	30	0.70 (0.48, 1.00)	–	–	–	–
Parkinson’s Disease (G20-G21)	20	1.43 (0.86, 2.24)	–	–	–	–
Radiological Technologist
All-cause (A00-Y89)	460	0.59 (0.54, 0.65)	970	0.56 (0.53, 0.60)	1430	0.57 (0.54, 0.60)
All CVD (I00-I99)	120	0.57 (0.47, 0.68)	150	0.47 (0.40, 0.55)	270	0.51 (0.45, 0.58)
All cancer (C00-C97)	150	0.61 (0.52, 0.72)	460	0.62 (0.56, 0.68)	610	0.62 (0.57, 0.67)
Dementia (F00, F01, F03)	10	0.80 (0.38, 1.47)	30	0.79 (0.53, 1.15)	40	0.80 (0.56, 1.09)
Parkinson’s Disease (G20-G21)	–	–	–	–	–	–
Radiation Therapist, Radiologist (Diagnostic/ Therapeutic)
All-cause (A00-Y89)	520	0.48 (0.44, 0.52)	120	0.51 (0.42, 0.61)	640	0.48 (0.45, 0.52)
All CVD (I00-I99)	140	0.41 (0.35, 0.49)	20	0.40 (0.23, 0.63)	160	0.41 (0.35, 0.48)
All cancer (C00-C97)	160	0.47 (0.40, 0.55)	50	0.57 (0.43, 0.75)	210	0.49 (0.43, 0.56)
Dementia (F00, F01, F03)	30	1.00 (0.69, 1.40)	–	–	–	–
Parkinson’s Disease (G20-G21)	–	–	–	–	–	–

*Includes dentists, dental assistants, dental hygienists, and dental therapist/nurse; Estimates were not presented when the observed number of deaths were either 0 or less then 10 and represented with a ‘dash’

Findings from the internal cohort analyses for the restricted cohort of workers with positive lifetime cumulative dose are presented in Table 5. Overall, the relative risk of all-cause mortality in the highest exposure category (75- < 500 mSv) was 1.16 (95% CI: 0.96, 1.41) compared to those in the unexposed group (0 mSv). The excess relative risk (ERR) per mSv was 0.001 (95% CI: 0.000, 0.003). When stratified by sex, the risk estimates were moderately higher among males.

Table 5.

Internal cohort dose–response analyses for all-cause mortality over categories of lifetime cumulative whole-body effective dose for medical workers in the NDR cohort, 5-year exposure lag, 1951–2018, by sex

All-cause mortality (ICD 10: A00-Y89)	Person-years	Number of deaths	Relative risk, RR (95% CI)
Both Sexes
Cumulative whole-body effective dose (mSv)
0	4,724,920	5970	Ref
> 0–< 2	1,788,100	4940	0.98 (0.95, 1.02)
2- < 5	355,440	1280	0.93 (0.88, 0.99)
5- < 50	325,360	1590	1.00 (0.94, 1.05)
50- < 75	14,300	90	0.93 (0.75, 1.16)
75- < 500	12,800	110	1.16 (0.96, 1.41)
ERR/mSv	7,220,920	13,980	0.001 (0.000, 0.003)
Male
Cumulative whole-body effective dose (mSv)
0	1,028,680	2300	Ref
> 0–< 2	562,880	2600	1.02 (0.96, 1.08)
2- < 5	123,540	710	0.95 (0.87, 1.03)
5- < 50	133,580	980	1.00 (0.93, 1.08)
50- < 75	7600	70	1.00 (0.79, 1.28)
75- < 500	8420	90	1.22 (0.98, 1.52)
ERR/mSv	1,864,700	6750	0.002 (0.000, 0.003)
Female
Cumulative whole-body effective dose (mSv)
0	3,696,240	3670	Ref
> 0–< 2	1,225,220	2340	0.96 (0.91, 1.01)
2- < 5	231,900	570	0.93 (0.85, 1.02)
5- < 50	191,780	610	1.01 (0.92, 1.10)
50- < 75	6700	20	0.78 (0.49, 1.22)
75- < 500	4380	20	1.01 (0.64, 1.59)
ERR/mSv	5,356,220	7230	0.001 ( −0.002, 0.004)

Background rates were adjusted for age, sex and calendar year. CI indicates Confidence Interval; NDR, National Dose Registry; ICD, International Classification of Diseases; RR, Relative risk; ERR, Excess relative risk; mSv, millisieverts

In the sensitivity analyses, we observed a similar overall reduction in mortality risk (SMR) in the cohort including workers with a zero cumulative dose by the end of follow-up, compared to the primary analytical cohort of workers with positive lifetime cumulative exposure (Supplemental Table S2). We noted no substantial changes in the SMRs when the follow-up was censored at age 90 (data not shown).Changing the censoring date to December 31, 2019 to avoid COVID related mortality in 2020 did not produce any substantial change to our risk estimates (data not shown).

Discussion

In this study, we investigated differences in mortality rates among a cohort of Canadian medical workers in comparison to the general population and described the trend in radiation exposure for both sexes and across different job classes using data from the Canadian National Dose Registry. We observed an overall decline in the mean annual dose and mean non-zero annual dose from 1948 to 2018. While all medical job classes experienced a decline in mean annual effective dose over time, nuclear medicine technologists showed the smallest relative reduction and consistently had the highest annual doses, remaining above 1 mSv throughout the follow-up period. We found a reduced overall mortality risk among exposed medical workers compared to the Canadian general population. Similar deficits were noted for cause-specific mortality among male and female workers and were consistent across job classes.

While the use of medical technologies that use ionizing radiation is rising (Boice Jr et al. 2020), the doses received by medical workers have exhibited a worldwide downward trend (Chartier et al. 2020). In our study, the mean lifetime cumulative effective dose among all exposed workers over the 70-year study period was 1.44 mSv, which is 1.7 times lower than the previously reported lifetime cumulative dose of 2.45 mSv among a cohort of medical workers from 1951 to 1987 (Zielinski et al. 2009, 2005). Although our study included earlier calendar years than the Korean and USRT studies (Villoing et al. 2021), in which higher exposure levels are likely, the overall mean cumulative effective dose in this study is lower than the South Korean medical workers (author’s calculation: mean cumulative badge dose, 7.1 mSv) (Choi et al. 2018) and USRT (median cumulative badge dose, 5.3 mSv) (Villoing et al. 2021). This discrepancy may stem from the differences in the duration of employment (exposure) among the cohorts, differences in technological features across calendar years or radiologic practices (wearing lead apron and dosimeter, position of dosimeter, dosimeter worn under or over the lead apron, etc.) in three healthcare settings. Nonetheless, it is challenging to directly compare the cumulative effective dose from our study with the organ dose estimates reported in Korean (Choi et al. 2018), as well as the Hp(10) dose from the USRT studies (Villoing et al. 2021). This difficulty arises from the lack of information on conversion factors needed to translate our calculated effective dose to the reported operational dose in the USRT study (Villoing et al. 2021) or derive organ specific doses as in the Korean analyses.

Female workers received a lower lifetime cumulative effective dose than male workers. This finding is consistent with a previous study of diagnostic radiation workers in South Korea (Lee et al. 2018a) and Canadian radiation workers (Sont et al. 2001). In general, males consistently received slightly higher doses annually than their female counterparts. In addition, a relatively higher average number of years monitored for radiation among male workers in this study indicates a longer duration of occupational exposure, resulting in a higher lifetime cumulative dose. The higher exposure among male workers across cohort may reflect differential employment in more highly exposed occupations, but may also be influenced by gender-based factors on protection practices and workload. Male workers tend to perform relatively hazardous practices and are exposed to higher radiation (Lee et al. 2015a). Indeed, a South Korean study found that male radiologic technologists tended to perform CT scanning, portable chest radiography, and C-arm radiography procedures (Kim et al. 2018). Procedures such as portable chest radiography and interventional radiography require technologists to hold or be close to patients throughout, exposing them to significantly higher radiation doses. In contrast, mammography was predominantly done by female workers (Kim et al. 2018). Sex differences in the frequency of radiologic activities were reported in studies among medical workers (Heo et al. 2016), dentists (Kim et al. 2016) in Korea and radiologic technologists (Yoshinaga et al. 1998) in Japan. Further evaluation of workload, protection practice, and safety compliance among Canadian male and female medical workers in different job categories would be instrumental in developing an effective, targeted prevention protocol and reducing potential detrimental health effects.

The downward trend in the mean annual and mean non-zero annual effective dose continued to be observed among Canadian medical workers exposed to radiation. Previous studies reported that the mean annual effective dose for medical and dental workers in Canada was 0.36 mSv and 0.045 mSv, respectively, during the period of 1970—1987 (Zielinski et al. 2009, 2005). The mean annual dose continued to decrease, reaching a very low level of 0.10 mSv or less since the 1990s. The mean annual effective dose of 0.08 mSv in our cohort in 2018 is lower than the South Korean (Noh et al. 2023) (approximately 0.4 mSv) and Chinese cohorts (Deng et al. 2020) (0.28 mSv) but comparable to the 0.08 mSv observed in European countries (Platform).) in the same year. The male predominance of medical workers can explain the higher annual effective dose in South Korea (Noh et al. 2023) and China (Deng et al. 2020) compared to Western countries like the U.S.A (Mohan et al. 2003) or Canada (Zielinski et al. 2009, 2005), where more females work in the medical sector.

We observed a decline in the mean annual effective dose across different medical professions. One exception was the nuclear medicine technologists, who had a consistent mean annual effective dose of over 1 mSv from 1948 to 2018. Since 2005, no nuclear medicine technologist in Canada has received annual effective doses exceeding 20 mSv, and over the past three years, no nuclear technologist has received annual effective doses greater than 10 mSv (Chen et al. 2021). In comparison to U.S. radiologic technologists who frequently conducted nuclear procedures, the median annual dose was 1.2 mSv, which aligns closely with our estimated 1.4 mSv for the same period from 1980 to 2015 (Villoing et al. 2021). The high doses among the nuclear medicine technologists were associated with a rapid increase in diagnostic nuclear medicine, in contrast to the declining use of therapeutic procedures (Van Dyke et al. 2016). Frequent performance of procedures like cardiac nuclear medicine and high-energy procedures, such as positron emission tomography (PET), was also noted (Villoing et al. 2021).

Our analysis found lower overall and cause-specific mortality rates than those observed in the Canadian general population, consistent with prior findings (Zielinski et al. 2009, 2005). During the study period between 1951 and 1987, overall and most cause-specific mortalities showed a reduced rate compared to the Canadian general population (Zielinski et al. 2009, 2005). The SMRs for all-cause mortality among the dental and medical workers were 0.53 (90% CI: 0.49–0.57) and 0.53 (90% CI: 0.51–0.55), respectively (Zielinski et al. 2009, 2005). These are consistent with our present findings. The combined SMRs for all cancers were also comparable to those of this study (SMR = 0.60; 95% CI: 0.58, 0.62). Similar to our findings, lower mortality rates have been observed among medical workers in the U.S. (Boice Jr et al. 2023), South Korea (Lee et al. 2024, 2018b), and France (Lopes et al. 2023). Our lower SMR of 0.54 (95% CI: 0.53, 0.56) for all-cause mortality was comparable to that of medical workers in the U.S.A (SMR = 0.60; 95% CI: 0.59, 0.61) (Boice Jr et al. 2023), relatively higher than in South Korea (SMR = 0.45; 95% CI: 0.43, 0.47) (Lee et al. 2018b) and France (SMR = 0.37; 95% CI: 0.35, 0.39) (Lopes et al. 2023). The relatively lower SMRs in the South Korean (Lee et al. 2018b) and French (Lopes et al. 2023) cohorts could be attributed to the younger age of the workers (11% vs. 30% born before 1960 in South Korea and Canada, respectively) and the shorter follow-up period (8 vs. 24 years of follow-up for the French and Canadian cohorts, respectively).

The mortality deficit in our cohort extended across cause-specific deaths, job classes and both sexes. There were fewer deaths observed from all cancers, solid and haematological cancers, in our cohort than expected from the death rates for the general population. These findings are similar to those of other studies in the U.S. (Boice Jr et al. 2023; Mohan et al. 2003), South Korea (Lee et al. 2018b), and France (Lopes et al. 2023). In a previous study in a cohort of medical workers in Canada (Zielinski et al. 2009), oesophageal cancer rates among female medical workers showed a slight, albeit statistically not significant, increase based on only three observed deaths (SMR = 1.40; 90% CI: 0.38, 3.62). This finding was not supported in this study, with an SMR of 0.44 (95% CI: 0.23, 0.75) based on 10 observed cases among female workers. With regards to other non-cancer mortality, including cardiovascular disease, our findings on decreased risk corroborated with those in the US (Boice Jr et al. 2023a; b), South Korea (Lee et al. 2018b) and France (Lopes et al. 2023). The findings on neurogenerative diseases are consistent with those of U.S. (Boice Jr et al. 2023a; b) medical workers. We noted lower SMRs for all-cause and cause-specific mortality across job classes, consistent with previous studies among radiological technologists in the US (Mohan et al. 2003) and Japan (Yoshinaga et al. 1999), British radiologists (Berrington et al. 2001), South Korean physicians (Lee et al. 2024) and the US physicians and technologists (Boice Jr et al. 2023a; b). We observed a similar reduction in mortality rates among male and female workers in our cohort compared to the general population. Sex-specific cancer mortality rates for both males and females were lower than those of the general population, a consistent finding observed in previous studies conducted in Canada (Zielinski et al. 2009, 2005) and the US (Boice Jr et al. 2023a; b). However, it is important to interpret differences in SMRs across various study populations or subgroups (such as sex) carefully, because SMRs are not appropriate for direct comparison between different populations or strata. This is due to factors like variations in age distribution, reference rates, and residual confounding (Rothman et al. 2008). SMRs could also be not comparable due to the possibility of differential mortality ascertainment by strata (e.g., for sex, surname changes among females following marriage, leading to fewer deaths than males).

The typical pattern of lower mortality rates among medical workers compared to the general population has been observed in other studies (Berrington et al. 2001; Boice Jr et al. 2023a, b; Lee et al. 2018b; Lopes et al. 2023; Yoshinaga et al. 1999) and is referred to as the healthy worker effect (Choi 1992). The healthy worker hire effect occurs when relatively healthier individuals are selected during the job recruitment process. The improved health status of medical workers compared to the general population is likely the result of several factors, including better access to medical care, higher socioeconomic status, and healthier lifestyles (e.g., non-smoking, a healthy diet, exercise) (Berrington et al. 2001; Boice Jr et al. 2023a, b; Lee et al. 2018b; Lopes et al. 2023; Yoshinaga et al. 1999). Our internal analysis within the cohort of medical workers did not find reduced mortality risks associated with increased cumulative radiation exposure, indicating a healthy worker effect. The incomplete ascertainment of mortality during record linkage could also explain the lower mortality risk among medical workers. Indeed, our SMR calculation by the decades of the follow-up period showed a substantially low SMR in the earlier decade of 1970–1979 (Supplemental Table S3). This could be due to the potential for the incompleteness of mortality ascertainment during record linkage in the earlier decades of the follow-up period. Furthermore, there was an underrepresentation of the earliest cohort members of NDR, with no deaths occurring between 1951 and 1969, which remains a limitation of this study. Another explanation for such deficits in mortality risk could be the healthy worker survivor effect (Arrighi and Hertz-Picciotto 1994). Employees with higher mortality risks (e.g., poor health conditions, unhealthy lifestyles, such as alcohol consumption) might exit the workforce sooner than their healthier counterparts. However, since employment status evolves over a worker's career and prior radiation exposure could affect current employment status, further evaluation using an appropriate analytical approach (e.g., g-estimation) (Chevrier et al. 2012) would provide more insight into the healthy worker survivor effect.

The major strengths of the present study are that the risk estimates are based on a cohort of a large number of medical workers, with a long observation period of more than 70 years and individual dose monitoring data. Our cohort included a substantial proportion of female workers, compared to other studies of medical workers, therefore enabling us to characterize radiation exposure and cause-specific mortality risks by sex.

Limitations of this study include measurement error in individual radiation dose and the absence of important lifestyle risk factors like smoking and socioeconomic status. Uncertainties in annual external radiation dose estimates cannot be evaluated due to the lack of information on working scenarios that account for the energy of the incident photons, geometric orientation or the use of a lead apron, among many other assessment factors. Additionally, the extent of underestimating the lifetime cumulative external dose remains unknown due to our inability to distinguish between an actual zero dose and a dose below the minimum threshold of 0.01 to 0.1 mSv, which is also reported as 0. In Canada, while recommended, there is no mandatory regulation for monitoring medical workers in the dental field (Ashmore et al. 1998; Zielinski et al. 2005). The lowest average annual dose observed in this study explains the incomplete dose assessment among dental workers and might be underestimated. However, such underestimation was reported to be unlikely to result in an overestimation of risk by more than 15–20% (Shin et al. 2005). Another source of measurement error in dose assessment may be not wearing the badge according to the protocol (e.g., on the truck, below the lead apron) or non-compliance with wearing the dosimeter (Boice Jr et al. 2023a; b). A shorter wearing period of over a year for the same exposure rate as in other professions, such as nuclear power plant workers, may not lead to a dose that exceeds the reporting threshold (Health Canada 2019). As a result, the dose may be reported as zero, leading to an underestimation of the actual dose amount (Health Canada 2019). As explained by Boice et al. (2023b, a) (Boice Jr et al. 2023a; b), physicians could avoid wearing a badge and continue performing as many fluoroscopic procedures as possible to qualify for a specialist status. Such an intentional act could underestimate the lifetime cumulative dose and result in classifying such personnel into low-dose categories or even exclusion from the study (Boice Jr et al. 2023a; b). In this present study, around 4.5% (n = 14,260) of radiation-exposed medical workers were excluded due to missing lifetime cumulative dose. However, while including these workers could potentially increase the power of the study findings, exclusions are less likely to affect the risk estimates significantly (Boice Jr et al. 2023a; b; Zielinski et al. 2005). Incomplete follow-up due to missed record linkages, particularly among older birth cohorts, may have contributed to under-ascertainment of deaths. However, a sensitivity analysis censoring follow-up at age 90 yielded similar SMRs, suggesting minimal impact on mortality estimates. The SMRs of our sensitivity analysis, censoring on December 31, 2019, to minimize the potential impact of COVID-19–related excess mortality, were marginally higher with substantial overlap in confidence intervals, suggesting that the overall conclusions remain robust. Finally, the study is limited by our inability to account for risk factors such as smoking or its surrogate, socioeconomic status (SES) as these data were not available. However, a reanalysis of cancer mortality among Canadian nuclear power plant (NPP) workers suggested that the direction and magnitude of risk estimates are less likely to be affected by residual confounding due to smoking (Zablotska et al. 2014).

In summary, this study demonstrated a decreasing trend in the external whole-body effective dose among Canadian medical workers exposed to radiation over 70 years of follow-up. Maintaining existing occupational radiation protection standards and practices is essential for reducing radiation exposure in this growing workforce. Particular attention should be given to professionals exposed to the highest doses of ionizing radiation, such as nuclear medicine technologists, who may require closer monitoring. Consistent with earlier findings examining patterns of mortality in Canadian medical workers exposed to radiation, this study demonstrates substantially reduced death rates among medical workers compared to the general population. However, our findings should be interpreted cautiously, considering potential biases due to a healthy worker effect. An extension of this study involving organ-specific absorbed dose and cause-specific mortality using internal cohort comparison is underway. Continued efforts are necessary to enhance our understanding of the radiation-induced health effects experienced by Canadian medical workers.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Funding

This project was funded by the Canadian Institutes of Health Research (Funding Application #: 487910).

Data availability

The data were analyzed within the secure computing environment of the Statistics Canada Research Data Centre. The confidential nature of the data do not allow for the data to be shared outside the research data centre network.

Declarations

Conflict of interest

The authors have no financial or non-financial interests that are directly or indirectly related to this work.