Mayak worker cohort: Characteristics and key results of epidemiological studies

INTRODUCTION

Hygienic regulation of ionizing radiation is based on understanding its medical consequences. For this reason, since the first years of practical use of ionizing radiation, permissible exposure levels have decreased by more than an order of magnitude: from 500 mSv per year in the 1930s to 20 mSv per year today¹. The primary reason for this gradual reduction in dose limits is related to the stochastic (carcinogenic) nature of the main adverse effects of ionizing radiation, which typically develop following long latency periods. To assess the risks associated with these effects, prolonged (and still ongoing) observation of irradiated populations is required—currently spanning a maximum of 70–75 years. During this period, methods for radiation-epidemiological studies have been developed, and estimates of radiogenic risk have been obtained (through epidemiological and radiobiological
research)².

The selection, quality assessment, and evaluation of scientific research results are conducted by the United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR). Based on continuously updated data on the relationship between cancer incidence, mortality, and ionizing radiation doses, UNSCEAR systematically publishes scientific reports on the levels and consequences of radiation exposure to human health and the environment. These reports are recognized as a reliable and comprehensive source of information by the international community and are widely used for risk assessment and radiation protection measures. Radiation safety recommendations are formulated by the International Commission on Radiological Protection (ICRP). In the USSR and later in the Russian Federation, occupational dose limits for radiation workers have always aligned with ICRP guidelines [1].

The Life Span Study (LSS) of atomic bomb survivors in Hiroshima and Nagasaki (Japan) remains the primary source of quantitative radiogenic risk estimates, due to its large cohort size (over 100,000 subjects) and wide range of radiation doses (up to 4 Gy) [2]. The cohort includes both males and females of various ages at exposure (from children to the elderly), enabling robust population risk assessments³. In its latest Publication 103, providing recommendations for the radiological protection of workers and the public, the ICRP states:

“Risk modeling was based on data from the LSS cohort of Japanese atomic bomb survivors, but epidemiological literature was also reviewed to compare other studies with LSS-derived estimates.”⁴

Thus, from a radiation safety perspective, the scientific community requires validation of LSS findings using data on the effects of occupational exposure among workers in radiation-hazardous industries.

The Mayak Production Association (PA) was the first nuclear industry enterprise in the USSR. The Mayak PA Personnel Registry was created as part of the Epidemiology Department to study long-term stochastic effects of occupational ionizing radiation exposure. Data collection began in the mid-1980s and continues up to the present [3].

The cohort derived from this registry differs from other similar cohorts [4–6], remaining the only global cohort demonstrating statistically significant effects from both alpha radiation (via incorporated plutonium) and external gamma exposure⁵.

In this research, we aim to assess radiogenic risks from prolonged occupational radiation exposure in the Mayak PA worker cohort, including the subcohort employed under normal radiation conditions.

MATERIALS AND METHODS

Inclusion criteria for the study cohort and subgroup stratification

A long-term retrospective epidemiological study on the incidence and mortality from malignant neoplasms (MN) was conducted using the medical-dosimetric registry of Mayak Production Association (PA) workers. Initially, the Mayak PA registry contained information exclusively on personnel working during the 1948–1972 period at three main production facilities (reactors, radiochemical and chemical-metallurgical plants) [7]. Subsequently, the registry was extended to include data on workers hired during the following decade [8], as well as those from two auxiliary facilities, i.e., the water treatment plant and the mechanical repair plant. The registry continues to be updated both by adding newly hired workers at these facilities, currently including individuals employed up to 2016 [3], and by collecting data on employees from other departments. As of today, the Mayak PA medical-dosimetric registry covers the data on workers employed at the main plants and other enterprise divisions in 1948–2016.

The Mayak worker cohort, which is currently available for study, is limited to workers from three main and two auxiliary production facilities hired in 1948–1982. This restriction is related to insufficient and lower-quality dosimetric monitoring of personnel from other Mayak PA departments, particularly regarding internal exposure from incorporated radionuclides.

At the time of commissioning the Mayak PA, knowledge about the effects of radiation on the human body was limited. The delayed manifestation of health consequences also contributed to a lag in implementing more stringent radiation exposure limits. In the USSR, radiation safety standards were based on ICRP recommendations. The authors in [9] provide detailed information on the evolution of dose limits for radiation workers — from initial levels of 0.1 R/day and 30 R/year to the annual limit of 50 mSv recommended by the ICRP⁶ and implemented through Regulation No. 333-60⁷.

The Mayak PA personnel registry initially identified four subcohorts based on the year of employment at the main production facilities: 1948–1953, 1954–1958, 1959–1963, and 1964–1972 [10][11]. Subsequently, the fifth subcohort (1973–1982) and workers from two auxiliary facilities were added [8]. Currently, based on actual radiation exposure levels and assessed health outcomes, two subcohorts have been distinguished:

• the 1948–1958 subcohort includes workers hired during the technology development phase with high occupational radiation exposures;
• the 1959–1982 subcohort includes workers hired during routine operations with exposure levels comparable to modern dose limits [8][12].

All radiogenic risk studies in the Mayak worker cohort have employed methodologies and software tools, particularly the Epicure⁸ statistical software package [13], consistent with those used in both the LSS cohort and other radiation worker cohorts worldwide. Tabulated data are presented with quantitative characteristics including median (M_e), minimum (min), and maximum (max) values.

RESULTS AND DISCUSSION

Cohort size and follow-up period

Table 1 presents the cohort and subcohort sizes along with the distribution of workers by sex, birth year, age at hiring, and employment duration. The cohort comprised 25,755 workers, including 25% females, with a wide range of birth years (1886–1965) and ages at employment initiation (18–69 years). The 1948–1958 hiring subcohort included 13,790 workers (53.5%), while the 1959–1982 subcohort contained 11,966 (46.5%). Due to sufficient availability of male specialists, females constituted only 20.7% in the latter subcohort, compared to 28.2% in the early post-war years. Most workers had already completed their employment at the enterprise — by 2018, 98% of workers had been discharged, including 100% from the first subcohort.

Information on the vital status of cohort members (specifically the year of departure from the city, location, death data) was collected and prepared for use in epidemiological studies through 2018 inclusive (Table 2). The vital status is known for 24,146 individuals (93.8%). Among those with the known vital status, 17,810 persons (73.8%) had died, with 89.0% deceased in the first decade of hire subcohort and 57.1% in the 1959–1982 hire subcohort. The increase in deaths in recent years (2009–2018) was substantial (23.3% of total deaths over the 70-year observation period). Extending the observation period through 31.12.2018 allowed accumulation of over 1 million person-years of follow-up for analysis of radiogenic mortality risk.

Cause-of-death and cancer incidence data

Cause of death was coded according to two International Statistical Classifications of Diseases⁹ and Related Health Problems, 9th and 10th revisions (ICD-9, ICD-10). Both codes are provided for each worker.

For all individuals who died in the city, information on the cause of death was obtained from medical sources or civil registry records. Due to the availability of medical information among those who died in the city, the proportion of unknown causes of death is 1.6% for the entire observation period and 2.7% for 2010–2018.

For individuals who left the city, obtaining information on the cause of death from official sources is currently virtually impossible. However, even before the adoption of the Federal Law “On Personal Data,”¹⁰ this was a challenging task. As a result, among those who left and died before the 2000s, the number of individuals with an unknown cause of death was ≈7%, while later—on average, about 50% (Fig.). Over the past 20 years, the primary source of data on the cause of death has remained personal contact with relatives.

The structure of causes of death differed slightly depending on the hiring period. On average, 47.8% of deaths were due to cardiovascular diseases, 24.3% to malignant neoplasms (MN), and 13.1% to external causes (Table 3).

Unlike mortality data, which were obtained for all members of the study cohort regardless of their place of residence, information on diseases is currently available only for the period when individuals resided in Ozersk. All cases were coded according to ICD-9 and ICD-10. Additionally, the data included morphological diagnoses of MN in accordance with the International Classification of Diseases for Oncology (ICD-O)¹¹.

A total of 4285 malignant neoplasm cases were diagnosed among 3805 workers in 1948–2018. Over the last 19 years of observation (2000–2018), the number of MN cases accounted for 49.2% (2107 cases)—nearly the same as during the previous 52 years (1948–1999; 2178 cases).

Dosimetric data

Dosimetric information represents fundamental data for epidemiological studies of radiogenic risk. Therefore, alongside cohort member identification, continuous updates of individual vital status data, and records of diagnosed diseases, in the 1990s, research began to revise and reconstruct absorbed doses (hereinafter referred to as doses) in specific organs from both external and internal exposure. As a result, five generations of dosimetric systems for external exposure dose assessments were sequentially developed (Doses-1999, Doses-2000, Doses-2005, Doses-2008, and Doses-2013), as well as seven generations for assessments of ²³⁹Pu body content and corresponding internal exposure doses (Doses-1999, Doses-2000, Doses-2005, Doses-2008, Doses-2013, Doses-2016, and Doses-2019) [14–21].

Since the launch of the first industrial reactor at Mayak PA in 1948, the enterprise personnel have been provided with individual dosimeters for measuring gamma radiation doses [14–16]. Starting from 1984, systematic measurements of the neutron dose component have been introduced. Among the study cohort members, dosimetric data on external exposure is available for all 25,755 workers (100%), with 80% of annual dose estimates based on individual dosimeter readings. About 29% of cohort members have at least one annual dose estimated using only indirect data. For 2063 workers (8.0%), the analysis of professional employment records confirmed the absence of occupational external exposure.

The sets of annual external exposure dose values in different generations of dosimetric systems differ primarily in the list of organs for which doses were assessed and the size of the cohort. In 1949–1958, the average annual gamma radiation doses for personnel (Doses-2013, individual dose equivalent — γHp10) exceeded 50 mSv, decreasing to 5–10 mSv in 1968–1989. Since 1990, the average annual dose has not exceeded 5 mSv. Overall, 10,304 individuals (40.1% of the cohort) received low doses. The mean cumulative gamma dose was 748 mGy for the 1948–1958 hire subcohort and 130 mGy for the 1959–1982 subcohort.

Annual gamma doses were estimated through 2007. Due to the cessation of participation of Mayak PA specialists in joint studies, access to external dose data from 2008 onward has been restricted.

An analysis of autopsy materials from cohort workers revealed that internal exposure in the Mayak PA cohort essentially involved dosimetry of inhaled ²³⁹Pu, compared to doses from uranium fission products, which were orders of magnitude lower [17][18]. Estimates of nuclide content and organ/tissue doses are based on urinary ²³⁹Pu activity measurements [19–21]. The latest Dose-2019 system includes dose estimates for 17 organs/tissues and lung compartments for 8395 workers. Cumulative doses varied significantly between primary plutonium deposition organs and systemic pool organs, with maximum values in bone surfaces and minimum values in stomach, intestines, and muscles.

The mean cumulative lung dose was 179.4 mGy (329.2 mGy for the 1948–1958 hiring period; 41.0 mGy for the 1959–1982 hiring period). In the first subcohort, 1394 workers (34.6%) received >100 mGy lung doses, compared to only 9.2% in the second subcohort. Conversely, 264 workers (6.5%) hired before 1959 and 1734 (39.7%) hired later received <5 mGy lung doses. Systemic organ doses were two orders lower: the mean stomach dose was 1.2 mGy, with >5 mGy doses found in 4.7% of examined workers (only 13 in the later subcohort).

Only 32.6% of workers in the study cohort underwent examination. As of 2018, approximately 2000 local residents remained available for testing, including < 200 early hires (first decade). For unexamined workers, doses were estimated using the Job Exposure Matrix (JEM) approach, covering 25,423 workers (98.7%).

The Mayak PA worker cohort remains the world’s primary source on health effects of occupational plutonium exposure. The main stochastic effect of inhaled plutonium compounds is lung cancer. Numerous Mayak studies employing various dosimetric systems, observation periods, and non-radiation factors have established lung cancer dose-response models and statistically significant risk estimates [22–24].

The estimated excess relative radiation risk (ERR) per 1 Gy dose to the lungs was 3.5–8 per 1 Gy for males aged 60 years. No deviations from linear dose-response relationships were found. Radiogenic risk values showed a stronger dependence on smoking status than on gender, although these factors demonstrated moderate correlation (r = 0.61) in the Mayak PA worker cohort. Additionally, the excess risk showed a statistically significant decline with an increase in age.

Studies of the Mayak PA cohort also revealed dose-dependent relationships between alpha radiation dose and MNs in other primary plutonium deposition organs (liver, bones). For liver cancer, a nonlinear dose response was observed, although apparently being driven exclusively by high-dose exposures.

For other solid tumors as well as lymphohematopoietic malignancies, neither incidence nor mortality outcomes showed demonstrable effects from incorporated plutonium exposure levels.

Beyond plutonium-related effects, the Mayak worker cohort has provided estimates of radiogenic cancer risks from external gamma exposure. The principal late effect of gamma radiation in this cohort was leukemia development. The radiation risk for leukemia (excluding chronic lymphocytic leukemia) was approximately 3 per 1 Gy dose to red bone marrow under a linear model [25–27]. However, the data were statistically significantly better described by nonlinear (purely quadratic or linear-quadratic) models incorporating effect modification by:

• age at exposure,
• time since exposure,
• attained age [26, 27].

For solid MN, the coefficient of excess relative risk per Gy (ERR/Gy) from external gamma radiation ranged 0.1–0.4 per Gy across various studies [28–30]. When examining the influence of non-radiation factors (sex, smoking, type of production, attained age, age at hire) as modifiers of radiogenic risk, no statistically significant differences were found.

When developing models to predict MN risk among workers at modern facilities, it is important to consider the significant difference between the current working conditions, including dose loads, and those during the formative period [31–40]. The assessment of radiogenic risk for solid MN incidence (excluding MN in primary plutonium deposition organs) in relation to combined occupational gamma and alpha radiation levels among workers employed under normal radiation conditions (1959–1982 hiring period) revealed an increase in MN incidence at external radiation doses of 0.5–1.0 Gy (relative risk RR = 0.15; 95% CI: -0.21–0.51) and at alpha radiation doses up to 0.005 Gy (RR = 0.30; 95% CI: 0.07–0.53). The linear coefficient of radiation risk for MN incidence (ERR/Gy) depending on gamma radiation dose was statistically significantly different from zero only at the 90% level (0.36; 95% CI: -0.02–0.85; 90% CI: 0.03–0.76) when alpha radiation dose was not accounted for [41]. Estimates of the linear ERR/Gy coefficient for alpha radiation dose were negative¹².

In the study of cancer mortality using a linear dose-response function, the excess risk coefficient was zero for alpha radiation dose and positive, although not statistically significant, for gamma radiation dose (ERR: 0.17/Gy; 95% CI: -0.24–0.68)¹³. When conducting an interval dose estimation, a positive and statistically significant excess risk was observed only in the high-dose range of external radiation above 0.5 Gy (ERR: 0.33/Gy; 95% CI: 0–0.82). When modeling only alpha dose intervals, a statistically significant positive association was found in the dose range up to 0.005 Gy; however, this excess risk was not confirmed when using a model accounting for both radiation types [42].

Thus, among workers employed under normal radiation conditions (1959–1982 hiring period), the attributable risk assessment for MNs (excluding tumors in primary plutonium deposition organs) suggests that only 1–5% of cases can be considered radiation-induced, and solely due to external gamma radiation exposure.

In the analysis of non-cancer mortality rates among workers hired in 1959–1982¹⁴, a comparison of various excess relative risk models based on external radiation exposure levels, both with and without consideration of internal exposure levels, showed no increase in mortality with an increase in radiation exposure. Indeed, no disease category showed a positive estimate of the ERR/Gy coefficient when using a linear dose-response relationship, nor was there a monotonic statistically significant increase in relative risk when using a nonparametric dose-response relationship.

The improvement in data approximation quality when using dose intervals was statistically significant at the 90% level only for the group of infectious and parasitic diseases: however, this was solely due to a positive estimate of excess risk in the dose interval up to 100 mGy (ERR = 0.6; 90% CI: 0.04–1.58). For the most representative category of circulatory system diseases, no dose-effect relationship was observed as well, with the only positive estimate of excess risk obtained for doses exceeding 0.5 Gy (ERR = 0.05; p > 0.5).

CONCLUSION

The Mayak Production Association Personnel Registry constitutes an authoritative source for epidemiological assessments of radiogenic risks associated with prolonged occupational radiation exposure at nuclear industrial facilities. Based on the worker cohort hired in 1948–1982, direct estimates of carcinogenic risk have been obtained for both external radiation doses and ²³⁹Pu intake. The observation of workers who began employment during 1959–1982 serves dual purposes. On the one hand, this allows the magnitude of dose-dependent carcinogenic risk from cumulative gamma radiation exposure to be assessed. On the other hand, this work highlights the need to extend both the observation period and the cohort itself to include personnel working under exposure conditions comparable to contemporary standards.