Estimation of the historical radiation dose of Korean radiation workers, 1961–1983

Accurate dosimetry plays a central role in measuring the radiation-induced health risk of radiation workers, who are typically exposed to protracted low doses and low rates of ionising radiation. Thus, exposure information, such as individual radiation doses over the entire employment period, is essential to estimate cumulative radiation doses by occupational exposure.

In Korea, worker radiation doses have been reported and managed in the National Dose Registry (NDR) in the Radiation Worker Information Service System, formerly named the Central Registry for Radiation Worker Information. In the NDR, radiation doses for workers are available from 1984; thus, dose information dating before 1984 remains insufficient. The lack of exposure information from before 1984 often makes it challenging to assess the association between radiation doses and health effects. Additionally, the decreasing trend of radiation doses since 1984 suggests that levels of occupational exposure may be higher in the remote past [1]; therefore, because they fail to consider radiation doses before 1984, cumulative radiation doses for workers before 1984 are likely to be underestimated. One way to address this issue of insufficient exposure information is to estimate undocumented past doses by dose reconstruction models developed by known radiation doses (e.g. badge doses) and their related factors to determine exposure levels [2].

Several studies have addressed historical dose reconstruction for the occupational radiation exposure of radiation workers [3–5]. The million workers study [3] reported dose reconstruction approaches to provide realistic estimates of organ-specific radiation dose. The rocketdyne/atomics radiation workers study [4] also performed organ-specific dose reconstruction from lifetime occupational exposure. The US radiation workers study [5] reconstructed doses using a job exposure matrix based on detailed work practices. However, since exposure characteristics differ among countries and occupation types, it is necessary to estimate the historical dose for a target population with consideration for its own characteristics of occupational exposure. Furthermore, previous studies of radiation workers in Japan [6], Canada [7], and Belgium [8] showed an association between radiation dose and lifestyles such as smoking and alcohol consumption, indicating that lifestyle factors can be used as supplementary variables to estimate occupational radiation doses.

In Korea, a radiation worker study was conducted in 2016 to construct a cohort and investigate its baseline characteristics [9, 10]. The study included a nationwide survey to obtain comprehensive individual information, such as demographics, occupational characteristics, and lifestyle factors, which was linked to the NDR for individual radiation doses and the National Cancer Registry [9, 10]. This study aimed to develop a dose reconstruction model using this collected individual information and to evaluate historical radiation doses before 1984.

2.1. Study population and data collection

The study population was derived from the cohort of 20 608 workers in the Korean Radiation Workers Study [9]. To develop a dose reconstruction model for radiation doses before 1984, which are not available in the NDR, data from 410 workers who had radiation doses that had been first documented in 1984 were used with the presumption that they had possibly been employed during or before 1984 (figure 1).

Zoom In Zoom Out Reset image size

Two information sources were used to develop a dose reconstruction model: a nationwide self-administered questionnaire survey of radiation workers in 2016–2017 and radiation doses from the NDR. The survey included information on occupational characteristics, such as work history (e.g. calendar year of hire, employment duration) and work practices (e.g. wearing a dosimeter, separation of the workplace from a radiation source, wearing protective equipment), lifestyle, and demographics. The NDR has individual workers' radiation dose equivalent (Hp [10]) from 1984 to 2016 measured by personal badge dosimeters. In addition to radiation doses, the NDR also contains additional information about the participating workers' sex, age, birth date, facility, and working duration. Radiation doses less than or equal to 0.1 mSv ('below recording level') were considered as zero. The sources of the survey and dosimetry data are described in detail elsewhere [9].

This study was approved by the Institutional Review Board of the Korea Institute of Radiological and Medical Sciences (IRB No. KIRAMS-2016-03-007-009) and all study participants provided informed written consent.

2.2. Data analysis

The demographic and occupational characteristics including mean reported annual dose of the study subjects are summarised in table 2. The majority of study subjects were men, and more than two-thirds of the workers were under age 30 when their radiation doses were first recorded. The mean annual dose (±standard deviation) was 2.0 mSv (±2.4), and higher radiation doses were observed in men. There was a significant difference in annual dose by facilities: radiation doses in industrial radiography facilities (2.8 mSv) and nuclear power plants (2.1 mSv) were higher than in other facilities (0.8 mSv). Figure 2 shows the mean annual reported doses as measured by personal badge dosimeters (i.e. radiation doses from the NDR) for each group gradually decreased over time from 1984 to 2016.

Zoom In Zoom Out Reset image size

Figure 3 compares the annual reported dose of 410 workers from 1984 to 2016 with the annual doses estimated using the reconstruction models in supplementary table 1. The mean estimated dose decreased from 4.67 mSv in 1984 to 0.6 mSv in 2016. The reported doses showed the same trend from 4.51 mSv in 1984 to 0.43 mSv in 2016, and a similar trend was observed for the four subgroups (supplementary figure 2). The mean estimated cumulative dose from 1984 to 2016 was 57.3 (±52.6) mSv, similar to the mean reported cumulative dose of 55.2 (±72.4) mSv.

Zoom In Zoom Out Reset image size

The distribution of annual estimated doses from 1961 to 1983 for the exposure scenario in each group is shown in figure 4, while figure 5 shows the estimated historical reconstruction doses for the 290 workers who were assumed to work before 1984. The mean estimated doses between 1961 and 1983 decreased from 11.08 to 4.82 mSv for exposure scenarios and 11.15–4.88 mSv for those workers.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The reconstruction model was developed to evaluate historical doses using dose-related variables, including age, sex, and employment characteristics, such as facility type and working hours. The reported doses from the NDR and the estimated radiation doses from the reconstruction model from 1984 to 2016 agreed reasonably well with a decreasing trend over the period. This decreasing trend is mainly due to strengthening the management system for radiation protection with the establishment of relevant legislation. Indeed, the dose limit for radiation workers in Korea was lowered from 50 mSv yr⁻¹ to 100 mSv for five consecutive years in 1998 (the revised dose limit was applied from 2003 with a grace period until 2002 and this regulation was elevated from an administrative rule to a Presidential decree.) [19].

The estimated annual historical doses for workers assumed to work before 1984 demonstrated downward trends from 11.15 to 4.88 mSv between 1961 and 1983 based on the dose reconstruction model. This temporal downward trend in the estimated annual doses was consistent with a decline in the average annual dose in US nuclear industry workers (from >10 mSv to <2 mSv) [20], German uranium miners (10–2 mSv) [21], and British nuclear workers (11–5 mSv) [22]. Additionally, it was consistent with the historical dose estimation for medical worker studies [11, 13, 14, 23, 24].

In addition to the reconstruction model-based historical dose estimation, historical doses, which were missing before the initiation of the NDR, may be imputed using the next observation carried backward (NOCB) (i.e. missing doses before 1984 are replaced by doses in 1984), assuming that exposure levels during the period of early nuclear activity before the initiation of the NDR held with the earliest documented radiation doses at the year of the initiation of the NDR. With this approach, the estimated cumulative doses for workers over the period from 1961 to 1983 were lower than those based on the reconstruction model (supplementary table 3). However, given that radiation doses tend to decrease over time in most radiation worker studies [20–22] including this study cohort, the estimated doses with the NOCB approach may be underestimated, particularly for workers who were employed in the earlier remote past.

This study has some limitations. First, the dose-related variables for the reconstruction model were derived from a cross-sectional survey. Therefore, the information we collected from the survey may not capture occupational characteristics for the entire employment period, including uncertainty due to possible changes of job duties during employment. However, considering that the fits of the reconstruction models remain constant for earlier employment years, temporal changes in occupational characteristics may not significantly influence the estimation of historical doses.

Second, there might be other occupational characteristics and/or information on internal exposure that were not included in this study but that could have refined the reconstruction model. Although we categorised workers into facility types, exposure characteristics could vary according to work duties within a single facility. Furthermore, the reconstruction model may not reflect a variability for a certain period (for example, a relatively large decrease of radiation doses in the early 1990s possibly due to the introduction of the International Commission on Radiological Protection 60 recommendation [25] containing reduction of dose limits). Thus, more caution is needed in the application of the reconstruction model when a high variability of radiation doses exists, and more detailed information on specific duties or procedures and internal exposure should be considered for a more sophisticated reconstruction model. Moreover, additional information about types of radiation sources (e.g. neutron, photon) and their fraction is also necessary for estimating organ-absorbed doses for future dose-response analyses.

Third, the reconstruction model was based on radiation doses (i.e. the dose registry) from 1984 to 2016, and the predictive ability of the model for historical radiation doses beyond the available period of the NDR could not be fully verified. Compared with the period-specific doses reported in the United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR) report [26] and the white paper on nuclear safety in Korea [27], our estimates were relatively high for the same period of 1975–1984. The average annual doses for Korean radiation workers from the above reports of which individual doses were not documented in the NDR due to inaccurate or insufficient personal information, were ranged from 3.49 to 4.67 mSv for nuclear power plants, 0.46–0.64 mSv for research facilities, and 5.22–7.43 mSv for industrial facilities, while those estimated doses in this study for the same period were 5.56–6.73 mSv, 1.89–3.27 mSv, and 5.63–9.32 mSv, respectively. This discrepancy might be due to the possibility of low compliance of wearing a personal dosimeter in early workers [28] and relatively low sensitivity for the early type of a personal dosimeter [29], which might result in that those reported doses were likely to be underestimated. However, we cannot rule out possible overestimation of historical doses from the reconstruction models in this study, particularly due to a higher weight of calendar year in estimating radiation doses in the earlier remote past. Furthermore, our estimations may have uncertainties originating from not only a small number of subjects but also from dose measurement sensitivity and accuracy which might be different depending on types of personal dosimeters and their technical specifications. Thus, caution must be exercised in interpreting estimated doses from the reconstruction models.

Lastly, Berkson-type errors might also have occurred as we estimated the pre-1984 period doses from observed determinants of exposure with a prediction model [30]. Thus, when reconstructed doses are used in epidemiologic studies, these errors should be explained by the range of estimated doses that would most likely be measured as individual doses [31].

Despite these limitations, our study has strengths in terms of data reliability. The NDR used in the reconstruction model is a government-run, single registry system, and the regulatory authority has controlled any potential variations in the measurement quality and interpretation algorithms among the different monitoring companies [12]. Additionally, we included various occupational characteristics, which were associated with occupational radiation doses, in the reconstruction models per facility-based subgroup, which could make a more reliable estimation of missing doses.

In summary, as the first study to evaluate historical doses of Korean radiation workers in nuclear-based facilities, we estimated radiation doses in the remote past when information on radiation doses was sparse or missing. Our findings will contribute to considering cumulative radiation doses over the entire employment period of workers and determining further dose-response relationships for radiation-related health effects.

This work was supported by the Korea Institute of Radiological and Medical Sciences, funded by the Nuclear Safety and Security Commission, Republic of Korea [50091-2021].