Doses for post-Chernobyl epidemiological studies: are they reliable?

The two screening cohorts consist of 11 732 and 13 204 individuals in Belarus and Ukraine, respectively, selected from people who met the following criteria: (1) they resided in April–May 1986 in the most contaminated Gomel and Mogilev Oblasts and Minsk-city in Belarus; and in the northern part of one of the three most contaminated oblasts of Ukraine: Kyiv, Zhytomyr or Chernihiv (see figure 1); (2) their age was 0–18 y at the time of the Chernobyl accident; and (3) they were subjected to direct thyroid measurements performed between 26 April and 30 June 1986. These unique cohorts provide the basis for the estimates of radiation-related risk of thyroid cancer and other thyroid diseases in the exposed population (Tronko et al 2006, Brenner et al 2011, Zablotska et al 2011, Ostroumova et al 2013).

Zoom In Zoom Out Reset image size

Essentially, the same approach was used to calculate thyroid doses due to ¹³¹I intake for members of Belarusian and Ukrainian cohorts. The methodology is shown in figure 2 and is described in detail elsewhere (Likhtarov et al 2014, Drozdovitch et al 2015). In brief, information on residential history and dietary habits collected during personal interviews (box 1 in figure 2) was combined with data on ¹³¹I deposition densities at places of residence (box 2 in figure 2) and ecological and biokinetic models (boxes 3 and 4 in figure 2) to calculate the 'ecological' ¹³¹I activity in the thyroid, A_ecol(k,T), at the time of the direct thyroid measurement, T (box 5 in figure 2) and the time integral of the ¹³¹I activity in the thyroid AI_ecol(k) (box 6 in figure 2). The 'ecological dose' D_ecol(k) (box 7 in figure 2) was then derived from AI_ecol(k) using the thyroid mass M_th(k) of the subject under consideration. The 'instrumental' ¹³¹I activity in the thyroid A_inst(k,T), which was derived from the individual direct thyroid measurement done at time T (box 9 in figure 2), was used to calibrate the 'ecological' ¹³¹I activity in the thyroid, A_ecol(k,T), at the time of the direct thyroid measurement (box 5 in figure 2). The 'instrumental thyroid dose', D_inst(k), which is the dose assigned to each cohort member for epidemiological analysis (box 8 in figure 2), was obtained as the product of the ecological dose and of the 'scaling factor', SF(k), which is the term used to refer to the ratio of the instrumental and ecological ¹³¹I activities in the thyroid at the time of the direct thyroid measurement, A_inst(k,T)/A_ecol(k,T), (box 10 in figure 2). In summary, the instrumental dose for subject k is calculated as follows:

$$\begin{eqnarray}&&{{D}_{\text{inst}}}(k)=K\times \left({{E}_{\text{th}}}/{{M}_{\text{th}}}(k)\right)\times A{{I}_{\text{ecol}}}(k)\times \left({{A}_{\text{inst}}}(k,T)/{{A}_{\text{ecol}}}(k,T)\right)={{D}_{\text{ecol}}}(k)\times \text{SF}(k),\end{eqnarray} \tag{ 1 }$$

where D_inst(k) and D_ecol(k) are values of instrumental and ecological thyroid-absorbed doses (mGy); K is a unit conversion coefficient equal to 13.82 (Bq kBq⁻¹ g kg⁻¹ J MeV⁻¹ s d⁻¹ mGy Gy⁻¹); E_th is the energy absorbed in the thyroid per decay of ¹³¹I (MeV); M_th(k) is the thyroid mass of the subject under consideration (g); AI_ecol(k) is the ecological time-integrated ¹³¹I activity in the thyroid (kBq d); A_inst(k,T) and A_ecol(k,T) are instrumental and ecological ¹³¹I activity in the thyroid, respectively, at the time of the direct thyroid measurement, T (kBq).

Zoom In Zoom Out Reset image size

The four most important components of the thyroid dose estimation procedure are: (1) the direct thyroid measurements; (2) the measurements of ¹³⁷Cs (and ¹³¹I) ground deposition densities; (3) the administration of personal interviews to all cohort members; and (4) the measurements of thyroid volumes among the populations of the contaminated areas.

2.1.1.1. Direct thyroid measurements.

2.1.1.2. Ground deposition densities of ¹³⁷Cs and ¹³¹I.

2.1.1.3. Personal interviews.

2.1.1.4. Thyroid mass.

The present state-of-the-art in dosimetry is to document the sources and quality of all input data, to establish a dosimetry error structure, to characterize each parameter used in dose calculations as a source of shared or unshared error, and to generate multiple sets of dose estimates for the entire study population. These multiple sets of cohort doses instead of a mean dose for each individual are being used to perform the dose-response analyses to estimate true risk (e.g. Li et al 2007, Stayner et al 2007, Little et al 2014, 2015, Land et al 2015, Stram et al 2015).

A Monte Carlo simulation procedure was used to estimate the uncertainties in instrumental thyroid doses received by the Belarusian (Drozdovitch et al 2015) and Ukrainian (Likhtarov et al 2014) cohort members. According to this procedure, 1000 sets of cohort instrumental thyroid doses, taking into account the classification of errors as shared or unshared, were calculated. This procedure is similar and generally consistent with the 2D Monte Carlo (2DMC) method (Simon et al 2015). For a specific dose realization, some of the model parameter values were common among the members of subgroups, i.e. shared among subjects of those groups, implying that any error made on this parameter was shared by all subjects to whom it applied. Subject-independent, or shared, parameters included parameters of the ecological model describing the temporal variation of ¹³¹I contamination in the ground, air, and foodstuff (box A in figure 2). Other uncertainties were considered to be subject-dependent, or unshared, including errors related to the ¹³¹I activity in the thyroid derived from the direct thyroid measurements (box B in figure 2), errors in assigning of thyroid-mass value to the study subject (box C in figure 2), biokinetic models of iodine in the human body (box D in figure 2), and the uncertainties attached to the imprecise answers on relocation history and individual diet from personal interviews (box E in figure 2).

Figure 3 shows the general scheme of calculation of 1000 sets of cohort doses taking into account shared and unshared errors. At the beginning of the calculation of each dose set for the entire cohort, values for all shared parameters were assigned. The same value for each shared parameter was used to calculate one dose set for all cohort members for whom this parameter was considered to be shared. This step introduced correlations in each cohort dose set between individual dose estimates of the study subjects who shared parameter values. In the process of dose-set simulation, the values of unshared parameters for each cohort member were sampled from their distributions and calculated, one dose realization for cohort member k, D_i,k. The set of doses from D_i,1 to D_i,N represents a set number i of cohort thyroid doses for N cohort subjects. The thousand realizations of dose across the cohort dose sets for cohort member k, represent the individual stochastic thyroid doses of that cohort member.

Zoom In Zoom Out Reset image size

Table 1 shows the distribution of the arithmetic means of individual stochastic thyroid doses calculated for the Belarusian and Ukrainian study subjects. The global arithmetic mean of individual mean thyroid doses for the entire cohort was 0.68 Gy in Belarus and 0.65 Gy in Ukraine. The highest individual arithmetic mean of the individual stochastic doses among the study subjects was 39 Gy in Belarus and 42 Gy in Ukraine.

The fitted distribution of the individual stochastic doses for each subject was found to be approximately lognormal and the geometric standard deviation (GSD) of that distribution was used to characterize the overall uncertainty for each individual. The GSDs varied from 1.3–5.1 with an arithmetic mean of 1.8 and a geometric mean of 1.7 over all subjects of the Belarusian cohort, and from 1.3–10.6 with an arithmetic mean of 1.6 and a geometric mean of 1.5 for the subjects of the Ukrainian cohort. In general, lower uncertainty in doses estimated in the Ukrainian cohort compared to the Belarusian cohort was mainly a result of higher quality estimates of ¹³¹I activity in the thyroid in the Ukrainian cohort and due to the uncertainty of the assigned dates of some thyroid activity measurements in Belarus. Table 2 shows the distribution of the GSDs attached to the individual stochastic thyroid doses for cohort members. The largest GSDs were associated with small ¹³¹I activity in the thyroid and were due to measurement uncertainties attached mainly to the calculated lower limit of the exposure rate in Belarus (Drozdovitch et al 2013b) and to measurement uncertainties associated with very small doses in Ukraine (Likhtarov et al 2014).

The sensitivity analysis performed for the Belarusian cohort shows that sources of unshared errors associated with estimates of ¹³¹I activity in the thyroid and thyroid mass are the major contributors to overall uncertainty in the instrumental thyroid dose estimates. Shared errors associated with parameters of the ecological model were found to be small contributors to the overall uncertainty in the instrumental doses. Little contribution of shared errors was caused by the nature of the instrumental dose calculation. Indeed, to calculate the instrumental thyroid dose, the ecological dose was calibrated by the scaling factor that is the ratio of the measured to ecological¹³¹I activity in the thyroid at the time of measurement (see figure 2). Calibration of the ecological dose by the scaling factor eliminated the shared errors associated with the parameters of the ecological model, as the same values of shared parameters of the ecological model were used during the calculation of the given cohort dose set to estimate both the ecological dose and the ecological activity of ¹³¹I in the thyroid. However, the component of uncertainty in the cohort doses caused by shared errors was found to be significant for the ecological doses. The left panel of figure 4 shows the cumulative percentage distribution of 1000 ecological doses for the Belarusian cohort. Wide distribution indicates that the sources of shared errors are important contributors to the uncertainty of the ecological doses. In contrast, the shared uncertainties play a minor role in the overall uncertainty in the instrumental thyroid dose estimates (right panel of figure 4).

Zoom In Zoom Out Reset image size

As described above, to estimate realistic (instrumental) doses, the ecological dose was calibrated using a scaling factor that is the ratio of the instrumental to ecological ¹³¹I activities in the thyroid at the time of the direct thyroid measurement. The scaling factor is an indicator of the agreement between the dose estimated using the model and questionnaire data and the dose derived from direct thyroid measurement. The closer the scaling factor is to one, the closer the ecological dose is to the instrumental dose.

A wide inter-individual variability of the scaling factors was observed both in the Belarusian and Ukrainian cohorts. Neither the dates of the direct thyroid measurement and relocation, age of the study subjects at the time of the accident nor the occurrence of stable iodine administration were correlated with the scaling factors. The median values of the scaling factors were 0.81 in Ukraine and 0.35 in Belarus, thus indicating a bias in the results of either the direct thyroid measurements or the ecological model used in Belarus. Although the observed bias does not influence the estimated instrumental dose, this was eliminated in the Belarusian cohort by multiplying the scaling factors by 2.9. Figure 5 shows the distribution of the corrected scaling factors in the Belarusian cohort. The observed wide distribution can be partitioned into 3 categories:

• For about 86% of the study subjects, the corrected scaling factors were in the range from 0.1–10. This range was expected, given (1) the uncertainties related to the derivation of the ¹³¹I activity in the thyroid from the direct thyroid measurement, and (2) the uncertainties in the parameter values used in the ecological model.
• For about 13% of the study subjects, the corrected scaling factors were in the intermediate ranges (i.e. from 10–100 and from 0.01–0.1). Although the main reasons for these intermediate values are usually not clear, it is believed that they are due to a combination of factors, including biases in parameter values, complex residential histories, and difficulties in processing the direct thyroid measurements. In those cases, the uncertainties are somewhat underestimated.
• For about 1% of the study subjects, the corrected scaling factors were found to be very high (i.e. greater than 100) or very low (i.e. less than 0.01), and biases in the parameter values of the ecological model are not sufficient to explain these extreme values. An analysis conducted by Drozdovitch et al (2013b) showed that the possible reasons for obtaining the very low or the very large scaling factors appear to be: (a) imprecise and incorrect answers provided during personal interviews, (b) an assignment of direct thyroid measurement to a wrong subject, and/or (c) clerical errors in the recording of the direct thyroid measurement results. It is acknowledged that in those cases the uncertainty associated with the instrumental thyroid dose value may have been substantially underestimated. However, to assess the doses to all the study subjects in the same manner, it was assumed that the direct thyroid measurements were correctly performed and processed.

Zoom In Zoom Out Reset image size

An evaluation of the importance of imprecise and incorrect answers provided during personal interviews was performed following analysis of the responses obtained during the first and the second personal interviews. These questionnaire responses were used to calculate two sets of ecological and two sets of instrumental thyroid doses due to ¹³¹I intake for each study subject. Figure 6 compares the individual ecological (left panel) and instrumental (right panel) thyroid doses calculated using information from the first and second interviews. As can be seen from figure 6(a), the ecological thyroid doses estimated for the same person using the results of different interviews were spread over five orders of magnitude. The ecological dose is directly proportional to the ¹³¹I deposition density in the settlement of residence, which together with the consumption pattern of contaminated foodstuffs determines the intake of ¹³¹I and, consequently, the ecological thyroid dose received by the study subject. Therefore, the low degree of agreement in the individual behavior and dietary data reported during two interviews led to the low degree of agreement between the ecological doses that were calculated using this information for each individual.

Zoom In Zoom Out Reset image size

Essentially, better agreement was found for the instrumental doses (figure 6(b)), which were obtained by calibrating the ecological doses based on the result of direct thyroid measurement. The mean ratio of the instrumental dose calculated using data from the first interview to that calculated using data from the second interview for the entire cohort was 1.1  ±  0.7 while the median of the ratio was 1.0. It should be noted that although two sets of instrumental doses calculated using the two different interview data are, for the majority of compared doses, rather consistent, a substantial difference (<0.1 or  >10 times) between the instrumental doses is observed for 23 out of 14 982 (0.15%) of the compared dose pairs (figure 6(b)). Further work is underway to understand the reasons for such discrepancies in the instrumental doses. However, it is important to conclude that, if the measurements are available for the study subjects, the quality of the individual behavioral and dietary data has, in general, a small influence on the quality of the retrospective exposure assessment (Drozdovitch et al 2016).

A population-based case-control study of thyroid cancer was carried out in the contaminated regions of Belarus (Gomel and Mogilev Oblasts) and Russia (Bryansk, Kaluga, Orel and Tula Oblasts) among people exposed to Chernobyl fallout in childhood and adolescence. The study included 1218 people from Belarus and 397 people from Russia who were 18 years old or younger at the time of the accident (Cardis et al 2005).

A working group, composed of dosimetrists and epidemiologists from Belarus, Russia, USA, France, Germany and Japan, was established to develop a unified approach to reconstruct individual thyroid doses due to ¹³¹I intake in the study subjects. The development of a common approach included four steps: (a) critical review of the four dosimetry models available for dose reconstruction in Belarus and Russia; (b) validation of the models by comparison of the modeled average age-specific thyroid doses for a settlement with average age-specific thyroid doses for a settlement derived from direct thyroid measurements; (c) validation of the two best performing models by comparison of the modeled individual thyroid doses accounting for the information obtained from personal interviews with individual thyroid doses derived from direct thyroid measurements; and (d) critical analysis of discrepancies observed in model predictions, and the selection of the best model for thyroid dose reconstruction (Drozdovitch et al 2010). The same steps were implemented for the models to calculate doses due to external irradiation and ingestion of Cs isotopes.

As a result of the validation study, a semi-empirical approach was selected to estimate the individual thyroid doses from ¹³¹I intake for the subjects of this case-control study. This approach was based on the relation between environmental contamination (¹³⁷Cs or ¹³¹I deposition density) and the thyroid dose estimated from direct thyroid measurements carried out in population groups of various ages (infants, children, adolescents and adults) in the territories of Belarus and Russia with different contamination levels. Table 1 shows the distribution of the individual modeled thyroid doses due to ¹³¹I intake for the study subjects. The arithmetic mean of the individual thyroid doses for the entire study was 0.49 Gy. The highest individual dose due to ¹³¹I intake among the study subjects was 9.5 Gy (Drozdovitch et al 2010).

The validity of the modeled thyroid doses was confirmed by comparison with the instrumental doses. Fortunately, direct thyroid measurements were available for 63 study subjects from Belarus and 18 from Russia. A rather wide range of ratios between those two sets of individual doses was observed. The mean of the ratios of the modeled thyroid dose to the instrumental dose was 2.0  ±  2.2 and the median of the ratios was 1.2.

A Monte Carlo simulation procedure was used to calculate a set of 1,000 cohort (study population) thyroid doses from ¹³¹I intake accounting for shared and unshared errors in the dosimetry model (Drozdovitch et al 2007). In general, the approach used in this study was similar to that described in section 2.1.2. The GSD of the individual stochastic doses varied from 1.7–4.0 with median values of 2.3 and 2.1 for the Belarusian and Russian study subjects, respectively. The parameters of the semi-empirical model (shared errors) and thyroid dose per unit of ¹³¹I intake (unshared error) were identified to be the most important contributors to the overall uncertainty. The sources of shared errors were found to be relatively important contributors to the uncertainty of the modeled doses calculated in this study.

A population-based case-control study including 66 cases of thyroid cancer (all aged less than 20 years old at the time of the accident) and 132 controls was carried out in Bryansk Oblast, which was the most contaminated region in Russia (Kopecky et al 2006). To reconstruct the thyroid doses due to ¹³¹I intake, a semi-empirical model was used. It was based on the relation between the environmental contamination (¹³⁷Cs deposition density and the ratio of ¹³¹I-to-¹³⁷Cs activity in deposition) and thyroid dose for adults derived from direct thyroid measurements carried out in 94 settlements in Belarus. The semi-empirical model was tested for application in Russia using thyroid dose estimates derived from direct thyroid measurements made in the Kaluga and Bryansk Oblasts (Stepanenko et al 2004).

Information about individual history of relocations, milk consumption and origin, and stable iodine administration was obtained during personal interviews of the study subjects, or their parents for subjects who were very young at the time of the accident. Dose adjustment factors were introduced to adjust the settlement average thyroid dose for adults to an individual dose, taking into account personal interview data.

The validity of the individual thyroid dose estimates in this study was confirmed by comparison with the instrumental doses for 19 people who were not study subjects, but resided in the study area and had reliable direct thyroid measurements. The mean ratio of the thyroid dose estimated using the semi-empirical model to that estimated using the direct thyroid measurement was 1.2  ±  0.99 and the GM of the ratio was 0.77; the distribution of the ratios was approximately lognormal with GSD  =  2.9 (Stepanenko et al 2004).

The uncertainty in individual thyroid doses was estimated using a Monte Carlo calculation procedure. In this procedure, the values for all parameters used in the thyroid dose calculation were sampled repeatedly from their distributions (10 000 times each) to compute 10 000 estimates of the thyroid dose. In the implementation of stochastic dose calculations all sources of errors were treated as unshared. As in the IARC-coordinated study (section 2.2.1), the parameters of the semi-empirical model and thyroid dose per unit of ¹³¹I intake were found to be the most important contributors to overall uncertainty (Stepanenko et al 2004). The estimated GSDs of the thyroid doses varied among the study subjects from 1.8–3.5 and the median GSD was 2.2 (Kopecky et al 2006).

A population-based case-control study of thyroid cancer was carried out in the contaminated regions of Belarus by the Research Institute of Radiation Medicine (Minsk, Belarus) and NCI (USA). This study was the first analytical epidemiology study conducted on the population exposed after Chernobyl. A group of 107 cases, who were children at the time of the accident and who were diagnosed with thyroid cancer before 1 March 1992, and a group of 214 controls matched with the cases by sex and year of birth, and, to some extent, location, were included in the study (Astakhova et al 1998).

Children included in the case-control study were divided into three basic groups according to the type of radiation information that was available for the thyroid dose assessment (Gavrilin et al 2004):

• Children with direct thyroid measurements made in May–June 1986. The estimates of thyroid doses that were derived from these direct thyroid measurements were called 'measured' doses;
• Children without direct thyroid measurements, but living in those settlements where a reasonably large number of inhabitants had direct thyroid measurements. The age distribution of the average thyroid doses can be estimated for the children of those settlements with relatively good accuracy; the thyroid doses estimated using this method were called 'passport' doses; and
• Children without direct thyroid measurements and living in those settlements where such measurements were not carried out. To assess thyroid doses for these children, a dosimetric model based on the relationship between the measured ¹³⁷Cs deposition or inferred ¹³¹I deposition and measured thyroid doses in comparable areas was developed (Gavrilin et al 1999). The thyroid doses obtained in this manner were called 'inferred' doses.

Table 1 shows the distribution of thyroid doses due to ¹³¹I intake reconstructed for the study subjects. The arithmetic mean of the individual thyroid doses for the entire study was 0.3 Gy. The highest individual dose due to ¹³¹I intake among the study subjects was 4.3 Gy. The uncertainties associated with the thyroid doses were not assessed analytically, but were assigned to each study subject depending on the group to which the subject belonged (Gavrilin et al 2004). For the group with 'measured' doses, the GSDs were estimated to vary from 1.7–2.6 depending on the quality of the direct thyroid measurement. For individuals with 'passport' doses, the GSDs ranged from 2.1–2.9 depending on the availability of information on individual milk consumption rates. The overall uncertainties attached to the estimates of the 'inferred' dose were characterized by GSDs of 2.2–2.8 if the milk consumption rate was relatively well known and 2.5–3.1 if the milk consumption rate was unknown.

A screening study of thyroid cancer and other thyroid diseases was conducted in a cohort of individuals exposed in utero to ¹³¹I fallout from the Chernobyl accident (Hatch et al 2009). The cohort comprised 2582 mother–child pairs in which the mother had been pregnant at the time of the Chernobyl accident on April 26, 1986 or in the following two months when ¹³¹I fallout was still present in the environment. Among these, 1494 were categorized as 'exposed'; a comparison group of 1 088 was considered 'relatively unexposed' (Likhtarov et al 2011). At the time of the accident, the women resided in some of the most contaminated territories of Ukraine: Chernihiv, Kyiv, Vinnitsa and the Zhytomyr Oblast.

To estimate the prenatal and postnatal thyroid doses due to ¹³¹I, the mother–child pairs were classified into three groups according to the availability of direct thyroid measurements:

• 720 exposed children with mothers who had direct thyroid measurement made in May–June 1986. Individual instrumental thyroid doses were calculated for these children;
• 774 exposed children whose mothers lived in contaminated settlements where thyroid activities were measured in other women, but not in their mothers; and
• 1088 children who belonged to a relatively unexposed group whose mothers resided in non-contaminated areas where direct thyroid measurements were not performed.

The estimation of the thyroid doses received by the subjects was based on (a) the results of the direct thyroid measurements performed on the mothers of the subjects or on women who lived in the same village as the mother of the subject; (b) responses on behavior and the consumption of milk, milk products and leafy vegetables provided during personal interviews of the mothers of the study subjects; (c) an ecological model of calculation of the ¹³¹I activity in the thyroid of the mother of the subject that takes into account the level of information available for that individual (Likhtarov et al 2005, 2006); and (d) a model of calculation of the in utero thyroid dose per unit intake of ¹³¹I by the mother (Berkovski 1999, ICRP 2001).

Table 1 shows the distribution of thyroid doses due to ¹³¹I intake reconstructed for this study. Individual in utero thyroid dose estimates were found to range from less than 1 mGy to 3.2 Gy, with an arithmetic mean of 0.072 Gy. The main sources of uncertainty in the dose estimates, which were identified but not quantified, are expected to be related to: (a) the transfer coefficient from ¹³¹I intake by the mother to the thyroid dose received by the child; (b) the determination of the gestational age at the time of the accident; and (c) the scaling factors for adjustment of the ecological dose in groups of children whose mothers did not have direct thyroid measurements (Likhtarov et al 2011).

The RADRUE method (Realistic Analytical Dose Reconstruction with Uncertainty Estimation) was developed and extensively tested by an international group of experts (Kryuchkov et al 2009). The main ideology behind the RADRUE technique is very straightforward: it is based on the calculation of the external dose as a product of the exposure rate and irradiation time, with shielding taken into account. Mathematically, the external dose D (mGy) absorbed in the bone marrow (or another organ) during a cleanup worker's trip can be inferred from the summation of the products of exposure rate, duration, and shielding factor during each time interval when the person was exposed to radiation:

$$\begin{eqnarray}&&D=\underset{i=1}{\overset{n}{\sum}}\,{{C}_{i}}\times P\left[x\left({{t}_{i}}\right),y\left({{t}_{i}}\right),{{t}_{i}}\right]\,\times \Delta {{t}_{i}}\times \,{{L}_{i}},\end{eqnarray} \tag{ 2 }$$

where C_i is the conversion coefficient from exposure to bone-marrow dose (mGy h⁻¹ per mR h⁻¹); P[x(t_i),y(t_i),t_i] is the exposure rate (mR h⁻¹) at the location [x(t_i), y(t_i)] and at time t_i, where and when the cleanup worker was present; ▵t_i is the time interval during which the cleanup worker performed a relatively brief but complete task; L_i is the shielding factor for the working conditions (sometimes called the 'location factor'); and n is the number of (usually unequal) time intervals considered in the calculation. Any time that an individual spent working, commuting, sleeping, or resting within the 70 km zone was included in the analysis.

An important element of the RADRUE method is a questionnaire which includes detailed questions about (a) the cleanup workers' routes to and from their work place(s) in the 30 km zone, (b) details about the work they performed, and (c) the locations of the residence and rest quarters used during their stay at Chernobyl. In the case of deceased, or otherwise incapacitated subjects, two proxies were interviewed: (1) a next-of-kin (usually the spouse or a grown-up child), who provided information on the lifestyle, and occupational history, besides the work in Chernobyl and the health data of the subject, and also gave the names of colleagues, who may have worked together with the subject during his cleanup mission, and (2) one (or several, if work periods overlapped only partially) identified colleague(s), to retrieve information on the cleanup activities that were carried out by the subject. Obviously, there is a concern about the accuracy of the information on work history obtained during the personal interview as so many years have passed since the accident. It should be noted that all interviewers were trained and, more importantly, in particular in cohort studies of the Ukrainian cleanup workers, they were experienced cleanup workers themselves. Supporting material, such as booklets with maps, schemes and photographs of the facilities and buildings within the 30 km zone at various periods of the cleanup activities, a dictionary of jargon and terms used by the cleanup workers, as well as photographs of dosimeters were used to help interviewees to remember important elements of their work and rest while in Chernobyl. During the personal interviews interviewers also used special questions to stimulate memory recall of a respondent.

Another key element of the RADRUE technique was a proper interpretation ('reading') of each questionnaire performed by an extremely knowledgeable expert-dosimetrist. Special exercises were performed to check the repeatability of the doses estimated using the results of two interviews of the same subject, and to evaluate the effects of using questionnaire responses from proxies to estimate the doses to cleanup workers (Kryuchkov et al 2009).

Dosimetry experts, familiar with the organization and conditions of work in the 30 km zone, reviewed this information in detail and reconstructed each cleanup worker's type, location and duration of activities and itineraries. These itineraries were entered into the computer program for the RADRUE calculations and linked with the data from an extensive database on exposure-rate measurements and interpolated values to calculate the organ doses. In some cases, because cleanup workers may have overestimated the amount of time they spent in specific high-exposure-rate areas, results from the use of RADRUE were much higher than the permissible dose levels in force at the time of the work. Therefore, a dose constraint procedure was introduced in RADRUE, and the doses were estimated with and without constraints. Both doses without and with constraint were used for epidemiological analyses, although the latter were thought to be more realistic and were applied in the main analyses. A detailed description of the models used to estimate the doses due to external irradiation can be found elsewhere (Kryuchkov et al 2009).

The validity testing involved various intercomparison exercises where RADRUE doses were compared by Kryuchkov et al (2009) to the most reliable dose estimates available for different groups of cleanup workers. These included 39 professional workers from the Ministry of Atomic Energy who wore calibrated personal thermoluminescence dosimeters (TLDs); 20 early responders (firemen, ChNPP workers from the reactor and turbine buildings, etc.) with dose estimates based on the quantitative analysis of unstable chromosome aberrations (dicentrics); and 68 other cleanup workers with dose estimates based on the measurement of electron paramagnetic resonance (EPR) in tooth enamel (table 3).

Figure 7 compares the doses calculated using the RADRUE method with the dose estimates based on personal TLD monitoring of the staff of the Administration of Construction No. 605 (AC-605) from the Ministry of Medium Machinery. The two sets of doses show reasonable agreement within the uncertainty range; the mean of the ratios of the RADRUE-estimated dose to TLD-based dose was 1.3, while the median of the ratios was 1.2. The RADRUE-estimated doses correlated well with the doses derived from analysis of unstable chromosome aberrations (r  =  0.80): however, the mean of the ratios was 0.83 (table 3). This difference could be due to the fact that radiation exposure resulting from the passing cloud of radioactive materials released during the first day of the accident was not considered in the RADRUE doses calculated for early responders. Although only a moderate correlation (r  =  0.38) between RADRUE-estimated doses and EPR-based doses was observed, the mean and the median of the ratios were close to 1.

Zoom In Zoom Out Reset image size

Two categories of uncertainties in the doses estimated by the RADRUE method were identified: 'intrinsic' and 'human factor' uncertainties (Chumak et al 2008). The first category includes uncertainties in exposure-rate data and soil-contamination measurements; uncertainties in the interpolation of these data in time and space; uncertainties in factors used to characterize the effectiveness of shielding; and imprecise answers from the questionnaire (e.g. failure to recall specifically the time of a particular cleanup mission). The second category includes uncertainties of recollection and reporting of the events related to the cleanup activities of a subject that occurred many years ago, especially when proxies were interviewed for deceased cleanup workers to estimate doses to cleanup workers. RADRUE takes into account only the 'intrinsic' component of uncertainty in the dose estimate, while 'human factor uncertainty' was not quantified (Kryuchkov et al 2009). Human factor uncertainty is being considered in depth in a separate study. Such a study, including the analysis of the work history reported by the cleanup workers shortly (weeks or months) after completion of their cleanup mission in 1986–1987, and of that obtained recently using the study questionnaire as well as the comparison of two sets of doses calculated using historical and recently reported work history, is currently underway.

In order to evaluate the intrinsic uncertainties, a Monte Carlo method was used to calculate 10 000 individual stochastic dose estimates for each study subject. No attempt was made to separate the shared and unshared errors in the calculations, but a sensitivity analysis was performed by Kryuchkov et al (2009) to evaluate the relative importance of the shared and unshared errors in the overall uncertainties of the RADRUE dose estimates. In brief, it was found that (1) the variability in the exposure rate contributed more than 96% to the overall uncertainty of the doses, while the variability of other parameters (location factor, frame duration, episode repetition, conversion coefficient) accounted for only a few percent of the overall uncertainty; and (2) the same location during the cleanup mission was shared by two cleanup workers at the same time in 0.13% instances and, as a result, the 'shared' dose represented less than 1% of the total dose among the study population. It is important to note that there are other sources of shared errors that have not been considered. For example, all the exposure rate measurements in the 4 km zone were made in 1986. Later, the exposure rates were extrapolated, and the uncertainties associated with that extrapolation were large. The doses to people working there after 1986 were all linked to the dataset used for the extrapolation. The same comment applies to the measurements made in other areas.

In the framework of the Ukrainian–American collaborative project, a case-control study of leukemia and related disorders was conducted in a cohort of 110 645 male Ukrainian cleanup workers of the Chernobyl accident, who were exposed to various radiation doses over the 1986–1990 time period. The study included 1000 subjects and consisted of two phases. Phase 1 of the study included 572 subjects (Romanenko et al 2008a, 2008b), while phase 2 covered 428 subjects, for both cases and controls (Zablotska et al 2013). As already mentioned in section 3.1, all individual bone-marrow doses were calculated based on personal cleanup history data collected from an interview with the subject himself, if he was alive, or with two proxies, if he was deceased (Chumak et al 2008, 2015). The arithmetic means of the individual stochastic estimates of the bone-marrow dose range from 3.7  ×  10⁻⁵ to 3260 mGy, with an arithmetic mean of 92 mGy for the entire study population (table 4). The GSDs (intrinsic uncertainty) of the individual stochastic external doses calculated by RADRUE varied among the study subjects from 1.2–5.9 with a mean equal to 2.0 (table 5).

A nested case-control study of hematological malignancies was conducted in cohorts of about 66 000 Chernobyl cleanup workers of 1986–1987 from Belarus, 65 000 from Russia and 15 000 from the Baltic countries. A total of 70 cases (40 leukemias, 20 non-Hodgkin lymphomas and ten other malignancies of lymphoid and hematopoietic tissue) and 287 matched controls were included in the study (Kesminiene et al 2008). The arithmetic means of the individual stochastic estimates of the bone-marrow doses among the study subjects ranged from 0.007–657 mGy (table 4). The GSDs of the individual stochastic external doses varied from 1.2–4.0 with a mean equal to 1.9 (table 5).

A collaborative case-control study nested within the cohorts of Belarusian, Russian and the Baltic cleanup workers was conducted by IARC to evaluate the radiation-induced risk of thyroid cancer (Kesminiene et al 2012). The study included 107 cases and 423 controls. In this study, in addition to the doses due to external irradiation, thyroid doses were also reconstructed for the following exposure pathways:

• Intake of ¹³¹I via inhalation during the period of work as a cleanup worker was calculated for the six study subjects who worked on the industrial site during the first 10 d after the accident;
• Inhalation of short-lived iodine and tellurium isotopes during the cleanup mission was estimated using the approach developed for the population of the 30 km zone around the Chernobyl nuclear power plant (Gavrilin et al 2004); and
• Intake of ¹³¹I with locally produced foodstuffs during residence between 26 April and 30 June 1986. Cleanup workers who were residents of the southeastern part of Belarus also received non-negligible doses at their places of residence if they lived in highly contaminated areas and consumed locally produced milk and/or vegetables.

The approach for the reconstruction of thyroid doses due to ¹³¹I was developed and validated by the results of direct thyroid measurements taken between 30 April 1986 and 5 May 1986 in a group of 624 early cleanup workers. For this purpose, the values of ¹³¹I activity in the thyroids were calculated with the approach used in the case-control study and then compared with those derived from the direct thyroid measurements. The mean of the ratios of those measured to the calculated activities was found to be 0.8  ±  1.1 and the median of the ratios was 0.5. The modeled doses from ¹³¹I inhalation for the study subjects were then calibrated against the median of the ratios of those measured to the calculated ¹³¹I activity in the thyroids of the group of 624 cleanup workers.

An estimation of the thyroid doses during residence and of their uncertainties was carried out using models developed previously for calculating the average thyroid doses due to ¹³¹I intake in adults who resided in the contaminated settlements of Belarus (Gavrilin et al 2004, Drozdovitch et al 2010). Such doses were not calculated for the Russian and the Baltic cleanup workers, who were sent on mission from distant areas with low or no contamination. The validity of the modeled thyroid doses was checked by comparison with the instrumental doses. The study subjects from Belarus resided in 33 settlements, where direct measurements of ¹³¹I activities in the thyroid were performed in May and the beginning of June 1986. Calculated residential thyroid doses from ¹³¹I intake were compared with the mean thyroid doses in residential settlements derived from direct measurements of ¹³¹I activities in the thyroid (figure 8). The mean of the ratios of the modeled-to-instrumental thyroid doses from ¹³¹I intake was 1.2  ±  0.5 and the median of the ratios was 1.1.

Zoom In Zoom Out Reset image size

Thyroid doses from different exposure pathways estimated in this study are shown in table 4. It should be noted that only doses due to external irradiation during mission, which were calculated using RADRUE, are given for all study subjects. For other exposure pathways mean and GM of doses are shown only for the study subjects who were exposed to that exposure pathway. As can be seen from the table, for Belarusian cleanup workers exposure from ¹³¹I intake during residence was the main contributor to the overall mean thyroid dose.

The GSDs of the individual stochastic external doses calculated by RADRUE varied from 1.1–5.8 with a mean of 1.9 (table 5), while for the residential doses from ¹³¹I intake GSDs varied from 1.9–2.5 with a mean of 2.2 (not shown). Doses from external radiation and ¹³¹I intake were found to be independent of each other, because the approach to estimate these doses differed considerably, and little correlation was found between the arithmetic means of the individual stochastic dose estimates from both sources (Spearman's rank correlation coefficient: r_s  =  0.096). It should be noted that uncertainty in the thyroid doses due to ¹³¹I intake during residence was evaluated taking into account the separation of shared and unshared errors in the dosimetry model.

The Research Center for Radiation Medicine (Kyiv, Ukraine) in collaboration with the U.S. (NCI) is conducting a case-control study of thyroid cancer nested in a cohort of Ukrainian cleanup workers. In this study, in addition to the doses due to external irradiation, the thyroid doses are also being reconstructed, as in the IARC-coordinated study, for (a) the intake of ¹³¹I and short-lived iodine and tellurium isotopes via inhalation during the period of work as a cleanup worker, and (b) the intake of ¹³¹I with locally produced foodstuffs during residence. Some cleanup workers, who were residents of the northern part of Ukraine, might have received non-negligible doses at their places of residence, while consuming locally produced milk and/or vegetables.

Approaches for the reconstruction of thyroid doses due to the intake of ¹³¹I and short-lived iodine and tellurium isotopes via inhalation during the period of work as a cleanup worker were the same as in the IARC-coordinated study of thyroid cancer. To estimate residential exposure from the ¹³¹I intake, the method developed by Likhtarov et al (2005) to estimate the thyroid doses in the population in the absence of direct thyroid measurements was used.

Thyroid doses from different exposure pathways estimated in this study are shown in table 4. As can be seen from the table, external irradiation during the cleanup mission was, on average, the main contributor to the thyroid dose. However, some cleanup workers received high doses (up to a few Gy) from ¹³¹I intake from cows' milk at their places of residence between 26 April and 30 June 1986, as they lived in highly contaminated areas in a 30 km zone around the Chernobyl accident site. It should be noted that the dose estimates are preliminary as the study is ongoing.

The GSDs (intrinsic uncertainty) of the individual stochastic external doses varied from 1.2–7.0 with a mean equal to 2.1 (table 5), while for the residential doses from ¹³¹I intake GSDs varied from 1.8–4.9 with a mean of 2.5 (not shown). Doses from external radiation and ¹³¹I intake were found to be independent, as the approach used to estimate these differed considerably and because little correlation was found between the arithmetic means of the individual stochastic dose estimates from both sources (r_s  =  −0.02). Uncertainty in the thyroid doses due to the ¹³¹I intake during residence was evaluated, taking into account the separation of shared and unshared errors in the dosimetry model.

Dosimetry in occupational studies has an advantage, because personnel having contact with ionizing radiation were monitored using personal dosimeters. However, occupational doses are also associated with uncertainties.

A study of errors in the recorded doses was conducted as a part of the 15 country nuclear industry workers' study (Cardis et al 2007) to evaluate the comparability of the recorded dose estimates across facilities and time, and to identify and quantify the sources of bias and uncertainties in the dose estimates (Thierry-Chef et al 2007). The major sources of errors in external radiation doses were found to be dosimetry technology, exposure conditions of the workplace, and calibration practices. These sources of systematic errors were characterized and quantified. Although the sources of the dose errors (mainly shared) were well quantified in the study, multiple dose realizations were calculated not for the entire cohort, because of the cohort size of 407 391 nuclear industry workers, but only for the Oak Ridge National Laboratory (ORNL), one of the study facilities included in the 15 country study (Stayner et al 2007). Monte Carlo simulations were used for this facility to randomly select a set of bias factors from the assigned distributions. The resulting bias factors were then used to estimate the annual 'corrected' doses for each worker in the study by dividing their annual recorded dose by the appropriate sampled bias factor. Ten thousand individual stochastic annual doses were generated for each individual in the ORNL cohort and used for estimating an excess relative risk and its confidence interval that reflects both the statistical sampling error and uncertainty in the measurement of exposures (Stayner et al 2007). Another source of uncertainties is the difficulties in accounting for exposure due to lower-energy photons and neutrons and from the intake of radionuclides. In particular, in the 15 country study workers with a potential for a substantial dose from these sources of exposure (⩾10% of their whole-body dose) were flagged and excluded from the main epidemiological analyses (Thierry-Chef et al 2007).

Updated and expanded methods of occupational exposure assessment were used in the International Nuclear Workers Study (INWORKS) conducted in France, the United Kingdom and the United States of America in a cohort of 308 297 workers (Thierry-Chef et al 2015). This approach accounts for differences in dosimeter response to predominant workplace energy and geometry of exposure, and for sex-specific dose conversion coefficients recently published by the ICRP (ICRP 2010). Workers potentially exposed to neutrons (neutron dose exceeded 10% of the external dose) and with internal contamination were identified and flagged (Thierry-Chef et al 2015). As in the 15 country study, individual stochastic doses were not calculated for the cohort.

The most comprehensive uncertainty assessment in the doses reconstructed for the studies of occupational exposure was done for the study of 110 374 U.S. radiologic technologists (USRT study) (Sigurdson et al 2003). The previous USRT dosimetry system (Simon et al 2006a) was revised and significant improvements were made to estimate the annual and cumulative individual organ doses received by medical radiologic technologists (Simon et al 2014). The USRT dosimetry system was designed to account for the separation of the shared and unshared errors, and allowed for the production of 1000 realizations of the cohort dose. Uncertainties in the organ doses for each study subject are characterized by the GSD of 1000 individual stochastic doses. For example, the GSDs of breast doses among the female study subjects varied from 1.2–6.4 (Simon et al 2014).

Obviously, there are limitations in occupational dosimetry including, among other factors, the processing of dosimetry readings below the detection limit, the lack of measurements in earlier years, and not accounting for neutron doses and doses from the intake of radionuclides.

Different measures of exposure have been used in the epidemiological studies of low-level medical exposure, in particular, of exposure to patients from computed tomography (CT) procedures. Some studies did not estimate the actual radiation exposure from each CT scan, because of the absence of the patients' data for dose assessment. In some of studies, the number of CT scans was used as a measure of a person's radiation exposure (e.g. Huang et al 2014). In others, an effective dose, not an absorbed dose, was used to assess risk estimates (e.g. in Mathews et al 2013). In the studies with the most advanced dosimetry, an absorbed dose in the target organs was reconstructed for the study subjects based either on a CT scan-dose matrix or on CT-model and machine settings (e.g. Pearce et al 2012, Meulepas et al 2014).

The actual absorbed doses from CT scans can vary substantially depending on the patient's age, sex, equipment type, shape and size of the patient's body, the use of contrast, protocols of scanning, purpose of examination, etc. Data for the machine settings that also influence the dose, such as milliampere seconds and peak kilovoltage, are typically not available for each of the study participants. Therefore, a dose that is typically averaged per scan by site, years of scan and by age, is used to characterize individual exposure (Mathews et al 2013). Other sources of uncertainty include the absence of records on repeated scans that were done because of patient movements; missing doses for CT examinations with unknown anatomical areas; and incomplete radiology data (Journy et al 2015, Krille et al 2015).

Although radiation exposure assessment in the medical exposure studies with the most advanced dosimetry (e.g. Pearce et al 2012, Meulepas et al 2014) is quite accurate compared to that in other areas of radiation epidemiology, uncertainties in the studies of low-level medical exposure are acknowledged, but not numerically estimated.

Uncertainties in the studies of environmental exposure, where thyroid doses were based only on modeling, are typically higher than those reported in the post-Chernobyl thyroid dose estimates based on direct thyroid measurements. A mean GSD of 2.2 (range 1.6–5.4) was reported by Kopecky et al (2004) for thyroid doses reconstructed for the participants of the Hanford thyroid disease study. For thyroid doses reconstructed in the Utah thyroid cohort study, as a consequence of radioactive fallout from atmospheric nuclear weapons tests, the GSDs varied from 1.5–8.5 with the majority of values lying in the range from 2–3 (Simon et al 2006b). A comprehensive dose reconstruction using the 2DMC method (Simon et al 2015) was conducted recently for a cohort of 2376 individuals exposed at age  <21 years to the Semipalatinsk Nuclear Test Site fallout in Kazakhstan (Land et al 2015). The GSDs of individual stochastic thyroid doses due to the intake of radioiodines varied up to 9.3 with a mean GSD of 3.7 and the majority of values lying in the range from 3–4 (derived from Land et al 2015).

The discussion on environmental studies is split into two parts: those that deal with thyroid cancer and other thyroid diseases, where ¹³¹I is the main radionuclide contributing to the thyroid dose, and the studies of all other cancer endpoints, for which external irradiation and internal irradiation due to ingestion of caesium isotopes play a major role.

4.2.1.1. Studies of thyroid cancer and other thyroid diseases.

4.2.1.2. Studies of other endpoints.

4.2.2.1. Leukemia studies.

4.2.2.2. Studies of thyroid cancer.