Ionizing radiation is one of the established risk factors for thyroid cancer, especially after exposure during childhood. Evidence of increased radiosensitivity of the thyroid in childhood and adolescence is available not only for external gamma- and x-ray radiation but also for internally incorporated radioiodines [1
]. Excessive risk of thyroid cancer remains elevated 50 or more years after exposure in the atomic bomb survivors who were below 20 years old at the time of bombardments, and 30 years after the Chernobyl radioactive fallout in the Ukrainian-American cohort of people who were below 18 years at the time of the accident [4
The Chernobyl nuclear power station accident resulted in a widespread release of substantial amounts of radionuclides into the atmosphere leading to exposure, predominantly to radioiodines, of large numbers of children and adolescents residing in the most contaminated areas of Belarus, Ukraine, and the Russian Federation (RF). Post-Chernobyl epidemiological studies had reported an increased risk of thyroid cancer, benign thyroid tumours, and thyroid dysfunction, mainly due to internal exposure to iodine-131 (131
I), although interaction between 131
I exposure and other potential non-radiation risk factors remained unclear [1
A population-based case–control study of thyroid cancer was carried out by the International Agency for Research on Cancer (IARC, WHO), in collaboration with the Sasakawa Memorial Health Foundation, in the most contaminated areas of Belarus and Russia [1
]. The study aimed to evaluate an association between exposure to 131
I and risk of thyroid cancer, and the effect of stable iodine supplementation for goiter prophylaxis on the radiation-related thyroid cancer risk among those exposed in childhood and adolescence. This study is, to our knowledge, the only analytical study that used a weighted soil iodine level in the residence of study participants as an environmental indicator of long-term stable iodine intake status. Results of the first analyses, focused on subjects who were less than 15 years old at the time of the Chernobyl accident, were published by Cardis et al. [1
]. Recently, some improvements were implemented in the 131
I dose reconstruction approach by incorporating revised age-specific thyroid mass values for the Belarusian children.
Since the contribution of radioisotopes other than 131I to the thyroid dose was very small and there were no changes in the methodology, we report only on the updated 131I thyroid dose–effect analysis using improved 131I thyroid dosimetry and on association between non-radiation factors, such as soil iodine deficiency, iodine supplementation, self-reported personal and family history of thyroid diseases, body mass index (BMI), and thyroid cancer risk, and also on their interaction with 131I-related thyroid cancer risk. In the current analyses, we also included study subjects who were 15–18 years old at the time of the accident.
Selected characteristics of the study participants are presented in Table 1
. In total, 298 thyroid cancer cases and 1934 controls were included in the study from two regions of Belarus (Gomel, Mogilev) and four regions of the RF (Bryansk, Kaluga, Orel, Tula). One case, confirmed as follicular adenoma by the study international pathologist panel, was excluded from the data analysis, together with the respective controls. Approximately 77% of all cases (228 out of 298) were from Belarus, with 192 out of 228 cases from Gomel region (84%). Approximately 63% of thyroid cancers were in women, and 58% of the cases were less than 5 years old at the time of the accident.
I thyroid doses (arithmetic means of 1000 individual stochastic doses) in the study subjects had a log-normal distribution, with 937 subjects having a 131
I dose <0.10 Gy. The mean (median) value of 131
I thyroid dose was 0.54 (0.29) Gy for study subjects in Belarus and 0.10 (0.02) Gy in the RF. Mean and median 131
I thyroid dose estimates by case or control status for six study regions are shown in Table 2
. The highest mean 131
I thyroid doses of 0.77 Gy and 0.58 Gy in cases and controls, respectively, were in Gomel region. Kaluga and Orel regions had the lowest mean 131
I thyroid dose, which was about 0.03 Gy for both cases and controls. The frequency distribution of study cases and controls by categories of 131
I absorbed thyroid dose is presented in Figure 1
Information on thyroid cancer diagnosis circumstances was available for 274 cases. In total, 140 out of 274 cases (51%) were diagnosed at screening, 72 (26%) were diagnosed at doctor’s consultations because of having thyroid-gland-related symptoms, and 62 (23%) were diagnosed at doctor’s consultations after referral for care due to non-thyroid related problems. Information on tumor size was systematically available only for 227 thyroid cancers from Belarus, among which 72 (32%) had tumor size less than 10 mm. Mean (median) tumor size was 14.5 (12) mm, ranging from 2 to 60 mm. Information on cancer staging based on TNM classification of malignant tumors describing tumor size, involvement of lymph nodes and presence of metastasis [10
] was available for 227 thyroid cancers from Belarus, among which 151 cases (66.5%) had regional lymph node metastasis, including 5 (3.3%) cases with distant metastasis.
2.1. Association Between 131I Thyroid Dose and Thyroid Cancer Risk
Twelve study subjects (3 cases and 9 controls) had estimated doses of above 5 Gy and were excluded from the analysis because we believed that their doses might be overestimated based on the comparison between model-based and measurement-based doses. Subsequently, we also had to exclude 15 control subjects who were matched to those three cases with doses higher than 5 Gy. Additionally, one case and three controls were excluded from the risk analysis because they were in the Ukraine on the April 26, 1986. The 6 respective controls to this case were excluded as well. Overall, 2195 subjects (294 cases and 1901 controls) were included in the thyroid cancer risk analysis.
Adjustment for self-reported personal history of benign nodules significantly improved the model goodness of fit (p < 0.05), so we retained this parameter in the model as a potential confounder of association between 131I thyroid dose and thyroid cancer.
Dose–effect associations between 131
I thyroid dose and thyroid cancer based on parametric and non-parametric risk models are shown in Figure 2
. We found a statistically significant linear dose–effect association with odd ratios (OR) at 1 Gy of 3.34 (95% CI: 2.04; 5.93), based on parameter estimates of β = 2.34 with p
for linear trend = 0.003. However, similar to the previously reported results, we observed a departure from linearity at a higher dose range from 2 to 5 Gy, where the linear quadratic model described the dose–response better than a simple linear model (p
test for non-linearity < 0.001), with somewhat higher OR at 1 Gy of 4.71 (95% CI: 2.23; 7.19) based on parameter estimates of β = 4.70 and ϒ = −1.00. At the 131
I thyroid dose range up to 2 Gy, there was no evidence of non-linearity in the dose response (p
= 0.49), and OR at 1 Gy was 5.12 (95% CI: 2.98; 9.41), based on parameter estimates of β = 4.12. Thyroid cancer OR estimates by 131
I thyroid dose categories based on non-parametric dose response analysis are presented in Table 3
. ORs were statistically significantly elevated for 131
I thyroid dose categories above 0.1 Gy, as compared with the reference category (p
< 0.05 in each case category above 0.1 Gy).
We performed a sensitivity analysis of thyroid cancer risk using the same excess odds ratio (EOR) linear quadratic model and dose range <5 Gy, but excluding 14 thyroid cancer cases having non-papillary subtypes (13 follicular and 1 medullar). We estimated an OR at 1 Gy of 4.58 (95% CI: 2.14; 7.03), showing almost no effect of exclusion of non-papillary thyroid cancers on thyroid cancer radiation risk estimate.
We also performed sensitivity analysis of risk estimates by tumor diameter (<10 mm vs. ≥10 mm) for cases from Belarus where information on tumour size was available (total of 223 cases in the dose range up to 5 Gy). Fitting the EOR model with linear quadratic dose response in the dose range up to 5 Gy, risk estimates were very similar for two groups: the OR at 1 Gy was 5.19 (95% CI: 1.67; 8.71) when cases with tumor size ≥10 mm were included (151 cases), compared to 5.17 (95% CI: 0.13; 10.48) when cases with tumor size smaller than 10 mm were included (72 cases). We did not see statistically significant association between thyroid dose and tumor diameter (p = 0.68 based on Spearman’s rank-order correlation test).
2.2. Association Between Selected Non-Radiation Risk Factors and Thyroid Cancer
We investigated associations between thyroid cancer and several independent non-radiation risk factors while adjusting for thyroid dose effect. Selected risk factors included self-reported personal and family history of thyroid disease, BMI, stable iodine intake status described by the average level of soil iodine in the settlement at the time of the accident, and iodine supplementation received in months and years following the accident (Table 4
). We studied the role of independent risk factors in the subjects with 131
I thyroid doses up to 2 Gy where there was no statistical evidence of non-linearity in the dose response. We found statistically significant associations between: thyroid cancer and self-reported personal history of benign nodules, with an OR of 14.26 (95% CI: 4.50; 45.18); history of any thyroid disease (except thyroid cancer), with an OR of 1.98 (95% CI: 1.30; 3.00); and family history of thyroid cancer, with an OR of 3.37 (95% CI: 1.38; 8.24). The overweight BMI category was also statistically significantly associated with thyroid cancer, with an OR of 1.87 (95% CI: 1.24; 2.82) (Table 4
Both iodine supplementation received regularly at school in the months and years following the accident and stable iodine intake status estimated by the average level of soil iodine in the settlement at the time of the accident showed a statistically significant association with thyroid cancer risk (Table 4
). Iodine supplementation had an inverse association with thyroid cancer risk, with an OR of 0.41 (95% CI: 0.25; 0.66) among subjects who consumed antistrumin compared to subjects who did not. The length of antistrumin consumption was less than one year for about 74% of cases and 68% of controls who reported on intake. The thyroid cancer OR estimate in subjects with deficient iodine intake status was 63% higher as compared to those with sufficient iodine intake (p
We observed elevated but non-statistically significant thyroid cancer ORs in subjects with self-reported personal history of goiter, family history of any endocrine disease, thyroid nodule, or goiter as compared to the subjects who did not report on personal or family history of these diseases (Table 4
2.3. Effect Modification of the Radiation Dose Response
We tested for possible interaction between 131
I thyroid dose and a few selected non-radiation risk factors (Table 5
). There was no statistically significant variation of 131
I-related thyroid cancer risk between men and women and by age at the accident (p
for heterogeneity = 0.78 and 0.15, respectively).
Iodine supplementation in the years after the accident had a statistically significant modifying effect on 131I-related risk of thyroid cancer (p = 0.05 for heterogeneity). Subjects who received iodine supplementation in the years after the accident had a significantly lower 131I-related risk of thyroid cancer with OR at 1 Gy of 0.65 (95% CI: 0.23; 1.81), while subjects with no iodine supplementation had an OR at 1 Gy of 3.62 (95% CI: 2.43; 5.40).
Non-statistically significantly increased OR for radiation-related risk of thyroid cancer was also observed among subjects with self-reported history of thyroid diseases other than thyroid cancer as compared to the subjects with no thyroid disease history (p
for heterogeneity = 0.16); in subjects with self-reported history of benign nodules as compared to subjects reported to be nodule-free; in subjects with family history of thyroid cancer as compared to those without family history of thyroid cancer; in overweight subjects as compared to those with normal BMI; and in subjects with deficient iodine intake status at the time of the accident as compared to those with sufficient iodine intake (p
for heterogeneity >0.5 in each case) (Table 5
We report a significant positive association between 131
I thyroid dose and thyroid cancer risk after exposure at ages below 18 years at the time of the Chernobyl fallout using improved thyroid dosimetry and an expanded dataset, following up on the earlier case–control study [1
I thyroid dose estimates were revised based on more precise age-specific thyroid mass values in Belarusian and Russian children, updated reduction factors for different types of milk and dairy products, and reclassified shared and unshared dose errors. This resulted in somewhat lower estimates of 131
I thyroid dose in the study subjects with 131
I median dose of 0.29 Gy compared to the previously reported median of 0.36 Gy in Belarus, and of 0.02 Gy as compared to the earlier reported median of 0.04 Gy the RF [1
We evaluated the shape and magnitude of the dose response for 131
I thyroid doses below 5 Gy and below 2 Gy, and potential modifying effect of some non-radiation risk factors on thyroid cancer 131
I-related risk. In the thyroid dose range up to 5 Gy, a linear quadratic model described data significantly better than a linear model, while there was no evidence of non-linearity in dose response in analysis restricted to 131
I doses below 2 Gy. Thyroid cancer OR at 1 Gy was 4.7 (95% CI: 2.2; 7.2) for 131
I doses up to 5 Gy based on the linear quadratic EOR model. This estimate is comparable to the previously reported OR at 1 Gy of 4.9 (95% CI: 2.2; 7.5), estimated over the entire range of total (131
I + external exposure) thyroid dose (up to 10.2 Gy), where almost 95% of the total dose was due to 131
I exposure [1
]. In addition, we calculated a similar OR at 1 Gy of 5.1 (95% CI: 3.0; 9.4) for 131
I doses below 2 Gy based on the linear EOR model, compared to the previously reported OR at 1 Gy of 5.2 (95% CI: 2.2; 8.2) for the same model and dose range [1
]. It should be emphasized that ORs in the present study were calculated using the mean of 1000 dose realizations for each subject rather than point estimates in the earlier study [1
], and hence our risk estimates are adjusted for errors in doses. However, within this study, we did not consider the dosimetric uncertainties in the confidence intervals of the risk estimates. Use of improved dose estimates that resulted in decrease of 131
I median dose for 0.07 Gy and 0.02 Gy in Belarus and the RF, respectively, did not have a high impact on the calculated ORs compared to previous analyses.
Our results are consistent with those reported from thyroid cancer and other thyroid disease studies in the Belarusian-American and Ukrainian-American screening cohorts of subjects who were below the age 18 at the time of the Chernobyl accident and who had direct thyroid activity measurements, with doses mainly from exposure to 131
]. A significant positive linear association between 131
I thyroid doses up to 10 Gy and thyroid cancer risk was established in the Ukrainian-American cohort with an EOR/Gy of 5.25 (95% CI: 1.70; 27.5)—that is, a OR of 6.25 at 1 Gy—for prevalent cases found during the first round of screening (from 1998 to 2000), and with an excess relative risk per grey (ERR/Gy) of 1.9 (95% CI: 0.43; 6.34)—that is, a relative risk (RR) at 1 Gy of 2.9—for incident cases diagnosed during second through fourth rounds of screening between 2001 and 2007 in Ukraine [3
]. Statistically significant 131
I-related thyroid cancer risk persisted in this cohort, even 30 years after the exposure with EOR/Gy of 1.36 (95% CI: 0.39; 4.15)—that is a OR at 1 Gy of 2.36—for cases diagnosed during the fifth cycle of the screening during 2012–2015 [6
]. In the Belarusian-American cohort, a linear dose response was demonstrated for 131
I doses up to 5 Gy with EOR/ Gy of 2.15 (95% CI: 0.81; 5.47)—that is, an OR of 3.15 at 1 Gy—and a linear-exponential dose response at the dose range up to 32.8 Gy, based on thyroid cancer prevalence cases diagnosed during screening in the period from 1996 through 2004 [2
]. In a cross-sectional thyroid ultrasound screening study of 2376 residents around the Semipalatinsk nuclear weapon test site exposed before the age of 21 years, there was a positive association between papillary thyroid cancers and thyroid dose from external and internal radiation sources, although this was not statistically significant and was based on a small number of cases (n
= 21) [13
]. No association was reported between combined benign and malignant thyroid-prevalent tumors and 131
I thyroid exposure in the Ozyorsk city residents exposed in childhood to atmospheric releases of radioiodines by the Mayak nuclear weapon production, although the results were based on five cases only, which lacked estimates of thyroid doses in the exposed population [14
Our risk estimates are also in agreement with thyroid cancer risks after external irradiation in childhood. In a pooled 12-study analysis of thyroid cancer risk after external environmental and medical exposures in childhood (<20 years at exposure), a relative risk (RR) at 1 Gy of 6.5 (95% CI: 5.1; 8.5) was reported with mean time since exposure of 28.9 years [5
]. In that analysis, the dose response was statistically significant and positive for thyroid doses up to 0.1 Gy, with no evidence for non-linearity, though a supralinear relation was seen in the 2–4 Gy dose region, which plateaued between 10–30 Gy and decreased at doses of 30 Gy or more. Among members of the Life Span Study (LSS) cohort of atomic bomb survivors who were exposed to acute external gamma- and neutron-radiation at age 10 years and reached the age of 60 years, an increased thyroid cancer risk was reported with ERR/ Gy of 1.28 (95% CI: 0.59; 2.70), with a linear dose response at thyroid doses up to 2 Gy [4
I-related thyroid cancer risk decreases with attained age and time since exposure, however it remains statistically significant [3
]. In our study, where cases were ascertained between 1992–1998 (mean time since exposure of 9.5 years), our reported risk estimates are fairly comparable with the estimates obtained from the first rounds of screening in Belarusian-American and Ukrainian-American cohorts in 1996–2004 and 1998–2000, respectively [2
We found a significant positive association between thyroid cancer risk and self-reported personal history of benign nodules, of any thyroid disease, and family history of thyroid cancer after adjustment for 131I thyroid dose effect.
Importantly, the indicators of iodine deficiency were statistically significantly associated with the risk of thyroid cancer after adjustment for 131
I thyroid dose effect. Subjects with deficient iodine intake at the time of the accident had a 63% higher risk of thyroid cancer compared to subjects with sufficient iodine intake; subjects who received iodine supplementation in the years following the accident had an almost 60% lower risk of thyroid cancer compared to those who did not. Other Chernobyl-related studies have reported no evidence of statistically significant thyroid cancer background risk in relation to urinary iodine concentration (used as an indicator of iodine intake, which reflects stable iodine consumption only for a short time before the measurement) [2
]. In our study, we used average stable iodine content in soil—estimated for villages where the study subjects resided—as a stable iodine intake indicator. It reflects long-term iodine consumption, because in the past, locally produced food was the main source of the daily iodine intake in the study areas. We also found a statistically significant positive association between thyroid cancer risk and increased BMI. A statistically significant association between increased BMI in childhood and increased risk of thyroid cancer in adulthood was reported in a large prospective cohort of children who had weight and height measurements in the period from 7 to 13 years of age [16
]. A positive association with BMI before the diagnosis was also observed in the case–control study of thyroid cancer risk in French Polynesia [17
]. However, in our study, the measurements were performed 12–16 years after the accident and after the thyroid cancer diagnosis, limiting our ability to discuss whether the observed association is causal or not.
Although thyroid nodules are quite common in clinical practice, about 7–15% of them are expected to be malignant [18
]. Our study showed a strong association between the self-reported personal history of benign thyroid nodules and thyroid cancer risk, with an OR of 14.29 (95% CI: 4.51; 45.25). However, the estimate was based on a small number of subjects (18 subjects in total, among whom 12 were thyroid cancer cases), and we had no means to validate the self-reported thyroid nodule history. In contrast, in the Ukrainian screening cohort, ultrasound-detected nodules were positively but not statistically significantly associated with increased background risk of thyroid cancer (RR = 2.44, 95% CI: 0.96; 6.19) [3
We did not observe statistically significant variation of the 131
I-related thyroid cancer risk by personal history of any thyroid disease excluding thyroid cancer (p
= 0.16), nor by family history of thyroid diseases (p
> 0.5). Although not statistically significantly, thyroid cancer 131
I-related risk was much higher in the study subjects with self-reported personal history of benign nodules (n
= 12) as compared to those who did not report on benign nodule history (n
= 270), with ORs at 1 Gy of 32.61 and 3.03, respectively (p
for heterogeneity >0.5). In the Belarusian-American thyroid screening cohort, thyroid cancer radiation-related risk was significantly higher in people with nodular or diffuse goiter in anamnesis or detected at screening compared to goiter-free subjects, and in subjects with enlarged thyroid volume compared to those with normal thyroid volume [2
]. By contrast, in the Ukrainian-American cohort, the radiation risk did not vary by presence of diffuse goiter, levels of serum thyroglobulin, or urinary iodine [3
]. Also, family history of nodular goiter was positively associated with thyroid cancer risk in Belarus (OR = 3.54, p
< 0.001) [2
Iodine deficiency has an impact on the dose–effect relationship by increasing 131
I thyroid uptake and stimulating thyroid cellular activity [20
]. Elucidation of the role of iodine deficiency on the radiation-induced thyroid cancer risk is very important, particularly in areas of iodine deficiency. Post-Chernobyl thyroid cancer studies, generally, lack reliable indicators of iodine deficiency of the study population at the time of the accident, and use diffuse goiter, thyroglobulin, or urinary iodine levels as an indicator of long-term deficiency, with the latter two reflecting only the most recent dietary intake [2
]. Similar to earlier published study [1
], we showed that 131
I-related thyroid cancer risk is somewhat higher in subjects with deficient soil iodine levels as compared with subjects with sufficient soil iodine levels, although in our study, the difference was not statistically significant.
Iodine supplementation affects iodine deficiency status. Since the areas contaminated after Chernobyl were known for being iodine-deficient, regional endocrinological dispensaries, particularly in Belarus, had stocks of potassium iodide (called antistrumin), which was used for goiter prophylaxis in the former Soviet Union. Antistrumin was mainly administered to evacuated children in Belarus, and only a few subjects are reported to have taken it in Russia. The distribution sometimes continued months and even years after the accident [1
]. Our data showed that subjects who reported regular iodine supplementation in schools and summer camps after the Chernobyl accident had lower risk of spontaneous thyroid cancer compared to the subjects who did not. More importantly, our results suggested that the iodine supplementation received in the years following the accident also diminished the 131
I-related thyroid cancer risk. As the study population was not systematically provided with stable iodine immediately after the accident, it is plausible that longer-term stable iodine supplementation in the deficient areas could have had a positive effect on the process of thyroid growth in children, resulting in a lower thyroid cancer incidence among them [1
]. This has important implications for radiation protection policy, particularly in areas of iodine deficiency in cases of nuclear accidents.
Our study has some limitations, including possible recall bias as a consequence of conducting interviews with the study subjects or their mothers 12–16 years after the accident. Recall bias may have led to under- or overestimation of reported milk and dairy product consumption, resulting in under- or overestimation of individual thyroid doses. It is difficult to evaluate whether such an error would be differential between cases and controls. Even if it is, however, because the thyroid dose reconstruction model depends on many other factors besides level of consumption, it is not clear what effect this would have on the dose–response relationship. Uncertainty in reported consumption is, however, a source of classical error in dose estimations, and would tend to bias risk estimates towards the null in continuous analyses. Another study limitation includes self-reported information on thyroid diseases when the cases could have a better recall than the controls, or when the case may have been under increased surveillance for thyroid cancer. Our assessment of 131I-related thyroid cancer risk variation by categories of potential effect modifiers was sometimes limited by relatively small numbers in some categories, preventing detection of interactions.
Comparison between the model-based individual thyroid doses and the doses based on individual direct thyroid measurements for 64 study subjects with available individual measurements showed a rather wide range of ratios between the two sets of doses [23
]. For 70.3% individuals, the correspondence between the two doses was within a factor of 3. The mean ± standard deviation of ratios of thyroid dose based on the model to the dose based on direct thyroid measurements was found to be 1.2 (±1.3; median of 0.8). There was a moderate positive correlation between model-based individual thyroid doses and the doses based on individual direct thyroid measurements (Spearman’s correlation coefficient = 0.50, p
< 0.001). The observed difference between model- and measurement-based thyroid doses could be explained by relatively large uncertainties in the doses calculated using the “semi-empirical” model and uncertainties associated with recalling the information on relocation history and individual diet of the study subjects collected more than 10 years after the Chernobyl accident. To account for the uncertainties associated with parameters of the semi-empirical model, thyroid mass-values, 131
I deposition densities, and imprecise responses to questions administered during the personal interview, a set of 1000 individual stochastic doses was generated using MC simulation for each study individual [23
]. We did not further address the dose uncertainty in dose–effect analysis.