Ascertainment bias are common in epidemiologic studies to assess the association between thyroid cancer risk and living near nuclear power plants because many thyroid cancers are diagnosed by chance through health examination. We surveyed the ultra sonography (USG) examination history and conducted thyroid and breast USG in residents living near nuclear power plants.
The study population comprised 2,421 residents living near nuclear power plants in Korea. Information on demographic characteristics, including diagnostic examination history, was collected by interview using questionnaires. USG examination was conducted to evaluate the presence of thyroid nodules and breast lesion. Study participants were divided into 3 groups according to the distance of their respective villages from a nuclear power plant. The proportions of USG examination history and prevalence of thyroid nodules and breast lesions were compared between groups.
Examination histories of thyroid USG were 23.1%, 13.7%, and 10.5% in men and 31.3%, 26.7%, 18.3% in women in the short, intermediate, and long distance groups, respectively. There were significant inverse associations between thyroid USG history and the distance from nuclear power plants (
Our results suggest that there may be an ascertainment bias in population-based studies examining the harmful effects of NPPs examination and researchers should pay attention to ascertainment bias resulted from differential health examination. Correction for ascertainment bias, active follow-up and examination for all study population to remove differential health examination is needed.
The thyroid is highly sensitive to radiation exposure [
Although several population-based studies have assessed the association between thyroid cancer risk and residence distance near NPPs, the results have been inconsistent due to low dose radiation, varying study designs, measurement error, and confounding factors, among others. Furthermore, because many thyroid cancers are diagnosed by chance through health examination, ascertainment bias or detection bias are common in epidemiologic studies.
Recently, a cohort study reported cancer risk in adult residents living near NPPs in Korea, prompting criticism [
We surveyed the thyroid ultrasonography (USG) examination history in residents living near NPPs to assess whether the examination rate varied according to the distance of residences from NPPs. We also conducted thyroid USG to assess whether the prevalence of thyroid nodules varied in patients according to the distance of their residence from NPPs.
In Korea, 25NPPs are being operated commercially in 4 areas (Busan, Gyeongju, Yonggwang, and Ulchin). The study population was selected from residents who live near the Wolsong NPPs in Gyeongju. Since 1983, five power plants have been operational in Gyeongju. Of the 17,066 residents aged 20 yr or older who lived within 20 km of NPP facilities, 2,421 were voluntarily recruited for this study in 2013.
We divided participants into 3 groups according to the distance of the village hall from the NPPs: short (within 5 km), intermediate (6–10 km), and long distance (11–20 km). Sample sizes were allocated; sampling fractions were 20%, 15%, and 10% for short, intermediate, and long distance groups, respectively. We informed villagers about our study through village foremen and enrolled the volunteer participants within the allocated sample size. And then well-trained interviewer visited house of each participants.
Information on demographic characteristics, life style, dietary habits, and health examination history was collected through interviews using questionnaires. Participants were asked to answer the question “Have you ever had a thyroid USG examination?” Participants were asked to respond “Yes”, “No”, or “I do not know”. Participants were asked the same question about history of abdominal USG, breast USG, carotid artery USG, cardiac USG, abdominal computer tomography (CT), chest CT, brain CT, and brain magnetic resonance imaging (MRI).
After excluding missing data, the final study sample was 2,414. Written informed consents were obtained from all participants, and this study was approved by the Ethics Committee of the Institute of Medicine at the Seoul National University.
After 2 months from questionnaire survey, enrolled participants were asked to assemble to have a health check. Of the 2,414 study participants, 1,976 had received a thyroid USG to check for thyroid nodules, and 1,275 women received a breast USG from three expert radiologists. The radiologists assessed the thyroid nodules as yes or no according to presence of nodule more than 1 mm and classified breast USG results according to the Breast Imaging Reporting and Data System (BI-RADS). In the BI-RADS system, C1 is negative, C2 is benign, C3 indicates probably benign, C4 is suspicious abnormality, C5 is highly suggestive of malignancy, and C6 indicates proven malignancy.
Analysis of variance and chi-square tests were conducted to identify differences between groups with regard to the means and proportions of baseline characteristics such as age, sex, cigarette smoking history, alcohol drinking history, household income, education status, regular exercise and self-rated health status. The proportions of participants who received thyroid USG, abdominal USG, breast USG, carotid artery USG, cardiac USG, abdominal CT, chest CT, brain CT, and brain MRI are expressed as percentages and 95% confidence intervals. To assess the association between examination history and distance of residence from NPP facilities in men and women, the
The short, intermediate, and long distance groups comprised 788, 1,313, and 313 study participants, respectively (
History of thyroid USG was 31.3%, 26.7%, and 18.3% among women in the short, intermediate, and long distance groups, respectively (
History of thyroid USG was 23.1%, 13.7%, and 10.5% among men in the short, intermediate, and long distance groups, respectively (
The prevalence of thyroid nodules among men in the short, intermediate, and long distance groups was 10.2%, 11.7%, and 17.0%, respectively (
We estimated the expected prevalence of thyroid nodules in the community by multiplying the proportion of those in the study population with a history of thyroid USG by the positive proportion of thyroid nodule detection in the study population (
There were differences in thyroid and breast USG history according to the distance of villages from the NPPs. However, there was no association between the distance from NPPs and prevalence of thyroid nodules.
Many studies of the effects of high-dose radiation exposure on thyroid cancer were conducted in relation to the Chernobyl accident, atomic bombing areas, and the Fukushima NPP accident. A case-control study conducted in Belarus and Russia reported that the estimated linear coefficient of the excess relative risk of thyroid cancer per gray was 4.5 (95% CI=1.2–7.8) [
However, there were doubts about whether the Chernobyl NPP accident resulted in an increase in thyroid cancer because of artifacts through active case finding, previous underreporting before the accident, unexpected geographic distribution between thyroid cancer and cesium contamination, short latency period, and decrease of thyroid cancer after initial increase [
Because of the harmful effects and epidemiologic discrepancies of radiation, the public has concerns about the safety of low-dose radiation exposure originating from NPPs. In recent decades, epidemiologic studies have been conducted to assess the association between living near NPPs and risk of thyroid cancer. A Korean prospective cohort study of 11,367 adults living within a 5 km radius of NPPs and 24,809 adults living far away from NPPs showed that women living within a 5 km radius of a NPP had a 1.9 times higher risk of developing thyroid cancer compared to the living far away from NPPs [
These discrepancies in results from epidemiologic studies might be caused by low-dose radiation, genetic susceptibility, dietary iodine intake, study designs, bias, definition of exposure, confounding factors, and statistical power [
The residential area of our study population overlapped with the previous Korean cohort study. In Korea, health service accessibility for thyroid USG is relatively good, and the cost for thyroid USG is 30–50 US dollars [
Since 1999, national screening services for gastric, hepatic, colon, breast, and cervical cancer have been provided free of charge or at low cost in Korea. Additionally, most hospitals have promoted their health examination program, including thyroid USG in their programs [
Our study had considerable limitations. One limitation is that our study population was not a representative sample because we recruited volunteer participants. This convenient sampling might result in selection bias. Second, there was the possibility of information bias. For example, recall bias could occur if participants in the short distance group had a tendency to recall previous USG examinations more than participants in the long distance group. Third, because we obtained information on USG examination history through questionnaire interviewing, there might be measurement error due to memory decay.
Our results suggest that there may be an ascertainment bias in population-based studies examining the harmful effects of NPPs examination and researchers should pay attention to ascertainment bias resulted from differential health examination. Correction for ascertainment bias, active follow-up and examination for all study population to remove differential health examination is needed.
There were differences in self-reported thyroid and breast ultrasonography history according to the distance of villages from the nuclear power plants. However, when we conducted thyroid and breast ultrasonography, there was no association between the distance from nuclear power plants and prevalence of thyroid nodules. Our results suggest that there may be an ascertainment bias in studies examining the health effects of nuclear power plants and that active follow-up and examination are needed to assess thyroid cancer in study populations.
C1 is negative, C2 is benign, C3 indicates probably benign, C4 is suspicious abnormality, C5 is highly suggestive of malignancy, and C6 indicates proven malignancy.
Legend: The expected prevalences of thyroid nodules in the community were estimated by multiplying the proportion of those in the study population with a history of thyroid USG (A) by the positive proportion of thyroid nodule detection in the study population (B). The expected prevalence of thyroid nodules and breast masses observed in the community was higher in the short distance group than in the intermediate or long distance groups.
National Cancer Information Center. Korea cancer registry statistics. 2012
Possibility of ascertainment bias according to distance of villages from NPPs. (A) Expected thyroid nodule prevalence of men in the community. (B) Expected thyroid nodule prevalence of women in the community. (C) Expected breast mass prevalence of women in the community.
Baseline Characteristics of Study Population
Variables | Distance of villages from NPPs | |||
---|---|---|---|---|
| ||||
Short (≤5 km) (N=788) | Intermediate (6–10 km) (N=1,313) | Long (11–20 km) (N=313) | ||
Age, mean (SD) | 60.7 (14.2) | 64.6 (10.5) | 62.9 (9.78) | <0.001 |
| ||||
Sex, n (%) | 0.132 | |||
Men | 265 (33.6) | 476 (36.3) | 125 (39.9) | |
Women | 523 (66.4) | 837 (63.7) | 188 (60.1) | |
| ||||
Smoking status, n (%) | 0.003 | |||
Never | 585 (74.2) | 886 (67.5) | 204 (65.2) | |
Past | 99 (12.6) | 234 (17.8) | 53 (16.9) | |
Current | 104 (13.2) | 193 (14.7) | 56 (17.9) | |
| ||||
Alcohol drinking, n (%) | 0.516 | |||
Never | 390 (49.5) | 628 (47.8) | 138 (44.1) | |
Past | 40 (5.1) | 78 (5.9) | 17 (5.4) | |
Current | 358 (45.4) | 607 (46.2) | 158 (50.5) | |
| ||||
Household income per month, n (%) | < 0.001 | |||
–$ 999 | 264 (33.5) | 609 (46.4) | 135 (43.1) | |
$ 1,000–1,999 | 198 (25.1) | 385 (29.3) | 89 (28.4) | |
$ 2,000– | 326 (41.4) | 319 (24.3) | 89 (28.4) | |
| ||||
Educated years, n (%) | <0.001 | |||
–5 years | 203 (25.8) | 363 (27.7) | 77 (24.6) | |
6–12 years | 343 (43.5) | 681 (51.9) | 171 (54.6) | |
13 years– | 242 (30.7) | 269 (20.5) | 65 (20.8) | |
| ||||
Regular exercise, n (%) | <0.001 | |||
No | 590 (74.9) | 972 (74.0) | 266 (85.0) | |
Yes | 198 (25.1) | 341 (26.0) | 47 (15.0) | |
| ||||
Self-rated health status, n (%) | 0.002 | |||
Healthy | 291 (37.0) | 397 (30.3) | 108 (34.5) | |
Intermediate | 284 (36.1) | 456 (34.8) | 103 (32.9) | |
Not healthy | 212 (26.9) | 459 (34.9) | 102 (32.6) |
SD, standard deviation.
Examination History According to Distance of Villages from NPPs in Women and Men
Examination history | Men, % (95% CI) | Women, % (95% CI) | ||||||
---|---|---|---|---|---|---|---|---|
|
| |||||||
Short (≤5 km) (N=265) | Intermediate (6–10 km) (N=476) | Long (11–20 km) (N=125) | Short (≤5 km) (N=523) | Intermediate (6–10 km) (N=837) | Long (11–20 km) (N=188) | |||
Thyroid USG | 23.1 (18.3–27.9) | 13.7 (10.8–16.6) | 10.5 (5.4–15.6) | 0.001 | 31.3 (27.8–34.8) | 26.7 (24.0–29.4) | 18.3 (13.2–23.5) | 0.015 |
| ||||||||
Abdominal USG | 43.9 (38.3–49.5) | 41.1 (36.9–45.3) | 36.8 (28.7–44.9) | 0.279 | 38.7 (35.0–42.4) | 36.1 (33.2–39.0) | 35.0 (28.6–41.4) | 0.732 |
| ||||||||
Breast USG | 41.7 (38.0–45.4) | 26.5 (23.8–29.2) | 30.1 (24.0–36.2) | <0.001 | ||||
| ||||||||
Carotid USG | 8.7 (5.5–11.9) | 4.9 (3.1–6.7) | 8.1 (3.5–12.7) | 0.501 | 6.2 (4.4–8.0) | 4.3 (3.1–5.5) | 4.3 (1.6–7.0) | 0.229 |
| ||||||||
Cardiac USG | 20.1 (15.6–24.7) | 22.7 (19.2–26.3) | 19.2 (12.6–25.8) | 0.842 | 18.1 (15.2–21.0) | 17.8 (15.5–20.1) | 14.0 (9.4–18.6) | 0.139 |
| ||||||||
Abdominal CT | 28.4 (23.3–33.5) | 20.8 (17.4–24.2) | 24.2 (17.0–31.4) | 0.109 | 16.8 (14.0–19.6) | 14.4 (12.3–16.5) | 17.1 (12.1–22.1) | 0.317 |
| ||||||||
Chest CT | 28.0 (22.9–33.1) | 21.3 (17.8–24.8) | 21.0 (14.2–27.8) | 0.054 | 16.2 (13.4–19.0) | 13.1 (11.1–15.2) | 14.4 (9.7–19.1) | 0.102 |
| ||||||||
Brain CT | 19.4 (14.9–23.9) | 14.3 (11.3–17.3) | 16.1 (9.9–22.3) | 0.245 | 13.7 (11.1–16.3) | 13.2 (11.1–15.3) | 11.8 (7.5–16.1) | 0.242 |
| ||||||||
Brain MRI | 23.5 (18.7–28.3) | 20.6 (17.2–24.0) | 20.3 (13.6–27.0) | 0.273 | 24.1 (20.9–27.3) | 22.3 (19.8–24.8) | 22.6 (17.0–28.2) | 0.059 |
USG, ultrasonography; CT, Computer tomography; MRI, Magnetic resonance imaging.
Prevalence of Thyroid Nodule and Breast Lesion According to Distance of Villages from NPPs
Distance to NPPs | ||||
---|---|---|---|---|
| ||||
Short (≤5 km) | Intermediate (6–10 km) | Long (11–20 km) | ||
Men | ||||
Thyroid nodule (+), n (%) | 22 (10.2) | 46 (11.7) | 16 (17.0) | 0.173 |
Thyroid nodule (−), n (%) | 193 (89.8) | 348 (88.3) | 78 (83.0) | |
OR (95% CI) |
1 | 1.16 (0.68–1.99) | 1.80 (0.90–3.61) | |
OR (95% CI) |
1 | 1.02 (0.58–1.76) | 1.81 (0.88–3.71) | |
| ||||
Women | ||||
Thyroid nodule (+), n (%) | 82 (18.6) | 122 (17.8) | 26 (17.6) | 0.166 |
Thyroid nodule (−), n (%) | 359 (81.4) | 565 (82.2) | 122 (82.4) | |
OR (95% CI) |
1 | 0.95 (0.69–1.29) | 0.93 (0.57–1.52) | |
OR (95% CI) |
1 | 0.75 (0.54–1.03) | 0.80 (0.49–1.31) | |
BI-RADS classification of breast mass | ||||
C1, n (%) | 381 (85.8) | 643 (94.0) | 138 (95.2) | |
C2–C3, n (%) | 59 (13.3) | 40 (5.9) | 7 (4.8) | |
C4, n (%) | 4 (0.90) | 1 (0.20) | 0 (0.0) | |
OR (95% CI) |
1 | 0.63 (0.39–1.01) | 0.49 (0.21–1.14) | |
OR (95% CI) |
1 | 0.11 (0.01–1.17) | 0.34 (0.04–3.04) | |
OR (95% CI) |
1 | 0.59 (0.37–0.93) | 0.45 (0.19–1.03) |
OR, odds ratio; CI, confidence interval; BI-RADS, Breast Imaging Reporting and Data System.
Crude odds ratio and 95% CI.
Adjusted for age, cigarette smoking, monthly household income, years of education, regular exercise and self-rated health status.
C2–C3 vs. C1.
C4 vs. C1: The OR was estimated using the Firth method.
C2–C4 vs. C1.