Abstract
Background:
The incidence of thyroid cancer is increasing worldwide due to an increased detection of small well-differentiated papillary thyroid carcinomas. This study aimed to evaluate the effect of screening with ultrasonography on deaths from thyroid cancer.
Materials and Methods:
We conducted a matched case–control study using data from two sources representative of the adult Korean population. Cases were selected from the National Epidemiologic Survey of Thyroid Cancer database, and controls were selected from the Korean National Cancer Screening Survey database. Controls were individually matched to case patients with respect to age, sex, and area with a ratio of 10:1. The primary outcome was death from thyroid cancer. Controls were required to have been alive on the date of thyroid cancer diagnosis in the corresponding case.
Results:
The analysis included 120 patients who died from thyroid cancer and 1184 controls. Compared with those who had never been screened, the odds ratios for death from thyroid cancer among those who had been screened were 1.44 (95% confidence interval [CI] 0.68–3.05) if cases with missing information on screening were excluded and 1.13 [CI 0.49–2.63] if all cases were included, and missing information was imputed. Stratification by sex, year of diagnosis, and histological type did not show any statistically significant relationships between screening with ultrasonography and death from thyroid cancer, regardless of the statistical model used.
Conclusions:
Screening for thyroid cancer with ultrasonography does not prevent death from thyroid cancer; therefore, screening asymptomatic adults for thyroid cancer is unwarranted.
Introduction
Globally, the incidence of thyroid cancer has increased in the past few decades (1). Despite the rapid increase in incidence, thyroid cancer mortality has remained stable or is declining in most countries. This discrepancy between incidence and mortality is due to increased detection of papillary thyroid cancer that has a better prognosis than other types of thyroid cancers (2). Although no changes in risk factors such as iodine intake or radiation exposure have been observed, the increase in thyroid cancer incidence has been attributed to the increased use of screening and diagnostic techniques such as thyroid ultrasonography and ultrasound-guided fine needle aspiration, respectively (3).
In Korea, the incidence of thyroid cancer has increased rapidly. A total of 44,631 new cases of thyroid cancer were diagnosed and there were 373 deaths from thyroid cancer in 2012 (4). Especially, the increase in the incidence has been particularly rapid among women, from 11.9 per 100,000 in 1999 to 122.1 per 100,000 in 2012. The proportion of thyroid cancer of the small-sized papillary type has steadily increased (5). Since 2004, thyroid cancer has been the most common cancer among women in Korea. In 2014, a heated debate began over the overdiagnosis and overtreatment of thyroid cancer (6).
Several guidelines and recommendations on screening, diagnosis, and treatment for thyroid cancer have been revised and updated recently (7 –9). After reviewing the evidence on the benefits and harms of screening for thyroid cancer in asymptomatic adults, the U.S. Preventive Service Task Force recommended against screening for thyroid cancer in asymptomatic adults (7). However, there remains an evidence gap regarding screening for thyroid cancer. The most important consideration when deciding whether or not to screen is whether screening reduces mortality (10). To date, no study has been conducted that has compared mortality from thyroid cancer in screened and unscreened populations directly (7).
There are several challenging issues in designing and conducting a study to evaluate the effectiveness of thyroid cancer screening: (1) ethical considerations, such as overdiagnosis and overtreatment, and (2) the large-scale sample size or long-term follow-up period required to obtain a sufficient number of health outcome events. The goal of this study was to investigate the associations between the uptake of thyroid ultrasonography and thyroid cancer mortality using two national surveys.
Materials and Methods
Case identification and control selection
We designed a case–control study using two data sources: cases were selected from the National Epidemiologic Survey of Thyroid cancer (NEST), and controls were selected from participants of the Korean National Cancer Screening Survey (KNCSS). The Korea Central Cancer Registry (KCCR), which contains >95% of all newly diagnosed malignancies in Korea (11), conducted the NEST in 2011 to investigate secular trends in the clinicopathological features of 5796 thyroid cancer patients in Korea (12). Since 2014, the data set has been available to the public (
Cases in this study were defined as adults who were newly diagnosed with thyroid cancer (International Classification of Disease, 10th Revision [ICD-10] codes C73) and who died from thyroid cancer by December 2014. For death ascertainment, all cases were followed for at least six years. The incidence of thyroid cancer and thyroid cancer mortality was confirmed by reviewing records maintained by the KCCR database and death certificates from Statistics Korea, respectively. Informed consent was waived for the cases on the grounds that the analysis used routinely collected data from the KCCR.
The National Cancer Screening Program (NCSP), launched in Korea in 2002 (13), currently provides cancer screening for gastric, colorectal, liver, breast, and cervical cancer. The KNCSS, a nationwide cross-sectional survey, which is conducted annually to investigate behavioral patterns with respect to cancer screening, focuses mainly on gastric, liver, colorectal, breast, and cervical cancers. In addition, the survey contains several questions on screening tests for other cancers such as lung, prostate, and thyroid cancers. Cancer-free men aged ≥40 years and cancer-free women aged ≥30 years were eligible to take part in the KNCSS. Details of the survey have been described elsewhere (14). Participants in the KNCSS provided informed consent.
Among the participants of KNCSS surveys conducted between September 2014 and April 2015, we used random sampling to select individual controls for each thyroid cancer case, who were alive on the date of death of the corresponding case. Statistical power in matched case–control studies depends on the prevalence of exposure among controls (15). If the prevalence is <15%, then >5 controls per case will help to increase statistical power. To maximize statistical power, 10 controls were individually matched to each case based on age, sex, and area of residence. This study was approved by the Institutional Review Board of the National Cancer Center, Korea (NCCNCS08129).
Exposure to thyroid ultrasonography
In this study, we defined exposure to screening for thyroid cancer as an ultrasonographic test of the thyroid conducted for the purpose of screening, not diagnosis. A retrospective review of medical records of thyroid cancer cases was performed by a trained medical record administrator, using a structured form. Based on the information in the medical record, the method of detection of thyroid cancer was categorized as detected on screening, detected clinically, or method of detection unknown. In controls, face-to-face interviews were performed by trained interviewers to collect information on their thyroid cancer screening history. After showing and explaining the pictures of an ultrasonographic procedure on thyroid, participants were asked, “Have you ever undergone screening for thyroid cancer” and “When were you screened?” We defined the index date as the date on which thyroid cancer was diagnosed in the case subject. To ensure similar follow-up times among each pair of matched cases and controls, we restricted the follow-up period for uptake of screening for thyroid cancer before the index date of the corresponding case. For cases, we defined ever-screened as screening detected on record review, and in controls, we defined ever-screened as answering “yes” to the interview question on thyroid cancer screening, with a screening date before the index date of the corresponding case.
Statistical analysis
To investigate the effects of thyroid cancer screening on prevention of death from thyroid cancer, case–control analyses were performed using conditional logistic regression models. We used conditional logistic regression to calculate odds ratios (ORs) with 95% confidence intervals. Some cases had missing information on the mode of detection of their thyroid cancer. We used two approaches for dealing with missing values: omitting all cases with missing information (Model I) and using multiple imputation (Model II). We used a fully conditional specification method as the imputation model (16). All p values <0.05 were considered statistically significant. All statistical analyses were conducted using SAS software, version 9.4 (SAS Institute, Cary, NC).
Results
Among the 120 cases, there were 3 cases for which we did not find 10 matched controls, finding 3, 5, and 6 controls, respectively. A total of 120 cases and 1184 controls were included in the analyses. The mean age of diagnosis among the cases was 65.4 years (range: 31 to 86 years), and 80 out of 120 cases (66.7%) were women. The most common histological type (67.5%) was papillary carcinoma (Table 1). Among the 1184 controls, the overall proportion of ever-screened was 9.4%.
Characteristics of the Study Participants
A total of 209 controls were not included because they were screened for thyroid cancer after the index date.
Overall, conditional logistic regression did not find a statistically significant association between screening and death from thyroid cancer using either Model I or Model II (Table 2). Stratified analyses using Model I found that participants in the age <65 years subgroup who had been screened had a significantly higher risk of death from thyroid cancer, but there were no significant associations between screening and death from thyroid cancer in any of the other subgroups. Stratified analyses using Model II did not find a statistically significant association between screening and risk of death from thyroid cancer in any of the subgroups.
The Risk of Death Due to Thyroid Cancer Among Participants Screened with Thyroid Ultrasonography Compared with Unscreened Participants
Model I: ORs and 95% CIs were determined using conditional logistic regression excluding cases with missing information on the mode of detection.
Model 2: ORs and 95% CIs were determined using conditional logistic regression with missing information on mode of detection replaced by imputed values.
CI, confidence interval; OR, odds ratio.
Discussion
This study demonstrates that thyroid screening with ultrasonography does not significantly affect the risk of death from thyroid cancer. Results from subgroup analyses by sex, age group, and histological type also did not show any associations between screening with ultrasonography and the risk of death due to thyroid cancer. Therefore, these findings provide evidence that there is no benefit to thyroid cancer screening with ultrasonography to reduce thyroid cancer mortality.
This study has several limitations. First, it had an observational study design and was not a randomized controlled trial. Therefore, we were unable to control fully for potential biases, such as misclassification bias. About 30% of cases had no information on the mode of tumor detection. Participants in the control group were asked to recall whether they had ever been screened for thyroid cancer. The long recall period could have caused recall bias. However, the temporal trends in thyroid cancer screening in cases and controls were similar to those reported in other studies (5,17).
Second, the number of cases may have provided insufficient statistical power to draw conclusions. We increased the number of matched controls per case to increase statistical power and minimized the number of cases who had to be excluded because of missing information on the mode of detection by using a multiple imputation model. We performed post hoc power analysis for conditional logistic regression. The results had various powers (from <10% to >90%) to detect the ORs. However, considering the annual number of deaths from thyroid cancer in Korea (111 men and 258 women in 2017), the number of cases in our study is not small (4).
Third, we could not adjust for socioeconomic status, which is known to be a strong risk indicator for death due to cancer in Korea, including death due to thyroid cancer. However, by using representative samples and by matching cases individually to randomly selected controls, we minimized the possibility of confounding due to differences in socioeconomic status.
This study had several strengths. To the best of our knowledge, this is the first study to investigate the associations between the uptake of thyroid ultrasonography and thyroid cancer mortality. Moreover, we used nationally representative data sets to source both cases and controls, and were able to control confounding and some biases, such as information bias and selection bias, methodologically. Our results could be applicable to other countries or regions with increasing trends of thyroid cancer incidence.
The NCSP, an organized cancer screening program launched in 2002, has included all Korean men and women of specific age ranges. Unlike the United States and Europe, opportunistic cancer screenings are also prevalent in Korea, in addition to organized screenings. The increase in the incidence of thyroid cancer has been associated with an increase in mammography screening for breast cancer as part of the NCSP. Owing to their dense breasts, women aged 40–60 years who participated in the NCSP were recommended to undergo breast ultrasonography, which often led to an opportunistic screening of the thyroid with ultrasonography (5,18 –20).
The incidence of thyroid cancer peaked in 2012 before the debate began on over-diagnosis in Korea and has subsequently sharply dropped. However, of all cancers, the incidence of thyroid cancer in Korea is still ranked second in women and 6th in men, and 26,051 new cases of thyroid cancer were diagnosed in 2016 (11). The Korean guidelines for screening and management for thyroid nodules and cancer were revised in 2015 (21,22). However, despite the debate concerning thyroid cancer screening, the general public is still not adequately informed about the latest recommendations, and many people undergo thyroid ultrasonography as opportunistic screening (23,24). More active interventions are needed to make providers and the population comply with the guidelines.
In conclusion, our study demonstrates that screening with thyroid ultrasonography does not prevent death from thyroid cancer. Our results strengthen the evidence that screening for thyroid cancer in asymptomatic adults is unwarranted. However, further research is needed that would reflect the changes in clinical practice made as a result of updated guidelines and recommendations related to thyroid cancer.
Footnotes
Author Disclosure Statement
No competing financial interests exist.
Funding Information
This study was supported by a Grant-in Aid for Cancer Research and Control from the National Cancer Center of Korea (Grant No. 1910233-1).