We used health claims data from the National Health Information Database (NHID) of South Korea from 2006 to 2019. The NHID, maintained by the National Health Insurance Service of South Korea, is a public database that covers more than 97% of the Korean population. It includes demographic data, socioeconomic status, lifestyle questionnaire responses, National Health Checkup data, inpatient and outpatient diagnoses based on the International Classification of Diseases (ICD) codes, and medical treatment information [14] . This retrospective cohort study was approved by the Institutional Review Board of Myongji Hospital (IRB No.2024-12-019) and adhered to the tenets of the Declaration of Helsinki. The requirement for informed consent was waived because all data were deidentified and anonymized.

Data of patients diagnosed with DTC between January 2008 and December 2014 were extracted from the NHID. A total of 199,340 patients with DTC were initially identified using the ICD codes for thyroid cancer (C73) and those who underwent thyroidectomy [16] . A 2-year washout period (January 2006 to December 2007) was applied to exclude any previous history of thyroid cancer or thyroidectomy; thus, we identified newly diagnosed cases of DTC. The exclusion criteria were defined as patients with (1) missing demographic data, (2) medullary thyroid cancer (MTC) or anaplastic thyroid cancer, (3) history of levothyroxine use or thyroidectomy (P4551, P4552, P4553, P4554, and P4561) during the washout period, and (4) history of AF before the thyroid cancer index date (Fig. 1). The date of thyroidectomy was defined as the index date. MTC was defined as patients who underwent calcitonin testing more than twice within 1 year after the C73 code was assigned, and anaplastic thyroid cancer was defined as patient death occurring within 1 year of that assignment [16,17] . In total, 113,797 patients with DTC were included in the final cohort.

The control group data were extracted from the NHID using a random number generator. For each patient with DTC, four thyroid cancer-free individuals with available National Health Checkup data were randomly selected and matched by age and sex. To prevent bias due to differences in follow-up periods, we matched the follow-up periods for both groups. Identical exclusion criteria were applied to the control cohort before matching them to those of the DTC cohort.

The primary end point was the first occurrence of AF during the follow-up period. AF was defined as at least two outpatient visits or at least one hospitalization with ICD codes I480–I484 and I489. This definition was adopted from previous epidemiological studies that used data from the NHID [18] .

Diabetes was defined as the presence of ICD-10 codes (E10–E14) in conjunction with the prescription of antidiabetic medications. Hypertension was defined as a combination of ICD-10 codes (I10) and the prescription of antihypertensive drugs. Dyslipidemia was identified using the ICD-10 code (E78), along with the prescription of dyslipidemia medications. Household income was categorized into quartiles, with absolute poverty defined as a monthly household income below the minimum cost of living. Smoking status was classified as nonsmoker, former smoker, or current smoker.

Descriptive statistics were presented as mean values with standard deviations for continuous variables, whereas categorical variables were represented as numerical values and percentages. Group comparisons were conducted using t-tests for continuous variables and chi-square tests for categorical variables. Cumulative incidence rates (IRs) of AF during the follow-up period were plotted using a cumulative incidence function. Follow-up started on the index date and was censored on the date of outcome occurrence, date of death, or the last day of follow-up (December 31, 2019).

Cox proportional hazards regression models were used to compute the adjusted hazard ratios (HRs) when patients in the DTC and control groups were compared. The HRs were adjusted for age, sex, diabetes, hypertension, dyslipidemia, smoking, drinking, income, and body mass index (BMI). Subgroup analyses were performed according to age group (20–39, 40–64, and ≥ 65 yr), sex (male or female), comorbidities, smoking status, alcohol intake, and BMI (< 25, ≥ 25 kg/m²). All statistical analyses were performed using SAS version 14.0, and results with p values < 0.05 were considered significant.

This study included 113,797 patients with DTC and 455,188 matched controls. A comparison of the baseline characteristics of patients with DTC and controls is presented in Table 1. The mean age of the patients with DTC was 49.6 ± 11.1 years, and 15.6% were male. Patients with DTC have a higher prevalence of hypertension, diabetes, and dyslipidemia. They also had a higher BMI and a lower proportion of smokers than those in the control group. Among the patients with DTC, 83.5% underwent total thyroidectomy and 44.2% received RAIT.

During a median follow-up period of 9.0 years (7.0–10.1 yr), a total of 9,184 newly developed cases of AF were noted, with 2,868 in the patients with DTC and 6,316 in the controls (Table 2). The cumulative incidence of AF was higher in the DTC group than in the control group (Fig. 2). The IR per 1,000 person-years of AF was 3.17 in patients with DTC and 1.65 in the control group. In multivariate Cox-regression analysis, patients with DTC had a 2.07-fold higher risk of developing AF (adjusted HR, 2.07; 95% confidence interval [CI], 1.98–2.17), compared to controls.

When analyzing the risk of AF according to the treatment method for DTC, the risk of AF was significantly increased in all treatment groups compared to that in the control group (Table 2, Fig. 2). IR of AF was 3.35 in the total thyroidectomy with RAI group, 3.25 in the total thyroidectomy without RAI, and 2.43 in the lobectomy group. The IR was 4.1 among patients with DTC who underwent total thyroidectomy with RAIT more than double. The adjusted HR for AF was 2.07 (95% CI, 1.95–2.20) in the total thyroidectomy with RAIT group, 2.20 (95% CI, 2.06–2.34) in the total thyroidectomy without RAIT group, and 1.93 (95% CI, 1.72–2.15) in the lobectomy group, compared to the control group.

Results of subgroup analyses according to age, sex, presence of diabetes, hypertension, dyslipidemia, obesity (BMI ≥ 25 kg/m²), and status of alcohol intake and smoking are presented in Table 3. The IR of AF was consistently higher in the DTC group than in the control group across all subgroup analyses. After adjusting for baseline demographic variables, patients with DTC still had a higher risk of AF than controls. In particular, the absolute IR of AF was higher in the older age group (≥ 65 yr) than in the younger ones (40–64 yr and 20–39 yr). However, HR increased in younger age groups. The HR for AF was highest among patients with DTC aged 20–39 years (HR, 5.20; 95% CI, 4.10–6.60), compared to those aged 40–64 years (HR, 2.38; 95% CI, 2.24–2.53) and those aged ≥ 65 years (HR, 1.79; 95% CI, 1.66–1.93). The HR for AF was also higher in males; patients without diabetes, hypertension, or dyslipidemia; smokers and those who consumed alcohol; and patients without obesity than in females; patients with diabetes, hypertension, and dyslipidemia; nonsmokers; nondrinkers; and patients with obesity. Significant interactions were observed between DTC and age, sex, diabetes, hypertension, alcohol intake, and BMI, whereas no significant interaction was found between DTC and dyslipidemia and smoking status.