Physical Activity Changes and the Risk of Incident Atrial Fibrillation in Patients With Type 2 Diabetes Mellitus: A Nationwide Longitudinal Follow-up Cohort Study of 1.8 Million Subjects

Type 2 diabetes mellitus (T2DM) is a common form of diabetes that is characterized by glucose homeostasis dysregulation and has a substantial disease burden worldwide (1). Notably, T2DM is associated with incident cardiovascular events such as myocardial infarction, stroke, atrial fibrillation (AF), and heart failure (2–6). Thereby, both pharmacologic and nonpharmacologic strategies for patients with T2DM have been investigated to improve clinical outcomes by managing cardiovascular morbidities (5,7). Despite these efforts, the impact of T2DM on cardiovascular diseases is still unacceptably high (8), suggesting the need for further tailored management to improve clinical prognosis.

Atrial fibrillation (AF) is the most common clinical arrhythmia and one of the leading causes of stroke and dementia (9,10). As patients with T2DM are at increased risk of incident AF (4,11), various studies have investigated the predisposing factors for incident AF in these patients. For example, alcohol consumption and smoking are significant risk factors for newly developed AF (12,13). Physical activity (PA) or regular exercise is also known as a protective factor for AF (12). However, few data are available on whether increment or decrement changes of PA in patients with T2DM could influence the risk of incident AF. Additionally, there is a lack of data on how much PA could be recommended as an optimal amount of PA to reduce incident AF risks in patients with T2DM.

Using nationwide epidemiologic data from the Korean National Health Insurance Service (NHIS) database, we aimed to investigate the prognostic value of PA alterations on incident AF risks in patients with T2DM. Second, we studied the cutoff levels of PA for incident AF prevention.

This study was approved by the Seoul National University Hospital Institutional Review Board (IRB No. E-2107-012-1232) and conducted in accordance with the Declaration of Helsinki. As anonymized data from NHIS were analyzed, the need for informed consent was waived.

We used data from the Korean NHIS database for this nationwide, population-based cohort study. The profile of the NHIS database was previously reported elsewhere (14,15). In brief, the NHIS is a single public insurer that covers the entire Korean population. It also recommends that Korean adults undergo general health examinations biannually. Based on this national health policy, we designed this study, which had a criterion of the 2-year interval between the first and second general health examinations. Thereby, individual demographic information, history of diagnoses, and results of the health examinations have been collected in the NHIS database. Individuals’ medical records of diagnoses were coded according to the ICD-10-CM.

The study design and flowchart of study subjects are shown in Fig. 1. We identified 2,746,988 patients diagnosed with T2DM in general health examinations between 1 January 2009 and 31 December 2012. Individuals who were newly diagnosed with T2DM at health examinations and those diagnosed with T2DM were both included. This study included adult patients (aged ≥30 years) who underwent both first and second general health examinations in a 2-year interval before 31 December 2012. We excluded subjects who had a history of AF before the second general health examination or during a 1-year lag from the second health examination. Patients with missing data during the public health examinations were excluded. Finally, 1,815,330 subjects with T2DM were included and followed up until December 2018. The index date was the date of the second health examination, and data on baseline characteristics were collected from the index date.

Patients with T2DM were defined as follows: 1) subjects having at least one claim per year for a prescription of antidiabetes medication under ICD-10-CM codes (e.g., E11–14), from the insurance claims data, or 2) subjects having a fasting blood glucose ≥126 mg/dL in the general health checkup with a prescription of oral hypoglycemic agents or insulin (14). Antidiabetes medications consisted of metformin, sulfonylureas, meglitinides, dipeptidyl peptidase 4 inhibitors, thiazolidinediones, α-glucosidase inhibitors, and insulin. Care was taken to exclude patients with type 1 diabetes (ICD-10-CM code E10). Medications were assessed at the index year, and T2DM duration was measured from the first diagnosis of T2DM up to the index date.

When individuals underwent the first and second general health examinations, data on the intensity and frequency of PA in each health examination were collected via a self-reported questionnaire. The survey questionnaire used for PA evaluation was adopted and modified from the International Physical Activity Questionnaire (IPAQ), which was reported by the World Health Organization. The validity and reliability of this survey have been validated in previous studies (16,17). Briefly, the questionnaire section on exercise was composed of three questions asking about the frequency of 1) light-intensity PA (i.e., walking slowly) for >30 min, 2) moderate-intensity PA (i.e., brisk-pace walking or bicycling leisurely) for >30 min, and 3) vigorous-intensity PA (i.e., running or climbing) for >20 min during a recent week.

In this study, regular PA “yes” was defined as undergoing vigorous-intensity PA for >20 min or moderate-intensity PA for >30 min, at least once weekly. On the contrary, regular PA “no” was defined as not engaging in any moderate- or vigorous-intensity PA, regardless of the amount of light-intensity PA (17). The regular PA status on the first and second general health examinations was used to divide the study population into four groups: persistent nonexercisers (no to no), new exercisers (no to yes), exercise dropouts (yes to no), and exercise maintainers (yes to yes).

To further explore the amount of PA that could be recommended to reduce the risks of incident AF, we analyzed the hazards of incident AF according to stratified energy expenditure, as previously reported (18,19). By calculating energy expenditure by METs and time, we stratified patients into <500, 500–999, 1,000–1,499, and ≥1,500 MET-min/week according to energy expenditure.

Demographic data, anthropometric data, and a previous history of hypertension, dyslipidemia, and T2DM-related factors, such as diabetes duration and history of diabetic retinopathy, were collected at the time of the index date. In addition, data on lifestyle behaviors, T2DM duration, use of insulin, and use of two or more oral hypoglycemic agents were also collected. Laboratory results were obtained from the second general health examination. Self-reported questionnaires evaluated lifestyle behaviors such as smoking and alcohol consumption.

We defined incident AF as the primary end point (Fig. 1A). This end point was defined based on the ICD-10-CM codes with additional conditions. Detailed definitions of comorbidities and study end point are provided in Supplementary Table 1 and were previously published elsewhere (20). Follow-up duration was defined as the interval between the index date and the first occurrence of the incident AF.

Data are presented as numbers and frequencies for categorical variables and as means ± SDs for continuous normally distributed variables. The χ² test or Fisher exact test was used for categorical variables, as appropriate. One-way ANOVA was used to analyze continuous variables between more than two groups. The annual event incidence rates were calculated as the number of events per 1,000 person-years. Multivariate Cox proportional hazard regression models were used to estimate hazard ratios (HRs) and corresponding 95% CIs for the associations between PA alteration and incident AF. Age, sex, BMI, hypertension, dyslipidemia, diabetes duration, glomerular filtration rates, fasting glucose level, use of insulin, use of oral hypoglycemic agents, income level, smoking history, and alcohol consumption level were adjusted as covariates. We analyzed risks of incident AF with a 1-year lag period. Cox models were used to conduct separate subgroup analyses according to age, sex, T2DM duration, presence of diabetic retinopathy, and use of insulin. A two-sided P value of <0.05 was considered statistically significant. All statistical analyses were performed using SAS 9.4 software (SAS Institute, Cary, NC).

Under reasonable request, data are available through approval and oversight by the Korean NHIS.

In total, 1,815,330 patients with T2DM (mean age, 58.4 ± 11.5 years; male, 1,112,222 [61.3%]) were included and analyzed. Hypertension was present in 1,029,451 subjects (56.7%), dyslipidemia in 789,750 (43.5%), and diabetic retinopathy in 194,603 (10.7%). Regarding diabetes medication, 155,161 patients (8.6%) were treated with insulin and 789,640 patients (43.5%) were treated with two or more oral hypoglycemic agents. Among these individuals, 408,369 participants (22.5%) were engaged in regular PA.

Patients were categorized into four groups according to the change of PA: persistent nonexerciser (n = 1,181,837 [65.1%]), exercise dropouts (n = 225,124 [12.4%]), new exerciser (n = 242,968 [13.4%]), and exercise maintainer (n = 165,401 [9.1%]). Table 1 demonstrates the baseline characteristics of each group. The persistent nonexerciser group showed female preponderance and short diabetes duration, while the exercise maintainer group showed male preponderance and long diabetes duration.

During the mean follow-up of 5.6 ± 1.3 years, 46,589 cases of incident AF occurred. Baseline characteristics according to incident AF are presented in Supplementary Table 2. In brief, patients with incident AF were older, more frequently diagnosed with hypertension, and had longer diabetes duration than those without incident AF.

The risk of incident AF was decreased in individuals who performed regular PA at any time of the first or second health examination compared with those without regular PA (Fig. 2A and Supplementary Table 3). Compared with the persistent nonexerciser group, both the exercise dropout group (adjusted HR 0.96, 95% CI 0.94–0.99) and new exerciser group (HR 0.95, 95% CI 0.93–0.98) had lower risks of incident AF. The exercise maintainer group showed a lower risk of incident AF than the persistent nonexerciser group (HR 0.91, 95% CI 0.89–0.94). As a sensitivity analysis, we additionally analyzed the data without a 1-year lag period for incident AF, and the data are presented in Supplementary Table 4, showing consistency with the main analysis.

We performed subgroup analyses according to age, sex, diabetes duration, presence of diabetic retinopathy, and use of insulin. Across all subgroups, the beneficial association of performing regular PA to prevent incident AF was observed: both the exercise dropout group and the new exerciser group tended to show lower risks of incident AF compared with the persistent nonexerciser group. The exercise maintainer group showed the lowest risk of incident AF among the groups (Supplementary Table 5).

There was a significant association between performing regular PA and age for incident AF (P = 0.004). Those with regular PA showed a lower risk of incident AF, especially those with age ≥65 years, whereas this association was attenuated in those with age <65 years. Also, the association between regular PA and the risk of AF seemed more prominent in women than in men (P = 0.053). Regarding diabetes-related factors, regular PA seemed to be associated with lower risks of incident AF in patients with longer diabetes duration (≥5 years), in those with diabetic retinopathy, and in those treated with insulin, although statistical significance was attenuated.

To identify the optimal amount of PA to reduce the risks of incident AF, we further stratified patients with T2DM according to energy expenditure (Fig. 2B and Table 2). Those with energy expenditure ≥1,500 MET-min/week in the new exerciser group showed a significantly lower risk of incident AF (HR 0.93, 95% CI 0.89–0.97) than the persistent nonexerciser group, while those with energy expenditure <1,500 MET-min/week in the new exerciser group did not show statistical significance. In the exercise maintainer group, those with both energy expenditure of 1,000 to <1,500 MET-min/week and ≥1,500 MET-min/week showed a significantly lower risk of incident AF compared with the persistent nonexerciser group (HR 0.92, 95% CI 0.88–0.96 and HR 0.91, 95% CI 0.86–0.95, respectively). Baseline characteristics of the persistent nonexerciser group, new exerciser group, and exercise maintainer group according to PA amount are presented in Supplementary Tables 6 and 7.

In this study of the associations between PA changes and the risk of incident AF, our principal findings are as follows: 1) the persistent, nonexerciser group had the highest risk of incident AF, whereas the exercise maintainer group demonstrated the lowest risk; 2) patients with regular PA seemed to be associated with decreased risks of AF in elderly patients, in those with longer diabetes duration (≥5 years), in those with diabetic retinopathy, and in those treated with insulin, while significant interaction was only observed between age and PA changes; and 3) optimal PA ranges based on energy expenditure of ≥1,500 MET-min/week and ≥1,000 MET-min/week were associated with lower risks of incident AF in those who initiated and who maintained regular PA, respectively.

As both T2DM and AF confer a substantial health care disease burden and cost, the association between both has been comprehensively investigated. In the Framingham Heart Study, T2DM had an independent association with increased risk of AF (10), which was reaffirmed in a meta-analysis (4). Longer T2DM duration and worse glycemic control further increase the risks of AF (21,22). Indeed, T2DM is not only independently associated with the risk of incident AF, but the coexistence of T2DM and AF also contributes to a grave prognosis (23). Although the use of metformin and sodium–glucose cotransporter 2 inhibitors proved to have therapeutic benefits in reducing incident AF risks (24,25), the use of some other oral hypoglycemic agents, such as sulfonylurea, dipeptidyl peptidase 4 inhibitors, insulin, and glucagon-like peptide 1 receptor agonists, have not been independently associated with reduced risks of incident AF (26–28).

Recently, risk-factor modification by lifestyle changes has become an important strategy to prevent AF (29,30). In addition to pharmacologic approaches to reduce incident AF, nonpharmacologic lifestyle modification approaches should be considered as part of the guideline-recommended (31) holistic or integrated care management approach to AF care (32,33). For example, alcohol abstinence has been associated with protective effects for incident AF patients with T2DM (13).

Nonetheless, the relationship between PA and T2DM is more complex. As one of the lifestyle management strategies for patients with T2DM, PA has demonstrated its beneficial effects on managing blood glucose levels and body weight (5). In addition, performing PA was associated with improved arterial elasticity, which also has protective effects against incident AF risks (34). On the contrary, excessive PA might lead to atrial enlargement, inflammation, remodeling, and autonomic activation, which could increase the risks of incident AF (35–37). Considering these together, a nonlinear association is suspected between PA and incident AF, whereby the appropriate amount of PA might reduce AF risks, while endurance exercise (i.e., marathons or long-distance triathlons) could increase AF risks.

In the current study, we found that PA alterations with starting and maintaining regular PA were both associated with a lower risk of incident AF, and optimal PA ranges based on energy expenditure that were associated with lower risks of incident AF can be defined. To our knowledge, there is a paucity of data that comprehensively explore the associations between PA alteration and incident AF in patients with T2DM. Our semiquantitative analysis could suggest a reference point for future prospective studies exploring the optimal intensity and amount of exercise to prevent AF.

There are also several limitations to this study. First, this study was an observational study, albeit a large one, and there could be unmeasured confounders, reverse causality, and biases that might have an influence on results. For example, there is a possibility of selection biases in individuals attending the first general health examination and undergoing the second general health examination within a 2-year interval. Unmeasured confounders might have an influence on results during the follow-up. However, given that performing randomized trials that require and prohibit regular PA are not likely, carefully managed observational studies could be valuable alternatives.

Second, we did not include patients with type 1 diabetes, so the results obtained could not be extrapolated to patients with type 1 diabetes. Further studies are suggested to explore the prognostic implication of PA changes in patients with type 1 diabetes.

Third, this study screened and included all participants based on general health examinations between 2009 and 2012. This time point is before the introduction of sodium–glucose cotransporter 2 inhibitors and glucagon-like peptide 1 analogs in Korea, both of which proved substantial cardiovascular risk reduction. Further, cardiorespiratory fitness and muscular strength could be considered as surrogate markers of PA. However, these variables would not be available in our claims database. Considering the positive association of cardiorespiratory fitness and exercise training with improved clinical outcomes (38), further studies are needed to explore the relationship between cardiorespiratory fitness, muscular strength, and risks of incident AF in patients with T2DM.

Finally, we could not analyze the relationship between PA and the type and burden of AF, as related data were not available in the NHIS database.

In this large nationwide cohort study, both initiating and maintaining regular PA were associated with lower risks of incident AF in patients with T2DM. When the amount of PA was stratified according to energy expenditure, optimal energy expenditures of ≥1,000 MET-min/week in exercise maintainers and ≥1,500 MET-min/week in new exercisers could be encouraged to reduce incident AF risks compared with nonexercisersy.

Funding. This work was supported in part by the Korea Medical Device Development Fund grant funded by the Korean government (the Ministry of Science and ICT, the Ministry of Trade, Industry and Energy, the Ministry of Health & Welfare, the Ministry of Food and Drug Safety) (Project No.: HI20C1662, 1711138358, KMDF_PR_20200901_0173), and by a grant from the Patient-Centered Clinical Research Coordinating Center (PACEN) funded by the Ministry of Health & Welfare, Republic of Korea (grant no. HC21C0028).

Duality of Interest. E.-K.C. reports research grants or speaking fees from Abbott, Bayer, BMS/Pfizer, Biosense Webster, Chong Kun Dang, Daewoong Pharmaceutical Co., Daiichi-Sankyo, DeepQure, DREAMTECH Co., Ltd., Jeil Pharmaceutical Co. Ltd, Medtronic, Samjinpharm, Seers Technology, and Skylabs. G.Y.H.L. is a consultant and speaker for BMS/Pfizer, Boehringer Ingelheim, and Daiichi-Sankyo. No fees were received personally. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. C.S.P. contributed to the conception and design of the work, data interpretation and analysis, and drafting of the manuscript. E.-K.C. contributed to conception, design, data acquisition and interpretation, and critical revision of the manuscript. K.-D.H. and J.Y. contributed to the data acquisition, analysis, and critical revision of the manuscript. H.-J.A., S.K., S.-R.L., S.O., and G.Y.H.L. contributed to the conception and design of the work and critically revised the manuscript. E.-K.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.