We retrospectively collected the longitudinal claims data of all individuals with CHD between 2000 and 2010 from the Taiwan National Health Insurance Research Database (NHIRD) (http://nhird.nhri.org.tw/date_01.html). The national health insurance program in Taiwan was launched in 1995, and it universally and successfully provides quality health care at an affordable cost. More than 99.6% of the residents of Taiwan are covered under the program, and medical records are stored in the NHIRD, which is updated biannually. All patients with major diseases, including ACHD, must be registered in the Registry for Catastrophic Illness Patients database (http://nhird.nhri.org.tw/date_01.html).

In Taiwan, suspected ACHDs are referred to cardiologists for echocardiographic diagnosis and treatment, and the majority of them are followed up at medical centers. The diagnosis of ACHDs, whether inpatient or outpatient, is made based on the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes, and the accuracy of the diagnosis is verified by a hospital-based insurance claims data to ensure its validity. Hospitals that file inaccurate claims may be penalized by the National Health Insurance Bureau. In addition, patients who receive a catastrophic illness certificate are exempted from copayments pertaining to their condition in Taiwan. The CHD diagnoses can be classified into cyanotic CHD (tetralogy of Fallot (TOF), ICD-9-CM code 745.2; common truncus, ICD-9-CM code 745.0; double-outlet right ventricle, ICD-9-CM code 745.11; or other cyanotic CHDs, ICD-9-CM codes 745.1, 745.12, 745.3, 746.1, 746.7, and 747.41) and noncyanotic CHD (ventricular septal defect (VSD), ICD-9-CM code 745.4; ostium- or secundum-type atrial septal defect (ASD), ICD-9-CM code 745.5; congenital pulmonary stenosis (PS), ICD-9-CM code 746.02; or other noncyanotic CHDs, ICD-9-CM codes 745.60, 745.6, 746.2, 746.3, 746.82, and 747.1). In addition, patent ductus arteriosus was not included because it is generally not considered to be ACHD.

AF diagnosis was ascertained if an ICD-9-CM code of 427.31 was listed in the secondary discharge diagnosis of stroke hospitalization, in at least one subsequent inpatient claim, or in at least two subsequent outpatient claims. The primary outcome of this study was hospitalization, with a principal discharge diagnosis of stroke events during the study period. The stroke refers to ischemic stroke only (ICD-9-CM codes 433-434), excluding hemorrhagic stroke, cryptogenic stroke and transient ischemic attack (TIA). The diagnostic codes of AF and strokes have been validated in previous NHIRD studies [19, 20, 21, 22, 23].

The comorbidities included hypertension, diabetes mellitus, obstructive sleep apnea, hypothyroidism, congestive heart failure, prior stroke or TIA or thromboembolism, vascular disease, Charlson Comorbidity Index score, and CHA${}_2$DS${}_2$-VASc score. The presence of each comorbidity was defined as having at least two outpatient diagnoses or anyone inpatient diagnosis in the previous year. However, there was one exception for the previous stroke as one component of CHA${}_2$DS${}_2$-VASc score. Previous stroke was defined as anyone inpatient diagnosis prior to the index date.

ACHDs were identified from the Registry for Catastrophic Illness database, which is a sub-database of the NHIRD, between 2000 and 2010. Among ACHDs, the date of the first AF diagnosis (either before or after the CHD diagnosis) was considered the cohort entry date in the AF group. We further assigned the cohort entry date of ACHDs with AF to ACHDs without AF. This assignment approach is termed as “prescription time-distribution matching” which is known to deal with the immortal time bias [24]. We further matched the two groups at a 1:1 ratio based on sex, age group (18–54, 55–64, 65–74, and $\ge$75 years), and CHD type (ASD, VSD, TOF, and others). Furthermore, patients aged 18 years or with ischemic stroke (including transient ischemic attack) before the cohort entry date were excluded. Finally, patients in the AF group were matched to those in the non-AF group through propensity score matching at a 1:1 ratio. The variables included in the calculation of propensity scores were sex, age, ACHD type, CHA${}_2$DS${}_2$-VASc score, obstructive sleep apnea (OSA), hypothyroidism, and the Charlson comorbidity index (CCI) score. The subsequent outcome comparisons were conducted on the matched cohort (Supplementary Fig. 1). Each patient was followed up from the index date to the date of event occurrence, death, or the end of database records (December 31, 2010), whichever occurred first.

The SAS statistical package (version 9.4, SAS Institute Inc., Cary, NC, USA) was used for all statistical analyses. The risk of stroke events between the AF ACHD and non-AF ACHD groups was compared using the Cox proportional hazard model. Further, patients who underwent mechanical valve surgery are obligatory treated with warfarin lifelong, therefore a sensitivity analysis by excluding the patients who received mechanical valve surgery was conducted (the matching was re-performed).

Furthermore, subgroup analysis was performed based on the age group and CHD type. Finally, after excluding patients receiving antiplatelet therapy, we evaluated whether warfarin use was associated with ischemic stroke occurrence in ACHDs with AF. ACHDs with AF were subgrouped based on warfarin use and CHA${}_2$DS${}_2$-VASc score (0–1 vs. $\ge$2). A p value of 0.05 was considered statistically significant.

A total of 36,530 ACHDs were enrolled. Compared with the general population, ACHDs have an increased prevalence of AF in the 55–65 age group (up to 12.1 times in men and 15.3 times in women; Fig. 1).

Fig. 1.

Difference in the prevalence of atrial fibrillation (Afib) between the general population and adults with congenital heart diseases (ACHDs). (A) Male population. (B) Female population.

The baseline characteristics of ACHDs in the AF and non-AF groups after matching are presented in Table 1. The mean age was approximately 54 years, with no significant differences in age, sex, CHA${}_2$DS${}_2$-VASc score , OSA, hypothyroidism, or CCI scores between the two groups. The risk of ischemic stroke events was higher in ACHDs with AF (adjusted hazard ratio [HR] = 1.45, 95% confidence interval [CI] = 1.11–1.92, p = 0.0093) than in those without AF (Table 2 and Fig. 2). Moreover, the analysis results remained consistent with the primary analysis even after excluding individuals who underwent mechanical valve surgery (adjusted hazard ratio [HR] = 1.82, 95% confidence interval [CI] = 1.31–2.54, p = 0.0004; Table 2).

Fig. 2.

Incidence of ischemic strokes in adults with congenital heart disease with and without atrial fibrillation (Afib) in the overall population and the population aged 50 years. (A) Overall ischemic stroke incidence. (B) Ischemic stroke incidence in the population aged 50 years. y/o, years/old.

In ACHDs aged 50 years, those with AF had a higher risk of ischemic stroke (adjusted HR = 1.46, 95% CI = 0.81–2.63, p = 0.2071) than those without AF (Table 2 and Fig. 2).

Among ACHDs with AF, the subgroups of ASD, VSD, TOF, and other CHDs demonstrated 1.57, 1.29, 2.29, and 1.66 times higher risk of ischemic stroke, respectively, compared to the corresponding subgroups in the non-AF group (Table 3, Fig. 3). Although the trend remains consistent, statistically significant ischemic stroke risk was observed only in the ASD subgroup with AF, even in individuals under the age of 50 (adjusted HR = 1.57, 95% CI = 1.11–2.21, p = 0.0097) (Table 3, Fig. 3).

Fig. 3.

Incidence of ischemic strokes in the population with atrial septal defect (ASD) and ventricular septal defect (VSD). (A) Overall ischemic stroke incidence (Left: ASD; Right: VSD). (B) ischemic stroke incidence in the ASD population aged 50 years. Afib, atrial fibrillation; y/o, years/old.

Mixed type CHD is defined as the presence of two or more diagnosis CHD subtypes in a patient, and the ratio of this mixed type CHD is approximately 33% of the overall ACHD population (Table 1). Patients with mixed type CHD with AF had higher ischemic stroke risks than patients with mixed type CHD without AF (HR = 1.62, 95% CI = 1.01–2.6, p = 0.0457; Table 3).

After the exclusion of patients who underwent antiplatelet therapy (in the original cohort before propensity score matching), patients with AF were divided into two subgroups according to the CHA${}_2$DS${}_2$-VASc score (0–1, $\ge$2, and overall). The baseline characteristics of warfarin therapy groups with lower and higher scores are listed (Supplementary Table 1). The annual incidence of ischemic stroke among those who use warfarin is presented in Table 4. Among those who do not use warfarin, the incidence of ischemic stroke among those with CHA${}_2$DS${}_2$-VASc scores of 0–1 and $\ge$2 was 1.17% and 2.56%; among those who use warfarin, the incidence of stroke among those with CHA${}_2$DS${}_2$-VASc scores 0–1 and $\ge$2 was 2.22% and 4.69%, respectively.

Our study showed that ACHDs (regardless of their sex or age) had a high prevalence of AF, particularly young women. However, the results of the AnTicoagulation and Risk Factors in Atrial Fibrillation (ATRIA) Study were different, wherein AF prevalence was 0.95% (95% CI, 0.94%–0.96%), and it was more common in men than in women (1.1% vs. 0.8%; p 0.001) [25]. According to a Swedish database study of ACHDs, the risk of intra-atrial re-entrant tachycardia or AF is 22-fold higher in patients with CHD than in matched controls, with a prevalence of 8.3% in 42-year-olds [26]. In the current study, we found that AF prevalence was higher in ACHDs than in the general population in Taiwan. Furthermore, in ACHDs, the prevalence of AF increases with age and remains higher than that in the general population, with the highest prevalence observed in individuals under the age of 65. This may explain the increased stroke risk among young ACHDs in Taiwan.

Furthermore, among ACHDs, patients with AF have a higher ischemic stroke risk than those without AF. The Framingham study in 1991 demonstrated that AF is an independent risk factor for stroke in the general population [14]. Although the mechanism is unclear, ACHDs have some unique risk factors that are suspected to be associated with thromboembolic events, such as cyanosis, Fontan circulation, intracardiac shunt, and heart defect complexity [27]. Thus, thromboembolic risk in ACHDs might be underestimated. In the current study, among ACHDs, those with AF had nearly 1.5 times higher risk of ischemic stroke than those without AF. This trend was the same in the younger population (age 50 years).

Previous data on ACHDs from Asia and Europe showed that the overall stroke incidence was higher in ACHDs than in the general population, but details regarding stroke subtypes were limited [15, 28]. In this study, the overall ACHDs with AF as well as ASD, mixed type CHD subgroups showed a significantly increased risk of ischemic stroke. Importantly, in the young population (age 50 years), the trend was the same in the ASD group, which has not been reported in previous studies. Therefore, our study suggests that increased ischemic stroke incidence in ACHDs with AF even at young age, particularly ASD.

Previous attempts to use the CHA${}_2$DS${}_2$-VASc score to predict thromboembolic risk in patients with CHD with AF have shown conflicting results [15]. Our study revealed that ACHDs with AF have a higher incidence of ischemic stroke, even if the CHA${}_2$DS${}_2$-VASc score is low (2). Traditionally, the CHA${}_2$DS${}_2$-VASc score has been used to determine whether patients with AF can benefit from anticoagulation therapy. In 2014, experts recommended the use of the CHA${}_2$DS${}_2$-VASc score to determine whether to use anticoagulation therapy in patients with CHD without prosthetic valves or significant valve disease [17]. However, in patients with CHD with prosthetic valves or significant valve disease and patients with moderate or severe CHD with intra-atrial re-entrant tachycardia or AF, using anticoagulants directly was recommended instead of CHA${}_2$DS${}_2$-VASc score evaluation [17]. However, because ACHDs are few, sufficient comparative studies to evaluate the benefits and safety of anticoagulants are lacking. Few reports are available on the use of anticoagulant drugs in ACHDs with AF. Our study showed an annual risk of ischemic stroke incidence of approximately 1.47% and 3.06% at CHA${}_2$DS${}_2$-VASc scores of 2 and $>$2, respectively. These results suggest that ACHDs with AF have a high risk of ischemic stroke, and that even those with a low CHA${}_2$DS${}_2$-VASc score.

This study has several limitations. First, information on many risk factors, such as smoking, obesity, metabolic syndrome, ventricular function, and severe valvular disease, was unavailable in the claims database, which may confound the results. Additionally, information on the International Normalized Ratio (INR) level and medical compliance of warfarin usage were unavailable in the database. Second, it is known that a patent foramen ovale (PFO) can increase the risk of ischemic stroke. However, the ICD-code based diagnosis cannot differentiate between ASD and PFO. To overcome this limitation and identify patients with ACHD in our study, we used the Registry for Catastrophic Illness database, a sub-dataset of the Taiwanese NHIRD that includes detailed clinical information and requires expert audited approval with formal medical echocardiography or cardiac catheterization reports. Using this rigorous criterion, we are confident that simple PFO cases were not classified as ACHD in our study. Third, detailed AF subtypes could not be validated from this cohort due to the limitations of the database. Moreover, information on the complexity and clinical functional status of CHD subtypes was limited. Fourth, it is important to acknowledge that this study has a retrospective observational design, which introduces the possibility of confounding factors influencing the outcomes of anticoagulation therapy. Despite the significantly high risk of ischemic stroke observed in ACHD patients with AF, as evidenced by the data presented in Table 4, it is likely that clinical and ethical biases have influenced the lack of protective effect observed with warfarin treatment. It should be noted that the decision to use warfarin in Taiwan is based on physician judgment, and this can be a challenging task given the higher propensity for bleeding and substantial fluctuations in therapeutic plasma levels among the Asian population. Close monitoring of anticoagulation therapy is crucial, especially in young populations [29]. Moreover, it is worth mentioning that the time spent within the therapeutic International Normalized Ratio (TTR) range is lower in Taiwan compared to other countries, even when considering data from randomized controlled trials such as RE-LY and ROCKET-AF [30, 31]. Furthermore, the baseline characteristics between the warfarin and non-warfarin groups were not balanced due to the limitation of small event numbers, which may have influenced the observed protective effect of warfarin. These findings raise questions about the potential benefits of anticoagulant therapy with warfarin in the ACHD population in Asia, specifically in Taiwan, and highlight the need for further prospective studies to address this issue. Additionally, in Taiwan, reimbursement policies do not allow the prescription of novel oral anticoagulants (NOACs) for stroke primary prevention in young adults with congenital heart disease (ACHDs) and atrial fibrillation. In 2020, a systematic review of the literature on NOAC use in ACHDs was conducted, and the results indicated that NOACs are safe and effective in ACHDs without mechanical prostheses [32]. Later, the international NOTE registry revealed that NOACs are safe and may be effective for thromboembolic prevention in adults with heterogeneous forms of congenital heart disease [33]. Further studies, such as the ongoing PROTECT AR (Apixaban in Adults With Congenital Heart Disease and Atrial Arrhythmias: the PROTECT-AR Study) trial, are necessary to confirm whether ACHDs with simple or complex diseases may benefit more from NOACs than from warfarin [34]. Finally, the study cohort sizes were small with few stroke events, but the number of ACHDs has been increasing in the past decade, and the management of clinically crucial issues such as AF and ischemic stroke in this population is important.