Annual atrial tachyarrhythmia burden determined by device interrogation in patients with cardiac implanted electronic devices is associated with a risk of ischaemic stroke independent of known risk factors

Abstract

OBJECTIVES

This study evaluated the risk of non-fatal ischaemic stroke associated with increased long-term cumulative duration of atrial tachycardia (AT).

METHODS

We retrospectively reviewed the records of 260 patients with cardiovascular implantable electronic devices capable of monitoring AT. Patients were separated into zero, low and high AT burden groups. The cut-off point between low and high AT burden was defined by the median value of AT burden in the non-zero AT burden groups (5% in 1 year, about 18 days annually). The primary outcome was non-fatal ischaemic stroke.

RESULTS

The mean patient age was 63.3 ± 13.7 years, the average follow-up was 7.0 years and 10 patients had strokes. Multivariate analysis showed only hypertension and a diagnosis of atrial fibrillation (AF) were associated with stroke. The risk of stroke in patients with hypertension was 12.57-fold higher than in those without hypertension, and was 20.81-fold higher in patients with paroxysmal AF and 162.59-fold higher in patients with chronic AF than in those without AF. Kaplan–Meier analysis showed that stroke-free survival was significantly different in the three AT burden groups (P = 0.002, long-rank test); the rate was greatest in the zero AT burden group, followed by the low AT burden group and was lowest in the high AT burden group.

CONCLUSIONS

Patients who accumulated an AT duration exceeding 5% (18 days) of the total time in any of the 1-year periods are more likely to have an ischaemic stroke than those who have a low or zero AT burden.

INTRODUCTION

Atrial fibrillation (AF) is a common sustained arrhythmia [1], and its prevalence increases with age from 0.1% in adults <55 years old to 9% in those ≥80 years old [1, 2]. AF is often associated with underlying cardiac disease [3]. Classic symptoms of AT/AF include palpitations, chest discomfort, dizziness, breathlessness and syncope; however, AF can occur without any signs or symptoms. Widely considered a potential cause of mortality, non-valvular AF is an independent risk factor for ischaemic stroke [2, 4], possibly owing to the atrio-embolic mechanism [5]. The most serious consequence of AF is embolization, where impaired atrial contraction facilitates the production of a blood clot and subsequent arterial obstruction may cause stroke. Treatment of AF focuses on decreasing the heart rate, restoring normal heart rhythm and anticoagulation to prevent blood clot formation.

Via the assistance of implantable cardiac pacemakers, implantable cardioverter defibrillators (ICDs), and other implantable monitors, physicians can thoroughly document brief, asymptomatic atrial tachycardia (AT)/AF episodes that are common in patients without other evidence of AF. Information such as the number of events and their cumulative duration can be gathered in every single device interrogation. By taking advantage of such a technology, several studies have observed an increased risk of stroke in patients with device-detected atrial tachyarrhythmias [6, 7]. Additionally, Mattioli et al. [8] reported that an increased ventricular pacing percentage was associated with an increased incidence of stroke and AF in patients with pacemakers. Therefore, detecting AT and the ventricular pacing percentage from cardiac implantable electronic devices (CIEDs) may yield information regarding risk stratification [9].

The aim of this study was to evaluate the risk of non-fatal ischaemic stroke associated with increased AT duration, independent of known risk factors, in patients with CIEDs.

MATERIALS AND METHODS

This retrospective study was approved by the Institutional Review Board of Chang Gung Memorial Hospital (certificate number: 100–2620B). To ensure the privacy of the study subjects, all data were reviewed anonymously. Because of the retrospective nature of the study, the requirement of informed consent was waived.

Patients

The medical records of a total of 2544 patients with symptomatic bradycardia (sinus node, atrioventricular node disease, paroxysmal AF or chronic AF) who had Class I or II American College of Cardiology/American Heart Association indications [10, 11] and received CIEDs at Chang Gung Memorial Hospital from December 1993 to May 2011 were reviewed. All patients who received pacemaker implantation and came back for follow-up during the study period and had complete medical records were included in this study. Excluded from this study were patients who had not visited the device clinic for more than 1 year and patients whose device implantation or replacement was performed at another institution. Also excluded from this study were patients with valvular heart disease. None of the included patients had undergone a concomitant ablation.

Device characteristics

The implanted devices were dual- and single-chamber pacemakers, implantable cardioverter defibrillators and cardiac resynchronization therapy devices. The devices used were manufactured by St Jude Medical (USA), Medtronic (USA) and Guidant (USA).

Data collection and follow-up

Based on a review of medical records, various clinical data were collected including clinical information related to known risk factors for stroke (i.e. heart failure, hypertension, age, diabetes mellitus and previous stroke), from initial pacemaker implantation to subsequent device replacement. Ischaemic stroke events were diagnosed by imaging studies and clinical signs and symptoms, with each stroke diagnosis confirmed by a neurologist. Transient ischaemic attacks without image confirmation were excluded from this study. AF was diagnosed by a cardiologist, who reviewed data from a 12-lead surface electrocardiogram (EKG) and device interrogation information. An EKG was performed during device implantation and each device clinic visit. Use of antithrombotic or antiplatelet therapy prescribed by the attending physician was also recorded.

Device data were retrieved from CIEDs during the follow-up in our clinic every 6 months and included the sum of AT (AT, atrial flutter and AF) duration, number of AT incidents, and the sum of atrial and ventricular percentages. AT was defined as an atrial contraction rate exceeding 170–180 bpm at initial implantation. This setting was lowered to 150 bpm if the clinician determined that a stricter interpretation of AT was necessary for patient safety and minimization of ventricular pacing.

Categorization

The patients were separated into three AT burden groups (i.e. zero, low and high) before performing data analysis. As in a previous study [7], this investigation identified the cut-off point between low and high AT burden by using the observed median value of the AT burden in the non-zero AT burden group. The median value was 5% in 1 year, a value that can be converted into an accumulation of AT duration of about 18 days annually.

Statistical analysis

Continuous variables were presented by mean and standard deviation, and categorical variables by count and percentage. Differences between the stroke and the stroke-free groups at baseline were compared using an independent two-sample t-test for the continuous variables, and χ² test or Fisher's exact test with Yate's correction if any cell number was <5 or close to zero for the categorical variables. Univariate Cox proportional hazards models were used to evaluate the correlation between patient variables and stroke event. A multivariate Cox proportional hazards model was performed to identify risk factors for stroke event; independent variables with P < 0.2 in the univariate Cox model were included in the multivariate model. Since some variables including aspirin, warfarin, atrial pacing percentage, ventricular pacing percentage and AT burden may change over time, a Cox model with time-dependent covariates was used in the univariate and multivariate analyses. Furthermore, multicollinearity was also evaluated by the variance inflationary factor (VIF) in the multivariate Cox model. Variables with VIF >5 were considered to have multicollinearity with other covariates and were excluded from the multivariate analyses. The results of Cox regression models were summarized by hazard ratio (HR) with 95% confidence interval (CI). The stroke-free survival and follow-up time among AT burden subgroups were determined using the Kaplan–Meier method, and results presented as a survival curve. In addition, the time-dependent covariates among different disease groups, including AV block and sick sinus syndrome, were examined using stratification analyses. Data were analysed using SPSS 18.0 statistics software (SPSS, Inc., Chicago, IL, USA). A P-value of <0.05 was considered statistically significant.

RESULTS

A total of 260 patients with complete data were included in this study, and their baseline characteristics are summarized in Table 1. The average follow-up period of all subjects was 7.0 years, and the total follow-up time was 1873 person-years. The mean age of the patients was 63.3 ± 13.7 years, and almost half were male (47.3%). Of the 260 patients, 10 had strokes. Patients in the stroke event group were significantly older than in the stroke event-free group (70.5 ± 6.5 vs 63.0 ± 13.8 years, P = 0.006). There were significant differences between the stroke event and stroke event-free groups at baseline with respect to stroke before CIED implantation (30.0 vs 6.4%, P = 0.029), hypertension (80.0 vs 40.4%, P = 0.019), left atrial (LA) diameter (43.3 ± 6.1 vs 38.0 ± 7.9 mm, P = 0.035), CHADS₂ score (2.0 ± 1.2 vs 0.9 ± 1.0, P < 0.001), CHA₂DS₂-VASc score (3.8 ± 1.5 vs 2.0 ± 1.4, P < 0.001), tachy-brady syndrome (50.0 vs 18.0%, P = 0.025), type of clinical AF (20.0, 50.0 and 30.0 vs 82.4, 16.8 and 0.8% for no AF, paroxysmal AF and chronic AF, respectively, P < 0.001), use of aspirin (60.0 vs 18.0%, P = 0.005), use of β-blockers (70 vs 24.4%, P = 0.005) and use of statins (30.0 vs 8.0%, P = 0.048).

Table 2 shows the HRs for the clinical characteristics associated with stroke event by univariate and multivariate Cox regression models based on 1873 person-years. Results of the univariate analysis showed that the time-independent covariates significantly associated with a stroke event were stroke before implantation (HR = 5.39, 95% CI 1.39, 20.89; P = 0.015), diabetes mellitus (HR = 4.76, 95% CI 1.18, 19.16; P = 0.028), hypertension (HR = 6.49, 95% CI 1.38, 30.59; P = 0.018), vascular disease (HR = 6.71, 95% CI 1.70, 26.46; P = 0.007), LA diameter ≥50 mm (HR = 6.39, 95% CI 1.34, 30.54; P = 0.020), CHADS₂ score (HR = 2.40, 95% CI 1.47, 3.89; P = 0.001), CHA₂DS₂-VASc score (HR = 2.18, 95% CI 1.45, 3.27; P < 0.001), tachy-brady syndrome (HR = 4.19, 95% CI 1.21, 14.47; P = 0.024), use of aspirin at baseline (HR = 6.70, 95% CI 1.89, 23.75; P = 0.003), diagnosed with paroxysmal AF (HR = 12.43, 95% CI 2.41, 64.14; P = 0.015, compared with no AF) and diagnosed with chronic AF (HR = 97.77, 95% CI 15.74, 607.31; P < 0.001, compared with no AF).

Univariate analysis showed that the time-dependent covariates associated with a stroke event were aspirin use (HR = 4.59, 95% CI 1.29, 16.35; P = 0.019), high AT burden (HR = 10.97, 95% CI 2.58, 46.69; P = 0.001, compared with zero AT burden) and daily increase in AT burden (HR = 1.01, 95% CI 1.004, 1.014; P = 0.001).

Stroke risk factors with a P-value of <0.2 in univariate analysis were input into the multivariate Cox regression model, and the results showed that only hypertension and a diagnosis of AF were associated with stroke. The risk of stroke in the hypertension subgroup was 12.57-fold (95% CI 1.14, 138.35; P = 0.039) higher than the risk in the group without hypertension. The risk of stroke in the paroxysmal AF subgroup was 20.81-fold (95% CI 1.42, 305.27; P = 0.027) higher than the risk in the group without AF. Moreover, the chronic AF subgroup had a 162.59-fold (95% CI 5.37, 4922; P = 0.003) higher risk of stroke than the group without AF.

The Kaplan–Meier survival curve of stroke-free survival among the AT burden subgroups based on the 1873 person-years data set is shown in Fig. 1. The stroke-free survival rate was significantly different between the three subgroups (P = 0.002, log-rank test); the stroke event-free survival rate was greatest in the zero AT burden subgroup, followed by the low AT burden subgroup and the lowest rate was seen in the high AT burden subgroup.

The HRs of time-dependent covariates for stroke event stratified by AV block and sick sinus syndrome are shown in Table 3. In the subgroup with AV block, associated with a stroke event were aspirin use (HR = 10.41, 95% CI 1.08, 100.56; P = 0.043). In the subgroup without AV block, high AT burden (HR = 16.96, 95% CI 1.83, 157.18; P = 0.013, compared with zero AT burden) and daily increase in AT burden (HR = 1.01, 95% CI 1.003, 1.016; P = 0.004) were associated with an elevated risk of a stroke event. In the subgroup with sick sinus syndrome, high AT burden (HR = 15.70, 95% CI 1.70, 144.86; P = 0.015, compared with zero AT burden) and daily increase in AT burden (HR = 1.01, 95% CI 1.003, 1.016; P = 0.004) were associated with an elevated risk of a stroke event. In the subgroup without sick sinus syndrome, aspirin use change over time was associated with an elevated risk of a stroke event (HR = 11.94, 95% CI 1.23, 115.58; P = 0.032).

DISCUSSION

This very long-term retrospective study is, to our knowledge, the first to explore the relationship between annual AT burden and strokes during a whole device lifetime. The incidence of a stroke during a mean follow-up of 7.0 years was 3.8% in the 266 patients. Patients who had accumulated an AT duration exceeding 5% of the total time in any of the 1-year periods (i.e. ∼18 days in 1 year) were more likely to have a stroke than those who had less AT burden. This trend became more pronounced in the patient subgroup with sick sinus syndrome than in the subgroup with AV block. This study also found that CHADS₂ or CHA₂DS₂-VASc scores were significantly related to the risk of a stroke. Among the various stroke risk factors in the CHADS₂ and CHA₂DS₂-VASc scores, hypertension was found to be the strongest in this study. This study may assist physicians and cardiac surgeons in the care of patients receiving device therapy. When the AT/AF burden is increased, patients may be informed of the increased risk of stroke, and appropriate treatment can be given.

Atrial tachyarrhythmia burden and stroke

This study attempted to define how device-recorded AT burden and stroke are related. Device-detected arrhythmias are commonly found upon device interrogation, and physicians must identify high-risk groups based on device interrogation data during each clinic visit. The AT burden can accurately reflect the level of stroke risk in these patients. Notably, this study differs from previous studies in that it describes, for the first time, the relationship between AT burden and ischaemic stroke on a very long-term basis (average follow-up 7.0 years, 1873 patient-years). Patients with an AT burden exceeding 5% of the total time in 1 year had a higher risk for non-fatal ischaemic stroke, and these patients tended to develop a stroke earlier than patients with a smaller AT burden. This finding is consistent with those of several studies that showed an increased duration of AT/AF is associated with an increased risk of thromboembolism. Healy et al. [6] analysed data from 2580 patients who had no previous diagnosis of AF, and who were implanted with either a cardiac pacemaker or defibrillator with a mean follow-up period of 2.5 years. After a 3-month follow-up of these patients, sub-clinical atrial tachyarrhythmia of more than 5 min was found to be an independent risk factor of ischaemic stroke and systemic embolism. Glotzer et al. [7] followed 2486 patients with cardiac implantable devices in the TRENDS study. With a 1.3-year follow-up, the authors reported that the risk of thromboembolism was 2.5-fold higher in patients who had an AT/AF burden of more than 5.5 h in the previous 30 days. Capucci et al. [12] observed a similar phenomenon; the risk of embolism was increased by 3.1-fold in patients with AF for more than 1 day in a study with a 22-month follow-up.

Characteristics associated with a stroke event

This study identified several clinical features that were related to a higher risk of stroke. Previous embolic events, hypertension and history of vascular disease were all associated with a stroke event, and these findings have been thoroughly examined in previous studies [2, 13]. Multivariate analysis of clinical features and patient outcomes revealed that hypertension and diagnosis of AF are independently associated with a higher risk of developing an ischaemic stroke. Previous studies have consistently demonstrated that previous stroke, hypertension, age and diabetes mellitus are strongly associated with embolic events in AF patients [14]. However, according to our results, hypertension is the only independent, or more hazardous risk factor of stroke that is considered in the CHADS₂ or CHA₂DS₂-VASc scoring systems. This finding is similar to that of a previous Taiwanese nationwide database analysis, which indicated that hypertension is the strongest predictor of stroke in the Taiwanese population with AF [15]. We hypothesize that only hypertension and diabetes mellitus are diseases that have a continuous or cumulative atherosclerotic effect in the CHADS₂ or CHA₂DS₂-VASc scoring system. As mentioned earlier, hypertension is highly predictive of macrovascular events such as strokes in patients with diabetes mellitus [16]. Over an extended time period, as is in this study, hypertension becomes the strongest predictor in the CHADS₂ or CHA₂DS₂-VASc score.

Tachy-brady syndrome is a form of sick sinus syndrome, and its relationship with stroke may be attributed to its tachyarrhythmic component. Univariate analysis in this study revealed that tachy-brady syndrome significantly increased the risk of thromboembolic events, while a similar tendency was not seen in the sick sinus syndrome group, a finding similar to that of previous studies [17]. This phenomenon may be related to the greater AT burden in patients with tachy-brady syndrome.

Atrial fibrillation and risk of stroke

Although patients with persistent AF would appear to be more likely to develop an ischaemic stroke, previous studies have failed to demonstrate that patients with paroxysmal and persistent AF have a different risk of ischaemic stroke [18, 19]. However, conventional intermittent EKG monitoring may underestimate the incidence of AF, since AF can often be asymptomatic [20]. Healey et al. [6] showed that although most patients implanted with either a pacemaker or ICD were free from symptomatic AF, many of them had sub-clinical atrial tachyarrhythmias. Implanted cardiac devices monitor AT continuously, allowing a more thorough examination of how AT burden and ischaemic stroke are related. Our multivariate analysis results indicate that chronic AF patients are more likely to develop strokes than those with paroxysmal AF. However, according to previous studies patients with paroxysmal or chronic AF have the same risk of developing a stroke, but those with chronic AF have worse outcomes after a stroke [21, 22]. The difference in findings between prior studies and ours may be explained by our unique approach to detect paroxysmal AF in which 12-lead resting EKG results and data from CIEDs were used. Therefore, our data should more accurately estimate the incidence in AF than intermittent monitoring alone. This finding could help clinicians to accurately identify high-risk patients and guide subsequent stroke prophylactic treatment.

Baseline clinical characteristics also profoundly impact the ability to identify patients who are at a higher risk of developing a stroke. Designed for predicting the risk of stroke in AF patients [23], the CHADS₂ score was significantly correlated with the risk of developing ischaemic stroke in this study. A previous study demonstrated an increase in the incidence of stroke among patients receiving ventricular pacing [8]. In this study, patients with an atrioventricular block tended to have a higher ventricular pacing percentage than those in the sick sinus syndrome group; however, the incidence of strokes was not different. Additionally, in patients with sick sinus syndrome, the incidence of stroke in patients who had a higher ventricular pacing was not increased.

Anticoagulation and antiplatelet therapy

In this study, 60% of the patients who experienced a stroke were prescribed aspirin upon device implantation; only one of them (10%) received oral anticoagulation therapy. In patients diagnosed as having paroxysmal AF, only 44.7 and 19.1% were administered aspirin and warfarin, respectively, as prevention and only 20.4 and 5.8% of all of the patients were prescribed oral antiplatelet and anticoagulation therapy, respectively. Based on current guidelines for stroke prevention [24], our patient group was not adequately treated. This inadequate treatment may be in part because some of the patient data analysed in this study were from as early as 1993. During this early period, the modern concept of stroke prevention in AF patients had not yet been developed. Because of a concern over a potential bleeding risk, clinicians during that period were more conservative about prescribing patients antiplatelet and anticoagulation therapy. In this study, aspirin was a risk factor for stroke in univariate time-dependent analysis. This seemingly contradictory finding could be because patients receiving aspirin treatment were at high risk for developing a stroke, and the aspirin prescribed was used for secondary prevention of a stroke. The findings of this study are in agreement with those of the AFFIRM investigators [25] in that patients with an increased risk of stroke should continue with anticoagulation therapy, or at least, antiplatelet therapy. Although previous studies have indicated the association between high AT burden and stroke risk, no consensus has been reached on treating patients with a high AT burden [2, 13].

Study limitations

This retrospective study has a number of limitations that should be considered. Patients with incomplete clinical records, who discontinued therapy, or had an incomplete follow-up were excluded; as a result, the patient number was rather limited (n = 260). In patients who were lost to follow-up, the possibility of stroke or death caused by stroke cannot be eliminated. In the setting of the implanted devices used in this study, atrial tachyarrhythmia refers to an atrial rate higher than 150–170 bpm; however, the exact threshold varies between different models. Therefore, our data cannot distinguish between the exact types of atrial tachyarrhythmia. It is important to note that, in our study, the pacemaker could only detect the presence of AT; the diagnosis of AF could not be made simply using pacemaker data. This limitation was recognized at the outset of the study, and is the reason the study focused on the impact of AT/AF burden rather than identifying whether AF or AT have an impact on patient outcome. Lastly, the use of antithrombotic therapy was not consistent in the patients included in the study. However, this accurately reflects the use of antithrombotic therapy in actual practice. Even physicians who specialize in treating arrhythmias will individualize treatment for AF patients, rather than managing them completely in accordance with guidelines. Thus, one can view this point as a highlight of the study as it accurately reflects the realities of the patient treatment.

CONCLUSIONS

This study found that patients who had accumulated an AT duration longer than 5% in any 1-year period during a whole device lifetime were more likely to have an ischaemic stroke than those who had a lower or zero AT burden. Other clinical features that were predictive of a stroke include hypertension and paroxysmal and chronic AF. These results may help to identify patients who will benefit from anticoagulation or antiplatelet therapy.

ACKNOWLEDGEMENTS

Sung-Hao Hung is appreciated for the preliminary design of this project. Chun-Chieh Wang and Hung-Ta Wo are commended for providing the raw patient data from patients. Ted Knoy is appreciated for his editorial assistance.

Conflict of interest: none declared.

REFERENCES

Author notes