Symptom Severity and Health-Related Quality of Life in Patients with Atrial Fibrillation: Findings from the Observational ARENA Study

Background: Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia and is associated with impaired health-related quality of life (HRQoL), high symptom severity, and poor cardiovascular outcomes. Both clinical and psychological factors may contribute to symptom severity and HRQoL in AF. Methods: Using data from the observational Atrial Fibrillation Rhine-Neckar Region (ARENA) trial, we identified medical and psychosocial factors associated with AF-related symptom severity using European Heart Rhythm Association symptom classification and HRQoL using 5-level EuroQoL- 5D. Results: In 1218 AF patients (mean age 71.1 ± 10.5 years, 34.5% female), female sex (OR 3.7, p < 0.001), preexisting coronary artery disease (CAD) (OR 1.7, p = 0.020), a history of cardioversion (OR 1.4, p = 0.041), cardiac anxiety (OR 1.2; p < 0.001), stress from noise (OR 1.4, p = 0.005), work-related stress (OR 1.3, p = 0.026), and sleep disturbance (OR 1.2, p = 0.016) were associated with higher AF-related symptom severity. CAD (β = −0.23, p = 0.001), diabetes mellitus (β = −0.25, p < 0.001), generalized anxiety (β = −0.30, p < 0.001), cardiac anxiety (β = −0.16, p < 0.001), financial stress (β = −0.11, p < 0.001), and sleep disturbance (β = 0.11, p < 0.001) were associated with impaired HRQoL. Conclusions: Psychological characteristics, preexisting CAD, and diabetes may play an important role in the identification of individuals at highest risk for impaired HRQoL and high symptom severity in patients with AF.

AF-related symptom severity in sufficiently large samples are needed to close this gap. In addition, HRQoL and AF-related symptom severity have rarely been studied as distinct but probably related subjective consequences of AF.
Accordingly, the aim of this study was to identify medical and psychological factors associated with AF-related symptom severity and impaired HRQoL in a large AF population. On the basis of previous research, we hypothesize, controlling for age, sex, and AF type, (a) an independent association between impaired HRQoL and higher AF-related symptom severity; (b) associations between higher AF symptom severity and impaired HRQoL with comorbidities and history of AF treatments (cardioversion, CA); and (c) associations between higher EHRA class and impaired HRQoL with sleep disturbance, higher perceived stress, and higher generalized and cardiac anxiety. The findings of this analysis may help to characterize those individuals at highest risk for high symptom severity and impaired HRQoL and to identify opportunities for screening or treatment that may reduce symptom burden and improve HRQoL in this population.

Study Design and Patient Population
The Atrial Fibrillation Rhine-Neckar Region (ARENA) Project is an observational study of the Foundation Institute for Myocardial Infarction Research (IHF) in cooperation with the Departments of Cardiology of the Hospitals in Ludwigshafen, Heidelberg, and Mannheim, as well as local resident cardiologists and the Heidelberg University Hospital (Department of Clinical Pharmacology and Pharmacological Epidemiology) in Germany. Its aim is to improve patient-centered care and prognosis, particularly in optimizing stroke prophylaxis in patients with AF [25]. Inclusion criteria for this study were residence in the polycentric Rhine-Neckar Metropolitan Region in Germany, confirmed diagnosis of atrial fibrillation, informed consent to participate in the ARENA-project, and age ≥ 18 years [25]. More than 5000 patients were screened, and 2777 patients were included in the ARENA study between August 2016 and December 2018. Of these, 1857 patients returned the baseline questionnaires. We analyzed baseline data of a subset of 1218 patients with complete data for the medical (except for the intake of metoprolol, new-onset AF, ejection fraction (EF), BMI, and HAS-BLED score) and psychosocial variables of interest.
Each subject's demographic profile and cardiac health status (including symptom severity of AF) were assessed directly after enrollment at the recruiting sites (baseline assessment) and were documented in an electronic case report form (eCRF). Questionnaires related to medications, HRQoL, and psychosocial characteristics were then mailed to participants, who completed and returned them by mail. The cohort will be contacted for follow-up once a year for up to 10 years via postal surveys and standardized telephone interviews.
All research was performed in accordance with the ethical principles of the Declaration of Helsinki [26]. Written informed consent was obtained from all the participants. The study team provided sufficient time to the participants to make a decision regarding participation in the study. The study was approved by the Ethics Committee of the Rhineland-Palatine state Medical Association (#837.366.15) and by the Ethical Review Committee of the University of Heidelberg (#B-F-2016-051) and the Mannheim Medical Center (#2016-613N-MA) in Germany. The investigators registered the ARENA study on ClinicalTrials.gov (NCT02978248) on 30 November 2016 (https://clinicaltrials.gov/ct2/show/NCT02978248, accessed on 4 February 2022).

Assessments
Demographic and clinical variables (sex, age, coronary artery disease (CAD), diabetes mellitus (DM), stroke/transient ischemic attack (TIA), chronic kidney disease, CHA 2 DS 2 -VASc score, HAS-BLED score, BMI, EF, new-onset AF, AF type, history of cardioversion/CA, and intake of metoprolol) were assessed at baseline. AF-related symptom severity was assessed by the EHRA symptom classification [27], which was validated and improved by the EHRA in 2014 [9]. This classification system assesses whether AF-related symptoms are present and to what extent they interfere with daily activities (EHRA class 1 'no symptoms', EHRA 2 'mild symptoms', EHRA 3 'severe symptoms', and EHRA 4 'disabling symptoms') [28]. According to the 2020 guidelines of the European Society of Cardiology (ESC), the assessment of EHRA symptoms is of clinical relevance for treatment decisions [10].
Psychosocial factors were assessed with a questionnaire designed specifically for this study. It included items assessing educational and occupational status, different sources of stress (work, home, financial, noise), sleep disturbance, generalized anxiety disorder (GAD) symptoms using the GAD-2 questionnaire [29], and two items from the German version of the Cardiac Anxiety Questionnaire (CAQ-2) [30]. The two items of GAD-2 and CAQ-2 were aggregated to sum scores ranging from 0 to 6 for GAD-2 and from 0 to 8 for CAQ-2. Stress related to home, home, work, and finances were measured with items developed for the INTERHEART study [31].
HRQoL was assessed with the German version of the EQ-5D-5L, a questionnaire for assessing health on five dimensions (mobility, self-care, usual activities, pain/discomfort, anxiety/depression) in five levels each [32]. The particular item values were converted into index-based values (utilities) using the German Value Set for the EQ-5D-5L [33], ranging from −0.661 to 1; higher scores indicate higher HRQoL.

Statistical Analyses
Sample characteristics are presented for the study population and split up by EHRA class. To avoid too many categories with too small numbers of observations, we combined EHRA classes 2a and 2b (i.e., mild to moderate symptoms) as well as EHRA classes 3 and 4 (severe to disabling symptoms) into a single category each. For dimensional variables, means and standard deviations are presented. As not all variables were distributed normally, groups were compared by Kruskal-Wallis tests for independent samples. For categorical variables, absolute numbers and percentages are displayed. Proportions are compared by chi-squared test.
Factors potentially associated with EHRA class were then analyzed using multinomial logistic regression. For analysis of HRQoL as measured by EQ-5D-5L, we used linear multiple regression. All three hypotheses were tested in a regression model controlling for age, sex, and AF type. For EHRA class, the parameters indicate whether the respective input variable affects the probability of having EHRA class 2 or EHRA class 3 or 4 rather than EHRA class 1 (asymptomatic, the reference category), independently of the other variables in the same model. Odds ratios are reported as indicators of effect size. For EQ-5D-5L, the parameter estimates indicate the change in raw values of EQ-5D-5L with each unit change of the respective predictor. As additional indicators of effect size, we report β-values for each parameter. These values were arrived at by z-transforming the continuous (but not the categorical) input variables and the outcome variable prior to model estimation. The reported β-values thus express how many standard deviations the EQ-5D-5L changes with one standard deviation change in the continuous input variables or with a change of categorical input variables from one level to the other, keeping all other variables in the model constant.
Finally, we performed a dropout analysis to compare the included subset of 1218 patients with confirmed AF and complete data for the variables of interest with dropout subset of 1557 with incomplete data, using chi-squared tests for categorical variables and Wilcoxon rank-sum tests for continuous variables.
All statistical analyses were performed with R, version 4.0.2.

Sample Characteristics
We included N = 1218 patients with complete data for all analyzed variables (mean age 71.1 ± 10.5 years, 34.5% female). The most frequent cardiac comorbidity was CAD (75.4%) followed by DM (22.2%). A history of cardioversion was reported in 30.2% of the study population, and 18.4% of the AF patients had a history of CA. Descriptive characteristics for the full sample and for individuals in each EHRA class are displayed in Table 1. In bivariate comparisons, EHRA class was associated with sex, age, AF type, DM, CHA 2 DS 2 -VASc score, new-onset AF, history of cardioversion, history of ablation, generalized and cardiac anxiety, sleep disturbance, stress from work, home, and noise, as well as with HRQoL. Note. AF = atrial fibrillation; BMI = body mass index; CA = catheter ablation; CAD = coronary artery disease; CAQ-2 = Cardiac Anxiety Questionnaire 2-item screener; DM = diabetes mellitus; EF = ejection fraction; EQ-5D-5L = EuroQoL-5D; GAD-2 = Generalized Anxiety Disorder 2-item screener; TIA = transient ischemic attack. Differences by EHRA class in categorical variables were tested with χ 2 tests. Differences in continuous variables were tested with Kruskal-Wallis rank sum tests. * The total number of analyzed participants was 1218 with complete data for medical and psychosocial variables, except of intake of metoprolol (n = 1077), new-onset AF (n = 1212), EF (n = 705), BMI (n = 1198), and HAS-BLED score (n = 1209).

Association of AF-Related Symptom Severity and HRQoL
Regarding the first hypothesis, AF-related symptom severity was not associated with HRQoL when controlling for age, sex, and AF type. Table 2 presents parameters of the multiple regression model predicting EQ-5D-5L by EHRA class.

Medical and Psychological Correlates of AF-Related Symptom Severity
Testing the second hypothesis (Table 3), the probability of having EHRA class 3 or 4 rather than 1 increased with CAD (OR 1.7, p = 0.020). The probability of having EHRA class 2 (rather than 1) increased with history of cardioversion (OR 1.4, p = 0.041). Regarding the third hypothesis (Table 4), the probability of having EHRA class 2 rather than 1 increased with sleep disturbance (OR 1.2, p = 0.016), stress from noise (OR 1.4, p = 0.005), and CAQ-2 (OR 1.2, p < 0.001). The probability of having EHRA class 3 or 4 rather than 1 increased with higher levels of stress related to work (OR 1.3, p = 0.026) and CAQ-2 (OR 1.2, p < 0.001). In a combined multinomial logistic regression model of the significant medical and psychosocial variables from the Tables 3 and 4, all the above-mentioned associations remained significant (Table S1).   Furthermore, the probability of having higher EHRA class increased with female sex. EHRA class 3 or 4 (rather than 1) was associated with persistent (versus paroxysmal) AF (OR 1.6, p = 0.033).

Dropout Analyses
In comparison to the dropout subset of patients with incomplete data, included patients were younger than dropouts (71.1 ± 10.5 y. versus 73.7 ± 11.1 y., p < 0.001); less likely to be female (34.5% versus 39.8%, p = 0.004); slightly less likely to be single (21.

Discussion
We investigated the interrelation of AF-related symptom severity and HRQoL, as well as their associations with medical and psychosocial factors in a sample of 1218 AF patients. Our findings confirmed some-but not all-of our hypotheses. Consistent with the hypotheses, severe to disabling AF-related symptoms (EHRA class 3 or 4 rather than 1) were associated with CAD, perceived stress at work, and cardiac anxiety. Furthermore, mild-to-moderate symptoms (EHRA class 2 rather than 1) were associated with history of cardioversion, cardiac anxiety, sleep disturbance, and stress from noise. As expected, the probability of having higher EHRA class increases with female sex. HRQoL was associated with sex only when psychological variables were not controlled for. Furthermore, impaired HRQoL was associated with CAD, DM, sleep disturbance, financial stress, and generalized and cardiac anxiety. Contrary to our hypotheses, there was no independent association between AF-related symptom severity and HRQoL when adjusting for age, sex, and type of AF. In what follows, we draw on prior literature to outline hypotheses about causal mechanisms potentially underlying the observed relationships.

Relationship between AF-Related Symptom Severity and HRQoL
We found no evidence for a detrimental effect of increased AF-related symptom severity on HRQoL. Although one small study found that AF-related symptom severity was not an independent predictor of HRQoL [22], most studies in this area have found an inverse relationship between EHRA symptom class and HRQoL [8,9,13,34]. This includes two large trials: Schnabel et al. [13] demonstrated (n = 6196) that EHRA symptom classification and its components were moderately correlated with individual EQ-5D-5L items, and ORBIT-AF (n = 10,087) found an inverse association between EHRA class and disease-specific HRQoL assessed by the Atrial Fibrillation Effect on Quality-of-Life questionnaire (AFEQT) [8].
There are several possible explanations for this. First, the current study controlled for multiple potential confounding variables (e.g., age, sex, AF type), while many of the prior trials that found relationships between AF-related symptom severity and HRQoL did not. It is possible that factors not included in those analyses explained-at least in part-the relationship between symptom severity and HRQoL. Second, the current trial included a generic measure of HRQoL, and it may be that this measure did not fully capture HRQoL in patients with AF. Consistent with this idea, Wynn et al. [9] found that while EHRA class was negatively associated with both AF-specific and generic HRQoL, AF-specific HRQoL (measured with the AFEQT) was much more strongly associated with EHRA class than generic HRQoL. Similarly, Essebag et al. [34] suggested that assessments of AF-related QoL may be more specific and sensitive to changes in AF-related symptom severity compared to generic HRQoL scales. However, generic instruments such as the EQ-5D are better suited for cost-effectiveness analyses. It remains unclear as to why patients in our study with EHRA classes 3-4 tended to report even better HRQoL in bivariate analyses than those in lower EHRA classes. Research studies that include both generic and AF-specific measures of HRQoL and that adjust for relevant confounders may be helpful to clarify the relationship between AF-related symptom severity and HRQoL.

Factors Associated with Increased AF-Related Symptom Severity
Our findings concerning associations of medical and psychosocial factors with AFrelated symptom severity appear consistent with existing literature. Next to medical factors such as history of cardioversion and CAD, prior work found that female sex [17,[35][36][37][38] and lower age [37] are related to symptom severity in AF patients. In the RACE II and the FRACTAL studies, women reported more severe AF-related symptoms [39,40]. Several explanations have been described for this observation. First, women with AF have more cardiac comorbidities such as hypertension, heart failure with preserved ejection fraction, and valvular heart disease [36]. Second, longitudinal analyses have found that women with AF have greater impairments in mental health and tendencies for somatization than men, although women continued to experience worse physical function and greater symptom severity even after controlling for somatization [41]. There may be sex differences in expressing and coping with AF symptoms, which may have important implications for treatment of AF.
Several psychological conditions, such as anxiety, depression, or perceived stress, have also been linked to AF-related symptom severity. Charitakis et al. [42] reported that anxiety, left atrial dilatation, and low-grade inflammation were significant predictors of arrhythmia-related symptoms in patients with AF. Another study [43] found that there was an association between a higher AF-related symptom severity and increased anxiety or depression. Interestingly, in both adjusted and unadjusted follow-up analyses, antiarrhythmic drug therapy or CA reduced AF symptom severity, but neither the perception of AF frequency nor the severity of anxiety or depression improved significantly with AF treatment [43]. In another study [44] investigating patients with persistent AF planned for cardioversion, higher levels of emotional distress (anxiety, depression, and perceived stress) were significantly associated with the number and frequency of reported AF symptoms. The authors suggest implementing screening and treatment of emotional distress as a patient-centered approach into cardiological care to reduce attentional bias toward bodily sensations and to influence the success rate of cardioversion.
In our study, disturbed sleep was associated with EHRA class 2 (rather than 1) but not with EHRA class 3 or 4 (rather than 1). Obstructive sleep apnea is an established risk factor for AF, promoting arrhythmogenesis and impairing treatment efficacy [45]. There are only a few studies reporting relationships of other sleep-related problems with AF incidence, such as short sleep duration or frequent nocturnal awaking [46][47][48]. Concerning AF-related symptom severity in particular, Szymanski et al. [21] found that poor sleep quality is highly prevalent in AF patients, and that its prevalence increases with higher EHRA class. Sleep quality may play a role in the pathogenesis and prediction of AF, potentially representing a novel target for prevention or treatment. Alternatively, sleep disturbance may be a somatic symptom of depression or anxiety disorder. In our analyses, we could not adjust for depression, and our anxiety scale (GAD-2) did not assess sleep-related anxiety symptoms.
Finally, there is some evidence that specific psychosocial stressors such as occupational stress [49] or exposure to traffic noise [50] or noise annoyance [51] are risk factors for increased AF-related symptom perception; however, to our knowledge, their relationships with AF symptom severity have not yet been investigated. Further studies should confirm the associations between the AF symptom severity and work-or noise-related stress [52]. Whether psychosocial interventions could impact the symptom severity of AF though a reduction of work-and noise-related stress requires further research.

Factors Associated with Impaired HRQoL
In our study, preexisting CAD and DM were associated with impaired HRQoL. Both cardiac symptoms and increased heart-related attention in CAD patients might help to explain the relationships between CAD and HRQoL. Furthermore, the presence of DM was associated with impaired HRQoL. This finding is consistent with a study among 192 patients with AF in which DM was a negative predictor of improvement in HRQoL between baseline and 1 year follow-up [53]. Metabolic changes and coping with the chronic disease and its behavioral and health consequences long term might be possible explanations for the association between DM and impaired HRQoL in AF patients. Surprisingly, we found no associations between HRQoL and AF interventions such as cardioversion or CA. However, AF interventions were reported retrospectively, and their benefits may have vanished over time.
The key finding of this study is that both higher AF-related symptom severity and impaired HRQoL-although poorly interrelated-are associated with several psychological variables. Interventions that target generalized and cardiac anxiety, stress, and sleep disturbance may be important for improvement of HRQoL in AF patients. For example, in the SMURF study, symptoms of anxiety and depression significantly predicted poor HRQoL in patients with AF [42]. One mechanism underlying the predictive value of anxiety and depression for poor HRQoL might be the overestimation of duration and frequency of AF episodes [42]. Another explanation for impaired HRQoL due to anxiety or depression might be the cognitive interpretation (e.g., catastrophizing) of bodily AF sensations and dysfunctional coping with AF symptoms. Furthermore, a study on patients with paroxysmal AF showed that trait anxiety, psychological stress events, and anxiety symptoms were strong determinants of poor HRQoL [54]. Similarly, Ong et al. [55] showed that anxiety sensitivity was associated with poorer HRQoL, greater symptom severity, and increased distress in AF patients, and the tested models explained 19-40% of the variance in HRQoL and distress [55]. The authors recommended focusing more strongly on the psychosocial aspects in AF management, as those might be critical determinants of patient s HRQoL [54,55].
From a clinical perspective, it may be useful to consider screening for psychosocial factors (e.g., stress; sleep quality; anxiety; and, specifically, cardiac anxiety) in patients with a diagnosis of AF, especially in cases of high AF-related symptom severity and impaired HRQoL. In cases of a positive psychosocial screen, the primary care providers or cardiologists could provide additional information (e.g., educational materials) about stress prevention, anxiety reduction, or sleep hygiene. Collaboration with psychosomatic/psychiatric specialists might be a next possible treatment option of persistent stress, anxiety, or sleep disorders in patients with AF. Such a holistic approach in AF management has the potential to reduce AF-related symptom severity and increase HRQoL.
Finally, an integrated approach including avoidance of stroke, better symptom management, and cardiovascular and comorbidity risk reduction (ABC pathway) could be a holistic way for clinical decision-making steps in AF management [56]. Such a novel approach might reduce AF-related symptom severity and increase HRQoL in AF patients.

Strengths and Limitations
This study has several strengths, the most important one being the large sample size of patients with AF. Another important strength is the concurrent analysis of both medical and psychosocial factors, as well as their associations with symptom severity and HRQoL in patients with AF. Furthermore, our sample consisted of a heterogeneous group of patients with paroxysmal, persistent, or permanent AF.
Notable limitations include the study's cross-sectional design (which prevents the identification of causal relationships), absence of a comparator group of participants without AF, and measurement of HRQoL using a generic scale (EQ-5D-5L) rather than an AF-specific one. In addition, the effect of depression as a potential predictor of symptom severity and HRQoL would have been of interest, but there was no measure of depression in this study. Furthermore, we did not consider data regarding oral anticoagulation from the ARENA study, and we did not adjust our statistical analyses for all important clinical AF factors (e.g., comorbidities or use of oral anticoagulation), which might have led to bias in our results. Finally, the analyzed subset showed several differences from the excluded population including age, sex, marital status, chronic kidney disease, stroke/TIA, newly diagnosed AF, EF, CHA 2 DS 2 -VASc score, HAS-BLED score, EHRA class, cardioversion, CA, DM, cardiac anxiety, and stress related to work. Therefore, the results should be generalized with caution. However, included and excluded patients were similar with respect to the remaining variables.

Conclusions
In conclusion, the novelty of this study is a multidimensional approach using medical data, history of AF treatments, and psychosocial characteristics and their associations with AF-related symptom severity and HRQoL. Our results indicate that it might be important to consider psychosocial factors such as generalized and cardiac anxiety, sleep disturbance, work-related stress, and stress from noise as possibly affecting AF-related symptom severity and HRQoL. Such a holistic approach in AF management has the potential to reduce AFrelated symptom severity and increase HRQoL, and it might inform clinical decisions for invasive treatments.
However, the relationships between psychosocial predictors and symptom severity of AF are still unclear and require longitudinal data analyses. Finally, the impact of psychosocial variables, age, sex, and factors of medical history on HRQoL should be better understood for optimizing the clinical treatment algorithms in AF patients and to improve their HRQoL.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/jcm11041140/s1, Table S1. Parameter estimates of a multiple regression model predicting EHRA scores concurrently from all significant predictors in the models reported in the main manuscript (compare Tables 3 and 4). Table S2. Parameter estimates of a multiple regression model predicting EQ-5D-5L scores concurrently from all significant predictors in the models reported in the main manuscript (compare Tables 5 and 6).