Epidemiology of subclinical atrial fibrillation in patients with cardiac implantable electronic devices: A systematic review and meta-regression

Background: In recent years, attention to subclinical atrial fibrillation (SCAF), defined as the presence of atrial high-rate episodes (AHREs), in patients with cardiac implantable electronic devices (CIEDs), has gained much interest as a determinant of clinical AF and stroke risk. We aim to perform a systematic review and meta-regression of the available scientific evidence regarding the epidemiology of SCAF in patients receiving CIEDs. Methods: PubMed and EMBASE were searched for all studies documenting the prevalence of AHREs in patients ( n = 100 or more, < 50% with history of AF) with CIEDs from inception to 20th August 2021, screened by two independent blind reviewers. This study was registered in PROSPERO: CRD42019106994. Results: Among the 2614 results initially retrieved, 54 studies were included, with a total of 72,784 patients. Meta-analysis of included studies showed a pooled prevalence of SCAF of 28.1% (95%CI: 24.3-32.1%), with high heterogeneity between studies (I 2 = 98%). A multivariable meta-regression was able to explain significant proportion of heterogeneity ( R 2 = 61.9%, p < 0.001), with age and follow-up time non-linearly, directly and independently associated with occurrence of SCAF. Older age, higher CHA 2 DS 2 -VASc score, history of AF, hypertension, CHF, and stroke/TIA were all associated with SCAF occurrence. Conclusions: In this systematic review and meta-regression analysis, SCAF was frequent among CIED recipients and was non-linearly associated with age and follow-up time. Older age, higher thromboembolic risk, and several cardiovascular comorbidities were associated with presence of SCAF.


Introduction
In 1958 Senning and Elmqvist implanted the first permanent cardiac pacemaker (PM) to Arne Larsson, who survived for more than 40 years and ultimately died from different causes [1].Since then, the use of cardiac implanted electronic devices (CIEDs) is widespread thanks to the technological advances that progressively introduced implanted cardioverter-defibrillator (ICD) and cardiac resynchronization therapy (CRT) with only pacing activity (CRT-P) or also defibrillation (CRT-D) [2], with increasing clinical indications [3][4][5].Despite some geographical variability [6], in recent years increasing use of CIEDs is evident [2,7].The increasing use of CIEDs and atrial leads for sensing purposes led to the identification of a clinical entity, the CIED-detected atrial highrate episodes (AHRE) [8].The occurrence of AHRE, which is nowadays assimilated as the term of 'subclinical atrial fibrillation (SCAF), is defined as asymptomatic atrial tachyarrhythmias detected only with long-term continuous cardiac monitoring and not through usual electrocardiographic means [8].In this clinical context, AHRE/SCAF has been associated with an increased risk of developing clinical AF and an increased risk of stroke and systemic embolism [8,9].Thus far, a highly variable incidence of AHRE/SCAF has been reported, basically depending on patients' clinical characteristics [8].
In this study, we performed a systematic review of all studies reporting about AHRE/SCAF prevalence in patients with CIEDs and provide pooled estimates to obtain a comprehensive assessment of its' epidemiology.Second, we performed a meta-regression analysis to investigate study setting and clinical factors that would be more likely associated with AHRE/SCAF.

Methods
This systematic review was performed according to the 'Meta-analysis Of Observational Studies in Epidemiology' guidelines [10] and reported according to the 'Preferred Reporting Items for Systematic Reviews and Meta-Analyses' guidelines [11].A protocol for this systematic review has been registered into the international prospective register of systematic reviews PROSPERO (Center for Reviews and Dissemination, University of York) (CRD42019106994).
We performed a systematic and comprehensive literature search PubMed and EMBASE databases from inception up to 20th August 2021.The search was performed combining the terms 'AHRE', 'SCAF', and 'CIED'.The full search strategy is reported in the Supplementary Materials (Supplemental Table 1).

Studies selection
Two co-authors (MB and GFR) independently screened the search results.Disagreements were resolved by discussion with a third author (MP).All articles retrieved from the searches were evaluated according to titles and abstracts sequentially.From here on, for simplicity and consistency, only the term SCAF will be used.
Full-text eligibility was assessed independently by two co-authors (MB and GFR).Disagreements were resolved by discussion with a third author (MP).All full-texts that: (i) reported data about the prevalence of SCAF in patients implanted with PM, ICD, CRT-P or CRT-D; (ii) evaluated SCAF according to a reliable assessment of episodes; (iii) included <50% of patients with a history of AF; and (iv) included at least 100 patients were included.Exclusion criteria were: (i) conference abstracts, letters, comments, case reports, editorials; (ii) studies not published in English; (iii) all the studies that defined SCAF only based on algorithms (e.g., atrial mode switch episodes).To improve consistency among the studies included in the analysis, we also excluded studies reporting only SCAF of very short or long duration (i.e., <30 s and >15 min).

Data extraction and quality assessment
Data were extracted independently by two of the co-authors (MB and GFR), alongside the supervision of a third author (MP).The following data were extracted: sample size, SCAF events and follow-up time, SCAF definition, type of CIED implanted, geographical location, study design and patients clinical characteristics in the overall cohort and according to SCAF presence (age, sex, body mass index [BMI], hypertension, diabetes, coronary artery disease [CAD], chronic heart failure [CHF], history of atrial fibrillation, history of stroke/transient ischemic attack [TIA], chronic kidney disease, left ventricular ejection fraction [LVEF], baseline treatments).SCAF definition was subdivided according to a atrial rate (<180 bpm, 180-200 bpm, >200 bpm and other/unclear definition) and a duration (<5 min, 5, 6 min, >6 min, and other/unclear definition) criterion, according to the cut-offs adopted by the included studies.
All studies were evaluated independently to assess risk of bias by two co-authors (MB and GFR), according to the Newcastle-Ottawa Scale for cross-sectional studies, composed of 5 items across three domains (Selection, Comparability, Outcome), with a maximum of 5 points.Any study with a score equal or less than 3 was categorized as high risk of bias.

Data synthesis and analysis
We calculated the pooled prevalence of SCAF as reported in the original studies included, with a generalised linear mixed model (random intercept logistic regression model) [12] using logit transformation of proportions.
Mean age, CHA 2 DS 2 -VASc, BMI and LVEF differences, number of males, and number of patients with a history of hypertension, diabetes mellitus, CAD, CHF, history of stroke/TIA, and history of atrial fibrillation were pooled and compared between SCAF and non-SCAF patients using random-effect models.For continuous outcomes, mean, SD and total number in each group were pooled and compared with the inverse variance method.Pooled estimates were reported as Odds Ratios (OR) and 95% confidence intervals (CI), or mean difference and 95% CI for continuous variables.
The inconsistency index (I 2 ) was calculated to measure heterogeneity.According to pre-specified cut-offs, low heterogeneity was defined as an I 2 of <25%, moderate heterogeneity as I 2 between 25 and 75%, and high heterogeneity when I 2 was >75%.
For the pooled prevalence rate of SCAF, a pre-specified sensitivity analysis was performed with a "leave-one-out" approach, in which all studies are removed iteratively one at a time to evaluate their influence on the pooled estimate and heterogeneity.
To account for potential sources of heterogeneity in the pooled prevalence of SCAF, we performed several subgroup analyses, according to geographical location, study design, risk of bias, atrial rate and duration cut-offs for SCAF, and type of SCAF definition (manual review vs. device-based).
We also performed a multivariable meta-regression, with relevant baseline characteristics of the included studies.To account for the potential non-linear relationship between continuous variables and pooled effect size, we fitted meta-regression with the use of restricted cubic splines [13], with default placement of 3 knots.All analyses were conducted with R version 4.0.3, using the 'meta', [14] 'metafor', [15] 'dmetar' [16] and 'rms' [17] packages.

Results
After the electronic search, we retrieved 1486 articles from PubMed and 1128 articles from EMBASE.After the selection process, a total of 205 full-text articles were assessed for inclusion in the systematic review and meta-analysis (Supplemental Fig. 1).Fifty-four papers were included in the analysis, with a total of 72784 patients.[1871] Baseline characteristics of the studies included are reported in Table 1.
Among the papers included 8 papers derived from randomized controlled trials; 30 from observational retrospective studies; and 16 Twenty-three studies involved patients implanted with PM; 7 studies with patients implanted with CRT; 4 studies involved patients implanted with ICD; and remaining 20 studies involved patients which were not selectively implanted with a specific type of CIED.Of note, 31 studies adopted manual confirmation of SCAF, while 13 studies used device-based definition, and the remaining 10 articles used other or unclear assessments.Different durations were used to define SCAF.Four studies used duration of ≥6 min; in 13 cohorts a duration of <5 min was adopted, while 32 used cut-offs between 5 and 6 min.In 5 studies, other or unclear duration of SCAF was reported.As for the atrial rate, 10 studies adopted a cut-off of less than 180 beats per min (bpm); 7 studies used a cut-off over 200 bpm, 22 studies used a cut-off comprised between 180 and 200 bpm, and in 15 studies other or unclear velocity cut-offs were used.Overall, 15 (28%) studies were considered at significant risk of bias.All the other studies (39 out of 53) were considered at low risk of bias (Supplemental Table 2).

SCAF prevalence
The overall prevalence of SCAF across the 54 included studies included was 28.1% (95% CI: 24.3-32.1%),with a high heterogeneity detected (I 2 =98.3%) (Fig. 1).According to the pre-specified 'leave-oneout' analysis reported in Supplemental Fig. 2, we did not find any significant influence of individual studies on pooled estimates or heterogeneity.
Subgroup analyses for the prevalence of SCAF are reported in Table 2.We did not observe any significant differences across the subgroups explored, except for a non-significant trend for higher SCAF prevalence in patients with previous history of AF.

Meta-regression analysis
A multivariable model comprised of cut-offs of SCAF duration and atrial rate, type of study, level of bias, the inclusion of patients with previous history of AF, and age and follow-up times (both modelled as restricted cubic splines) was able to explain a significant proportion of the heterogeneity observed (R 2 =61.9%, p<0.001;Supplementary Table 3).A graphical representation of the marginal relationship between follow-up times and age, modelled as restricted cubic splines, and the prevalence of AHRE is reported in Fig. 2, panel A and B, respectively.Prediction of SCAF prevalence according to the multivariable model, according to mean age and different follow-up times, is reported in Fig. 3.

Clinical variables associated with SCAF
We calculated pooled estimates for several characteristics comparing patients with SCAF vs. patients without SCAF (Table 3).The occurrence of SCAF was significantly associated with older age, higher CHA 2 DS 2 -VASc score, and clinical history of atrial fibrillation, hypertension, heart failure, and history of stroke/transient ischemic attack; however, male sex, body mass index, diabetes and coronary artery disease were not associated with SCAF.We did not observe any significant association between SCAF and pharmacological treatments.

Discussion
This systematic review and meta-regression analysis about the epidemiology of SCAF in patients with CIEDs found an overall prevalence of 28.1% of patients presenting AHREs among the over 72000 patients with CIEDs.While no significant differences were found in prevalence among the subgroups examined, meta-regression analysis found that both patients' age and length of follow-up were significantly, independently and non-linearly associated with SCAF prevalence.Furthermore, the analysis of clinical characteristics revealed that patients presenting SCAF, beyond having a more frequent AF history, were older, more likely hypertensive, affected with congestive heart failure and with an history of stroke/transient ischemic attack.Accordingly, SCAF patients had an overall higher CHA 2 DS 2 -VASc score [Central Illustration].
Our findings have important epidemiological and clinical implications.First, the pooled estimate we provided adds an important piece of information in understanding the natural course of this condition.Thus far, no accurate evaluation of SCAF epidemiology has been provided by any study.Available studies have shown large variability in reporting the occurrence of SCAF [72] and differences in studies design, patients' characteristics and follow-up duration limit the generalizability of any of the previous studies to the general population of patients receiving a CIED implant.
The finding that 28% of patients receiving a CIED report SCAF after receiving the device is highly relevant.First, it gives us a reliable appraisal of the real "size of the problem".Second, it allows us to make some considerations in terms of patients' care.Indeed, the presence of SCAF significantly increases the risk of developing incident clinical AF.Some recent estimates showed that presence of SCAF entails more than 3-fold higher risk of clinical AF [9].Overall, the findings that almost 1 in every 3 patients receiving a CIED could be at high risk for developing clinical AF clearly underlines the importance of performing more accurate monitoring for this condition and the need for proper dissemination of information about its impact on patients.
Furthermore, as shown in the ASSERT trial and further underlined by other studies and meta-analyses, the presence of SCAF increases the risk of stroke by more than 2 fold and the risk of all-cause death and other outcomes, particularly in those patients with a high baseline thromboembolic risk [8,9,22,29,73].Data from European Society of Cardiology countries reported that almost 750,000 patients received a CIED [2], extrapolating our findings suggests that more than 210,000 patients would report SCAF and consequently have such a higher risk for clinical AF and stroke, also progressively increasing with age and throughout time, with important implications for patients' management and health-care services resources use.These numbers highlight the need for further data regarding the use of oral anticoagulants in patients with SCAF.While not all the guidelines discuss this issue, use of oral anticoagulant is recommended only in patients with high stroke risk and longer (≥24 h) episodes of AHRE [72,74,75].Uncertainty still exists in patients with shorter AHRE episodes.Two currently ongoing randomized trials will provide important evidence to elucidate this important issue [76,77].
Our findings regarding the relationship between increasing age and SCAF occurrence underlines how SCAF shares similar risk factors with clinical AF, and extends our knowledge on SCAF epidemiology beyond the estimate of prevalence per se.Indeed, is well-known that increasing older age is a pivotal risk factor for AF [78], and our multivariate analysis elucidates that the contribution of increasing age to the risk of developing SCAF is pivotal.Also, the non-linear association between follow-up time and SCAF prevalence suggests that the observation period is likely to play a role in determining the prevalence observed in CIED recipients.The combination of increasing age and longer follow-up, concurrently with the higher prevalence of comorbidities, which we found associated with presence of SCAF, could underline how these factors combine together to determine the development of an atrial arrhythmogenic substrate [78].Indeed, also the association with increasing age could be a proxy of a generally more complex and impaired clinical situation with no specific risk factor driving the occurrence of SCAF (as underlined by the results of meta-regression analysis), but with the overall progressively higher amount of exposure to risk factors determining the arrhythmogenic substrate.As the occurrence of clinical AF is multifactorial and related to various risk factors [79], the relationship between age, exposure and comorbidities together lead to SCAF, through the mechanism of developing atrial cardiomyopathy and further strengthening the relationship between SCAF, clinical AF and higher risk of thromboembolic events [78,[80][81][82].
These findings may have significant implications in the management of patients receiving CIEDs, underlining the importance of close observation for SCAF occurrence.Further studies are required to expand and corroborate these hypotheses, and to clarify whether specific subgroups of patients may have a different risk(s) of SCAF.Indeed, it should be noted that few studies focused on very elderly CIED recipients, and therefore our estimates for this group of patients were broad, underlining the need for specific studies aiming at evaluating the epidemiology of SCAF in elderly patients, as well as other high-risk subgroups which may experience a different burden of SCAF.
We also observed that patients presenting with SCAF were found to have a higher burden of risk factors and comorbidities, eventually resulting in a higher CHA 2 DS 2 -VASc score, conferring greater thromboembolic risk compared to patients without SCAF.This evidence supports the current knowledge about the increased risk of stroke and thromboembolic events in patients with SCAF, underlining the need for more data about oral anticoagulant and supporting current recommendations on prescribing oral anticoagulant drugs in SCAF patients with very high thromboembolic risk [72].
In the latest European Society of Cardiology clinical guidelines on the management of patients with AF, the nosological entity of SCAF is by definition identifiable only in patients who have no previous history of AF [72].In our meta-analysis we also included patients with previous history of AF.While also other meta-analyses about SCAF included patients with previous history of AF [9], the approach of excluding patients with history of AF has been only recently adopted, hence excluding all studies with such patients would have excluded large part of previous evidence about SCAF.While overall 22/54 studies (41%) included patients with previous history of AF, only 10/54 studies (19%) included more than 25% of patients with previous history of AF, thus with a limited contribution of the group of previously diagnosed AF patients in most of these cohorts.While the subgroup analysis reported a non-significant trend in higher SCAF prevalence in those patients with previous history of AF, the presence of SCAF was clinically more associated with the previous history of AF.

Limitations
The main limitation is the high heterogeneity reported in our pooled estimates.However, high heterogeneity is a common concern in epidemiological meta-analysis exploring the prevalence of several conditions, in which we expect the results to vary consistently in each study [83][84][85].This also reflect the clinical heterogeneity found in clinical practice, with SCAF prevalence highly likely to be influenced by several determinants.Furthermore, we performed additional analyses to account for heterogeneity, including the multivariable meta-regression which allows us to account for roughly 60% of the observed heterogeneity in the pooled estimate for SCAF prevalence.Although we cannot exclude the contribution of other, unaccounted confounders in influencing our findings, the results of the meta-regression clearly underline the impact of age and follow-up time in determining the occurrence of SCAF, representing, together with the epidemiological data about the clinical profile of SCAF patients, the most relevant evidence provided by our work.

Conclusions
In this systematic review and meta-regression analysis, SCAF increased with age and decreased over longer follow-up times, both being independently associated with its prevalence.The presence of SCAF is associated with higher age, more prevalent comorbidities, and higher thromboembolic risk.Each curve is a graphical representation of SCAF incidence rate at each year of follow-up according to patients' age.

Funding
No funding was received by any authors for the development of this project.

Declaration of Competing Interest
JSH reports research grants and speaking fees from Boehringer-Ingelheim, Bayer, BMS/Pfizer and Servier; DL has received investigator-initiated educational grants from Bristol-Myers Squibb (BMS), has been a speaker for Boehringer Ingelheim, and BMS/Pfizer and has consulted for BMS, Boehringer Ingelheim, and Daiichi-Sankyo; GB received small speaker's fees from Medtronic, Boston, Boehringer Ingelheim and Bayer; SB reports esearch grant from MSD not relevant to the manuscript; GYHL has been consultant and speaker for BMS/Pfizer, Boehringer Ingelheim and Daiichi-Sankyo.No fees are directly received personally.All other authors have nothing to declare.

Fig. 2 .
Fig. 2. Marginal Relationship between Follow-up times and Mean Age of the included studies and SCAF prevalence.Panel A) Years of Follow-Up; Panel B) Age.

Fig. 3 .
Fig. 3. Prevalence of SCAF according to mean age at different follow-up times.Each curve is a graphical representation of SCAF incidence rate at each year of follow-up according to patients' age.

Table 1
General Characteristics of the Studies Included in the Systematic Review.

Table 2
Subgroup Analyses for SCAF Prevalence.
from observational prospective studies.Of the included studies, 18 studies were conducted in Europe; 18 in Asia; 8 in North America, and 10 in other geographical locations, including multinational studies.

Table 3
Association between clinical characteristics and SCAF Presence.
Central Illustration.Epidemiology of Subclinical Atrial Fibrillation in Patients with CIEDs: A Systematic Review and Meta-RegressionLegend: CIEDs= Cardiac Implanted Electronic Devices; IR= Incidence Rate; SCAF= Subclinical Atrial Fibrillation.