Background
Recent studies have suggested a high prevalence of subclinical atrial fibrillation (AF) in various patient populations, and interest in AF screening has increased. However, knowledge about episode duration is scarce, and risk factors for short or long subclinical AF episodes have yet to be recognized. The aim of the study was to assess AF by long-term continuous screening and to investigate predictors of episodes lasting ≥6 minutes, ≥5.5 hours, or ≥24 hours, respectively.
Methods
A total of 597 patients aged ≥70 years and diagnosed with ≥1 of hypertension, diabetes, previous stroke, or heart failure were recruited from the general population to receive implantable loop recorder with remote monitoring. Exclusion criteria included history of AF or cardiac implantable electronic device. AF episodes were adjudicated by senior cardiologists.
Results
During 40 (37; 42) months of continuous monitoring, AF was detected in 209 (35%) of the patients. The cumulative incidences at 3 years were 33.8% (30.2%-37.8%), 16.1% (13.4%-19.4%), and 5.7% (4.1%-7.9%) for AF episodes lasting ≥6 minutes, ≥5.5 hours, and ≥24 hours, respectively. Slower resting sinus rate and higher body mass index, N-terminal prohormone of brain natriuretic peptide, and troponin T at baseline were independently associated with AF detection. Addition of these markers to a model of sex, age, and comorbidities improved prediction of AF episodes ≥24 hours (time-dependent area under the receiver operating characteristic curve 79% vs 65%, P = .037).
Conclusions
A considerable burden of previously unknown AF was detected when long-term monitoring was applied in at-risk patients. Biomarkers were associated with AF incidence and improved prediction of long AF episodes.
Ischemic stroke is an increasing health problem worldwide. At least 20% of ischemic strokes are attributable to atrial fibrillation (AF), and another 30% of ischemic strokes are cryptogenic, possibly related to undiagnosed AF. Oral anticoagulation (OAC) is well established as an effective treatment for stroke prevention in at-risk patients diagnosed with AF. However, as AF is often asymptomatic, many patients remain undiagnosed. Approximately 30% of a general population of pacemaker or cardioverter defibrillator patients will have new-onset AF during 2-3 years following implantation. , Although the majority of these AF episodes are short-lasting and asymptomatic, the ASymptomatic atrial fibrillation and Stroke Evaluation in pacemaker patients and the atrial fibrillation Reduction atrial pacing Trial (ASSERT) and other studies have found that such AF episodes are associated with risk of stroke. , Since the ASSERT was published, screening for AF has received increasing attention. Still, the bulk of our knowledge about AF and stroke risk is derived from patients with symptomatic or rather long-lasting AF episodes, and mass screening cannot yet be recommended. , One problem is that an appropriate screening population has yet to be recognized. A post hoc analysis of ASSERT found that the increased risk of stroke was driven by patients with AF episodes lasting >24 hours, and a recent consensus document suggested the compromise of OAC in patients with device-detected AF episodes lasting ≥5.5 hours. Similarly, recent data from the Veterans Health Administration showed increasing rates of stroke with increasing episode duration. Predictors for such AF episodes remain unknown.
The ongoing LOOP study randomizes individuals with CHA 2 DS 2 VASc score of ≥2 to AF screening with implantable loop recorder (ILR) screening or control. The study’s power calculation assumes that 30% of participants receiving ILR will have AF detected and that stroke can be reduced by OAC in these patients. The primary end point will be published according to protocol when the required number of events is reached.
The current substudy had two aims: First, we sought to assess incidence of AF using very long-term continuous monitoring in a large general population at risk. In this regard, we wished to validate if the AF detection rate among the earliest included participants in the LOOP study matched the study’s power calculation. Second, we sought to investigate predictors of AF episodes of shorter and longer duration.
Methods
Study design
The LOOP study is an ongoing, investigator-initiated, multicenter controlled trial. A detailed description of the study design has been published previously. In brief, participants from the general population residing in 3 of the 5 administrative regions of Denmark are identified by administrative registries and receive a letter of invitation from 1 of the 4 study centers. Eligible subjects are ≥70 years old and have ≥1 of the following stroke risk factors: hypertension, diabetes, heart failure, or previous stroke. Exclusion criteria include OAC or contraindication to OAC and any history of AF. At the initial screening visit, study eligibility is confirmed, and a baseline evaluation is performed including detailed medical history, height and weight, and blood pressure and sinus rate measurement after 10 minutes of supine rest. Prevalent AF is ruled out by 12-lead electrocardiogram. Blood is sampled for measurement of creatinine, N-terminal prohormone of brain natriuretic peptide (NT-proBNP), high-sensitivity C-reactive protein (hs-CRP), and troponin T. In the smallest center, NT-proBNP was not routinely measured for logistic reasons, and in the 2 smallest centers, troponin I was measured instead of troponin T. Subjects are then randomized in a 1:3 ratio to receive ILR (Reveal LINQ, Medtronic) with continuous electrocardiographic monitoring or control.
The programmable parameters of the ILRs are set to the standard for detection of “suspected AF” according to manufacturer’s recommendations: AF detection “less sensitive,” ectopy rejection “nominal,” blank after sense 150 milliseconds, sensing threshold decay delay 150 milliseconds, and R-wave sensitivity 0.035 mV, although in participants with R-wave amplitude below 0.3 mV even after repositioning, sensitivity would be reprogrammed at the discretion of the implanting physician, and in participants with false AF alerts due to premature beats during follow-up, ectopy rejection would be reprogrammed to “aggressive” on the discretion of the monitoring physician. All participants receiving ILR are followed by automated remote transmissions to an online database, where any new arrhythmia episodes are reviewed daily by an experienced medical doctor. AF adjudication is obtained by at least 2 senior cardiologists independently reviewing new-onset AF episodes lasting ≥6 minutes. The length of the rhythm strip used for adjudication is 2 minutes for all AF episodes. When AF is confirmed, OAC is initiated, and monitoring continues for further detection of longer AF episodes or other arrhythmias as adjudicated by minimum 1 experienced medical doctor. Patients are questioned about AF-related symptoms at the index episode.
The heart rhythm monitoring continues until end of battery life, study withdrawal, or death but is expected to last minimum 3 years according to the ILR manufacturer’s projected battery longevity.
In the current analysis, all LOOP study participants receiving ILR until June 1, 2015, were included, and data acquisition and AF adjudication concluded on December 1, 2018. Thus, the time span from implantation to last day of possible interrogation was minimum 42 months.
The primary end point of the current analysis was time to first adjudicated AF episode lasting ≥6 minutes, whereas time to first AF episode lasting ≥30 minutes, ≥1 hour, ≥5.5 hours, ≥12 hours, and ≥ 24 hours were secondary end points. In this way, participants reaching the end point of AF lasting 6 minutes at 1 day were subsequently followed for AF episodes of longer durations, whereas participants who debuted with AF lasting 24 hours at 1 day also reached the end points of AF ≥12 hours, ≥5.5 hours, ≥1 hour, etc, on that day. Participants were censored on the date of the last ILR interrogation.
All study participants gave written informed consent. A centralized, online case report file system was used for storage of study data. The LOOP study was approved by the Ethics Committee of the Capital Region of Denmark (H-4-2013-025) and the Danish Data Protection Agency (2007-58-0015). The trial is registered at ClinicalTrials.gov ( NCT02036450 ).
Statistics
Continuous variables were presented as means and SDs for normally distributed variables, and medians and quartiles 1 and 3 (Q1; Q3) for non-normally distributed variables, whereas categorical variables were presented as frequencies and corresponding percentages. Any prevalences, cumulative incidences, or rates were presented as percentage (95% CI) and ratios as ratio (95% CI).
A power calculation was conducted for analyses of association between baseline variables and time to event. Assuming that 30% of the population would reach the primary end point during follow-up, inclusion of 600 participants would yield 80% power to detect a hazard ratio (HR) of 1.5 or higher for the half of the population with the predictor (eg, biomarker above median) compared to without the predictor (eg, biomarker below median), with 2-sided equality and 5% risk of type I error.
Incidence rates and exact CIs were derived from a Poisson distribution. To account for the competing risk of death, the cumulative incidences were estimated, plotted, and groupwise compared in a multistate fashion.
To assess potential predictors of AF, association analyses were performed. Baseline variables were analyzed with univariate cause-specific Cox regression, and after considering clinical and statistical significance ( P < .1) in the univariate analyses, multivariable Cox regression models were constructed to investigate the association between baseline variables and AF episodes lasting ≥6 minutes, ≥5.5 hours, and ≥ 24 hours independently of sex, age, and comorbidities (heart failure, hypertension, diabetes, previous stroke, cardiac valvular disease, and previous acute myocardial infarction and/or coronary arterial bypass graft surgery [CABG]), constituting a basic model. Schoenfeld and Martingale residuals were assessed to validate the proportional hazards and linearity assumption, respectively. To comply with these assumptions for valid Cox regression, CHA 2 DS 2 VASc and CHADS 2 score were grouped as ≤3 and >3, and ≤1 and >1, respectively. NT-proBNP and hs-CRP were logarithmized to normalize the distribution.
Additionally, to evaluate the added discriminative value over the basic model from the physical and biochemical markers identified in the association studies as possible predictors, risk prediction analyses were conducted for the end points of AF episodes lasting ≥6 minutes, ≥5.5 hours, and ≥24 hours. The 42-month risk of AF was predicted using cause-specific Cox regression. Receiver operating characteristics (ROC) curves were drawn, and the time-dependent area under the ROC curve (AUC) was calculated for each model. Differences in AUC between models were calculated, whereas Brier scores were used to assess model calibration. Furthermore, to extrapolate the predictive value of the models from the study population, the risk prediction analyses were repeated after splitting the data set in training and test sets. The training set comprised two thirds of the population, and analyses were repeated in 1,001 splits at random. The split corresponding to the median AUC of the models without the predictors was used to analyze the difference between the basic model and the model with the predictors. All association and risk prediction were performed as complete-case analyses, meaning that missing variables in the proposed model were dropped.
Lastly, to further evaluate the relationship between continuous variables and AF episodes ≥6 minutes, a sliding-windows approach was used to plot the incidence rate against the median value of the variable in subgroups comprising 10% of the study population with sequentially overlapping steps the size of 2.5% of the population.
For supplementary analyses, in the association analyses, instead of adjusting for sex, age, and all individual comorbidities, CHA 2 DS 2 VASc and CHADS 2 score, respectively, were entered into a model adjusting for each of the following baseline variables: body mass index (BMI), resting sinus rate, NT-proBNP, and troponin T.
Analyses were performed using the R software, https://www.R-project.org/ , R Core Team (2017), including the survival , cmprsk , epiR , evobiR , timereg, pec , riskRegression , and ggplot2 packages.
Funding
This investigator-initiated study was supported by The Innovation Fund Denmark ( 12-135225 ), The Research Foundation for the Capital Region of Denmark (no grant number), The Danish Heart Foundation ( 11-04-R83-A3363-22625 ), Aalborg University Talent Management Programme (no grant number), Arvid Nilssons Fond (no grant number), Skibsreder Per Henriksen, R. og Hustrus Fond (no grant number), and Medtronic (no grant number). The authors are solely responsible for the design and conduct of this study, all study analyses, the drafting and editing of the paper, and its final contents.
Results
Population and follow-up
A total of 597 study participants received ILR from February 26, 2014, until June 1, 2015, and were included in the current analysis. Mean age was 76 (±4) years, 57% were men, and mean CHA 2 DS 2 VASc score was 3.9 (±1.2) ( Table I ). NT-proBNP was missing in 65, troponin T in 198, blood pressure in 1, resting sinus rate in 3, hs-CRP in 1, and creatinine in 2 participants.
| AF during follow-up | No AF | AF ≥ 6 min | AF ≥ 30 min | AF ≥ 1 h | AF ≥ 5.5 h | AF ≥ 12 h | AF ≥ 24 h | All |
|---|---|---|---|---|---|---|---|---|
| n (% of all) | 388 (65) | 209 (35) | 180 (30) | 159 (27) | 99 (17) | 61 (10) | 37 (6) | 597 (100) |
| Male sex (%) | 218 (56) | 123 (59) | 105 (58) | 91 (57) | 63 (64) | 42 (69) | 25 (68) | 341 (57.1) |
| Age, y (SD) | 76.0 (4.0) | 77.1 (4.5) | 77.0 (4.4) | 77.1 (4.6) | 77.1 (4.7) | 76.8 (4.4) | 77.2 (4.6) | 76.4 (4.2) |
| CHA 2 DS 2 VASc score (SD) | 3.9 (1.2) | 4.1 (1.2) | 4.0 (1.2) | 4.1 (1.2) | 4.1 (1.2) | 4.0 (1.1) | 4.0 (1.1) | 3.9 (1.2) |
| CHADS 2 score (SD) | 2.2 (1.1) | 2.4 (1.1) | 2.4 (1.1) | 2.4 (1.0) | 2.5 (1.0) | 2.4 (1.0) | 2.4 (1.0) | 2.3 (1.1) |
| Heart failure (%) | 19 (4.9) | 6 (2.9) | 5 (2.8) | 4 (2.5) | 2 (2.0) | 2 (3.3) | 1 (2.7) | 25 (4.2) |
| Hypertension (%) | 350 (90) | 190 (91) | 165 (92) | 146 (92) | 90 (91) | 56 (92) | 34 (92) | 540 (90.5) |
| Diabetes (%) | 114 (29) | 60 (29) | 51 (28) | 46 (29) | 30 (30) | 18 (30) | 11 (30) | 174 (29.1) |
| Previous stroke (%) | 64 (16) | 43 (21) | 37 (21) | 34 (21) | 23 (23) | 16 (26) | 9 (24) | 107 (17.9) |
| Previous transient ischemic attack, n (%) | 46 (11.9) | 21 (10.0) | 19 (10.6) | 16 (10.1) | 10 (10.1) | 5 (8.2) | 3 (8.1) | 67 (11.2) |
| Previous systemic embolism, n (%) | 25 (6.4) | 17 (8.1) | 15 (8.3) | 14 (8.8) | 10 (10.1) | 5 (8.2) | 2 (5.4) | 42 (7.0) |
| Previous AMI (%) | 38 (9.8) | 20 (9.6) | 17 (9.4) | 16 (10.1) | 12 (12.1) | 8 (13.1) | 5 (13.5) | 58 (9.7) |
| Previous CABG (%) | 27 (7.0) | 14 (6.7) | 14 (7.8) | 13 (8.2) | 11 (11.1) | 9 (14.8) | 7 (18.9) | 41 (6.9) |
| Valvular heart disease (%) | 13 (3.3) | 13 (6.2) | 8 (4.4) | 6 (3.8) | 4 (4.0) | 2 (3.3) | 1 (2.7) | 26 (4.4) |
| β-Blockers, n (%) | 95 (24) | 49 (23) | 45 (25.0) | 41 (25.8) | 29 (29.3) | 23 (37.7) | 15 (40.5) | 144 (24.1) |
| Calcium channel blockers, n (%) | 124 (32) | 85 (41) | 72 (40.0) | 63 (39.6) | 43 (43.4) | 32 (52.5) | 17 (45.9) | 209 (35.0) |
| Nondihydropyridine type, n (%) | 5 (1.3) | 9 (4.3) | 8 (4.4) | 6 (3.8) | 3 (3.0) | 2 (3.3) | 0 (0.0) | 14 (2.3) |
| ACEi, ARB, or renin inhibitors, n (%) | 235 (61) | 121 (58) | 105 (58.3) | 93 (58.5) | 57 (57.6) | 35 (57.4) | 24 (64.9) | 356 (59.6) |
| Lipid-lowering drugs, n (%) | 201 (52) | 121 (58) | 107 (59.4) | 97 (61.0) | 65 (65.7) | 39 (63.9) | 23 (62.2) | 322 (53.9) |
| Diuretics, n (%) | 112 (29) | 65 (31) | 57 (31.7) | 52 (32.7) | 34 (34.3) | 19 (31.1) | 12 (32.4) | 177 (29.6) |
| Platelet inhibitors, n (%) | 188 (48) | 107 (51) | 93 (51.7) | 84 (52.8) | 54 (54.5) | 36 (59.0) | 23 (62.2) | 295 (49.4) |
| Glucose-lowering drugs, n (%) | 98 (25) | 51 (24) | 45 (25.0) | 40 (25.2) | 27 (27.3) | 17 (27.9) | 11 (29.7) | 149 (25.0) |
| Systolic BP, mm Hg (SD) | 151.9 (19) | 151.6 (18) | 151.7 (18) | 152.2 (18) | 151.9 (17) | 149.6 (18) | 153.4 (17) | 151.8 (18.7) |
| Diastolic BP, mm Hg (SD) | 85.0 (12) | 84.6 (11) | 84.7 (11) | 84.6 (11) | 84.5 (11) | 84.1 (11) | 86.0 (12) | 84.9 (11.6) |
| Resting sinus rate, beat/min (SD) | 72.5 (13) | 68.8 (11) | 68.7 (12) | 68.6 (12) | 68.0 (12) | 68.2 (13) | 67.1 (13) | 71.2 (12.4) |
| Height, cm (SD) | 170.5 (8.7) | 171.3 (8.7) | 171.2 (8.8) | 171.0 (8.8) | 171.6 (8.5) | 172.9 (8.6) | 172.9 (9.0) | 170.8 (8.7) |
| Body mass index, kg/m 2 (SD) | 27.4 (4.4) | 27.8 (4.8) | 27.6 (4.5) | 27.8 (4.6) | 28.3 (4.8) | 28.4 (4.7) | 29.3 (4.9) | 27.6 (4.6) |
| Creatinine, μmol/L (SD) | 86.7 (23) | 87.9 (24) | 88.3 (25) | 87.9 (24) | 89.8 (24) | 92.1 (25) | 91.7 (24) | 87.1 (23.6) |
| NT-proBNP, pmol/L (Q1; Q3) | 14 (8; 26) | 20 (12; 34) | 20 (12; 34) | 21 (12; 34) | 21 (12; 34) | 24 (13; 38) | 24 (13; 44) | 16 (9; 28) |
| hs-CRP, mg/L (Q1; Q3) | 2 (1; 3) | 2 (1; 4) | 2 (1; 4) | 2 (1; 4) | 2 (1; 4) | 2 (1; 4) | 2 (1; 4) | 2 (1; 4) |
| Troponin T, ng/L (SD) | 13.8 (5.4) | 15.8 (7.3) | 15.8 (7.9) | 15.9 (8.3) | 16.4 (8.9) | 17.5 (10.4) | 19.0 (12.1) | 14.6 (6.48) |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
