Atrial Fibrillation Burden: Moving Beyond Atrial Fibrillation as a Binary Entity

Authors, Journal, Affiliations, Type, DOI

• Authors: Lin Y. Chen, MD, MS, FAHA (Chair); Mina K. Chung, MD, FAHA (Vice Chair); Larry A. Allen, MD, MHS, FAHA; Michael Ezekowitz, MBChB, DPhil, FAHA; Karen L. Furie, MD, MPH; Pamela McCabe, PhD, RN; Peter A. Noseworthy, MD; Marco V. Perez, MD; Mintu P. Turakhia, MD, MAS
• Journal: Circulation 2018;137. DOI 10.1161/CIR.0000000000000568. Approved by the AHA Science Advisory and Coordinating Committee on 27 November 2017 and the AHA Executive Committee on 28 February 2018
• Affiliations: AHA Council on Clinical Cardiology; Council on Cardiovascular and Stroke Nursing; Council on Quality of Care and Outcomes Research; and Stroke Council
• Type: AHA scientific statement — literature review with expert consensus, not a guideline. Formally structured around AF burden as a concept: definition and measurement, burden–stroke relationships, non-stroke outcomes, risk-factor modification, and a tabulated knowledge-gap section (its Table 5)
• DOI: 10.1161/CIR.0000000000000568
• Note on pagination: the source PDF carries placeholder page numbers (137:00–00, e•••–e•••) because it was published online-first. No definitive page range appears anywhere in the document, so none is asserted here

Overview

The argument of this statement is in its title: cardiology had been studying atrial fibrillation as a binary — present or absent — and treating it as such, when the biologically relevant variable is how much AF a person has. The authors define AF burden ideally as the proportion of time spent in AF during a monitoring period, note that continuous monitoring is what that definition requires and therefore why the literature is sparse, and review what is known when burden is proxied by AF pattern (paroxysmal/persistent/permanent) or by device-detected duration. Their most consequential conclusions are that stroke risk is not well characterised as a function of increasing burden — the strongest evidence supports only that persistent AF carries more risk than paroxysmal — and that the thresholds used throughout the CIED literature were arbitrarily prespecified rather than empirically derived. A structured risk-factor management programme is the one intervention shown to reduce AF burden; higher burden is not reliably associated with worse quality of life. The statement remains the reference framing for the burden-versus-binary debate that ARTESIA and NOAH-AFNET 6 later tested.

Keywords

AHA Scientific Statements; atrial fibrillation; lifestyle; risk factors; stroke

Key Takeaways

Definition of AF Burden

• AF burden is "the amount of AF that an individual has." Considering it means treating AF as a quantitative rather than binary entity
• Many definitions are possible — the duration of the longest AF episode, or the number of episodes in a monitoring period — but the best is the proportion of time in AF during a monitoring period, expressed as a percentage
• That definition is rarely used because it requires continuous ambulatory monitoring. Consequently most studies instead use AF type or pattern (paroxysmal, persistent, permanent) or device-detected AF duration from cardiac implantable electronic devices (CIEDs)

AF Pattern or Type and Risk of Stroke

• Guidelines recommend assessing stroke risk from the individual patient risk profile, not from AF type, pattern or burden. The constraint is that the anticoagulation decision itself is binary: risk attributable to patient characteristics (e.g. the CHA₂DS₂-VASc variables) was thought to overwhelm risk attributable to AF pattern. On its own, a particular pattern is unlikely to spare a high-risk patient from anticoagulation or justify it in a low-risk one
• Direct oral anticoagulant trials consistently favour paroxysmal over persistent AF, even after adjustment:
  • ARISTOTLE: adjusted HR 0.70 (95% CI 0.51–0.93), P=0.015
  • ROCKET-AF: adjusted HR 0.79 (0.63–1.0), P=0.048
  • ENGAGE-AF: adjusted HR 0.79, P=0.015
  • RE-LY: stroke rates numerically lower in paroxysmal (1.32%) than persistent (1.55%), but no formal adjusted comparison and paroxysmal patients had lower CHADS₂ scores
• Older anticoagulation-era data agree: SPORTIF III/V showed annual stroke/systemic embolism of 1.73% for persistent vs 0.93% for paroxysmal (adjusted HR 1.87, 1.04–3.36, P=0.037); AMADEUS showed a higher composite of cardiovascular death and stroke/systemic embolism in nonparoxysmal AF (HR 1.68, 1.08–2.55, P=0.02)
• But SPORTIF III/V found similar stroke rates among patients with ≥2 stroke risk factors, "which suggests that at the higher end of the risk spectrum, the paroxysmal AF pattern is not associated with significantly lower stroke risk"
• Among aspirin-treated patients (ACTIVE-A, AVERROES): annual ischaemic stroke 2.1% paroxysmal, 3.0% persistent, 4.2% permanent; adjusted HR 1.83 for permanent vs paroxysmal (P<0.001) and 1.44 for persistent vs paroxysmal (P=0.02)
• Two older antiplatelet studies found no difference — SPAF (intermittent 3.2% vs sustained 3.3%) and ACTIVE-W (adjusted RR 0.94, 0.63–1.40, P=0.755) — though both were comparatively small, and in ACTIVE-W a greater proportion of persistent-AF patients received warfarin (85% vs 65%), which could attenuate any pattern difference
• Registry data are mixed, and confounded by differential treatment. The Loire Valley AF Project found clinical factors, not AF pattern, independently increased stroke risk; EORP-AF and J-RHYTHM showed higher death rates with nonparoxysmal AF that did not survive multivariable adjustment; the Euro Heart Survey found comparable thromboembolic risk across patterns. Conversely, the Fushimi AF Registry found paroxysmal AF independently associated with lower stroke/systemic embolism, and a meta-analysis of 12 studies (nearly 100,000 patients) found a multivariable adjusted HR of 1.38 (1.19–1.61) for nonparoxysmal vs paroxysmal AF
• Stroke severity: persistent AF has been associated with more severe neurological deficits and poorer outcomes after cardioembolic stroke, and possibly higher recurrent stroke risk — yet paroxysmal AF remains the most common pattern among patients presenting with acute stroke, reflecting its higher prevalence and that many such patients are untreated

Device-Detected AF Duration and Stroke Risk

• CIEDs are the only practical source of continuous long-term AF data, tracking cumulative burden with date, time, duration, cycle length and day-level burden
• AHREs are an imperfect surrogate for AF. False detection arises from far-field R-wave oversensing or runs of premature atrial complexes or atrial tachycardia. Implantable loop recorders and cardiac monitors relying on R-R irregularity have lower sensitivity and specificity than limb-lead ambulatory ECG or CIEDs with atrial leads. In ASSERT, 17.3% of AHREs >190 bpm lasting >6 minutes were false positives for AF
• Every burden threshold in this literature was chosen arbitrarily, not derived from risk:
  • MOST (316 patients): AHRE >220 bpm with a 5-minute cutoff (chosen to avoid oversensing false positives) → death or nonfatal stroke HR 2.79 (1.51–5.15), P=0.0011. Shorter and much longer episodes were never evaluated
  • TRENDS (2,846 patients): stroke risk doubled with high AT/AF burden (≥5.5 h on any single day in a 30-day window) but not with low burden. The 5.5-hour threshold was the median burden, not an empirical risk threshold
  • ASSERT: subclinical AT (>6 min at ≥190 bpm) → ischaemic stroke HR 1.76 (0.99–3.11); episodes >6 h HR 2.00 (1.13–3.55). On follow-up analysis, only episodes >24 h increased risk (adjusted HR 3.24, 1.51–6.95), while 6 min–6 h (HR 0.75, 0.29–1.96) and 6–24 h (HR 1.32, 0.40–4.37) were not significantly different from no subclinical AT
• The SOS AF project is the most direct test of dose–response, and it does not show one. Pooled individual patient data from 5 prospective studies, 10,016 patients without permanent AF. Annual stroke risk by burden category: 0–<5 min 0.24%, ≥5 min–<1 h 0.08%, ≥1–<6 h 0.34%, ≥6–<12 h 0.26%, ≥12–<23 h 0.30%, ≥23 h 0.23%. The lowest burden category has nearly the highest risk, and there is no monotonic gradient across a >250-fold range of burden
• Turakhia case-crossover analysis (9,850 CIED patients, Veterans Administration, 2002–2012): AF burden ≥5.5 h on a given day raised short-term stroke risk 4- to 5-fold; risk was highest in the 5–10 days after the episode and was no longer elevated by 30 days. Critically, when the cutoff was varied from 30 minutes to 6 hours in increments, no risk threshold was identified, and the majority of strokes were temporally dissociated from AF — so although a transient risk increase exists, the overall attributable risk was low
• The statement's own conclusions on this section: cutoffs were arbitrarily prespecified rather than empirically derived; the effect of very brief episodes (<5–6 min) on stroke risk is unknown; the functional form of the burden–stroke relationship (threshold versus monotonicity, and the duration of any transient risk) is not well characterised; most ischaemic strokes in CIED patients are temporally discordant from AF; and these studies may have included atrial flutter, whose stroke risk may differ

AF Type, AF Duration, and Non-Stroke Outcomes

• Non-stroke outcomes reviewed: heart failure, cognitive impairment and dementia, myocardial infarction, chronic kidney disease progression, and quality of life
• Heart failure: HF and AF are highly concordant overlapping epidemics — AF occurs in more than half of individuals with HF, and HF in more than a third with AF
• Consistent pattern: compared with paroxysmal AF, nonparoxysmal AF is associated with higher prevalence and incidence of heart failure and higher mortality, while data are less clear for device-detected AHRE
• Cognitive function: one small cross-sectional study found persistent AF associated with lower cognitive function than paroxysmal AF, but data on AF burden and cognition are scarce — explicitly flagged as a gap
• Quality of life runs counter to intuition: "higher AF burden is not necessarily related to lower quality of life; new-onset AF is associated with worse quality of life than reported in patients with permanent AF." Interventions aimed at reducing AF burden have not translated into improved quality of life

Atherosclerotic and Lifestyle Risk Factors and AF Burden

• A structured and comprehensive risk-factor management programme targeting atherosclerotic risk factors, weight loss, and maintenance of a healthy weight is effective in reducing AF burden — the statement's one clear interventional conclusion
• Tables 3 and 4 tabulate the non-randomised studies/registries and RCTs of comprehensive risk-factor management on AF burden

Key Knowledge Gaps and Future Directions (the statement's Table 5)

• Validation of definitions and measures of AF burden — no agreed operational definition exists
• The threshold of AF burden that increases stroke risk enough to warrant anticoagulation — the central unanswered question, and the one ARTESIA and NOAH-AFNET 6 later addressed
• The mechanisms underlying the weak temporal correlation between AF and stroke — why most strokes are not proximate to AF episodes
• The effect of very brief AF episodes (<5–6 min) on stroke risk
• The relationship between AF burden and cognitive function
• Whether reducing AF burden improves outcomes, given that it does not reliably improve quality of life

Limitations of the document

Stated and evident in the statement itself

• This is a scientific statement, not a guideline. It makes no formal recommendations, assigns no Class of Recommendation or Level of Evidence, and performs no quantitative synthesis. It is a narrative review with expert consensus, and its strength lies in framing rather than in pooled estimates
• The pivotal evidence, by the authors' own account, does not support a dose–response relationship. The SOS AF pooled analysis of 10,016 patients shows no monotonic gradient in annual stroke risk across a >250-fold range of AF burden (0.24%, 0.08%, 0.34%, 0.26%, 0.30%, 0.23%), and the Turakhia case-crossover found no identifiable threshold when the cutoff was varied from 30 min to 6 h. The statement's framing — "burden, not binary" — is therefore better supported for pattern (paroxysmal vs persistent) than for quantified burden, which is the concept the title advances
• Nearly every threshold in the literature is arbitrary, which the authors state directly and repeatedly. TRENDS used the median burden as its cutoff; ASSERT's 6-minute and 24-hour cutoffs came from manufacturer categories; MOST's 5-minute cutoff was chosen to avoid oversensing. Any "threshold" quoted from this literature is consequently a convention, not a validated risk point
• The pattern-versus-burden comparison is confounded by treatment. Anticoagulation rates and other management differ systematically by AF pattern, and patient characteristics trend toward higher risk in persistent AF — making rigorous adjustment critical and the registry results correspondingly unstable in both directions
• Device-detected AF is an imperfect surrogate. 17.3% of ASSERT's AHREs at >190 bpm were false positives; implantable monitors relying on R-R irregularity have lower sensitivity and specificity; and several of the key studies may have included atrial flutter, whose stroke risk may not equal that of AF
• Publication date 2018 — the statement predates the trials that tested its central question. ARTESIA and NOAH-AFNET 6 (both 2023–2024) directly randomised device-detected AF to anticoagulation or control, and ARTESIA's stroke reduction (HR 0.63) alongside NOAH-AFNET 6's neutral result (HR 0.81) is exactly the evidence this statement called for. The statement's framing of the threshold question remains current; its evidence base does not
• Disclosures are substantial and tabulated. The writing group discloses relationships across multiple device and pharmaceutical manufacturers (including Boston Scientific, Amgen, Janssen, Novartis, Elsevier, UpToDate and others), with research grants from NIH, AHA and PCORI

Key Concepts Mentioned

• concepts/AF-Burden — the concept this statement defines and argues for
• concepts/Subclinical-AF — device-detected AF/AHRE, the setting where burden is measured continuously; the statement's TRENDS/ASSERT/SOS AF material belongs to both pages
• entities/Atrial-Fibrillation — AF pattern (paroxysmal/persistent/permanent) as the proxy for burden
• concepts/CHA2DS2-VA — the risk score that guidelines use instead of burden when deciding on anticoagulation
• concepts/Poststroke-AF-Monitoring — monitoring yield after stroke, where detection and burden intersect
• concepts/AF-CARE — the "Avoid stroke" pillar, whose threshold this statement problematises
• AHRE (atrial high-rate episode) — device-detected atrial tachyarrhythmia used as an AF surrogate, with a 17.3% false-positive rate in ASSERT
• SOS AF project — the largest pooled burden–stroke analysis, and the source of the absent dose–response finding
• Temporal dissociation of AF and stroke — the observation that most strokes are not proximate to AF episodes

Key Entities Mentioned

• entities/Atrial-Fibrillation — primary entity
• American Heart Association — issuing body; Councils on Clinical Cardiology, Cardiovascular and Stroke Nursing, Quality of Care and Outcomes Research, and Stroke
• Lin Y. Chen (University of Minnesota) — Chair; Mina K. Chung (Cleveland Clinic) — Vice Chair
• Trials and registries discussed: ARISTOTLE, ROCKET-AF, ENGAGE-AF, RE-LY, SPORTIF III/V, AMADEUS, ACTIVE-A, AVERROES, SPAF, ACTIVE-W, ASSERT, MOST, TRENDS, SOS AF, Loire Valley AF Project, EORP-AF, J-RHYTHM, Fushimi AF Registry, Euro Heart Survey

Wiki Pages Updated

• Created: wiki/sources/afburden-aha-2018.md
• Created: wiki/concepts/AF-Burden.md
• Updated: wiki/concepts/Subclinical-AF.md
• Updated: wiki/entities/Atrial-Fibrillation.md
• Updated: wiki/sourceindex.md
• Updated: wiki/wikiindex.md