#atrial-fibrillation #heart-failure #cardio-oncology #perioperative #acute-coronary-syndrome #stroke

AF in Special Populations

Contents. #Definition · #AF in heart failure · #AF in cancer · #AF after noncardiac surgery · #AF and ACS · #AF and stable CAD · #AF and acute ischaemic stroke · #Contradictions / Open Questions · #Sources

Definition

General AF management assumes a patient whose comorbidities and competing antithrombotic needs can be managed at leisure. This page covers the settings where that assumption fails — heart failure, active cancer, the perioperative period, recent acute coronary syndrome, stable coronary disease on long-term antithrombotic therapy, and the weeks after an ischaemic stroke. In each, the ordinary rules for anticoagulation, rate control or rhythm control are modified by a competing risk, and the modifications rest on weaker evidence than the parent guideline implies.

Key Concepts

AF in heart failure

Prevalence and prognosis. AF affects 25–50% of HF patients, rising with age and HF severity. When AF causes HF (tachycardiomyopathy) the course is more favourable than AF developing in established chronic HF; AF onset in chronic HF worsens prognosis including stroke. (sources/HF-ESC-2021, rating: very high)
Anticoagulation. DOACs are preferred over VKAs for all HF plus AF except moderate-to-severe mitral stenosis or mechanical valves (Class I). Long-term OAC should be considered at CHA₂DS₂-VASc ≥1 (men) or ≥2 (women) (Class IIa); LA appendage closure may be considered when OAC is contraindicated (Class IIb).
Rate control. Lenient control (resting <110 bpm) is an acceptable initial approach. Beta-blockers are first-line (Class IIa); digoxin or digitoxin may be added if rate remains high or beta-blockers are not tolerated (Class IIa); intravenous amiodarone is used for NYHA IV or haemodynamic instability; AV node ablation is Class IIb if medical rate control fails.
Rhythm control. Electrical cardioversion is Class I in haemodynamically unstable patients. Amiodarone is the only safe antiarrhythmic in HFrEF — propafenone, flecainide and dronedarone are associated with worse outcomes. Drug-based rhythm control versus rate control failed to show benefit in several HF trials.
Early rhythm control in HF. In the 28.6% of EAST-AFNET 4 patients with HF, early rhythm control reduced cardiovascular death, stroke, HF hospitalisation and ACS — but ablation was used in only 19.4% of the rhythm-control arm. (sources/EAST-AFNET4-NEJM-2020, rating: very high)
CASTLE-AF. In 363 patients with paroxysmal or persistent AF, LVEF <35% and an implanted device, catheter ablation versus medical therapy reduced the primary endpoint of all-cause death or HF hospitalisation from 44.6% to 28.5% (HR 0.62; 95% CI 0.43–0.87; P=0.007), with cardiovascular death, all-cause death and HF hospitalisation each individually reduced. The population was highly selected (363 of 3,013 screened), unblinded, with crossovers and low event counts; ablation is Class IIa where AF is clearly associated with worsening HF despite medical therapy. (sources/HF-ESC-2021)

AF in cancer

Epidemiology. AF occurs in 2–16% of patients during active cancer therapy; post-lung-surgery incidence is highest (6–32%); every cancer type carries higher AF risk than controls. (sources/Cardio-Oncology-ESC-2022, rating: very high)
Bidirectionality. Cancer patients have more AF and AF patients have more cancer, driven by shared risk factors of age, inflammation and metabolic disease; a cohort of >15,000 patients confirmed higher prevalent AF in cancer. (sources/arrhythmia-cardio-oncology-aha-2021, rating: very high)
Outcome penalty. Cancer-associated AF carries 2× the stroke and thromboembolic risk and 6× the HF risk of AF alone. (sources/Cardio-Oncology-ESC-2022)
Risk scores fail. CHA₂DS₂-VASc likely underestimates thromboembolic risk — Danish registry data show higher 2-year stroke risk in cancer patients at scores 0–1 than in non-cancer counterparts — and HAS-BLED ignores thrombocytopenia and intracranial metastases. Cancer-specific algorithms are needed and do not exist.
The TBIP framework organises the decision:

Letter	Domain	Content
T	Thrombotic risk	CHA₂DS₂-VASc plus cancer type and stage modifiers; OAC recommended at ≥2 (men) / ≥3 (women) (I/C), considered at 1 / 2 (IIa/C), may be considered at 0 / 1 after bleeding assessment (IIb/C)
B	Bleeding risk	Thrombocytopenia; GI or GU malignancy; intracranial lesions; eGFR <30; HAS-BLED
I	Interactions	P-glycoprotein and CYP3A4 interactions with anticancer agents
P	Preferences	Patient preference and drug availability

(Table: sources/Cardio-Oncology-ESC-2022)

Very high bleeding risk — platelets <25,000/µL, active or recent major bleeding, intracranial lesions — means no anticoagulation, with regular reassessment.
Agent choice. NOAC preferred over LMWH and VKA (Class IIa/B). LMWH preferred over NOAC in unoperated GI or GU cancer, CrCl <15 mL/min, platelets <50,000, and major drug interactions. The ARISTOTLE cancer subgroup (n=1,236) confirmed apixaban superior to warfarin for stroke and systemic embolism; ENGAGE AF-TIMI 48 showed similar findings for edoxaban. (sources/arrhythmia-cardio-oncology-aha-2021)
Rate control. Beta-blockers preferred, particularly where cancer-therapy-related cardiac dysfunction is a risk; diltiazem and verapamil should be avoided where possible because of interactions and negative inotropy. Lenient control (<110 bpm) is acceptable.
Rhythm control. Antiarrhythmics may prolong QTc and interact with anticancer agents. Ablation is considered in selected patients — HF or LV dysfunction with uncontrolled symptoms — weighing cancer status and prognosis in a multidisciplinary meeting. LAA occlusion is Class IIb/C when long-term OAC is contraindicated and life expectancy exceeds 1 year.
Cancer surgery. Post-operative AF in cancer surgery carries 4–5× the recurrence risk over 5 years and long-term thromboembolic risk comparable to non-surgical AF, so anticoagulation should be considered in at-risk patients.
Antiplatelets are not a substitute for anticoagulation in cancer-associated AF (Class III).
BTK inhibitors. Ibrutinib carries a relative risk of 4.69 for AF (meta-analysis of 8 RCTs, n=2,580; incidence 3.5–16%) through BTK, Tec and PI3K inhibition plus enhanced sarcoplasmic reticulum calcium automaticity. Acalabrutinib (4.1%) and zanubrutinib (2%) have lower rates through greater BTK selectivity. Pulse-taking or a rhythm strip at every visit (I/C) and transthoracic echo in anyone who develops AF (I/C) are recommended, and ibrutinib should be interrupted for DAPT and 3–7 days before invasive procedures.

AF after noncardiac surgery

Incidence and consequence. New-onset post-operative AF occurs in 0.4–40% after noncardiac surgery depending on procedure and risk, and carries HR 2.00 for 30-day stroke in the NSQIP registry — comparable to the stroke risk of known pre-operative AF. (sources/periop-aha-2024, rating: very high)
39% recur at 5 years, so distinguishing transient perioperative stress arrhythmia from unmasked paroxysmal AF matters clinically.
Acute management follows standard AF care: beta-blockers or calcium channel blockers first-line for rate, cardioversion for haemodynamic instability.
Anticoagulation (COR 2a) for AF persisting beyond 24–48 h, balancing post-operative bleeding against stroke risk and individualised with CHA₂DS₂-VASc.
Follow-up (COR 2a). All patients with new-onset POAF should have structured cardiology follow-up; Danish registry data showed a 48% reduction in thromboembolism with OAC in post-surgical AF, and prolonged ambulatory monitoring is reasonable where recurrence is suspected.

AF and ACS

Triple therapy becomes dual quickly. In patients with ACS requiring OAC, aspirin should be stopped 1–4 weeks after PCI and a P2Y12 inhibitor continued with the OAC (Class I/B-R). (sources/ACS-AHA-2025, rating: very high)
Trial basis. RE-DUAL PCI, AUGUSTUS, ENTRUST-AF PCI and PIONEER AF-PCI all showed that dropping aspirin reduces major bleeding without increasing all-cause or cardiovascular death, stroke or overall MACE.
Stent thrombosis risk is front-loaded — about 80% of events occur within 30 days of PCI — so aspirin may be continued up to 30 days in patients at high risk.
Agent choices. DOAC preferred over VKA except in mechanical valves or rheumatic mitral stenosis with MVA ≤2.0 cm². Clopidogrel is the preferred P2Y12 inhibitor because the prasugrel and ticagrelor trials excluded OAC patients and more potent agents theoretically increase bleeding.
Duration. P2Y12 therapy continues at least 12 months after aspirin withdrawal, and can be shortened in patients with multiple bleeding risk factors.

AF and stable CAD

In AF with stable coronary disease at least 6–12 months after the index event or PCI, the question is whether OAC monotherapy alone preserves ischaemic protection while removing antiplatelet bleeding. See concepts/AF-Stable-CAD-Antithrombotic for the full synthesis.

AFIRE found rivaroxaban monotherapy non-inferior for ischaemic events and superior for MACE and bleeding, but used non-standard doses (15/10 mg, Japanese approval) in predominantly low-risk patients; the 2023 AHA/ACC guideline gives DOAC monotherapy COR 2b/C-LD beyond 1 year. (sources/CCS-AHA-2023, rating: very high)
EPIC-CAD (n=1,040, standard-dose edoxaban, 12 months):

Outcome	Edoxaban alone	Edoxaban + antiplatelet	Effect
Net adverse clinical events	6.8%	16.2%	HR 0.44 (0.30–0.65); P<0.001; NNT 10.6
Major ischaemic events	1.6%	1.8%	HR 1.23; not significant
Major or clinically relevant non-major bleeding	4.7%	14.2%	HR 0.34 (0.22–0.53)
Major bleeding	1.3%	4.5%	HR 0.32 (0.14–0.73)

(Table: sources/edoxaban-af-cad-nejm-2024, rating: high)

EPIC-CAD was open-label and East Asian only, the net-benefit endpoint favours less potent antithrombotic therapy, and it was underpowered for ischaemic outcomes.
ADAPT AF-DES (2026) extends this to a DES-specific population: 960 patients at least one year after a second- or third-generation stent, NOAC monotherapy versus NOAC plus clopidogrel, net adverse clinical events 9.6% versus 17.2% (HR 0.54; non-inferior and superior), driven by bleeding (5.2% versus 13.2%) with no stent-thrombosis signal. (sources/adapt-af-des-nejm-2026, rating: high)

AF and acute ischaemic stroke

Early OAC (COR 2a). ELAN (n=2,013, open-label, 103 sites) compared early with later DOAC initiation using imaging-defined stroke severity: 30-day composite 2.9% versus 4.1% (risk difference −1.18 percentage points, 95% CI −2.84 to 0.47), with symptomatic ICH 0.2% in both arms and recurrent ischaemic stroke 1.4% versus 2.5%. The trial tested no formal hypothesis — the CI crosses zero, and the recommendation rests on the interval being narrow enough to exclude more than a 0.5-point harm. (sources/elan-doac-timing-nejm-2023, rating: very high; sources/ais-aha-2026, rating: very high)
Timing by severity — trial windows and guideline advice:

Stroke severity	ELAN early arm	ELAN later arm	Guideline position
Minor	≤48 h	Day 3–4	Early initiation reasonable
Moderate	≤48 h	Day 6–7	Early initiation reasonable
Major	Day 6–7	Day 12–14	Delay ≥14 days if NIHSS >15 or large infarct

(Table: sources/elan-doac-timing-nejm-2023; sources/ais-aha-2026)

The severe-stroke row is the least well supported. ELAN excluded parenchymal haematoma at randomisation, enrolled a population with median NIHSS 3, and did not test initiation before day 6 in major stroke; the ≥14-day advice comes from consensus rather than from a randomised arm. (sources/elan-doac-timing-nejm-2023)
Full trial detail and the component outcomes: entities/Ischemic-Stroke.
Heparin bridging is harmful (COR 3). It offers no benefit and increases intracranial haemorrhage; hold heparin and start OAC when the appropriate timing is reached.
DOAC preferred over VKA except mechanical valves and rheumatic mitral stenosis with MVA ≤2.0 cm². (sources/AF-AHA-2023, rating: very high) See entities/Ischemic-Stroke for acute management and concepts/Poststroke-AF-Monitoring for detection strategies.

Contradictions / Open Questions

CASTLE-AF versus CABANA. CASTLE-AF's HR of 0.62 in HFrEF contrasts with CABANA's null all-comers result. The HF trial screened 3,013 patients to enrol 363, was unblinded, and had substantial crossover and few events — the benefit is plausible but the effect size is uncertain. (sources/HF-ESC-2021, rating: very high)
Cancer has no validated risk model. TBIP is a consensus construct, not a validated score, and both CHA₂DS₂-VASc and HAS-BLED are known to fail in cancer. Which patients benefit from anticoagulation, and at what platelet count the trade-off inverts, remains undefined. (sources/Cardio-Oncology-ESC-2022)
Perioperative AF anticoagulation is empirical. The recommendation to start OAC after 24–48 h rests on registry data showing a 48% reduction in thromboembolism, not on a randomised trial, and the 39% five-year recurrence rate is the strongest argument for treating. (sources/periop-aha-2024)
Clopidogrel preference in ACS with OAC is extrapolation. Prasugrel and ticagrelor were not tested in OAC patients; the preference follows pharmacology and bleeding risk rather than head-to-head evidence. (sources/ACS-AHA-2025)
Anticoagulation timing in severe stroke remains unsettled. ELAN's severe-stroke subgroup was small and excluded haemorrhagic transformation; whether early OAC is safe in large infarcts is unresolved.
Guideline drift in HF. ESC 2021 uses CHA₂DS₂-VASc (OAC Class IIa at ≥1 men, ≥2 women), while ESC 2024 adopts CHA₂DS₂-VA, so the HF guideline's threshold and the AF guideline's threshold disagree for women whose only risk factor is sex. (sources/HF-ESC-2021, sources/AF-ESC-2024)

Connections

Part of entities/Atrial-Fibrillation — the parent overview, which carries the general epidemiology, pathophysiology and management framework
concepts/AF-CARE — the four-pillar framework these populations are managed within
concepts/CHA2DS2-VA — the risk score whose thresholds differ between the guidelines cited here
concepts/NOAC-AF-Management — dosing, renal thresholds and interactions that drive agent choice in these groups
concepts/AF-Stable-CAD-Antithrombotic — AFIRE, EPIC-CAD and the OAC-monotherapy question in full
concepts/DAPT-Strategies — the antithrombotic combinations used after PCI
concepts/Cancer-Associated-Arrhythmia — the drug-by-drug cancer arrhythmia framework
concepts/Cardio-Oncology — HFA-ICOS risk stratification and cancer therapy toxicity
concepts/Perioperative-Cardiovascular-Assessment — the wider perioperative framework around POAF
concepts/Poststroke-AF-Monitoring — detection of AF after stroke
concepts/Antiarrhythmic-Drugs — the drugs whose safety differs by population
concepts/Catheter-Ablation-AF — ablation evidence including CASTLE-HTx in end-stage HF
entities/HFrEF — heart failure with reduced ejection fraction
entities/Ischemic-Stroke — acute stroke management and OAC timing
entities/Acute-Coronary-Syndrome — the ACS pathway
entities/LAAO — LAA occlusion, including where it is recommended in these populations