AF is the most common heart rhythm disorder, with a prevalence of 2–4% worldwide, and it is a leading cause of emergency department presentations and unplanned hospitalisations.1,2 The clinical impact of AF varies significantly, both across patients and over time within individuals. Heart failure (HF), a common sequela of AF, affects up to 40% of AF patients and represents the leading cause of mortality in this group.3
The relationship between AF and HF is complex. AF can precipitate HF by increasing the ventricular rate, promoting irregular ventricular rhythm and impairing atrial mechanical function. Conversely, elevated left ventricular (LV) pressures from HF can promote atrial structural remodelling, initiating AF. Both conditions often begin insidiously and sub-clinically, complicating the determination of the initiating factor. AF onset is temporally associated with the risk of HF hospitalisation, precipitating concurrent diagnosis.4 Clarifying this ‘chicken-and-egg’ dilemma is clinically important, as patients with AF-mediated HF may significantly benefit from timely rhythm control therapies such as catheter ablation (CA).5
However, not all patients with AF and HF will benefit from rhythm control and in the absence of guidelines for prospective stratification, many physicians limit the use of CA to patients in whom they are confident the HF is AF-mediated (Figure 1 ).6
Limitations of Current Definitions
Tachycardia-induced cardiomyopathy is a reversible cardiomyopathy that resolves on cessation of the tachycardia.7 However, tachycardia is not quantitatively defined and can be especially difficult in the setting of AF. The time course, clinical manifestation and cellular and neurohormonal changes can also vary between diagnosed patients.
AF-induced cardiomyopathy (AIC) is a composite mechanism that can also only be retrospectively diagnosed when LV systolic dysfunction (LVSD) persists after rate control in AF but reverses after rhythm control.8 Therefore, establishing prerequisite heart rate (HR) thresholds to define tachycardia and rate control is a prerequisite for practically defining AIC. Definitions based on retrospectively detected features are helpful for classification and downstream management in the setting of arrhythmia recurrence. However, prospectively distinguishing tachycardia-induced cardiomyopathy from AIC would help to personalise index treatment strategies, stratifying patients who are more likely to respond to rate control rather than rhythm control.
Retrospective, observational studies sought to identify pre-CA markers in patients whose left ventricular ejection fraction (LVEF) on echocardiography improves after CA, an objective, quantifiable surrogate for cardiomyopathy.9 Features such as left atrium (LA) size or co-existent causes of LVSD, such as known HF aetiology or electrical dyssynchrony of ventricular contraction have been proposed as predictors of a lack of response to CA. However, these studies are confounded by patient selection and by the association of these features with AF recurrence, which may affect post-CA remodelling.10 While these features are associated with response to rhythm control, many others may benefit from CA that are difficult to identify prospectively. In addition, AF can mediate cardiomyopathy in the presence of other HF causes and so excluding patients from CA based on these co-existent causes may result in undertreatment.
Understanding the direct mechanisms through which AF can cause cardiomyopathy may help to identify tools to guide stratification when both are contemporaneously diagnosed. This narrative review aims to summarise the mechanisms underlying AIC and to propose practical tools to rule in patients who would benefit from CA.
Mechanisms of AF-induced Cardiomyopathy
Understanding the mechanisms by which AF causes LVSD may help identify at-risk patients. Three key potential contributors are proposed: increased ventricular rate; variation in ventricular diastolic interval (R-R irregularity); and loss of active atrial contraction.
There is also variable expression of LVSD in patients with AF. Whether this is due to the heterogeneity in the above characteristics or an intrinsic vulnerability in some patients to LVSD in AF is unknown. Therefore, an additional mechanism for AF-mediated LVSD should be considered:
Structural Vulnerability or Predisposition
Defining Tachycardia During AF: Looking Beyond the Mean
Untreated AF is associated with a rapid ventricular rate in the absence of significant conduction system disease. In pacing models, rapid atrial pacing causes a dilated cardiomyopathy with LVSD, which is reversed on cessation.11 Tachycardia-induced cardiomyopathy in the context of AF is a well-established component phenomenon with demonstrable reverse remodelling with rate control alone.12,13 Potential mechanisms have been proposed, including impaired LV filling during a shortened diastolic interval, increased myocardial demand resulting in ischaemia, pressure overload or a combination of these7
Translating this in clinical trials of patients with AF and HF has been challenging due to difficulties defining and achieving meaningful, sustained rate control. AIC and AF-mediated cardiomyopathy are distinguished from an isolated tachycardia-induced cardiomyopathy when the denoted improvement from sinus rhythm is compared to LVEF during rate-controlled AF.8 However, in the absence of a clear definition of rate control in the context of HF, comparisons between AIC and solely tachycardia-induced cardiomyopathy in the setting of AF cannot be objectively made.
A sub-analysis of the RACE II trial in HF patients demonstrated no greater improvement in ventricular remodelling or clinical HF outcomes with a strict rate-control target of mean HR <80 BPM.14 As well as being challenging to achieve, the mean HR is an inappropriate characterisation of the non-parametric R-R interval distribution during AF. During AF, the R-R interval histogram is commonly positively skewed and the mean HR overestimates the efficacy of rate control as a result (Figure 2 ). The achieved difference in mean HR compared to the lenient rate control cohort was only 10 BPM, suggesting minimal difference in HR profiles between the two groups.15
As a result, the effective threshold to define rate control is unclear. In the absence of demonstrable benefit from intensive rate control, contemporary guidelines propose a mean rate of 100 or 110 BPM not specific to patients with HF.16
The irregular rhythm results in a non-parametric distribution of R-R intervals which is better represented using a non-parametric, quantitative measure. As a result, our group has tested a novel, non-parametric marker of HR distribution, called the Restitution Threshold Index (RTI). This is the percentage of R-R intervals that are shorter than 660 ms during a 10-minute resting recording during AF in an outpatient clinic setting (Figure 3). Where patients were already on β-blockers or other atrioventricular (AV)-nodal blocking drugs, this was documented, and no acute changes were made before recording. The 660 ms threshold was derived empirically based on its capacity to discriminate between patients with AF with concurrent LVSD (LVEF <50%) and those with preserved LVEF (>50%).17 Sensitivity analyses demonstrated robust predictive performance across thresholds of 400–1,200 ms and an optimal cut-off at 660 ms for differentiating the two arms (Figure 4). Internal validation against alternative Holter recording durations and protocols was robust and cross-validation in an external dataset is on-going. This threshold value of 660 ms is in keeping with in silico and preclinical haemodynamic trials that demonstrate R-R intervals below this threshold impair myocardial contractility and LV output.18,19
RTI reflects the absolute proportion of implicated beats in a non-parametric distribution quantified by a single value. The mean value is only appropriate for normally distributed samples and in the AFHF trial cohort, it did not discriminate between the two groups.17 The RTI also prospectively predicted patients with AIC with a positive predictive value of 0.97 and an area under the receiver operator curve (AUROC) of 0.74. Whether RTI can serve as a definitional threshold for clinically meaningful tachycardia during AF sufficient to induce LV systolic dysfunction remains to be determined. Further work is needed to validate the optimal cut-off through interval imaging studies during stepwise rate control. If such a nadir can be confirmed, RTI may offer a prospective, rate-independent marker of AIC, reflecting intrinsic susceptibility to AF-mediated dysfunction rather than simply rate burden.
Measuring Irregularity of the Ventricular Rate
The effect of R-R interval irregularity on haemodynamic and myocardial function has been tested in preclinical models. Regular versus irregular stimulation of ventricular cardiomyocytes was tested in rats while maintaining a constant mean HR.20 Significant downregulation in the mRNA and protein expression required for intracellular calcium handling was observed in irregularly paced cardiomyocytes, which is associated with impaired contractile function. Although the mean HR (120 BPM) was the same in both rat models, irregular stimulation ranged from 72 BPM to 192 BPM, with a proportion of short R-R intervals indicative of tachycardia. However, this may reflect a greater positive skew in the R-R interval distribution, i.e. a higher RTI. Therefore, it is unclear whether the effect of irregularity reflects a greater burden of shorter R-R intervals driving the dysfunction.
The impact of RR interval variability during AF is dynamic and appears greater at higher ventricular rates. Human studies in patients with AV nodal block and LV dysfunction demonstrated that irregular ventricular pacing significantly reduced contractile function at higher heart rates (120 BPM), but not at lower rates (80 BPM).21 This rate-dependent effect likely relates to changes in the myocardial force-interval relationship: at higher rates, short R-R intervals cause larger fluctuations in contractility, negatively affecting cardiac function, while at lower rates variations in interval length have a minimal effect (Figure 5).
Beat-to-beat haemodynamic studies during AF have also shown this curvilinear relationship with cardiac output and characterised ventricular rate irregularity as the relative burden of short, culprit R-R intervals that impair cardiac output versus beats above the threshold that do not, underscoring the rationale for the RTI.18
Atrial Contraction
The LA serves as a reservoir for pulmonary venous return during ventricular systole. During early diastole, it serves as a passive conduit of venous blood to the LV, which is then boosted by active atrial contraction during ventricular end-diastole in sinus rhythm. This ‘atrial kick’ enhances ventricular systolic output via the Frank-Starling mechanism without requiring continuous high LA pressure during diastole. AF limits the LA to a ‘pipeline function’ status in the haemodynamic chain. This may be expected to result in reduced LV filling and reduced LVEF. Phasic assessment of LV filling using cardiac MRI in healthy volunteers quantified a 10–40% contribution from active LA contraction.22 In a pacing-dependent dog model, atrial pacing increased LV filling by 22%.23
However, these models do not account for physiological compensation and chronic adaptation of passive LA function during AF and so overestimate the net loss from fibrillating atria. The deleterious effect of an absent atrial kick on LVEF is complex and depends on individual LA size, pressure and fibrosis burden, which are also influenced by AF. These factors also influence the extent of reverse remodelling and recovered atrial function, especially if compounded by iatrogenic LA fibrosis in the context of extensive CA. Therefore, measuring ‘hibernating’ LA active function during AF may poorly correlate with the resumptive function. The LA also exerts neurohormonal homeostatic effects, and the differential effects of AF on neurohormonal mediators of cardiac function are unclear.24
Structural Vulnerability to AF-induced Cardiomyopathy
Many patients tolerate AF, with only a proportion developing LVSD in the context of AF. Although rhythm-related features may contribute to this, it is reported anecdotally that rapid ventricular rates that are well above the consensus definition of tachycardia do not always result in cardiomyopathy. This suggests that individual features may also predispose to the development of AIC and even protect against it.
A double-hit hypothesis with genetic vulnerability has been shown to underlie both alcohol-induced cardiomyopathy and peripartum cardiomyopathy.25,26 Loss-of-function mutations in the titin gene are attributed to the risk of dilated cardiomyopathy and early onset AF. However, a genetic basis for AIC has not been explored. This may be limited by the challenge of delineating a clean cohort of patients with AIC as it can only be diagnosed retrospectively.
Identifying phenotypic vulnerabilities to AIC is also limited by the absence of prospective imaging in sinus rhythm before AF onset. However, retrospective characterisation is feasible by re-imaging patients after restoration of sinus rhythm and AIC is confirmed. Our group have reported that patients with AIC demonstrate sub-clinical features of cardiac dysfunction even though LVEF normalisation is achieved.27 This can include impaired global LV longitudinal strain, with a relative apical sparing pattern and LV diastolic dysfunction, characterised by abnormal LA reservoir strain.28 Although these features may reflect an underlying cardiomyopathic process that manifests as occult cardiomyopathy in the context of AF, they may reflect incomplete reverse remodelling of the interval AIC. A prospective study would be challenging but would help to clarify their prognostic value further.
The prospective CAMERA-MRI trial demonstrated an association between the absence of LV late gadolinium enhancement (LGE) on baseline cardiac MRI and reverse remodelling with improvement in function after CA.29 This suggests that AIC is not associated with LV LGE, a surrogate marker of myocardial fibrosis. However, the LV LGE did regress at the 6-month post-CA timepoint when the primary endpoint was reassessed, and whether this would continue over time is unknown.30
Contractile reserve on stress echocardiography is the capacity to increase LVEF during exercise or pharmacological stress. This feature has been associated with an improvement in LVEF in patients with LVSD undergoing HF treatments, such as cardiac resynchronisation therapy.31 A contractile reserve of ≥5% has also been shown to be associated with an improvement in LVEF after CA and can prospectively predict AF-mediated cardiomyopathy with an AUROC of 0.85.32 No static measure on echocardiography has been prospectively shown to be associated with improvement in cardiac function after CA. Whether contractile reserve on stress echocardiography and LGE on cardiac MRI present different methods to characterise the same vulnerability or different phenotypic features is unknown and further multi-modal studies in this patient group would be helpful.
Towards a Prospective Unified Definition of AF-induced Cardiomyopathy
Many physicians will only consider CA of AF when they are confident that it is the cause of the HF. Therefore, developing prospective risk stratification tools to characterise AIC is essential.
A combination of rhythm-based and imaging-based parameters has been discussed and their stratifying value has been considered independently. However, the mechanism of AF-mediated cardiomyopathy may be complex and reflect a synergistic effect of rhythm-related factors with structural vulnerabilities, best characterised by multiple measures that capture the different mechanisms of effect.
This may include a measure of ventricular rate burden, using RTI in combination with contractile reserve on exercise echocardiography or LGE burden on cardiac MRI. This could be considered in combination with the patient’s clinical history, which could identify possible competing aetiologies.
Clinical Implementation
Clinical risk stratification tools should provide an estimate of the likelihood of treatment response. Validated tools should ideally be integrated into a patient’s care pathway to support informed decision-making. While the decision to pursue CA or not should primarily be the patient’s, it is the clinician’s responsibility to use available tools to provide sufficient information for the patient to make that decision. Prioritisation of access to CA is another consideration in a resource-limited setting. Estimating the likelihood of response is important because the benefits of restoring sinus rhythm appear to be time-sensitive in HF patients and can be considered during triage of the caseload.5
Implementing risk stratification tools also relies on their accessibility and affordability. RTI is a cheap, non-invasive measure obtained from a 10-minute ECG recording. The investigation can be performed at the bedside or in the outpatient setting. The demonstrated reproducibility suggests a high correlation between RTI measurements recorded in the same environment. To enable independent testing and reproducibility, an open-access web application (https://restitutionthreshold.com) is available that can derive RTI from RR-interval-level or voltage-time ECG data. The reported clinical usefulness of RTI is based on a single-centre prospective study. External evaluation of its application on uncontrolled Holter recordings is on-going and validation studies in independent cohorts are needed.
Different institutions will have access to different tests, so the pathway should also be designed with local availability in mind. MRI and specialist echocardiography services are resource-limited, with some services and regions having no access to them. Therefore, it may be most efficient to measure the RTI initially and if the result is uncertain or further characterisation would be appropriate, an imaging test can then be arranged (Figure 6).
A risk stratification tool’s negative predictive value is important when the patient’s and the physician’s bias is to act. We evaluated the compounding value of sequential testing. This might be used to avoid exposing the patient to the unnecessary risk of a procedure with little chance of benefit.
Expedited Rhythm Control without Stratification
The prognostic benefit of rhythm control treatments in persistent AF, including CA, appears to be time-sensitive, with early treatment associated with the greatest likelihood of structural improvement and prognostic benefit in patients with HF.5,33 Stratifying patients can delay treatment decisions. Further, stratifying tools can only provide a probabilistic outcome and patients with severe symptoms and low probability of benefit may still make an informed decision to pursue invasive rhythm control treatments if available. As the safety profile of CA continues to improve with the introduction of cardio-selective pulsed field ablation and advanced peri-procedural imaging tools such as intracardiac echocardiography, the risk–benefit balance for undertaking the procedure should be made using up-to-date information and with input from the electrophysiologist who would undertake it. Therefore, the decision to undertake risk stratification tests must be made in collaboration with the patient and specialist to ensure that the results will inform decision-making.
Conclusion
Current retrospective definitions of AIC, although conceptually accurate, have limited practical value. Prospective stratification using the RTI and complementary imaging markers addresses these limitations by quantifying critical aspects of AF-induced ventricular dysfunction and underlying myocardial vulnerability.
While further prospective validation is essential, a multi-dimensional approach offers a strategy toward personalised treatment decisions based on an individual’s AF and structural phenotype. Future studies should focus on refining measures based on the pathophysiology of AIC and investigating genetic predispositions to enhance the clinical applicability and accessibility of prospective risk stratification when AF and HF are diagnosed together.
Clinical Perspective
- AF can cause left ventricular systolic dysfunction through several mechanisms beyond simple tachycardia. These may include R-R interval irregularity, impaired atrial contraction and intrinsic myocardial vulnerability, which highlights the need for mechanistic, rather than rate-based, assessment.
- The RTI, obtained from a 10-minute ECG performed at the bedside or outpatient clinic, provides a practical measure of ventricular rate burden that correlates with AF-mediated ventricular dysfunction and may guide early identification of patients likely to benefit from rhythm control.
- Imaging markers such as contractile reserve on exercise echocardiography or absence of late gadolinium enhancement on cardiac MRI reflect underlying structural vulnerability and complement rhythm-based metrics.
- A combined approach using the Restitution Threshold Index and imaging markers may allow clinicians to prospectively identify patients in whom early catheter ablation is most likely to reverse left ventricular dysfunction.