One‐year outcomes of rate versus rhythm control of atrial fibrillation in the Kerala‐AF Registry

Abstract Background There is ongoing debate around rate versus rhythm control strategies for managing atrial fibrillation (AF), however, much of the data comes from Western cohorts. Kerala‐AF represents the largest prospective AF cohort study from the Indian subcontinent. Objectives To compare 12‐month outcomes between rate and rhythm control strategies. Methods Patients aged ≥18 years with non‐transient AF were recruited from 53 hospitals across Kerala. Patients were stratified by rate or rhythm control. The primary outcome was a composite of all‐cause mortality, arterial thromboembolism, acute coronary syndrome or hospitalization due to heart failure or arrhythmia at 12 months. Secondary outcomes included bleeding events and individual components of the primary. Predictors of the composite outcome were analysed by logistic regression. Results A total of 2901 patients (mean age 64.6 years, 51% female) were included (2464 rate control, 437 rhythm control). Rates of the primary composite outcome did not differ between groups (29.7% vs 30.0%; p = .955), nor did any component of the primary. Bleeding outcomes were also similar (1.6% vs 1.9%; p = .848). Independent predictors of the primary composite outcome were older age (aOR 1.01; p = .013), BMI <18 (aOR 1.51; p = .025), permanent AF (aOR 0.78; p = .010), HFpEF (aOR 1.40; p = .023), HFrEF (aOR 1.39; p = .004), chronic kidney disease (aOR 1.36; p < .001), and prior thromboembolism (aOR 1.31; p = .014). Conclusion In the Kerala‐AF registry, 12‐month outcomes did not differ between rate and rhythm control cohorts.


Despite atrial fibrillation (AF) being very common in clinical practice
globally, most study data come from Western cohorts.The Indian state of Kerala has seen rapid development and epidemiological transition recently, resulting in improved longevity and consequently a shift in mortality risk from communicable to non-communicable causes such as cardiovascular disease. 1,2The estimated mortality rate due to cardiovascular disease in Kerala is now higher than in many Western countries, 1 emphasizing the need for effective treatment and prevention.
The Kerala-AF registry 3 was designed to provide a prospective analysis of patterns, treatment, and outcomes of AF in the Kerala region, with which to inform changes in practice.This is the largest prospective AF cohort from the Indian subcontinent.
5][6] The 'B' component of the pathway involves a decision on whether to pursue restoration of sinus rhythm (rhythm control) or accept a state of permanent AF and focus on normalizing the heart rate (rate control).
In this ancillary analysis from the Kerala-AF registry, we assessed 12-month outcomes, as stratified by rate versus rhythm control strategies.

| Study design and cohort
The design of the Kerala-AF registry has been described previously. 3Briefly, patients aged ≥18 years with documented AF were recruited between April 2016 and April 2017 across 53 hospitals in the Kerala region of India.Patients were recruited during attendance or admission to a recruiting hospital.Patients with transient AF (e.g., due to myocardial infarction, sepsis, or post-operative AF) were excluded, as were critically ill patients with life expectancy <30 days.
The study was conducted in accordance with the Indian Council of Medical Research guidelines, and the ethical principles of the Declaration of Helsinki.All participants provided informed consent and only de-identified data were shared for analysis.

| Definition of rate versus rhythm control
Sites were asked to report rate or rhythm control on the study proforma, however discrepancies were noted during data analysis.
Hence, for the purpose of this analysis, rhythm control was defined as (a) use of class I antiarrhythmic drugs; (b) use of catheter ablation; or (c) site definitively reported a rhythm control strategy.Thus, 117 patients classified as rhythm control were classified as having 'permanent AF'.As the definition of permanent AF means that rhythm control has been discontinued, these patients were reclassified to 'persistent AF'.

| Study endpoints
The primary outcome in this analysis was a composite of major adverse cardiac events (MACE)-defined as all-cause mortality, cerebrovascular accident (CVA), transient ischaemic attack (TIA), systemic embolism (SE), acute coronary syndrome (ACS) or hospitalization due to heart failure or arrhythmia-over 12 months of follow-up.
Secondary outcomes included a composite of bleeding events, including gastrointestinal (GI) bleeds, intracranial haemorrhage (ICH), or other minor bleeding events.Individual endpoints within the primary composite outcome were also reported as secondary outcomes.

| Statistical analysis
Continuous variables were described as mean ± standard deviation and compared using t-tests.As all continuous variables were reasonably normally distributed, medians and non-parametric tests were not applied.Categorical data were described using counts and/or percentages and compared using two-sample proportion tests, Chi square tests or Fisher's exact test.
A loss to follow-up was considerable in this dataset, we applied an algorithm to determine which patients to include in each outcome analysis, as follows: (a) if the patient had the event of interest, regardless of follow-up duration, they were included; (b) if the patient completed 12-month follow-up without having the event, they were included; (c) if the patient died during follow-up, they were included; (d) otherwise they were excluded.This approach ensured that patients with a known outcome, despite loss to follow-up, could be included in analysis.Sensitivity analyses to assess the impact of loss to follow-up were performed to ensure robustness of findings; (a) assuming all lost-to-follow-up patients remained event-free, and (b) assuming all lost-to-follow-up patients experienced the primary composite endpoint.
Missing data were handled either by removal of the variable, simple imputation, or multivariable imputation by chained equations (MICE).As hospital ID was missing for 5 patients, these were classified into a separate 'unknown hospital' category.Patients with apparent considerable data entry errors were excluded.
The primary outcome was analyzed by logistic regression.AF management strategy, along with variables with p < .1 on univariable regression were entered into a multivariable logistic regression model.To account for clustering by hospital, a marginal regression model using generalized estimating equations with an exchangeable covariance structure was applied.p-Values <.05 were considered statistically significant.Statistical analysis was performed in Python and R.

| RE SULTS
The original Kerala-AF dataset included 3421 patients.One record was removed due to data entry errors.Using the algorithm described above to account for loss to follow-up, the primary outcome cohort consisted of 2901 patients (mean age 64.6 years; 51% female), of whom 2464 were assigned to a rate control strategy and 437 were assigned to a rhythm control strategy.The cohort sizes for secondary outcomes differed slightly due to the application of the loss to follow-up algorithm for these different outcome measures.
Demographic and clinical differences between the primary outcome cohorts are shown in Tables 1 and 2. Hypertension was more common in the rhythm control group.Left atrial size was larger in the rate control group, as was history of thromboembolism, and valvular AF.Reasons for index consultation were similar in most cases, though the rate control group was more often related to valvular heart disease (7.8% vs 3.4%; p = .002).Most anticoagulation in both groups was achieved using Vitamin K anticoagulants (VKAs) such as Warfarin.Non-vitamin K antagonist oral anticoagulants (NOACs) were infrequently used in either arm.Overall the cohort was high at risk for adverse events, with a high prevalence of chronic kidney disease and prior thromboembolism.
There was also no difference in all-cause mortality (rate control 14.9% vs rhythm control 15.0%; p = .948).Causes of death were mostly classified as cardiac in both groups (74.2% vs 78.5%; p = .539).Stroke accounted for 14.7% of deaths in the rate control group and 7.7% in the rhythm control group (p = .169).The remainder were classified as 'other' or 'unknown'.The composite bleeding outcome was also no different between groups (rate control 1.6% vs rhythm control 1.9%; p = .848).

| Predictors of the composite MACE outcome
The results of univariable and multivariable logistic regression analyses, predicting the composite MACE outcome, are shown in Table 3.
On univariable regression, age (p < .001),heart failure, both HFpEF and HFrEF (p < .001),diabetes mellitus (p = .011),ischaemic heart disease (p < .001),chronic kidney disease (p < .001),prior thromboembolism (p = .023)and use of class III antiarrhythmics (p = .001)were associated with increased risk of the composite MACE outcome.BMI was analyzed categorically as the linearity assumption was not met.As compared to a reference of normal BMI (18-24), being underweight significantly increased the risk of composite MACE events (p = .013).The other weight categories showed no significant association.Rhythm control strategy, as compared to rate control, was non-significant (OR 0.98; 95% CI 0.79-1.21;p = .841).

| Sensitivity analyses
Sensitivity analyses, shown in the supplementary material, demonstrated minor changes to statistical significance, but did not substantially alter our overall findings.In the minimal loss effect model (assuming all patients lost-to-follow-up remained event-free), the sole difference was that permanent AF was no longer significantly associated with the composite MACE outcome, however, the trend remained in the same direction.In the maximal loss effect model (assuming all patients lost-to-follow-up experienced a composite MACE outcome), underweight BMI became non-significant and overweight BMI became significant, and prior thromboembolism became nonsignificant, though the overall point estimates and confidence intervals were broadly similar to the main analysis.Additionally, the point estimate of catheter ablation was much closer to no effect.

| DISCUSS ION
The main findings from this ancillary analysis from the largest prospective AF cohort from the Indian subcontinent, are as follows: (i) Notably, the overall MACE and mortality rates in our study were very high; more than might be expected for this region.This is likely, in part, reflective of the fact that patients were recruited at the time of hospital admission, resulting in an overall 'sicker' cohort.This is demonstrated by the high rates of comorbidities such as hypertension, diabetes mellitus and chronic renal disease, as shown in Table 1.

| The impact of rhythm control
0][11] Nonetheless, older studies failed to demonstrate any benefit to rhythm control over a rate control strategy. 12This may be due to evolution in the landscape of preventative medicine, as well as improvements and more widespread use of techniques such as catheter ablation.
Ablation is known to be more effective than antiarrhythmic drugs at maintaining sinus rhythm and, once the small up-front it may be speculated that this could reflect a symptom-based approach to AF management-that is, asymptomatic patients may not be referred for rhythm control-however, reported symptoms were somewhat more common in the permanent group (42% vs 39.8%) and were no different between the rate and rhythm control cohorts (86% vs 85%; p = .683).As the very long-term effects of AF, particularly on neurocognitive outcomes, are increasingly recognized, 13 these figures may reflect a need to promote a rhythm control strategy in the younger and healthier populations.
Furthermore, comprehensive management following the Atrial fibrillation Better Care (ABC) pathway, especially in terms of cardiovascular risk optimization-cannot be overlooked.The findings in the Kerala-AF registry may relate to relative undertreatment.
For example, NOACs were rarely utilized and many patients were not anticoagulated despite elevated stroke risk scores.Given the high rates of cardiovascular disease in the Kerala state, primary preventative efforts are essential.A recent survey showed that many preventative medications are either unaffordable or unavailable within Kerala, 14 demonstrating the need for improvement.
Our analysis demonstrates that, at this point in time and over 1 year of follow-up, rhythm control of AF does not confer a prognostic advantage in the Kerala-AF cohort.The focus of healthcare improvements going forward should likely be on primary prevention to optimize overall cardiovascular risk.Improving access to modern therapies, such as NOACs and catheter ablation, may also be of benefit.

| The effect of BMI on cardiovascular risk
In our study, underweight BMI was associated with an increased risk of composite MACE outcomes (aOR 1.51 [95% CI 1.05-2.18];p = .025).This may be explained by malnourishment, which is known to be associated with increased cardiovascular risk. 15,16Alternatively, patients with chronic health conditions, such as cancer, frequently lose weight rapidly and thus may fall into the underweight category, resulting in bias towards mortality risk.
Notably, the point estimates for overweight and obese BMIs were inversely associated with the primary composite outcome, but not to statistical significance.It is well established that obesity is a risk factor for cardiovascular disease.In the past, the concept of the "obesity paradox" has been described-that is, that obesity may somehow be protective in those with established cardiovascular disease.This paradox is partly due to the fact that BMI fails to account for body composition-for example, a bodybuilder may be classed as overweight due to high volume of lean muscle.A recent study showed that alternative measures of adiposity, such as waist-toheight ratio, may be more reliable. 17 general, promoting a target normal BMI (18-24) by eating a healthy diet and regularly exercising, remains the most logical recommendation.TA B L E 3 (Continued) Demographic differences between rate and rhythm cohorts.