Across 30 years, the question shifted three times

The first question posed in atrial fibrillation care was simple and powerful: should rhythm be restored to sinus? Since the arrhythmia was the disease, correcting it should benefit the patient — a line of reasoning that even had mechanistic backing.

In 2002, that question was answered no. For roughly the next 20 years, rate control became the default approach to atrial fibrillation management.

But the question did not disappear — it was replaced. The first replacement went from "what stops it" to "what solidifies it": anticoagulation became the main battleground, and warfarin was replaced by DOACs. The second went from "what to add" to "how far can you pull back": when coronary artery disease coexisted, the trend moved toward trimming antithrombotic therapy from three drugs to two.

And a third shift is now underway. Ablation is reviving rhythm control in a different form, while implanted devices have revealed a target that did not exist in the ECG era — asymptomatic arrhythmias. From AFFIRM to ARTESiA, 31 trials arranged across six acts.

Act I: restoring sinus rhythm did not change survival

In 2002, two trials appeared in the NEJM almost simultaneously: AFFIRM and RACE.

AFFIRM compared rhythm control with rate control in 4,060 patients with atrial fibrillation at high risk of stroke or death (mean age 69.7±9.0 years). The primary endpoint was all-cause mortality. Five-year mortality was 23.8% vs 21.3%, HR 1.15 (95%CI 0.99–1.34), P=0.08. No significant difference emerged, but if anything the direction favoured against the rhythm group. Hospitalisations and drug adverse events were also more frequent in the rhythm group.

RACE chose a more restricted population: 522 patients with persistent atrial fibrillation that had recurred after electrical cardioversion. What happened here is instructive. At 2.3 years, 39% of the rhythm group were in sinus rhythm versus 10% of the rate group — a clear difference in maintaining sinus rhythm. And yet the primary composite was 17.2% vs 22.6%, with rate control shown to be non-inferior.

In other words, "whether sinus rhythm was maintained" and "patient outcome" did not align within the same trial. It was about as clear a demonstration of the limits of a surrogate endpoint as one could ask for.

Same result in heart failure, and at strict heart-rate targets

The 2008 AF-CHF trial selected the population where rhythm control might be expected to do the most good: 1,376 patients with LVEF ≤35%, heart failure symptoms, and a history of atrial fibrillation. Since atrial fibrillation was known to be a predictor of mortality in heart failure patients, the inference that suppressing it would improve outcomes had reasonable grounding.

The result was completely negative. Cardiovascular death was 27% vs 25%, HR 1.06 (95%CI 0.86–1.30), P=0.59. All-cause mortality was 32% vs 33%, stroke 3% vs 4%, worsening heart failure 28% vs 31%. No prespecified subgroup showed either strategy to be superior. Being a predictor and being modifiable by intervention turned out to be two different things.

So how strictly should rate control be pursued? RACE II (2010) asked this question. In 614 patients with permanent atrial fibrillation, patients were assigned to lenient management (resting heart rate target below 110/min) or strict management (resting below 80/min and moderate exertion below 110/min). The 3-year cumulative primary composite was 12.9% vs 14.9%, absolute difference −2.0 points (90%CI −7.6 to 3.5) — non-inferior.

The real story of this trial lies not in the non-inferiority itself but in the effort required. The heart-rate target was achieved in 97.7% of the lenient group versus 67.0% of the strict group. Total number of visits was 75 in the lenient group (median 0) versus 684 in the strict group (median 2). Trying to lower heart rate strictly increased visits ninefold, still left a third of patients below target, and changed nothing in outcome.

After "how far to lower" comes "with what." RATE-AF (2020, 160 patients) directly compared low-dose digoxin with bisoprolol in older patients with permanent atrial fibrillation and dyspnoea. The primary endpoint, quality of life at 6 months (SF-36 PCS), showed no difference: adjusted mean difference 1.4 (95%CI −1.1 to 3.8), P=.28. Resting heart rate also did not differ: 76.9 vs 74.8/min.

Yet symptom classification differed. A 2-grade improvement on the EHRA classification was reported by 53% on digoxin versus 9% on bisoprolol, adjusted odds ratio 10.3 (95%CI 4.0–26.6). At least one adverse event occurred in 25% vs 64%. Despite reaching the same heart rate, how patients felt and the burden of side effects differed. This must be read with caution, though: only 160 patients, and these were secondary endpoints among 17 items at 6 months and 20 at 12 months.

The question shifts for the first time: from stopping to solidifying

If rhythm control had no advantage, what exactly was being treated in atrial fibrillation? The answer was thromboembolism prevention. And the next question followed naturally — anticoagulation is a nuisance, so could antiplatelet therapy substitute for it?

ACTIVE W (2006) tested this in patients with atrial fibrillation and one or more stroke risk factors, comparing oral anticoagulant (3,371 patients) with clopidogrel plus aspirin (3,335 patients). Designed to test non-inferiority, the result was not merely non-inferiority missed but clear inferiority. The primary endpoint occurred at 3.93% vs 5.60% per year, relative risk 1.44 (1.18–1.76), p=0.0003. Superiority of anticoagulation was clear, and the trial was stopped early.

What's notable is that among patients already on anticoagulation at baseline, the advantage of anticoagulation was even larger (relative risk 1.50), and major bleeding risk was also lower with anticoagulation (interaction P=0.03). There is no advantage — in either efficacy or safety — to switching stably anticoagulated patients to antiplatelet therapy.

The compromise of "aspirin if warfarin can't be used" was also rejected, by AVERROES in 2011. In 5,599 patients deemed unsuitable for vitamin K antagonists, apixaban cut stroke and systemic embolism to less than half that of aspirin (1.6% vs 3.7% per year, HR 0.45, 95%CI 0.32–0.62). And major bleeding was 1.4% vs 1.2% per year, HR 1.13, P=0.57 — not increased. Intracranial haemorrhage was 11 events vs 13 events. This asymmetry — large gains in benefit with almost no increase in harm — was the reason for early termination.