Review of the RACE Trial

Rate Control versus Rhythm Control in Patients with Recurrent Persistent Atrial Fibrillation

N Engl J Med 2002;347:1834-1840

Background: By the late 1990s, the management of persistent atrial fibrillation rested on a clinical preference for restoring sinus rhythm. Repeated electrical cardioversion combined with prophylactic antiarrhythmic drugs was the standard approach, on the assumption that sinus rhythm itself was prognostically beneficial; it would prevent tachycardia-induced cardiomyopathy and heart failure, reduce the risk of stroke, and potentially allow discontinuation of anticoagulation. The competing strategy was simply to accept the arrhythmia, control the ventricular rate with AV nodal blocking agents, and continue anticoagulation indefinitely. The rhythm-first preference was widely held but not evidence-based, and was complicated by the well-documented limitations of antiarrhythmic drugs: frequent recurrences of atrial fibrillation, proarrhythmia, and extracardiac toxicity. A separate concern was that the rhythm itself might not be the principal determinant of prognosis, that the underlying cardiovascular substrate, not the surface ECG, drove the outcomes that mattered.

The RACE trial was designed to test, prospectively and as a noninferiority comparison, whether rate control was at least as effective as rhythm control in preventing major cardiovascular events in patients with recurrent persistent atrial fibrillation. It was published in the same issue of the New England Journal of Medicine as AFFIRM.

Patients: Eligible patients had recurrent persistent atrial fibrillation or flutter (defined as non-self-terminating arrhythmia requiring electrical cardioversion to obtain sinus rhythm), in whom oral anticoagulation was not contraindicated. Patients were required to have undergone one electrical cardioversion during the previous two years, with a maximum of two. Arrhythmia lasting longer than one year was an exclusion, as were NYHA class IV heart failure, current or prior amiodarone treatment, and existing pacemaker. The trial deliberately enriched for patients with documented recurrence — these were not first-presentation patients but patients in whom maintaining sinus rhythm had already proved difficult at least once.

Baseline Characteristics: The trial enrolled 522 patients across 31 centers in the Netherlands and randomized them to rate control (n=256) or rhythm control (n=266). The mean age was 68 years and 63.4% were men. Atrial fibrillation was the index arrhythmia in 93%, with atrial flutter in 7%. The median total duration of atrial fibrillation was approximately 320 days, with the current episode having lasted a median of about 33 days. Coronary artery disease was present in 27.4%, prior myocardial infarction in 15%, and valvular disease in 17%. About half the cohort (50.2%) had a history of heart failure, though NYHA class was I or II in 97%. Cardiomyopathy was present in 9%, hypertension in 49%, chronic obstructive lung disease in 20%, diabetes in 10.5%, and prior cerebrovascular accident in 14%. The mean heart rate was 90 bpm, and about a quarter of patients had a baseline rate above 100 bpm. Echocardiographic findings showed moderately enlarged left atria (mean 45 mm) and preserved fractional shortening (30%). Ninety percent of patients in the rate-control arm and 91% in the rhythm-control arm had at least one risk factor for stroke. The two groups were generally well matched, with one significant imbalance: hypertension was substantially more common in the rhythm-control group (55% vs. 43%; p=0.007). The authors performed a sensitivity analysis adjusting for this imbalance, which did not change the conclusion.

Procedures: The trial was open-label. Patients were seen at 1, 3, 6, 12, and 24 months and at study end. In the rate-control arm, digitalis, non-dihydropyridine calcium-channel blockers, and beta-blockers were used alone or in combination, targeting a resting heart rate below 100 bpm. Cardioversion or AV node ablation with pacemaker implantation was permitted only for intolerable symptoms, unacceptable drug adverse effects, or progressive LV dysfunction.

In the rhythm-control arm, the protocol followed a stepwise sequence. Patients underwent electrical cardioversion without prior antiarrhythmic drug pretreatment, then received sotalol (160-320 mg daily based on body weight and renal function). If atrial fibrillation recurred within six months, sotalol was replaced by flecainide (200-300 mg daily) or propafenone (450-900 mg daily) and cardioversion was repeated. If recurrence happened within six months on this second drug, amiodarone was loaded (600 mg daily for four weeks) and cardioversion was repeated, with maintenance at 200 mg daily thereafter. This sequential drug protocol is more structured than the physician-choice approach used in AFFIRM and is worth noting as a methodological strength.

Anticoagulation with acenocoumarol or fenprocoumon (target INR 2.5-3.5 — higher than the modern 2.0-3.0 target) was required from four weeks before until four weeks after each cardioversion. After one month of sustained sinus rhythm, anticoagulation could be stopped or switched to aspirin. In practice, anticoagulation use during follow-up was 96-99% in the rate-control arm and 86-99% in the rhythm-control arm, so the planned discontinuation occurred in a meaningful but limited fraction of rhythm-control patients. Mean follow-up was 2.3 years, with a maximum of three years.

The trial was supported by the Center for Health Care Insurance and the Interuniversity Cardiology Institute, with an unrestricted grant from 3M Pharma.

Endpoints: The primary endpoint was a composite of death from cardiovascular causes, heart failure (defined as left or right ventricular failure necessitating hospitalization), thromboembolic complications, bleeding (defined as hemoglobin drop >2 g/L, transfusion or hospitalization necessary, or fatal), pacemaker implantation, and severe adverse effects of antiarrhythmic drugs (including torsade de pointes, sustained ventricular tachycardia or fibrillation, 1:1 AV conduction during flutter, third-degree AV block, sick sinus syndrome, digitalis intoxication, and drug-induced heart failure). All deaths were considered cardiovascular unless an unequivocal noncardiac cause was identified — an approach that biases mortality numerator estimates upward but does so equally in both arms. An independent committee blinded to treatment assignment adjudicated all endpoints.

Analysis was performed based on the intention-to-treat principle. The noninferiority margin was prespecified as a 10-percentage-point absolute difference in the upper bound of the 90% confidence interval, with a target sample of 260 patients per group based on an assumed 30% primary event rate at follow-up.

The objective was to show the noninferiority of rate control compared to rhythm control.

Results: At the end of follow-up, 39% of patients in the rhythm-control arm were in sinus rhythm and 44% were in atrial fibrillation (with 18% scheduled for further cardioversion). In the rate-control arm, only 10% were in sinus rhythm at end of follow-up - half through cardioversion for intolerable symptoms and half spontaneously. Patients in the rhythm-control arm underwent a median of 2 electrical cardioversions (range 0-7). The mean resting heart rate was 82 bpm in the rate-control arm and 73 bpm in the rhythm-control arm, but this difference reflected the underlying rhythm (sinus rhythm: 66 bpm; atrial fibrillation: 85 bpm) rather than a true treatment-related effect on rate.

The primary endpoint occurred in 44 patients (17.2%) in the rate-control arm and 60 patients (22.6%) in the rhythm-control arm. The absolute difference was −5.4 percentage points (90% CI −11.0 to 0.4), confirming noninferiority of rate control and approaching superiority. The adjusted analysis accounting for the hypertension imbalance produced an absolute difference of −4.2 percentage points (90% CI −10.0 to 1.5). The hazard ratio for rate control versus rhythm control was 0.73 (90% CI 0.53-1.01; p=0.11). The Kaplan-Meier curves separated modestly throughout follow-up, with the rhythm-control arm accumulating events at a slightly faster pace.

Decomposition of the composite endpoint shows where the difference lives. Cardiovascular death was essentially identical (7.0% vs. 6.8%). Heart failure was not significantly different (3.5% with rate control vs. 4.5% with rhythm control). Thromboembolic complications occurred in 14 rate-control patients versus 21 rhythm-control patients (5.5% vs. 7.9%; absolute difference −2.4 percentage points). Bleeding was numerically higher in the rate-control arm (4.7% vs. 3.4%) but not significantly so. The two components that drove the rhythm-control disadvantage were severe adverse effects of antiarrhythmic drugs (0.8% vs. 4.5%; absolute difference −3.7 percentage points; 90% CI −6.0 to −1.4) and pacemaker implantation (1.2% vs. 3.0%; absolute difference −1.8 percentage points). In other words, the rhythm-control strategy did not produce more deaths or strokes — it produced more drug toxicity and more pacemaker implantations, both of which are direct consequences of the strategy itself.

The thromboembolism findings deserve particular attention. Of 35 thromboembolic complications, all occurred in patients with stroke risk factors. Six of these events — all in the rhythm-control arm — occurred after anticoagulation had been stopped because sinus rhythm appeared maintained, and in five of those six the patient was still in sinus rhythm at the time of the stroke. Twenty-three events occurred during inadequate anticoagulation (INR <2.0). Seventy-three percent of patients with thromboembolic events had atrial fibrillation at the time of the event. Twenty of 21 bleeding episodes occurred during anticoagulation, with INR >3 in 17 cases. This pattern — strokes occurring after warfarin discontinuation in apparently rhythm-controlled patients — is the same finding that AFFIRM produced in the same year and has held up as one of the most clinically actionable results in the AF literature.

Among patients in the rhythm-control arm, outcomes did not differ significantly according to whether sinus rhythm was maintained or atrial fibrillation recurred at end of follow-up. This is a critical observation: the success of rhythm control as measured by the surface ECG did not translate into prognostic benefit.

Post hoc subgroup analyses showed two notable signals. Among women, rhythm control was associated with markedly worse outcomes than rate control (32.0% vs. 10.5%; absolute difference −21.5 percentage points; 90% CI −30.8 to −12.1). Among patients with hypertension, rhythm control was also associated with worse outcomes (30.8% vs. 17.3%; absolute difference −13.5 percentage points; 90% CI −22.2 to −4.9). These were not prespecified analyses, the subgroups are small, and the appropriate framing is hypothesis-generating. But the directional consistency with subsequent literature suggesting that women have higher rates of antiarrhythmic-related proarrhythmia (particularly torsade with QT-prolonging drugs) makes the female-sex finding worth noting.

Conclusions: In patients with recurrent persistent atrial fibrillation who had already required at least one prior cardioversion, rate control was noninferior to rhythm control for the prevention of major cardiovascular events. The point estimate favored rate control by roughly 5 absolute percentage points over 2.3 years, with the upper bound of the confidence interval at +0.4% — barely consistent with the null and well within the prespecified noninferiority margin. The results of this trial, together with those of AFFIRM trial, established the empirical basis for the view that pharmacologic rhythm control offers no clear advantage over rate control and is associated with a higher rate of adverse events.

Several aspects of the trial deserve careful consideration. First, the rhythm-control disadvantage was driven by treatment-related harms (antiarrhythmic drug adverse effects, pacemaker implantation) rather than by failure to prevent stroke or death. This is an important interpretive point: the trial does not show that sinus rhythm itself is prognostically neutral — it shows that the strategy of pursuing sinus rhythm with antiarrhythmic drugs produces drug toxicity that offsets any potential benefit.

Second, the noninferiority framing is technically appropriate but worth keeping in view. The trial does not establish that the two strategies produce identical outcomes; it establishes that lenient rate control falls within 10 percentage points of rhythm control, and the point estimate happens to favor rate control. The 90% confidence interval extends from −11.0 to +0.4, meaning the data are consistent with rate control being up to 11 percentage points better and essentially exclude rate control being clinically meaningfully worse. This is a stronger result than the typical noninferiority trial, where the point estimate often sits closer to the null.

Third, the stroke distribution is the most clinically actionable finding. The rhythm-control arm had numerically more thromboembolic events, and the excess was concentrated in patients in whom anticoagulation had been discontinued because sinus rhythm appeared restored — patients who then had strokes while still in sinus rhythm. This pattern, replicated in AFFIRM, established the principle that anticoagulation decisions should be driven by stroke risk factors rather than by apparent rhythm status. The 2.5-3.5 INR target used in RACE was actually higher than the modern 2.0-3.0 target, and most strokes still occurred at INR <2.0 while most bleeds occurred at INR >3.0 — illustrating how narrow the therapeutic window is and how much harm can come from intermittently subtherapeutic or supratherapeutic anticoagulation. Of note, vitamin K antagonists are now used far less frequently for atrial fibrillation, having largely been supplanted by non–vitamin K oral anticoagulants (NOACs).

Fourth, the enriched study population matters for generalization. RACE enrolled patients who had already failed at least one cardioversion. The authors are appropriately cautious in noting that the conclusion does not necessarily extend to patients seen for the first time with atrial fibrillation, and that rhythm control may still be indicated for patients with serious symptoms. The trial is best read as establishing that persistence with rhythm control after documented failure is not warranted, rather than as a blanket rejection of rhythm control at first presentation.

A framing that the authors do not explicitly offer but we think is instructive: RACE and AFFIRM together demonstrated that the 1990s consensus around aggressive rhythm control was wrong as applied to most patients. In neither trial did maintenance of sinus rhythm translate into prognostic benefit on hard endpoints; in both trials, the drugs used to maintain sinus rhythm produced detectable harm; and in both trials, the most clinically important finding was that anticoagulation must be continued regardless of apparent rhythm. The clinical implication that has held up best is not “rate control is better” — it is “rhythm is not the right thing to chase, anticoagulation is the right thing to maintain, and symptoms are the right reason to attempt rhythm control.” RACE was a smaller trial than AFFIRM (522 vs. 4,060 patients) and produced numerically more favorable results for rate control, but the convergent finding from both trials is what gave the conclusion its durability.