Methods
Study Population
AF-CHF and AFFIRM studies share many similar methodological features as open-labelled randomized multicenter trials that compared rhythm and rate control strategies in patients with nonpermanent AF. Similarities in design and data reporting rendered these 2 trials particularly suitable for merging patient-level records. Detailed inclusion criteria have been previously described. A summary of key inclusion and exclusion criteria is presented in Table 1 of the online only data supplement http://onlinelibrary.wiley.com/store/10.1111/jce.12535/asset/supinfo/jce12535-sup-0001-supinfo.docx?v=1&s=8d071af8eb09733f39c5cc232f16bcc7b4f755e9. In brief, AFFIRM included a total of 4,060 patients ≥65 years of age or with additional risk factors for stroke. Only 23% had a history of heart failure. In contrast, the AF-CHF trial enrolled 1,376 patients, all of whom had a left ventricular ejection fraction (LVEF) ≤35% and either New York Heart Association (NYHA) class II-IV symptoms, hospitalization for CHF in the preceding 6 months, or an LVEF <25%. The current analysis, based on individual patient-level data, includes all subjects from AFFIRM and AF-CHF trials who were randomized to rate control and those randomized to rhythm control who received amiodarone as the initial antiarrhythmic agent.
Rhythm- and Rate-control Strategies
In AFFIRM, rhythm control was achieved with a variety of drugs at the treating physician's discretion. Amiodarone, at a maintenance dose of 200 or 300 mg per day, was administered as first-line therapy to 612 of 2,033 (30.1%) patients randomized to rhythm control. Patients randomized to rhythm control in AF-CHF systematically received amiodarone as the initial antiarrhythmic agent, with a maintenance dose of 200 mg per day. In both studies, electrical cardioversion was recommended for patients assigned to rhythm control not in sinus rhythm at enrolment. Strict parameters were recommended for patients randomized to rate control (i.e., maximal heart rate of 80 bpm at rest and <110 bpm on exertion). Follow-up visits were initially every 4 months in both studies and continued on an "as needed" basis after 5 years of follow-up in AFFIRM, and every 6 months after 4 years of follow-up in AF-CHF.
Outcomes
The primary outcome of the current combined study was time to first recurrence of AF. Patients were stratified according to their underlying LVEF, classified as: >50% (normal), 30–49% (mild to moderate dysfunction), and <30% (severe dysfunction). Factoring in an initial stabilization period that allowed for cardioversion, the 4-month follow-up visit was defined as time zero. Patients not in sinus rhythm at time zero were not considered for the primary outcome.
Overall AF burden, as defined by the proportion of time spent in AF during the course of the study, was considered a secondary outcome. It was quantified using previously described AFFIRM methodology. In short, each patient's heart rhythm was determined at baseline and at every follow-up visit. If the same rhythm was recorded at 2 consecutive visits, the patient was considered to have maintained that rhythm during the intercurrent interval. If different rhythms were recorded at consecutive follow-up visits, half the time interval was assigned to each rhythm. Additional outcomes included all-cause death, cardiovascular death, time to first hospitalization, and time to first cardiovascular hospitalization. For these outcomes, time zero was defined as time at randomization.
AFFIRM and AF-CHF study protocols were approved by each participating center's institutional review board. All participants provided written informed consent. In addition, local institutional review board approval was obtained to conduct this analysis on pooled patient-level data. The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written.
Statistical Analysis
Continuous variables are summarized as mean ± standard error. Categorical variables are represented by frequencies and percentages. All analyses were performed according to the intention-to-treat principle. Baseline comparisons stratified by treatment group and LVEF were performed using independent Student's t tests, one-way analysis of variance, or chi-square tests, where appropriate.
Freedom from recurrent AF and secondary time-to-event outcomes (i.e., all-cause hospitalization, cardiovascular hospitalization, all-cause death, and cardiovascular death) were plotted using the Kaplan–Meier method, with comparisons by log rank tests. Censoring occurred at the last contact visit with a study investigator or upon withdrawal from the study. Factors associated with each outcome were assessed in univariate and multivariate Cox regression models, after verifying proportionality assumptions. These models considered the 22 baseline covariates listed in Table 1. Variables associated with a P value <0.2 in univariate analyses were included in an automated backward selection multivariate Cox regression model. Factors associated with AF burden were assessed in linear regression models that considered the same baseline covariates. Variables significant at the 0.2 level in univariate analyses were included in an automated backward selection multivariate linear regression model. In addition, a term for study (i.e., AFFIRM vs. AF-CHF) and a first-order interaction term evaluating study-specific effects (amiodarone*study) were forced in each multivariate regression models. Statistical significance was defined as a 2-tailed value P value <0.05. All analyses were performed using SAS software version 9.2 (SAS Institute, Cary, NC, USA).