The best magazine
Methods
The ORBIT-AF study is a contemporary registry of outpatients in the United States with AF managed by a variety of providers, including internists, cardiologists, and electrophysiologists. A nationally representative sample of sites was invited to participate, with diversity across practice-type and geography. An adaptive design was used to ensure provider and geographic heterogeneity. However, enrollment was not formally stratified. Site selection and management was performed by the Duke Clinical Research Institute. Site investigators enrolled consecutive patients with AF meeting inclusion and exclusion criteria. Eligible patients included those 18 years of age or older, with electrocardiographic evidence of AF, providing informed consent, and able to follow-up. Patients with life expectancy of less than 6 months or AF secondary to reversible conditions were excluded. The medical record served as the primary source of data, which was entered into a web-based case report form. Data collection focused on demographics, past medical history, type of AF and prior interventions, ongoing antithrombotic therapy (with monitoring), vital signs, laboratory studies, electrocardiographic findings, and echocardiographic findings. Prior and incident electrophysiology interventions are also captured, including both catheter-based and surgical ablations for AF and atrial flutter. It is important to note that the inclusion criteria mandated a diagnosis of atrial fibrillation. Patients with atrial flutter only were not eligible for ORBIT AF. Details about the ORBIT-AF registry have been described previously.
The ORBIT-AF case report form specifically asked each treating physician to state the management strategy for each patient, as indicated by a mutually-exclusive check box (rate control vs. rhythm control). For the purpose of this analysis, patients were stratified by strategy (rate control or rhythm control), regardless of the type of AF (new onset, paroxysmal, persistent, or longstanding persistent AF). Baseline characteristics were compared between the two groups, including demographics, medical history, procedures, medical therapies, vital signs, and laboratory studies. Contraindications to anticoagulation were also collected. Risk scores for stroke (eg, CHADS2) were calculated from baseline clinical data. The data are presented as frequencies and percentages for categorical variables and medians (interquartile range) for continuous variables (except where appropriate). The chi-square test for categorical variables and the Wilcoxon rank sum test for continuous variables were used for univariate comparisons.
In order to determine factors associated with rhythm control (versus rate control), a multivariable logistic regression model was constructed for the binary outcome of AF management strategy (with rate control as the reference group). Candidate variables included demographics, medical history, echocardiographic assessment, physician-assessed stroke and bleeding risks, vital signs, laboratory studies, functional status, provider care specialty, and enrolling site region, but not current therapies or symptoms. Identification of provider specialty was not mutually exclusive—patients managed by a primary care provider and multiple specialists were identified as having multiple providers.
Missing data was multiply-imputed and final estimates and standard errors reflect the combined analysis over five imputed datasets (all the candidate variables were missing < 5% except for the following: electrocardiographic evidence of left ventricular hypertrophy [7%], serum creatinine [7%], hemoglobin [10%], hematocrit [11%], left ventricular ejection fraction [20%], and left atrial diameter [26%], and posterior wall thickness [38%]). Model selection using backward selection with a stay criteria of 0.05 was used to obtain a set of factors in which each factor was independently associated with AF management strategy. The model was fit using logistic generalized estimating equations method with exchangeable working correlation matrix to account for within-site clustering because patients at the same site are more likely to have similar responses relative to patients at other sites (ie, within-center correlation for responses). The resulting model was subsequently used to adjust for confounders, and identify factors associated with treatment strategy selection.
The above-described multivariable model was also used to derive adjusted rates of rate-control therapies, including β-blockers, calcium-channel blockers, and digoxin.
All statistical analyses of the aggregate, de-identified data were performed by the Duke Clinical Research Institute using SAS software (version 9.2 and 9.3, SAS Institute, Cary, NC). All P values were 2 sided. The ORBIT-AF Registry is approved by the Duke Institutional Review Board, and all participating sites obtained institutional review board approval pursuant to local requirements. All subjects provided written, informed consent. The authors had access to the primary data, and take full responsibility for the validity.
Source: ...