Methods

The study cohort was obtained from the Thomas Reuters MarketScan Commercial Claims and Encounters Database. This database consists of inpatient, outpatient, and prescription claims data from US employers who provide health plans for their employees and employees’ spouses and dependents. The MarketScan database has primarily been used for health care utilization and outcome studies of a variety of diseases, including AF. For purposes of this analysis, we obtained data on all patients with an inpatient or outpatient encounter that included a diagnosis of AF ( International Classification of Diseases, Ninth Revision, Clinical Modification [ ICD-9-CM ], code 427.31) from January 1, 2006, to December 31, 2010. The Marketscan database used for this study does not include claims data from patients ≥65 years. ICD-9-CM and current procedural terminology codes for diagnosis and procedures of interest were obtained from previously published studies. This study was reviewed by the Duke University Health System Institutional Review Board and determined to be exempt from the review of the IRB.

We only included patients with available individual-level and pharmacy benefit data. We identified the first inpatient or outpatient encounter with a diagnosis of AF and considered this the index AF encounter. We excluded patients without at least 6 months of continuous health plan enrollment before the “index AF encounter.” Using National Drug Codes, we selected patients who filled a prescription for a ≥30-day supply for oral formulations of Vaughan Williams class Ic (propafenone and flecainide) and class III (amiodarone, dofetilide, sotalol, and dronedarone) AADs within 14 days after the end of the index AF encounter; the date of this initial prescription claim was considered the “index AAD prescription date.” Patients were then excluded from the final study cohort if they were <18 years, had a prescription claim for any AAD in the 6 months before the index AAD prescription, or had any of the following diagnoses in the 6 months before the index AAD prescription: CAD ( ICD-9-CM codes 410 to 414, 429.2, V45.81), heart failure ( ICD-9-CM codes 428.xx 402.01, 402.11, 402.91, 404.01, 404.03, 404.11, 404.13, 404.91, 404.93, and 398.91), cardiomyopathy ( ICD-9-CM codes 425.0, 425.1, 425.2, 425.3, 425.5, 425.7, 425.8, 425.9), ventricular arrhythmia ( ICD-9-CM codes 427.1, 427.4x, 427.5), or heart transplant or left ventricular assist device ( ICD-9-CM codes 37.5x, 33.6, 37.6x, V42.1).

The study cohort was divided into treatment groups based on the index prescription: class Ic drugs (propafenone or flecainide), amiodarone, dofetilide, sotalol, and dronedarone. Propafenone and flecainide were considered to be very similar in pharmacologic activity and indication; therefore, these 2 drugs were combined into 1 treatment group. The class III drugs were thought to be pharmacologically distinct; therefore, they were considered separate treatment groups. Because very few patients were prescribed dofetilide during the study period, the dofetilide treatment group was excluded from the study, leaving 4 main treatment groups: (1) class Ic, (2) amiodarone, (3) sotalol, and (4) dronedarone.

Dronedarone was first marketed in the United States in July 2009. Being a new AAD, we recognized that the availability of dronedarone could differentially affect AAD drug selection before and after its marketing; consequently, the analyses of factors associated with class Ic drugs versus each class III drug were done in 2 distinct time periods: (1) before July 1, 2009, and (2) after June 30, 2009. Likewise, a separate multinomial logistic regression model was created for each time period. In the first model (index AAD prescriptions before July 1, 2009), we determined factors associated with amiodarone versus class Ic drugs and sotalol versus class Ic drugs. In the second model (index AAD prescriptions after June 30, 2009), we determined factors associated with amiodarone versus class Ic, sotalol versus class Ic, and dronedarone versus class Ic. In each model, we used a backward selection process of variables.

Candidate variables for each model were the same and included age; gender; inpatient versus outpatient index AF encounter; geographic region; AF as primary diagnosis versus secondary diagnosis; hospitalization in the 6 months before index prescription; year of index prescription; electrical cardioversion during index AF encounter; cardiac ablation during index AF encounter; number of days from index AF encounter to index prescription, rate-controlling drug use (ß blocker, calcium channel blocker, or digoxin), QT-prolonging medication use, or anticoagulant use concomitant with or within 6 months before the index prescription; and during the 6 months before the index prescription, a history of atrial flutter, bradyarrhythmias, other atrial arrhythmias, diabetes, hypertension, chronic or acute rheumatic heart disease, pacemaker, renal impairment, liver disease, thyroid disease, pulmonary disease, cancer, stroke, cerebral hemorrhage, depression, obesity, nonrheumatic valvular disease, bleeding, or cardiothoracic surgery. Continuous variables are presented as medians and 25th and 75th percentiles, and categorical variables are presented as percentages.

To evaluate longitudinal use following the index AAD prescription, we first assessed changes in AAD drug use in the 12 months after the index prescription in the overall study cohort. We determined initiation of another AAD treatment by the presence of a prescription claim for a ≥30-day supply of an AAD in a different group from the index AAD prescription (class Ic, amiodarone, sotalol, and dronedarone) during the 12 months after the index AAD prescription date. We used the Kaplan-Meier method to determine rates and 95% confidence intervals for rates of AAD change for each of the 4 treatment groups. In patients without an AAD change, we explored discontinuation of the index AAD. Using the day’s supply from each prescription claim of the index AAD, we identified gaps in prescription fills for the index AAD. A gap of ≥90 days during the 12 months after the index AAD prescription was considered a discontinuation of the index AAD. A prescription claim with a dispensed day supply of <0 day on any subsequent AAD prescription claim was considered a claim error, and the patient was excluded from our analysis.