K. Jack Ishak, PhD

Atrial fibrillation is the most common arrhythmia requiring hospitalization.

 Approximately 350,000 hospital admissions are attributable to it within the United States annually, and the costs for inpatient care account for the largest proportion (44%) of the total direct medical cost ($6.65 billion per year) for treatment of atrial fibrillation.

 Atrial fibrillation also has been reported to be associated with excess mortality and significant adverse impact on quality of life.

Long-term management of atrial fibrillation involves controlling the ventricular rate or attempting a return to sinus rhythm and then maintaining it. Both treatment strategies may be accompanied by therapy for prevention of thromboembolism.

 Although the strategy of maintaining sinus rhythm has not yet shown any survival advantage over controlling ventricular rate in the management of patients with atrial fibrillation,

 it is still frequently recommended for those patients with disabling symptoms.

 Amiodarone and sotalol are the most effective pharmacologic treatments for maintaining sinus rhythm

 and have been increasingly used in clinical practice.

Persistence with antiarrhythmic therapy is expected to decline over time due to adverse events associated with these drugs. Several clinical studies have shown that patients on amiodarone may stop the treatment because of adverse events such as pulmonary toxicity, gastrointestinal symptoms, and thyroid disorders.

 Patients receiving sotalol may discontinue the drug because of cardiac events such as torsade de pointes and bradycardia.

 Lack of efficacy is another potential cause of discontinuation. Although the drugs have been proven more effective than placebo in preventing recurrence, about 35% of the patients on amiodarone and 60% on sotalol may experience recurrent atrial fibrillation within 1 year.

Many studies have been conducted to examine the efficacy and safety of these drugs.

 However, little is known about persistence with rhythm-control medications over the longer term in actual practice and whether persistence has any impact on health outcomes. This retrospective database analysis examined the rates of persistence among patients starting on amiodarone or sotalol and quantified the effect of persistence on clinical outcomes.

The data used for this study were assembled from the PHARMetrics database.

 Patients with a diagnosis of atrial fibrillation (

International Classification of Diseases, Ninth Revision

] code 427.31) recorded between January 1, 1999, and December 31, 2005, were selected. Because reporting of claims may be incomplete in the last few months preceding the data cut, only claims reported before June 30, 2005, were used for this study.

The study cohort was selected from patients covered by Medicare Risk, Medicaid, or commercial insurance for at least 1 year and collecting a first (index) prescription for amiodarone or sotalol any time following 1 year of continuous insurance coverage. Patients were included if they met the following criteria: age 40 years or older at the time of the index prescription; established monotherapy with the index prescription, defined as collecting at least 1 additional prescription within 1 refill period plus 1 month (or 60 days if the former was shorter) following the end of the first index prescription for the same drug without concurrent use of other antiarrhythmic agents; naïve to amiodarone or sotalol, defined as no prescription for these drugs in the year preceding the index prescription; and no coverage gap between the index and second prescription.

All patients included in the analyses were followed from the index prescription until the earliest of the following events: the end of the administrative data (ie, June 30, 2005); any change (ie, switch, loss, or gap) in insurance coverage; addition of or switch to other antiarrhythmic agents; or restarting treatment after having become nonpersistent with the index treatment. It is important to note that the PHARMetrics database does not contain date of death, but the end of follow-up for these patients was noted by the end of insurance coverage.

All pharmacy claims for the index drug collected after the date of the index prescription were identified by using National Drug Codes.20 Successive prescriptions for each patient were examined to assess whether and when a patient became nonpersistent, which was defined as permanent discontinuation of the index prescription, or a gap between prescriptions lasting the equivalent of 1 refill period plus 1 month, or 60 days (if the former was shorter). The minimum gap of 60 days was used to avoid classifying patients as not persisting following a short prescription. The last date of the prescription prior to the gap was considered the time of stopping. If an index prescription was collected after a patient became nonpersistent, it was considered a restart in treatment and follow-up time was censored at the time of restarting therapy.

Patients may collect multiple prescriptions at the same time or refill a new prescription before the end of the previous prescription. To account for these situations, the following assumptions were made. First, if a patient collected multiple prescriptions of the index drug at the same dose on the same day (eg, in case of travel), it was assumed that the prescriptions would be taken in sequence; the duration was determined by summing the durations of each collected prescription. Second, if the collected prescriptions were for different doses, it was assumed that the pills were to be taken together to achieve a higher total dose. The duration of the prescriptions was assumed to be that of the shortest prescription. Third, when a prescription was collected before the end of the current prescription, the new prescription was assumed to start at the end of the current prescription. Finally, when a prescription was collected after the end of the current prescription, the new prescription was assumed to start on the day it was collected.

Patients stopping treatment may have switched to another antiarrhythmic drug. To determine whether a switch occurred, the gap was scanned for prescriptions of a drug other than the index treatment (ie, amiodarone or sotalol). The 60-day period preceding the end of the current prescription also was considered in case a switch had occurred during that time interval. Those switches may be missed if subsequent prescriptions for the new drug are collected prior to the end of the last index prescription.

The analyses examined the benefit of persistence with treatment on atrial fibrillation—related hospitalizations and cardioversions. These clinical outcomes were chosen because they are the medical interventions commonly received by patients with atrial fibrillation. The occurrence and timing of these clinical events were recorded by processing the management and facility claim data. Relevant hospitalizations were identified by an ICD-9 code of 427.31 in the primary or secondary diagnoses; cardioversions, defined as elective electrical conversions of arrhythmia, were recorded if the 

 codes 92960 and 92961 were found during a relevant hospitalization.

The study cohort was described at baseline in terms of age, sex, and comorbidities (eg, myocardial infarction, stroke, diabetes). As some patients may have had multiple years of data, depending on the timing of the index prescription, only data obtained during the 365 days preceding the index prescription were considered. Continuous factors were summarized in terms of means and standard deviations; categorical factors, in terms of counts and proportions of patients. Comparisons were made between patients establishing treatment versus those not establishing treatment (ie, receiving only 1 prescription) and between patients receiving amiodarone versus those receiving sotalol.

Stopping rates for patients were calculated by using the Kaplan-Meier approach and compared between index treatments by using the log-rank test.

 The switching patterns of patients stopping treatment were examined. 

To assess the impact of persistence on adverse outcomes, patients’ follow-up time was divided into persistent and off-treatment periods. Because follow-up starts at the time of the first prescription, all patients are initially persistent. Patients who remained on treatment had persistent time only. For those who stopped treatment, the time between the index prescription and stopping was counted as persistent time, whereas the time from stopping to end of follow-up was counted as nonpersistent time. The first hospitalization or cardioversion during the follow-up window was identified for each patient, and the timing of the event was calculated relative to the index prescription. Time to event for patients who were neither hospitalized nor had cardioversion was censored at the end of follow-up. The effect of persistence with treatment on these outcomes was examined with a proportional hazards model in which persistence, categorized as a bivariate variable, was included as a time-dependent covariate; baseline patient characteristics and comorbidities also were included as potential confounders. These were retained in the model if their effect estimates were significant at the 5% level or if they altered the estimate of persistence effect. The underlying assumption of a proportional hazard model also was tested by including an interaction term of time and persistence in the model. 

At some time between January 1999 and June 2005, 557,786 patients were diagnosed with atrial fibrillation. Only 6% of them collected 1 or more prescriptions for the index drug and had at least 1 year of insurance coverage preceding the index prescription date. Of these patients, about 24% (n = 8193) met all the inclusion criteria for this study. 

The baseline characteristics of the patients are summarized in 

. Those establishing treatment (study cohort) and those starting but not establishing treatment (ie, those who had only 1 prescription) were similar in terms of age and sex. Patients who did not establish treatment, however, were more likely to have comorbidities compared with the study cohort. Among the study cohort, those starting on amiodarone were more likely to be male and to have comorbidities (Table 1).

The study cohort was followed for a mean 1.26 years (SD = 0.96). Mean follow-up was 1.03 years (SD = 0.82) for those who remained on the index prescription during the follow-up period and 1.5 years (SD = 1.03) for those who became nonpersistent. Patients on amiodarone were followed for an average of 1.28 years (SD = 0.97) versus 1.21 years (SD = 0.92) for patients on sotalol.

Fifty-two percent (n = 4273) of the study cohort discontinued treatment during the study period. The median time to discontinuation was 345 days. 

 shows the distribution of discontinuation time over the first 3 years of the study. Patients on sotalol stayed on treatment substantially longer than those on amiodarone (median time to nonpersistence: 570 vs 281 days; P <.001). The persistence rate was 44% at 1 year and declined to 20% after 3 years of treatment with amiodarone; 60% of patients taking sotalol remained on treatment at 1 year, but only 30% remained on treatment at 3 years.

Switching Patterns After Stopping Treatment

Of those patients who did not establish treatment (ie, collecting only 1 prescription), the majority (76%) never filled another prescription for antiarrhythmic agents after the index prescription. Only 8% added or switched to another treatment, and 16% restarted on the same or new drug after the defined duration for persistence. Treatment patterns after stopping varied based on the treatment received beforehand. Patients with an index prescription for amiodarone were more likely to stop antiarrhythmic therapy completely (81% vs 58% for sotalol; P <.001), whereas those starting on sotalol were more likely to switch to another antiarrhythmic drug (28% vs 3% for amiodarone; P <.001).

Of those patients who did establish treatment, nonpersistence was observed in 52% (n = 4273). A large proportion (62%) of these patients never collected another prescription for antiarrhythmic drugs; only 10% added or switched to another treatment, and about 28% restarted on the same or another antiarrhythmic drug at a later time. Patients on amiodarone were more likely to stop antiarrhythmic therapy completely (66% vs 47%; P <.001) or restart the same treatment (27% vs 18%; P <.001), whereas patients on sotalol were more likely to add or switch to another treatment (30% vs 5%; P <.001).

Effect of Persistence on Hospitalizations and Cardioversions

There were 1652 (20%) patients with at least 1 relevant hospitalization and 202 (2.5%) with at least 1 cardioversion during the study period. 

 summarizes the results of the proportional hazards model relating persistence to the risk of hospitalization and cardioversion. The results of the univariate model indicate that persistence with treatment was associated with a slight but statistically nonsignificant decrease in risk of hospitalizations (crude hazard ratio [HR] = 0.92, 95% confidence interval [CI] = 0.78, 1.09); a strong protective association was observed for cardioversions (crude HR = 0.44, 95% CI = 0.29, 0.67). Adjusting for potential confounders led to stronger associations; in the case of hospitalizations, the estimate was borderline statistically significant (adjusted HR = 0.86, 95% CI = 0.79, 1.02), whereas the adjusted HR for cardioversion was 0.37 (95% CI = 0.25, 0.56). The test for nonproportionality of the effect of persistence was not statistically significant, suggesting that the measured association does not change noticeably over time. Amiodarone also was associated with lower risk of cardioversions by 22% (adjusted HR = 0.78, 95% CI = 0.57, 1.06) and hospitalizations by 10% (adjusted HR = 0.90, 95% CI = 0.80, 1.01) compared with sotalol, but these associations fell just short of statistical significance.

This study indicates that only a small portion of patients diagnosed with atrial fibrillation were managed with antiarrhythmic agents. This finding was expected as rhythm control is mainly recommended for patients with disabling symptoms.

 Patients who did not establish the initial treatment (collected only 1 prescription) tended to have more comorbidities than those who did establish the initial treatment. That may be because these drugs were used only short term to prevent atrial fibrillation after a heart attack or cardiac surgery.

 In these cases, patients may have obtained only 1 prescription and thus did not meet the criterion of at least 2 continuous prescriptions for establishing the initial treatment. For patients establishing treatment, amiodarone was more likely to be used than sotalol. Selection of an appropriate antiarrhythmic agent for a patient, however, is highly dependent on safety, individualized to whatever underlying heart problems may be present. In patients with lone atrial fibrillation, sotalol and class IC antiarrhythmic agents often are recommended as first-line treatments, followed by amiodarone as an alternative.

 In patients with only coronary artery disease, sotalol is the preferred initial agent as it is associated with less long-term toxicity than amiodarone. Amiodarone is frequently suggested as initial treatment for patients with heart failure or left ventricular hypertrophy caused by hypertension. In this study, patients starting on amiodarone tended to have more comorbidities than those starting on sotalol. This prescribing pattern seems to be consistent with the current guidelines.

This study shows that persistence was generally poor, especially with amiodarone. Lack of efficacy and drugrelated side effects are likely to be important reasons for the overall poor persistence. Drug-related side effects, however, may be the key contributing factor to a higher stopping rate with amiodarone, as it has been reported to have a greater risk of causing adverse side effects than sotalol.

 Unfortunately, this study was not able to properly examine the effect of drug-related side effects on persistence because the link between use of the drugs and side effects could not be definitively confirmed based on the available data. Future research efforts to investigate this relationship may have to rely on data from systematic chart review to confirm such a linkage. Another factor potentially contributing to the higher stopping rate with amiodarone may have been individual patients’ health conditions, as patients taking amiodarone tended to be sicker and, in turn, were likely to die earlier than those taking sotalol. The data for this study, nevertheless, do not provide specific death information to validate this hypothesis. However, more patients on amiodarone (50%) were censored because of the end of insurance coverage than those on sotalol (45%), indirectly supporting this postulation.

A large proportion of patients did not collect any prescription after they became nonpersistent with the initial therapy. Of those who did (about one-third of them), patients taking sotalol were more likely to switch to or add another antiarrhythmic drug, whereas patients on amiodarone tended to restart on the same drug. Because sotalol often is recommended as first-line treatment according to the existing practice guidelines,

 it is not surprising to see more patients on this drug switching to or adding another treatment. Nevertheless, amiodarone often is used as the only or final pharmacologic option for rhythm-control strategy, which may be the main reason that a higher percentage of patients on amiodarone restart the same treatment after initial discontinuation. These findings seem to suggest that there may be a need for other, better pharmacologic alternatives, especially for nonpersisters with amiodarone, because about one-third of them restarted rhythm-control therapy at a later time.

Findings from this study were consistent with existing evidence that amiodarone is more effective than sotalol in preventing clinical events related to atrial fibrillation.

 Patients receiving amiodarone in our study were less likely than those receiving sotalol to experience cardioversion, although the reduction in hospitalizations, while clinically relevant, was not statistically significant. 

The public health burden of atrial fibrillation is substantial and expected to rise over the next few decades.

 Hospitalization is one of the main reasons for the associated high cost.

 A recent study further suggests that recurrences also are one of the major cost drivers, increasing annual healthcare cost by about $1600 per recurrence.

 In addition, atrial fibrillation is associated with impaired quality of life.

 The study findings imply that keeping patients on treatment may maintain their quality of life and could potentially reduce the consumption of healthcare resources associated with atrial fibrillation.

This study has 3 key limitations. First, follow-up for the majority of patients was censored before the end of the administrative study period, resulting in an average follow-up of 1.26 years. Longer follow-up may have provided more power for analyses of rarer events such as recurrence. Limited power also may have contributed to the nonsignificant test result for nonproportionality of the effect of persistence. Second, treatment may have been stopped by physicians for certain reasons, such as shortterm use of antiarrhythmic medications for prevention of atrial fibrillation after undergoing major cardiac surgery. This could not be determined from the data, however, and thus could not be incorporated in the analyses. Finally, the time to hospitalization or cardioversion due to recurrent atrial fibrillation is very likely to be based on the patients’ ability to determine their arrhythmia status between clinical evaluations. The estimated effects of persistence on these end points from this study, however, could not be controlled for patients’ arrhythmia status (ie, symptomatic or asymptomatic) in the analyses because of lack of data. Despite that, the effects shown in this study should not be attenuated if the analyses had controlled for this covariate because the majority of the recurrent atrial fibrillation tends to be asymptomatic as indicated by literature.

Persistence with amiodarone or sotalol was generally poor. Poor persistence was associated with an increased rate of hospitalizations and cardioversions, which could be potentially associated with greater utilization of healthcare resources. A new treatment with an improved persistence profile may help to reduce or slow the rising healthcare cost associated with atrial fibrillation.

Articles

Retrospective analysis of claims data showed that persistence with amiodarone and sotalol was generally poor.