Advertisement

Amiodarone versus dronedarone for atrial fibrillation – a retrospective cohort study

  • Author Footnotes
    1 Funding disclosure: JMB is a research scholar supported by Les Fonds de Recherche Québec Santé
    James M. Brophy
    Correspondence
    Corresponding author: (James M Brophy MD PhD)
    Footnotes
    1 Funding disclosure: JMB is a research scholar supported by Les Fonds de Recherche Québec Santé
    Affiliations
    McGill University Health Center Centre for Health Outcomes Research (CORE) 5252 Boul. de Maisonneuve West Room 2B.37 Montreal QC H4A 3S5
    Search for articles by this author
  • Lyne Nadeau
    Affiliations
    McGill University Health Center Centre for Health Outcomes Research (CORE) 5252 Boul. de Maisonneuve West Room 2B.37 Montreal QC H4A 3S5
    Search for articles by this author
  • Author Footnotes
    1 Funding disclosure: JMB is a research scholar supported by Les Fonds de Recherche Québec Santé
Open AccessPublished:October 10, 2022DOI:https://doi.org/10.1016/j.cjco.2022.09.008

      Abstract

      Background

      Atrial fibrillation is one of the most common arrhythmias but the optimal drug choice for a rhythm control strategy remains uncertain.

      Methods

      This was a retrospective cohort claims database study using the Truven Health Market Scan Commercial Claims and Encounters and Medicare Supplemental databases. Patients with a new diagnosis of atrial fibrillation and a discharge date between 2011 and 2015 were included. The exposure variables of interest were a discharge prescription for amiodarone or dronedarone. The average treatment effect for the composite of total mortality or a repeat cardiovascular (CV) hospitalization was the primary outcome. Sensitivity analyses with other treatment effect metrics were performed. Baseline covariate imbalance between the groups were adjusted using propensity score methods with inverse probability weighting.

      Results

      There were 1735 patients discharged on amiodarone and 338 discharged on dronedarone with a median follow-up time of 357 days. There were 43 (12.7%) CV hospitalizations in the dronedarone group and 146 (8.4%) in the amiodarone group (Risk difference (RD) 4.3%, 95% CI 0.4% - 8.3%, p =0.02). There were 4 (1.2%) deaths in the dronedarone group and 31 (1.8%) deaths with amiodarone (RD -0.6%, 95% CI -2.1% - 0.9%, p =0.6). After adjusting for baseline covariates, the dronedarone hazard ratio for the composite endpoint was 1.47 (95%CI 1.01 - 2.12). This result was generally robust to sensitivity analyses.

      Conclusion

      In this incident cohort of patients hospitalized for atrial fibrillation, compared to those discharged on amiodarone, patients receiving a dronedarone discharge prescription had an increase in the composite endpoint of recurrent CV hospitalizations and death over a median 1 year follow-up.

      Keywords

      2. Introduction

      Atrial fibrillation is one of the most common arrhythmias and a progressive worldwide increase in its incidence, prevalence, and burden of disease has been observed in the 21st century [
      • Chugh S.S.
      • Havmoeller R.
      • Narayanan K.
      • Singh D.
      • Rienstra M.
      • Benjamin E.J.
      • Gillum R.F.
      • Kim Y.H.
      • McAnulty Jr., J.H.
      • Zheng Z.J.
      • Forouzanfar M.H.
      • Naghavi M.
      • Mensah G.A.
      • Ezzati M.
      • Murray C.J.
      Worldwide epidemiology of atrial fibrillation: a global burden of disease 2010 study.
      ]. The rising incidence is likely multifactorial due to increasing longevity with an associated increase in comorbidities as well as improved detection strategies including the use of wearables. Atrial fibrillation may exist as paroxysmal, persistent or permanent forms. Although some atrial fibrillation patients are asymptomatic, many do present to the hospital with symptoms of palpitations, dyspnea or reduced exercise tolerance. Atrial fibrillation complications may include systemic or cerebral thromboembolism and heart failure as well as an increase in overall mortality [
      • Andersson T.
      • Magnuson A.
      • Bryngelsson I.L.
      • Frobert O.
      • Henriksson K.M.
      • Edvardsson N.
      • Poci D.
      Allcause mortality in 272,186 patients hospitalized with incident atrial fibrillation 1995-2008: a Swedish nationwide long-term case-control study.
      ]. There are numerous therapeutic options for the treatment of atrial fibrillation including ablation techniques, but the majority of patients still receive some form of pharmacological treatment either as part of a rhythm or rate control strategy. A recent large randomized study, EAST AFNET[
      • Kirchhof P.
      • Camm A.J.
      • Goette A.
      • Brandes A.
      • Eckardt L.
      • Elvan A.
      • Fetsch T.
      • van Gelder I.C.
      • Haase D.
      • Haegeli L.M.
      • Hamann F.
      • Heidbuchel H.
      • Hindricks G.
      • Kautzner J.
      • Kuck K.H.
      • Mont L.
      • Ng G.A.
      • Rekosz J.
      • Schoen N.
      • Schotten U.
      • Suling A.
      • Taggeselle J.
      • Themistoclakis S.
      • Vettorazzi E.
      • Vardas P.
      • Wegscheider K.
      • Willems S.
      • Crijns Hjgm
      • Breithardt G.
      East-Afnet Trial Investigators
      Early rhythm-control therapy in patients with atrial fibrillation.
      ], showed a rhythm-control therapy with either antiarrhythmic drugs or atrial fibrillation ablation was associated with a lower risk of adverse cardiovascular outcomes than usual care among patients with early atrial fibrillation and cardiovascular conditions. Although the choice of antiarrhythmic drug was not standardized in this study and although there has been a longstanding debate about the optimal rhythm control agent, amiodarone has been a common choice despite its many known toxic side effects.
      Dronedarone is a congener of amiodarone, with the omission of its iodine molecule, developed with the intention of reducing the likelihood of toxic side effects and consequently a safer alternative to amiodarone and was also a popular choice in EAST AFNET[
      • Kirchhof P.
      • Camm A.J.
      • Goette A.
      • Brandes A.
      • Eckardt L.
      • Elvan A.
      • Fetsch T.
      • van Gelder I.C.
      • Haase D.
      • Haegeli L.M.
      • Hamann F.
      • Heidbuchel H.
      • Hindricks G.
      • Kautzner J.
      • Kuck K.H.
      • Mont L.
      • Ng G.A.
      • Rekosz J.
      • Schoen N.
      • Schotten U.
      • Suling A.
      • Taggeselle J.
      • Themistoclakis S.
      • Vettorazzi E.
      • Vardas P.
      • Wegscheider K.
      • Willems S.
      • Crijns Hjgm
      • Breithardt G.
      East-Afnet Trial Investigators
      Early rhythm-control therapy in patients with atrial fibrillation.
      ]. The only comparative randomized clinical trial (RCT) of these two drugs showed dronedarone was better tolerated with fewer side effects than amiodarone but was less effective in maintaining sinus rhythm[
      • Le Heuzey J.Y.
      • De Ferrari G.M.
      • Radzik D.
      • Santini M.
      • Zhu J.
      • Davy J.M.
      A short-term, randomized, double-blind, parallel-group study to evaluate the efficacy and safety of dronedarone versus amiodarone in patients with persistent atrial fibrillation: the dionysos study.
      ]. However the small sample size and relatively short 6 month follow-up precluded any comparison of clinical outcomes and hospitalizations were not recorded with the primary end point being recurrent atrial fibrillation [
      • Le Heuzey J.Y.
      • De Ferrari G.M.
      • Radzik D.
      • Santini M.
      • Zhu J.
      • Davy J.M.
      A short-term, randomized, double-blind, parallel-group study to evaluate the efficacy and safety of dronedarone versus amiodarone in patients with persistent atrial fibrillation: the dionysos study.
      ]. Moreover placebo controlled RCTs have shown an increased risk of death or hospitalization with dronedarone in patients with low ejection fraction [
      • Kober L.
      • Torp-Pedersen C.
      • McMurray J.J.
      • Gotzsche O.
      • Levy S.
      • Crijns H.
      • Amlie J.
      • Carlsen J.
      Group Dronedarone Study
      Increased mortality after dronedarone therapy for severe heart failure.
      ] or with permanent atrial fibrillation [
      • Connolly S.J.
      • Camm A.J.
      • Halperin J.L.
      • Joyner C.
      • Alings M.
      • Amerena J.
      • Atar D.
      • Avezum A.
      • Blomstrom P.
      • Borggrefe M.
      • Budaj A.
      • Chen S.A.
      • Ching C.K.
      • Commerford P.
      • Dans A.
      • Davy J.M.
      • Delacretaz E.
      • Di Pasquale G.
      • Diaz R.
      • Dorian P.
      • Flaker G.
      • Golitsyn S.
      • Gonzalez-Hermosillo A.
      • Granger C.B.
      • Heidbuchel H.
      • Kautzner J.
      • Kim J.S.
      • Lanas F.
      • Lewis B.S.
      • Merino J.L.
      • Morillo C.
      • Murin J.
      • Narasimhan C.
      • Paolasso E.
      • Parkhomenko A.
      • Peters N.S.
      • Sim K.H.
      • Stiles M.K.
      • Tanomsup S.
      • Toivonen L.
      • Tomcsanyi J.
      • Torp-Pedersen C.
      • Tse H.F.
      • Vardas P.
      • Vinereanu D.
      • Xavier D.
      • Zhu J.
      • Zhu J.R.
      • Baret-Cormel L.
      • Weinling E.
      • Staiger C.
      • Yusuf S.
      • Chrolavicius S.
      • Afzal R.
      • Hohnloser S.H.
      Pallas Investigators
      Dronedarone in high-risk permanent atrial fibrillation.
      ].
      The increased tendancy to pursue a rhythm control strategy coupled with the uncertainty regarding the optimal drug choice due to a lack of randomized comparative trial data for meaningful patient outcomes with antiarrhythmic drugs were the motivation for this observational study seeking to determine the comparative drug effectiveness of amiodarone and dronedarone in a real world population of patients with an incident hospitalization for atrial fibrillation.

      3. Methods

      Design and Population

      To optimize the possibility of drawing causal inferences from this observational data, we have followed the target trial emulation approach of Hernan and colleagues [
      • Hernan M.A.
      • Robins J.M.
      Using big data to emulate a target trial when a randomized trial is not available.
      ],[
      • Hernan M.A.
      • Sauer B.C.
      • Hernandez-Diaz S.
      • Platt R.
      • Shrier I.
      Specifying a target trial prevents immortal time bias and other self-inflicted injuries in observational analyses.
      ]. This retrospective claims database study was conducted using the Truven Health Analytics MarketScan Research Databases (2011-2015) with U.S. commercial and Medicare supplemental claims. The dataset included all inpatient medical claims and outpatient medical claims, plus enrollment information, all within this 5 year period, on individuals with a hospitalization for atrial fibrillation (inpatient medical claim with primary diagnosis using International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 427.x) or ICD-10 diagnosis codes I46, I47, I48, I49, I51.8, R00.1, R00.8, R01.2. Subjects had to be continuously enrolled for at least 12 months before the index date with no prior atrial fibrillation diagnosis. The discharge date of the first observed hospital claim was the assigned index date for this incident cohort. The cohort was restricted to patients who received a prescription for isolated dronedarone or amiodarone anti-arrhythmic therapy within 2 days of discharge, thereby avoiding any potential immortal time bias[
      • Hernan M.A.
      • Sauer B.C.
      • Hernandez-Diaz S.
      • Platt R.
      • Shrier I.
      Specifying a target trial prevents immortal time bias and other self-inflicted injuries in observational analyses.
      ]. Specifically, patients receiving concomitant therapy with additional atrial fibrillation drug therapies including digoxin, sotalol, flecainide, propafenone, dofetilide, mexiletine or disopyramide were excluded. The primary outcome was a repeat cardiovascular hospitalization or within hospital death occurring following the initial hospital discharge. The project received ethical approved from the McGill University Health Center research ethics board.

      Propensity score methods

      As patients have not been randomized to amiodarone or dronedarone, propensity scores with inverse probability of treatment weights were used to balance the baseline characteristics of the two groups. Propensity scores are the probability of being assigned to a certain treatment, conditional on baseline characteristics and estimated using logistic regression. The idea is to apply a weight to each propensity score to balance the dronedarone and amiodarone groups with regards to the propensity score and therefore hopefully balance the individual covariates. Essentially this means for any given propensity score, the choice of drug is a random process, at least as far as the measured confounders are concerned. Ultimately, the propensity score will be used to estimate the treatment effect. It is important to realize that several treatment effects may be calculated and of interest. Also to prevent individual measurements from assuming an over-sized importance in the pseudo-populations, observations with weights greater than 5 have been censored.
      • 1.
        The Average Treatment Effect (ATE) is the most common measure and what is estimated in a randomized study where the target population is the whole population, both treated and controlled and was the prespecified primary analysis. However, this is not always the medically or scientifically appropriate population as it assumes that every participant can be switched from their current treatment to the opposite, which isn’t always possible. Therefore as sensitivity analyses, additional treatment effects were investigated.
      • 2.
        The Average Treatment Effect Among the Treated (ATT) estimates the treatment effect with the treated population as the target population.
      • 3.
        The Average Treatment Effect Among the Controls (ATC) estimates the treatment effect with the controlled population as the target population.
      • 4.
        The Average Treatment Effect Among the Evenly Matchable (ATM) estimates the treatment effect with a matched population as the target population. The estimated population is nearly equivalent to the cohort formed by one-to-one pair matching.
      The weights for these different effect measures are calculated as follows where the propensity score for participant i is defined as ei and the treatment assignment is Zi, where Z=1 indicates the participant received the treatment (dronedarone) and Z=0 indicates they received the control (amiodarone).ωATE=Ziei+1Zi1ei
      ωATT=eiZiei+ei(1Zi)1ei


      ωATC=(1ei)Ziei+(1ei)(1Zi)1ei


      ωATM=min{ei,1ei}Ziei+(1Zi)(1ei)


      The success of the propensity score methods was assessed by examining standardized mean difference between the groups and by graphical inspection of the propensity score histograms.

      Statistical analyses

      Multivariate logistic regression was used to model the propensity score treatment model. Cox proportional hazards regression was used to model time to the composite outcome using the datasets created by the inverse probability of treatment weighted propensity scores described above. All analyses were performed in R[

      R. Core Team. R: A language and environment for statistical computing. manual, 2021 2021. URL https://www.R-project.org/.

      ].

      4. Results

      During the study period there were 52,164 hospitalizations for atrial fibrillation and 1735 patients were discharged on amiodarone and 338 discharged on dronedarone (Table 1). None of these patients received other anti-arrhythmic drugs on discharge. Table 2 compares their baseline characteristics. In general, the amiodarone group was older (67.0 (11.3) versus 63.9 (11.4)) and had more comorbidites. The low rate of baseline anticoagulants is to be expected as this was an incident cohort of new atrial fibrillation cases. The discharge rate of oral anticoagulants was increased approximately 10 fold (Table 2).
      Table 1Discharge medications for 52,164 patients hospitalized for atrial fibrillation
      ExposureNumber
      BB12,723
      Sotalol1,072
      Amio1,735
      Dronedarone338
      CCB5,450
      Digoxin406
      Other546
      None29,894
      *Abbreviations; BB=Beta blockers, Amio = Amiodarone, CCB=Calcium channel blockers
      Table 2Demographic and Baseline Characteristics
      Amiodarone (N=1735)Dronedarone (N=338)Overall (N=2073)
      Age
      Mean (SD)67.0 (11.3)63.9 (11.4)66.5 (11.4)
      Median [Min, Max]

      Sex
      65.0 [40.0, 89.0]63.0 [40.0, 89.0]65.0 [40.0, 89.0]
       Male1069 (61.6%)191 (56.5%)1260 (60.8%)
       Female666 (38.4%)147 (43.5%)813 (39.2%)
      Previous AMI154 (8.9%)15 (4.4%)169 (8.2%)
      Congestive heart failure335 (19.3%)20 (5.9%)355 (17.1%)
      Peripheral vascular disease90 (5.2%)7 (2.1%)97 (4.7%)
      Chronic obstructive lung disease

      CHADS_VASC
      324 (18.7%)48 (14.2%)372 (17.9%)
       Mean (SD)2.22 (1.48)1.84 (1.44)2.16 (1.48)
       Median [Min, Max]2.00 [0, 7.00]2.00 [0, 6.00]2.00 [0, 7.00]
      Renal disease93 (5.4%)16 (4.7%)109 (5.3%)
      Cancer213 (12.3%)20 (5.9%)233 (11.2%)
      Diabetes261 (15.0%)46 (13.6%)307 (14.8%)
      Cerebrovascular Disease150 (8.6%)21 (6.2%)171 (8.2%)
      Amiodarone (N=1735)Dronedarone (N=338)Overall (N=2073)
      0551 (31.8%)196 (58.0%)747 (36.0%)
      1514 (29.6%)82 (24.3%)596 (28.8%)
      2311 (17.9%)30 (8.9%)341 (16.4%)
      3155 (8.9%)18 (5.3%)173 (8.3%)
      454 (3.1%)4 (1.2%)58 (2.8%)
      5+150 (8.6%)8 (2.4%)158 (7.6%)
      ACE-I334 (19.3%)67 (19.8%)401 (19.3%)
      ARB249 (14.4%)40 (11.8%)289 (13.9%)
      Statins565 (32.6%)91 (26.9%)656 (31.6%)
      OAC admission67 (3.9%)9 (2.7%)76 (3.7%)
      OAC on discharge628 (36.2%)155 (45.9%)783 (37.8%)
      Charlson Index
      AMI = acute myocardial infarction, ACE-I = angiotensin converting enzyme inhibitors, ARB = angiotensin receptor blockers, OAC = oral anticoagulants
      Table 3 shows the outcomes according to discharge drug exposure. There were more repeat CV hospitalizations in the dronedarone group (146 (8.4%) versus 43 (12.7%), risk difference (RD) = 4.3%, 95% confidence interval (CI) 0.3% - 8.3%) but with no statistically significant difference in total mortality (31 (1.8%) vs. 4 (1.2%), RD = 0.6%, 95% CI -0.9% - 2.1%). Although not prespecified in our protocol, in a post hoc analysis we examined individual CV hospitalization causes. The majority of the recurrent hospitalization had atrial fibrillation as the primary diagnosis (89 (5.13%) with amiodarone versus 39 (11.54%) with dronedarone, RR2.25, 95% CI 1.57 - 3.22). There were no statistically significant differences between the drugs for other CV hospitalizations including acute coronary syndromes (19 (1.09%) with amiodarone versus 1 (0.30%) with dronedarone, RR 0.27, 95% CI 0.04 - 2.01) and congestive heart failure (39 (2.25%) with amiodarone versus 3 (0.89%) with dronedarone, RR 0.39, 95% CI 0.12 - 1.27).
      Table 3Outcome data on repeat hospitalizations and deaths
      AmioDrondaroneOverall
      (N=1735)(N=338)(N=2073)
      Outcomes
       No outcomes1558 (89.8%)291 (86.1%)1849 (89.2%)
       CV hospitalization146 (8.4%)43 (12.7%)189 (9.1%)
       Deaths31 (1.8%)4 (1.2%)35 (1.7%)
      Followup_time
       Mean (SD)465 (387)492 (405)470 (390)
       Median [Min, Max]356 [3.00, 1450]363 [4.00, 1460]357 [3.00, 1460]
      The Kaplan Meier survival curves for the composite endpoint is shown in Figure 1. There was no statistically significant difference between these 2 curves (log rank test, p = 0.1) for this unadjusted analysis. However given the baseline imbalances observed in Table 2, adjusted time to event analyses are required.
      The success of the propensity score model in adjusting for the baseline imbalances is shown in Figure 2 where the standardized mean differences are displayed and are judged acceptable (standardized mean differences < 0.1). The pseudo-populations created by the inverse probability weighting are shown graphically to have very similar propensity score distributions for all measured confounders (Figure 3).
      Figure thumbnail gr2
      Figure 2Standardized mean differences from models with different treatment effect estimands
      Figure thumbnail gr3
      Figure 3Histogram of propensity scores according to treatment received (dronedarone in dark green and amiodarone in dark blue). The weighted pseudo-population that is shown in the light colors and the propensity score histograms are similar for both treatment arms in these pseudo-populations.
      The Cox proportional model for these pseudo populations revealed a dronedarone hazard ratio of 1.47 (95%CI 1.01 - 2.12) for the average treatment effect for the primary outcome of recurrent CV hospitalization or death. Table 4 shows that a consistently increased dronedarone risk was observed when evaluating other average treatment effects.
      Table 4Cox proportional hazards for varying treatment effect estimands
      GroupCharacteristicHR195% CI1p-value
      ATEDronedarone vs. amiodarone1.410.97, 2.050.075
      ATTDronedarone vs. amiodarone1.521.08, 2.150.018
      ATCDronedarone vs. amiodarone1.250.86, 1.820.2
      ATMDronedarone vs. amiodarone1.521.08, 2.150.018
      1: HR = Hazard Ratio, CI = Confidence Interval
      *: Abbreviations; ATE = Average Treatment Effect, ATT = Average Treatment Effect Among the Treated, ATC = CAverage Treatment Effect Among the Controls, ATM = Average Treatment Effect Among the Evenly Matchable

      5. Discussion

      In this cohort of patients with an incident hospitalization for atrial fibrillation, an increased risk for recurrent CV hospitalizations or death was observed in patients receiving a discharge prescription for dronedarone compared to amiodarone. This increased risk was observed even after adjusting for baseline imbalances between the 2 groups in this non-randomized study. These results were robust to different average treatment effect measures.
      While RCTs remain the gold standard to evaluate drug efficacy, there has only been one small study of 504 patients[
      • Le Heuzey J.Y.
      • De Ferrari G.M.
      • Radzik D.
      • Santini M.
      • Zhu J.
      • Davy J.M.
      A short-term, randomized, double-blind, parallel-group study to evaluate the efficacy and safety of dronedarone versus amiodarone in patients with persistent atrial fibrillation: the dionysos study.
      ] that directly compared dronedarone to amiodarone and with a combined 7 deaths and no data reported on recurrent hospitalizations, this provides only minimal information as to the relative safety and efficacy of these two drugs. In an attempt to overcome this limitation, a mixed treatment comparison (MTC) including an additional 10 placebo controlled RCTs has provided an indirect comparison of clinical outcomes between dronedarone and amiodarone [
      • Freemantle N.
      • Lafuente-Lafuente C.
      • Mitchell S.
      • Eckert L.
      • Reynolds M.
      Mixed treatment comparison of dronedarone, amiodarone, sotalol, flecainide, and propafenone, for the management of atrial fibrillation.
      ]. This study found no statistically significant difference in mortality when amiodarone was compared to dronedarone (OR 2.52, 95% CI 0.72 - 8.90). Not withstanding the limitations of the indirect comparisons in a MTC, the data for mortality remains very limited as reflected by the wide confidence intervals. While recurrences of atrial fibrillation were less frequent with amiodarone (OR 0.41, 95% CI 0.29 - 0.57) compared to dronedarone, repeat hospitalizations were not systematically recorded in these trials and therefore a mixed treatment effect for our composite endpoint can’t be estimated. In this context, data from well performed non-experimental designs may be helpful. A general rule for minimizing bias from observational studies is to try and design them to emulate a RCT[
      • Hernan M.A.
      • Robins J.M.
      Using big data to emulate a target trial when a randomized trial is not available.
      ],[
      • Hernan M.A.
      • Sauer B.C.
      • Hernandez-Diaz S.
      • Platt R.
      • Shrier I.
      Specifying a target trial prevents immortal time bias and other self-inflicted injuries in observational analyses.
      ] and our study has followed these principles.
      Our study has several strengths including its incident design, avoiding possible prevalence bias. The low anticoagulation rates on admission are a testament to our incident cohort design. Discharge anticoagulation rates were 10 times those observed on admission, but still lower than expected given current guidelines[
      • January C.T.
      • Wann L.S.
      • Calkins H.
      • Chen L.Y.
      • Cigarroa J.E.
      • Cleveland Jr., J.C.
      • Ellinor P.T.
      • Ezekowitz M.D.
      • Field M.E.
      • Furie K.L.
      • Heidenreich P.A.
      • Murray K.T.
      • Shea J.B.
      • Tracy C.M.
      • Yancy C.W.
      Members Writing Group
      2019 aha/acc/hrs focused update of the 2014 aha/acc/hrs guideline for the management of patients with atrial fibrillation: A report of the american college of cardiology/american heart association task force on clinical practice guidelines and the heart rhythm society.
      ]. The reasons for this low rate are beyond the scope of the current research question but may reflect patient delays in filling out patient prescriptions. Nevertheless, the rate of discharge anticoagulant use was statistically higher in the dronedarone group (46% vs. 36%, p =0.001) so this imbalance can’t readily explain the observed outcome differences in the two exposure groups. By limiting our study to population to patients prescribed dronedarone or amiodarone in isolation of other anti-arrhythmic drugs should provide a sharper exposure contrast. Care in our study design has eliminated possible immortal time bias[
      • Hernan M.A.
      • Sauer B.C.
      • Hernandez-Diaz S.
      • Platt R.
      • Shrier I.
      Specifying a target trial prevents immortal time bias and other self-inflicted injuries in observational analyses.
      ]. Our propensity score methods suggest that measured confounding variables have been appropriately adjusted. Propensity score methods also separate the exposure and outcome models minimizing false positive errors from researcher degrees of freedom in the model selection process. Our results are also concordant with published RCTs that suggest concerns with dronedarone’s safety profile in patients with heart failure and permanent AF [
      • Connolly S.J.
      • Camm A.J.
      • Halperin J.L.
      • Joyner C.
      • Alings M.
      • Amerena J.
      • Atar D.
      • Avezum A.
      • Blomstrom P.
      • Borggrefe M.
      • Budaj A.
      • Chen S.A.
      • Ching C.K.
      • Commerford P.
      • Dans A.
      • Davy J.M.
      • Delacretaz E.
      • Di Pasquale G.
      • Diaz R.
      • Dorian P.
      • Flaker G.
      • Golitsyn S.
      • Gonzalez-Hermosillo A.
      • Granger C.B.
      • Heidbuchel H.
      • Kautzner J.
      • Kim J.S.
      • Lanas F.
      • Lewis B.S.
      • Merino J.L.
      • Morillo C.
      • Murin J.
      • Narasimhan C.
      • Paolasso E.
      • Parkhomenko A.
      • Peters N.S.
      • Sim K.H.
      • Stiles M.K.
      • Tanomsup S.
      • Toivonen L.
      • Tomcsanyi J.
      • Torp-Pedersen C.
      • Tse H.F.
      • Vardas P.
      • Vinereanu D.
      • Xavier D.
      • Zhu J.
      • Zhu J.R.
      • Baret-Cormel L.
      • Weinling E.
      • Staiger C.
      • Yusuf S.
      • Chrolavicius S.
      • Afzal R.
      • Hohnloser S.H.
      Pallas Investigators
      Dronedarone in high-risk permanent atrial fibrillation.
      ],[
      • Kober L.
      • Torp-Pedersen C.
      • McMurray J.J.
      • Gotzsche O.
      • Levy S.
      • Crijns H.
      • Amlie J.
      • Carlsen J.
      Group Dronedarone Study
      Increased mortality after dronedarone therapy for severe heart failure.
      ].
      Of note, in contrast to our findings a comparative effectiveness study of dronedarone published in 2020 [
      • Goehring Jr., E.L.
      • Bohn R.L.
      • Pezzullo J.
      • Tave A.K.
      • Jones J.K.
      • Bozzi S.
      • Tamayo Rcsg
      • Sicignano N.
      • Naccarelli G.V.
      Outcomes associated with dronedarone use in patients with atrial fibrillation.
      ] reported a decreased hazard ratio (HR = 0.86, 95% CI 0.78 to 0.95) for the composite outcome of CV hospitalization and death. However this cohort is substantially different from ours in that it was derived from an earlier time period, involved older patients with more comorbidities, was not an incident atrial fibrillation cohort (50% were on anticoagulants on admission) and most importantly the comparative group was a mixture of multiple different antiarrhythmic drugs including not only amiodarone but also sotalol, flecainide, propafenone, anddofetilide. Another comparative effectiveness study from Germany[
      • Ehrlich J.R.
      • Look C.
      • Kostev K.
      • Israel C.W.
      • Goette A.
      Impact of dronedarone on the risk of myocardial infarction and stroke in atrial fibrillation patients followed in general practices in germany.
      ] reported a lower risk of myocardial infarction and stroke in atrial fibrillation patients taking dronedarone but again the comparative group was a heterogeneous mixture of multiple anti-arrhythmic drugs. These group and exposure differences may well explain these discordant findings.
      Our study does have limitations. First while we have more outcomes than what has been reported in the clinical trials, the sample size nevertheless lacks the statistical power to examine mortality as a separate outcome. While our propensity score adjustments have successfully taken into consideration 16 baseline risk factors, there is no guarantee that unmeasured potential confounders, such a smoking, obesity and socio-economic determinants, are equally distributed between the groups. However, in order for these non-measured variables to be confounders, physicians must be basing their choice of anti-arrhythmic drug on these characteristics. While this may be unlikely for some lifestyle variables, it may be an issue for financial / insurance or other socio-economic variables that can also influence outcomes. Another limitation is the unexpectedly large number of patients who did not fill a prescription upon discharge. While an additional 9,248 did eventually fill an anti-arrhythmic drug prescription at a later date, they are not included as their introduction would introduce an immortal time bias. We have no means of knowing how many other patients received but elected not to fill their prescription possibly introducing a selection bias. Also this analysis follows an intention to treat paradigm where group assignment is determined at hospital discharge and exposure misclassification may occur over time if patients do not follow the same treatments. Therefore while these data sources have been used extensively in clinical research[
      • Patel M.D.
      • Wu D.
      • Chase M.R.
      • Mavros P.
      • Heithoff K.
      • Hanson M.E.
      • Simpson Jr., R.J.
      Antiplatelet therapy and clinical outcomes following myocardial infarction among patients in a u.s. employer-based insurance database.
      ],[
      • Kulaylat A.S.
      • Schaefer E.W.
      • Messaris E.
      • Hollenbeak C.S.
      Truven health analytics marketscan databases for clinical research in colon and rectal surgery.
      ], exposure, covariate and outcome misclassifications remain a possibility. This is perhaps especially true for mortality where the data set only records deaths that occur during a hospitalization. It is also worth recalling that our median follow-up time is only one year and the observed hazard rates may change with a longer follow-up.
      In conclusion, this study demonstrated that dronedarone compared to amiodarone treatment following a hospitalization for incident atrial fibrillation is associated with worse outcomes. Given the large burden of disease with atrial fibrillation, there is a pressing need to reproduce and expand these research findings in different settings.

      References

        • Andersson T.
        • Magnuson A.
        • Bryngelsson I.L.
        • Frobert O.
        • Henriksson K.M.
        • Edvardsson N.
        • Poci D.
        Allcause mortality in 272,186 patients hospitalized with incident atrial fibrillation 1995-2008: a Swedish nationwide long-term case-control study.
        Eur Heart J. 2013; 34 (ISSN 1522-9645 (Electronic) 0195-668X (Linking). doi: 10.1093/eurheartj/ehs469. URL https://www.ncbi.nlm.nih.gov/pubmed/23321349): 1061-1067
        • Chugh S.S.
        • Havmoeller R.
        • Narayanan K.
        • Singh D.
        • Rienstra M.
        • Benjamin E.J.
        • Gillum R.F.
        • Kim Y.H.
        • McAnulty Jr., J.H.
        • Zheng Z.J.
        • Forouzanfar M.H.
        • Naghavi M.
        • Mensah G.A.
        • Ezzati M.
        • Murray C.J.
        Worldwide epidemiology of atrial fibrillation: a global burden of disease 2010 study.
        Circulation. 2014; 129 (ISSN 1524-4539 (Electronic) 0009-7322 (Linking). doi: 10.1161/CIRCULATIONAHA.113.005119. URL https://www.ncbi.nlm.nih.gov/pubmed/24345399): 837-847
        • Connolly S.J.
        • Camm A.J.
        • Halperin J.L.
        • Joyner C.
        • Alings M.
        • Amerena J.
        • Atar D.
        • Avezum A.
        • Blomstrom P.
        • Borggrefe M.
        • Budaj A.
        • Chen S.A.
        • Ching C.K.
        • Commerford P.
        • Dans A.
        • Davy J.M.
        • Delacretaz E.
        • Di Pasquale G.
        • Diaz R.
        • Dorian P.
        • Flaker G.
        • Golitsyn S.
        • Gonzalez-Hermosillo A.
        • Granger C.B.
        • Heidbuchel H.
        • Kautzner J.
        • Kim J.S.
        • Lanas F.
        • Lewis B.S.
        • Merino J.L.
        • Morillo C.
        • Murin J.
        • Narasimhan C.
        • Paolasso E.
        • Parkhomenko A.
        • Peters N.S.
        • Sim K.H.
        • Stiles M.K.
        • Tanomsup S.
        • Toivonen L.
        • Tomcsanyi J.
        • Torp-Pedersen C.
        • Tse H.F.
        • Vardas P.
        • Vinereanu D.
        • Xavier D.
        • Zhu J.
        • Zhu J.R.
        • Baret-Cormel L.
        • Weinling E.
        • Staiger C.
        • Yusuf S.
        • Chrolavicius S.
        • Afzal R.
        • Hohnloser S.H.
        • Pallas Investigators
        Dronedarone in high-risk permanent atrial fibrillation.
        N Engl J Med. 2011; 365 (ISSN 1533-4406 (Electronic) 0028-4793 (Linking). doi: 10.1056/NEJMoa1109867. URL https://www.ncbi.nlm.nih.gov/pubmed/22082198): 2268-2276
        • Ehrlich J.R.
        • Look C.
        • Kostev K.
        • Israel C.W.
        • Goette A.
        Impact of dronedarone on the risk of myocardial infarction and stroke in atrial fibrillation patients followed in general practices in germany.
        Int J Cardiol. 2019; 278 (ISSN 1874-1754 (Electronic) 0167-5273 (Linking). doi: 10.1016/j.ijcard.2018.11.133. URL https://www.ncbi.nlm.nih.gov/pubmed/30528621): 126-132
        • Freemantle N.
        • Lafuente-Lafuente C.
        • Mitchell S.
        • Eckert L.
        • Reynolds M.
        Mixed treatment comparison of dronedarone, amiodarone, sotalol, flecainide, and propafenone, for the management of atrial fibrillation.
        Europace. 2011; 13 (ISSN 1532-2092 (Electronic) 1099-5129 (Linking). doi: 10.1093/europace/euq450. URL https://www.ncbi.nlm.nih.gov/pubmed/21227948): 329-345
        • Goehring Jr., E.L.
        • Bohn R.L.
        • Pezzullo J.
        • Tave A.K.
        • Jones J.K.
        • Bozzi S.
        • Tamayo Rcsg
        • Sicignano N.
        • Naccarelli G.V.
        Outcomes associated with dronedarone use in patients with atrial fibrillation.
        Am J Cardiol. 2020; 135 (ISSN 1879-1913 (Electronic) 0002-9149 (Linking). doi: 10.1016/j.amjcard.2020.08.026. URL https://www.ncbi.nlm.nih.gov/pubmed/32861738): 77-83
        • Hernan M.A.
        • Robins J.M.
        Using big data to emulate a target trial when a randomized trial is not available.
        Am J Epidemiol. 2016; 183 (ISSN 1476-6256 (Electronic) 0002-9262 (Linking). doi: 10.1093/aje/kwv254. URL https://www.ncbi.nlm.nih.gov/pubmed/26994063): 758-764
        • Hernan M.A.
        • Sauer B.C.
        • Hernandez-Diaz S.
        • Platt R.
        • Shrier I.
        Specifying a target trial prevents immortal time bias and other self-inflicted injuries in observational analyses.
        J Clin Epidemiol. 2016; 79 (ISSN 1878-5921 (Electronic) 0895-4356 (Linking). doi: 10.1016/j.jclinepi.2016.04.014. URL https: //www.ncbi.nlm.nih.gov/pubmed/27237061): 70-75
        • Kirchhof P.
        • Camm A.J.
        • Goette A.
        • Brandes A.
        • Eckardt L.
        • Elvan A.
        • Fetsch T.
        • van Gelder I.C.
        • Haase D.
        • Haegeli L.M.
        • Hamann F.
        • Heidbuchel H.
        • Hindricks G.
        • Kautzner J.
        • Kuck K.H.
        • Mont L.
        • Ng G.A.
        • Rekosz J.
        • Schoen N.
        • Schotten U.
        • Suling A.
        • Taggeselle J.
        • Themistoclakis S.
        • Vettorazzi E.
        • Vardas P.
        • Wegscheider K.
        • Willems S.
        • Crijns Hjgm
        • Breithardt G.
        • East-Afnet Trial Investigators
        Early rhythm-control therapy in patients with atrial fibrillation.
        N Engl J Med. 2020; 383 (ISSN 1533-4406 (Electronic) 0028-4793 (Linking). doi: 10.1056/NEJMoa2019422. URL https://www.ncbi.nlm.nih.gov/pubmed/32865375): 1305-1316
        • Kober L.
        • Torp-Pedersen C.
        • McMurray J.J.
        • Gotzsche O.
        • Levy S.
        • Crijns H.
        • Amlie J.
        • Carlsen J.
        • Group Dronedarone Study
        Increased mortality after dronedarone therapy for severe heart failure.
        N Engl J Med. 2008; 358 (ISSN 1533-4406 (Electronic) 0028-4793 (Linking). doi: 10.1056/NEJMoa0800456. URL https://www.ncbi.nlm.nih.gov/pubmed/18565860): 2678-2687
        • Kulaylat A.S.
        • Schaefer E.W.
        • Messaris E.
        • Hollenbeak C.S.
        Truven health analytics marketscan databases for clinical research in colon and rectal surgery.
        Clin Colon Rectal Surg. 2019; 32 (ISSN 1531-0043(Print)1530-9681(Linking). doi: 10.1055/s-0038-1673354. URL https://www.ncbi.nlm.nih.gov/ pubmed/30647546): 54-60
        • Le Heuzey J.Y.
        • De Ferrari G.M.
        • Radzik D.
        • Santini M.
        • Zhu J.
        • Davy J.M.
        A short-term, randomized, double-blind, parallel-group study to evaluate the efficacy and safety of dronedarone versus amiodarone in patients with persistent atrial fibrillation: the dionysos study.
        J Cardiovasc Electrophysiol. 2010; 21 (ISSN 1540-8167 (Electronic) 1045-3873 (Linking). doi: 10.1111/j.1540-8167.2010.01764.x. URL https://www.ncbi.nlm.nih.gov/pubmed/20384650): 597-605
        • Patel M.D.
        • Wu D.
        • Chase M.R.
        • Mavros P.
        • Heithoff K.
        • Hanson M.E.
        • Simpson Jr., R.J.
        Antiplatelet therapy and clinical outcomes following myocardial infarction among patients in a u.s. employer-based insurance database.
        J Manag Care Spec Pharm. 2017; 23 (ISSN 2376-1032 (Electronic). doi: 10.18553/jmcp.2017.23.6.684. URL https://www.ncbi.nlm.nih.gov/pubmed/28530518): 684-690
      1. R. Core Team. R: A language and environment for statistical computing. manual, 2021 2021. URL https://www.R-project.org/.

        • January C.T.
        • Wann L.S.
        • Calkins H.
        • Chen L.Y.
        • Cigarroa J.E.
        • Cleveland Jr., J.C.
        • Ellinor P.T.
        • Ezekowitz M.D.
        • Field M.E.
        • Furie K.L.
        • Heidenreich P.A.
        • Murray K.T.
        • Shea J.B.
        • Tracy C.M.
        • Yancy C.W.
        • Members Writing Group
        2019 aha/acc/hrs focused update of the 2014 aha/acc/hrs guideline for the management of patients with atrial fibrillation: A report of the american college of cardiology/american heart association task force on clinical practice guidelines and the heart rhythm society.
        Heart Rhythm. 2019; 16 (ISSN 1556-3871 (Electronic) 1547-5271 (Linking). doi: 10.1016/j.hrthm.2019.01.024. URL https://www.ncbi.nlm.nih.gov/pubmed/30703530): e66-e93