JOURNAL OF THE AMERICAN COLLEGE OF CARDIOLOGY
VOL. 65, NO. 25, 2015
ª 2015 BY THE AMERICAN COLLEGE OF CARDIOLOGY FOUNDATION PUBLISHED BY ELSEVIER INC.
ISSN 0735-1097/$36.00 http://dx.doi.org/10.1016/j.jacc.2015.04.045
Digoxin Use and Subsequent Outcomes Among Patients in a Contemporary Atrial Fibrillation Cohort Larry A. Allen, MD, MHS,* Gregg C. Fonarow, MD,y DaJuanicia N. Simon, MS,z Laine E. Thomas, PHD,z Lucas N. Marzec, MD,* Sean D. Pokorney, MD, MBA,z Bernard J. Gersh, MB, CHB, DPHIL,x Alan S. Go, MD,k Elaine M. Hylek, MD, MPH,{ Peter R. Kowey, MD,# Kenneth W. Mahaffey, MD,** Paul Chang, MD,yy Eric D. Peterson, MD, MPH,z Jonathan P. Piccini, MD, MHS,z for the ORBIT-AF Investigators
ABSTRACT BACKGROUND Although digoxin has long been used to treat atrial fibrillation (AF) and heart failure (HF), its safety remains controversial. OBJECTIVES This study sought to describe digoxin use over time in patients with AF who were stratified by the presence or absence of HF, to characterize the predictors of digoxin use and initiation, and to correlate digoxin use with outcomes. METHODS Longitudinal patterns of digoxin use and its association with a variety of outcomes were assessed in a prospective outpatient registry conducted at 174 U.S. sites with enrollment from June 2010 to August 2011. RESULTS Among 9,619 patients with AF and serial follow-up every 6 months for up to 3 years, 2,267 (23.6%) received digoxin at study enrollment, 681 (7.1%) were initiated on digoxin during follow-up, and 6,671 (69.4%) were never prescribed digoxin. After adjusting for other medications, heart rate was 72.9 beats/min among digoxin users and 71.5 beats/min among nonusers (p < 0.0001). Prevalent digoxin use at registry enrollment was not associated with subsequent onset of symptoms, hospitalization, or mortality (in patients with HF, adjusted hazard ratio [HR] for death: 1.04; without HF, HR: 1.22). Incident digoxin use during follow-up was not associated with subsequent death in patients with HF (propensity adjusted HR: 1.05), but was associated with subsequent death in those without HF (propensity adjusted HR: 1.99). CONCLUSIONS After adjustment for detailed clinical factors, digoxin use in registry patients with AF had a neutral association with outcomes under most circumstances. Because of the multiple conflicting observational reports about digoxin’s safety and possible concerns in specific clinical situations, a large pragmatic trial of digoxin therapy in AF is needed. (J Am Coll Cardiol 2015;65:2691–8) © 2015 by the American College of Cardiology Foundation.
From the *University of Colorado School of Medicine, Aurora, Colorado; yUniversity of California, Los Angeles, California; zDuke Clinical Research Institute, Durham, North Carolina; xMayo Clinic College of Medicine, Rochester, Minnesota; kKaiser Permanente Northern California, Oakland, California; {Boston University School of Medicine, Boston, Massachusetts; #Lankenau Heart Institute and the Jefferson Medical College, Philadelphia, Pennsylvania; **Stanford University School of Medicine, Palo Alto, California; and yyJanssen Scientific Affairs, Raritan, New Jersey. ORBIT-AF is sponsored by Janssen Scientific Affairs, LLC, Raritan, New Jersey. This project was supported in part by funding from the Agency of Healthcare Research and Quality through cooperative agreement number 1U19 HS021092. Dr. Allen is supported by the National Heart, Lung, and Blood Institute of the National Institutes of Health (NIH) under Award Number K23HL105896. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. Dr. Allen has served as a consultant for Amgen, Inc., Janssen Pharmaceuticals, and Novartis. Dr. Fonarow has received research grants or other research support from the NIH and the Agency for Healthcare Research and Quality; has consulted for Novartis, Amgen, Bayer, Gambro, Medtronic, and Janssen Pharmaceuticals; and has received support from the Ahmanson Foundation and the Corday Foundation. Dr. Pokorney has received educational grant support from AstraZeneca; research support from Gilead and Boston Scientific; and modest support from the advisory board of Janssen Pharmaceuticals. Dr. Kowey has been a paid consultant for Johnson & Johnson on multiple projects, including the development of rivaroxaban; has served as a consultant for the American College of Cardiology, AstraZeneca, Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, Cubist, Eli Lilly, Elsevier, Forest, GlaxoSmithKline, Medtronic, Omthera, Protola Pharma, and
2692
Allen et al.
JACC VOL. 65, NO. 25, 2015 JUNE 30, 2015:2691–8
Digoxin in Atrial Fibrillation: The ORBIT-AF Study
C
ABBREVIATIONS AND ACRONYMS AF = atrial fibrillation CI = confidence interval eGFR = estimated glomerular filtration rate
ardiac glycosides, such as digoxin,
contemporary cohort of patients with AF who were
have been used for decades to treat
stratified by the presence or absence of HF, to char-
patients with atrial fibrillation (AF)
acterize predictors of digoxin use and initiation, and
and those with heart failure (HF) to slow
to clarify the association of digoxin use with heart
atrioventricular nodal conduction and in-
rate, health-related quality of life (HRQOL) measures,
crease cardiac inotropy (1). With the develop-
hospitalization, and survival.
ment of alternative treatments for AF (2) and
EHRA = European Heart
HF (3), as well as concerns about digoxin’s
Rhythm Association
potential
HF = heart failure
proarrhythmic
properties
METHODS
and
long-term effects on cardiac remodeling (4),
We used data from the ORBIT-AF (Outcomes Registry
reduced ejection fraction
prescribing digoxin has decreased and is no
for Better Informed Treatment of Atrial Fibrillation)
HR = hazard ratio
longer recommended as first-line therapy
study to assess the use of digoxin and its association
for either disease (3,5). However, there
with outcomes. Details of the ORBIT-AF study design
quality of life
remain unmet needs for the treatment of
have been published previously (17). Briefly, ORBIT-
LVEF = left ventricular ejection
many subgroups of AF patients, including
AF was a U.S.-based, prospective outpatient registry
fraction
those with HF, which has prompted calls
of AF conducted at 176 sites nationwide. The Duke
OR = odds ratio
for renewed use of digoxin in certain clinical
Clinical Research Institute was responsible for ORBIT-
situations (6).
AF site selection and study management. Eligible
HFrEF = heart failure with
HRQOL = health-related
SEE PAGE 2699
patients were 18 years of age and older with electrocardiographically confirmed AF. Enrolling providers
Effectiveness and safety data for digoxin are rela-
included cardiologists, electrophysiologists, and pri-
tively limited. The only large randomized trial of
mary care providers. Site personnel entered infor-
digoxin, the DIG (Digitalis Investigation Group) trial,
mation
showed no effect on mortality, but digoxin did reduce
history, cardiovascular risk factors, AF management
hospitalization among patients with heart failure and
strategy, cardiac imaging, and provider characteris-
a reduced ejection fraction (HFrEF) (7). This trial only
tics into a standardized, web-based collection form.
enrolled patients in sinus rhythm, was conducted
The presence or absence of HF and New York Heart
between 1991 and 1993, and raised safety concerns at
Association (NYHA) functional class were determined
higher serum concentrations and in certain sub-
at baseline by medical record review. Following initial
groups, including women (8–10). A more recent
enrollment, longitudinal information was collected
observational analysis of patients with incident
during clinic visits at approximately 6-month intervals
HFrEF under routine care found that digoxin use was
for up to 36 months, and included information on
independently associated with higher mortality (11).
medication regimens, procedures, hospitalizations,
There are currently no large randomized trials of
quality of life, and vital status. We excluded patients
digoxin in patients with AF. Two post-hoc non-
who were missing information as to whether they were
randomized analyses of data from the large AFFIRM
taking digoxin. Written informed consent was ob-
(Atrial Fibrillation Follow-up Investigation of Rhythm
tained from all study participants. The Duke Institu-
Management) trial came to conflicting conclusions
tional Review Board approved the ORBIT-AF Registry;
(12,13). Post-hoc analysis of other AF trials have
all participating sites obtained approval from local
shown higher mortality associated with digoxin use
institutional review boards before entering patient
(14), as have real-world data from a large incident AF
data.
cohort from the Veterans Administration (15) and 2 large health maintenance organizations (16).
on
demographic
characteristics,
medical
Medication use was collected prospectively at each study visit, including a field specific for
Due to limited and conflicting data, we set out
digoxin. Dose and blood levels were not collected.
to describe digoxin use over time among a large
The follow-up visit date at which digoxin was
Spring Publishing; and has received research grants from Medtronic and St. Jude Medical. Dr. Peterson has received research funding from Janssen Pharmaceuticals, Sanofi, Genentech, Daiichi-Sankyo, Eli Lilly, and AstraZeneca; and has served as a consultant for Boehringer Ingelheim, AstraZeneca, and Janssen Pharmaceuticals. Dr. Piccini has received research grants from ARCA biopharma, Boston Scientific, Gilead, Janssen Pharmaceuticals, ResMed, St. Jude Medical; and has served as a consultant for Bayer, ChanRx, Johnson & Johnson, Medtronic, and Spectranetics. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Listen to this manuscript’s audio summary by JACC Editor-in-Chief Dr. Valentin Fuster. Manuscript received January 10, 2015; revised manuscript received March 31, 2015, accepted April 14, 2015.
Allen et al.
JACC VOL. 65, NO. 25, 2015 JUNE 30, 2015:2691–8
Digoxin in Atrial Fibrillation: The ORBIT-AF Study
first reported was defined as the time period of
estimated by the model predicted heart rate, with and
initiation.
without digoxin, with the adjustment variables set
The primary outcome of interest was all-cause
equal to their population average.
death. Additional outcomes of interest included heart
Associations between prevalent digoxin use at
rate, symptoms, HRQOL, all-cause hospitalization, and
baseline and subsequent all-cause death, all-cause
the composite of all-cause hospitalization and death.
hospitalization, cardiovascular hospitalization, and
Symptoms were measured using the European Heart
onset of symptoms were assessed in unadjusted and
Rhythm Association (EHRA) score of AF-related
adjusted analysis. According to the pre-specified
symptoms (18). HRQOL was assessed by the Atrial
analysis plan, primary analyses were stratified a pri-
Fibrillation Effect on Quality-of-Life questionnaire
ori by the presence or absence of HF. Time to first
in a subset of patients at baseline, and at 12 and
reported symptoms was measured at 6-month visit
24 months (19).
intervals, and a discrete time Cox model was used for
STATISTICAL
ANALYSIS. Characteristics
between
this outcome only; otherwise, exact event dates were
patients were described as frequency and percent for
used. The potential for clustering of patient outcomes
categorical
(interquartile
within a site was handled by adding a random effect
ranges) for continuous variables. The characteristics
for site (multivariable Cox frailty model). Models
were compared using the chi-square test for cate-
were adjusted for all covariates listed in Online
variables
and
median
gorical variables and the Wilcoxon rank sum test for
Table 3, which were determined to have: 1) partic-
continuous variables. The cohort was divided into
ular clinical relevance, determined a priori; or 2) a
those taking digoxin at study enrollment (prevalent
statistically significant association with any of the
use), those initiated on digoxin during follow-up
outcomes under evaluation, as previously identified
(incident use), and those not on digoxin at any time
by backward selection with stay criteria of 0.05. The
during the study. Characteristics among the groups
same set of covariates was used for adjustment of all
were compared using Pearson chi-square tests for
outcomes. Adjusted associations for outcomes were
categorical variables and Wilcoxon rank sum tests for
displayed as HRs (95% CI).
continuous variables.
Associations between incident digoxin use in
To examine factors associated with prevalent
follow-up and subsequent all-cause death, all-cause
digoxin use, a multivariable hierarchical logistic
hospitalization, and cardiovascular hospitalization
regression model was constructed using backward
were assessed through propensity score matching
selection for the binary outcome of digoxin use at
between patients initiated on digoxin at follow-up. It
baseline (see the covariates in Online Table 1 [54 pre-
was unusual to have subsequent follow-up in subjects
specified clinical and demographic characteristics
who initiated digoxin at the final follow-up period
and a random effect for the enrolling site] followed by
(30 to 36 months); therefore, digoxin initiation was
the inclusion criterion of p < 0.05 final model cova-
restricted to occur between 6 and 24 months. Anal-
riates in Online Table 2a). Prevalent digoxin users at
yses were conducted separately for patients with and
baseline were excluded from this model. Because
without HF. Each case (incident digoxin use) was
digoxin use was measured at 6-month visit intervals, a
matched to 3 control subjects (noninitiators) using
second multivariable, discrete time Cox frailty model
sequential stratification matching (20), which identi-
was constructed for the time to the first report of
fied matches from the same point in follow-up at
digoxin initiation (final model covariates in Online
which digoxin was initiated and used all available
Table 2b). Patients were censored from the risk set
covariate information up to that point (including HF
when they were lost to follow-up (mainly due to stag-
status). The criteria for matching was a single pro-
gered entry into the cohort). A third discrete time Cox
pensity score, obtained from a logistic regression
frailty model was constructed for digoxin discontinu-
model for digoxin initiation. Matching was conducted
ation among prevalent digoxin users (final model
sequentially, starting at 6 months and moving for-
covariates in Online Table 2c). Results were presented
ward through follow-up. At each visit, patients who
as odds ratios (ORs) and/or hazard ratios (HRs) with
initiated digoxin were matched to other subjects
corresponding 95% confidence intervals (CIs) and
who were still under follow-up at the same time, but
p values.
who had not yet started digoxin. The criteria for
Baseline heart rate was compared for baseline
identifying a match was “closeness” on a single pro-
digoxin use using a linear regression model that
pensity score value calculated at each visit. To be
accounted for other rate control medications (i.e.,
considered a match, patients had to have a difference
beta-blockers, verapamil, diltiazem, sotalol, and
in propensities no greater than a caliper of 20% of a
amiodarone). The adjusted mean heart rate was
SD. Standardized differences were used to evaluate
2693
2694
Allen et al.
JACC VOL. 65, NO. 25, 2015 JUNE 30, 2015:2691–8
Digoxin in Atrial Fibrillation: The ORBIT-AF Study
the success of propensity matching at achieving bal-
compared with 32.6% of patients who never received
ance. Outcomes assessment began immediately after
digoxin.
the time period of initiation, and the model was fit
Factors independently associated with digoxin use
using stratified Cox regression, stratified on the
at baseline included the following: rate control
matched pair (21).
strategy and absence of previous ablation; permanent
Pre-defined secondary analyses were performed
AF; worse HF functional class and LVEF; sinus node
in patient subgroups with renal function (esti-
dysfunction; larger left atria; lower diastolic blood
mated glomerular filtration rate [eGFR]