See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/327613917

Digoxin and clinical outcomes in the Global Rheumatic Heart Disease Registry

Article  in  Heart (British Cardiac Society) · September 2018

DOI: 10.1136/heartjnl-2018-313614

CITATIONS READS

3 125

8 authors, including:

Liesl Zuhlke Mark E Engel

University of Cape Town University of Cape Town
150 PUBLICATIONS   10,188 CITATIONS    108 PUBLICATIONS   2,562 CITATIONS   

SEE PROFILE SEE PROFILE

Sumathy Rangarajan Bongani M Mayosi

McMaster University University of Cape Town
129 PUBLICATIONS   10,524 CITATIONS    483 PUBLICATIONS   41,459 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

A systematic review of standard echocardiography versus handheld echocardiography for the detection of subclinical rheumatic heart disease View project

Prevention of Rheumatic Heart Disease in Africa View project

All content following this page was uploaded by Liesl Zuhlke on 21 October 2018.

The user has requested enhancement of the downloaded file.


►► Additional material is
published online only. To view
please visit the journal online
(http://d​ x.​doi.o​ rg/​10.​1136/​
heartjnl-2​ 018-​313614).
1
Department of Cardiology,
All India Institute of Medical
Sciences, New Delhi, India
2
Indian Institute of Public
Health, Gurgaon, India
3
Department of Paediatrics and
Child Health, Red Cross War
Memorial Children’s Hospital
and University of Cape Town,
Cape Town, South Africa
4
Department of Medicine,
Groote Schuur Hospital and
University of Cape Town, Cape
Town, South Africa
5
Population Health Research
Institute, Hamilton Health
Sciences and McMaster
University, Hamilton, Ontario,
Canada
Correspondence to
Dr Ganesan Karthikeyan,
Department of Cardiology,
All India Institute of Medical
Sciences, New Delhi 110029,
India; ​karthik2010@​gmail.​com
BMM is deceased.
Received 21 May 2018
Revised 27 July 2018
Accepted 1 August 2018
© Author(s) (or their
employer(s)) 2018. No
commercial re-use. See rights
and permissions. Published
by BMJ.
To cite: Karthikeyan G,
Devasenapathy N, Zühlke L,
et al. Heart Epub ahead of
print: [please include Day
Month Year]. doi:10.1136/
heartjnl-2018-313614
Original research article
Digoxin and clinical outcomes in the Global
Rheumatic Heart Disease Registry
Ganesan Karthikeyan,1 Niveditha Devasenapathy,2 Liesl Zühlke,3,4
Mark Emmanuel Engel,4 Sumathy Rangarajan,5 Koon K Teo,5 Bongani M Mayosi,4
Salim Yusuf,5 on behalf of the Global Rheumatic Heart Disease Registry (REMEDY)
Investigators
Abstract
Objective  Digoxin is widely used in patients with
rheumatic heart disease (RHD) despite a lack of data on
its impact on clinical outcomes. We aimed to determine
the association of digoxin use on clinical outcomes in
patients with RHD.
Methods  We performed a retrospective analysis of
the association of digoxin use with mortality at 2 years
in a large RHD registry. Secondary outcomes were
recurrent heart failure (HF) and hospitalisation for any
cause. We assessed associations using multivariable
logistic regression in the entire cohort and in subgroups
of patients with atrial fibrillation (AF) and HF. We also
estimated average treatment effects from propensity-
adjusted analyses using inverse probability treatment
weighting.
Results Information on digoxin use at baseline was
available for 98.7% (3298/3343) of patients. In the
overall population, digoxin was significantly associated
with mortality (OR 1.63, 95% CI 1.30 to 2.04,
p<0.0001) and recurrent HF (OR 1.48, 95% CI 1.07 to
2.04, p=0.019). On propensity-weighted analyses, this
effect was markedly attenuated (OR 1.05, 95% CI 1.01
to 1.09, p=0.005). Patients in sinus rhythm without HF
had a higher propensity-adjusted odds of death with
digoxin use (OR 1.06, 95% CI 1.01 to 1.12, p=0.015),
but those with both AF and HF had lower mortality (OR
0.88, 95% CI 0.80 to 0.98, p=0.019).
Conclusion  Digoxin use is associated with higher
mortality in patients with RHD, but this is greatly
attenuated on propensity adjustment, indicating the
presence of substantial treatment bias. The adjusted
estimates may therefore not be reliable, and large
randomised trials are needed to determine the true effect
of digoxin in patients with RHD.
Introduction
Rheumatic heart disease (RHD) is an important cause
of morbidity and mortality in less developed coun-
tries. A recent Global Burden of Disease Study report
estimates the current worldwide prevalence to be over
33 million cases and the number of deaths attributable
to RHD to be nearly 320 000 annually.1 Endemic
regions in low/middle-income countries account
for three-quarters of the burden of disease and
mortality. Mortality is higher in patients with atrial
fibrillation (AF), heart failure (HF) and poor func-
tional class. In the Global Rheumatic Heart Disease
Karthikeyan G, et al. Heart 2018;0:1–7. doi:10.1136/heartjnl-2018-313614
Valvular heart disease
Heart: first published as 10.1136/heartjnl-2018-313614 on 12 September 2018. Downloaded from http://heart.bmj.com/ on 29 September 2018 by guest. Protected by copyright.
Registry (REMEDY), which recruited 3343 patients
from 12 African countries, India and Yemen, over a
fifth of patients were in AF, a third were in HF and
nearly a quarter were in New York Heart Association
(NYHA) class III or IV at presentation.2 Mortality in
this population was 16.9% at 2 years.3 Markers of
severe valve disease, including HF and poor func-
tional class, were the strongest predictors of death.3
While the definitive treatment of patients with signif-
icant valve involvement is surgical or interventional
correction of the valvular abnormality, many patients
in low/middle-income countries do not receive timely
surgery or balloon valvuloplasty.3 Patients therefore
receive medical therapy for control of symptoms
while awaiting definitive treatment.
Even though there are no recommendations
for the use of digoxin in patients with RHD,4 5 it
is commonly prescribed. In the REMEDY study,
nearly 35% of patients were on treatment with
digoxin.2 Clinicians use digoxin with the intent to
control heart rate, both in patients with AF and also
in those who are in sinus rhythm and have signif-
icant mitral stenosis, where a reduction in heart
rate may translate into lower transvalvular pressure
gradients and improvement in symptoms.6 Further,
as in patients without valve disease, digoxin forms
part of the treatment regimen of patients with RHD
who are in HF. However, there are no large studies,
either observational or randomised, evaluating the
efficacy and safety of digoxin on clinical outcomes
in this population. Among patients in HF without
valve disease, one large trial indicated that digoxin
had a neutral effect on mortality but significantly
reduced hospitalisations for HF.7  However, more
recent data from large observational studies suggest 
that digoxin use is associated with increased
mortality.8 9 Patients with non-valvular AF are typi-
cally elderly and have a high prevalence of coro-
nary artery disease and may be at greater risk of
adverse effects due to digoxin (such as life-threat-
ening ventricular arrhythmia). On the other hand,
patients with RHD are young (median age 28 years
in REMEDY), and have a lower prevalence of hyper-
tension, diabetes and atherosclerotic cardiovas-
cular disease, and so the effects of digoxin may be
different in this population. We performed a retro-
spective analysis of observational data on digoxin
use in the REMEDY study in order to understand its
association with mortality and recurrent HF among
patients with RHD.
  1
 Valvular heart disease
Methods
Study design, participants and outcomes
The design, baseline characteristics and 2-year outcomes of the
REMEDY study have been published previously.2 3 10 Briefly,
REMEDY was a prospective, multicentre, international, hospi-
tal-based registry which enrolled 3343 patients with a clin-
ical diagnosis of RHD, who were seen at 25 hospitals in 14
countries (12 African countries, Yemen and India). The key
outcomes of interest were death, HF, stroke, transient isch-
aemic attack or non-central nervous system systemic embolism
at 2 years.
For the purposes of this analysis, we considered patients to
be digoxin users if they were on the drug at baseline. The prin-
cipal outcome was mortality at 2 years. Other outcomes studied
were recurrent HF, all hospitalisations and the incidence of
valve surgery or percutaneous interventions for valve disease.
We also assessed the effect of digoxin use on the composite
outcomes of death or recurrent HF, and death, recurrent HF
or hospitalisation. The severity of valve lesions and other echo-
cardiographic variables were described using standard criteria.2
Patients were considered to have severe valve disease if at least
one of the involved valves had severe disease. HF at baseline
and during the study (recurrent HF) was diagnosed if any two of
the following criteria were present: (1) symptoms (dyspnoea on
exertion or at rest, orthopnoea,paroxysmal nocturnal dyspnoea
or ankle oedema) or signs (rales, increased jugular venous pres-
sure or ankle oedema) of congestive HF; (2) radiologic signs of
pulmonary congestion; and (3) treatment with diuretics. Surgery
for valve disease was defined as performance of valve repair,
or replacement of any affected valve with a tissue, or mechan-
ical prosthesis. Percutaneous valve interventions consisted of
balloon dilatation of stenosed mitral, aortic or tricuspid valves.
Patients who had at least one overnight admission to hospital
were considered to have a hospitalisation event. Details and defi-
nitions of the baseline and outcome measures used have been
published previously.2 3 10
Statistical methods
We compared the baseline characteristics of digoxin users and
non-users using independent t-tests and Χ2 tests as appropriate.
We performed univariable and multivariable logistic regression
to assess the effect of digoxin on the following outcomes: death,
recurrent HF, hospitalisation, surgical or percutaneous valve
intervention and on the composites of death or recurrent HF, and
death, recurrent HF or hospitalisation. The variables included
in the multivariable model were decided a priori based on their
relevance to prognosis and were identical to the ones used previ-
ously in our primary analysis of outcomes.3 They were: age, sex,
presence of AF or atrial flutter, NYHA class, history of HF or
HF at enrolment, previous heart valve surgery or intervention,
a history of stroke or IE, severe disease (severe valve disease
in at least one affected valve), multivalve involvement and use
of secondary prophylaxis. We performed subgroup analyses to
assess effect modification by the presence of AF or HF at base-
line, on the association of digoxin on the principal outcomes. In
addition, we tested for interactions between gender, body mass
index (BMI) and concomitant beta-blocker use with digoxin , by
adding appropriate interaction terms in the model. Finally, we
also assessed the effect of digoxin use in the subpopulation of
patients who had both AF and HF at baseline when compared
with those who had neither.
2 Karthikeyan G, et al. Heart 2018;0:1–7. doi:10.1136/heartjnl-2018-313614
Propensity-adjusted analysis
Heart: first published as 10.1136/heartjnl-2018-313614 on 12 September 2018. Downloaded from http://heart.bmj.com/ on 29 September 2018 by guest. Protected by copyright.
We generated propensity scores for each patient using logistic
regression with digoxin as the dependent variable. The vari-
ables used in this model (listed in figure 2) were selected based
on their potential association with digoxin use and clinical
outcomes, in accordance with expert recommendations.11–14
After generating the propensity scores we tested for similarity
of distribution, adequate overlap of the distributions (common
support) (online supplementary figure S1) and ensured balance
of covariates between the two groups (digoxin and no digoxin)
within blocks of propensity scores. After creating a balanced
propensity score, we used a weighting method (inverse prob-
ability treatment weights, IPTW) and a matching method (1:2
calliper matching with replacement, with calliper width set at
0.2×SD of the logit of the propensity score) to ensure balance
of covariates between the two groups.14 Balance was assessed by
the standardised percentage of bias for each covariate. An overall
mean standardised percentage of bias <5%, or a standardised
mean difference <0.1, is considered to indicate good balance.
We estimated average treatment effects of digoxin on outcomes
using the IPTW as the primary analysis and corroborated the
results using those obtained from 1:2 calliper matching. Further,
if statistically significant subgroup effects were present, we
generated separate propensity scores for subgroups of patients
who were in AF, HF or both at baseline (variables listed in online
supplementary table S2) and followed similar procedures to
obtain effect estimates of digoxin use on outcomes.
The effect of digoxin use on outcomes was estimated using
ORs and their 95% CIs. A p value of <0.05 was considered
statistically significant. All analyses were done using Stata V.14
(StataCorp, College Station, Texas) and the propensity anal-
yses were performed using the pscore, psmatch2, dr and teffects
packages.
Results
Of the 3343 patients in REMEDY, 3298 (98.7%) had informa-
tion on baseline digoxin use, and 1144 (34.7%) of these were
on digoxin. The proportion of patients lost to follow-up among
those on and those not on digoxin was similar (figure 1). Patients
who were on digoxin at baseline were more likely to reside in
low-income countries, were older, in AF, HF and poor func-
tional class, and were more likely to be on diuretics and ACE
inhibitors (or angiotensin receptor blockers).(online supplemen-
tary table S1) Overall, 51% (1709/3343) of patients had severe
valve disease. The prevalence of clinical and echocardiographic
markers of severe valve involvement was higher among those
prescribed digoxin (table 1). After propensity score weighting,
we achieved excellent balance between the two groups as
evidenced by the small (<5%) standardised mean difference
between covariates (figure 2). The variables used to generate the
propensity scores and the percentage bias reduction are depicted
in figure 2. Balance of covariates for the 1:2 calliper matched
analysis is shown in online supplementary figures S2 and S3.
Effect of digoxin on clinical outcomes
All patients
Among all patients, digoxin use was significantly associated
with mortality (OR 1.63, 95% CI 1.30 to 2.04, p<0.0001)
and recurrent HF (OR 1.48, 95% CI 1.07 to 2.04, p=0.019)
at 2 years (table 2). There was no impact of digoxin on hospi-
talisations or the performance of percutaneous or surgical
interventions. The composite outcome of death or HF, and
death, HF or hospitalisations was also significantly increased

Figure 1  Flow of patients contributing to the digoxin versus no digoxin analysis. Numbers are for patients using digoxin at the time of enrolment
into the study. REMEDY, Global Rheumatic Heart Disease Registry.
with digoxin use (table 2). The association of digoxin with
mortality was not modified by low BMI, female gender or
the concomitant use of beta-blockers (p values for interaction
0.14, 0.08 and 0.35, respectively).
On propensity score weighted analyses, the effect of digoxin
on mortality was markedly attenuated, though remaining nomi-
nally statistically significant (OR 1.05, 95% CI 1.01 to 1.09,
p=0.005). The effect on the composite outcomes was similarly
attenuated but the association with recurrent HF was no longer
statistically significant (table 2). Propensity score matched anal-
ysis by the 2:1 calliper matching method yielded identical results
(online supplementary table S3).
Patients in AF
Digoxin was used by 454 (41%) patients in AF and 548 (31.5%)
patients without AF. More patients taking digoxin died both
among patients in AF (113, 24.9% vs 94, 14.4%) and among
those in sinus rhythm (148, 27% vs 133, 10.7%), compared
with those not on digoxin. In multivariable analyses, digoxin
was associated with a significantly higher odds of death among
patients in sinus rhythm than among patients in AF (p for inter-
action=0.03), which remained significant, although markedly
attenuated on propensity adjustment (OR 1.07, 95% CI 1.03
to 1.11, p=0.001) (table 3). Propensity weight adjustment
showed no significant effect on mortality with digoxin use
among patients in AF (OR 1.01, 95% CI 0.95 to 1.07, p=0.87)
(table 3). For recurrent HF, and the composite of death or HF,
there was no significant difference in the effect of digoxin among
those with and without AF at baseline (online supplementary
table S4 and S5).
Patients in HF
About half of the patients (532, 50.2%) who were in HF,
or who had history of HF, were on digoxin compared with
about a quarter of those who did not have HF (481, 25.8%).
Karthikeyan G, et al. Heart 2018;0:1–7. doi:10.1136/heartjnl-2018-313614
Valvular heart disease
Heart: first published as 10.1136/heartjnl-2018-313614 on 12 September 2018. Downloaded from http://heart.bmj.com/ on 29 September 2018 by guest. Protected by copyright.
A greater proportion of patients taking digoxin died, both
among patients with (156, 29.3% vs 98, 18.6%) and without
HF (110, 22.9% vs 130, 9.4%). Similar to the effect in patients
with AF, digoxin use was associated with a greater odds of
death among patients who did not have HF compared with
those who had HF (table 3). On propensity adjustment, this
remained significant though markedly attenuated (OR 1.08,
95% CI 1.04 to 1.12, p<0.0001). Propensity weight adjusted
analysis suggested a neutral effect of digoxin use on mortality
among patients with HF (OR 0.99, 95% CI 0.94 to 1.06,
p=0.83) (table 3). Similar results were seen for recurrent HF,
and the composite of death or HF (online supplementary table
S4 and S5).
Patients with both AF and HF
A significant proportion of patients (412, 14.1%) had both HF
and AF at baseline. Over half of these patients (245, 59.5%)
were on digoxin. Nearly a quarter of the patients without
either HF or AF (265/1151, 23%) also received digoxin.
Among patients with both HF and AF, a similar proportion
of digoxin users died compared with non-users (65, 26.5%
vs 42, 25.2%). However, among those with neither HF nor
AF, the proportion of deaths among digoxin users was greater
than among non-users (58, 21.9% vs 77, 8.7%). This asso-
ciation persisted in multivariable analyses (OR for patients
not in AF or HF 2.26, 95% CI 1.5 to 3.42; OR for patients
with both HF and AF 0.71, 95% CI 0.43 to 1.17, p for inter-
action <0.0001). However, there was no difference in the
odds of developing recurrent HF. (online supplementary table
S4) Propensity score adjustment suggested persistent harm in
patients with neither AF nor HF, and a protective effect of
digoxin in mortality for patients who had both AF and HF (OR
0.88, 95% CI 0.80 to 0.98, p=0.019) (table 3). Results for the
composite outcome of death or HF were broadly similar in
this group of patients (online supplementary table S5).
3
 Valvular heart disease
the three drugs, digoxin produced the least improvement in

Heart: first published as 10.1136/heartjnl-2018-313614 on 12 September 2018. Downloaded from http://heart.bmj.com/ on 29 September 2018 by guest. Protected by copyright.
Table 1  Comparison of baseline characteristics of patients who
subjective symptoms and peak treadmill exercise capacity. In
completed 2-year follow-up
the Control of Rate versus Rhythm in Rheumatic Atrial Fibril-
Baseline characteristics Digoxin No digoxin P values lation Trial study, although rate control regimens were not
Participants, n (%) 1014 (34.7) 1908 (65.3) systematically evaluated, digoxin was used as add-on therapy
Mean age in years (SD) 32.3 (16) 30.2 (15.8) 0.0009 to diltiazem in 15% of patients in the rate control arm to
Female 671 (66.2) 1276 (67) 0.687 optimise treatment.15 However, neither of these studies was
Schooling beyond primary level 415 (41.1) 875 (46.2) 0.010 designed or powered to evaluate hard outcomes in relation to
Country income group <0.0001 digoxin use.
 LIC 468 (46.2) 491 (25.7)
 LMIC 271 (26.7) 867 (45.4) Digoxin and mortality
 UMIC 275 (27.1) 550 (28.8) Numerous observational studies and a few randomised
Severe disease* 658 (64.9) 793 (41.6) <0.0001 trials have addressed the effect of digoxin use and mortality
Multivalve disease 714 (70.4) 1037 (54.4) <0.0001 in patients with non-valvular AF and HF, but have yielded
Atrial fibrillation 454 (45.3) 652 (34.5) <0.0001 contrasting results. Observational studies have in general
Heart failure 532 (52.5) 527 (27.6) <0.0001 suggested that digoxin use is associated with about a 20% higher
NYHA III/IV 315 (31.6) 353 (18.8) <0.0001
mortality.8 9 16 On the other hand, digoxin has been shown
Prior intervention or surgery 221 (21.8) 512 (26.9) 0.003
to have a neutral effect on mortality in randomised trials.7 17
Much of this difference is likely to be due to confounding
Prior stroke 77 (7.6) 134 (7) 0.557
by indication (prescription/treatment bias) in observational
Prior infective endocarditis 70 (6.9) 104 (5.5) 0.113
studies.18 In a metaregression of all-cause mortality according
Secondary prophylaxis at enrolment 454 (47) 1085 (58.5) <0.0001
to risk of bias, Ziff et al showed an increase in the risk of
Mitral valve area less than 1.5 cm2 327 (32.3) 513 (26.9) 0.002
mortality with increasing risk of bias.17 Propensity-adjusted
Dilated left atrium (≥50 mm) 562 (59.7) 555 (31.5) <0.0001 analyses may help reduce confounding due to known vari-
Dilated left ventricle† 483 (49.1) 552 (30.2) <0.0001 ables to some extent. Expectedly, studies employing propen-
Decreased ejection fraction 337 (34.5) 368 (20.2) <0.0001 sity matched analysis show a considerably attenuated risk of
Tricuspid regurgitation 449 (44.3) 416 (21.8) <0.0001 death associated with digoxin, with some reporting a neutral
Pulmonary arterial hypertension 338 (33.7) 452 (23.8) <0.0001 effect and others showing a smaller increase in risk than that
Concomitant beta-blocker use 410 (40.6) 737 (38.6) 0.3 obtained after conventional multivariable adjustment.17 In our
Calcium channel blocker use 52 (5.2) 95 (5.0) 0.819 analysis, we found an OR of 1.63 for death on multivariable
Diuretic use 915 (90.2) 1073 (56.2) <0.0001 adjustment which attenuated markedly to 1.05 in the propen-
ACE inhibitors/receptor blockers 385 (38.2) 523 (27.4) <0.0001 sity-weighted analysis, which is similar to the risk ratios (RR)
*The presence of severe disease in at least one of the involved valves. obtained by Ziff et al (RR 1.61 and 1.18).17 The large differ-
†Left ventricular end diastolic dimension >50 mm in children and >55 mm in adults. ence in risk between the propensity-adjusted and unadjusted
LIC, low-income country; LMIC, lower middle-income country; NYHA, New York analyses highlights the substantial prescription bias in digoxin
Heart Association functional class; UMIC, upper middle-income country. use in observational studies.19

Discussion Digoxin, recurrent HF and hospitalisation

In this large cohort of patients with symptomatic RHD,
digoxin use was associated with a small but nominally signif-
icant increase in mortality. This association was primarily
driven by the effect of digoxin among patients without either
AF or HF. By contrast, among patients with conventional
indications for digoxin use such as AF or HF, propensity-ad-
justed analyses suggested a neutral effect on mortality, and a
potential reduction in mortality in those who had both AF and
HF. These results are based on a well-powered (494 deaths),
robust, propensity-adjusted analysis in a cohort of well-char-
acterised patients with RHD. However, these  results need to
be interpreted with caution because of the non-randomised
and retrospective nature of the analyses, and the large differ-
ence between the unadjusted and propensity-adjusted esti-
mates , which highlight the magnitude of bias.
Digoxin in RHD
To our knowledge, this is the only study relating digoxin use
to clinical outcomes in patients with RHD. The few previous
studies have enrolled small numbers of patients and have
reported subjective or surrogate outcomes such as symptom
relief and exercise capacity. In an open-label, cross-over trial,
Ahuja et al randomised 10 patients with AF to receive digoxin,
verapamil or metoprolol and compared the efficacy of these
agents in improving symptoms and exercise capacity.6 Of
4 Karthikeyan G, et al. Heart 2018;0:1–7. doi:10.1136/heartjnl-2018-313614
Unlike in previous studies of non-valvular AF and HF,17 we
did not observe any benefit of digoxin on recurrent HF or
hospitalisation. There could be several possible explanations
for this discrepancy. First, even if in reality digoxin reduced
HF, its effects may be masked by the substantial differences
in inherent risks between patients prescribed digoxin and
those who were not. Second, worsening HF in patients with
RHD may be due to progression of valve disease or occur-
rence of new valve disease which are unlikely to be affected by
digoxin. We did not collect information on cause of admission
to hospital during follow-up. Hospitalisation in patients with
valve disease could be due to progression of valve disease,
admission for performance of urgent interventions or due
to complications such as stroke, infective endocarditis or
recurrence of rheumatic fever, none of which are affected by
digoxin use.
Digoxin use in patients without AF or HF
Unlike patients without valve disease, in whom digoxin is
only used in those with AF or HF, a significant proportion
of patients with RHD receive digoxin in the absence of these
conditions (primarily with a view to reducing transvalvular
pressure gradients and symptoms in patients with mitral
stenosis). In this study, treating physicians appear to have
prescribed digoxin in the absence of AF or HF to patients
 Valvular heart disease

Heart: first published as 10.1136/heartjnl-2018-313614 on 12 September 2018. Downloaded from http://heart.bmj.com/ on 29 September 2018 by guest. Protected by copyright.
Figure 2  Standardised percentage of bias across covariates before and after inverse probability treatment weighting. Figure illustrates the balance
of covariates between treatment and comparison groups before and after inverse probability weighting. Standardised mean differences of covariates
after matching (indicated by red solid circles) are well within the prespecified ±5 margin indicating good comparability between the two groups.
NYHA, New York Heart Association; P/H/O, Past history of stroke.

who had severe disease, larger left atrial and left ventric-
ular dimensions, along with diuretics and ACE inhibitors or
receptor blockers (online supplementary table S6). We found
a significant increase in the odds of death with digoxin use in
patients without either AF or HF. Digoxin appeared to confer
a significant reduction in mortality among those with both AF
and HF which persisted in propensity-adjusted analysis. These
results are consistent with previous studies which showed that
the benefits of digoxin are most pronounced in patients with
severe HF,7 and the risk of death is highest in those with no
HF and the lowest baseline cardiac risk.20 The mechanisms of
increased mortality with digoxin among low-risk patients are
unclear. It is plausible that among patients in AF and HF, and
the accompanying neuroendocrine derangements, the salutary
effects of digoxin may balance its deleterious effects such as
proarrhythmia, while low-risk patients may only be exposed

Table 2  Clinical outcomes with digoxin use

OR† by inverse probability
Crude OR (95% CI) aOR* (95% CI) treatment weighted analysis
Digoxin No digoxin P values P values (95% CI) P values
Clinical outcomes n=1014 n=1908 n=2922 n=2730 n=2453
Death 266 (26.2) 228 (12)   2.62 (2.15 to 3.19)   1.63 (1.30 to 2.04) 1.05 (1.01 to 1.09)
<0.0001 <0.0001 0.005
Recurrent HF 109 (10.8) 90 (4.7)   2.43 (1.82 to 3.25) 1.48 (1.07 to 2.04) 1.02 (1.00 to 1.04)
<0.0001 0.019 0.114
Any hospitalisation 219 (21.6) 313 (16.4) 1.40 (1.16 to 1.70) 1.05 (0.85 to 1.31) 1.00 (0.96 to 1.03)
0.001 0.640 0.929
Percutaneous or surgical 114 (11.2) 155 (8.1) 1.43 (1.11 to 1.85) 1.04 (0.78 to 1.38) 1.01 (0.98 to 1.04)
intervention 0.006 0.789 0.556
Death or recurrent HF 323 (31.9) 277 (14.5)   2.75 (2.29 to 3.31)   1.68 (1.36 to 2.08) 1.06 (1.03 to 1.1)
<0.0001 <0.0001 0.001
Death, recurrent HF or 434 (42.8) 478 (25.1)   2.24 (1.90 to 2.63)   1.49 (1.24 to 1.79) 1.06 (1.02 to 1.11)
hospitalisation <0.0001 <0.0001 0.006
*aOR, OR adjusted for age, gender, education, atrial fibrillation, severity of valve disease, multiple valve involvement, New York Heart Association (NYHA) class, heart failure in
the past or at enrolment, previous valve surgery or intervention, secondary penicillin prophylaxis; addition of prior stroke and infective endocarditis produced identical results. The
effect of digoxin on mortality was not modified by low body mass index (BMI) (p value for interaction=0.142), female gender (p value for interaction=0.084) or concomitant use
of beta-blockers (p value for interaction=0.335).
†Average treatment effects (ATE) generated from inverse probability treatment weighting.
HF, heart failure.

Karthikeyan G, et al. Heart 2018;0:1–7. doi:10.1136/heartjnl-2018-313614 5

 Valvular heart disease

Heart: first published as 10.1136/heartjnl-2018-313614 on 12 September 2018. Downloaded from http://heart.bmj.com/ on 29 September 2018 by guest. Protected by copyright.
Table 3  Effect of digoxin on mortality by presence of AF, HF or both at baseline
Multivariable logistic regression analysis Inverse probability treatment weighted analysis
OR OR
n Digoxin No digoxin (95% CI) P values n Digoxin No digoxin (95% CI) P values
Atrial fibrillation
 Patients in AF 1058 424 634 1.23 942 376 566 1.01 0.866
(0.88 to 173) (0.95 to 1.07)
 Patients in sinus rhythm 1672 507 1165 2 0.03* 1549 489 1060 1.07 0.001
(1.5 to 2.7) (1.03 to 1.07)
Heart failure
 Patients in HF 1009 503 506 1.26 877 448 429 0.99 0.834
(0.92 to 1.73) (0.94 to 1.06)
 Patients without HF 1721 428 1293 2.09 0.024* 1641 428 1213 1.08 <0.0001
(1.53 to 2.86) (1.04 to 1.12)
Atrial fibrillation and heart failure
 Patients in both AF and HF 403 237 166 0.71 348 210 138 0.88 0.019
(0.43 to 1.17) (0.80 to 0.98)
 Patients in sinus rhythm 1066 241 825 2.26 <0.0001* 970 228 742 1.06 0.015
without HF (1.5 to 3.42) (1.01 to 1.12)
 Overall 2730 931 1799 1.63 <0.0001 2453 842 1611 1.05 0.005
(1.3 to 2.04) (1.01 to 1.09)
*P values for interaction.
AF, atrial fibrillation; HF, heart failure.

to its deleterious effects without deriving any benefit.20 21 assist in designing large trials of the use of digoxin in patients
These data may potentially be used to guide the choice of with RHD.
patient populations for inclusion in randomised controlled
trials of digoxin in RHD. Key messages

What is already known on this subject?

Limitations
►► Among patients with non-valvular atrial fibrillation (AF),
The most important limitation of our study is that this was a
observational data suggest that digoxin may increase
retrospective analysis of observational data, and was therefore
mortality. However, there are no data relating digoxin use to
subject to all the biases associated with this design, particularly
clinical outcomes in patients with rheumatic heart disease
confounding by indication (prescription bias).19 We attempted
(RHD).
to mitigate imbalances between the digoxin and no digoxin
groups using propensity score weighting and matching, but
What might this study add?
there were large differences in the risk associated with digoxin
►► This is the first report of the effect of digoxin on clinical
use between multivariable and propensity-adjusted analyses.
outcomes in patients with RHD. These data suggest a possible
In the presence of large biases, we believe that no amount of
association of digoxin and mortality in this population,
statistical adjustment will be sufficient to yield reliable effect
particularly among those without AF or heart failure (HF).
estimates. The substantial attenuation of the apparent excess
mortality raises the possibility that the residual excess risk
How might this impact on clinical practice?
may be spurious.19 We only included variables present at base-
►► The attenuation of the association of digoxin with mortality
line in our analyses, and did not account for the occurrence of
with propensity adjustment indicates the presence of
AF or HF, and initiation of digoxin during follow-up. Given
substantial treatment bias in observational analyses, and
that this study was performed in low-resource settings, we did
highlights the need for randomised trials of digoxin in RHD. In
not measure serum digoxin levels, which may be related to
the meantime, it may be prudent for clinicians to avoid using
mortality.22 Finally, we did not collect data on cause of death.
digoxin in patients with RHD who neither have AF or HF.
This information would have provided useful insights into the
mechanisms contributing to increased mortality with digoxin.
Acknowledgements  The authors acknowledge the substantial contribution
made by the following individuals: Addis Ababa, Ethiopia: Araya Gidey Desta, Bekele
Conclusion Alemayehu Shasho, Dufera Mekonnen Begna; New Delhi, India: Jitender Sharma,
Digoxin use appears to be associated with a slightly higher Gaurav Purohit; Nairobi, Kenya: Christine Yuko Jowi; Windhoek, Namibia: Henning du
mortality in patients with RHD, which was mainly seen in Toit, Masomi Kaaya, Liina Sikwaya, Andreas Wilberg; Abeokuta, Nigeria: Tanimowo
patients who did not have AF or HF. But these estimates are Sunkanmi; Jos, Nigeria: Ludu Audu, Charity Durojaiye-Amodu, Ngozi Elekwa, Ogechi
based on a retrospective, propensity-adjusted analysis and Maduka , Oludolapo Marcaulay, Shamsudeen Mohammed, Halim Odiachi; Cape
may not be reliable due to the substantial biases associated Town, South Africa: Dylan Barth, Patrick Commerford, Rezeen Daniels, Veronica
with digoxin prescription and its use. However, these data Francis, Felicia Gili, Alexia Joachim, John Lawrenson, Carolise Lemmer, Nonkululeko
suggest that patients with RHD who have AF, and/or HF may Koyana, Katya Mauff, Kathryn Manning, Wendy Matthiassen, Alet Meiring, Peggy
potentially derive some benefit from digoxin use. Our data Mgwayi, Lwazi Mhlanti, Simpiwe Nkepu, Mpiko Ntsekhe, Janine Saaiman, Unita
provide the impetus and background information that can September, Kathie Walker, Marnie van de Wall; Polokwane, South Africa: Priscilla

6 Karthikeyan G, et al. Heart 2018;0:1–7. doi:10.1136/heartjnl-2018-313614


Adolf, Jabulani Mbokazi, Susan Perkins; Maputo, Mozambique: Neusa Jessen;
Khartoum, Sudan: Tagwa Eltahir. The authors also thank Dr Shrikant Bangdiwala
(Hamilton, Canada) for reviewing the statistical methods. Dedication: The authors
wish to dedicate this manuscript to Prof Bongani Mayosi, who was a dear
colleague, friend and mentor. This work would not have been possible
without his efforts.
Collaborators  The REMEDY Investigators: clinical investigators and sites: Egypt:
Benha University: Azza Abul Fadl (Principal Investigator); Cairo University: Sahar S
Sheta (Principal Investigator). Ethiopia: University of Jimma: Abraham Haileamlak
(Principal Investigator); University of Addis Ababa: Senbeta G Abdissa, Dufera
M Begna, Wandimu Daniel, Araya G Desta, Dejuma Yadeta Goshu (Principal
Investigator), Bekele A Shasho. Kenya: University of Nairobi: Bernard Gitura, Stephen
Ogendo (Principal Investigator). Malawi: University of Malawi: Neil Kennedy
(Principal Investigator). Mozambique: Eduardo Mondlane University: Albertino
Damasceno (Principal Investigator); Instituto Nacional de Saúde: Ana Olga Mocumbi
(Principal Investigator). Namibia: Windhoek Hospital: Christopher Hugo-Hamman
(Principal Investigator). Nigeria: University of Abuja: Dike Ojji (Principal Investigator);
University of Jos: Ganiyu A Amusa, Fidelia Bode-Thomas (Principal Investigator),
Christopher C Yilgwan, Olukemi Ige, Basil Okeahialam; Kano University: Mahmoud
U Sani (Principal Investigator); University of Ibadan: Okechukwu S Ogah (Principal
Investigator); Federal Medical Centre, Abeokuta: Taiwo Olunuga; University College
Hospital, Ibadan: Abiodun M Adeoye, Okechukwu Ogah (Principal Investigator).
Rwanda: King Faisal Hospital: Joseph Mucumbitsi (Principal Investigator). South
Africa: University of Cape Town: Blanche Cupido; University of Limpopo: Phindile
Mntla (Principal Investigator), Christopher Sutton (Principal Investigator), Rajeev
Misra. Sudan: Alzaiem Alazhari University: Ahmed ElSayed (Principal Investigator),
Ahmed S Ibrahim; Ahmed Gasim Teaching Hospital: Huda HM Elhassan (Principal
Investigator). Uganda: Makerere University: Peter Lwabi, Charles Mondo (Principal
Investigator), Emmy Okello. Yemen: University of Sana’a: Mohammed M Al-Kebsi
(Principal Investigator). Zambia: University of Lusaka: John Musuku (Principal
Investigator).
Contributors  Conception and design: GK, ND. Data acquisition: LZ, MEE, SR,
GK. Data analysis: ND. Data interpretation: all authors. Drafting of manuscript: GK.
Critical revision for important intellectual content: all authors. Approval of final
version: all authors. All authors agree to be accountable for all aspects of the work
in ensuring that questions related to the accuracy or integrity of any part of the work
are appropriately investigated and resolved.
Funding  REMEDY was funded principally by the Canadian Network and Centre for
Trials Internationally (CANNeCTIN) to GK as part of the Clinical Research Initiative
of Canadian Institutes of Health Research. The other sources of funding were the
South African Medical Research Council, Lily and Ernst Hausmann Trust, the Else
Kroner Fresenius Foundation, the University of Cape Town, the National Research
Foundation of South Africa, Harold and Ethel Pupkewitz Heart Foundation (Namibia)
and the World Heart Federation. The Jos site was funded by the Jos University
Teaching Hospital, the Heart Aid Trust and Faith Alive Foundation. The Sudan sites
had partial funding from Sheikan Insurance. LZ was funded in part by the Discovery
Foundation, a US National Institutes of Health Fogarty International Clinical
Research Fellowship, Thrasher Research Fund Early Career Award, Wellcome Trust
Clinical Infectious Disease Research Initiative (CIDRI) Research Officer Award, the
Hamilton Naki Clinical Scholarship and by Medtronic Foundation through support
to Rheumatic Heart Disease Action. SY is funded by the Marion Burke Chair of the
Heart and Stroke Foundation of Canada.
Competing interests  None declared.
Patient consent  Not required.
Karthikeyan G, et al. Heart 2018;0:1–7. doi:10.1136/heartjnl-2018-313614
View publication stats
Valvular heart disease
Ethics approval  Ethics committees of all participating sites approved the study
Heart: first published as 10.1136/heartjnl-2018-313614 on 12 September 2018. Downloaded from http://heart.bmj.com/ on 29 September 2018 by guest. Protected by copyright.
protocol.
Provenance and peer review  Not commissioned; externally peer reviewed.
References
1 Watkins DA, Johnson CO, Colquhoun SM, et al. Global, regional, and national burden
of rheumatic heart disease, 1990-2015. N Engl J Med 2017;377:713–22.
2 Zühlke L, Engel ME, Karthikeyan G, et al. Characteristics, complications, and gaps in
evidence-based interventions in rheumatic heart disease: the Global Rheumatic Heart
Disease Registry (the REMEDY study). Eur Heart J 2015;36:1115–22.
3 Zühlke L, Karthikeyan G, Engel ME, et al. Clinical outcomes in 3343 children and
adults with rheumatic heart disease from 14 low- and middle-income countries: two-
year follow-up of the global rheumatic heart disease registry (the REMEDY Study).
Circulation 2016;134:1456–66.
4 Nishimura RA, Otto CM, Bonow RO, et al. AHA/ACC guideline for the management
of patients with valvular heart disease: a report of the American College of
Cardiology/American Heart Association Task Force on Practice Guidelines. Circulation
2014;2014:e521–643.
5 Vahanian A, Alfieri O, Andreotti F, et al. Guidelines on the management of valvular
heart disease (version 2012). Eur Heart J 2012;33:2451–96.
6 Ahuja RC, Sinha N, Saran RK, et al. Digoxin or verapamil or metoprolol for heart rate
control in patients with mitral stenosis--a randomised cross-over study. Int J Cardiol
1989;25:325–31.
7 Digitalis Investigation Group. The effect of digoxin on mortality and morbidity in
patients with heart failure. N Engl J Med 1997;336:525–33.
8 Washam JB, Stevens SR, Lokhnygina Y, et al. Digoxin use in patients with atrial
fibrillation and adverse cardiovascular outcomes: a retrospective analysis of the
Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared with Vitamin K
Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation (ROCKET
AF). Lancet 2015;385:2363–70.
9 Vamos M, Erath JW, Hohnloser SH. Digoxin-associated mortality: a systematic review
and meta-analysis of the literature. Eur Heart J 2015;36:1831–8.
10 Karthikeyan G, Zühlke L, Engel M, et al. Rationale and design of a Global Rheumatic
Heart Disease Registry: the REMEDY study. Am Heart J 2012;163:535–40.
11 Brookhart MA, Schneeweiss S, Rothman KJ, et al. Variable selection for propensity
score models. Am J Epidemiol 2006;163:1149–56.
12 Brookhart MA, Wyss R, Layton JB, et al. Propensity score methods for confounding
control in nonexperimental research. Circ Cardiovasc Qual Outcomes 2013;6:604–11.
13 Austin PC. An introduction to propensity score methods for reducing the effects of
confounding in observational studies. Multivariate Behav Res 2011;46:399–424.
14 Garrido MM, Kelley AS, Paris J, et al. Methods for constructing and assessing
propensity scores. Health Serv Res 2014;49:1701–20.
15 Vora A, Karnad D, Goyal V, et al. Control of rate versus rhythm in rheumatic atrial
fibrillation: a randomized study. Indian Heart J 2004;56:65–73.
16 Ouyang AJ, Lv YN, Zhong HL, et al. Meta-analysis of digoxin use and risk of mortality
in patients with atrial fibrillation. Am J Cardiol 2015;115:901–6.
17 Ziff OJ, Lane DA, Samra M, et al. Safety and efficacy of digoxin: systematic review and
meta-analysis of observational and controlled trial data. BMJ 2015;351:h4451.
18 Bavendiek U, Aguirre Davila L, Koch A, et al. Assumption versus evidence: the case of
digoxin in atrial fibrillation and heart failure. Eur Heart J 2017;38:ehw577.
19 Yusuf S, Wittes J, Bailey K, et al. Digitalis--a new controversy regarding an old drug.
The pitfalls of inappropriate methods. Circulation 1986;73:14–18.
20 Hallberg P, Lindbäck J, Lindahl B, et al. Digoxin and mortality in atrial fibrillation: a
prospective cohort study. Eur J Clin Pharmacol 2007;63:959–71.
21 Whitbeck MG, Charnigo RJ, Khairy P, et al. Increased mortality among patients taking
digoxin--analysis from the AFFIRM study. Eur Heart J 2013;34:1481–8.
22 Lopes RD, Rordorf R, De Ferrari GM, et al. Digoxin and mortality in patients with atrial
fibrillation. J Am Coll Cardiol 2018;71:1063–74.
7