J A C C : H E A R T F A I L U R E V O L . 6 , N O . 2 , 2 0 1 8

ª 2 0 1 8 B Y T H E A M E R I C A N C O L L E G E O F C A R D I O L O G Y F O U N D A T I O N

P U B L I S H E D B Y E L S E V I E R

CLINICAL RESEARCH

Low-Dose Aspirin in Heart FailureNot Complicated by Atrial Fibrillation

A Nationwide Propensity-Matched Study

Christian Madelaire, MD,a Gunnar Gislason, MD, PHD,a,b,c Søren L. Kristensen, MD, PHD,d Emil L. Fosbøl, MD, PHD,e

Jenny Bjerre, MD,a Maria D’Souza, MD,a Finn Gustafsson, MD, PHD, DMSC,e Lars Kober, MD, DMSC,e

Christian Torp-Pedersen, MD, DMSC,f,g Morten Schou, MD, PHDa

JACC: HEART FAILURE CME/MOC

This article has been selected as the month’s JACC: Heart Failure

CME/MOCactivity, availableonlineat http://www.acc.org/jacc-journals-cme

by selecting the JACC Journals CME/MOC tab.

Accreditation and Designation Statement

The American College of Cardiology Foundation (ACCF) is accredited by

the Accreditation Council for Continuing Medical Education (ACCME) to

provide continuing medical education for physicians.

The ACCF designates this Journal-based CME/MOC activity for a

maximum of 1 AMA PRA Category 1 Credit(s). Physicians should only

claim credit commensurate with the extent of their participation in

the activity.

Method of Participation and Receipt of CME/MOC Certificate

To obtain credit for JACC: Heart Failure CME/MOC, you must:

1. Be an ACC member or JACC subscriber.

2. Carefully read the CME/MOC-designated article available online and

in this issue of the journal.

3. Answer the post-test questions. At least 2 out of the 3 questions

provided must be answered correctly to obtain CME/MOC credit.

4. Complete a brief evaluation.

5. Claim your CME/MOC credit and receive your certificate electronically

by following the instructions given at the conclusion of the activity.

CME/MOC Objective for This Article: Upon completion of this activity, the

learner should be able to: 1) identify the increased risk of myocardial

infarction and heart failure hospitalization among patients with a first

diagnosis of heart failure and without a history of atrial fibrillation who

ISSN 2213-1779/$36.00

From the aDepartment of Cardiology, Cardiovascular Research Center, G

Denmark; bNational Institute of Public Health, University of Southern Denm

miology and Outcomes Research, The Danish Heart Foundation, Copenhage

Hospital, Copenhagen, Denmark; eDepartment of Cardiology, Rigshospital

Science, and Technology, Aalborg University, Aalborg, Denmark; and th

Biostatistics, Aalborg University Hospital, Aalborg, Denmark. Dr. Torp-Peder

Bayer. Dr. D’Souza has received grants from the Danish Heart Foundation

reported that they have no relationships relevant to the contents of this pap

Manuscript received August 18, 2017; revised manuscript received Septemb

are prescribed aspirin, compared to those who were not prescribed

aspirin; 2) compare the reduced risk of the composite outcome among

patients who are on beta-blockers or statin at baseline (vs. those without

beta-blockers at baseline) and those prescribed aspirin; and 3) recognize

that while the data presented here describes association, it does not

imply causation, which must be determined through a well-conducted

randomized controlled clinical trial.

CME/MOC Editor Disclosure: Editor-in-Chief ChristopherM.O’Connor,MD,

has received consultant fees/honoraria from AbbVie, Inc., Actelion

Pharmaceuticals Ltd., Bayer, Bristol-Myers Squibb, Cardiorentis, Merck &

Co., Inc., ResMed, and Roche Diagnostics; and has ownership interest in

Biscardia, LLC. Executive Editor Mona Fiuzat, PharmD, has received

research support from ResMed, Gilead, Critical Diagnostics, Otsuka, and

Roche Diagnostics. Tariq Ahmad, MD, MPH, has received a travel scholar-

ship fromThoratec.AbhinavSharma,MD,has receivedsupport fromBayer-

CanadianCardiovascular Society,Alberta InnovatesHealth Solution, Roche

Diagnostics, andTakeda.Mitchell Psotka,MD,PhD, andKishanParikh,MD,

have no relationships relevant to the contents of this paper to disclose.

Author Disclosures: Dr. Torp-Pedersen has received grants and speaker

honoraria from Bayer. Dr. D’Souza has received grants from the Danish

Heart Foundation and the VELUX Foundation. All other authors have

reported that they have no relationships relevant to the contents of this

paper to disclose.

Medium of Participation: Print (article only); online (article and quiz).

CME/MOC Term of Approval

Issue date: February 2018

Expiration date: January 31, 2019

https://doi.org/10.1016/j.jchf.2017.09.021

entofte and Herlev University Hospital, Hellerup,

ark, Copenhagen, Denmark; cCardiovascular Epide-

n, Denmark; dDepartment of Cardiology, Bispebjerg

et, Copenhagen, Denmark; fDepartment of Health,

e gDepartment of Cardiology and Epidemiology/

sen has received grants and speaker honoraria from

and the VELUX Foundation. All other authors have

er to disclose.

er 14, 2017, accepted September 19, 2017.

J A C C : H E A R T F A I L U R E V O L . 6 , N O . 2 , 2 0 1 8 Madelaire et al.F E B R U A R Y 2 0 1 8 : 1 5 6 – 6 7 Aspirin and Heart Failure

157

Low-Dose Aspirin in Heart FailureNot Complicated by Atrial Fibrillation

A Nationwide Propensity-Matched Study

Christian Madelaire, MD,a Gunnar Gislason, MD, PHD,a,b,c Søren L. Kristensen, MD, PHD,d Emil L. Fosbøl, MD, PHD,e

Jenny Bjerre, MD,a Maria D’Souza, MD,a Finn Gustafsson, MD, PHD, DMSC,e Lars Kober, MD, DMSC,e

Christian Torp-Pedersen, MD, DMSC,f,g Morten Schou, MD, PHDa

ABSTRACT

OBJECTIVES This study sought to assess safety and effectiveness of low-dose aspirin in heart failure (HF) not

complicated by atrial fibrillation.

BACKGROUND Despite lack of evidence, low-dose aspirin is widely used in patients with HF and sinus rhythm with and

without prior ischemic heart disease.

METHODS The study included 12,277 patients with new-onset HF during 2007 to 2012 who had no history of atrial

fibrillation. Of 5,450 patients using low-dose aspirin at baseline, 3,840 were propensity matched to non-aspirin

users in a 1:1 ratio. Propensity-matched Cox models were calculated with respect to the primary composite outcome

of all-cause mortality, myocardial infarction, and stroke and the secondary outcomes of bleeding and HF

readmission.

RESULTS The composite outcome occurred in 1,554 (40.5%) patients in the aspirin group and 1,604 (41.8%) patients

in the non-aspirin group. Aspirin use was not associated with an altered risk of composite outcome (hazard ratio [HR]:

0.98; 95% confidence interval [CI]: 0.91 to 1.05), but it was associated with an increased risk of myocardial infarction

(HR: 1.34; 95% CI: 1.08 to 1.67), whereas no differences were observed in all-cause mortality and stroke. An increased

risk of HF readmission was observed in the aspirin group (HR: 1.25; 95% CI: 1.17 to 1.33). No difference in bleeding

was observed. In subgroup analyses on the basis of a history of ischemic heart disease, the results were similar to the

main result.

CONCLUSIONS No association was detected between low-dose aspirin use and the composite outcome of all-cause

mortality, admission for myocardial infarction, and admission for stroke in patients with HF with no history of atrial

fibrillation. Aspirin use was associated with an increased risk of readmission for HF. (J Am Coll Cardiol HF 2018;6:156–67)

© 2018 by the American College of Cardiology Foundation.

H eart failure (HF) is considered to be a pro-thrombotic condition (1,2), and antithrom-botic therapy, in particular with aspirin, is

frequently used in ischemic HF and may also be usedin nonischemic HF. The effect of aspirin taken for ashort duration after an MI is well proven (3), whereasthe effect of long-term use in ischemic heart disease(IHD) is undocumented (4). Similarly, the frequentuse of aspirin in patients with HF is not backed bytrial evidence and is thus controversial (5).

In clinical randomized trials, aspirin was testedagainst placebo, warfarin, and clopidogrel in 4,368patients with HF with sinus rhythm (6–9). Only 1study included a no-antithrombotic therapy arm (6).The study was underpowered and found no

significant differences among aspirin, warfarin, andno antithrombotic therapy, although there was atrend toward an increased risk of the composite ofall-cause mortality and cardiovascular hospitaliza-tion in the aspirin arm. In the 2 other studies, nosignificant differences in all-cause and cardiovascu-lar events were found among aspirin, warfarin, andclopidogrel. The safety and efficacy of long-term useof low-dose aspirin in patients with HF are thereforeunknown. In contrast to these clinical trials, in amore recent, relatively small (n ¼ 1,476) retrospec-tive cohort study of unselected patients with HF,low-dose aspirin (75 mg/day) was associated with asignificantly reduced mortality risk (10). A largernumber of patients is needed before any conclusions

ABBR EV I A T I ON S

AND ACRONYMS

AF = atrial fibrillation

ATC = Anatomical Therapeutic

Chemical

CI = confidence interval

HF = heart failure

HR = hazard ratio

ICD = International

Classification of Diseases

ICD-10 = International

Classification of Diseases-

10th Revision

IHD = ischemic heart disease

MI = myocardial infarction

Madelaire et al. J A C C : H E A R T F A I L U R E V O L . 6 , N O . 2 , 2 0 1 8

Aspirin and Heart Failure F E B R U A R Y 2 0 1 8 : 1 5 6 – 6 7

158

can be drawn with respect to major adverseevents such as bleeding (11) and HF admis-sions for fluid retention induced by changesin renal function (12). An ongoing trial iscomparing rivaroxaban to placebo in pa-tients with HF and sinus rhythm andreduced ejection fraction (13). Ideally, theresults will contribute to an understandingof whether antithrombotic therapy has aplace in the treatment of HF.

SEE PAGE 168

With all study results taken together, therole of aspirin in patients with HF remainsunclear in terms of both benefit and safety.

The aim of the present study was to evaluate the as-sociation between the long-term use of aspirin(75 mg/day) and the risk of the composite endpoint ofdeath, myocardial infarction (MI), and stroke, as wellas the risk of hospital admission for bleeding oradmission for HF in a large, real-life unselected pop-ulation of patients with HF with no history of atrialfibrillation (AF).

METHODS

DATA SOURCES. This was a retrospective, registry-based cohort study using nationwide Danish regis-tries. In Denmark, all citizens are provided with apersonal registration number at birth or immigration,and this number allows us to follow a person amongdifferent registries. In the present study data wereobtained from the National Patient Registry (14), theDanish Registry of Medicinal Product Statistics (theNational Prescription Registry) (15), and the DanishCivil Registry (16).

Although low-dose aspirin can be bought inDenmark without a prescription, the proportion soldby prescription in the general population hasincreased over the years. Since 2007, the percentagehas been 90% or more (17).

PATIENT POPULATION. The study cohort consistedof all Danish residents >18 years of age who wereadmitted to a hospital in the period 2007 to 2012 witha first diagnosis of HF (International Classification ofDiseases-10th Revision [ICD-10] codes I50, I42, I110,J819) and no history of AF, who were still alive at thestudy’s start, 90 days after discharge. We identifiedall discharged patients who claimed a prescriptionfor low-dose (75/150 mg/day) aspirin (AnatomicalTherapeutic Chemical [ATC] code B01AC06) beforethe study’s start, whether or not these patients had ahistory of previous aspirin use. Starting the study 90days (3 months) after discharge was considered

optimal because aspirin was typically prescribed inpackages of 100 pills. By doing this, it avoided thepossibility that some members of the study popula-tion, taking aspirin, could carry a prescription frombefore the initial admission over into the studyperiod and accidentally appear as non-aspirin users.Severity of HF was estimated on the basis of dailydosage of loop diuretic, as described previously (18).Comorbidities of interest were IHD, cerebrovasculardisease, chronic renal failure, chronic obstructivepulmonary disease, and gastric ulcer, which wereidentified from ICD diagnoses, and diabetes, whichwas identified from ATC codes for antidiabetic med-ications. Comorbidities were considered to be presentif a patient had at least 1 hospital or outpatientcontact (or drug prescription for diabetes) in the past5 years before the index hospitalization. Baselinemedical therapy was defined as collected pre-scriptions in the past 6 months before the indexhospitalization. All ICD-10 and ATC codes used arelisted in the Online Tables 1 and 2.

EXCLUSION CRITERIA. History of AF was defined asan AF diagnosis (ICD-10 code I48) or other cardiacdysrhythmia (ICD-10 codes I46, I49). Thus, patientswith diagnosed AF and/or other cardiac dysrhythmiasand patients who had a permanent pacemakerimplanted before or during the initial hospital admis-sionwere excluded; patients treatedwith amiodarone,class 1c antiarrhythmic drugs, and sotalol were alsoexcluded. Further, we excluded all patients receivingwarfarin because they could bemisclassified as non-AFpatients if they received warfarin as thrombus pro-phylaxis for AF but without an AF diagnosis. Patientswith prosthetic heart valves and patients using otheranticoagulant agentsor antithrombotic treatmentwereexcluded.

FOLLOW-UP AND STUDY OUTCOMES. The primaryoutcome was the composite of all-cause mortality,admission for MI, or admission for stroke, whichevercame first. Secondary outcomes were the individualcomponents of the primary composite outcome,hospital admissions for bleeding events, and hospitalreadmission for HF. The study population was fol-lowed until death, emigration, or the end of study,whichever came first.

STATISTICS. Differences in baseline characteristicswere tested with chi-square and Kruskal-Wallis testsfor categorical and continuous variables, respec-tively. Cox proportional hazard models were used tomodel time to event for each outcome. Interactionanalyses were conducted considering age, sex, year ofinclusion, HF severity, comorbidities, and concomi-tant medical treatment at baseline.

TABLE 1 Patients’ Baseline Characteristics

Unmatched Propensity-Matched

Total(N ¼ 12,277)

Aspirin(n ¼ 5,450)

Non-Aspirin(n ¼ 6,827) p Value

Aspirin(n ¼ 3,840)

Non-Aspirin(n ¼ 3,840) p Value

Demographics

Age, yrs 76 (65–84) 77 (67–85) 74 (62–84) <0.001 78 (67–85) 78 (67–85) 0.82

Female 6,306 (51.4) 2,632 (48.3) 3,674 (53.8) <0.001 1,995 (52.0) 2,039 (53.1) 0.31

Comorbidities

IHD 3,412 (27.8) 2,229 (40.9) 1,183 (17.3) <0.001 1,047 (27.3) 1,047 (27.3) 1.00

Previous MI 1,051 (8.6) 744 (13.7) 307 (4.5) <0.001 290 (7.6) 280 (7.3) 0.66

CVD 717 (5.8) 323 (5.9) 394 (5.8) 0.72 250 (6.5) 266 (6.9) 0.47

CRF 506 (4.1) 218 (4.0) 288 (4.2) 0.55 160 (4.2) 174 (4.5) 0.43

COPD 2,520 (20.5) 1,013 (18.6) 1,507 (22.1) <0.001 774 (20.2) 812 (21.2) 0.28

Diabetes 2,189 (17.8) 1,174 (21.5) 1,015 (14.9) <0.001 742 (19.3) 770 (20.1) 0.42

Gastric ulcer 657 (5.4) 239 (4.4) 418 (6.1) <0.001 177 (4.6) 178 (4.6) 0.96

Concomitant medical treatment

Previous aspirin use 5,326 (43.4) 3,905 (71.7) 1,421 (20.8) <0.001 2,695 (70.2) 1,153 (30.0) <0.001

ACE inhibitor/ARB 7,707 (62.8) 3,907 (71.7) 3,800 (55.7) <0.001 2,560 (66.7) 2,566 (66.8) 0.88

BB 6,033 (49.1) 3,289 (60.4) 2,744 (40.2) <0.001 2,015 (52.5) 1,989 (51.8) 0.55

MRA 2,621 (21.4) 1,272 (23.3) 1,349 (19.8) <0.001 829 (21.6) 830 (21.6) 0.98

Digoxin 697 (5.7) 380 (7.0) 317 (4.6) <0.001 261 (6.8) 252 (6.6) 0.68

Loop 8,657 (70.5) 3,977 (73.0) 4,680 (68.6) <0.001 2,784 (72.5) 2,814 (73.3) 0.44

Thiazid 2,926 (23.8) 1,353 (24.8) 1,573 (23.0) 0.02 955 (24.9) 978 (25.5) 0.55

Statins 5,162 (42.1) 3,220 (59.1) 1,942 (28.5) <0.001 1,792 (46.7) 1,798 (46.8) 0.89

Values are median (interquartile range) or n (%).

ACE ¼ angiotensin-converting enzyme; ARB ¼ angiotensin receptor blocker; BB ¼ beta-blocker; COPD ¼ chronic obstructive pulmonary disease; CRF ¼ chronic renal failure;CVD¼ cerebrovascular disease; IHD¼ ischemic heart disease; Loop¼ loop diuretic; MI ¼myocardial infarction; MRA¼mineralocorticoid receptor antagonist; Thiazid¼ thiazidediuretic.

J A C C : H E A R T F A I L U R E V O L . 6 , N O . 2 , 2 0 1 8 Madelaire et al.F E B R U A R Y 2 0 1 8 : 1 5 6 – 6 7 Aspirin and Heart Failure

159

The models fulfilled the proportional hazardassumption and linearity of continuous variables. Allp values reported were 2-sided, and p < 0.05 wasconsidered statistically significant. The p values oncumulative incidences with competing risk calcula-tions were calculated using the Fine and Graymethod. Statistical calculations were performed withSAS software version 9.4 (SAS Institute, Inc., Cary,North Carolina) and RStudio version 00.99.903(R Foundation, Vienna, Austria).

PROPENSITY-MATCHED ANALYSES. We carried outa propensity-matched analysis, using the GreedyMatch Macro (Mayo Clinic, Rochester, Minnesota) tomatch the patients in the aspirin group with non-users on the basis of propensity score. We calcu-lated the propensity of receiving aspirin before 90days after discharge from hospital (study start),considering sex, age, calendar year, HF severity,comorbidities, and concomitant medical treatment aslisted in Table 1, but not previous use of aspirin. Thepropensity score was calculated using logisticregression, and matching was carried out in sub-groups on the basis of IHD and age groups, with amaximum distance equivalent to one-fourth standarddeviation. After merging subgroups, 3,840 patientsusing aspirin were matched to 1 control each. The

distribution of propensity score was equal in theaspirin and non-aspirin groups (Kruskal-Wallis,p ¼ 0.87). The patients in the propensity-matchedpopulation were older than the overall study popu-lation, and IHD- and HF-related medications wereslightly more frequent. The matched aspirin andnon-aspirin groups appeared to be similar in terms ofsex, comorbidities, and concomitant medical therapyafter propensity matching, thereby supporting thematching being successful.

RESULTS

PATIENTS’ CHARACTERISTICS. From 2007 through2012, 59,838 patients ($18 years of age) had a diag-nosis of HF during a hospital admission. During theinitial hospital stay, 5,701 patients died. Afterexcluding 34,902 patients because of a history ofarrhythmia and/or anticoagulant therapy, 7,277 pa-tients had an event before inclusion, 90 days subse-quent to discharge. Among those events were 2,929primary composite outcomes, including 2,616 deaths,4,627 HF readmissions, and 126 bleeding events. Atotal of 12,277 patients met the inclusion criteria, and5,450 of those patients were treated with aspirin atbaseline, as shown in Figure 1.

FIGURE 1 Flowchart Showing Selection of Patients for the Main Study

HF ¼ heart failure; MI ¼ myocardial infarction.

Madelaire et al. J A C C : H E A R T F A I L U R E V O L . 6 , N O . 2 , 2 0 1 8

Aspirin and Heart Failure F E B R U A R Y 2 0 1 8 : 1 5 6 – 6 7

160

Previous use of aspirin was common and was muchmore frequent in the aspirin group (72%) than in thenon-aspirin group (21%). The propensity-matchedanalyses comprised 7,680 patients, as describedearlier. The patients were followed for a median of 1.7years. Baseline characteristics are presented inTable 1.

STUDY OUTCOMES, UNMATCHED POPULATION. Atotal of 4,817 patients (39.2%) met the compositeoutcome of all-cause mortality, MI, or stroke duringfollow-up. Of these patients, 2,160 were using aspirin,thereby yielding an unadjusted hazard ratio (HR) of1.03 with a 95% confidence interval (CI) of 0.97 to 1.09.

The components all-cause mortality and strokeappeared neutral, whereas aspirin use was associatedwith an increased risk of MI (HR: 1.82; 95% CI: 1.52 to2.19). A total of 3,582 patients had at least 1 HFreadmission (HR for aspirin use vs. nonuse: 1.25;95% CI: 1.17 to 1.33), and 750 patients had a bleedingevent (HR for aspirin use vs. nonuse: 1.18; 95% CI:1.02 to 1.36). Primary and secondary outcomes arelisted in detail in Table 2.

STUDYOUTCOMES,PROPENSITY-MATCHEDANALYSES. Inthe propensity-matched population, 3,158 patients(41.1%) met the composite outcome of all-cause

mortality, MI, or stroke (HR: 0.98; 95% CI: 0.91 to1.05). We observed no significant differences for theindividual components all-cause mortality andstroke. However, we observed an association betweenaspirin use and increased risk of MI (HR: 1.34; 95% CI:1.08 to 1.67). Of the 321 MI events, only 44 (13.7%)were described as ST-segment elevation MI, whereas170 (53.0%) were described as non-ST-segmentelevation. However, 107 (33.3%) were unspecified.The primary outcomes are shown in Table 2 andFigure 2.

We observed a significantly increased risk of HFreadmission in the aspirin-treated group (HR: 1.10;95% CI: 1.01 to 1.19). When considering death as acompeting risk, the 5-year cumulative incidence ofHF readmissions was 41.3% in the aspirin group and35.1% in the non-aspirin group (p ¼ 0.024). No asso-ciation between aspirin and an altered risk ofbleeding was observed. The results of the propensity-matched analyses are shown in Table 2 and Figure 3.

SUBGROUP ANALYSES, PROPENSITY-MATCHED

ANALYSES. Subgroup analyses were conducted forthe primary composite outcome considering age, sex,IHD, and diabetes. The results were similar to the mainresults in all subgroups. In patientswith history of IHD,

TABLE 2 Primary and Secondary Outcomes for the Unmatched Population and the

Matched Subpopulation

Aspirin(n ¼ 5,450)

Non-Aspirin(n ¼ 6,827)

Hazard Ratio(95% CI) p Value

Unmatched unadjusted results

Primary composite outcome

Death from any cause or firstadmission for MI or stroke

2,160 (39.6) 2,657 (38.9) 1.03 (0.97–1.09) 0.33

Death from any cause 1,918 (35.2) 2,412 (35.3) 0.99 (0.93–1.05) 0.71

First admission for MI 284 (5.2) 197 (2.9) 1.82 (1.52–2.19) <0.001

First admission for stroke 217 (4.0) 254 (3.7) 1.06 (0.89–1.28) 0.50

Secondary outcomes

First HF readmission 1,755 (32.2) 1,827 (26.8) 1.25 (1.17–1.33) <0.001

First admission for bleeding 365 (6.7) 385 (5.6) 1.18 (1.02–1.36) 0.02

Propensity-matched results

Primary composite outcome

Death from any cause or firstadmission for MI or stroke

1,554 (40.5) 1,604 (41.8) 0.98 (0.91–1.05) 0.53

Death from any cause 1,400 (36.5) 1,441 (37.5) 0.98 (0.91–1.05) 0.51

First admission for MI 183 (4.8) 138 (3.6) 1.34 (1.08–1.67) 0.009

First admission for stroke 151 (3.9) 164 (4.3) 0.92 (0.74–1.15) 0.47

Secondary outcomes

First HF readmission 1,181 (30.8) 1,103 (28.7) 1.10 (1.01–1.19) 0.03

First admission for bleeding 252 (6.6) 224 (5.8) 1.13 (0.94–1.35) 0.18

Values are n (%), unless otherwise indicated.

CI ¼ confidence interval; HF ¼ heart failure; MI ¼ myocardial infarction.

J A C C : H E A R T F A I L U R E V O L . 6 , N O . 2 , 2 0 1 8 Madelaire et al.F E B R U A R Y 2 0 1 8 : 1 5 6 – 6 7 Aspirin and Heart Failure

161

the risk of the composite outcome was an HR of 0.91with a 95% CI of 0.79 to 1.04, using nonusers asreference. The tendency in this subgroup was toward alower risk (or greater benefit) associated with aspirinuse, although it was not significant. The results of thesubgroup analyses are illustrated in Figure 4.

INTERACTION ANALYSES. Analyses assessing in-teractions with the propensity-matched results werecarried out for the primary composite outcome withrespect to age, sex, calendar year, diabetes, IHD, andHF severity, estimated by loop diuretic dosages, andall drugs listed in Table 1. We observed a significantinteraction between previous aspirin use and theprimary outcome (p ¼ 0.04), with a significantlyreduced risk in the aspirin group compared with thecontrol subjects among the patients who were aspirinnaïve at baseline. However, the age distributions inthe 2 aspirin-naïve groups were uneven. To addressthis issue, we conducted a matched sensitivity anal-ysis (see later). We further observed a significantinteraction with HF severity (p ¼ 0.03) becausepatients who received #40 mg/day loop diuretics hada reduced risk of the primary outcome if they were inthe aspirin group (HR: 0.78; 95% CI: 0.64 to 0.94),driven by a significantly reduced risk of stroke (HR:0.56; 95% CI: 0.35 to 0.89). We observed a significantinteraction with beta-blockers (BB) and statins,meaning that patients using these drugs had areduced risk of reaching the primary compositeoutcome if they were in the aspirin group. This wasdriven by reduced risk of all-cause mortality andstroke, whereas the risk of MI remained significantlyhigher among the aspirin users. Drug-drug in-teractions with the primary outcome are illustrated inFigure 5A, whereas interactions with the secondaryoutcome HF readmissions are illustrated in Figure 5B.

SENSITIVITY ANALYSIS. Because a significant inter-action with previous aspirin use was observed, we con-ducted a propensity-matched sensitivity analysis onlyincluding naïve users of aspirin. This analysiscomprised 1,373 patients who started using aspirin atbaseline and 1,373 matched control subjects, with aneven age distribution in the 2 groups. HR for the pri-mary composite outcome was 1.08 with a 95% CI of0.96 to 1.22, and in general the results for the primaryand secondary outcomes were neutral and similar tothose of the primary study population.

We conducted a per protocol–type analysis inwhich patients from the aspirin group were excludedif they stopped therapy and patients from the controlgroup were excluded if they started using aspirin. Theresults were similar to the overall results, so eventhough a relatively large proportion of the patients in

the non-aspirin group crossed over, it had no effecton the main results.

Finally, a propensity-matched analysis with inclu-sion 7 days after primary hospital discharge wasconducted. It comprised 3,163 cases and 3,163 non-aspirin control subjects. No difference in the risk ofreaching the primary composite outcome wasobserved (HR: 0.92; 95% CI: 0.85 to 1.01). The sameapplied for the risk of all-cause mortality and MI.However, the risk of admission for stroke was reducedin the aspirin group (HR: 0.75; 95% CI: 0.57 to 0.98). Asin the main analyses, the risk of having an HF read-mission was slightly increased in the aspirin group(HR: 1.13; 95% CI: 1.01 to 1.25), whereas no difference inthe risk of bleeding was observed. Among the controlsubjects, 890 (28%) were included as aspirin users inthe main analysis. Of those aspirin users, 86% usedaspirin before the primary hospital admission.

DISCUSSION

PREVIOUS STUDIES. Our results differ from thosepresented by Bermingham et al. (10), who reportedthat low-dose aspirin was associated with a signifi-cantly reduced risk of death and HF readmission. Thestudy included AF, which was shown to be over-represented in the non-aspirin group. Naturally, thefrequency of warfarin use was also considerably

FIGURE 2 Propensity-Matched Cumulative Probabilities for the Primary Composite Outcome of All-Cause Mortality, MI, and Stroke and

the Individual Components of the Composite

(A to D) The curves represent the aspirin group (red line) and the non-aspirin group (dashed blue line). Hazard ratios are calculated with

propensity-matched Cox regression analyses. CI ¼ confidence interval; MI ¼ myocardial infarction.

Madelaire et al. J A C C : H E A R T F A I L U R E V O L . 6 , N O . 2 , 2 0 1 8

Aspirin and Heart Failure F E B R U A R Y 2 0 1 8 : 1 5 6 – 6 7

162

higher in this group—in fact, higher than the fre-quency of AF. The propensity model considered age,sex, IHD, AF, diabetes, hypertension, and dyslipide-mia, whereas factors such as concomitant medicaltherapy and severity of HF were left out. This mayhave influenced the results, and other factors such asyounger age, fewer patients, and number of eventsmay explain the difference between their results andours.

The results of the present study are in accord withthe results of WASH (Warfarin/Aspirin Study in Heartfailure) (6), in which aspirin was tested againstwarfarin and no antithrombotic treatment. The trialwas underpowered, and the primary purpose was toexamine whether it would be feasible to conduct a

larger trial. The trends, however, were that therewere no significant differences among the 3 groupswith respect to the primary composite outcome of all-cause mortality, nonfatal MI, and nonfatal stroke. In alarger trial, WATCH (Warfarin and Antiplatelet Ther-apy in Chronic Heart Failure) (7), aspirin wascompared with warfarin without a no-antithrombotictherapy arm. The study was terminated prematurelybecause of slow enrollment. However, the in-vestigators concluded that no significant differencesbetween aspirin and warfarin treatment were foundin terms of safety or effectiveness. The same resultswere found in WARCEF (Warfarin Versus Aspirin inReduced Cardiac Ejection Fraction) (8), which is thelargest trial conducted to address the issue until now,

FIGURE 3 Propensity-Matched Cumulative Incidence of Heart Failure Readmission

With 95% Confidence Intervals for the Aspirin and Non-Aspirin Groups, With Death

Considered Competing Risk Factor

The p value is calculated using the Fine and Gray method.

J A C C : H E A R T F A I L U R E V O L . 6 , N O . 2 , 2 0 1 8 Madelaire et al.F E B R U A R Y 2 0 1 8 : 1 5 6 – 6 7 Aspirin and Heart Failure

163

including only patients with sinus rhythm. None ofthe trials included a placebo arm, although there wasa no-antithrombotic therapy arm in WASH. All thesetrials comprised patients 60 to 65 years of age, mostlymale, and with a reduced ejection fraction, featuresthat stand in contrast to our real-life cohort of olderpatients with HF with a high proportion of femalepatients. The event rates in the present study werehigher than in the studies mentioned earlier (e.g.,16.78 deaths per 100 patient-years in the presentstudy compared with 6.57 deaths per 100 patient-years in WARCEF, most likely because of the agedifferences). In light of the low probability that anadequately powered trial evaluating long-term use ofaspirin in HF will ever be conducted, our analyses addknowledge and cast further doubt on whether aspirinis useful in HF, even in the presence of IHD.

We observed a significant interaction with BB. Inthe subgroup of patients using BB, aspirin was asso-ciated with a slightly reduced risk of meeting theprimary composite outcome, driven by all-causemortality and stroke. Considering internationalguidelines, BB users from our cohort could have been

FIGURE 4 Propensity-Matched Event Rates and Cox Regression Analyses Regarding Risk of the Primary Composite Outcome in Various

Subgroups

Findings are plotted as hazard ratios with 95% confidence intervals (CIs) for patients treated with aspirin. Hazard ratio 1 represents the

non-aspirin control subjects in the respective groups. IHD ¼ ischemic heart disease.

FIGURE 5 Drug-Drug Interactions

Drug-drug interactions and propensity-matched event rates and Cox regression analyses regarding risk of (A) the primary composite outcome

and (B) the secondary outcome of HF readmission in subgroups on the basis of concomitant medical therapy, plotted as hazard ratios with

95% confidence intervals (CIs) for patients treated with aspirin. A hazard ratio of 1.00 represents the non-aspirin control subjects in the

respective groups. ACE-I ¼ angiotensin-converting enzyme inhibitors; ARB ¼ angiotensin receptor blockers; BB ¼ beta-blockers; Loop ¼loop diuretics; MRA ¼ mineralocorticoid receptor antagonists; Thiazid ¼ thiazide diuretics.

Madelaire et al. J A C C : H E A R T F A I L U R E V O L . 6 , N O . 2 , 2 0 1 8

Aspirin and Heart Failure F E B R U A R Y 2 0 1 8 : 1 5 6 – 6 7

164

more likely to have reduced ejection fraction thanpatients who were not receiving BB therapy, and thisfinding may indicate that there could be a slightpositive effect in this subgroup of patients with HF.However, because we did not have access to

information about left ventricular ejection fraction,further investigation was not possible in this study.

METHODOLOGICAL CONSIDERATIONS. In the un-matched population of the present study, the patients

J A C C : H E A R T F A I L U R E V O L . 6 , N O . 2 , 2 0 1 8 Madelaire et al.F E B R U A R Y 2 0 1 8 : 1 5 6 – 6 7 Aspirin and Heart Failure

165

using aspirin were slightly older than the non-aspirincontrol subjects. The frequency of IHD was higherand the frequency of HF-associated medication use,including loop diuretics, was higher. The meandosage of loop diuretics did not differ. Because low-dose aspirin is indicated in patients with IHD, wewere surprised that only 65% of the patients withknown IHD were using aspirin at baseline.

There was, as expected, an overweight of previousaspirin use in the aspirin group. Previous use ofaspirin was not included in the propensity scorecalculation, although this probably could havecontributed to reduce the risk of confounding byindication (i.e., patients given aspirin were morelikely to have ischemia even though they did not havea diagnosis of ischemia). Conversely, patients with ahistory of ischemia but who were not given aspirincould represent a population that does not tolerateaspirin very well. However, it seems likely that itwould have introduced further selection bias and asmaller population. Furthermore, because there is nosolid evidence to support or reject aspirin use in HF,we would expect that if aspirin is prescribed once, itis likely that the treatment will be continued if thepatient is admitted to hospital, without furtherconsideration about indication. All considered, webelieve that we chose the lesser of 2 evils.

We observed a significantly increased risk ofHF readmission in the aspirin group. This could bepartly explained by impaired renal function causedby aspirin (12), which subsequently could contributeto salt and water retention and worsened HF. It haspreviously been suggested that aspirin could reducethe effect of angiotensin-converting enzyme in-hibitors when these drugs are used simultaneously(5,19). In the present study, we did not observe asignificant interaction with angiotensin-convertingenzyme inhibitors or angiotensin receptor blockers,but we observed a trend to no risk difference be-tween the aspirin and non-aspirin groups amongpatients not using angiotensin-converting enzymeinhibitors or angiotensin receptor blockers(Figure 5B). The increased risk is in accord with thetrends found in the WASH trial (6). Although our re-sults could be affected by residual confounding, itseems biologically plausible that aspirin could in-crease the risk of HF readmission. If so, on the basisof the event rates for HF readmission (1,181 [30.8%]in the aspirin group and 1,103 [28.7%] in thenon-aspirin group), the number needed to harm was1/(0.308 � 0.287) ¼ 48.

We observed a significantly increased risk ofadmission for MI in the aspirin group, although rela-tively few events occurred. A biologically plausible

explanation could be that aspirin may contribute tohemorrhage in existing plaques and thus contribute toprogression and instability of the plaque, which in-creases the risk of MI. However, at this extent, it isdifficult to reject that some level of confounding byindication and misclassification could contribute tothe observation (i.e., patients prescribed aspirin forsome reason have a higher frequency of undiagnosedor unregistered IHD). Moreover, it is not unlikely thatsome patients with a history of IHD or known aspirinuse aremore likely to have a diagnosis of MI when theyare admitted to the hospital with dyspnea andincreased troponin than are patients with no history ofIHD and/or no known aspirin use. In the IHD subgroup,comprising 2,094 patients, there was still a tendencytoward an increased risk of MI, although it was notsignificant (Figure 4).

We observed more bleeding events in the aspiringroup compared with the control group (n ¼ 252[6.6%] vs. n ¼ 224 [5.8%]), but the difference was notsignificant. However, for the safety outcome ofbleeding, we considered only hospital admissions,and therefore, minor bleeding events treated athome, in general practice, or in the emergencydepartment were not encountered. Thus, the actualevent rate was most likely higher.

Because we ended up including only 12,277 patientsof the 59,838 diagnosed with HF, it may be argued thatthe study population was selected and that the studyhas limited generalizability. However, the focus of thestudy was on patients with sinus rhythm and noconcomitant anticoagulant or antithrombotic therapy,and that focus cost the study 34,902 patients. Asdescribed earlier, to keep misclassification at a mini-mum, we had to postpone inclusion to 3 months afterthe index hospitalization. Unfortunately, this cost anadditional 8,338 patients because theymet an outcomewithin this period. It is likely that these patients wereamong the frailest, and thus we may have introducedsome survival or selection bias. However, this wasthought to be equal in both groups, which a sensitivityanalysis further supported.

STUDY STRENGTHS. Strengths of this study are thesample size and the completeness of data, with anationwide unselected cohort of patients with HF. InDenmark, all residents, regardless of socioeconomicstatus, health insurance, and geography, are coveredby a public health care system, ensuring completeregistration of outpatient contacts and hospital ad-missions. The registries used in the present studyhave previously been validated as tools for epidemi-ological research (14–16). Identifying patients with HFin the National Patient Registry has been shown to

PERSPECTIVES

COMPETENCY IN MEDICAL KNOWLEDGE: Use

of daily low-dose aspirin in patients with HF and no

history of AF and no other antiplatelet or anticoagu-

lant therapy was not associated with reduced risk of

death, MI, or stroke.

TRANSLATIONAL OUTLOOK: Because low-dose

aspirin is widely used in patients with HF regardless of

knownAFor IHD andwithout solid evidence supporting

this use, a randomized, blinded clinical trial is needed.

Madelaire et al. J A C C : H E A R T F A I L U R E V O L . 6 , N O . 2 , 2 0 1 8

Aspirin and Heart Failure F E B R U A R Y 2 0 1 8 : 1 5 6 – 6 7

166

have a very high specificity of 99%, although sensi-tivity was lower (20). Because part of the drug ex-penses in Denmark are reimbursed by thegovernment-financed health care system, all Danishpharmacies are required to register each dispensedprescription in the National Prescription Registry,which ensures complete and precise registration. Thediagnoses of interest, MI and stroke, have previouslybeen found to have high positive predictive values inthe National Patient Registry (21,22).

STUDY LIMITATIONS. A noticeable limitation of thestudy is the sparse clinical data provided by theregistries. This limitation introduces potentialmisclassification and unmeasured and residualconfounding. It was not possible to assess informa-tion about ejection fraction or New York Heart As-sociation functional classification. This was partiallycompensated for by collecting the dosages of loopdiuretics and thereby determining relative HFseverity (18). Fifty percent of the patients were fe-male, and we may therefore have a relatively highpercentage of patients with HF with preserved ejec-tion fraction. In a Danish registry including patientswith HF with both sinus rhythm and AF (23), 40.4%of the patients had preserved ejection fraction, andthis is also likely to be the case in the present study.Lacking electrocardiography data, sinus rhythm wasdefined as the absence of admissions or outpatientcontact for AF or other arrhythmic diseases, as wellas the absence of relevant treatment (warfarin). TheAF diagnosis has previously been validated with ahigh positive predictive value in the Danish regis-tries (24). Finally, it may be argued that theincreased risk of an HF admission is the result ofconfounding by indication, a concept supported byour observation of an increased risk of MI in theaspirin-treated group. However, the 2 groups of pa-tients had the same mortality risk and the same

degree of severity of HF evaluated by doses of loopdiuretics, and the observed association between low-dose aspirin and the risk of an HF admission is bio-logically plausible.

CONCLUSIONS

In the present study, an association between low-dose aspirin and a reduced risk of the compositeendpoint of death, MI, and stroke was not found, norwas an increased risk of bleeding observed. However,in these older adult patients with HF, a slightlyincreased risk of HF readmission was detected. Thefindings cast further doubt on the benefit associatedwith the use of aspirin in patients with HF. A pro-spective, randomized clinical trial could be relevant.

ADDRESS FOR CORRESPONDENCE: Dr. ChristianMadelaire, Department of Cardiology, CopenhagenUniversity Hospital, Herlev and Gentofte Cardiovas-cular Research Unit 1–Post 635, Kildegårdsvej 28,DK-2900 Hellerup, Denmark. E-mail: christian.madelaire.rasmussen@regionh.dk.

RE F E RENCE S

1. Palka I, Nessler J, Nessler B, Piwowarska W,Tracz W, Undas A. Altered fibrin clot properties inpatients with chronic heart failure and sinusrhythm: a novel prothrombotic mechanism. Heart2010;96:1114–8.

2. Gurbel PA, Tantry US. Antiplatelet and antico-agulant agents in heart failure: current status andfuture perspectives. J Am Coll Cardiol HF 2014;2:1–14.

3. Randomised trial of intravenous streptokinase,oral aspirin, both, or neither among 17,187 casesof suspected acute myocardial infarction: ISIS-2.ISIS-2 (Second International Study of InfarctSurvival) Collaborative Group. Lancet 1988;2:349–60.

4. Cleland JG. Is aspirin “the weakest link” incardiovascular prophylaxis? The surprising lack ofevidence supporting the use of aspirin for cardio-vascular disease. Prog Cardiovasc Dis 2002;44:275–92.

5. Cleland JG. Is aspirin useful in primary preven-tion? Eur Heart J 2013;34:3412–8.

6. Cleland JG, Findlay I, Jafri S, et al. The Warfarin/Aspirin Study in Heart failure (WASH): a random-ized trial comparing antithrombotic strategies forpatients with heart failure. Am Heart J 2004;148:157–64.

7. Massie BM, Collins JF, Ammon SE, et al.,WATCH trial investigators. Randomized trial of

warfarin, aspirin, and clopidogrel in patients withchronic heart failure: the Warfarin and AntiplateletTherapy in Chronic Heart Failure (WATCH) trial.Circulation 2009;119:1616–24.

8. Homma S, Thompson JLP, Pullicino PM, et al.,WARCEF Investigators. Warfarin and aspirin inpatients with heart failure and sinus rhythm.N Engl J Med 2012;366:1859–69.

9. Cokkinos DV, Haralabopoulos GC, Kostis JB,Toutouzas PK, HELAS investigators. Efficacy ofantithrombotic therapy in chronic heart failure:the HELAS study. Eur J Heart Fail 2006;8:428–32.

10. Bermingham M, Shanahan MK, O’Connell E,et al. Aspirin use in heart failure: is low-dose

J A C C : H E A R T F A I L U R E V O L . 6 , N O . 2 , 2 0 1 8 Madelaire et al.F E B R U A R Y 2 0 1 8 : 1 5 6 – 6 7 Aspirin and Heart Failure

167

therapy associated with mortality and morbiditybenefits in a large community population? CircHeart Fail 2014;7:243–50.

11. Derry S, Loke YK. Risk of gastrointestinal hae-morrhage with long term use of aspirin: meta-analysis. BMJ 2000;321:1183–7.

12. Segal R, Lubart E, Leibovitz A, et al. Early andlate effects of low-dose aspirin on renal functionin elderly patients. Am J Med 2003;115:462–6.

13. Zannad F, Greenberg B, Cleland JG, et al.Rationale and design of a randomized,double-blind, event-driven, multicentre studycomparing the efficacy and safety of oral rivar-oxaban with placebo for reducing the risk ofdeath, myocardial infarction or stroke in subjectswith heart failure and significant coronary arterydisease following an exacerbation of heart fail-ure: the COMMANDER HF trial. Eur J Heart Fail2015;17:735–42.

14. Schmidt M, Schmidt SA, Sandegaard JL,Ehrenstein V, Pedersen L, Sørensen HT. TheDanish National Patient Registry: a review ofcontent, data quality, and research potential. ClinEpidemiol 2015;7:449–90.

15. Kildemoes HW, Sørensen HT, Hallas J. TheDanish National Prescription Registry. Scand JPublic Health 2011;39 7 Suppl:38–41.

16. Schmidt M, Pedersen L, Sørensen HT.The Danish Civil Registration System as a tool

in epidemiology. Eur J Epidemiol 2014;29:541–9.

17. Schmidt M, Hallas J, Friis S. Potential ofprescription registries to capture individual-leveluse of aspirin and other nonsteroidal anti-inflammatory drugs in Denmark: trends in utili-zation 1999–2012. Clin Epidemiol 2014;6:155–68.

18. Eshaghian S, Horwich TB, Fonarow GC.Relation of loop diuretic dose to mortality inadvanced heart failure. Am J Cardiol 2006;97:1759–64.

19. Dagenais GR, Pogue J, Fox K, Simoons ML,Yusuf S. Angiotensin-converting-enzyme in-hibitors in stable vascular disease without leftventricular systolic dysfunction or heart failure: acombined analysis of three trials. Lancet 2006;368:581–8.

20. Kümler T, Gislason GH, Kirk V, et al. Ac-curacy of a heart failure diagnosis in adminis-trative registers. Eur J Heart Fail 2008;10:658–60.

21. Krarup LH, Boysen G, Janjua H, Prescott E,Truelsen T. Validity of stroke diagnoses in aNational Register of Patients. Neuroepidemiology2007;28:150–4.

22. Madsen M, Davidsen M, Rasmussen S,Abildstrom SZ, Osler M. The validity of thediagnosis of acute myocardial infarction in

routine statistics: a comparison of mortalityand hospital discharge data with the DanishMONICA registry. J Clin Epidemiol 2003;56:124–30.

23. Gustafsson F, Torp-Pedersen C, Brendorp B,Seibaek M, Burchardt H, Køber L, DIAMOND StudyGroup. Long-term survival in patients hospitalizedwith congestive heart failure: relation to preservedand reduced left ventricular systolic function. EurHeart J 2003;24:863–70.

24. RixTA,Riahi S,OvervadK,Lundbye-ChristensenS,Schmidt EB, Joensen AM. Validity of thediagnoses atrial fibrillation and atrial flutter in aDanish patient registry. Scand Cardiovasc J 2012;46:149–53.

KEY WORDS antithrombotic therapy,aspirin, heart failure, sinus rhythm

APPENDIX For supplemental tables, pleasesee the online version of this paper.

Go to http://www.acc.org/jacc-journals-cme totake the CME/MOC quizfor this article.