J A C C : H E A R T F A I L U R E V O L . 4 , N O . 1 , 2 0 1 6

ª 2 0 1 6 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 ND A T I O N I S S N 2 2 1 3 - 1 7 7 9 / $ 3 6 . 0 0

P U B L I S H E D B Y E L S E V I E R h t t p : / / d x . d o i . o r g / 1 0 . 1 0 1 6 / j . j c h f . 2 0 1 5 . 0 8 . 0 0 7

Impact of Dietary Sodium Restriction onHeart Failure Outcomes

Rami Doukky, MD, MSC,*yzx Elizabeth Avery, MS,*x Ashvarya Mangla, MD,*zx Fareed M. Collado, MD,zZeina Ibrahim, MD,y Marie-France Poulin, MD,z DeJuran Richardson, PHD,*xk Lynda H. Powell, PHD*x

JACC: HEART FAILURE CME

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

activity, available online at http://www.acc.org/jacc-journals-cme by

selecting the CME tab on the top navigation bar.

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 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 Certificate

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

1. Be an ACC member or JACC subscriber.

2. Carefully read the CME-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 credit.

4. Complete a brief evaluation.

5. Claim your CME credit and receive your certificate electronically by

following the instructions given at the conclusion of the activity.

CME Objective for This Article: After reading this article, the reader should

be able to discuss: 1) the clinical characteristics observed in association

with sodium restriction; 2) the association between sodium restriction

From the *Department of Preventive Medicine, Rush University Medical Cen

Stroger, Jr. Hospital of Cook County, Chicago, Illinois; zDivision of Cardiolog

xRush Center for Urban Health Equity, Rush University Medical Center, Chica

Computer Science, Lake Forest College, Lake Forest, Illinois. The Heart Fai

was funded by the National Heart, Lung, and Blood Institute (NHLBI) (HL065

Health Equity, which is funded by the NHLBI, grant number 1P50HL105189-

and received research funding from Astellas Pharma. All other authors have

the contents of this paper to disclose.

Manuscript received July 14, 2015; accepted August 6, 2015.

and clinical outcomes in chronic HF patients as observed in the present

propensity-matched analysis; and 3) the implications of these data

related to clinical practice and future research.

CME Editor Disclosure: Deputy Managing Editor Mona Fiuzat, PharmD,

FACC, has received research support from ResMed, Gilead, Critical Di-

agnostics, Otsuka, and Roche Diagnostics. Tariq Ahmad, MD, MPH, has

received a travel scholarship from Thoratec. Robert Mentz, MD, has

received a travel scholarship from Thoratec; research grants from Gilead;

research support from ResMed, Otsuka, Bristol-Myers Squibb, AstraZe-

neca, Novartis, and GlaxoSmithKline; and travel related to investigator

meetings from ResMed, Bristol-Myers Squibb, AstraZeneca, Novartis, and

GlaxoSmithKline. Adam DeVore, MD, has received research support from

the American Heart Association, Novartis Pharmaceuticals, Thoratec, and

Amgen.

Author Disclosures: The Heart Failure Adherence and Retention Trial

(NCT00018005) was funded by the National Heart, Lung, and Blood

Institute (NHLBI) (HL065547). This study is part of the Rush Center for

Urban Health Equity, which is funded by the NHLBI, grant number

1P50HL105189-01. Dr. Doukky has served on the advisory board for and

received research funding from Astellas Pharma. 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 Term of Approval

Issue date: January 2016

Expiration date: December 31, 2016

ter, Chicago, Illinois; yDivision of Cardiology, John H.

y, Rush University Medical Center, Chicago, Illinois;

go, Illinois; and the kDepartment of Mathematics and

lure Adherence and Retention Trial (NCT00018005)

547). This study is part of the Rush Center for Urban

01. Dr. Doukky has served on the advisory board for

reported that they have no relationships relevant to

J A C C : H E A R T F A I L U R E V O L . 4 , N O . 1 , 2 0 1 6 Doukky et al.J A N U A R Y 2 0 1 6 : 2 4 – 3 5 Sodium Restriction and HF Outcome

25

Impact of Dietary Sodium R

estriction onHeart Failure Outcomes

ABSTRACT

OBJECTIVES This study sought to evaluate the impact of sodium restriction on heart failure (HF) outcomes.

BACKGROUND Although sodium restriction is advised for patients with HF, data on sodium restriction and HF

outcomes are inconsistent.

METHODS We analyzed data from the multihospital HF Adherence and Retention Trial, which enrolled 902 New York

Heart Association functional class II/III HF patients and followed them up for a median of 36 months. Sodium intake was

serially assessed by a food frequency questionnaire. Based on the mean daily sodium intake prior to the first event

of death or HF hospitalization, patients were classified into sodium restricted (<2,500 mg/d) and unrestricted

($2,500 mg/d) groups. Study groups were propensity score matched according to plausible baseline confounders.

The primary outcome was a composite of death or HF hospitalization. The secondary outcomes were cardiac death and

HF hospitalization.

RESULTS Sodium intake data were available for 833 subjects (145 sodium restricted, 688 sodium unrestricted), of whom

260 were propensity matched into sodium restricted (n ¼ 130) and sodium unrestricted (n ¼ 130) groups. Sodium

restriction was associated with significantly higher risk of death or HF hospitalization (42.3% vs. 26.2%; hazard ratio [HR]:

1.85; 95% confidence interval [CI]: 1.21 to 2.84; p ¼ 0.004), derived from an increase in the rate of HF hospitalization

(32.3% vs. 20.0%; HR: 1.82; 95% CI: 1.11 to 2.96; p ¼ 0.015) and a nonsignificant increase in the rate of cardiac death

(HR: 1.62; 95% CI: 0.70 to 3.73; p¼ 0.257) and all-cause mortality (p¼ 0.074). Exploratory subgroup analyses suggested

that sodium restriction was associated with increased risk of death or HF hospitalization in patients not receiving

angiotensin-converting enzyme inhibitor or angiotensin receptor blocker (HR: 5.78; 95% CI: 1.93 to 17.27; p ¼ 0.002).

CONCLUSIONS In symptomatic patients with chronic HF, sodium restriction may have a detrimental impact on

outcome. A randomized clinical trial is needed to definitively address the role of sodium restriction in HF management.

(A Self-management Intervention for Mild to Moderate Heart Failure [HART]; NCT00018005) (J Am Coll Cardiol HF

2016;4:24–35) © 2016 by the American College of Cardiology Foundation.

H eart failure (HF) continues to increase inprevalence with an enormous impact onmorbidity and mortality (1). The treatment

of HF involves both pharmacologic and nonphar-macologic approaches (2). Traditionally, one of thecornerstones of nonpharmacological managementin HF has been restricting dietary sodium intake.Data supporting this approach are inconsistent, assome studies have shown benefit (3,4), whereasothers demonstrated better outcomes with sodiumliberalization (5–7). This controversy has manifestedin the American College of Cardiology Foundation(ACCF)/American Heart Association (AHA) guidelinesfor the management of HF. The 2009 guideline gavesodium restriction in patients with symptomatic HFa Class I recommendation (recommended) to reducecongestive symptoms with Level of Evidence: C(expert consensus) (8). Other societal guidelines is-sued similar recommendations (9). More recently,

the 2013 ACCF/AHA guidelines downgraded therecommendation for sodium restriction to Class IIa(reasonable) with Level of Evidence: C (2).

In this study, we investigated the impact of sodiumrestriction on HF outcomes in patients enrolled in theHART (Heart Failure Adherence and Retention Trial),a behavioral intervention trial that assessed the effi-cacy of self-management counseling versus educa-tion alone in symptomatic HF patients (10). Wehypothesized that if sodium restriction was protec-tive, there would be a difference in clinical outcomesand HF symptoms between patients with low versushigh sodium intake.

METHODS

We analyzed data from HART (10), which was a mul-tihospital, partially blinded, behavioral randomizedcontrolled trial, funded by the National Institutes of

ABBR EV I A T I ON S

AND ACRONYMS

ACCF = American College of

Cardiology Foundation

ACEi = angiotensin-converting

enzyme inhibitor

AHA = American Heart

Association

ARB = angiotensin receptor

blocker

CI = confidence interval

CKD = chronic kidney disease

HF = heart failure

HR = hazard ratio

NYHA = New York Heart

Association

SF-36 = 36-Item Short Form

Health Survey

Doukky et al. J A C C : H E A R T F A I L U R E V O L . 4 , N O . 1 , 2 0 1 6

Sodium Restriction and HF Outcome J A N U A R Y 2 0 1 6 : 2 4 – 3 5

26

Health (HL065547). HART assessed the im-pact of self-management counseling versuseducation alone on the primary outcome ofdeath or HF hospitalization in patients withsymptomatic HF. Details of the intervention,patient enrollment, eligibility, and results ofthe main trial were reported elsewhere(10,11). The study enrolled from 10 centers inthe Chicago, Illinoismetropolitan area and thebehavioral intervention was conducted byRush University Medical Center. The trial wasapproved by the institutional review board ofeach participating institution and was regis-tered on clinicaltrials.gov (NCT00018005).

Briefly, HART enrolled HF patients withNew York Heart Association (NYHA) func-tional class II or III symptoms, having HFwith reduced ejection fraction (HFrEF) or

preserved ejection fraction (HFpEF). Reduced systolicfunction was defined as left ventricular ejectionfraction #40%. Eligible patients had to have HFsymptoms for no less than the prior 3 months andeither: 1) ejection fraction #40%; or 2) diuretic agenttherapy for at least 3 months and $1 previous HFhospitalization. This was a null trial; it showed nosignificant impact of the self-management interven-tion on the composite of death or HF-related hospi-talization (10). For the purpose of the current study,we assessed the impact of sodium intake on HF out-comes in the HART patients over a median follow-upof 36 months.

SEE PAGES 36 AND 39

SODIUM INTAKE ASSESSMENT. A standardized foodfrequency questionnaire was used to assess sodiumintake at baseline and in annual follow-up visits atyear 1, 2, and 3 (12). The questionnaire was developedand tested at Stanford University (Stanford, California)and had been used in multiple behavioral HF clinicaltrials, sponsored by the National Institutes of Health(10–14). The questionnaire queries the intake of 57commonly consumed food items in the Americandiet during the course of the preceding week, withparticular emphasis on high sodium content meals.Each food item is weighted according to frequencyof consumption and sodium content. Based on pa-tients’ responses, estimates of daily sodium intake(milligrams) were calculated after adding anassumed baseline sodium consumption of 1,250 mg/d,derived from essential food items in the American diet(e.g., bread and meats), as detailed in Online Table 1.Because sodium intake can vary, particularly duringand after HF hospitalizations, we analyzed sodiumintake as a time-dependent variable, averaging intake

reported in all study visits preceding the first adverseevent of death or HF hospitalization. We elected not tocategorize the patients on the basis of a single sodiumintake value because sodium consumption tends tovary over time.

Based on the reported average sodium consump-tion, we divided the cohort into 2 groups: 1) lowersodium intake (<2,500 mg/d), labeled “restricted”;and 2) higher sodium intake ($2,500 mg/d), labeled“unrestricted”. Recommendations from various clin-ical societies and existing guidelines were consideredin deciding this cutoff value (9).

CLINICAL AND PSYCHOSOCIAL DATA. During base-line and yearly visits, data were gathered on de-mographics, psychosocial characteristics, medicalcomorbidities, medication usage and adherence, andNYHA functional class. Socioeconomic status wasdefined as low if the patient’s annual household in-come was <$30,000 or if the highest attained educa-tion was high school or less. Medication adherence forkey HF medications was assessed as the proportion ofpills consumed relative to the prescribed amountusing electronic pill bottle cap over the course of amonth. To uniformly adjust for diuretic agent usage,the dosages of various loop diuretic agents wereconverted into furosemide dose equivalent (OnlineTable 2). During each visit, the 6-min walk distancewas measured. Depression was defined by self-reported established diagnosis or scoring $10 on theGeriatric Depression Screening Scale (15). Chronickidney disease (CKD) was defined as glomerularfiltration rate <60 ml/min/1.73 m2 (Cockcroft-Gaultformula) or dialysis therapy. Coronary artery diseasewas defined as a prior history of coronary revascu-larization or confirmed myocardial infarction.Quality of life was assessed using health and func-tioning subscale of the Quality of Life Index, Car-diac Version-IV (modified) (16) and 2 subscales fromthe 36-Item Short Form Health Survey (SF-36) (17).We indexed the burden of 12 HF symptoms talliedusing the cardiopulmonary subscale of the HFSymptom Checklist (modified) (Online Table 3) (18).

OUTCOME ASSESSMENT. The primary outcome wasa composite of death or HF hospitalizations, as inHART. Secondary outcomes were cardiac death andHF hospitalization. The median follow-up was 36months (interquartile range: 27 to 36 months). Theoutcomes were determined by a blinded adjudicationcommittee (10). All patients or their family members(in the case of death) were contacted every 3 monthsby telephone to ascertain the occurrence of death orhospitalization. Reports of death were confirmed bymedical records, death certificates, or queries from

J A C C : H E A R T F A I L U R E V O L . 4 , N O . 1 , 2 0 1 6 Doukky et al.J A N U A R Y 2 0 1 6 : 2 4 – 3 5 Sodium Restriction and HF Outcome

27

the Social Security Death Index. HF admissions wereadjudicated by the presence of shortness of breath,peripheral edema, or chest radiographic evidence ofpulmonary edema without an alternative diagnosis.HF admissions were confirmed if the patient respon-ded to HF therapy or had a documented decrease inleft ventricular function. Cardiac death was definedas death caused by myocardial infarction, arrhyth-mias, or pump failure.

STATISTICAL ANALYSIS. Basic statistical methods. Thechi-square test was used to compare dichotomousvariables, which were expressed as numbers (per-centages). The 2-tailed Student t test was used tocompare normally distributed continuous variables,which were expressed as mean � SD. The Wil-coxon test was used to compare skewed contin-uous data.Propens i ty score match ing . Because patients werenot randomly assigned to sodium restriction, wematched patients according to their propensity tobeing sodium restricted. Multivariate logistic regres-sion model (propensity model) was fit to calculate theprobability of sodium restriction based on 32 baselinevariables listed in Table 1. Additionally, 2 interactionterms, “CKD * angiotensin-converting enzyme inhib-itor (ACEi) or angiotensin receptor blocker (ARB) use”and “CKD * spironolactone use”, were included inthe propensity model to balance the clinical implica-tions of utilizing these agents. The resultant proba-bilities were then transformed into propensity scorelogits. Six-minute walk distance and quality of lifewere not included in the propensity model due tocollinearity with NYHA functional class and SF-36scores. Serum creatinine was not included in pro-pensity model as it was accounted for in CKD status.Obesity was not included due to association withobstructive sleep apnea. Each patient in the sodium-restricted group was then matched, in 1:1 ratio, to apatient in the unrestricted group with a propensityscore within a caliper width of 0.2 * SD of the pro-pensity score logits, creating a propensity-matchedcohort. Software (Python, Version 2.6.7, Python Soft-ware Foundation) was used for propensity matching.The absolute standardized differences in baselinecovariates were calculated pre- and post-propensitymatching; covariates with >10% difference aftermatching were considered suboptimally matched.Outcome ana lyses . Kaplan-Meier curves and thelog-rank test were used to compare cumulative eventrates in the entire cohort and in the propensity-matched cohort. Risk was expressed as hazard ratio(HR) and 95% confidence interval (CI), calculatedusing univariate and multivariate Cox regression

models. To confirm the findings of the propensity-matched analyses in the entire cohort, we analyzedoutcomes using multivariate Cox proportional haz-ards models, adjusting for the calculated propensityscores. The proportional hazards assumption withrespect to Cox regression modeling was confirmedusing log-minus-log survival plots.

Subgroup analyses . In exploratory, hypothesis-generating analyses, we studied the impact ofsodium restriction on the primary outcome in pre-defined subgroups of sex, age, ethnicity (AfricanAmerican vs. others), NYHA functional class (II vs. III),HF with reduced versus preserved ejection fraction,CKD, and the use of key HF medications (ACEi/ARB,spironolactone, b-blocker) before primary outcomeevents. Using multivariate Cox regression models inthe propensity-matched cohort, we tested the impactof an interaction between sodium restriction and eachof the aforementioned subgroup strata on the primaryoutcome. To confirm the results in the entire cohort,subgroup analyses were also performed in the entirecohort using multivariate Cox regression models,adjusted for the propensity scores.

Longi tud ina l ana lys i s of t ime-vary ing outcomemeasures . Mixed-effects regression modeling inpropensity-matched cohort was used to assess theinfluence of sodium restriction on the trajectory ofthe time-varying outcomes of 6-min walk distance,cardiopulmonary HF symptoms, quality of life, SF-36physical function score, SF-36 energy and vitalityscore, and daily loop diuretic agent dose. Due toits skewed distribution, the 6-min walk distance wasanalyzed after a logarithmic transformation. Themixed-effects models used the following structure:fixed effects included sodium intake group, timesince baseline, sodium group by time interaction, andcovariates with >10% absolute standardized dif-ference between the propensity-matched groups(Figure 1); random effects included intercept andtime. Due to missing data, by design, from the third-year visit, we analyzed continuous outcome dataonly over the first 2 years of the study.

Two-tailed p values <0.05 were considered signif-icant. Software systems PASW 18.0 (IBM Corp,Armonk, New York) and SAS version 9.3 (SAS Insti-tute, Cary, North Carolina) were used for statisticalanalyses.

POWER ANALYSIS. Based on the observed rate of theprimary outcome in HART (36.4%), we calculated,post-hoc, that the available propensity-matchedcohort provided the study 80% power to detect$37% difference in the rate of the primary outcomebased on the chi-square test (2-tailed a ¼ 0.05).

TABLE 1 Baseline Characteristics of Sodium Intake Groups

Characteristic

Before Propensity Matching (N ¼ 833) After Propensity Matching (N ¼ 260)

Restricted(<2,500 mg/d)

(N ¼ 145)

Unrestricted($2,500 mg/d)

(N ¼ 688)p

Value*

Restricted(<2,500 mg/d)

(N ¼ 130)

Unrestricted($2,500 mg/d)

(N ¼ 130)p

Value*

Propensity score logit –0.74 � 0.48 –1.08 � 0.43 <0.001 –0.787 � 0.440 –0.790 � 0.435 0.956

Sociodemographics

Age, yrs† 64 � 13 63 � 13 0.655 64 � 13 63 � 13 0.632

Male† 60 (41.4) 379 (55.1) 0.003 55 (42.3) 58 (44.6) 0.707

African American† 43 (29.7) 242 (35.2) 0.203 40 (30.8) 39 (30.0) 0.893

Low socioeconomic status† 94 (64.8) 431 (62.6) 0.621 85 (65.4) 84 (64.6) 0.897

Clinical

HFrEF† 108 (74.5) 530 (77.0) 0.510 97 (74.6) 99 (76.2) 0.773

NYHA functional class III (vs. class II)† 47 (32.4) 210 (30.5) 0.654 41 (31.5) 35 (26.9) 0.413

HART treatment arm† 73 (50.3) 343 (49.9) 0.915 64 (49.2) 62 (47.7) 0.804

Systolic blood pressure, mm Hg 118 � 21 121 � 20 0.260 119 � 22 119 � 20 0.839

Third heart sound† 1 (0.7) 36 (5.2) 0.016 1 (0.8) 1 (0.8) >0.999

Jugular venous distention† 12 (8.3) 57 (8.3) 0.997 12 (9.2) 11 (8.5) 0.827

Cardiopulmonary symptoms index† 0.55 � 0.58 0.58 � 0.56 0.269 0.55 � 0.59 0.50 � 0.48 0.451

SF-36, physical function† 45.4 � 25.5 49.1 � 24.6 0.104 46.3 � 25.2 48.3 � 24.8 0.511

SF-36, energy and vitality† 48.6 � 23.9 46.2 � 23.5 0.270 48.7 � 23.0 50.0 � 23.4 0.650

Quality of Life–health score 4.4 � 1.1 4.2 � 1.0 0.139 4.4 � 1.1 4.3 � 1.0 0.531

6-min walk distance, m 246 � 144 264 � 132 0.196 248 � 143 269 � 121 0.484

Comorbidities

Coronary artery disease† 83 (57.2) 374 (54.4) 0.526 73 (56.2) 72 (55.4) 0.901

Atrial fibrillation† 59 (40.7) 273 (39.7) 0.821 56 (43.1) 47 (36.2) 0.254

Hypertension† 104 (71.7) 517 (75.1) 0.390 93 (71.5) 93 (71.5) >0.999

Diabetes mellitus† 62 (42.8) 272 (39.5) 0.472 54 (41.5) 46 (35.4) 0.308

Tobacco use† 10 (6.9) 69 (10) 0.242 9 (6.9) 8 (6.2) 0.802

Chronic kidney disease† 76 (52.4) 280 (40.7) 0.010 67 (51.5) 66 (50.8) 0.901

Stroke† 26 (17.9) 69 (10.0) 0.007 20 (15.4) 16 (12.3) 0.473

Depression† 49 (33.8) 260 (37.8) 0.365 47 (36.2) 47 (36.2) >0.999

Chronic lung disease† 18 (12.4) 96 (14.0) 0.624 14 (10.8) 9 (6.9) 0.275

Obstructive sleep apnea† 25 (17.2) 118 (17.2) 0.979 23 (17.7) 23 (17.7) >0.999

Body mass index, kg/m2 31.2 � 8.9 31.0 � 7.3 0.787 31.4 � 9.2 31.7 � 7.7 0.789

Medical treatment

Medication adherence, %† 75 � 38 74 � 34 0.788 66 � 34 67 � 34 0.827

ACEi/ARB use† 126 (86.9) 596 (86.6) 0.931 112 (86.2) 114 (87.7) 0.713

Beta-blocker use† 116 (80.0) 480 (69.8) 0.013 104 (80.0) 110 (84.6) 0.330

Spironolactone use† 44 (30.3) 187 (27.2) 0.439 36 (27.7) 47 (36.2) 0.143

Loop diuretic agent, mg/d† 72 � 66 59 � 59 0.015 65 � 61 63 � 61 0.499

Thiazide second diuretic agent use† 6 (4.1) 41 (6.0) 0.388 5 (3.8) 5 (3.8) >0.999

Statin use† 75 (51.7) 347 (50.4) 0.778 67 (51.5) 73 (56.2) 0.455

Aspirin use 69 (47.6) 326 (47.4) 0.965 63 (48.5) 66 (50.8) 0.710

Biochemical

Hemoglobin, g/dl† 13.0 � 1.6 13.1 � 1.6 0.670 13.1 � 1.6 13.1 � 1.7 0.728

Serum sodium, mmol/dl† 139.6� 3.1 140.1 � 3.1 0.053 139.6 � 3.1 140.0 � 2.6 0.362

Serum albumin, g/dl† 4.1 � 0.5 4.1 � 0.5 0.097 4.1 � 0.5 4.1 � 0.6 0.432

Serum creatinine, mg/dl 2.1 � 3.2 1.6 � 1.3 0.153 1.9 � 2.5 1.8 � 2.1 0.243

Values are mean � SD or n (%). *t Test (normally distributed continuous variables), chi-square test (dichotomous data), Wilcoxon test (skewed data). †Covariates included inthe propensity model.

ACEi ¼ angiotensin-converting enzyme inhibitor; ARB ¼ angiotensin receptor blocker; HART ¼ Heart Failure Adherence and Retention Trial; HFrEF ¼ heart failure withreduced ejection fraction; NYHA ¼ New York Heart Association; SD ¼ standard deviation; SF-36 ¼ 36-item short form questionnaire.

Doukky et al. J A C C : H E A R T F A I L U R E V O L . 4 , N O . 1 , 2 0 1 6

Sodium Restriction and HF Outcome J A N U A R Y 2 0 1 6 : 2 4 – 3 5

28

RESULTS

Sodium intake data were available for 833 of 902(92%) subjects enrolled in HART. The median

sodium intake was 3,336 mg/d (interquartile range:2,701 to 4,237 mg/d; range 1,250 to 15,678 mg/d).Based on the mean sodium intake before primaryoutcome events (death or HF hospitalization),

FIGURE 1 Absolute Standardized Differences in Baseline Covariates Between Sodium-Restricted and Unrestricted Patients Before and

After Propensity Score Matching

ACEi ¼ angiotensin-converting enzyme inhibitor; ARB ¼ angiotensin receptor blocker; HART ¼ Heart Failure Adherence and Retention Trial;

HFrEF ¼ heart failure with reduced ejection fraction; NYHA ¼ New York Heart Association; SF-36 ¼ 36-Item Short Form Health Survey.

J A C C : H E A R T F A I L U R E V O L . 4 , N O . 1 , 2 0 1 6 Doukky et al.J A N U A R Y 2 0 1 6 : 2 4 – 3 5 Sodium Restriction and HF Outcome

29

patients were classified into restricted (n ¼ 145[17.4%]) and unrestricted (n ¼ 688 [82.6%]) sodiumintake groups. The baseline characteristics of thestudy groups are summarized in Table 1. Notably,sodium-restricted patients were more frequentlywomen and were more likely to have CKD, priorstroke, been treated with a b-blocker, and receivedhigher doses of loop diuretic agents.

During a median follow-up of 36 months, therewere 163 (19.6%) deaths, 105 (12.6%) cardiac deaths,and 199 (23.9%) HF hospitalizations; 303 (36.4%)subjects had $1 events of death or HF hospitalization.As presented in Table 2, sodium-restricted patients

had a borderline significant increase in the rate ofdeath or HF hospitalization (p ¼ 0.054) and statisti-cally significant increase in HF hospitalizations (p ¼0.033). The rates of death and cardiac death weresimilar between the study groups (Table 2).

A total of 130 (90%) of the restricted sodium intakepatients could be propensity matched to unrestrictedpatients, resulting in a propensity-matched cohort of260 patients (130 sodium restricted, 130 sodium un-restricted). After matching, there was no significantdifference in the mean propensity score betweenthe matched groups (p ¼ 0.956) and the balancebetween the study groups markedly improved, as

TABLE 2 Impact of Sodium Restriction on Heart Failure Outcomes

Events

Crude Even Rates Unadjusted Risk Adjusted Risk*

Total Restricted Unrestricted HR (95% CI) p Value HR (95% CI) p Value

Entire cohort, N ¼ 833

Death 163 (19.6) 30 (20.7) 133 (19.3) 1.09 (0.73–1.62) 0.676 1.19 (0.76–1.85) 0.448

Cardiac death 105 (12.6) 16 (11.0) 89 (12.9) 0.86 (0.51–1.47) 0.589 1.03 (0.58–1.83) 0.928

HF hospitalization 199 (23.9) 44 (30.3) 155 (22.5) 1.44 (1.03–2.01) 0.033† 1.44 (1.002–2.06) 0.049†

Death or HF hospitalization 303 (36.4) 62 (42.8) 241 (35.0) 1.32 (0.995–1.74) 0.054 1.37 (1.01–1.86) 0.042†

PS-Matched cohort, N ¼ 260

Death 38 (14.6) 24 (18.5) 14 (10.8) 1.83 (0.94–3.53) 0.074 1.69 (0.87–3.31) 0.123

Cardiac death 23 (8.8) 14 (10.8) 9 (6.9) 1.62 (0.70–3.73) 0.257 1.54 (0.66–3.58) 0.319

HF hospitalization 68 (26.2) 42 (32.3) 26 (20.0) 1.82 (1.11–2.96) 0.015† 1.68 (1.02–2.75) 0.040†

Death or HF hospitalization 89 (34.2) 55 (42.3) 34 (26.2) 1.85 (1.21–2.84) 0.004† 1.72 (1.12–2.65) 0.014†

Values are n (%) unless otherwise indicated. HR and CI were derived from Cox proportional hazard models. *In the entire cohort analyses adjustment was for the PS; in thepropensity-matched cohort analysis adjustment was for covariates with >10% absolute standardized difference between the propensity-matched groups (Figure 1).†Statistically significant.

CI ¼ confidence interval; HF ¼ heart failure; HR ¼ hazard ratio (restricted vs. unrestricted); PS ¼ propensity score.

Doukky et al. J A C C : H E A R T F A I L U R E V O L . 4 , N O . 1 , 2 0 1 6

Sodium Restriction and HF Outcome J A N U A R Y 2 0 1 6 : 2 4 – 3 5

30

none of the baseline characteristics were significantlydifferent (Table 1). The propensity-matched groupswere also well balanced in baseline characteristics notincluded in the propensity model, such as body massindex, systolic blood pressure, quality of life, 6-minwalk distance, and serum creatinine. The absolutestandardized differences between the propensity-matched groups was <10% for the majority ofbaseline covariates (Figure 1).

In the propensity-matched cohort, there were 89(34.2%) events of death or HF hospitalization duringfollow-up. As illustrated in Figure 2 and Table 2, so-dium restriction was associated with a statisticallysignificant increase in the rates of death or HF hospi-talization (42.3% vs. 26.2%; HR: 1.85; 95% CI: 1.21 to2.84; p ¼ 0.004) and HF hospitalization (32.3% vs.20.0%; HR: 1.82; 95% CI: 1.11 to 2.96; p ¼ 0.015), astatistically nonsignificant increase in the rate of car-diac death (10.8% vs. 6.9%; HR: 1.62; 95% CI: 0.70 to3.73; p ¼ 0.257), and a trend toward increased risk ofall-cause mortality (18.5% vs. 10.8%; HR: 1.62; 95% CI:0.94 to 3.53; p ¼ 0.074). To ensure that these findingswere not confounded by suboptimally matchedbaseline covariates, we analyzed the risk of adverseevents after adjusting for covariates with post-matching absolute standardized difference >10%(stroke, b-blocker use, tobacco use, serum sodium,spironolactone use, lung disease, diabetes, NYHAclass III, statin use, atrial fibrillation, ACEi/ARB use).After adjustment, sodium restriction continued to beassociated with increased risk of death or HF hospi-talization (adjusted HR: 1.72; 95% CI: 1.12 to 2.65;p ¼ 0.014) and HF hospitalization (adjusted HR: 1.68;95% CI: 1.02 to 2.75; p ¼ 0.040), and nonsignificantincrease in the risk of cardiac death (p ¼ 0.319) andall-cause mortality (p ¼ 0.123), as shown in Table 2.

The findings of the propensity-matched analyseswere confirmed in the entire cohort. As presented inTable 2, after adjusting for propensity scores, sodiumrestriction was associated with a significant increasein the rates of death or HF hospitalization (HR: 1.37;95% CI: 1.01 to 1.86; p ¼ 0.042) and HF hospitalization(HR: 1.44; 95% CI: 1.002 to 2.06; p ¼ 0.049), but therewas no significant increase in the rate of cardiac death(p ¼ 0.928) or all-cause mortality (p ¼ 0.448).

SUBGROUP ANALYSES. The HRs for sodium restric-tion were consistently >1.0 in all subgroups, indi-cating increased risk (Figure 3). Notably, there was asignificant interaction between sodium restrictionand ACEi/ARB use (interaction p ¼ 0.021). This wasconfirmed when we tested for this interaction in theentire cohort adjusting for propensity scores (inter-action p ¼ 0.009). This finding indicates a differen-tial effect of sodium restriction, in that those whoare not treated with an ACEi/ARB agent had asignificantly increased risk of adverse outcome,whereas those treated with an ACEi/ARB had nosignificant increase in risk (Figure 3). Furthermore,there was a trend toward a significant interactionbetween sodium restriction and NYHA class (inter-action p ¼ 0.138), in that sodium restriction wasassociated with a significantly greater hazard ofadverse outcome in patients with NYHA class IIsymptoms, whereas the association was weaker inclass III patients (Figure 3).

TIME-VARYING HF OUTCOME MEASURES. In thepropensity-matched study groups, there were nosignificant baseline differences in the mean 6-minwalk distance, quality of life score, SF-36 physicalfunction score, SF-36 energy and vitality score,cardiopulmonary HF symptoms index, and loop

FIGURE 2 Impact of Sodium Restriction on Heart Failure Outcomes in the Propensity-Matched Cohort

Adj HR ¼ adjusted hazard ratios for covariates with >10% absolute standardized difference between the propensity-matched groups; CI ¼ confidence interval;

HF ¼ heart failure; HR ¼ hazard ratio.

J A C C : H E A R T F A I L U R E V O L . 4 , N O . 1 , 2 0 1 6 Doukky et al.J A N U A R Y 2 0 1 6 : 2 4 – 3 5 Sodium Restriction and HF Outcome

31

diuretic agents daily dose (Table 1). As illustrated inFigure 4, during 2 years of follow-up, there was nosignificant difference between the study groups inthe trajectory of change in 6-min walk distance,quality of life, physical function, energy andvitality, HF cardiopulmonary symptoms, and loopdiuretic agent dosage (group * time interactionp values $0.242).

SENSITIVITY ANALYSIS. Because sodium intake inthe unrestricted group spanned a wide range, weanalyzed outcomes among patients who had moder-ate sodium intake (2,500 to 3,999 mg/d, n ¼ 440)versus those with high sodium intake ($4,000 mg/d,n ¼ 248), adjusting for propensity scores. There wasa nonsignificant increase in the risk of death orHF hospitalization among the high versus moderatesodium intake patients (HR: 1.07; 95% CI: 0.80 to 1.42;p ¼ 0.642).

PATIENTS WITH MISSING SODIUM INTAKE DATA. Inall, 69 (7.6%) patients failed to complete the foodfrequency questionnaire, thus lacked sodium intakedata. As compared with patients with available so-dium intake data, they were more commonly AfricanAmerican (p ¼ 0.012), less likely to be married(p ¼ 0.047), and poorly adherent (mean, 15%) to keyHF medical therapies (Online Table 4). These subjectshad higher rates of death or HF hospitalization, all-cause mortality, and cardiac and noncardiac death

(all p values #0.001), but had similar rates of HFhospitalization (p ¼ 0.903) (Online Table 5).

DISCUSSION

In these analyses of data from HART, there was nodemonstrable evidence that dietary sodium restric-tion is associated with lower rate of death or HFhospitalization. In fact, dietary sodium restrictionwas associated with increased risk of adverse out-comes, particularly HF hospitalization. Although theincrease in the rates of cardiac death and all-causemortality among sodium-restricted patients were notstatistically significant due to limitation in statisticalpower, the HRs of these events were commensuratewith that of HF hospitalizations. Moreover, sodiumrestriction was not associated with a discernableimpact on 6-min walk distance, quality of life, physicalfunction, energy and vitality, or cardiopulmonarysymptoms. These findings were not confounded bydifference in diuretic agent utilization. Given theobservational nature of this hypothesis-generatingstudy, a causal effect of sodium restriction cannot beestablished. Nonetheless, our findings challenge thepresent convention and press the need for multicenterrandomized trial to definitively address the role ofsodium restriction in HF management.

For decades, dietary sodium restriction has beenthe cornerstone of HF management alongside medical

FIGURE 3 Impact of Sodium Restriction on Death or Heart Failure Hospitalization by Subgroups of the Propensity-Matched Cohort

ACEi¼ angiotensin-converting enzyme inhibitor; ARB¼ angiotensin receptor blocker; CI¼ 95%confidence interval; CKD¼ chronic kidney disease;

HFpEF¼ heart failure with preserved ejection fraction; HFrEF¼ heart failure with reduced ejection fraction; NYHA¼ New York Heart Association.

Doukky et al. J A C C : H E A R T F A I L U R E V O L . 4 , N O . 1 , 2 0 1 6

Sodium Restriction and HF Outcome J A N U A R Y 2 0 1 6 : 2 4 – 3 5

32

therapy. The data supporting this recommendationare thin, leading to inconsistent guidelines fromvarious professional societies (9). A considerablebody of published reports indicates that sodium re-striction is associated with detrimental hemodynamic

and neurohormonal changes, such as decrease incardiac index and renal perfusion and activation ofthe renin-angiotensin-aldosterone system and sym-pathetic activity (9,19,20). Although some studieshave shown benefit with sodium restriction (3,4),

FIGURE 4 Impact of Sodium Restriction on Time-Varying Outcome Measures in the Propensity-Matched Cohort

The p values represent test of significance of between-groups difference in the trajectory of time-varying outcome measures (group * time

interaction), adjusted for covariates with >10% absolute standardized difference between the propensity-matched groups. †Furosemide dose

equivalent. SF-36 ¼ 36-Item Short Form Health Survey.

J A C C : H E A R T F A I L U R E V O L . 4 , N O . 1 , 2 0 1 6 Doukky et al.J A N U A R Y 2 0 1 6 : 2 4 – 3 5 Sodium Restriction and HF Outcome

33

others have reported better outcomes with sodiumliberalization (5–7). Our study provides quality,multicenter observational evidence that low sodiumintake may worsen HF outcomes, in agreement withsingle-center clinical trials (5–7).

It is plausible that increased event rates observedin the sodium-restricted group are due to reversecausality bias, such that sicker patients were morecompliant with dietary restriction. The aforemen-tioned explanation of our findings is unlikely givenour systematic propensity matching according towide range of plausible confounders, achieving well-

balanced groups in terms of key surrogate measuresof HF severity and outcome, such as 6-min walk,quality of life, physical functioning, medicationadherence, and others. Moreover, patients were wellmatched in serum albumin, hemoglobin, body massindex, systolic blood pressure, kidney function,congestive HF symptoms and signs, and diureticagent use; thus malnutrition, hypotension (end-stageHF), kidney disease, and fluid status were unlikelyconfounders to low sodium intake. For added rigor,we performed post-matching adjustment for anyresidual difference between the matched groups.

PERSPECTIVES

COMPETENCY IN MEDICAL KNOWLEDGE:

Despite limited and conflicting data, sodium restric-

tion is generally recommended for patients with

chronic HF. In this observational study of patients

enrolled in HART, we demonstrated that sodium

restriction (<2,500 mg/d) was associated with

increased risk of the composite endpoint of death

or HF hospitalizations and HF hospitalization. Sodium

restriction was not associated with improved quality

of life, physical functioning, 6-min walk distance, or

symptoms.

TRANSLATIONAL OUTLOOK: These findings

further question the value of sodium restriction in the

management of patients with chronic symptomatic

HF, and pressing the need for multicenter randomized

trials to definitively address this matter.

Doukky et al. J A C C : H E A R T F A I L U R E V O L . 4 , N O . 1 , 2 0 1 6

Sodium Restriction and HF Outcome J A N U A R Y 2 0 1 6 : 2 4 – 3 5

34

Importantly, dietary sodium intake data in this cohortwere not completely observational, because dietarysodium restriction was a major element of the inter-vention and control arms of HART (10,11).

In exploratory subgroups analyses, sodium re-striction appeared to be particularly detrimental inthe small subgroup of patients not being treatedwith ACEi/ARB, suggesting that renin-angiotensin-aldosterone system blockade may mitigate theneurohormonal effects of sodium restriction (9). Thisis also consistent with the neurohormonal studiesdemonstrating activation of the renin-angiotensin-aldosterone system in sodium-restricted patients(9,19,20). It is also notable that sodium restriction wasdetrimental among patients with NYHA class IIsymptoms, but had a minimal effect on those withclass III symptoms. This observation is somewhatconsistent with that of Lennie et al (5), who reportedin an observational cohort that sodium restriction(<3,000 mg/d) was associated with higher event ratesof death or cardiac hospitalizations among patientswith class I to II symptoms but lower event ratesamong those with class III to IV symptoms. It isplausible that excessive sodium restriction in euvo-lemic class II patients has led to activation of therenin-angiotensin-aldosterone system causing dete-rioration of HF, whereas sodium restriction may havevalue in keeping hypervolemic class III patients fromfurther fluid retention. The ACEi/ARB and NYHAclass subgroup analyses suggest that the observedfindings in the entire cohort were mediated throughneurohormonal modulation rather than unrecog-nized confounders.

The study cohort is composed of outpatients withchronic symptomatic HF. Therefore, the investigationprovides no insight on the impact of sodium restric-tion in patients with acute decompensated HF, pa-tients with severe class IV symptoms, or those withminimal class I symptoms. Additionally, the studyprovides no information on the effect of “sodiumbinges” on HF outcome. Furthermore, 7.6% of HARTpatients lacked sodium intake data; these patientswere poorly compliant to key HF therapies and hadpoor cardiac and noncardiac outcomes. They seem torepresent challenging HF patients with more funda-mental compliance problems than sodium restriction,thus their outcomes were not solely dependent ontheir sodium intake.

STUDY LIMITATIONS. We lacked data on implantabledevices and plasma B-type natriuretic peptide levels;these are potentially important covariates. We alsolacked caloric intake data, which may confound pa-tients’ outcomes. However, we matched the patients

based on surrogates for nutritional status, suchas serum albumin and physical functioning, andensured excellent match in body mass index. Deter-mining sodium intake from food frequency ques-tionnaire rather than gold standard 24-h urinarysodium (9,21), small sample size, and nonrandomizeddesign are notable limitations.

CONCLUSIONS

In patients with chronic HF, low dietary sodiumintake (<2,500 mg/d) does not seem to reduce the riskof death or HF hospitalization compared with highersodium intake ($2,500 mg/d). Low sodium intakeseems to be associated with increased risk of HFhospitalization. Our findings support further down-grade of the ACCF/AHA sodium restriction recom-mendation in patients with chronic HF to class IIb(efficacy less well established, conflictive evidence),and press the need for multicenter randomized trialto definitively address the role of sodium restrictionin HF management.

ACKNOWLEDGMENT The authors sincerely thankGuillaume Lambert, PhD, for his contribution in thepropensity score matching.

REPRINT REQUESTS AND CORRESPONDENCE TO:

Dr. Rami Doukky, Division of Cardiology, John H.Stroger, Jr. Hospital of Cook County, 1901 WestHarrison Street, Suite 3620, Chicago, Illinois 60612.E-mail: rami_doukky@rush.edu.

J A C C : H E A R T F A I L U R E V O L . 4 , N O . 1 , 2 0 1 6 Doukky et al.J A N U A R Y 2 0 1 6 : 2 4 – 3 5 Sodium Restriction and HF Outcome

35

RE F E RENCE S

1. Go AS, Mozaffarian D, Roger VL, et al. Heartdisease and stroke statistics–2014 update: a reportFrom the American Heart Association. Circulation2014;129:e28–292.

2. Yancy CW, Jessup M, Bozkurt B, et al. 2013ACCF/AHA guideline for the management of heartfailure: a report of the American College ofCardiology Foundation/American Heart Associa-tion Task Force on Practice Guidelines. J Am CollCardiol 2013;62:e147–239.

3. Son YJ, Lee Y, Song EK. Adherence to a sodium-restricted diet is associated with lower symptomburden and longer cardiac event-free survival inpatients with heart failure. J Clin Nurs 2011;20:3029–38.

4. Philipson H, Ekman I, Forslund HB, Swedberg K,Schaufelberger M. Salt and fluid restriction iseffective in patients with chronic heart failure. EurJ Heart Fail 2013;15:1304–10.

5. Lennie TA, Song EK, Wu JR, et al. Three gramsodium intake is associated with longer event-freesurvival only in patients with advanced heartfailure. J Card Fail 2011;17:325–30.

6. Paterna S, Gaspare P, Fasullo S, Sarullo FM, DiPasquale P. Normal-sodium diet compared withlow-sodium diet in compensated congestive heartfailure: is sodium an old enemy or a new friend?Clin Sci (Lond) 2008;114:221–30.

7. Paterna S, Fasullo S, Parrinello G, et al. Short-term effects of hypertonic saline solution in acuteheart failure and long-term effects of a moderatesodium restriction in patients with compensatedheart failure with New York Heart Associationclass III (class C) (SMAC-HF Study). Am J Med Sci2011;342:27–37.

8. Hunt SA, Abraham WT, Chin MH, et al. 2009Focused update incorporated into the ACC/AHA

2005 guidelines for the diagnosis and manage-ment of heart failure in adults: a report ofthe American College of Cardiology Foundation/American Heart Association Task Force on PracticeGuidelines developed in collaboration with theInternational Society for Heart and Lung Trans-plantation. J Am Coll Cardiol 2009;53:e1–90.

9. Gupta D, Georgiopoulou VV, Kalogeropoulos AP,et al. Dietary sodium intake in heart failure. Circu-lation 2012;126:479–85.

10. Powell LH, Calvin JE Jr., Richardson D, et al.Self-management counseling in patients withheart failure: the Heart Failure Adherence andRetention Randomized Behavioral Trial. JAMA2010;304:1331–8.

11. Powell LH, Calvin JE Jr., Mendes de Leon C,et al. The Heart Failure Adherence and RetentionTrial (HART): design and rationale. Am Heart J2008;156:452–60.

12. West JA, Miller NH, Parker KM, et al.A comprehensive management system for heart fail-ure improves clinical outcomes and reduces medicalresource utilization. Am J Cardiol 1997;79:58–63.

13. Mangla A, Doukky R, Powell LH, Avery E,Richardson D, Calvin JE Jr. Congestive heart fail-ure adherence redesign trial: a pilot study. BMJOpen 2014;4:e006542.

14. DeBusk RF. MULTIFIT: a new approach to riskfactor modification. Cardiol Clin 1996;14:143–57.

15. Yesavage JA, Brink TL, Rose TL, et al. Devel-opment and validation of a geriatric depressionscreening scale: a preliminary report. J PsychiatrRes 1982;17:37–49.

16. Ferrans CE, Powers MJ. Quality of life index:development and psychometric properties. ANSAdv Nurs Sci 1985;8:15–24.

17. Ware JE Jr., Sherbourne CD. The MOS 36-ItemShort Form Health Survey (SF-36). I. Conceptualframework and item selection. Med Care 1992;30:473–83.

18. Grady KL, Meyer PM, Dressler D, et al.Longitudinal change in quality of life andimpact on survival after left ventricular assistdevice implantation. Ann Thorac Surg 2004;77:1321–7.

19. Alvelos M, Ferreira A, Bettencourt P, et al. Theeffect of dietary sodium restriction on neurohu-moral activity and renal dopaminergic response inpatients with heart failure. Eur J Heart Fail 2004;6:593–9.

20. Damgaard M, Norsk P, Gustafsson F, et al.Hemodynamic and neuroendocrine responses tochanges in sodium intake in compensated heartfailure. Am J Physiol Regul Integr Comp Physiol2006;290:R1294–301.

21. Arcand J, Floras JS, Azevedo E, Mak S,Newton GE, Allard JP. Evaluation of 2 methods forsodium intake assessment in cardiac patients withand without heart failure: the confounding effectof loop diuretics. Am J Clin Nutr 2011;93:535–41.

KEY WORDS heart failure, Heart FailureAdherence and Retention Trial, outcome,salt restriction, sodium restriction

APPENDIX For supplemental tables,please see the online version of this article.

Go to http://www.acc.org/jacc-journals-cme to take the CMEquiz for this article.