JOURNAL OF THE AMERICAN COLLEGE OF CARDIOLOGY VOL. 73, NO.

12, 2019

ª 2019 PUBLISHED BY ELSEVIER ON BEHALF OF THE

AMERICAN COLLEGE OF CARDIOLOGY FOUNDATION

Impact of Exercise Rehabilitation on

Exercise Capacity and Quality-of-Life in
Heart Failure
Individual Participant Meta-Analysis

Rod S. Taylor, PHD,a Sarah Walker, PHD,b Neil A. Smart, PHD,c Massimo F. Piepoli, MD, PHD,d Fiona C. Warren, PHD,e
Oriana Ciani, PHD,b,f David Whellan, MD, MHS,g Christopher O’Connor, MD,h Steven J. Keteyian, PHD,i
Andrew Coats, DM,j Constantinos H. Davos, MD,k Hasnain M. Dalal, MD,b,l Kathleen Dracup, PHD,m
Lorraine S. Evangelista, PHD,n Kate Jolly, PHD,o Jonathan Myers, PHD,p Birgitta B. Nilsson, PT, PHD,q
Claudio Passino, MD,r Miles D. Witham, PHD,s Gloria Y. Yeh, MD,t on behalf of the ExTraMATCH II Collaboration

ABSTRACT

BACKGROUND Previous systematic reviews have indicated that exercise-based cardiac rehabilitation (ExCR) for
patients with heart failure (HF) has a beneficial effect on health-related quality-of-life (HRQoL) and exercise capacity.
However, there is uncertainty regarding potential differential effects of ExCR across HF patient subgroups.

OBJECTIVES The authors sought to undertake an individual participant data (IPD) meta-analysis to: 1) assess the
impact of ExCR on HRQoL and exercise capacity in patients with HF; and 2) investigate differential effects of ExCR ac-
cording to a range of patient characteristics: age, sex, ethnicity, New York Heart Association functional class, ischemic
etiology, ejection fraction, and exercise capacity.

METHODS A single dataset was produced, comprising randomized trials where ExCR (delivered for 3 weeks or more)
was compared with a no exercise control group. Each trial provided IPD on HRQoL or exercise capacity (or both), with
follow-up of 6 months or more. One- and 2-stage meta-analysis models were used to investigate the effect of ExCR
overall and the interactions between ExCR and participant characteristics.

RESULTS IPD was obtained from 13 trials for 3,990 patients, predominantly (97%) with reduced ejection fraction HF.
Compared with the control group, there was a statistically significant difference in favor of ExCR for HRQoL and exercise
capacity. At 12-month follow-up, improvements were seen in 6-min walk test (mean 21.0 m; 95% confidence interval:
1.57 to 40.4 m; p ¼ 0.034) and Minnesota Living With HF score (mean improvement 5.9; 95% confidence interval: 1.0 to
10.9; p ¼ 0.018). No consistent evidence was found of differential intervention effects across patient subgroups.

CONCLUSIONS These results, based on an IPD meta-analysis of randomized trials, confirm the benefit of ExCR on HRQoL
and exercise capacity and support the Class I recommendation of current international clinical guidelines that ExCR should
be offered to all HF patients. (Exercise Training for Chronic Heart Failure [ExTraMATCH II]: protocol for an individual
participant data meta-analysis; PROSPERO: international database of systematic reviews CRD42014007170)
(J Am Coll Cardiol 2019;73:1430–43) © 2019 Published by Elsevier on behalf of the American College of Cardiology Foundation.

Listen to this manuscript’s

From the aInstitute of Health Research, College of Medicine and Health, University of Exeter and Institute of Health and Well
audio summary by
Being, University of Glasgow, Glasgow, United Kingdom; bInstitute of Health Research, University of Exeter Medical School,
Editor-in-Chief
Exeter, United Kingdom; cUniversity of New England, Armidale, New South Wales, Australia; dCardiology Unit, Guglielmo da
Dr. Valentin Fuster on
Saliceto Hospital, Piacenza, Italy; eExeter Collaboration for Academic Primary Care, Institute of Health Research, University of
JACC.org.
Exeter Medical School, Exeter, United Kingdom; fCentre for Research on Health and Social Care Management, Bocconi University,
Milan, Italy; gDepartment of Medicine, Sidney Kimmel Medical College, Philadelphia, Pennsylvania; hDuke Clinical Research
Institute, Durham, North Carolina; iDepartment of Medicine, Henry Ford Hospital, Detroit, Michigan; jIRCCS, San Raffaele, Pisana,
Italy; kCardiovascular Research Laboratory, Biomedical Research Foundation, Academy of Athens, Athens, Greece; lResearch,
m
Development & Innovation, Royal Cornwall Hospital, Truro, United Kingdom; School of Nursing, University of California
San Francisco, San Francisco, California; nUniversity of California Irvine, Irvine, California; oInstitute of Applied Health Research,

ISSN 0735-1097/$36.00 https://doi.org/10.1016/j.jacc.2018.12.072

JACC VOL. 73, NO. 12, 2019 Taylor et al. 1431
APRIL 2, 2019:1430–43 Exercise-Based HF Rehabilitation: HRQoL and Exercise Capacity

H eart failure (HF) is a major public health trial-level patient characteristics (age, sex, ABBREVIATIONS

problem with substantial morbidity and ejection fraction) and ExCR on either exercise AND ACRONYMS

mortality and is a burden to patients and capacity or HRQoL. However, such analyses
6MWT = 6-min walk test
health systems (1). Whereas survival after HF diag- are highly prone to study-level confounding
CI = confidence interval
nosis has improved, prognosis remains poor; 30% to (ecological fallacy) and should be interpreted
ExCR = exercise-based cardiac
40% of patients die within a year of diagnosis (2). Pa- with great caution. Uncertainty, therefore,
rehabilitation
tients living with HF experience marked reductions in remains as to whether there are differential
HF = heart failure
their exercise capacity, which has detrimental effects effects of ExCR on exercise capacity and
HFpEF = heart failure with
on their health-related quality-of-life (HRQoL). HRQoL across HF patient subgroups (11). In-
preserved ejection fraction
With increasing numbers of people living longer dividual participant data (IPD) meta-analysis
HFrEF = heart failure with
with symptomatic HF, the effectiveness and accessi- is increasingly being recognized as the gold reduced ejection fraction
bility of health services for HF patients have never standard approach for assessing intervention HRQoL = health-related
been more important. Exercise-based cardiac reha- subgroup effects (11,12). Although a previous quality-of-life

bilitation (ExCR) is widely recommended in clinical IPD meta-analysis (ExTraMATCH [Exercise IPD = individual participant (or
guidelines as integral to the comprehensive care of Training Meta-Analysis of Trials in Chronic patient) data

HF patients (3–7). ExCR is a process by which pa- Heart Failure]) reported the impact of ExCR MLHFQ = Minnesota Living
With Heart Failure
tients, in partnership with health professionals, are on clinical events (death and hospitalization),
Questionnaire
encouraged and supported to achieve and maintain it did not consider the outcomes of exercise
NYHA = New York Heart
optimal physical health (3). In addition to exercise capacity or HRQoL (13). Association
training, it is now accepted that ExCR programs Using IPD meta-analysis, this ExTra-
peak VO2 = peak oxygen
should be comprehensive and include education and MATCH II study aimed to assess the impact of uptake
psychological care, as well as including advice on ExCR on HRQoL and exercise capacity, and to
health and lifestyle behavior change (3,4). investigate differential effects of ExCR across sub-
groups of patients with HF.
SEE PAGE 1444

Systematic reviews and trial-level data meta- METHODS

analyses have shown ExCR offers important health
benefits for HF patients compared with control pa- This study was conducted and reported in accor-
tients (8–10). On the basis of data from 26 randomized dance with the Preferred Reporting Items for a
trials with a median follow-up of 12.4 months, Uddin Systematic Review and Meta-analysis of Individual
et al. (9) reported a mean improvement in peak oxy- Participant Data (PRISMA-IPD) statement and cur-
gen uptake (peak V O 2) of 2.79 ml/kg/min (95% confi- rent guidance on the use of IPD (14,15). Our full
dence interval [CI]: 2.05 to 3.53 ml/kg/min) following study protocol has been published elsewhere and is
ExCR. The 2014 Cochrane review reported a clinically registered on the PROSPERO database of systematic
important improvement across 13 random controlled review protocols (CRD42014007170) (16,17). The clinical
trials in disease-specific HRQoL as assessed by the events results have been published elsewhere (18).
Minnesota Living With Heart Failure Questionnaire SEARCH STRATEGY AND SELECTION CRITERIA.
(MLHFQ) up to 12-month follow-up (mean score 5.8 Trials were identified from the original ExTraMATCH
points; 95% CI: 9.2 to 2.4 points) compared with IPD meta-analysis carried out in 2004 and updated
control patients (8). Using meta-regression analysis, with trials identified in the 2014 Cochrane systematic
these meta-analyses found no association between review of ExCR for HF (8,13). The Cochrane

University of Birmingham, Birmingham, United Kingdom; pVA Palo Alto Health Care System/Stanford University, Stanford,
California; qDivision of Medicine, Oslo University Hospital and Faculty of Health Sciences, Oslo Metropolitan University, Oslo,
Norway; rFondazione G. Monasterio and Scuola Superiore Sant’Anna, Pisa, Italy; sNIHR Newcastle Biomedical Research Centre,
Institute of Neuroscience, Faculty of Medical Sciences, Newcastle University, Newcastle, United Kingdom; and the tBeth Israel
Deaconess Medical Center, Department of Medicine, Boston, Massachusetts. This study presents independent research funded by
the National Institute for Health Research Health Technology Assessment Programme (NIHR-HTA 15/80/30). The views expressed
are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health and Social Care. The funders
peer review process informed the study protocol. The sponsor of the study had no role in data interpretation, or writing of the
report. Prof. Taylor and Dr. Dalal are co-chief investigators and Prof. Jolly a co-investigator on an NIHR-funded program grant (RP-
PG-1210-12004). Prof. Jolly is funded in part by NIHR CLAHRC West Midlands. Dr. Coats has received personal fees from Actimed,
AstraZeneca, Faraday, Gore, Impulse Dynamics, Menarini, Novartis, Nutricia, Resmed, Respicardia, Servier, Stealth Peptides,
Verona, and Vifor. All other authors have reported that they have no relationships relevant to the contents of this paper to
disclose.

Manuscript received August 29, 2018; revised manuscript received November 27, 2018, accepted December 11, 2018.
1432 Taylor et al. JACC VOL. 73, NO. 12, 2019

Exercise-Based HF Rehabilitation: HRQoL and Exercise Capacity APRIL 2, 2019:1430–43

F I G U R E 1 PRISMA-IPD Flow Diagram

2014 Cochrane review

n = 33 trials (n = 46 publications)

RCTs included from

Excluded trials, n = 14
ExTraMA TCH I analysis
Less than 50 patients in trial, n = 14
n = 4 trials (n = 4 publications)

ExTraMATCH II IPD Meta-analysis

n = 23 trials met inclusion criteria
[n = 4,398 patients]
Excluded trials, n = 4
Datasets destroyed / lost, n = 2
Trial group uncontactable, n = 1
Patients duplicated in another study, n = 1
ExTraMATCH II IPD meta-analysis
n = 19 trials (3,900 patients) provided data
n = 13 trials (14 comparisons, 3,332 patients) provided
either HRQ oL or exercise capacity data

HRQoL IPD meta-analysis Exercise capacity IPD meta-analysis

MLHF Peak VO2, directly reported
n = 8 trials provided data (759 patients) provided data n = 7 trials provided data (2,685 patients) provided data
All HRQoL outcomes 6MWT, directly reported
n = 9 trials, 10 comparisons (3,000 patients) provided data n = 8 trials provided data (2,717 patients) provided data
All exercise capacity outcomes
n = 13 trials, 14 comparisons (3,332 patients) provided data

A PRISMA-IPD (Preferred Reporting Items for Systematic Reviews and Meta-Analyses of Individual Participant Data) flow diagram to show selection and synthesis of
ExTraMATCH (Exercise Training for Chronic Heart Failure) II study data. 6MWT ¼ 6-min walk test; HRQoL ¼ health-related quality-of-life; MLHF ¼ Minnesota Living
with Heart Failure Questionnaire; peak VO2 ¼ peak oxygen uptake; RCT ¼ randomized controlled trial.

review searched the following electronic databases: may also include health education and/or a psycho-
Cochrane Central Register of Controlled Trials (CEN- logical intervention); 3) a comparator arm that did not
TRAL) in the Cochrane Library, EMBASE, MEDLINE, prescribe an exercise intervention; 4) a minimum
CINAHL, PsycINFO, and the NHS Centre for Reviews follow-up of 6 months; and 5) and a sample size of more
and Dissemination (CRD). Conference proceedings than 50 (to ensure that the logistical effort in obtaining,
were searched on Web of Science. Trial registers cleaning, and organizing the data was commensurate
(Controlled-trials.com and ClinicalTrials.gov) and with the contribution of the dataset to the analysis)
reference lists of all eligible trials and identified sys- (19,20).
tematic reviews were also checked. No language DATA MANAGEMENT. Principal investigators of
limitations were imposed. Details of the search studies were invited by e-mail to participate in this
strategy used are reported elsewhere (16,17). IPD meta-analysis and share their anonymized trial
Trials were included if they met the following data. Patients in the clinical trials providing data gave
criteria: 1) randomized trials of adult patients (18 years their consent on entry to the original clinical trial. All
of age and older) with a diagnosis of HF with reduced included datasets had ethical approval and consent
ejection fraction (HFrEF) or HF with preserved ejection from their sponsors; they were not required to seek
fraction (HFpEF) based on objective assessment of left additional ethical approval for the inclusion of their
ventricular ejection fraction and clinical findings; 2) data in this analysis. The complete list of all reques-
ExCR intervention that delivered an aerobic exercise ted variables and details on collaboration with prin-
training component involving the lower limbs, lasting cipal investigators are reported in the study protocol
a minimum of 3 weeks, either alone or as part of a (8). Data from each trial were checked on range,
comprehensive cardiac rehabilitation program (which extreme values, internal consistency, missing values,
JACC VOL. 73, NO. 12, 2019 Taylor et al. 1433
APRIL 2, 2019:1430–43 Exercise-Based HF Rehabilitation: HRQoL and Exercise Capacity

and consistency with published reports. Trial in-

T A B L E 1 Baseline Characteristics of Patients
vestigators were contacted about data discrepancies
or missing information. Each anonymized dataset ExCR Control All
(n ¼ 1,662) (n ¼ 1,670) (N ¼ 3,332)
was saved in its original format and then converted
Age, yrs 60.9  13.2 61.2  13.5 61.1  13.4
and combined into 1 overall master dataset. All files
Sex
were stored on a secure, password-protected com- Male 1,187 (71.4) 1,237 (74.1) 2,424 (72.8)
puter server managed and in accordance with the Female 475 (28.6) 433 (25.9) 908 (27.3)
data management standard operating procedures of Baseline ejection fraction, % 27.0  8.8 26.9  8.7 26.9  8.8
Exeter Clinical Trials Unit, a U.K. Clinical Research HFrEF <45% 1,721 (96.8) 1,744 (97.5) 3,465 (97.1)

Collaboration (UKCRC) registered clinical trials unit. HFpEF $45% 57 (3.2) 45 (2.5) 102 (2.9)
NYHA functional status
Access to data at all stages of cleaning and analysis
Class I 20 (1.2) 25 (1.5) 45 (1.4)
was restricted to the Exeter research team (R.S.T.,
Class II 1,002 (61.2) 1,032 (62.8) 2,034 (62.0)
S.W., F.C.W., and O.C.). Class III 597 (36.5) 569 (34.6) 1,166 (35.5)
Class IV 19 (1.2) 18 (1.1) 37 (1.1)
SPECIFICATION OF OUTCOMES, SUBGROUPS, AND
Etiology
RISK OF BIAS ASSESSMENT. HRQoL and exercise
Ischemic 892 (54.9) 884 (54.1) 1,776 (54.5)
capacity data were obtained from trial investigators at Nonischemic 732 (45.1) 750 (45.9) 1,482 (45.5)
the patient level. HRQoL was recorded as 1 of 3 vali- Ethnicity
dated measures: 1) Minnesota Living with Heart White 1,085 (69.3) 1,117 (70.9) 2,202 (70.1)
Failure Questionnaire (MLHFQ) (21); 2) Kansas City Nonwhite 480 (30.7) 458 (29.1) 938 (30.0)
Cardiomyopathy Questionnaire (22); and 3) Guyatt MLHFQ 35.6  23.7 33.6  25.6 34.6  24.7
Peak VO2, ml/kg/min 15.0  4.5 15.1  4.7 15.0  4.6
Chronic Heart Failure scale (23). The first analysis was
6MWT, m 362.6  109.3 362.5  112.1 362.6  110.7
performed using only MLHFQ data; the second anal-
ysis used a standardized score calculated from any of Values are mean  SD or n (%).
the 3 aforementioned measures. Because MLHFQ re- 6MWT ¼ 6-min walk test; EF ¼ ejection fraction; ExCR ¼ exercise-based cardiac rehabilitation;
HFpEF ¼ heart failure with preserved ejection fraction; HFrEF ¼ heart failure with reduced
ports higher HRQoL as a lower score, the scales of the ejection fraction; MLHFQ ¼ Minnesota Living With Heart Failure Questionnaire; NYHA ¼ New York
Kansas City Cardiomyopathy Questionnaire and Heart Association; VO2 ¼ oxygen uptake.

Guyatt Heart Failure score (which report higher

HRQoL as a higher score) were reversed before stan-
Study quality and risk of bias were assessed using the
dardizing so that the directionality would be the same
TESTEX quality assessment tool (25).
as MLHFQ. Therefore, for both the MLHFQ score and
standardized HRQoL score, an improvement in STATISTICAL ANALYSIS. A detailed statistical anal-
HRQoL is shown by a reduction in the overall score. ysis plan was prepared (available from the authors).
Exercise capacity was recorded as 1 of 4 validated All analyses were carried out according to the prin-
exercise capacity measures: 1) peak VO 2 (ml/kg/min); 2) ciple of intention to treat (i.e., patients analyzed as
distance (meters) walked in a 6-min walk test (6MWT); randomized) and included all patients providing the
3) distance (meters) walked in an incremental shuttle data required for each model. All 1-stage and 2-stage
walk test; and 4) cycle ergometer Watts. Two of these analyses used random effects models as the overall
measures, peak V O2 and 6MWT, were analyzed as dataset is likely to include a high degree of clinical
separate outcomes. A third outcome, a standardized heterogeneity across the individual trials due to dif-
exercise capacity score for patients with any validated ferences in population, exercise-based rehabilitation
exercise capacity measure, was also analyzed. The intervention, and comparator intervention (26). All
large HF-ACTION (Exercise Training Program to results are reported as a between-group mean dif-
Improve Clinical Outcomes in Individuals With ference (ExCR-control) with a 95% CI and p value.
Congestive Heart Failure) trial (24) provided data on The primary analyses comprised 1-stage and 2-stage
both peak V O2 and 6MWT, and was included in all an- IPD meta-analyses carried out at 2 follow-up times:
alyses, with the peak VO2 measure taking precedence 6 and 12 months. For all analyses, we used the obser-
for the standardized exercise capacity score. vation at, or closest before, the analysis time. Using
We also sought IPD on the following pre-defined this criterion, more trials had available data at 12-
subgroups: age, sex, ejection fraction (HFpEF [$45% month follow-up than at 6-month follow-up. There-
ejection fraction] vs. HFrEF [<45% ejection fraction]), fore, we have regarded the 12-month data analyses as
New York Heart Association (NYHA) functional class, being the primary analyses. The results at 12-month
HF etiology (ischemic vs. nonischemic), ethnicity follow-up are reported ahead of the 6-month results in
(white vs. nonwhite), and baseline exercise capacity. order to optimize the number of trials included.
1434 Taylor et al. JACC VOL. 73, NO. 12, 2019

Exercise-Based HF Rehabilitation: HRQoL and Exercise Capacity APRIL 2, 2019:1430–43

capacity) over the duration of each trial. These

T A B L E 2 Characteristics of Included Studies and Interventions
(14 Comparator Studies)
models used outcome data at all available time
points. Adjustments for baseline values of the
Study characteristics
Publication year
outcome measure were made; no other covariates
1990 to 1999 0 (0) were included in the model. This model included a
2000 to 2009 9 (64) time by treatment interaction term.
2010 to 2012 5 (36) To test the robustness of the primary analyses, pre-
Unpublished 0 (0) specified sensitivity analyses were carried out. First,
Main study location
each primary analysis was repeated after exclusion of
Europe 9 (64)
the largest trial, the HF-ACTION study (24). Second,
North America* 5 (36)
aggregate data from studies that did not provide IPD
Single-study center
Single 10 (71) was added and the impact on meta-analysis conclu-
Multiple 4 (29) sions assessed. We checked for potential small-study
Sample size bias by assessing funnel plot asymmetry and using
0 to 99 8 (57) the Egger test (27). Additional plots of the results of
100 to 999 5 (36) the 1-stage IPD meta-analysis models, stratified by
$1,000 1 (7)
patient characteristics, are presented in order to give
Duration of latest follow-up, weeks
the reader a visual representation of the differential
HRQoL outcomes 33 (26–104)
Exercise capacity outcomes 26 (9–520)
effect of ExCR in each subgroup. All analyses were
Intervention characteristics undertaken using Stata version 14.2 software (Stata-
Intervention type Corp, College Station, Texas).
Exercise-only programs 9 (64)
Comprehensive programs 5 (36) RESULTS
Type of exercise
Aerobic exercise only 10 (71)
SELECTION AND INCLUSION OF STUDIES. Of the 23
Aerobic plus resistance training 4 (29)
trials identified either in the ExTraMATCH IPD meta-
Dose of intervention
Duration of intervention, weeks 24 (4–120) analysis (13) or the 2014 Cochrane systematic review
Frequency, sessions per week 3 (2–7) of ExCR for HF (8,16), we were unable to include data
Length of exercise session, min 30 (15–60) from 3 trials (n ¼ 355): for 2 trials, data were no longer
Exercise intensity, range 40% to 70% peak VO2 available (28,29), and the investigators of the third
11 to 15 Borg rating
Setting trial could not be contacted (30).
Center-based only 9 (64) Of the 20 trials remaining, 1 trial (31) was excluded
Home-based only 5 (36) due to an overlap between patients included in
another identified trial (32). Thirteen studies pro-
Values are n (%), median (range), or range. *HF-ACTION study (24) was categorized
as North America but was also delivered to a small number of patients in France. vided anonymized IPD for analysis of HRQoL and
HRQoL ¼ health-related quality-of-life; peak VO2 ¼ peak oxygen uptake.
exercise capacity outcomes (24,32–43). Published
trial-level data were available for an additional 5 trials
One-stage IPD models used a hierarchical random for each of the HRQoL (28,29,44–46) and exercise
effects regression model, adjusted for the baseline capacity analyses (28–30,44,45). In addition to
value of the outcome measure. We ran a series of comparing usual care to an intervention arm of usual
models to estimate the overall treatment effect and to care plus ExCR, Gary et al. (35) also compared the
investigate potential interactions between ExCR and effects of cognitive behavior therapy to cognitive
pre-defined patient subgroups (age, sex, left ventric- behavior therapy plus ExCR. For the purpose of
ular ejection fraction [<45% or $45%], heart failure analysis from this point forward, this will be
etiology [ischemic vs. nonischemic], NYHA functional described as 1 trial providing 2 comparators and be
class [I/II vs. III/IV], and baseline exercise capacity analyzed as separate trials from this point forward.
[16,17]). Each model investigated 1 interaction effect For the HRQoL analysis, 9 trials (including 10
only. We used 2-stage random effects models as a comparator groups) provided data for 3,000 patients
sensitivity analysis to estimate the effect of ExCR. The (1,496 ExCR, 1,504 control) with a median follow-up
s2 and I2 statistics were reported alongside the asso- of 33 weeks (24,34,35,38–43). For the exercise capac-
ciated p value for the results of the main analyses. ity analysis, 13 trials (14 comparator groups) provided
The secondary analyses used a random effects hi- 3,332 patients (1,662 ExCR, 1,670 control) with a me-
erarchical model that took account of the repeated dian follow-up of 26 weeks (24,32–43). Figure 1 sum-
measurement of the outcome (HRQoL or exercise marizes the study selection process.
JACC VOL. 73, NO. 12, 2019 Taylor et al. 1435
APRIL 2, 2019:1430–43 Exercise-Based HF Rehabilitation: HRQoL and Exercise Capacity

C ENTR AL I LL U STRA T I O N Exercise-Based Heart Failure Rehabilitation: Health-Related Quality-of-Life and

Exercise Capacity at 12 Months

A Study Effect (95% CI)

Dracup (2007) –2.19 (–9.09 to 4.70)

Gary (2010 - usual care) –3.51 (–17.28 to 10.27)
Gary (2010 - CBT) –9.54 (–22.37 to 3.29)
Jolly (2009) 1.35 (–4.02 to 6.71)
Nilsson (2008) –6.78 (–13.05 to –0.50)
Passino (2006) –23.40 (–28.87 to –17.94)
Witham (2012) 0.85 (–3.13 to 4.83)
Yeh (2011) –3.09 (–9.31 to 3.14)
Overall –5.73 (–12.38 to 0.92)

–20 0 20
Favors Exercise-Based Favors Control
Cardiac Rehabilitation

B Study Effect (95% CI)

Dracup (2007) 14.62 (–15.31 to 44.55)

Gary (2010 - usual care) 42.79 (–15.06 to 100.64)
Gary (2010 - CBT) 91.74 (41.52 to 141.96)
HF-ACTION (2009) 9.24 (1.99 to 16.48)
Nilsson (2008) 77.22 (47.58 to 106.87)
Witham (2005) 5.18 (–17.41 to 27.76)
Witham (2012) –2.03 (–26.14 to 22.07)
Yeh (2011) 1.25 (–24.71 to 27.20)
Overall 23.98 (5.30 to 42.65)

–100 0 100
Favors Control Favors Exercise-Based
Cardiac Rehabilitation
Taylor, R.S. et al. J Am Coll Cardiol. 2019;73(12):1430–43.

A forest plot from the 2-stage individual participant data meta-analysis model to (A) Minnesota Living with Heart Failure Questionnaire and (B) 6-min walk test,
directly reported. CBT ¼ cognitive behavioral therapy; CI ¼ confidence interval; HF-ACTION ¼ Exercise Training Program to Improve Clinical Outcomes in Individuals
With Congestive Heart Failure.

STUDY, PATIENT, AND TRIAL CHARACTERISTICS. preserved ejection fraction heart failure (defined as
Patient baseline characteristics were well balanced ejection fraction >45%). Most patients were in NYHA
between ExCR and control patients (Table 1). The ma- functional class II (62%) or III (36%). Studies were
jority of patients were male (73%) with a mean age of 61 published between 2000 and 2012 across Europe and
years. The mean baseline left-ventricular ejection North America. Sample size ranged from 50 to 2,130
fraction was 27%; fewer than 3% of patients had patients. All trials evaluated an aerobic exercise
1436 Taylor et al. JACC VOL. 73, NO. 12, 2019

Exercise-Based HF Rehabilitation: HRQoL and Exercise Capacity APRIL 2, 2019:1430–43

F I G U R E 2 Effect of ExCR on HRQoL and Exercise Capacity at 6 Months: 2-Stage IPD Meta-Analysis

A
Study Effect (95% CI)

Dracup (2007) –1.66 (–8.37 to 5.04)

Gary (2010 - usual care) –3.51 (–17.28 to 10.27)
Gary (2010 - CBT) –9.54 (–22.37 to 3.29)
Jolly (2009) 0.25 (–4.99 to 5.49)
Nilsson (2008) –6.78 (–13.05 to –0.50)
Witham (2012) 0.85 (–3.13 to 4.83)
Yeh (2011) –3.09 (–9.31 to 3.14)
Overall –1.72 (–4.15 to 0.70)

–20 0 20
Favors ExCR Favors Control

B
Study Effect (95% CI)

Dracup (2007) –0.10 (–0.39 to 0.19)

Gary (2010 - usual care) –0.16 (–0.75 to 0.43)
Gary (2010 - CBT) –0.41 (–0.97 to 0.15)
HF-ACTION (2009) –0.11 (–0.17 to –0.05)
Jolly (2009) 0.01 (–0.22 to 0.24)
Nilsson (2008) –0.30 (–0.57 to –0.02)
Witham (2005) 0.04 (–0.40 to 0.49)
Witham (2012) 0.04 (–0.16 to 0.24)
Yeh (2011) –0.16 (–0.44 to 0.12)
Overall –0.10 (–0.15 to –0.05)

–1 0 1
Favors ExCR Favors Control

C
Study Effect (95% CI)

Dracup (2007) 0.04 (–1.26 to 1.34)

Hambrecht (2000) –2.16 (–4.43 to 0.10)

HF-ACTION (2009) 0.47 (0.24 to 0.71)

Mueller (2007) 4.48 (2.35 to 6.60)

Passino (2006) 1.57 (0.66 to 2.49)

Yeh (2011) –0.02 (–1.02 to 0.98)

Overall 0.69 (–0.24 to 1.62)

–5 0 5
Favors Control Favors ExCR

The blue circle is centered on the point estimate of the effect of ExCR in each trial, with the horizontal line showing the 95% confidence interval
(CI) of this estimate. An arrow to either the left or right shows that the CI extends beyond the area shown in the forest plot. The size of the
blue square around the point estimate is proportional to the weight that the individual trial contributes to the meta-analysis. The diamond and
vertical red line show the overall estimate of the effect of ExCR in the 2-stage meta-analysis. (A) Minnesota Living with Heart Failure
Questionnaire (MLHFQ). (B) All HRQoL measures (standardized score). (C) Peak VO2, directly reported. (D) 6MWT, directly reported. (E) All
exercise capacity measures (standardized score). CBT ¼ cognitive behavioral therapy; ExCR ¼ exercise-based cardiac rehabilitation; HF-ACTION ¼
Exercise Training Program to Improve Clinical Outcomes in Individuals With Congestive Heart Failure; other abbreviations as in Figure 1.

Continued on the next page

JACC VOL. 73, NO. 12, 2019 Taylor et al. 1437
APRIL 2, 2019:1430–43 Exercise-Based HF Rehabilitation: HRQoL and Exercise Capacity

F I G U R E 2 Continued

D
Study Effect (95% CI)

Dracup (2007) 5.19 (–28.39 to 38.78)

Gary (2010 - usual care) 42.79 (–15.06 to 100.64)
Gary (2010 - CBT) 91.74 (41.52 to 141.96)
HF-ACTION (2009) 18.14 (11.60 to 24.68)
Nilsson (2008) 77.22 (47.58 to 106.87)
Witham (2005) 5.18 (–17.41 to 27.76)
Witham (2012) –2.03 (–26.14 to 22.07)
Yeh (2011) 1.25 (–24.71 to 27.20)
Overall 24.35 (6.13 to 42.57)

–100 0 100
Favors Control Favors ExCR

E
Study Effect (95% CI)

Dracup (2007) 0.15 (–0.09 to 0.38)

Gary (2010 - usual care) 0.32 (–0.09 to 0.72)
Gary (2010 - CBT) 0.65 (0.30 to 0.99)
Giannuzzi (2003) 0.77 (0.52 to 1.01)
Hambrecht (2000) –0.39 (–0.80 to 0.02)
HF-ACTION (2009) 0.09 (0.04 to 0.14)
Jolly (2009) 0.16 (–0.04 to 0.37)
Mueller (2007) 0.86 (0.45 to 1.27)
Nilsson (2008) 0.53 (0.33 to 0.73)
Passino (2006) 0.32 (0.13 to 0.51)
Witham (2005) 0.10 (–0.09 to 0.28)
Witham (2012) –0.04 (–0.22 to 0.14)
Yeh (2011) 0.00 (–0.20 to 0.21)
Overall 0.26 (0.12 to 0.40)

–1 0 1
Favors Control Favors ExCR

intervention; 4 also included resistance training a maximum score of 15 (Online Table 1). The criteria of
(34,38,40,41). Four trials (5 comparators) were con- allocation concealment and physical activity moni-
ducted in an exclusively home-based setting toring in the control groups were met in only 2 (24,38)
(34,35,38,43); all other trials delivered ExCR in a and 3 studies (24,34,42), respectively. The other TES-
center-based setting. The “dose” of exercise training TEX criteria were each met in at least 50% of trials.
varied across studies; average session duration ranged EFFECT OF INTERVENTION ON OUTCOMES. One-
from 15 to 60 min (including warmup and cooldown); stage meta-analysis showed a significant improve-
minimum number of sessions per week was 2, with a ment in HRQoL for those on the ExCR intervention
maximum of 7; exercise intensity equivalent ranged compared with control, as assessed by the MLHFQ, at
from 40% to 70% peak V O2 ; and the duration of inter- 12-month follow-up: (mean improvement 5.9; 95% CI:
vention ranged from 4 to 120 weeks (Table 2). 1.0 to 10.9; p ¼ 0.018; s 2 ¼ 77; I2 ¼ 88%) (Online Table 2)
QUALITY OF INCLUDED TRIALS. The overall quality and standardized HRQoL score (mean improvement
of included trials was judged to be moderate to good, 0.20 SD units; 95% CI: 0.03 to 0.37; p ¼ 0.020; s 2 ¼ 0.07;
with a median TESTEX (25) score of 11 (range 9 to 14) of I2 ¼ 85%) (Online Table 3). Similar results were seen at
1438 Taylor et al. JACC VOL. 73, NO. 12, 2019

Exercise-Based HF Rehabilitation: HRQoL and Exercise Capacity APRIL 2, 2019:1430–43

F I G U R E 3 Effect of ExCR on HRQoL and Exercise Capacity Across Patient Subgroups 12 Months

A
Coefficient (95% CI) p-Value for Interaction

Age (years)
Less than 60 –7.76 (–13.7, –1.88) 0.912
60 and over –5.02 (–10.8, 0.75)
Sex
Male –7.42 (13.6, –1.27) 0.592
Female –3.44 (–8.10, 1.22)
Ejection fraction (%)
HFrEF –6.48 (–13.3, 0.32) 0.165
HFpEF –6.07 (–15.0, 2.89)
NYHA functional class
I/II –3.95 (–8.67, 0.77) 0.061
III/IV –10.2 (–16.7, –3.68)
Heart failure etiology
Ischemic –5.50 (–11.7, 0.74) 0.477
Non-ischemic –7.04 (–13.7, –0.37)
Ethnic group
White –5.68 (–11.1, –0.24) 0.169
Non-white –5.29 (–13.1, 2.47)
Peak VO2 (ml/kg/min)
Less than 15 –12.1 (–24.2, 0.12) 0.262
15 and over –3.29 (–6.96, 0.38)

Overall –5.94 (–10.9, –1.01)

–20 –15 –10 –5 0 5

B
Coefficient (95% CI) p-Value for Interaction

Age (years)
Less than 60 –0.16 (–0.25, –0.08) 0.734
60 and over –0.17 (–0.35, 0.01)
Sex
Male –0.27 (–0.47, –0.07) 0.775
Female –0.13 (–0.23, –0.02)
Ejection fraction (%)
HFrEF –0.24 (–0.48, –0.01) 0.340
HFpEF –0.23 (–0.54, 0.07)
NYHA functional class
I/II –0.18 (–0.33, –0.02) 0.505
III/IV –0.23 (–0.04, –0.01)
Heart failure etiology
Ischemic –0.19 (–0.41, 0.02) 0.611
Non-ischemic –0.30 (–0.52, –0.08)
Ethnic group
White –0.18 (–0.36, –0.01) 0.787
Non-white –0.12 (–0.22, –0.01)
Peak VO2 (ml/kg/min)
Less than 15 –0.48 (–0.88, –0.08) 0.230
15 and over –0.10 (–0.23, 0.02)

Overall –0.20 (–0.37, –0.03)

0
.8

.2
.4
.6

–0
–0
–0

–0

A plot to show the effect size for health-related quality-of-life outcomes (A and B) and exercise capacity outcomes (C and D), stratified by
patient characteristics. All results are reported as a between group mean difference (ExCR-control) with a 95% confidence interval (CI) from
1-stage meta-analyses carried out by strata. The p values given are from the interaction tests in the main 1-stage meta-analysis. (A) Minnesota
Living with Heart Failure Questionnaire (MLHFQ). (B) All HRQoL measures (standardized score). (C) Peak VO2, directly reported. (D) 6MWT,
directly reported. Abbreviations as in Figure 1.

Continued on the next page

JACC VOL. 73, NO. 12, 2019 Taylor et al. 1439
APRIL 2, 2019:1430–43 Exercise-Based HF Rehabilitation: HRQoL and Exercise Capacity

F I G U R E 3 Continued

C
Coefficient (95% CI) p-Value for Interaction

Age (years)
Less than 60 1.12 (–0.31, 2.56) 0.646
60 and over 0.90 (–0.47, 2.28)
Sex
Male 0.92 (–0.59, 2.42) 0.036
Female 0.99 (0.59, 1.39)
Ejection fraction (%)
HFrEF 1.15 (–0.49, 2.79) 0.897
HFpEF 1.22 (–0.10, 2.54)
NYHA functional class
I/II 0.78 (–0.63, 2.18) 0.318
III/IV 0.38 (–0.32, 1.08)
Heart failure etiology
Ischemic 1.20 (–0.55, 2.96) 0.577
Non-ischemic 0.77 (–0.57, 2.11)
Ethnic group
White 1.30 (–0.32, 2.92) 0.800
Non-white 0.47 (0.08, 0.86)
Peak VO2 (ml/kg/min)
Less than 15 1.12 (0.14, 2.10) 0.332
15 and over 0.88 (–0.72, 2.47)

Overall 1.01 (–0.42, 2.44)

0 1 2

D
Coefficient (95% CI) p-Value for Interaction

Age (years)
Less than 60 11.9 (1.98, 21.9) 0.911
60 and over 21.0 (1.36, 40.6)
Sex
Male 11.1 (–12.6, 34.7) 0.034
Female 23.4 (12.4, 34.4)
Ejection fraction (%)
HFrEF 22.0 (0.04, 43.9) 0.560
HFpEF 59.1 (20.3, 98.9)
NYHA functional class
I/II 21.5 (–2.21, 45.1) 0.847
III/IV 14.6 (3.74, 25.5)
Heart failure etiology
Ischemic 21.6 (–1.73, 45.0) 0.510
Non-ischemic 6.60 (–21.0, 34.2)
Ethnic group
White 22.1 (3.35, 40.8) 0.044
Non-white 1.43 (–9.83, 12.7)
6MWT at baseline
Less than 360m 16.2 (6.54, 25.8) 0.176
360m and over 12.0 (–13.4, 37.5)

Overall 21.0 (1.57, 40.4)

0 20 40 60 80

6-month follow-up. Two-stage meta-analysis results Compared with control, treatment effects from the
were comparable and are presented graphically for 12- 1-stage meta-analysis at 12-month follow-up showed a
month follow-up (Central Illustration, Online Figure 3) statistically significant improvement with ExCR in
and 6-month follow-up (Figure 2). exercise capacity as assessed by 6MWT (mean
1440 Taylor et al. JACC VOL. 73, NO. 12, 2019

Exercise-Based HF Rehabilitation: HRQoL and Exercise Capacity APRIL 2, 2019:1430–43

difference: 21.0 m; 95% CI: 1.6 to 40.4; p ¼ 0.034; 3990 randomized patients, predominantly (97%) with
s2 ¼ 491; I2 ¼ 78%) (Online Table 4) and standardized reduced ejection fraction HF, showed some evidence
exercise capacity score (mean difference 0.27 SD units; that ExCR improves both exercise capacity and
95% CI: 0.11 to 0.43; p ¼ 0.001; s 2 ¼ 0.08; I 2 ¼ 91%) HRQoL compared with no exercise control 12-month
(Online Table 5). No significant difference in peak V O2 follow-up, with weaker evidence for a treatment ef-
at 12 months was observed: 1.01 (95% CI: 0.42 to 2.44; fect at 6-month follow-up. The magnitude of the
p ¼ 0.168; s 2 ¼ 2.17; I 2 ¼ 94%) (Online Table 6). treatment effect of ExCR on MLHFQ score observed at
In the repeated measures analyses for each 12-month follow-up was not only statistically signifi-
HRQoL and exercise capacity outcome, a significant cant, but also clinically important (47), with a mean
interaction between ExCR and time was observed between-group difference of >5 points, favoring the
(Online Figure 1). In sensitivity analyses, the results ExCR group. Also, there was an increase of $16 m in
of the analyses excluding the HF-ACTION study, were the 6MWT in the ExCR group, which may also be
broadly consistent with the overall results (Online clinically significant (48). Interaction analyses
Tables 3 to 6). Similar results were found with the showed that younger patients responded better to
addition of the trial-level aggregate data to the ExCR in terms of improved HRQoL; women and white
2-stage model at 12-month follow-up. patients had a better exercise capacity response.
There was no evidence of significant small study However, the interactions between ExCR and age,
bias for the 5 outcomes studied (Online Figure 2). sex, and ethnicity were not consistent across health
DIFFERENTIAL EFFECTS ACROSS SUBGROUPS. An- outcomes, different analyses, and time points. The
alyses revealed no consistent interaction between the findings should therefore be considered hypothesis
effect of ExCR and the pre-defined subgroups sex, generating.
ejection fraction, NYHA functional class, HF etiology, We believe this to be the first IPD meta-analysis to
ethnicity, and baseline exercise capacity for either assess the impact of ExCR on HRQoL and exercise
HRQoL or exercise capacity (Figure 3, Online Tables capacity outcomes for patients with HF. The observed
2 to 6). beneficial effects of ExCR on these outcomes are
A differential effect of ExCR across ages was broadly consistent with previous trial-level (aggre-
observed in the standardized HRQoL score analysis at gate data) meta-analyses (8–10,49). The improvement
6-month follow-up, with a differential reduction in (reduction) in MLHFQ score was similar to that re-
HRQoL in the ExCR group compared with the control ported by the 2014 Cochrane meta-analysis (5.8;
group (i.e., an increase in standardized HRQoL score) 95% CI: 2.4 to 9.2) (8). The improvements in exercise
as age increased (0.006 SD units, 95% CI: 0.002 to capacity outcomes observed in our analyses were
0.011; p ¼ 0.006) (Online Table 3). To put this into lower than those seen in trial-level meta-analyses
context, based on an SD of 24 for MLHFQ score, this (6MWT 41.1 m; 95% CI: 16.7 to 53.6 m [31]; peak V O 2
equates to a mean increase of 1.4 in MLHFQ score (i.e., 2.79 ml/kg/min; 95% CI: 2.05 to 3.53 ml/kg/min) (9).
a reduction in HRQoL) for an increase of 10 years in We found no consistent evidence of HF patient sub-
patient age, in the ExCR group compared with the group effects, in accord with trial-level meta-regres-
control group. sion analyses (8,9). Within trial subgroup analyses
Interaction analyses for the 1-stage model at from the HF-ACTION trial found no differential effect
12 months showed differential effects of ExCR by sex, of ExCR on HRQoL across patient characteristics (50).
with women showing greater benefit from ExCR than A post hoc analysis of the same trial cohort reported a
men for each of peak V O 2 (0.57 ml/kg/min; 95% CI: significant interaction between ExCR and ethnic
0.04 to 1.11 ml/kg/min; p ¼ 0.036) (Online Table 6) group with regard to 6MWT distance at 3-month
and 6MWT (14.9 m; 95% CI: 1.2 to 28.7 m; p ¼ 0.034) follow-up (adjusted p ¼ 0.02), with mean improve-
(Online Table 4). Differential effects of ExCR were ment compared with control of 26 m (95% CI: 18 to 34
also seen between ethnic groups (Online Table 4); m) in white HF patients versus 11 m (95% CI: 0 to 21
white patients showed a greater improvement with m) in black HF patients, in the same direction as the
ExCR in 6MWT distance compared with nonwhite current study (51).
patients: 14.2 m (95% CI: 0.40 to 28.0; p ¼ 0.044). STUDY LIMITATIONS. IPD meta-analysis has a
number of strengths relative to traditional trial-level
DISCUSSION meta-analysis, including: reduction in ecological
biases; the ability to check and transform data to
We undertook an IPD meta-analysis to assess the common scores or measures; consistent methods of
impact of ExCR on exercise capacity and HRQoL in analysis across trials, and improved power to detect
patients with HF. Analyses of data from 13 trials in overall and subgroup effects. In this study, we used
JACC VOL. 73, NO. 12, 2019 Taylor et al. 1441
APRIL 2, 2019:1430–43 Exercise-Based HF Rehabilitation: HRQoL and Exercise Capacity

a 1-stage meta-analysis approach to compare the of HF patients in practice is approximately 10

outcomes between ExCR and control groups across years older (54).
all included trials. This approach adjusts the
CONCLUSIONS
between-group comparisons of outcomes at follow-
up for the baseline outcome score; this is impor-
Provision of ExCR to patients with HFrEF produces
tant here as many of the included studies were
clinically important benefits in HRQoL and exercise
small and therefore subject to chance differences in
capacity. Although we did observe some differences in
baseline score. Given these considerable advan-
the treatment effect of ExCR with age, sex, and
tages, meta-analyses that are based on IPD have
ethnicity, these subgroup effects were not consistent
been called the gold standard of systematic
across outcomes, time points, and analyses; hence, our
review (12).
findings do not endorse limiting ExCR interventions to
An increasingly recognized challenge of IPD meta-
subgroups of HF patients. However, due to the low
analysis is that of obtaining IPD from study in-
numbers of women and nonwhite patients partici-
vestigators (15,52). A recent systematic review across
pating in ExCR, the ExTraMATCH II study would sup-
a total of 122 IPD meta-analyses found the average
port the increasing representation of these groups.
meta-analysis located only 61% (95% CI: 46% to
These results, based on an IPD meta-analysis of ran-
74%) of eligible datasets (53). In this study, we were
domized trials, support the Class I recommendation of
able to retrieve patient data for all 13 trials with
current international clinical guidelines that ExCR
exercise capacity data; HRQoL data were available in
should be offered to all HF patients and the need to
9 of 13 (69%) trials for 89% (2,970 of 3,332) of par-
improve current poor uptake of ExCR in this popula-
ticipants. Although our level of data retrieval com-
tion. Future data collection in this field requires a
pares favorably with this recent systematic review,
consensus on the definition, collection, and reporting
we recognize that incomplete data capture is a lim-
of core outcomes, including a defined minimum stan-
itation of our study, which may have introduced bias
dardized set of outcomes that should be measured and
to our HRQoL analyses. Furthermore, we observed
reported in all clinical trials in specific areas of health
high levels of statistical heterogeneity for the out-
or health care (55). Additionally, we call for capture of
comes of MLHFQ and 6MWT, likely to be due to the
data on patient-level adherence to exercise training
variation in population and intervention character-
during the ExCR intervention period. Future trials
istics across the individual trials. Reassuringly, the
should be extended to include more women, older
inclusion of published results of trials for which no
patients, and more patients with HFpEF, as well as
IPD were available did not change main effects. Due
patients with comorbid conditions. More generally,
to limited published data on patient characteristics,
the research community should continue to imple-
we were unable to perform any sensitivity analyses
ment policies that encourage primary study authors to
using subgroup data.
make their datasets available, either by depositing
Further important limitations of this analysis
their datasets in publicly available repositories or
were the small number of patients with HFpEF
sharing with IPD meta-analysis collaborations when
that contributed to this analysis and the lack of
directly requested.
data on patient-level ExCR dose. We did not have
patient-level data on ExCR dose received, so we ACKNOWLEDGMENTS The authors acknowledge the

were unable to explore the effect of patient contribution of patient data from Dr. Rebecca Gary,
adherence to the rehabilitation program, or dura- Dr. Rainer Hambrecht, the late Dr. Romualdo Belar-
tion, frequency, or intensity of ExCR undertaken dinelli, and the late Dr. Pantaleo Giannuzzi. The au-
by an individual patient. Trials that include larger thors thank Tim Eames, Exeter Clinical Trials Unit, for
proportions of patients with HFpEF would enable his advice and support on data management for this
us to address the question of whether ExCR has a study.
differential effect in such patients compared to
those with HFrEF. Improved reporting of patient- ADDRESS FOR CORRESPONDENCE: Prof. Rod Taylor,
level data on adherence to ExCR will enable the Institute of Health Research, University of Exeter
investigation of any dose–response effect of ExCR. Medical School, College House, St. Luke’s Campus,
With regard to generalizability and application to Heavitree Road, Exeter EX1 2LU, United Kingdom.
clinical practice, the average age of participants in E-mail: [email protected]. Twitter: @ExeterECG,
this study was 61 years, whereas the average age @rod_ExeterECG, @Sarah1003Walker.
1442 Taylor et al. JACC VOL. 73, NO. 12, 2019

Exercise-Based HF Rehabilitation: HRQoL and Exercise Capacity APRIL 2, 2019:1430–43

PERSPECTIVES

COMPETENCY IN PATIENT CARE AND TRANSLATIONAL OUTLOOK: Future trials should

PROCEDURAL SKILLS: Meta-analysis of data from evaluate the effect of exercise-based cardiac rehabilita-
previous studies suggests that exercise-based cardiac tion in contemporary populations of patients with heart
rehabilitation improves health-related quality-of-life and failure.
exercise capacity in patients with heart failure, irrespec-
tive of patient characteristics.

REFERENCES

1. Braunwald E. The war against heart failure: the reviews of the effects of health care. Ann N Y Acad health status measure for heart failure. J Am Coll
Lancet lecture. Lancet 2015;385:812–24. Sci 1993;703:156–65. Cardiol 2000;35:1245–55.

2. Mosterd A, Hoes AW. Clinical epidemiology of 13. ExTraMATCH Collaborative. Exercise training 23. Guyatt GH, Nogradi S, Halcrow S, Singer J,
heart failure. Heart 2007;93:1137–46. meta-analysis of trials in patients with chronic Sullivan MJJ, Fallen EL. Development and testing
heart failure (ExTraMATCH). BMJ 2004;328:189. of a new measure of health status for clinical trials
3. British Association for Cardiovascular Preven-
in heart failure. J Gen Intern Med 1989;4:101–7.
tion and Rehabilitation. The BACPR Standards and 14. Stewart LA, Clarke M, Rovers M, et al.
Core Components for Cardiovascular Disease Pre- Preferred reporting items for a systematic review 24. O’Connor CM, Whellan DJ, Lee KL, et al. Effi-
vention and Rehabilitation 2017. 3rd edition. and meta-analysis of individual participant data: cacy and safety of exercise training in patients
Available at: http://www.bacpr.com/resources/ the PRISMA-IPD statement. JAMA 2015;313: with chronic heart failure: HF-ACTION randomized
6A7_BACR_Standards_and_Core_Components_2017. 1657–65. controlled trial. JAMA 2009;301:1439–50.
pdf. Accessed August 1, 2018. 25. Smart NA, Waldron M, Ismail H, et al. Valida-
15. Tierney J, Vale C, Riley R, et al. Individual
4. Dalal HM, Doherty P, Taylor RS. Cardiac reha- participant data (IPD) meta-analyses of rando- tion of a new tool for the assessment of study
bilitation. BMJ 2015:351. mised controlled trials: guidance on their use. quality and reporting in exercise training studies:
PLoS Med 2015;12:e1001855. TESTEX. Int J Evid Based Healthc 2015;13:9–18.
5. Yancy CW, Jessup M, Bozkurt B, et al. 2013
26. Higgins J, Thompson S. Quantifying hetero-
ACCF/AHA guideline for the management of heart 16. Taylor R, Piepoli M, Smart N, et al. Exercise
geneity in a meta-analysis. Stat Med 2002;21:
failure: executive summary. a report of the training for chronic heart failure (ExTraMATCH II):
1539–58.
American College of Cardiology Foundation/ protocol for an individual participant data meta-
American Heart Association Task Force on Practice analysis PROSPERO. 2014. Available at: http:// 27. Egger M, Smith GD, Schneider M, Minder C.
Guidelines. J Am Coll Cardiol 2013;62:1495–539. www.crd.york.ac.uk/PROSPERO/display_record.php? Bias in meta-analysis detected by a simple,
ID¼CRD42014007170. Accessed August 1, 2018. graphical test. BMJ 1997;315:629–34.
6. Kondamudi N, Haykowsky M, Forman DE,
Berry JD, Pandey A. Exercise training for preven- 17. Taylor RS, Piepoli MF, Smart N, et al. Exercise 28. Davidson PM, Cockburn J, Newton PJ, et al.
tion and treatment of heart failure. Prog Car- training for chronic heart failure (ExTraMATCH II): Can a heart failure-specific cardiac rehabilitation
diovasc Dis 2017;60:115–20. protocol for an individual participant data meta- program decrease hospitalizations and improve
analysis. Int J Cardiol 2014;174:683–7. outcomes in high-risk patients? Eur J Prev Cardiol
7. Fletcher GF, Landolfo C, Niebauer J, Ozemek C,
2010;17:393–402.
Arena R, Lavie CJ. Promoting physical activity and 18. Taylor RS, Walker S, Smart NA, et al. Impact of
exercise: JACC Health Promotion Series. J Am Coll exercise-based cardiac rehabilitation in patients 29. Austin J, Williams R, Ross L, Moseley L,
Cardiol 2018;72:1622–39. with heart failure (ExTraMATCH II) on mortality Hutchison S. Randomised controlled trial of car-
and hospitalisation: an individual patient data diac rehabilitation in elderly patients with heart
8. Taylor RS, Sagar VA, Davies EJ, et al. Exercise-
meta-analysis of randomised trials. Eur J Heart Fail failure. Eur J Heart Fail 2005;7:411–7.
based rehabilitation for heart failure. Cochrane
2018;20:1735–43. 30. Klecha A, Kawecka-Jaszcz K, Bacior B, et al.
Database Syst Rev 2014;4:CD003331.
Physical training in patients with chronic heart
19. Riley RD, Lambert PC, Abo-Zaid G. Meta-
9. Uddin J, Zwisler AD, Lewinter C, et al. Pre- failure of ischemic origin: effect on exercise ca-
analysis of individual participant data: rationale,
dictors of exercise capacity following exercise- pacity and left ventricular remodeling. Eur J Prev
conduct, and reporting. BMJ 2010;340:c221.
based rehabilitation in patients with coronary Cardiol 2007;14:85–91.
heart disease and heart failure: a meta-regression 20. Ahmed I, Sutton AJ, Riley RD. Assessment of
31. Myers J, Goebbels U, Dzeikan G, et al. Exercise
analysis. Eur J Prev Cardiol 2016;23:683–93. publication bias, selection bias, and unavailable
training and myocardial remodeling in patients
data in meta-analyses using individual participant
10. Smart N, Marwick TH. Exercise training for with reduced ventricular function: one-year
data: a database survey. BMJ 2012;344:d7762.
patients with heart failure: a systematic review of follow-up with magnetic resonance imaging. Am
factors that improve mortality and morbidity. Am 21. Bilbao A, Escobar A, García-Perez L, Navarro G, Heart J 2000;139 Part 1:252–61.
J Med 2004;116:693–706. Quirós R. The Minnesota Living with Heart Failure
32. Mueller L, Myers J, Kottman W, et al. Exercise
Questionnaire: comparison of different factor
11. Thompson S, Higgins J. How should meta- capacity, physical activity patterns and outcomes
structures. Health Qual Life Outcomes 2016;14:23.
regression analyses be undertaken and inter- six years after cardiac rehabilitation in patients
preted? Stat Med 2002;21:1559–73. with heart failure. Clin Rehabil 2007;21:923–31.
22. Green CP, Porter CB, Bresnahan DR,
12. Chalmers I. The Cochrane Collaboration: pre- Spertus JA. Development and evaluation of the 33. Belardinelli R, Georgiou D, Cianci G, Purcaro A.
paring, maintaining, and disseminating systematic Kansas City Cardiomyopathy Questionnaire: a new 10-year exercise training in chronic heart failure: a
JACC VOL. 73, NO. 12, 2019 Taylor et al. 1443
APRIL 2, 2019:1430–43 Exercise-Based HF Rehabilitation: HRQoL and Exercise Capacity

randomized controlled trial. J Am Coll Cardiol age with heart failure. Am J Cardiol 2005;95: 49. Ciani O, Piepoli M, Smart N, et al. Validation of
2012;60:1521–8. 1120–4. exercise capacity as a surrogate endpoint in
exercise-based rehabilitation for heart failure: a
34. Dracup K, Evangelista LS, Hamilton MA, et al. 41. Witham MD, Fulton RL, Greig CA, et al. Effi-
meta-analysis of randomized controlled trials.
Effects of a home-based exercise program on cacy and cost of an exercise program for func-
J Am Coll Cardiol HF 2018;6:596–604.
clinical outcomes in heart failure. Am Heart J tionally impaired older patients with heart failure:
2007;154:877–83. a randomized controlled trial. Circ Heart Fail 2012; 50. Flynn KE, Piña IL, Whellan DJ, et al. Effects of
5:209–16. exercise training on health status in patients with
35. Gary RA, Dunbar SB, Higgins MK,
chronic heart failure: HF-ACTION randomized
Musselman DL, Smith AL. Combined exercise and 42. Yeh GY, McCarthy EP, Wayne PM, et al. Tai chi
controlled trial. JAMA 2009;301:1451–9.
cognitive behavioral therapy improves outcomes exercise in patients with chronic heart failure: a
in patients with heart failure. J Psychosom Res randomized clinical trial. Arch Intern Med 2011;171: 51. Mentz RJ, Bittner V, Schulte PJ, et al. Race,
2010;69:119–31. 750–7. exercise training, and outcomes in chronic heart
failure: findings from Heart Failure–a Controlled
36. Giannuzzi P, Temporelli PL, Corra U, Tavazzi L. 43. Passino C, Severino S, Poletti R, et al. Aerobic
Trial Investigating Outcomes in Exercise
Antiremodeling effect of long-term exercise training decreases b-type natriuretic peptide
TraiNing (HF-ACTION). Am Heart J 2013;166:
training in patients with stable chronic heart fail- expression and adrenergic activation in patients with
488–95.
ure: results of the Exercise in Left Ventricular heart failure. J Am Coll Cardiol 2006;47:1835–9.
Dysfunction and Chronic Heart Failure (ELVD-CHF) 52. Stewart LA, Parmar MKB. Meta-analysis of the
44. Belardinelli R, Georgiou D, Cianci G, Purcaro A.
trial. Circulation 2003;108:554–9. literature or of individual patient data: is there a
Randomized, controlled trial of long-term mod-
difference? Lancet 1993;341:418–22.
37. Hambrecht R, Gielen S, Linke A, et al. Effects erate exercise training in chronic heart failure:
of exercise training on left ventricular function and effects on functional capacity, quality of life, and 53. Polanin JR. Efforts to retrieve individual
peripheral resistance in patients with chronic heart clinical outcome. Circulation 1999;99:1173–82. participant data sets for use in a meta-analysis
failure: a randomized trial. JAMA 2000;283: result in moderate data sharing but many data
45. McKelvie RS, Teo KK, Roberts R, et al. Effects
3095–101. sets remain missing. J Clin Epidemiol 2018;98:
of exercise training in patients with heart failure:
157–9.
38. Jolly K, Taylor RS, Lip GY, et al. A randomized the Exercise Rehabilitation Trial (EXERT). Am
Heart J 2002;144:23–30. 54. Cowie MR, Wood DA, Coats AJ, et al. Incidence
trial of the addition of home-based exercise to
and aetiology of heart failure; a population-based
specialist heart failure nurse care: the Birmingham 46. Willenheimer R, Rydberg E, Cline C, et al. Ef-
study. Eur Heart J 1999;20:421–8.
Rehabilitation Uptake Maximisation study for pa- fects on quality of life, symptoms and daily ac-
tients with Congestive Heart Failure (BRUM-CHF) tivity 6 months after termination of an exercise 55. Williamson PR, Altman DG, Blazeby JM, et al.
study. Eur J Heart Fail 2009;11:205–13. training programme in heart failure patients. Int J Developing core outcome sets for clinical trials:
Cardiol 2001;77:25–31. issues to consider. Trials 2012;13:132.
39. Nilsson BB, Westheim A, Risberg MA. Long-
term effects of a group-based high-intensity aer- 47. American Thoracic Society. Minnesota Living
obic interval-training program in patients with with Heart Failure Questionnaire 2004. Available KEY WORDS exercise capacity, heart
chronic heart failure. Am J Cardiol 2008;102: at: http://qol.thoracic.org/sections/instruments/ failure, MLHFQ, QoL, quality-of-life,
1220–4. ko/pages/mlwhfq.html. Accessed August 1, 2018. rehabilitation
40. Witham MD, Gray JM, Argo IS, Johnston DW, 48. Bohannon RW, Crouch R. Minimal clinically
Struthers AD, McMurdo ME. Effect of a seated important difference for change in 6-minute walk
exercise program to improve physical function and test distance of adults with pathology: a system- A PP END IX For supplemental figures and ta-
health status in frail patients > or ¼ 70 years of atic review. J Eval Clin Pract 2017;23:377–81. bles, please see the online version of this paper.