ORIGINAL RESEARCH ARTICLE

High-Intensity Interval Training in Patients With

Heart Failure With Reduced Ejection Fraction

ORIGINAL RESEARCH
ARTICLE
BACKGROUND: Small studies have suggested that high-intensity interval Øyvind Ellingsen, MD, PhD*
training (HIIT) is superior to moderate continuous training (MCT) in Martin Halle, MD*
Viviane Conraads, MD, PhD†
reversing cardiac remodeling and increasing aerobic capacity in patients Asbjørn Støylen, MD, PhD
with heart failure with reduced ejection fraction. The present multicenter Håvard Dalen, MD, PhD
Charles Delagardelle, MD
trial compared 12 weeks of supervised interventions of HIIT, MCT, or a Alf-Inge Larsen, MD, PhD
recommendation of regular exercise (RRE). Torstein Hole, MD, PhD
Alessandro Mezzani, MD, PhD
Emeline M. Van Craenenbroeck, MD, PhD
METHODS: Two hundred sixty-one patients with left ventricular ejection Vibeke Videm, MD, PhD
fraction ≤35% and New York Heart Association class II to III were randomly Paul Beckers, PhD
Jeffrey W. Christle, PhD
assigned to HIIT at 90% to 95% of maximal heart rate, MCT at 60% to 70% Ephraim Winzer, MD, PhD
of maximal heart rate, or RRE. Thereafter, patients were encouraged to Norman Mangner, MD
Felix Woitek, MD
continue exercising on their own. Clinical assessments were performed at Robert Höllriegel, MD
baseline, after the intervention, and at follow-up after 52 weeks. Primary Axel Pressler, MD
Tea Monk-Hansen, MD, PhD
end point was a between-group comparison of change in left ventricular Martin Snoer, MD, PhD
end-diastolic diameter from baseline to 12 weeks. Patrick Feiereisen, PhD
Torstein Valborgland, MD
John Kjekshus, MD, PhD
RESULTS: Groups did not differ in age (median, 60 years), sex (19% Rainer Hambrecht, MD
women), ischemic pathogenesis (59%), or medication. Change in left Stephan Gielen, MD
Trine Karlsen, PhD
ventricular end-diastolic diameter from baseline to 12 weeks was not Eva Prescott, MD, DMSc*
different between HIIT and MCT (P=0.45); left ventricular end-diastolic Axel Linke, MD*
For the SMARTEX Heart Failure Study
diameter changes compared with RRE were −2.8 mm (−5.2 to −0.4 mm; (Study of Myocardial Recovery After

P=0.02) in HIIT and −1.2 mm (−3.6 to 1.2 mm; P=0.34) in MCT. There Exercise Training in Heart Failure)
Group
was also no difference between HIIT and MCT in peak oxygen uptake *Drs Ellingsen, Halle, Prescott, and Linke
(P=0.70), but both were superior to RRE. However, none of these changes contributed equally.

was maintained at follow-up after 52 weeks. Serious adverse events were †Deceased.

not statistically different during supervised intervention or at follow-up at 52 Correspondence to: Øyvind Ellingsen,
MD, PhD, Department of Circulation and
weeks (HIIT, 39%; MCT, 25%; RRE, 34%; P=0.16). Training records showed Medical Imaging, Prinsesse Kristinas Gate
that 51% of patients exercised below prescribed target during supervised 3, 7030 Trondheim, PO Box 8905 MTFS,
NO-7491 Trondheim, Norway. E-mail
HIIT and 80% above target in MCT. oyvind.ellingsen@ntnu.no

Sources of Funding, see page 848

CONCLUSIONS: HIIT was not superior to MCT in changing left ventricular
remodeling or aerobic capacity, and its feasibility remains unresolved in Key Words: exercise ◼ heart failure

patients with heart failure. © 2017 The Authors. Circulation is

published on behalf of the American
Heart Association, Inc., by Wolters Kluwer
CLINICAL TRIAL REGISTRATION: URL: http://www.clinicaltrials.gov. Health, Inc. This is an open access article
Unique identifier: NCT00917046. under the terms of the Creative Commons
Attribution Non-Commercial-NoDervis
License, which permits use, distribution,
and reproduction in any medium, provided
that the original work is properly cited,
the use is noncommercial, and no
modifications or adaptations are made.

Circulation. 2017;135:839–849. DOI: 10.1161/CIRCULATIONAHA.116.022924 February 28, 2017 839

Ellingsen et al

discussions of whether HIIT should be included in stan-

Clinical Perspective dard care of patients with chronic heart failure.
This background formed the basis for a larger ran-
What Is New? domized controlled multicenter trial, the SMARTEX Heart
• The present multicenter trial did not confirm results Failure Study (Study of Myocardial Recovery After Exer-
from a small previous study that indicated that high- cise Training in Heart Failure), to test the hypothesis that
intensity interval training is superior to moderate HIIT is superior to MCT with regard to improvement of
continuous training (MCT) in reversing left ventricu- left ventricular dimensions and exercise capacity.
lar remodeling and increasing aerobic capacity.
• In both groups, results were only moderately bet-
ter than a recommendation of regular exercise; METHODS
improvements were not maintained at the 52-week
follow-up. Study Design
• Numeric differences in serious adverse events at 52 The SMARTEX Heart Failure Study is an investigator-initiated
weeks suggested a favor of MCT, but the study was randomized controlled clinical trial conducted at 9 European
not powered to compare safety. centers (Antwerp, Copenhagen, Leipzig, Luxembourg,
• Fifty-one percent of high-intensity interval training Munich, Stavanger, Trondheim/Levanger, Veruno, and
patients exercised below prescribed heart rate, and Ålesund) between June 2009 and July 2014. The final patient
80% of MCT exercised above their target. was randomized July 1, 2013, and had the 52-week follow-up
on July 22, 2014. The Clinical Trials database registration
What Are the Clinical Implications? reports 268 patients enrolled in the Web case report form
database. However, 7 randomizations were error entries dur-
• Given that high-intensity interval training was not ing initial testing and demonstration of the database. Thus,
superior to MCT in reversing remodeling or improv- the correct number of patients randomized was 261. The
ing secondary end points, and considering that trial was approved by the Regional Committee for Medical
adherence to the prescribed exercise intensity and Health Ethics of Central Norway and by national and local
based on heart rate may be difficult to achieve, even committees where required. Informed written consent was
when supervised and performed in centers experi- obtained from all participants. Details of rationale, design,
enced in cardiac rehabilitation, MCT remains the methods, sample size, randomization, and organization have
standard exercise modality for patients with chronic previously been published.8 Data management and statisti-
heart failure. cal analyses were performed by the coordinating center with
oversight by the steering committee (Ø.E., M.H., A.L., E.P.),
whose members had full access to all data and vouch for the

C
urrent guidelines recommend exercise training as accuracy and completeness of data and analyses.
an adjunctive therapy in patients with chronic heart
failure.1 A universal agreement on exercise prescrip- Patients and Interventions
tion does not exist; thus, an individualized approach, in- Patients were enrolled from outpatient heart failure clinics,
cluding behavioral characteristics, personal goals, and referrals to cardiac rehabilitation, public announcements, and
preferences, is recommended.2,3 At present, moderate screening of eligible patients in hospital registries. Eligible
continuous endurance exercise is the best described patients with symptomatic (New York Heart Association class
and established form of training because of its well- II–III), stable, pharmacologically optimally treated chronic
heart failure were randomized 1:1:1 to a 12-week program
demonstrated efficacy and safety.2 This advice is based
of HIIT, MCT, or recommendation of regular exercise (RRE),
mainly on a large multicenter exercise intervention trial stratified by study center and pathogenesis (ischemic versus
(HF-ACTION [Heart Failure: A Controlled Trial Investigat- nonischemic). Stratification by center was performed to avoid
ing Outcomes of Exercise Training]) with 2331 patients bias from unobserved treatment differences, and stratification
with heart failure, which observed a moderate reduction by pathogenesis was performed to allow possible post hoc
of symptoms, improvement of exercise capacity, and a analysis of the influence on left ventricular end-diastolic diam-
reduction of hospital readmissions for heart failure.4 eter (LVEDD) changes. Exercise training protocols have been
Exercise of high submaximal intensity performed in described elsewhere.5,8 Briefly, HIIT and MCT had 3 supervised
intervals of 1 to 4 minutes, also called high-intensity in- sessions per week on a treadmill or bicycle. HIIT included four
terval training (HIIT), has been tested in a small study of 4-minute intervals aiming at 90% to 95% of maximal heart rate
separated by 3-minute active recovery periods of moderate
patients with heart failure with reduced ejection fraction,
intensity. HIIT sessions lasted 38 minutes including warm-up
showing that HIIT was superior to moderate continuous
and cool-down at moderate intensity. MCT sessions aimed at
training (MCT) in improving exercise capacity, quality of 60% to 70% of maximal heart rate and lasted 47 minutes.
life, endothelial function, and left ventricular diameter HIIT and MCT sessions were estimated to obtain similar energy
and ejection fraction.5 The results were better than those expenditures.9 Patients randomized to RRE were advised to
observed in previous studies and meta-analyses of pa- exercise at home according to current recommendations and
tients with chronic heart failure.6,7 They also prompted attended a session of moderate-intensity training at 50% to

840 February 28, 2017 Circulation. 2017;135:839–849. DOI: 10.1161/CIRCULATIONAHA.116.022924

 High-Intensity Interval Training in Heart Failure

70% of maximal heart rate every 3 weeks.8 In all 3 groups, Scale 14. Physical activity was assessed with the International
there were no supervised training sessions after the 12-week Physical Activity Questionnaire (7-day short form), but its reli-
interventions, but the investigators had telephone contact with ability and validity have not been established in patients with
the participants every 4 weeks to register clinical events and heart failure, so the data have not been included. The measure-
to encourage physical activity. ments were taken at baseline, at 12 weeks immediately after
the intervention, and at follow-up 52 weeks from the start of
the training program.
Clinical Assessments
Screening procedures and clinical assessments before and
after exercise interventions were performed at local study Statistical Analysis
centers as previously described.8 Briefly, medical history, Power calculations for the main end point (comparison of

ORIGINAL RESEARCH
anthropometrics, physical examination including fasting blood the groups with respect to change in LVEDD from baseline to
sampling, quality-of-life questionnaires, cardiopulmonary exer- 12 weeks) have been detailed in a previous article on ratio-

ARTICLE
cise testing, and echocardiography were performed, in addi- nale and design.8 We estimated that a total number of 200
tion to prespecified substudies. patients, randomized 1:1:1 between RRE, MCT, and HIIT,
Echocardiography data were acquired according to stan- would be sufficient to detect a reduction of LVEDD of 3.0 mm
dard operation procedures of the study, stored digitally, and between HIIT and MCT and 5.0 mm between HIIT and RRE.
transferred as DICOM (digital imaging and communications in Calculations were based on LVEDD of 70 mm, coefficient
medicine) files or raw data to the core laboratory in Trondheim, of variance of 0.04, statistical power of 0.90, and value of
Norway. Analyses were performed by 1 of 2 expert echocar- P=0.05, adjusted for 3 comparisons by the Bonferroni cor-
diographers (A.S. and H.D.) blinded to group assignment but rection. Unless otherwise specified, data are presented as
not always to time point of assessment on EchoPAC SW (ver- median with 95% confidence interval of the median because
sion BT 11–13; GE Ultrasound, Horten, Norway). LVEDD was many variables were nonnormally distributed or as observed
measured at the tip of the mitral leaflet in the 2-dimensional numbers with percentages.
parasternal long-axis view.10 Repeatability was tested by Bland- The main end-point analysis was prespecified as mixed-
Altman analyses of the first 25 baseline assessments between models linear regression with robust standard errors, with
the 2 investigators. There was no bias (0.3 mm), and the coef- 12-week values used as outcome and baseline values used
ficient of variation was 4.1%. as adjustment variables, and included adjustments for center,
Cardiopulmonary exercise testing was performed with ischemic or nonischemic pathogenesis, and height. Model fit
standard equipment for indirect calorimetry in an incremen- was checked by residual plots, and estimated contrasts are
tal protocol until exhaustion on either a treadmill or a bicycle presented with 95% confidence intervals and P values cor-
ergometer, depending on exercise training equipment. The rected for 3 pairwise comparisons with the Scheffé method.
protocol comprised a 10- or 20-W increase in workload ≈1 For comparisons including the 52-week data, similar analyses
minute, starting at 20 or 40 W, respectively. For comparisons were performed. A 2-sided value of P≤0.05 was considered
per patient, the baseline, 12-week, and 52-week tests were statistically significant.
performed with the same protocol. The mean of the 3 highest To monitor adherence to training intensity, heart rate and
10-second consecutive measurements was identified as peak workload were recorded during training sessions, and aver-
oxygen uptake (Vo2peak). Respiratory quotient and other related age training intensity was calculated as percentage of maximal
values are reported from this time point. Cardiopulmonary heart rate at baseline. For HIIT and MCT patients, regression
exercise testing personnel were not blinded to assignment to models including average percentage of maximal heart rate
intervention group, but analysis was performed separately by during training (continuous variable or lowest versus highest
an independent investigator (P.B.). quartile), atrial fibrillation (yes versus no), or smoking (present
versus former/never) were developed.
End Points The study was not powered to assess differences in safety
or clinical events; therefore, SAEs were not a prespecified end
The primary end point was comparison of groups with respect
point.8 However, safety is an important concern in this popula-
to change in LVEDD from baseline to the 12-week assess-
tion, especially when performing exercise. With acknowledg-
ment by echocardiography. Key secondary end points were
ment of the limitations of post hoc analysis,11 χ2 tests for
change in left ventricular ejection fraction and Vo2peak; the latter
cardiovascular, noncardiovascular, and total SAEs during the
was also considered a measure of training effect. Safety was
training intervention period and during follow-up were per-
assessed by the rate of serious adverse events (SAEs) defined
formed with no corrections for multiple testing. Statistical
as all-cause and cardiovascular death, worsening heart failure
analyses were performed with Stata (version 13.1, StataCorp,
requiring hospitalization or intensified diuretic treatment, atrial
and ventricular arrhythmias, unstable angina, inappropriate College Station, TX).
implantable cardioverter-defibrillator shocks, and other events
leading to hospital admission or clinical evaluation. Events were
considered training related when occurring during or within 3
RESULTS
hours of supervised exercise training sessions. An indepen- Patient Population and Adherence to
dent blinded end-point committee (J.K., R.H., and S.G.) clas-
Intervention
sified all events. Quality of life was assessed with the Kansas
City Cardiomyopathy Questionnaire, the Hospital Anxiety and After initial exclusions and withdrawals, 231 patients
Depression Scale, the Global Mood Scale, and the Type D were included in HIIT, MCT, or RRE. Nine dropped

Circulation. 2017;135:839–849. DOI: 10.1161/CIRCULATIONAHA.116.022924 February 28, 2017 841

Ellingsen et al

Figure 1. Study enrollment, randomization, and follow-up.

Enrollment was stopped when it was estimated that at least 200 patients would complete the 12-week assessments according
to protocol. Two hundred fifteen patients came to follow-up assessments and were included in the intention-to-treat analysis; 207
of these were included in per-protocol analysis. Two hundred two patients came to the 52-week assessments and fulfilled the cri-
terion of having completed either echocardiography or cardiopulmonary exercise testing. LVEF indicates left ventricular ejection
fraction; and SAE, serious adverse event.

out because of SAEs, and 7 withdrew or were lost sessions, leaving 207 patients included in the per-
to follow-up (Figure 1). Two hundred fifteen patients protocol analysis that yielded equivalent results (data
were assessed after 12 weeks and were included in not shown).
the intention-to-treat analysis reported here. Median Baseline characteristics were similar in all groups, al-
adherence to supervised training was 35 (34–36) ses- though more RRE patients had a history of hypertension
sions of 36 possible in HIIT and MCT and 4 (3–4) of 4 (Table 1). Median age was 60 years (interquartile range
in RRE. Eight patients completed <24 of 36 exercise [IQR] 53–70 years); 71% were in New York Heart Asso-

842 February 28, 2017 Circulation. 2017;135:839–849. DOI: 10.1161/CIRCULATIONAHA.116.022924

 High-Intensity Interval Training in Heart Failure

Table 1.  Patient Characteristics at Baseline

Characteristics RRE (n=73) MCT (n=65) HIIT (n=77)
Age, y 60 (55–65) 60 (58–65) 65 (58–68)
Women, n (%) 14 (19) 12 (19) 14 (18)
Heart failure <12 mo, n (%) 14 (19) 7 (11) 14 (18)
NYHA class, n (%)
 II 54 (74) 41 (63) 55 (71)
 III 19 (26) 24 (37) 22 (29)

ORIGINAL RESEARCH
Left ventricular ejection fraction, % 30 (28–32) 29 (26–32) 29 (26–31)

ARTICLE
Ischemic origin, n (%) 41 (56) 39 (60) 46 (60)
 Previous myocardial infarction 32 (44) 36 (55) 44 (57)
 Previous CABG 17 (23) 14 (22) 20 (26)
 Previous PCI 33 (45) 23 (35) 32 (42)
Device therapy, n (%)
 Pacemaker 2 (3) 0 (0) 2 (3)
 Implantable cardioverter-defibrillator 31 (43) 38 (59) 27 (35)
 Cardiac resynchronization therapy 3 (4) 1 (2) 1 (1)
Atrial fibrillation, n (%)
 Chronic 6 (8) 8 (12) 14 (18)
 Paroxysmal 13 (18) 5 (8) 11 (14)
History of hypertension, n (%) 36 (49) 24 (37) 22 (29)
History of diabetes mellitus, n (%) 14 (19) 21 (32) 16 (21)
History of COPD, n (%) 4 (6) 8 (12) 4 (5)
Current smoking, n (%) 35 (48) 32 (49) 38 (49)
Alcohol drinks per week, n 1 (1–2) 2 (1–3) 1 (1–2)
Medications, n (%)
 ACE inhibitor/ARB 70 (96) 60 (92) 71 (92)
 β-Blocker 71 (97) 61 (94) 73 (95)
 Aldosterone receptor antagonist 39 (53) 34 (52) 49 (64)
 Diuretic 51 (70) 49 (75) 58 (75)
 Digoxin or digitoxin 6 (8) 8 (12) 17 (22)
 Statin 45 (62) 47 (72) 50 (65)
Body mass index, kg/m2 27.7 (25.7–28.3) 27.5 (26.6–29.7) 27.6 (26.3–28.7)
Systolic blood pressure, mm Hg 120 (116–124) 119 (112–122) 115 (110–120)
Diastolic blood pressure, mm Hg 75 (70–80) 73 (70–80) 71 (70–87)
NT-proBNP, ng/L 895 (635–1110) 976 (725–1348) 1052 (837–1472)
Values are median with 95% confidence interval of the median, because most of the characteristics were nonnormally
distributed, or number (percent) as indicated. There were no significant differences between the groups except for history
of hypertension (χ2 test, P=0.04). ACE indicates angiotensin converting enzyme; ARB, angiotensin receptor blockers; CABG,
coronary artery bypass graft; COPD, chronic obstructive pulmonary disease; HIIT, high-intensity interval training; MCT,
moderate continuous training; NT-proBNP, N-terminal prohormone of brain natriuretic peptide; PCI, percutaneous coronary
intervention; and RRE, recommended regular exercise.

ciation class II, and the rest were in class III. All patients 24%–34%), and median Vo2peak was 17.1 mL·kg−1·min−1
were considered to be on optimal medical treatment. (IQR, 14.2–20.3 mL·kg−1·min−1) with no difference be-
Only 19% of the patients were women (Table 1). Median tween groups at baseline (Table 2 and Tables I and II in
left ventricular ejection fraction at baseline was 29% (IQR, the online-only Data Supplement).

Circulation. 2017;135:839–849. DOI: 10.1161/CIRCULATIONAHA.116.022924 February 28, 2017 843

Ellingsen et al

Table 2.  Main Echocardiography and Cardiopulmonary Testing Measures at Baseline, 12 weeks, and 52
Weeks With Unadjusted Changes
RRE (n=73) MCT (n=65) HIIT (n=77)
Baseline 12 wk 52 wk Baseline 12 wk 52 wk Baseline 12 wk 52 wk
LVEDD, mm 68 69 66 69 67 64 68 63 63
(67 to 69) (65 to 71) (63 to 67) (66 to 72) (65 to 70) (61 to 70) (65 to 70) (62 to 68) (62 to 66)
LVEF, % 30 28 28 29 27 33 29 31 28
(28 to 32) (27 to 30) (27 to 32) (26 to 32) (25 to 31) (26 to 37) (26 to 31) (29 to 31) (26 to 32)
Vo2peak, 18.4 17.4 18.2 16.2 17.0 16.4 16.8 18.2 17.1
mL·kg−1·min−1 (16.8 to 19.6) (15.7 to 19.8) (15.8 to 20.0) (15.3 to 18.7) (15.7 to 19.6) (15.0 to 18.6) (15.8 to 17.8) (16.3 to 20.0) (15.5 to 18.6)
NT-proBNP, 895 821 626 976 821 698 1052 909 813
ng/L (635 to 1110) (594 to 1079) (419 to 1116) (725 to 1348) (580 to 1169) (544 to 1021) (837 to 1472) (722 to 1484) (585 to 1615)
Change Baseline to Baseline to Baseline to Baseline to Baseline to Baseline to
12 wk 52 wk 12 wk 52 wk 12 wk 52 wk
 LVEDD, mm 0.0 −2.0 −1.0 −3.0 −2.0 −3.0
(0.0 to 2.0) (−4.0 to 0.0) (−2.0 to 1.0) (−7.0 to −1.4) (−3.6 to −1.0) (−5.0 to −1.0)
 LVEF, % −0.6 1.1 0.7 0.7 1.7 −0.2
(−2.4 to 1.4) (−0.8 to 3.0) (−1.8 to 2.6) (−1.5 to 4.4) (0.0 to 4.5) (−3.1 to 2.8)
 Vo2peak, mL· −0.1 −0.4 1.1 1.2 0.9 0.1
kg−1·min−1 (−0.9 to 0.4) (−1.3 to 0.4) (0.5 to 1.7) (−0.2 to 1.4) (0.0 to 1.4) (−0.4 to 1.0)
 NT-proBNP, 9 −25 2 −101 19 112
ng/L (−43 to 112) (−108 to 76) (−91 to 97) (−130 to 30) (−76 to 129) (−24 to 236)
Values are median with 95% confidence interval of the median. There were no differences between the groups at baseline (Kruskal-Wallis test, P=0.68,
0.83, 0.21 and 0.30). Additional echocardiography and cardiopulmonary testing outcomes are presented online (Tables I and II in the online-only Data
Supplement). HIIT indicates high-intensity interval training; LVEDD, left ventricular end-diastolic diameter; LVEF, left ventricular ejection fraction; MCT,
moderate continuous training; NT-proBNP, N-terminal prohormone of brain natriuretic peptide; RRE, recommended regular exercise; and Vo2peak, peak
oxygen uptake.

Training Intensity Compared With Protocol Targets between-group differences in changes in LVEDD from
Heart rate and workload were monitored at all centers dur- baseline to 12 weeks (primary end point), are given in
ing the supervised training sessions (Figure 2). Average Table 3. Change of LVEDD in HIIT was not significantly
heart rate during sessions remained unchanged in the 12 different from that in MCT (−1.2 mm; −3.6 to 1.2 mm;
weeks of supervised training, indicating constant relative P=0.45) but larger than in RRE (−2.8 mm; −5.2 to
exercise intensity during interventions (Figure 2A). Work- −0.4 mm; P=0.02), whereas the change in MCT was
load during intervals in the HIIT group was consistently not significantly different from the change in RRE (−1.6
33 W (IQR, 24–42 W; P<0.001), higher than during con- mm; −4.2 to 1.1 mm; P=0.34; Table 3). There were no
tinuous exercise in MCT (Figure 2B). Median relative train- other significant differences in echocardiographic mea-
ing intensity based on maximal heart rate was 90% (IQR, surements or in prohormone of brain natriuretic peptide
88%–92%) in HIIT and 77% (74%–82%) in MCT. Thus, the (Table 3 and Table I in the online-only Data Supplement).
difference in training intensity was only 10% (8%–13%) with Change in Vo2peak in HIIT was not significantly dif-
adjustment for center and pathogenesis (Figure 2C) com- ferent from MCT (−0.4 mL·kg−1·min−1; −1.7 to 0.8
pared with the protocol target difference of 27.5%. The mL·kg−1·min−1; P=0.70) but was 1.4 mL·kg−1·min−1
training records showed that 51% of the patients in the (0.2−2.6 mL·kg−1·min−1; P=0.02) larger than in RRE.
HIIT group exercised at a lower intensity than prescribed, Change in Vo2peak was 1.8 mL·kg−1·min−1 (0.5−3.0
whereas 80% of those in the MCT group trained at a higher mL·kg−1·min−1; P=0.003) larger in MCT compared with
intensity than the protocol target (Figure 2D). RRE (Table 3 and Table II in the online-only Data Supple-
ment). There were no differences in respiratory quo-
tient between groups at Vo2peak at baseline, 12 weeks,
Echocardiography and Cardiopulmonary or 1 year, indicating similar levels of effort during test-
Exercise Testing ing (Table II in the online-only Data Supplement). At the
Table 2 presents the crude within-group changes in 1-year follow-up, there were no differences in primary or
main results from baseline to 12 and 52 weeks. Re- secondary end points between the groups (Table 3 and
sults of the prespecified primary analyses, the adjusted Tables I–III in the online-only Data Supplement).

844 February 28, 2017 Circulation. 2017;135:839–849. DOI: 10.1161/CIRCULATIONAHA.116.022924

 High-Intensity Interval Training in Heart Failure

ORIGINAL RESEARCH
ARTICLE
Figure 2. Training intensity during the 12-week intervention.
A, Heart rate during training. Average heart rate during the 12-week intervention, estimated as weekly mean (SD) during moder-
ate continuous training (MCT) and during the last 2 minutes of high-intensity interval training (HIIT). Constant difference between
groups: 16 bpm (10–22 bpm; P<0.001). B, Workload. Average workload estimated as for heart rate. Difference between
groups: 33 W (24–42 W; P<0.001). C, Training intensity. Average relative training intensity (percentage of maximal heart rate)
estimated as for heart rate: HIIT, 90% (88%–92%); MCT, 77% (74%–82%); difference, 10% (8%–13%; P<0.001). Some of the
variability in estimated training intensity probably results from variation in maximal heart rate. Comparing baseline and follow-
up assessments in individual patients revealed differences that seemed randomly distributed and independent of intervention
group, center, and whether the patients had sinus rhythm or atrial fibrillation (data not shown). Shaded areas mark boundaries of
prescribed training intensity: HIIT, 90% to 95%; MCT, 60% to 70%. D, Training intensity on target. Distribution of average training
intensity during the 12-week intervention; MCT, left histogram; HIIT, right histogram. Shaded areas mark boundaries for pre-
scribed training intensity. Fifty-one percent of HIIT patients exercised below their prescribed training intensity, and 80% of MCT
patients exercised above theirs. Density scales the height of the bars so that the sum of their areas equals 1.00.

Sensitivity analyses exploring factors that might was also not associated with atrial fibrillation (P=0.22).
have influenced the changes of LVEDD in response An alternative model for LVEDD changes from baseline
to exercise did not identify predictors of response. to 12 weeks excluding patients with atrial fibrillation
Change of LVEDD in HIIT and MCT was not associated gave results comparable to the results from the model
with average percentage of maximal heart rate dur- with all patients, albeit with slightly larger effect sizes
ing supervised training sessions when added to the for HIIT versus RRE (−3.7 mm; −6.7 to −0.8 mm;
regression model (P=0.52) or when used to substitute P=0.009) and for HIIT versus MCT (−2.0; −4.9 to 0.9
for intervention group (P=0.24). Likewise, findings mm; P=0.23). Smoking was not significantly associ-
were similar when comparing patients in the highest ated with change in LVEDD (P=0.26). There was no dif-
and lowest quartiles of achieved percentage of maxi- ference in exercise intensity assessed as percentage
mal heart rate during training and when using these of maximal heart rate during sessions between cen-
quartiles as a categorical variable. Change in LVEDD ters (P=0.61) or between training on treadmill (89%;

Circulation. 2017;135:839–849. DOI: 10.1161/CIRCULATIONAHA.116.022924 February 28, 2017 845

Ellingsen et al

Table 3.  Main Outcomes

HIIT vs MCT HIIT vs RRE MCT vs RRE
Adjusted contrasts, baseline to 12 wk
 LVEDD, mm −1.2 (−3.6 to 1.2) −2.8 (−5.2 to −0.4)* −1.6 (−4.2 to 1.1)
 LVEF, % 1.5 (−2.1 to 5.1) 2.5 (−0.5 to 5.5) 0.9 (−2.7 to 4.6)
  Vo2peak, mL·kg ·min
−1 −1 −0.4 (−1.7 to 0.8) 1.4 (0.2 to 2.6)* 1.8 (0.5 to 3.0)†
 NT-proBNP, ng/L −95 (−729 to 538) −52 (−489 to 384) 43 (−500 to 587)
Adjusted contrasts, baseline to 52 wk
 LVEDD, mm −0.1 (−2.9 to 2.7) −0.7 (−3.7 to 2.4) −0.5 (−3.6 to 2.5)
 LVEF, % −1.3 (−3.7 to 1.1) −0.3 (−2.8 to 2.1) 0.9 (−1.7 to 3.6)
  Vo2peak, mL·kg ·min
−1 −1 0.1 (−1.8 to 2.0) −0.3 (−2.3 to 1.6) −0.4 (−2.3 to 1.5)
 NT-proBNP, ng/L −6 (−528 to 517) 33 (−462 to 529) 39 (−491 to 569)
Between-group comparisons calculated as adjusted contrasts from models adjusted for center, pathogenesis, and height and
adjusted for multiple comparisons with the Scheffé procedure. Values are median with 95% confidence interval of the median.
HIIT indicates high-intensity interval training; LVEDD, left ventricular end-diastolic diameter; LVEF, left ventricular ejection fraction;
MCT, moderate continuous training; NT-proBNP, N-terminal prohormone of brain natriuretic peptide; RRE, recommended regular
exercise; and Vo2peak, peak oxygen uptake.
*P=0.02.
†P=0.003.

82%–92%) versus bicycle (85%; 83%–88%; P=0.20), Supplement. Three events occurred during or within 3
whether HIIT and MCT were analyzed jointly or sepa- hours of supervised exercise in the HIIT group. One pa-
rately (data not shown). tient had ventricular arrhythmia with cardiac arrest dur-
ing supervised exercise in week 1, was successfully
resuscitated, and stopped the exercise program. This
Quality of Life patient had refused cardioverter-defibrillator implanta-
There were no within-group or between-group differenc- tion before inclusion. Another patient had inappropri-
es in the quality-of-life measures Kansas City Cardiomy- ate implantable cardioverter-defibrillator discharge
opathy Questionnaire, Hospital Anxiety and Depression unrelated to arrhythmia during supervised exercise in
Scale, Global Mood Scale, or Type D Scale 14 at base- week 12 and stopped the exercise program. A third
line, 12 weeks, or 52 weeks (Table III in the online-only patient experienced dizziness within 3 hours after su-
Data Supplement). pervised exercise, with no detectable cardiovascular
cause, and continued the exercise program without
any reoccurrences.
Serious Adverse Events
There were no statistically significant differences be-
tween groups in total number of patients with SAEs or DISCUSSION
cardiovascular SAEs during the 12-week intervention, al- The present study is the first randomized multicenter
though SAEs were numerically higher in HIIT, followed by trial evaluating HIIT in chronic heart failure with reduced
MCT and RRE (Table 4). During the follow-up period from ejection fraction. It compares HIIT with the 2 most prev-
week 13 to 52, there was a possible trend (uncorrect- alent exercise prescriptions: a supervised program of
ed P=0.10) for more patients admitted to hospital with MCT or RRE. The main finding was that 12 weeks of HIIT
cardiovascular events in HIIT (n=19) and RRE (n=17) was not superior to MCT with respect to left ventricu-
compared with MCT (n=8), mainly because of fewer ad- lar reverse remodeling assessed as change in LVEDD.
missions for heart failure worsening in MCT (Table 4). Although there was a statistically significant difference
This was also reflected in the 52-week total number of in remodeling between HIIT and RRE at 12 weeks, im-
patients with SAEs: HIIT, 32 (39%); RRE, 26 (34%); and mediately after the supervised exercise intervention, its
MCT, 18 (25%; P=0.16). The corresponding number of clinical importance is uncertain.
fatal events at 52 weeks was as follows: HIIT, 3; RRE, 1; The effects of HIIT were less than expected from our
and MCT 3. working hypothesis8 and from a previous study by Wis-
Details of diagnoses and time of events, including løff et al5 on which it was based. In the present study,
multiple diagnoses or multiple admissions in single pa- change in LVEDD by HIIT was −2.8 mm relative to RRE
tients, are reported in Table IV in the online-only Data compared with −4.5 mm from our working hypothesis

846 February 28, 2017 Circulation. 2017;135:839–849. DOI: 10.1161/CIRCULATIONAHA.116.022924

 High-Intensity Interval Training in Heart Failure

Table 4.  Serious Adverse Events and −7.7 mm reported by Wisløff et al.5 In contrast,
the change in LVEDD was −1.6 mm with MCT, which is
RRE MCT HIIT
(n=76), (n=73), (n=82), similar to our prestudy estimate of −1.5 mm8 and −0.9
Events* n (%) n (%) n (%) mm observed in the previous study.5 Post hoc analy-
ses indicated that the magnitude of change in LVEDD
Cardiovascular, weeks 1–12 5 (7) 6 (8) 9 (11)
may be larger with sinus rhythm compared with atrial
 Fatal 0 1 0 fibrillation, but the present study was not powered to
 Ventricular arrhythmia, life 0 1 1 undertake a formal comparison. Change in LVEDD was
threatening chosen as the primary end point to represent change
in cardiac remodeling because the echocardiograph-

ORIGINAL RESEARCH
 Ventricular arrhythmia, other 0 0 1
ic measurement of diameter is more robust and has
 Worsening heart failure 2 3 4
less relative variation than left ventricular volume es-

ARTICLE
 Other nonfatal 3 1 3 timates.10 The present study indicated nonsignificant
Cardiovascular, weeks 13–52 17 (22) 8 (11) 19 (23) changes in LVEDV of −19 mL with HIIT and −13 mL
 Fatal 0 0 2 with MCT. This is in contrast to the large change of
−45.2 mL previously observed with HIIT but similar
 Ventricular arrhythmia, life 2 1 1
threatening
to −15.2 mL observed with MCT5 and to a −11.5-mL
change with MCT in a meta-analysis.7 The change in
 Ventricular arrhythmia, other 2 1 3 Vo2peak in the present study was 1.4 mL·kg−1·min−1 with
 Worsening heart failure 13 3 11 HIIT and 1.8 mL·kg−1·min−1 with MCT relative to RRE.
 Other nonfatal 4 3 4 This is markedly less than the previously observed 6.0
Noncardiovascular, weeks 1–12 2 (3) 3 (4) 6 (7)
mL·kg−1·min−1 with HIIT but similar to 1.9 mL·kg−1·min−1
with MCT5 and 2.1 mL·kg−1·min−1 in a meta-analysis.12
 Fatal 0 1 0 Nonetheless, it is slightly larger than 0.6 mL·kg−1·min−1
 Nonfatal 2 2 6 in the large HF-ACTION trial.4 It is notable that the dif-
Noncardiovascular, weeks 13–52 7 (9) 2 (3) 3 (4) ferences observed at 12 weeks, immediately after the
 Fatal 1 1 1
supervised interventions, were not maintained at the
52-week assessment, suggesting that the exercise
 Nonfatal 6 1 2 prescriptions had not been followed as closely in the
Total, weeks 1–12 7 (9) 9 (12) 14 (17) unsupervised period as in the supervised period.
 Fatal 0 2 0 Several factors may account for the small effect size
 Nonfatal 7 7 14
of HIIT in the present study. Multicenter studies are more
resistant to bias known to affect the study effects, espe-
Total, weeks 13–52 22 (29) 10 (14) 22 (27) cially compared with a single-center setting. Differences
 Fatal 1 1 3 in the study population between our trial and the previ-
 Nonfatal 21 9 19 ous one by Wisløff and coworkers5 might also contrib-
ute to the different effects: All patients in the previous
Total, weeks 1–52 26 (34) 18 (25) 32 (39)
study had ischemic heart failure, although fewer than
 Cardiovascular 21 (28) 13 (18) 24 (29) a third had a previous revascularization compared with
 Noncardiovascular 8 (11) 5 (7) 9 (11) 60% in the present study. Furthermore, patients in the
HIIT indicates high-intensity interval training; MCT, moderate continuous study by Wisløff et al5 were on average 15 years older
training; and RRE, recommended regular exercise. There was no significant and had a much lower Vo2peak at baseline compared with
difference between the groups during the 12-week training intervention the present trial.
in terms of cardiovascular, noncardiovascular, or total number patients with Another factor that may have reduced the response
serious adverse effects (χ2 test, P=0.61, 0.37, and 0.33, respectively). During to HIIT is lower training intensity during intervals. In our
the 13- to 52-week follow-up, there was a trend for higher numbers of patients trial, 51% of the HIIT patients unexpectedly trained at
with cardiovascular events in HIIT compared with MCT (χ2 test, P=0.10) as a lower intensity than prescribed, whereas 80% in the MCT
result of fewer hospitalizations for worsening of heart failure in the MCT group group trained at higher intensity. Despite the experience
but not compared with the RRE group. This same trend is also reflected in of cardiac rehabilitation of the participating centers, HIIT
the number of cardiovascular events during weeks 1 to 52 and the total
at 90% to 95% of maximal heart rate appeared difficult
number of events during weeks 13 to 52 and 1 to 52 (χ2 test, P=0.21,
0.06, and 0.16, respectively; P values not corrected for multiple tests).
to achieve in a multicenter cohort of patients, whereas
*Number of patients (percent) with serious adverse effects, defined as fatal intensities at 60% to 70% of maximal heart rate in MCT
events and events leading to hospitalization or clinical evaluation. Patients with seemed too low. It may be speculated that the response
multiple diagnoses or multiple events are counted only once; thus, accumulated to HIIT might have been larger if target intensity had
data are sometimes less than the respective sums. A detailed list of diagnoses been achieved; however, we could not detect any dif-
and time of events is presented in Table IV in the online-only Data Supplement. ference of change in LVEDD when comparing the upper

Circulation. 2017;135:839–849. DOI: 10.1161/CIRCULATIONAHA.116.022924 February 28, 2017 847

Ellingsen et al

and lower quartiles of percentage of maximal heart rate Conclusions

during exercise in the present study. The present multicenter trial did not confirm the hypoth-
To date, only 1 study has had sufficient statistical
esis that a 12-week program of supervised HIIT was
power to assess safety with exercise training in patients
superior to MCT in reducing left ventricular remodeling
with heart failure. The HF-ACTION study demonstrated a
in patients with stable heart failure. None of the inter-
moderate reduction in the composite end point of mor-
ventions led to deterioration of cardiac function com-
tality and hospital admissions after 1 year.4 By design,
pared with RRE, and both exercise programs increased
the present study was too small to assess differences
aerobic capacity, an important prognostic parameter of
in safety among HIIT, MCT, and RRE8; however, numeric
heart failure, to a similar extent. However, these posi-
differences in clinical events could generate hypotheses
tive changes were smaller than expected and were not
and identify issues for special attention in future studies
maintained at follow-up after 52 weeks. Numeric differ-
and follow-up. During the 12-week supervised training
ences in readmissions for worsening of heart failure
program, the number of patients with SAEs was small
suggested a favor of MCT relative to HIIT and RRE, but
in all groups, and few patients withdrew from exercise
the study was not powered to assess safety. Train-
training. These observations concur with comparisons of
ing records showed that exercise intensities >90% of
safety of HIIT versus MCT in patients with coronary ar-
maximal heart rate were not achieved in a significant
tery disease that detected no differences in SAEs.13,14 In
proportion of the patients. Thus, further studies are
contrast, there was a numeric difference in patients with
needed to define the role of HIIT as an alternative exer-
SAEs during the follow-up period from week 13 to 52 as
a result of more hospitalizations for worsening of heart cise modality in patients with heart failure with reduced
failure in HIIT and RRE compared with MCT. This trend ejection fraction.
was not statistically significant and resulted from post
hoc subgroup analyses because SAEs were not a pre-
specified end point. Hence, conclusions or recommenda- ACKNOWLEDGMENTS
tions could not be based on this finding, but the numeric Jennifer Adam, Elena Bonanomi, Silvia Colombo, Christian
difference should receive attention in future trials. Have Dall, Ingrid Granøien, Kjersti Gustad, Anne Haugland, Ju-
lie Kjønnerød, Marti Kristiansen, Jorun Nilsen, Maren Redlich,
Anna Schlumberger, and Kurt Wuyts performed exercise test-
Limitations ing and training; Rigmor Bøen, Marianne Frederiksen, Eli Gran-
One of the important objectives of moving from a small viken, Loredana Jakobs, Adnan Kastrati, Nadine Possemiers,
Hanne Rasmussen, Liv Rasmussen, and Johannes Scherr per-
proof-of-principle study to a type II multicenter study was
formed patient screening, inclusion, and clinical assessments;
to test whether the effect size would be conserved in
Volker Adams, Ann-Elise Antonsen, Wim Bories, Geert Frederix,
a setting that is closer to a real-world clinical setting. Malou Gloesner, Vicky Hoymans, and Hielko Miljoen collected
Although several measures were taken to ensure qual- data; Hanna Ellingsen and Maria Henningsen performed data
ity and consistency, including supervised training ses- monitoring; and Lars Køber and Christian Torp-Pedersen moni-
sions based on heart rate monitoring, the differences in tored safety.
training intensity between HIIT and MCT were less than
intended and partly overlapped. This was an unexpected
finding, suggesting that the HIIT prescription of 90% to SOURCES OF FUNDING
95% of maximal heart rate may be too high and the MCT
This work was supported by St. Olavs Hospital; Faculty of Med-
prescription of 60% to 70% too low for some patients.
icine, Norwegian University of Science and Technology; Nor-
For future studies, we suggest that exercise intensities wegian Health Association; Danish Research Council; Central
should be regularly adapted to improvements in exercise Norwegian Health Authority/Norwegian University of Science
capacity and to worsening of symptoms or changes of and Technology; Else-Kröner-Fresenius-Stiftung, and Société
medication. Repeated assessment of maximal heart rate Luxembourgeoise pour la recherche sur les maladies cardio-
and more emphasis on adjusting workload according to vasculaires.
perceived level of effort might also be helpful. We ex-
perienced that questionnaires were of limited value for
assessing physical activity outside supervised sessions DISCLOSURES
and recommend accelerometer recordings, particularly
Dr Halle reports grants from the Else-Kröner-Fresenius Founda-
in the unsupervised follow-up period. Furthermore, in tion for the present work and is on the advisory board of Novar-
future studies, women should be a focus because only tis, Sanofi-Aventis, and Merck Sharp & Dohme Corp. outside of
19% of the patients in this study were women. Although the present study. Dr Linke reports grants and personal fees
not unusual in similar studies, this sex bias was unin- from Medtronic and Claret Medical, as well as personal fees
tended and constitutes a limitation of the generalization from Edwards, St. Jude Medical, Bard, and Symetis, all outside
of the results. of the present study.

848 February 28, 2017 Circulation. 2017;135:839–849. DOI: 10.1161/CIRCULATIONAHA.116.022924

 High-Intensity Interval Training in Heart Failure

AFFILIATIONS ease (part III). Eur J Prev Cardiol. 2012;19:1333–1356. doi:

10.1177/2047487312437063.
From St. Olavs Hospital, Trondheim University Hospital, Norway 3. Taylor RS, Sagar VA, Davies EJ, Briscoe S, Coats AJ, Dalal H,
(Ø.E., A.S., H.D., V.V., T.K.); K.G. Jebsen Center for Exercise in Lough F, Rees K, Singh S. Exercise-based rehabilitation for heart
Medicine, Department of Circulation and Medical Imaging, Nor- failure. Cochrane Database Syst Rev. 2014;4:CD003331.
wegian University of Science and Technology, Trondheim, Nor- 4. O’Connor CM, Whellan DJ, Lee KL, Keteyian SJ, Cooper LS, Ellis
way (Ø.E., H.D., T.K.); Department of Prevention, Rehabilitation SJ, Leifer ES, Kraus WE, Kitzman DW, Blumenthal JA, Rendall DS,
and Sports Medicine, Technische Universität München, Klinikum Miller NH, Fleg JL, Schulman KA, McKelvie RS, Zannad F, Piña
rechts der Isar, Germany (M.H., J.W.C., A.P.); DZHK (German IL; HF-ACTION Investigators. Efficacy and safety of exercise train-
ing in patients with chronic heart failure: HF-ACTION randomized
Center for Cardiovascular Research), Partner Site Munich Heart
controlled trial. JAMA. 2009;301:1439–1450. doi: 10.1001/
Alliance, Munich, Germany (M.H.); Else-Kröner-Fresenius Preven- jama.2009.454.

ORIGINAL RESEARCH
tion Center, Klinikum rechts der Isar, Munich, Germany (M.H.); 5. Wisløff U, Støylen A, Loennechen JP, Bruvold M, Rognmo Ø, Haram
Antwerp University Hospital, Edegem, Belgium (V.C., E.M.V.C., PM, Tjønna AE, Helgerud J, Slørdahl SA, Lee SJ, Videm V, Bye A,

ARTICLE
P.B.); University of Antwerp, Belgium (V.C., E.M.V.C., P.B.); De- Smith GL, Najjar SM, Ellingsen Ø, Skjaerpe T. Superior cardiovas-
partment of Circulation and Medical Imaging (A.S., T.H.) and cular effect of aerobic interval training versus moderate continu-
Department of Laboratory Medicine, Children’s and Women’s ous training in heart failure patients: a randomized study. Circu-
Health (V.V.), NTNU–Norwegian University of Science and Tech- lation. 2007;115:3086–3094. doi: 10.1161/CIRCULATIONAHA.
nology, Trondheim, Norway; Department of Medicine, Levanger 106.675041.
Hospital, Nord-Trøndelag Hospital Trust, Norway (H.D.); Centre 6. Giannuzzi P, Temporelli PL, Corrà U, Tavazzi L; ELVD-CHF Study
Group. Antiremodeling effect of long-term exercise training in
Hospitalier de Luxembourg, Luxembourg (C.D., P.F.); Depart-
patients with stable chronic heart failure: results of the Exercise
ment of Cardiology, Stavanger University Hospital, Norway (A.-
in Left Ventricular Dysfunction and Chronic Heart Failure (ELVD-
I.L., T.V.); Department of Clinical Science, University of Bergen, CHF) Trial. Circulation. 2003;108:554–559. doi: 10.1161/01.
Norway (A.-I.L., T.V.); Ålesund Hospital, Møre og Romsdal Health CIR.0000081780.38477.FA.
Trust, Norway (T.H.); Cardiac Rehabilitation Division, Salvatore 7. Haykowsky MJ, Liang Y, Pechter D, Jones LW, McAlister FA,
Maugeri Foundation IRCCS, Scientific Institute of Veruno, Italy Clark AM. A meta-analysis of the effect of exercise training on
(A.M.); Division of Cardiovascular Medicine, Stanford Center for left ventricular remodeling in heart failure patients: the benefit
Inherited Cardiovascular Disease, CA (J.W.C.); Department of depends on the type of training performed. J Am Coll Cardiol.
Cardiology, Herzzentrum, Universität Leipzig, Germany (E.W., 2007;49:2329–2336. doi: 10.1016/j.jacc.2007.02.055.
N.M., F.W., R. Höllriegel, A.L.); Department of Cardiology, Bis- 8. Støylen A, Conraads V, Halle M, Linke A, Prescott E, Ellingsen Ø.
pebjerg Hospital, University of Copenhagen, Denmark (T.M.-H., Controlled study of myocardial recovery after interval training in
heart failure: SMARTEX-HF: rationale and design. Eur J Prev Car-
M.S., E.P.); University of Oslo, Rikshospitalet University Hos-
diol. 2012;19:813–821. doi: 10.1177/1741826711403252.
pital, Norway (J.K.); Department of Cardiology and Angiology, 9. Rognmo Ø, Hetland E, Helgerud J, Hoff J, Slørdahl SA. High
Klinikum Links der Weser, Bremen, Germany (R. Hambrecht); intensity aerobic interval exercise is superior to moderate in-
and Department of Cardiology, Angiology and Intensive Care, tensity exercise for increasing aerobic capacity in patients
Klinikum Lippe, Detmold, Germany (S.G.). with coronary artery disease. Eur J Cardiovasc Prev Rehabil.
2004;11:216–222.
10. Thorstensen A, Dalen H, Amundsen BH, Aase SA, Stoylen A. Re-
FOOTNOTES producibility in echocardiographic assessment of the left ventricu-
lar global and regional function, the HUNT study. Eur J Echocar-
Received April 9, 2016; accepted December 15, 2016. diogr. 2010;11:149–156. doi: 10.1093/ejechocard/jep188.
The online-only Data Supplement is available with this arti- 11. Brookes ST, Whitely E, Egger M, Smith GD, Mulheran PA, Peters
cle at http://circ.ahajournals.org/lookup/suppl/doi:10.1161/ TJ. Subgroup analyses in randomized trials: risks of subgroup-
CIRCULATIONAHA.116.022924/-/DC1. specific analyses; power and sample size for the interaction
Circulation is available at http://circ.ahajournals.org. test. J Clin Epidemiol. 2004;57:229–236. doi: 10.1016/j.jcline-
pi.2003.08.009.
12. Haykowsky MJ, Timmons MP, Kruger C, McNeely M, Taylor DA,
Clark AM. Meta-analysis of aerobic interval training on exercise
REFERENCES capacity and systolic function in patients with heart failure and
1. Ponikowski P, Voors AA, Anker SD, Bueno H, Cleland JG, Coats AJ, reduced ejection fractions. Am J Cardiol. 2013;111:1466–1469.
Falk V, González-Juanatey JR, Harjola VP, Jankowska EA, Jessup M, doi: 10.1016/j.amjcard.2013.01.303.
Linde C, Nihoyannopoulos P, Parissis JT, Pieske B, Riley JP, Rosano 13. Conraads VM, Pattyn N, De Maeyer C, Beckers PJ, Coeckel-
GM, Ruilope LM, Ruschitzka F, Rutten FH, van der Meer P; Authors/ berghs E, Cornelissen VA, Denollet J, Frederix G, Goetschalckx
Task Force Members. 2016 ESC guidelines for the diagnosis and K, Hoymans VY, Possemiers N, Schepers D, Shivalkar B, Voigt
treatment of acute and chronic heart failure: the Task Force for the JU, Van Craenenbroeck EM, Vanhees L. Aerobic interval train-
diagnosis and treatment of acute and chronic heart failure of the ing and continuous training equally improve aerobic exercise
European Society of Cardiology (ESC) developed with the special capacity in patients with coronary artery disease: the SAINTEX-
contribution of the Heart Failure Association (HFA) of the ESC. Eur CAD study. Int J Cardiol. 2015;179:203–210. doi: 10.1016/j.
Heart J. 2016;37:2129–2200. doi: 10.1093/eurheartj/ehw128. ijcard.2014.10.155.
2. Vanhees L, Rauch B, Piepoli M, van Buuren F, Takken T, Börjes- 14. Rognmo Ø, Moholdt T, Bakken H, Hole T, Mølstad P, Myhr NE, Grims-
son M, Bjarnason-Wehrens B, Doherty P, Dugmore D, Halle M; mo J, Wisløff U. Cardiovascular risk of high- versus moderate-in-
Writing Group, EACPR. Importance of characteristics and mo- tensity aerobic exercise in coronary heart disease patients. Circu-
dalities of physical activity and exercise in the management lation. 2012;126:1436–1440. doi: 10.1161/CIRCULATIONAHA.
of cardiovascular health in individuals with cardiovascular dis- 112.123117.

Circulation. 2017;135:839–849. DOI: 10.1161/CIRCULATIONAHA.116.022924 February 28, 2017 849