High‐Intensity Interval Training Is Associated With Improved 10‐Year Survival by Mediating Left Ventricular Remodeling in Patients With Heart Failure With Reduced and Mid‐Range Ejection Fraction

Abstract

Background

This study aimed to assess the left ventricular (LV) remodeling response and long‐term survival after high‐intensity interval training (HIIT) in patients with various heart failure (HF) phenotypes during a 10‐year longitudinal follow‐up.

Methods and Results

Among 214 patients with HF receiving guideline‐directed medical therapy, those who underwent an additional 36 sessions of aerobic exercise at alternating intensities of 80% and 40% peak oxygen consumption ($\dot{\mathrm{V}}$O_2peak) were considered HIIT participants (n=96). Patients who did not undergo HIIT were considered participants receiving guideline‐directed medical therapy (n=118). Participants with LV ejection fraction (EF) <40%, ≥40% and <50%, and ≥50% were considered to have HF with reduced EF, HF with mid‐range EF, and HF with preserved EF, respectively. $\dot{\mathrm{V}}$O_2peak, serial LV geometry, and time to death were recorded. In all included participants, 10‐year survival was better (P=0.015) for participants who underwent HIIT (80.3%) than for participants receiving guideline‐directed medical therapy (68.6%). An increased $\dot{\mathrm{V}}$O_2peak, decreased minute ventilation carbon dioxide production slope, and reduced LV end‐diastolic diameter were protective factors against all‐cause mortality. Regarding 138 patients with HF with reduced EF (P=0.044) and 36 patients with HF with mid‐range EF (P=0.036), 10‐year survival was better for participants who underwent HIIT than for participants on guideline‐directed medical therapy. Causal mediation analysis showed a significant mediation path for LV end‐diastolic diameter on the association between HIIT and 10‐year mortality in all included patients with HF (P<0.001) and those with LV ejection fraction <50% (P=0.006). HIIT also had a significant direct association with 10‐year mortality in patients with HF with LV ejection fraction <50% (P=0.027) but not in those with LV ejection fraction ≥50% (n=40).

Conclusions

Reversal of LV remodeling after HIIT could be a significant mediating factor for 10‐year survival in patients with HF with reduced EF and those with HF with mid‐range EF.

Nonstandard Abbreviations and Acronyms

2D

2‐dimensional

aHR

adjusted hazard ratio

CMA

causal mediation analysis

CO_ex

cardiac output during exercise

CPET

cardiopulmonary exercise test

Da–vO₂

systemic arteriovenous O₂ difference

GDMT

guideline‐directed medical therapy

HFmrEF

heart failure with mid‐range ejection fraction

HFpEF

heart failure with preserved ejection fraction

HFrEF

heart failure with reduced ejection fraction

HIIT

high‐intensity interval training

LVEDD

left ventricular end‐diastolic diameter

LVESD

left ventricular end‐systolic diameter

$\dot{\mathbf{V}}$CO₂

minute carbon dioxide production

$\dot{\mathbf{V}}$ _E

minute ventilation

$\dot{\mathbf{V}}$O₂

minute oxygen consumption

$\dot{\mathbf{V}}$O_2peak

peak oxygen consumption

Clinical Perspective.

What Is New?

• This 10‐year retrospective cohort study demonstrated that high‐intensity interval training is associated with a favorable prognosis for patients with heart failure (HF) by mediating the left ventricular remodeling process.

• Varied long‐term survival was observed in patients with different HF phenotypes.

• High‐intensity interval training–induced reversal of left ventricular remodeling was not identified in patients with HF with preserved ejection fraction.

What Are the Clinical Implications?

• High‐intensity interval training may reverse left ventricular remodeling to improve long‐term survival for patients with HF with reduced and mid‐range ejection fraction.

• However, this nonpharmacological mediation path for left ventricular geometry on the association between high‐intensity interval training and long‐term survival in patients with HF with preserved ejection fraction requires further investigation.

Heart failure (HF), a clinical syndrome consisting of orthopnea, peripheral edema, and fatigue, is caused by structural and/or functional abnormalities of the heart. ¹ In the past decades, HF mortality was high, with 1‐year mortality of approximately 20% and 5‐year mortality of up to 50% to 60%. ² In the United States, the total cost for HF was estimated to be $30.7 billion in 2012. Projections suggest that by 2030 the economic burden of HF will increase by 127% to $69.8 billion. ³ Therefore, care for patients with HF has become a great challenge in modern medicine, owing to its high mortality and growing medical costs.

High‐intensity interval training (HIIT) is characterized by exercising at ≥80% of one's peak oxygen consumption ($\dot{\mathrm{V}}$O_2peak) interspersed with 40% to 50% of $\dot{\mathrm{V}}$O_2peak to allow for recovery. ⁴ HIIT‐induced alternating aerobic and anaerobic metabolic responses ⁵ have been shown to increase aerobic capacity, ⁴ improve red blood cell deformability, ⁶ decrease cardiac chamber dimensions, ⁴ , ⁷ reduce cardiac fibrosis, ⁸ and increase 8‐month survival ⁷ in patients with HF. Thus, HIIT has been advocated as a beneficial therapy for patients with HF. ⁴

Different HF phenotypes include HF with reduced ejection fraction (HFrEF), HF with mid‐range EF (HFmrEF), and HF with preserved EF (HFpEF) based on the 2‐dimensional (2D) echocardiographic assessment of left ventricular EF (LVEF). ¹ Several reports have demonstrated favorable changes of cardiac geometry after HIIT in patients with HFrEF ⁴ , ⁷ ; however, available observations regarding associations between HIIT and LV geometry as well as the long‐term prognosis of patients with HFmrEF and those with HFpEF are limited. ¹

A typical causal mediation analysis (CMA) seeks to identify the mechanism underlying the observed relationship between an independent and a dependent variable by including a third hypothetical mediator. Therefore, this tool allows investigators to discover the potential mechanism of treatment and address competing explanations. ⁹ As HIIT has many health benefits for patients with HF, ⁴ , ⁷ we hypothesized a favorable long‐term clinical outcome for patients with HF after HIIT. The present work also highlighted the mediating role of HIIT‐associated alterations in LV geometry in the survival of patients with HF during long‐term follow‐up.

Methods

Participants

The data that support the findings of this study are available from the corresponding author upon reasonable request. The institutional review board of a tertiary care hospital approved this retrospective cohort study protocol and waived the documentation of informed consent. Patients with typical HF signs and symptoms ¹⁰ and New York Heart Association functional classification II to IV, who were hospitalized in a tertiary care hospital because of acute cardiogenic pulmonary congestion ¹¹ between January 1, 2009 and May 31, 2022, were initially surveyed. All primarily included patients with HF had stable clinical presentations for ≥4 weeks and received individualized patient education under optimized guideline‐directed medical therapy (GDMT) by our HF care team. One‐on‐one education with the case manager was initiated once a patient was enrolled. The HF care team provided individualized education related to HF and methods for self‐monitoring, optimized guideline‐based HF medication, and further laboratory assessments. ¹¹ In the primarily included patients, we excluded individuals younger than 20 years and those older than 80 years, or with pregnancy, cardiac transplantation within the next 6 months, moderate to severe chronic obstructive pulmonary disease, decompensated HF, noncardiac disease prohibiting cycling exercise, or absolute contraindications for exercise suggested by the American College of Sports Medicine. ¹² Patients with HF with an estimated glomerular filtration rate of <30 mL/min per 1.73 m² were also excluded because most of them were under hemodialysis and could not have regular exercise training.

Clinical Assessment

We recorded baseline age, sex, body mass index, comorbidities, serum B‐type natriuretic peptide levels, incremental cardiopulmonary exercise test (CPET) findings, and 2D echocardiography measurements in all participants. The physical and mental component scores of the Medical Outcomes Study 36‐item Short Form were used to evaluate quality of life before initiating CPET.

Exercise Training

Among the included patients with HF, those who underwent an additional 36 sessions (2 or 3 sessions weekly) of a supervised bicycle ergometer (Ergoselect 150P, Ergoline GmbH) training as in the previous protocol ⁴ were classified as HIIT participants. They exercised alternatively at 3‐minute intervals of 80% $\dot{\mathrm{V}}$O_2peak and 3‐minute intervals of 40% $\dot{\mathrm{V}}$O_2peak for 30 minutes in each session. The other participants who were unwilling to have supervised HIIT were classified as participants with GDMT.

Echocardiography

Baseline 2D echocardiography images of all participants were acquired at end‐expiration using a 2‐ to 5‐MHz tightly curved array ultrasound transducer (Vivid 7, General Electric Healthcare, or Phillips IE33, Philips Healthcare) to measure the left ventricular end‐diastolic diameter (LVEDD), left ventricular end‐systolic diameter (LVESD), and LVEF for all participants. ¹³ Among all enrolled participants with HIIT and GDMT, patients with HF who had LVEF <40%, ≥40% and <50%, and ≥50% were considered to have HFrEF, HFmrEF, and HFpEF, respectively. ¹

Cardiopulmonary Exercise Test

All participants underwent incremental CPET within 1 week before HIIT. ⁴ Minute ventilation ($\dot{\mathrm{V}}$ _E), carbon dioxide ventilation ($\dot{\mathrm{V}}$CO₂), and oxygen consumption ($\dot{\mathrm{V}}$O₂) were measured breath‐by‐breath using a computer‐based system (CareFusion MasterScreen CPX, CareFusion Corp.). $\dot{\mathrm{V}}$O_2peak, $\dot{\mathrm{V}}$ _E–$\dot{\mathrm{V}}$CO₂ slope, systemic arteriovenous O₂ difference (Da–vO₂), and peak cardiac output during the exercise (CO_ex) were defined as described in a previous protocol ⁴ and the American College of Sports Medicine guidelines. ¹⁴ The detailed CPET procedure is presented in Data S1.

Follow‐Up

All participants were followed up until May 31, 2022 or until death during the observational period. HIIT participants with different HF phenotypes underwent secondary CPET, quality‐of‐life assessment, and 2D echocardiography within 1 week after completing 36 sessions of HIIT. After completing the exercise training, the HIIT participants received the same treatment program as the GDMT participants until the end of follow‐up. For all participants, 2D echocardiography was performed during the long‐term follow‐up at an interval of 3 to 12 months. The dates and causes of death were also documented.

Heatmaps for Geometric Changes During Follow‐Up

Heatmaps were used to present changes of cardiac geometry in different HF phenotypes after HIIT during the long‐term follow‐up. The change in cardiac geometry was defined by the following equation, where _F/U indicates follow‐up:

$$\text{ΔLVEF}={\text{LVEF}}_{\mathrm{F}/\mathrm{U}}–{\text{LVEF}}_{\text{baseline}}$$ (1)

$$\text{ΔLVEDD}={\text{LVEDD}}_{\mathrm{F}/\mathrm{U}}–{\text{LVEDD}}_{\text{baseline}}$$ (2)

$$\text{ΔLVESD}={\text{LVESD}}_{\mathrm{F}/\mathrm{U}}–{\text{LVESD}}_{\text{baseline}}$$ (3)

Statistical Analysis

Data are presented as mean (95% CI) or number (percentage). Baseline demographic information and differences in LV geometric changes between the HIIT and GDMT participants with different HF phenotypes were compared using Student t test for continuous variables, and χ² test was conducted to assess categorical variables. In addition, clinical information assessed in the study, including cardiorespiratory fitness, LV geometry, and questionnaire for quality of life, were considered as possible confounders according to our serial works ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ¹⁵ and other prior publications ¹⁶ , ¹⁷ for related HF studies.

Regarding HIIT participants, ANOVA was used to estimate the differences in cardiorespiratory fitness and LV geometry after exercise training among the 3 HF phenotypes. Furthermore, differences in the above evaluations between 2 of the 3 phenotypes were estimated using the Bonferroni post hoc test. A paired t test was used to assess alterations of cardiorespiratory fitness and LV geometry for each HF phenotype after HIIT. To mitigate the risk of accepting results for a policy decision that may be false‐positive, generalized estimating equation with adjustment for prespecified confounders was performed due to multiple observations for each patient. Estimation of physiological adaptations of patients with different HF phenotypes after HIIT and associations between HIIT and the above variables in our included participants were also processed with adjustment for confounders.

The primary outcome measure was the change in $\dot{\mathrm{V}}$O_2peak measured after the HIIT in patients with different HF phenotypes. ⁸ With a mean SD value of the main criterion of 0.2, statistical power (1‐β) of 80%, and α value of 0.05 in a bilateral hypothesis, we estimated that to obtain a 20% change in $\dot{\mathrm{V}}$O_2peak with HIIT versus GDMT with G*Power software, ¹⁸ we needed ≥10 participants in each group.

To avoid immortal time bias, the index date for HIIT participants was the date of exercise completion, whereas the index date for GDMT participants was the date of the initial CPET examination. ⁷ Kaplan–Meier survival curves for patients with HF based on their exercise status and the different HF phenotypes were assessed using log‐rank test. Moreover, we performed multivariable Cox regression adjusted for potential confounders to investigate the clinical predictors of the 10‐year survival. LVEDD was selected for linear regression after conducting data exploration (Data S1) for the adjusted multivariable Cox model.

A CMA was performed to examine the potential mediating role of LVEDD measured during follow‐up in the association between HIIT at the index date and long‐term survival. The conceptual model depicting a path diagram is shown in the Figure S1. Mediational analyses were performed by using the lavaan package ¹⁹ of R software (R version 4.0.3). The lavaan package was used with the outcome variable considered as a binary variable. The path of indirect association was estimated by the delta method (or the Sobe method) ²⁰ based on the framework proposed by Baron and Kenny. ²¹ The average causal mediation analysis and average direct associations were calculated with the CIs of path coefficients estimated by the quasi‐Bayesian Monte Carlo method based on normal approximation. ²² A series of univariate and multiple linear regression were performed to identify potential mediation pathways. Since distinct clinical similarities between HFmrEF and HFrEF have been reported, ²³ participants were further collapsed into patients with HF with LVEF <50% and those with LVEF ≥50% for CMA. LVEDD findings of all participants at the index date were included in the mediation analyses. Statistical analyses were performed using R software and SAS version 9.4 (SAS Institute Inc). All statistical assessments were 2‐tailed and considered significant at P<0.05.

Results

Patient Characteristics for Different Phenotypes

Among the 973 patients with HF, 330 were initially screened and 214 eligible candidates were included in the study. Of the enrolled patients, 118 patients with HF (73 with HFrEF, 19 with HFmrEF, and 26 with HFpEF) who were unwilling to undergo additional HIIT were considered GDMT participants. The remaining 96 patients with HF (65 with HFrEF, 17 with HFmrEF, and 14 with HFpEF) who completed an additional 36 sessions of supervised HIIT for 3 to 4 months were considered HIIT participants. All HIIT participants exercised at a prescribed intensity during the supervised exercise training period. After completing exercise training, all HIIT participants received the same GDMT and were monitored by our HF care team until the end of the follow‐up (Figure 1). No significant differences in baseline demographics were observed between the HIIT and GDMT participants before phenotyping (Table 1).

Figure 1. Flow diagram of participant selection and follow‐up.

Patients with stable heart failure (HF) signs and symptoms were primarily included and further screened by exclusion criteria. Cardiopulmonary exercise test and 2‐dimensional (2D) echocardiographic examination were assessed for the high‐intensity interval training (HIIT) participants within 1 week after completing 36 sessions of exercise trainings (immediate physiological adaptations to HIIT in red transparent panel). The same guideline‐directed medical therapy (GDMT) program was applied for HIIT participants after ceasing HIIT and GDMT participants during follow‐up. Serial 2D echocardiographic examinations were performed for all included patients with HF during the 10‐year follow‐up to evaluate the chronic clinical outcomes after HIIT (green transparent panel) on patients with HF. HFmrEF indicates heart failure with mid‐range ejection fraction; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction; and LVEF, left ventricular ejection fraction.

Table 1.

Baseline Demographics of All Included Patients With HF With and Without HIIT

Variable	GDMT	HIIT	P value
	(n=118)	(n=96)
Age, y	59.8 (57.9–61.8)	59.5 (56.8–62.2)	0.808
Women/men, n	34/84	39/67	0.824
BMI, kg/m2	25.0 (24.4–25.7)	25.7 (24.8–26.6)	0.227
HF duration, mo	8.18 (5.32–11.0)	6.34 (3.06–9.62)	0.405
Follow‐up duration, y	6.72 (6.23–7.21)	6.50 (6.09–6.91)	0.512
SBP, mm Hg	127 (123–132)	123 (118–127)	0.166
DBP, mm Hg	78 (75–81)	75 (72–78)	0.175
HR, beats per min	81 (78–83)	78 (75–80)	0.109
LVEF, %	38.9 (36.0–41.8)	34.6 (31.5–37.7)	0.051
BNP, pg/mL	695 (503–887)	686 (532–840)	0.941
$\dot{\mathrm{V}}{\mathrm{O}}_{2\text{peak}}$, MET	5.14 (4.89–5.39)	4.83 (4.61–5.06)	0.080
${\dot{\mathrm{V}}}_{\mathrm{E}}–{\dot{\mathrm{V}}}_{\mathrm{CO}2}$ slope	33.3 (31.8–34.9)	34.5 (32.9–36.2)	0.293
NYHA functional class, n
I	10	3	0.306
II	58	56
III	48	35
IV	2	2
SF‐36
PCS	48.1 (46.6–49.6)	46.3 (44.6–47.9)	0.428
MCS	45.1 (43.2–47.0)	44.6 (42.7–46.6)	0.428
Cause, n (%)
CAD	61 (52)	44 (46)	0.394
Cardiomyopathy	20 (17)	25 (26)	0.105
Comorbidities, n (%)
Hypertension	63 (53)	50 (52)	0.849
Arrhythmia	40 (34)	26 (27)	0.283
Diabetes	43 (36)	35 (36)	0.998
Hyperlipidemia	42 (36)	27 (28)	0.245
Medication, n (%)
ACEI/ARB	97 (82)	81 (84)	0.673
β‐Blocker	96 (81)	77 (80)	0.832
CCB	17 (14)	20 (21)	0.216
Diuretics	73 (62)	50 (52)	0.150
MRA	17 (14)	19 (20)	0.295

Values are mean (95% CI) or number (percentage), unless otherwise indicated.

ACEI indicates angiotensin‐converting enzyme inhibitor; ARB, angiotensin receptor blocker; BMI, body mass index; BNP, B‐type natriuretic peptide; CAD, coronary artery disease; CCB, calcium channel blocker; DBP, diastolic blood pressure; GDMT, guideline‐directed medical therapy; HF, heart failure; HIIT, high‐intensity interval training; HR, heart rate; LVEF, left ventricular ejection fraction; MCS, mental component score; MET, metabolic equivalent (3.5 mL O₂/min/kg); MRA, mineralocorticoid receptor antagonist; NYHA, New York Heart Association; PCS, physical component score; SBP, systolic blood pressure; SF‐36, Medical Outcomes Study Short‐Form 36; $\dot{\mathrm{V}}$CO₂, minute carbon dioxide production; $\dot{\mathrm{V}}$ _E, minute ventilation; and $\dot{\mathrm{V}}$O_2peak, peak oxygen consumption.

Both the HIIT and GDMT participants were further classified according to HF phenotypes. Baseline clinical characteristics were similar between the HIIT and GDMT participants with each HF phenotype, except for the greater prevalence of atrial fibrillation (P=0.018) in the GDMT than in the HIIT participants and the greater prevalence of type 2 diabetes (P=0.034) in the HIIT than in the GDMT participants in those with HFmrEF (Table 2).

Table 2.

Baseline Demographics of Patients With Different HF Phenotypes

Variable	HFrEF		P value	HFmrEF		P value	HFpEF		P value
	GDMT	HIIT		GDMT	HIIT		GDMT	HIIT
	(n=73)	(n=65)		(n=19)	(n=17)		(n=26)	(n=14)
Age, y	57.8 (55.5–60.2)	59.7 (56.5–63.0)	0.890	60.1 (54.3–65.9)	61.1 (54.4–67.8)	0.824	65.3 (62.0–68.7)	66.2 (59.7–72.7)	0.797
Women/men, n	16/57	14/51	0.957	8/11	9/8	0.516	10/16	6/8	0.787
BMI, kg/m2	25.5 (24.7–26.4)	25.5 (24.3–26.6)	0.413	23.8 (22.2–25.3)	26.1 (23.9–28.4)	0.092	26.4 (24.7–28.0)	26.3 (24.6–28.1)	0.988
HF duration, mo	11.0 (6.6–15.4)	8.0 (3.3–12.8)	0.371	5.78 (2.54–9.03)	3.50 (1.04–5.95)	0.287	2.06 (1.34–2.78)	1.99 (1.02–2.95)	0.900
Follow‐up duration, y	7.62 (6.81–8.42)	7.79 (7.17–8.41)	0.743	7.20 (8.45–8.95)	7.01 (6.09–7.93)	0.435	9.03 (7.90–10.2)	8.99 (7.86–10.12)	0.962
SBP, mm Hg	123 (118–129)	120 (114–125)	0.389	133 (124–143)	127 (115–139)	0.408	135 (126–143)	133 (122–143)	0.775
DBP, mm Hg	77 (73–81)	76 (72–79)	0.574	80 (72–88)	76 (70–82)	0.410	80 (76–84)	74 (66–81)	0.142
HR, beats per min	82 (78–85)	78 (75–81)	0.115	83 (79–87)	77 (73–82)	0.094	76 (71–81)	76 (68–84)	0.901
LVEF, %	28.5 (26.9–30.0)	26.1 (24.0–28.3)	0.081	44.5 (42.9–46.0)	44.3 (42.9–45.7)	0.862	64.1 (60.9–67.3)	62.3 (58.0–66.7)	0.515
BNP, pg/mL	765 (524–1005)	792 (594–990)	0.864	485 (170–801)	491 (202–781)	0.979	429 (209–649)	351 (188–514)	0.581
$\dot{\mathrm{V}}{\mathrm{O}}_{2\text{peak}}$, MET	5.11 (4.79–5.44)	4.91 (4.63–5.19)	0.361	5.31 (4.73–5.90)	4.88 (4.36–5.39)	0.284	5.08 (4.57–5.59)	4.42 (4.61–6.09)	0.109
${\dot{\mathrm{V}}}_{\mathrm{E}}{\dot{–\mathrm{V}}}_{\mathrm{CO}2}$ slope	34.0 (32.0–36.1)	34.8 (32.8–36.8)	0.612	33.0 (30.9–35.1)	31.2 (29.3–33.1)	0.240	31.5 (28.4–34.6)	37.5 (31.6–43.4)	0.061
NYHA functional class
I	5	2	0.633			0.532	5	1	0.592
II	33	35		14	14		11	7
III	33	26		5	3		10	6
IV	2	2
SF‐36
PCS	47.3 (45.4–49.3)	46.5 (44.4–48.7)	0.459	48.9 (45.6–52.2)	44.4 (41.2–47.7)	0.330	49.7 (46.9–52.6)	47.5 (44.0–50.9)	0.424
MCS	43.8 (41.3–46.2)	45.7 (43.4–47.9)	0.484	49.0 (45.1–52.8)	44.2 (39.0–49.5)	0.208	45.8 (41.8–49.9)	40.2 (35.5–45.0)	0.424
Cause, n (%)
CAD	38 (52)	29 (45)	0.383	10(53)	9 (53)	0.985	13 (50)	6 (43)	0.666
Cardiomyopathy	18 (25)	18 (28)	0.685	2 (11)	6 (35)	0.074	0 (0)	1 (7)	0.168
Comorbidities, n (%)
Hypertension	36 (49)	31 (48)	0.849	10 (53)	13 (76)	0.137	17 (65)	6 (43)	0.445
AF	19 (26)	16 (25)	0.849	13 (68)	4 (24)	0.018*	8 (31)	6 (43)	0.445
Diabetes	24 (33)	22 (34)	0.582	2 (11)	7 (41)	0.034*	13 (50)	6 (43)	0.666
Hyperlipidemia	24 (33)	15 (23)	0.202	3 (16)	7 (41)	0.090	15 (58)	5 (36)	0.185
Medication, n (%)
ACEI/ARB	64 (88	57 (88)	0.997	17 (89)	14 (82)	0.537	16 (62)	10 (71)	0.730
β‐Blocker	64 (88)	54 (83)	0.601	13 (68)	13 (76)	0.590	19 (73)	10 (71)	1.000
CCB	9 (12)	12 (18)	0.317	5 (26)	4 (24)	0.847	3 (12)	4 (29)	0.176
Diuretics	50 (68)	34 (52)	0.077	10 (53)	6 (35)	0.296	13 (50)	10 (71)	0.191
MRA	14 (19)	16 (25)	0.440	2 (11)	3 (18)	0.498	1 (4)	0 (0)	0.651

Values were given as mean (95% CI) or number (percentage), unless otherwise indicated.

ACEI indicates angiotensin converting enzyme inhibitor; AF, atrial fibrillation; ARB, angiotensin receptor blocker; BMI, body mass index; BNP, B‐type natriuretic peptide; CAD, coronary artery disease; CCB, calcium channel blocker; DBP, diastolic blood pressure; GDMT, guideline‐directed medial therapy; HF, heart failure; HFmrEF, heart failure with mid‐range ejection fraction; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction; HIIT, high‐intensity interval training; HR, heart rate; LVEF, left ventricular ejection fraction; MCS, mental component score; MET, metabolic equivalent (3.5 mL O₂/min/kg); MRA, mineralocorticoid receptor antagonist; NYHA, New York Heart Association; PCS, physical component score; SBP, systolic blood pressure; SF‐36, Medical Outcomes Study 36‐Item Short Form; ${\dot{\mathrm{V}}}_{\mathrm{CO}2}$, minute carbon dioxide production; ${\dot{\mathrm{V}}}_{\mathrm{E}}$, minute ventilation; and $\dot{\mathrm{V}}$O_2peak, peak oxygen consumption.

^*Statistical differences were assessed by χ² test.

Various Immediate Physiological Adaptations to HIIT for Different HF Phenotypes

Regarding HIIT participants with different HF phenotypes, $\dot{\mathrm{V}}$O_2peak significantly increased by 15% to 21% (P≤0.001) after completing 36 sessions of HIIT. Patients with HFrEF also showed a significant decrease in $\dot{\mathrm{V}}$ _E–$\dot{\mathrm{V}}$ _CO2 slope (P<0.001) and a significant increase in CO_ex (P=0.040) and Da–vO₂ (P=0.009) after exercise training. However, significant changes in cardiorespiratory fitness were not observed in patients with the other 2 HF phenotypes after the intervention (Table 3).

Table 3.

Cardiorespiratory Fitness and LV Geometry in Patients With Different HF Phenotypes Before and After HIIT

Variable		HFrEF	HFmrEF	HFpEF	P value
		(n=65)	(n=17)	(n=14)
$\dot{\mathrm{V}}{\mathrm{O}}_{2\text{peak}}$, MET	Pre‐HIIT	4.91 (4.63–5.19)	4.88 (4.36–5.39)	4.42 (3.92–4.93)	0.502
	Post‐HIIT	5.94 (5.56–6.31)	5.60 (5.01–6.19)	5.35 (4.61–6.09)	0.991
	P value	<0.001∥	<0.001∥	0.001∥
$\dot{\mathrm{V}}$ E‐$\dot{\mathrm{V}}$ CO2 slope	Pre‐HIIT	34.8 (32.8–36.8)	31.2 (29.3–33.1)	37.5 (31.6–43.4)	0.061
	Post‐HIIT	32.4 (30.7–34.1)	32.1 (29.6–34.5)	34.5 (30.9–38.2)	0.865
	P value	<0.001∥	0.398	0.074
Da–vO2, mL/dL	Pre‐HIIT	12.6 (11.5–13.7)	11.7 (10.0–13.5)	9.91 (7.63–12.2)	0.067
	Post‐HIIT	14.2‡ (13.1‐15.2)	11.9 (10.5–13.4)	11.7‡ (9.65‐13.8)	0.004§
	P value	0.009∥	0.743	0.228
COex, L/min	Pre‐HIIT	10.4 (9.36–11.5)	10.5 (8.77–12.1)	11.5 (8.89–14.1)	0.370
	Post‐HIIT	11.0 (9.90–12.2)	11.2 (9.75–12.7)	11.8 (9.38–14.2)	0.179
	P value	0.040∥	0.161	0.876
PCS	Pre‐HIIT	46.5 (44.4–48.7)	44.4 (41.2–47.7)	47.5 (44.0–50.9)	0.486
	Post‐HIIT	51.5 (49.7–53.4)	51.9 (48.6–55.2)	49.9 (44.8–55.0)	0.844
	P value	0.483	0.419	0.411
MCS	Pre‐HIIT	45.7 (43.3–48.0)	44.2 (39.0–49.5)	40.3 (35.5–45.0)	0.242
	Post‐HIIT	48.4 (46.1–50.7)	45.2 (42.4–48.0)	44.7 (38.4–51.0)	0.504
	P value	0.458	0.371	0.411
LVEF, %	Pre‐HIIT	26.1*, ‡ (24.0–28.3)	44.3*, † (42.9–45.7)	62.3†, ‡ (58.0–66.7)	<0.001§
	Post‐HIIT	47.2 (43.1–51.3)	52.0 (46.3–57.7)	58.5 (51.6–65.4)	0.007§
	P value	<0.001∥	0.035∥	0.030∥
LVEDD, mm	Pre‐HIIT	63.3*, ‡ (60.6–66.0)	55.4*, † (52.1–58.7)	50.7†, ‡ (46.3–55.0)	<0.001§
	Post‐HIIT	59.5 (56.8–62.1)	54.4 (51.1–57.7)	54.5 (50.4–58.6)	0.232
	P value	0.002∥	0.585	0.049∥
LVESD, mm	Pre‐HIIT	55.3*, ‡ (52.4–58.2)	43.1*, † (40.4–45.7)	32.2†, ‡ (27.6–36.9)	<0.001§
	Post‐HIIT	44.9 (41.6–48.2)	39.8 (35.8–43.8)	37.4 (33.1–41.8)	0.088
	P value	<0.001∥	0.155	0.151

Values are mean (95% CI).

CO_ex indicates peak cardiac output during exercise; Da–vO₂, systemic arteriovenous O₂ difference; HIIT, high‐intensity interval training; LV, left ventricular; LVEDD, left ventricular end‐diastolic dimension; LVEF, left ventricular ejection fraction; LVESD, left ventricular end‐systolic dimension; MCS, mental component score; MET, metabolic equivalent in 3.5 mL O₂/min/kg; PCS, physical component score; ${\dot{\mathrm{V}}}_{\mathrm{E}}$, minute ventilation; $\dot{\mathrm{V}}$CO₂, minute carbon dioxide production; and $\dot{\mathrm{V}}$O_2peak, peak oxygen consumption.

^*Statistical differences between patients with heart failure with reduced ejection fraction (HFrEF) and those with HF with mid‐range ejection fraction (HFmrEF) were assessed by Bonferroni post hoc test.

^† Statistical differences between patients with HFmrEF and those with HF with preserved ejection fraction (HFpEF) were assessed by Bonferroni post hoc test.

^‡ Statistical differences between patients with HFrEF and those with HFpEF were assessed by Bonferroni post hoc test.

^§ Statistical differences were assessed by generalized estimating equation, adjusted for age, sex, disease duration, diabetes, and atrial fibrillation.

^∥ Statistical differences were assessed by paired t test.

LVEF in patients with HFrEF (P<0.001) and HFmrEF (P=0.035) was significantly increased after HIIT. However, LVEDD (P=0.002) as well as LVESD (P<0.001) in patients with HFrEF was significantly decreased after HIIT. Furthermore, patients with HFpEF showed a significant decrease in LVEF (P=0.030) accompanied by a significant increase in LVEDD (P=0.049) and a nonsignificant increase in LVESD after completing HIIT (Table 3).

Varied Chronic Clinical Features of Patients With Different HF Phenotypes after HIIT

Patients with HF who have different phenotypes responded differently to HIIT during the long‐term follow‐up. In addition, the heatmaps of the longitudinal follow‐up for cardiac geometry changes showed increased LVEF associated with decreased LVEDD and LVESD in patients with HFrEF and those with HFmrEF after HIIT (Figure 2). However, this trend was not observed in the patients with HFpEF. The mean change in LVEF (ΔLVEF) increased significantly in patients with HFrEF (P=0.041) and those with HFmrEF (P=0.012) during the long‐term follow‐up. For patients with HFpEF, ΔLVEF was not significantly different between GDMT and HIIT participants during the follow‐up; however, both mean ΔLVEDD (P=0.004) and ΔLVESD (P=0.003) increased significantly after HIIT during the follow‐up (Table S1).

Figure 2. Left ventricular (LV) remodeling responses to high‐intensity interval training (HIIT) and guideline‐directed medical therapy (GDMT) for participants with different heart failure (HF) phenotypes.

Heatmaps show the changes in LV ejection fraction (ΔLVEF), LV end‐diastolic diameter (ΔLVEDD), and LV end‐systolic diameter (ΔLVESD) as the differences in parameters measured during the longitudinal follow‐up to the baseline assessments for different HF phenotypes. HFmrEF indicates heart failure with mid‐range ejection fraction; HFpEF, heart failure with preserved ejection fraction; and HFrEF, heart failure with reduced ejection fraction.

HIIT Was Associated With Improved 10‐Year Survival for Patients With HFrEF and Those With HFmrEF

The 10‐year survival, adjusted by multivariate Cox regression model, for all HIIT participants was 80.3%, which was significantly better (P=0.015) than that for GDMT participants (68.6%) (Figure 3A). Among patients with HFrEF, the adjusted 10‐year survival was significantly better (P=0.044) for HIIT participants (80.0%) than for GDMT participants (74.0%) (Figure 3B). Among patients with HFmrEF, the adjusted 10‐year survival was significantly better (P=0.036) for HIIT participants (88.2%) than for GDMT participants (57.9%) (Figure 3C). Among patients with HFpEF, the 10‐year survival for HIIT (92.9%) and GDMT (73.1%) participants was not significantly different (Figure 3D). Adjusted Cox regression results are detailed in Table S2.

Figure 3. Survival for all included participants according to different heart failure (HF) phenotypes.

A, Ten‐year survival for all patients with HF who underwent high‐intensity interval training (HIIT) versus guideline‐directed medical therapy (GDMT). B, Patients with HF with reduced ejection fraction (HFrEF). C, Patients with HF with mid‐range ejection fraction (HFmrEF). D, Patients with HF with preserved ejection fraction (HFpEF).

Regarding patients with HF who had LVEF <50%, the adjusted 10‐year survival was significantly better (P=0.015) for HIIT participants (81.7%) than for GDMT participants (70.7%) (Figure S2). For all GDMT participants, the adjusted 10‐year survival for patients with HFrEF was significantly better (P=0.036) than that of patients with HFmrEF. For all HIIT participants, the survival of patients with different HF phenotypes was similar after completing HIIT (Figure S3).

$\dot{\mathbf{V}}{\mathbf{O}}_{\mathbf{\text{2peak}}}$, $\dot{\mathrm{V}}$E–$\dot{\mathrm{V}}$ _CO2 slope, and LVEDD Were Significant Risk Factors for 10‐Year Mortality of Patients With HF

For all patients with HF, our results showed that HIIT reduced the risk of 10‐year all‐cause mortality and each 1‐metabolic equivalent increase in $\dot{\mathrm{V}}$O_2peak after HIIT might confer a significant reduction in the all‐cause mortality risk by ≈51% (adjusted hazard ratio [aHR], 0.49 [95% CI, 0.35–0.67], P<0.001). However, risks of mortality increased with an increased $\dot{\mathrm{V}}$ _E–$\dot{\mathrm{V}}$ _CO2 slope (aHR, 1.08 [95% CI, 1.042–1.12], P<0.001), greater LVEDD (aHR, 1.06 [95% CI, 1.03–1.09], P<0.001), and older age (aHR, 1.06 [95% CI, 1.02–1.10], P=0.003) (Figure 4).

Figure 4. Relative risk of all‐cause mortality in all included patients with heart failure during the 10‐year follow‐up.

Age, peak oxygen consumption ($\dot{\mathrm{V}}$O_2peak), minute ventilation ($\dot{\mathrm{V}}$ _E) minute carbon dioxide production ($\dot{\mathrm{V}}$ _CO2) slope, and left ventricular end‐diastolic diameter (LVEDD) in red dots were significant risk factors for all‐cause mortality. aHR indicates adjusted hazard ratio; BMI, body mass index; and CO_ex, peak cardiac output during exercise.

For patients with HFrEF, $\dot{\mathrm{V}}$O_2peak was the only significant risk factor (aHR, 0.38 [95% CI, 0.15–0.95], P=0.039) for 10‐year all‐cause mortality (Figure S4). In addition, significant mortality risk factors could not be identified in patients with HFmrEF and HFpEF. Most of our patients with HF died because of atrial fibrillation and aggravation of HF. Causes of death did not differ significantly between the exercise and nonexercise groups for each phenotype (Figure S5).

Mediators for the 10‐Year Mortality of Patients With Different HF Phenotypes

Although a significant indirect mediating role of LVEDD in the association between HIIT and 10‐year all‐cause mortality (indirect path coefficient, cardio –0.067 [95% CI, −0.112 to −0.022], P=0.003) during adjusted univariate linear regression (Table S3), this analysis focused on determining a relationship between one independent variable and one dependent variable without considering interactions between different independent variables. Therefore, adjusted multiple linear regression was used to estimate the association between different independent variables and the 10‐year all‐cause mortality in all included participants. Results showed a significant indirect mediating role of LVEDD in the association between HIIT and 10‐year all‐cause mortality (indirect path coefficient, −0.201 [95% CI, −0.303 to −0.099], P<0.001) (Figure 5; Table S3).

Figure 5. Causal mediation analysis was assessed by adjusted multiple linear regression for the association between high‐intensity interval training (HIIT) and 10‐year all‐cause mortality in all included patients with heart failure (HF) (red) and patients with HF with left ventricular ejection fraction (LVEF) <50% (blue).

The direct association between HIIT and mortality was observed in patients with HF with an LVEF <50%, but not in all included patients with HF. The associations between HIIT and mortality were mediated by left ventricular end‐diastolic diameter (LVEDD) in the above patient classification. IA indicates indirect association; m1, pathway coefficient (95% CI) between HIIT and the mediating factor; and m2, pathway coefficient (95% CI) between the mediating factor and long‐term mortality.

After collapsing our participants into 2 groups according to LVEF (ie, LVEF <50% and ≥50%), 174 patients with LVEF <50% (82 HIIT participants and 92 GDMT participants) and 40 patients with LVEF ≥50% (14 HIIT participants and 26 GDMT participants) were included in the mediation path analysis. Univariate linear regression results for the mediating role of LV geometry in the association between HIIT and 10‐year all‐cause mortality in the above 2 groups are shown in Table S4. A significant direct association between HIIT and 10‐year all‐cause mortality (direct path coefficient, 0.360 [95% CI, 0.042–0.360], P=0.027) was observed during adjusted multiple linear regression. A significant indirect mediating role of LVEDD in the association between HIIT and 10‐year all‐cause mortality (indirect path coefficient, −0.282 [95% CI, −0.484 to −0.080], P=0.006) was also identified by adjusted multiple linear regression analysis (Figure 5; Table S5). Among patients with LVEF ≥50%, neither the direct nor the indirect associations were statistically significant for any measure of LV remodeling response (Table S5).

Discussion

This investigation is the first to our knowledge to demonstrate that HIIT is associated with a better 10‐year survival in patients with HF. An increased aerobic capacity and decreased ventilation response to CO₂ production and LVEDD after HIIT were promising protective factors against mortality in patients with HF. We also highlighted that patients with HF with different phenotypes had various long‐term clinical outcomes after HIIT. In the present work, HIIT could indirectly affect 10‐year all‐cause mortality in all included patients with HF in the study. This exercise strategy might directly exert health benefits or could affect the 10‐year all‐cause mortality by mediating the LV remodeling for patients with HFrEF and HFmrEF. In contrast with patients with HFrEF and those with HFmrEF, the reversal of the LV remodeling after HIIT could not be identified in patients with HFpEF.

A systematic review of 1.5 million patients with HF over the past 70 years indicated that the estimated survival without exercise intervention at 5 and 10 years after HF diagnosis was 56.7% and 34.9%, respectively. ²⁴ The study also showed a similar 5‐year survival in patients with HFrEF and HFpEF. In the ECHOES (Echocardiographic Heart of England Screening) study, the 10‐year survival was 30.8% for patients with HFrEF. ²⁵ Without HIIT, the 10‐year all‐cause mortality has been reported to be 39% for patients with HFrEF, 25% for patients with HFmrEF, and 22% for patients with HFpEF. ²⁶ The results vary substantially across studies; however, all of the above scientific reports have documented high mortality in patients with HF. In our participants, the prevalence of atrial fibrillation was significantly greater in the GDMT patients than in the HIIT patients with HFmrEF, which could increase the risk of death ²⁷ and result in a poor prognosis. After HIIT, the long‐term survival for patients with HFmrEF progressed to match that of the other 2 phenotypes. These observations suggest that atrial fibrillation should be optimally managed in patients with HFmrEF.

Our work has shown that HIIT provided a robust stimulus of ≈15% to 21% increase of $\dot{\mathrm{V}}$O_2peak in patients with the 3 different HF phenotypes, which was similar to findings found in several previous studies. ⁴ , ⁷ , ⁸ Aerobic capacity was not followed for patients with HF who underwent GDMT because clinicians did not refer them for secondary CPET examination in the study. However, the decrease of $\dot{\mathrm{V}}$O_2peak in GDMT participants could be expected, because a significant decrease of $\dot{\mathrm{V}}$O_2peak has been reported in patients with HFrEF without HIIT. ⁴ Consistent evidence has shown that exercise training improves aerobic capacity and health‐related quality of life (class I, Level of Evidence A). Exercise‐based cardiac rehabilitation also reduces the risk of all‐cause hospital admissions ²⁸ and should be considered for patients with HF. ¹ The clinical outcomes of exercise on patients with different HF phenotypes have been explored in several trials, but few reports are available on patients with HFmrEF. ¹ Either a nonsignificant reduction in all‐cause mortality ²⁹ or benefits for 8‐month survival after exercise training has been reported. ⁷ The conflicting observations of the exercise benefits for patients with HF are possibly caused by the heterogeneity of the included participants, classifications of cardiac functions, or HF care programs. ³⁰ Therefore, the advantage of exercise on the long‐term survival of patients with different HF phenotypes warrants further investigation. In the present study, HIIT was associated with favorable long‐term clinical outcomes in patients with HF who had an LVEF <50%. Patients with HFpEF developed an increase of LV dimensions as well as aerobic capacity but slowly decreased LV contractility after HIIT. Although 10‐year survival of patients with HFpEF was better for HIIT than for GDMT participants, statistical significance was not observed in them. The inconclusive results could be related to the small sample size in the study. Thus, long‐term clinical outcomes for patients with HFpEF after HIIT require further investigation.

Older age is a known independent predictor of mortality in patients with HF. ³¹ In this study, older age predicted a poor prognosis for 10‐year all‐cause mortality in all patients with HF but not for each phenotype, possibly owing to the strict control of the age range in the different phenotypes. An objective assessment of cardiorespiratory fitness has been increasingly used to indicate the surrogate efficacy of therapeutic end points in clinical trials. ³² $\dot{\mathrm{V}}$O_2peak is the strongest predictor of death risk among patients with cardiovascular events. ³³ In our observations, aerobic capacity was an important predictor of all‐cause mortality in patients with HF, especially in patients with HFrEF, and was consistent with results of previous studies. ³¹ , ³² , ³³ An increased $\dot{\mathrm{V}}$ _E–$\dot{\mathrm{V}}$CO₂ slope indicates an increased ventilatory drive during exercise and a poor prognosis in patients with HF. ³⁴ Our long‐term observations support this hypothesis. With a reduced cardiac output accompanied by HF, the sympathetic nervous system is overactivated to compensate for the impaired systolic function. Peripheral vasoconstriction and an enlarged heart result in edematous lung parenchyma and reduced lung compliance. This phenomenon contributes to a reduced aerobic capacity and increased breathing to expel CO₂ from the lungs during CPET. ³⁵ Aerobic exercise training has been proposed to normalize the imbalanced autonomic system, promote cardiorespiratory fitness, and improve long‐term survival in patients with HF. ³⁵

Increased LV wall stress is associated with pathological LV remodeling and is characterized by chamber enlargement and the development of contractile dysfunction. ³⁶ The progression of this process results in a poor prognosis, whereas the reversal of remodeling is associated with a better prognosis in patients with HF. ³⁷ Therefore, LV dimensions have long been used as a clinical indicator for the survival of patients with HF. ²⁵ Moreover, high end‐diastolic wall stress is expected to cause episodic hypoperfusion of the subendocardium, resulting in the worsening of LV function. ³⁸ Reductions in LVEDD and LVESD accompanied by various changes in LVEF were observed in patients with HFrEF after HIIT at 50% to 95% of $\dot{\mathrm{V}}$O_2peak ⁴ , ⁷ or moderate‐intensity continuous training at 60% to 70% of $\dot{\mathrm{V}}$O_2peak ³⁹ have been reported. In the study, findings concerning LV dimension measurements during follow‐up highlighted that LVEDD was a significant mortality risk factor for all included patients with HF.

HIIT at alternating 40% and 80% of $\dot{\mathrm{V}}$O_2peak ⁷ , ¹⁵ or moderate‐intensity continuous training at 60% $\dot{\mathrm{V}}$O_2peak ¹⁵ induced a lesser degree of alterations in LV dimensions with almost no change in LVEF for patients with HFpEF. In our patients with HFrEF and HFmrEF, HIIT reversed cardiac remodeling, similar to previous findings in patients with HFrEF. ⁴ , ⁷ , ³⁹ Nevertheless, patients with HFpEF showed a trend of increased LV dimensions and a significant reduction in LV contractility immediately after HIIT (Table 3). Progressive dilatation of the LV chamber associated with decreased LVEF became dominant in this HF phenotype during the long‐term follow‐up (Table S1). Whether this phenomenon is a typical finding of exercise‐induced physiological cardiac remodeling after an increase in the preload ⁴⁰ or a regression of pathological cardiac remodeling in patients with HFpEF is yet to be determined. A randomized clinical trial of patients with HFpEF who underwent different therapeutic strategies suggested that neither HIIT nor moderate‐intensity continuous training bring about a better clinical outcome compared with guideline‐based physical activity. ⁴¹ In our long‐term observations for limited patients with HFpEF, HIIT did not affect their survival. Therefore, the benefit of HIIT for HFpEF is not persuasive and further investigation for therapeutic approach in these patients is indicated.

Our CMA results identified a significant mediating role of cardiac remodeling in the association between HIIT and long‐term mortality. Significant reversal of cardiac remodeling in patients with HFrEF after HIIT has been observed immediately after 12 weeks of supervised aerobic exercise training. ⁴ , ⁷ , ¹⁷ These exercise‐induced physiological adaptations in patients under optimal medical treatment maintain the degree of change and do not deform subsequently during the long‐term follow‐up. ¹⁷ The HIIT‐associated reversal of LV remodeling for patients with HF who have an LVEF <50% in this study was similar to that observed in patients treated with angiotensin‐converting enzyme inhibitors or β‐blockers. ⁴² , ⁴³ A period of HIIT may modify lifestyle and directly bring long‐term benefits for patients with LVEF <50%. However, further investigations are required to support this point of view. LVEDD, an indicator of high‐risk HF, ⁴⁴ played an essential mediating role in the association between HIIT and the long‐term mortality of patients with HFrEF and HFmrEF. A long follow‐up time interval ⁴⁵ between the mediator (LV dimensions) and outcome (mortality) can decrease the causal mediation strength. Although increasing use of mediation analyses for nonnormally distributed continuous variables in health care studies are observed, these models are ill‐adapted to time‐to‐event outcomes as well as low reporting rates of underlying assumpations. ⁴⁶ Therefore, CMA for time‐to‐event outcomes analysis for patients with HF can only be accomplished on the premise of having a reliable mathematical model.

Limitations

The current study has few limitations. The first limitation is the study design. A retrospective study cannot completely exclude potential bias for defining health benefits of aerobic exercise training on patients with HF. Second, the small sample size of patients with HFmrEF and HFpEF might have affected the power of the statistical comparisons and estimation of cardiac mortality. Therefore, studies with larger sample sizes for patients with these 2 HF phenotypes are required to confirm our findings. Third, the physical activities, especially in the community, of both HIIT and GDMT participants should be followed, because the factor may be a confounder affecting long‐term survival. Finally, the guidelines and treatment strategies have changed in the follow‐up span of 14 years (2009–2022), which may affect clinical outcomes of our included patients with HF who underwent GDMT. For example, the European Cardiology Society introduced the HFmrEF phenotype since 2016, which is relatively unexplored compared with studies for patients with HFrEF and those with HFpEF. ²³ Diagnostic options for HFpEF have continuously evolved. ⁴⁷ Future dedicated research should encompass the GDMT and clinical outcomes of patients with HFmrEF and those with HFpEF after HIIT.

Conclusions

This 10‐year retrospective cohort study demonstrated that HIIT was associated with different long‐term clinical outcomes in patients with different HF phenotypes. The reversal of pathological LV remodeling participated in the transmittance of alterations from HIIT to long‐term survival and was associated with improved aerobic capacity in patients with HF with an LVEF <50% after HIIT. Protective factors against mortality were an increased $\dot{\mathrm{V}}$O_2peak, a decreased $\dot{\mathrm{V}}$ _E–$\dot{\mathrm{V}}$CO₂ slope, and LVEDD in all patients with HF. An improvement in aerobic capacity after HIIT could be beneficial for the long‐term survival of patients with HFrEF. HIIT was associated with an increased aerobic capacity in patients with HFpEF; however, this exercise strategy might induce persistent LV remodeling and did not affect their long‐term survival during the 10‐year follow‐up. Therefore, the present work does not support the HIIT‐associated long‐term clinical benefits for patients with HFpEF. With an in‐depth analysis of the associations between HIIT and the clinical outcomes of patients with all HF phenotypes, this study could provide insights for managing HF in patients with different phenotypes.

Sources of Funding

The work was supported by the Ministry of Science and Technology, Taiwan (MOST 109‐2314‐B‐182‐026‐MY3), and the Keelung Chang Gung Medical Research Program (CMRPG2H0231) for pure academic interests.

Disclosures

None.

Supporting information

Data S1

Tables S1–S5

Figures S1–S5