Clinical Predictors of Adherence to Exercise Training among Individuals with Heart Failure: The HF-ACTION Study

Structured Abstract

Purpose.

Suboptimal adherence is a major limitation to achieving the benefits of exercise interventions, and our ability to predict and improve adherence, is limited. The purpose of this analysis was to identify baseline clinical and demographic characteristics predicting exercise training adherence in the HF-ACTION study cohort.

Methods.

Adherence to exercise training, defined by the total number of min of exercise performed per wk, was evaluated in 1,159 participants randomized to the HF-ACTION exercise intervention. More than 50 clinical, demographic and exercise testing variables were considered in developing a model of the min/wk endpoint for mo 1-3 (supervised training) and 10-12 (home-based training).

Results.

In the multivariable model for mo 1-3, younger age, lower income, more severe mitral regurgitation, shorter six minute walk distance, lower exercise capacity, and Black or African American race was associated with poorer exercise intervention adherence. No variable accounted for >2% of the variance and the adjusted R² for the final model was .14. Prediction of adherence was similarly limited for mo 10-12.

Conclusions.

Clinical and demographic variables available at the initiation of exercise training provide very limited information for identifying patients with heart failure who are at risk for poor adherence to exercise interventions.

Condensed Abstract

The purpose of this analysis was to identify baseline clinical and demographic characteristics predicting exercise training adherence. We found clinical and demographic variables available at the initiation of exercise training provide very limited information for identifying patients with heart failure who are at risk for poor adherence to exercise interventions.

INTRODUCTION

Cardiac rehabilitation (CR) improves exercise capacity and quality of life, while reducing risk of hospitalization in patients with chronic heart failure (HF).^1,2 Participation in CR is now recommended by HF guidelines^3,4 and supported by the Centers for Medicare and Medicaid Services⁵ for patients with HF with reduced ejection fraction (HFrEF).

Unfortunately, one of the major limiting factors to successful delivery of CR is poor exercise training adherence; with adherence being especially challenging for individuals with HF.^6-8 Among patients with ischemic heart disease, those with concomitant HF participate in CR significantly less than those without HF.⁹ Furthermore, exercise training adherence is consistently lower than adherence to other HF-related self-care behaviors.^10-12 Even HF patients who believe exercise to be an important health behavior frequently show poor exercise adherence;¹² resulting in an increased risk for adverse clinical outcomes.¹³ Identifying factors associated with non-adherence is essential to helping HF patients realize the full benefits of exercise programs.

Heart Failure: A Controlled Trial Investigating Outcomes of Exercise Training (HF-ACTION) is the largest randomized controlled trial of exercise training to date. As a result, this study is uniquely positioned to offer potentially valuable insights into the clinical and demographic factors influencing exercise intervention adherence. As in other studies of exercise training, ^14-16 maintaining exercise adherence proved to be challenging during both supervised and home exercise training periods.¹⁷ The purpose of the current analysis was to provide a more comprehensive description of patient adherence characteristics and identify clinical and demographic predictors of exercise training adherence.

METHODS

A detailed description of the HF-ACTION trial design and the results of the primary analysis have been published previously.¹⁷ HF-ACTION was a multicenter, randomized controlled trial of exercise training and usual care versus usual care alone in 2,331 patients with chronic, stable, New York Heart Association (NYHA) Class II-IV HF and low left ventricular ejection fraction (LVEF ≤ 35%) in the context of optimal medical therapy. Participants in HF-ACTION were randomized from April 2003 through February 2007 within the United States, Canada, and France. Exclusion criteria included major comorbidities or limitations that could interfere with exercise training, recent (within six wks) or planned (within six mo) major cardiovascular events or procedures, performance of regular exercise training, or use of devices that limited the ability to achieve target heart rates (HR). The current analysis was conducted using the cohort randomized to the intervention (exercise training) arm only (Figure 1). The HF-ACTION study was approved by all local Institutional Review Boards and all participants provided informed consent. Eligible participants were randomized 1:1 using a permuted block randomization scheme, stratified by clinical center and heart failure etiology (ischemic vs. nonischemic).

Figure 1.

Participants in the HF-ACTION Trial Included in the Adherence Analysis

At the baseline visit, prior to randomization, demographics, past medical history, current medications, physical exam, and the most recent laboratory tests were obtained. All participants underwent baseline cardiopulmonary exercise testing (CPX). The HF-ACTION testing protocol was uniform and rigorous. A CPX was completed using commercially available metabolic carts and motor driven treadmills employing a modified Naughton protocol.¹⁸ Peak oxygen uptakte ($\stackrel{.}{\mathrm{V}}{\mathrm{O}}_{2\text{peak}}$) was determined as the highest oxygen consumption normalized to body weight (mL/kg/min) for a given 15- or 20- sec interval within the last 90 sec of exercise or the first 30 sec of recovery, whichever was greater. Test results were reviewed by investigators to identify significant arrhythmias or ischemia that would prevent safe exercise training, to determine an appropriate exercise prescription, and to establish training HR ranges. Additionally at baseline, participants completed the six-minute walk test (6MWT). Each 6MWT was conducted in a standardized format, with specific instructions provided in the HF-ACTION manual of operations, modeled after prior studies.^19,20 Each HF-ACTION site was instructed to measure a 20-25 meter indoor course and to position a chair at either end, providing participants a place to rest, if needed.

EXERCISE TRAINING

Participants randomized to the exercise training arm initially participated in a structured, group-based, supervised exercise program, with a goal of three sessions per wk for a total of 36 sessions in three mo. During the supervised training phase, participants performed walking on a treadmill, or stationary cycling as their primary training mode. Exercise was initially prescribed at 15 to 30 min per session at a HR corresponding to 60% of HR reserve (maximal HR on CPX minus resting HR), three times per wk. After six sessions, the duration of the exercise was increased to 30-35 min, and intensity was increased to 70% of HR reserve. Participants began home-based exercise after completing 18 supervised sessions and fully transitioned to home exercise after 36 supervised sessions. Participants were provided home exercise equipment (cycle or treadmill [ICON]) and HR monitors (Polar USA, Inc.). The target training regimen for home exercise was five times per wk for 40 min at a HR of 60% to 70% of HR reserve. Occasional “booster” sessions of supervised exercise were offered to participants with poor adherence during the period of home-based training.

ADHERENCE

The Adherence Core Laboratory was established for HF-ACTION to develop adherence strategies and related materials, support the implementation of adherence strategies at enrolling sites, and monitor exercise adherence at the individual and site level.²¹ Adherence data were regularly monitored by the Adherence Core Laboratory. When participants were not meeting their exercise prescription, as defined above, additional adherence measures were implemented, including targeted phone calls and periodic supervised “booster” sessions during the home-based training period. Sites with low exercise adherence were also provided with additional training, including site visits.

Adherence strategies included print reminders (calendars, newsletters, cards), phone calls to reinforce exercise training goals and identify barriers to adherence, involvement of family members and friends for support, logistical assistance including funds to assist with transportation and child care needs, and other incentives (e.g., t-shirts, mugs). To further promote adherence to home-based training, HR monitors were provided to participants.

Adherence was assessed using an assortment of measures including attendance at facility-based exercise sessions, completion of physical activity logs for home-based exercise, telephone follow-up logs, HR monitoring data, and self-reported percentage of time at or greater than the prescribed training range. For this analysis, total min of exercise per wk versus prescribed min/wk was the primary outcome of adherence. Target exercise min/wk for mo 1-3 was 90 min, and for mo 10-12 was 120 min.

STATISTICAL ANALYSIS

For descriptive purposes, adherence categories were established a priori based upon the total min of exercise per wk. Over mo 1-3, participants were considered fully adherent if they achieved an average of at least 90 min of exercise training per wk, partially adherent if 45-89 min were achieved, poorly adherent if 15-44 min were achieved, and non-adherent if fewer than 15 min were achieved. Baseline characteristics in each of these categories were described. Participants who dropped out of the study were considered separately.

The primary analysis focused on adherence to the exercise intervention for mo 1-3 (predominantly supervised exercise). All assumptions required for regression analysis were assessed prior to modelling. A model of the min/wk (adherence) as a continuous variable was developed by considering over 50 clinical, demographic and exercise testing variables—most of the collected clinical variables of the study—as potential predictors (Supplementary Table 1). A bootstrapped backwards selection algorithm using more than 1,000 bootstrap samples in total was performed for selection of the final model. A secondary analysis of exercise training adherence in mo 10-12 (predominantly home-based exercise) was also performed, however not included in the primary results of this analysis (Supplementary Table 2). Variables for the two time periods were identical, except min/wk in mo 1-3 was added as a potential predictor in the model of adherence in mo 10-12. As a sensitivity analysis, variables from the 1-3 mo and 10-12 mo models were used to predict dropout using multivariable logistic regression, however these results are not included in the primary results of this analysis (Supplementary Tables 3 and 4). Missing data among the covariates were filled in using a multiple imputation procedure.

RESULTS

All 1,159 participants randomized to the exercise intervention in HF-ACTION were included in this analysis. After accounting for death, dropout and missing adherence data, 1142 participants were available for the primary 3-mo analysis and 974 participants for the 12-mo analysis (Figure 1). For the 3-mo analysis, the median participant age was 59 yr, 30% were women, median LVEF was 25% and 51% had HF from ischemic etiology. Among those randomized to the exercise intervention, most were either “very satisfied” (79%) or “somewhat satisfied” (13%) with their study arm assignment.

There was a wide range of adherence throughout the study, with the majority of participants continuing to exercise to some extent (Figure 2). Median exercise time per wk was 77 min during mo 1-3 (goal ≥ 90 min/wk) and peaked at 95 min (goal ≥ 120 min/wk) during the transition to the home-based exercise training (mo 4-6). After this, min of exercise per wk gradually decreased over the first 12 mo, and then became relatively stable at approximately 50-60 min/wk—approximately 2 training sessions per wk—for the remainder of the study (Figure 2).

Figure 2.

Median Min/Wk of Exercise. The bar indicates exercise adherence goal for supervised (mo 1-3; 90 min) and home-based (mo 4-36; 120 min) training. Median exercise time was closest to goal during supervised training (mo 1-3; 77 min). Exercise time was greatest when participants first transitioned to home-based training (mo 4-6) and then gradually decreased over the first yr of follow-up, and continued across the 36 mo intervention. Missing data was assumed to be no exercise.

By adherence category, the distribution of key clinical and demographic variables for supervised training is displayed in Table 1. Participants who were older, of White race, and not taking a loop diuretic tended to have greater adherence. Full and partial adherers also tended to have better functional performance, including lower NYHA class; greater 6MWT distance; longer CPX duration; and greater peak oxygen uptake. Variables that appeared similar across adherence categories included LVEF, presence or absence of atrial fibrillation, device therapy and medical therapy other than loop diuretics.

Table 1.

Baseline Characteristics by Adherence Category for Mo 1-3

	Full Adherersa (n=482)	Partial Adherersb (n=333)	Poor Adherersc (n=176)	Non-Adherersd (n=151)
Age,yr†	61 (53, 69)	59 (52, 68)	57 (48, 67)	55 (46, 66)
Sex
Female (341)	125 (37)	109 (32)	58 (17)	49 (14)
Male (801)	357 (45)	224 (28)	118 (15)	102 (13)
Race
Black/African-American (373)	117 (31)	106 (28)	66 (18)	84 (23)
White (688)	335 (48)	201 (29)	94 (14)	58 (8)
LVEF, %e	25 (21, 30)	24 (20, 30)	24 (20, 29)	25 (19, 31)
Baseline NYHA classf
II (712)	334 (47)	211 (30)	91 (13)	76 (11)
III or IV (430)	148 (34)	122 (28)	85 (20)	75 (17)
Etiology of HFg
Ischemic (587)	260 (44)	178 (30)	84 (14)	65 (11)
Non-ischemic (555)	222 (40)	155 (28)	92 (17)	86 (16)
Diabetes mellitus
Yes (370)	133 (36)	117 (32)	61 (16)	59 (16)
No (772)	349 (45)	216 (28)	115 (15)	92 (12)
Serum creatinine	1.1 (1.0, 1.4)	1.2 (1.0, 1.5)	1.2 (1.0, 1.5)	1.2 (1.0, 1.5)
Atrial fibrillation/flutter
Yes (244)	108 (44)	70 (29)	39 (16)	27 (11)
No (898)	374 (42)	263 (29)	137 (15)	124 (14)
ACE inhibitorh
Yes (860)	366 (43)	251 (29)	133 (15)	110 (13)
No (282)	116 (41)	82 (29)	43 (15)	41 (15)
Angiotensin II receptor blocker
Yes (283)	117 (41)	88 (31)	40 (14)	38 (13)
No (859)	365 (42)	245 (29)	136 (16)	113 (13)
Beta Blocker
Yes (1075)	455 (42)	315 (29)	165 (15)	140 (13)
No (67)	27 (40)	18 (27)	11 (16)	11 (16)
Aldosterone receptor antagonist
Yes (518)	204 (39)	160 (31)	78 (15)	76 (15)
No (624)	278 (45)	173 (28)	98 (16)	75 (12)
Loop diuretic
Yes (883)	351 (40)	265 (30)	139 (16)	128 (14)
No (259)	131 (51)	68 (26)	37 (14)	23 (9)
ICDi
Yes (485)	210 (43)	145 (30)	73 (15)	57 (12)
No (657)	272 (41)	188 (29)	103 (16)	94 (14)
Bi-ventricular pacemaker
Yes (212)	92 (43)	62 (29)	36 (17)	22 (10)
No (930)	390 (42)	271 (29)	140 (15)	129 (14)
Six-Minute Walk Distance (meter)	389 (320, 450)	366 (287, 433)	343 (287, 427)	332 (259, 391)
CPX duration (minute) j	10.2 (7.9, 12.3)	9.5 (6.9, 12.0)	8.2 (6.2, 11.0)	8.0 (5.5, 10.6)
$\stackrel{.}{\mathrm{V}}{\mathrm{O}}_{2\text{peak}}$ k	15.1 (12.4, 18.1)	14.0 (11.1, 17.6)	13.1 (10.8, 16.9)	13.5 (10.7, 17.0)

^aFull adherers, ≥ 90 min/wk

^bpartial adherers, 45-89 min/wk

^cpoor adherers, 15-44 min/wk

^dnon-adherer, <15 min/wk.

^†Data presented as n (%) or median (25^th, 75^th).

Abbreviations:

^hACE, Angiotensin Converting Enzyme

^jCPX, Cardiopulmonary Exercise Test

^gHF, Heart Failure

ⁱICD, Implantable Cardioverter Defibrillator

^eLVEF, Left Ventricular Ejection Fraction

^fNYHA, New York Heart Association Class

^k$\stackrel{.}{\mathrm{V}}{\mathrm{O}}_{2\text{peak}}$, Oxygen Uptake

The multivariable model utilizing clinical, demographic, and exercise testing variables demonstrated limited ability to predict exercise adherence for the supervised exercise period (mo 1-3) with an overall adjusted R² of .143 (selected variables shown in Table 2). Variables (partial R²) associated with worse adherence in the model included younger age (.017), more severe mitral regurgitation (.015), lower income (.015), and not enrolled in France (.021). Other variables associated with worse adherence but explaining less than 1% of the variance in min/wk of exercise included: Black or African American race (.004), lower LVEF (.005), shorter CPX duration (.004), shorter 6MWT distance (.008), and hospitalizations within the last 6 mo (.006). Baseline variables evaluated and not associated with adherence included sex, etiology of HF, NYHA class, use or dose of HF related medications (including beta-blocker, angiotensin converting enzyme inhibitors, or diuretic), body mass index, smoking status, comorbid disease (including atrial fibrillation, diabetes mellitus, peripheral vascular disease, and chronic obstructive pulmonary disease), resting HR, resting blood pressure, and education level. A scatter-plot of actual min of exercise training per wk and the min of exercise per wk predicted by the model for mo 1-3 illustrates the limited predictive power of the multivariable model (Figure 3). A summary of these findings is displayed in Figure 4.

Table 2.

Multivariable Model of Baseline Demographic and Clinical Variables as Predictors of Adherence to Exercise Training in Mo 1-3

Parameter	Estimate	Standard Error	t	P-value	Partial R2
Region=France (vs. non-France)	47.0	9.5	4.9	< .0001	.021
Income (vs. < $15K)
$15-$25K	13.4	5.5	2.5	.014	.015
$25-$35K	20.3	5.9	3.5	.0005
$35-$50K	16.2	6.4	2.5	.012
$50-$75K	17.5	6.4	2.8	.006
$75-$100K	22.4	7.3	3.1	.002
$100K +	9.2	7.9	1.2	.24
Age	.87	.19	4.5	<.0001	.017
Race (vs. white)
Black or African American	−8.7	3.9	−2.2	.027	.004
Other race (not black or white)	−3.4	7.3	−0.5	.65
Marital Status (vs. Married)
Widowed	−1.0	6.2	−0.2	.87	.012
Divorced	−15.7	5.2	−3.0	.003
Separated	6.2	8.8	0.7	.48
Single (never married)	−3.3	5.9	−0.6	.57
Living with a partner	7.7	9.2	0.8	.41
Decline to answer	−14.2	22.7	−0.6	.53
Employment status (vs. full-time)
Part-time	−5.3	7.8	−0.7	.49	.008
Student	16.4	32.0	0.5	.61
Homemaker	−4.1	11.1	−0.4	.71
Volunteer	−28.7	32.1	−0.9	.37
Disabled	10.1	5.3	1.9	.057
Unemployed	7.5	8.1	0.9	.35
Retired	8.2	5.6	1.5	.14
LVEFa	2.6	1.1	2.5	.014	.005
Mitral Regurgitation Grade (vs. None)
Trivial	.23	6.7	.03	.97	.015
Mild	−9.2	5.3	−1.73	.084
Mild to Moderate	−10.8	7.8	−1.4	.17
Moderate	−11.3	6.0	−1.9	.058
Moderate to Severe	−23.2	8.2	−2.8	.005
Severe	−16.5	8.1	−2.0	.042
Season at Randomization (vs. Winter)
Spring	−6.3	4.6	−1.4	.17	.009
Summer	−3.5	4.6	−0.8	.45
Autumn	−13.6	4.5	−3.0	.002
Six-minute walk distance	.057	.020	2.9	.004	.008
CPXb duration	4.1	2.0	2.0	.042	.004
BUNc	−0.18	.092	−2.0	.046	.003
Weber Class: (vs. A: $\stackrel{.}{\mathrm{V}}{\mathrm{O}}_{2\text{peak}}$ > 20)
B (16 < $\stackrel{.}{\mathrm{V}}{\mathrm{O}}_{2\text{peak}}$ ≤ 20)	8.1	7.3	1.1	.27	.004
C (10 < $\stackrel{.}{\mathrm{V}}{\mathrm{O}}_{2\text{peak}}$ ≤ 16)	12.7	7.6	1.7	.097
D ($\stackrel{.}{\mathrm{V}}{\mathrm{O}}_{2\text{peak}}$ ≤ 10)	14.8	9.7	1.5	.13
Hospitalizations in last 6 mo (vs. 0)
One	8.2	3.9	2.1	.03	.006
Two or more	−4.6	5.0	−0.9	.36

Abbreviations:

^cBUN, Blood Urea Nitrogen

^bCPX, Cardiopulmonary Exercise Test

^aLVEF, Left Ventricular Ejection Fraction

Figure 3.

Predicted vs Actual Min/Wk of Exercise Training in Mo 1-3.

A scatter plot of predicted compared to actual min/wk of exercise in mo 1-3 is widely distributed around the identity line. The prediction model was developed by considering over 50 clinical, demographic and exercise testing variables measured prior to initiation of exercise training.

Figure 4.

Variance in exercise adherence explained and unexplained by study variables. Multivariate and univariate analyses combined considered > 50 baseline variables as predictors of exercise adherence (min of exercise per wk). These variables accounted for only approximately 14% of the variance in adherence, leaving 86% of the variance unexplained.

Baseline demographic and clinical characteristics were similarly limited as predictors of adherence during home-based training. The multivariable model to predict min of exercise training during mo 10-12 had an overall R² value of just 0.207; with exercise min/wk during mo 1-3 being the strongest predictor of adherence.

DISCUSSION

To identify predictors of exercise adherence, this study examined a wide range of clinical and demographic variables among participants randomized to exercise training in HF-ACTION. Adherence (exercise min/wk) varied and generally declined as the study progressed. However, most participants continued to exercise to some extent throughout the study. Younger age, Black or African American race, lower income, more symptomatic HF, more severe mitral regurgitation and lower exercise capacity were among the variables associated with decreased adherence. No variable included in our models explained more than approximately 2% of the variance in adherence. Thus, the ability of multivariable models to predict adherence at both 1-3 mo and at 10-12 mo was very limited (R²: 0.143 and 0.207, respectively).

Understanding the factors influencing exercise adherence is critical to both the successful delivery of CR^6,7 and to successful dissemination of exercise interventions.²² Exercise training adherence is especially challenging for patients with HF^9-12; potentially due to frequent intercurrent illness and hospitalization, HF symptoms, and comorbidites including depression, cognitive impairment, and musculoskeletal disorders.⁷ Unfortunately, previous methodologies for determining exercise adherence have varied—and at times have not been reported in clinical trials of exercise training in patients with HF.^7,21,23 Although we have clearly defined exercise adherence in the present analysis, the inability of such a wide range of baseline clinical and demographic variables to predict more than about 14% of the variability in adherence is striking.

Previous research has similarly noted both limitations in predicting exercise training adherence ^24,25 among other patient populations, in addition to early adherence proving to be the strongest predictor of subsequent adherence.^26,27 In a study of exercise training in 213 community dwelling older adults, Rejeski and colleagues found baseline demographics, disease burden, symptoms, physical function, cognition and social status accounted for only 10-13% of the variance in adherence during the initial supervised (mo 1-2), transitional (mo 3-6) and maintenance (mo 7-12) phases of exercise training.²⁶ When adherence early in the study was combined with baseline variables, the model accounted for 21% and 46% of the variance in exercise training adherence during the transitional (mo 3-6) and maintenance (mo 7-12) phases. The predictive power of prior exercise adherence underscores the importance of closely monitoring and promptly addressing problems with adherence early during the intervention phase ²⁶; yet it does not provide insight into the optimal structure and content of such adherence interventions. In addition, this finding suggests a ramp-up phase prior to participants achieving the prescribed exercise, combined with an initial period of supervised exercise before prescribing home exercise, may be useful to reinforce the exercise habit and in identifying participants at risk for poor adherence. ¹⁷

The STRRIDE (Studies of Targeted Risk Reduction Intervention through Defined Exercise) trials – which examined the differential effects of exercise amount, intensity, and mode on cardiometabolic health among individuals with overweight or obesity and dyslipidemia or prediabetes ^28-30 – conducted an analysis exploring determinants and timing of dropout from the exercise intervention, as well as variation in exercise intervention adherence.³¹ The STRRIDE trials contained a ramp-up period to allow for gradual adaptation to the exercise prescription; however, a majority of individuals who dropped out from one of the exercise interventions did so during this phase. Yet, if an individual made it past this initial ramp period, adherence to the exercise intervention remained high (>75%) and steady across the intervention period. Thus, although a ramp-up period may be important to increase participant adherence, the progression of the exercise and the duration of the ramp period may critical for increasing adherence and decreasing dropout among individuals with overweight or obesity and dyslipidemia or prediabetes. These findings are important to consider especially among a HF population, which has a high rate of obesity, and for whom may find exercise prescriptions too lofty to achieve.³¹

Overall, the findings from the present analysis suggest many of the variables influencing exercise adherence are currently unrecognized or poorly understood. Further research is needed to explore a broader range of variables, from system level factors and the physical environment to genetic determinants; the complex interactions of multiple variables; the influence of these variables in important sub-populations; and changes over time and in response to exercise training interventions.^23,32,33 Alternative adherence strategies and attempts to understand the participant experience during these interventions should also be explored.²² Strategies, like CR-specific incentive programs, offer the advantage of impacting multiple adherence barriers with a single intervention.^34,35 System level changes, such as innovative delivery models or alternative training methods also warrant further investigation.^6,22 Further, conducting motivational interviewing and health coaching techniques targeting participant motivation, competing commitments, physical discomfort, self-efficacy and overall exercise experience could provide insight into what the participant is feeling during the intervention which in turn may provide information for a more optimal intervention approach for improving adherence among the HF population.^36-42

The present analysis has several strengths including a large sample size, a well phenotyped population, randomized controlled trial study design, and a very strong adherence program. Important to note as well are the limitations of the analysis. To allow more consistent and accurate monitoring of adherence rates, the definition of adherence used in this study was developed shortly after the start of the trial. Also, the adherence definition did not incorporate all aspects of the exercise prescription (frequency, duration, intensity and mode)⁷ Adherence for mo 10-12 was based largely on self-report, which may not be accurate; thus, there could be differences in predictors of adoption and maintenance of exercise training not captured in this analysis. Given there was an Adherence Core designed to mitigate dropout and improve adherence to the protocol, adherence within this trial may not be generalizable to other interventions. Finally, this study was limited to patients with chronic, stable HFrEF and who met criteria for participation in a clinical trial. Our ability to assess system level factors influencing adherence to exercise programs outside the structure of a clinical trial was limited.

CONCLUSIONS

Baseline clinical and demographic variables were poor predictors of exercise training adhrence. Poorer adherence in mo 1-3 was associated with younger age, lower income, more severe mitral regurgitation, shorter 6MWT distance, lower exercise capacity, and Black or African American race; adherence in mo 10-12 were similarly associated with the same variables. However, the greatest predictor of adherence in mo 10-12 was adherence during in mo 1-3, suggesting if an individual is able to get through the initial mo of exercise they will most likely adhere through at least 12 mo. Nonetheless, additional research is needed to identify stronger predictors of exercise adherence. This research will facilitate the development of strategic interventions which target exercise training adherence both in clinical trials and in clinical practice.

Key Perspective.

What is novel?

• Few studies have assessed characteristics and predictors of exercise training adherence, especially among individuals with heart failure.

• Though baseline clinical and demographic characteristics are important to assess when predicting exercise training adherence, they provide very limited information for identifying patients with heart failure who are at risk for poor adherence to exercise interventions.

What are the clinical and/or research implications?

• The findings from the present analysis suggest that many of the variables influencing exercise adherence are currently unrecognized or poorly understood.

• Implying, future research is needed to explore a broader range of variables, from system level factors and the physical environment to genetic determinants; the complex interactions of multiple variables; the influence of these variables in important sub-populations; and changes over time and in response to exercise training or adherence interventions.