Measuring long-term physical activity in patients with heart failure and its relation to health-related outcomes

Abstract

Aim

Current physical activity guidelines often rely on self-reported or short-term device-based monitoring. We analysed long-term activity data via acceleration and continuous optical heart rate measurement and hypothesised a temporal association between volume and intensity and health-related outcomes in heart failure.

Methods

This prospective observational trial reanalysed longitudinal data from 156 patients with heart failure at baseline, 6 months, and 12 months. A multisensory activity tracker captured daily step count and continuous heart rate data over a 1-min sampling interval from patients with New York Heart Association (NYHA) Class I (n = 66) and Class II/III (n = 90) over 12 months. Linear mixed-effects models and univariate linear regression models were used to assess the associations between step count, moderate physical activity (MPA) and vigorous physical activity (VPA) and health-related outcomes [cardiorespiratory fitness, body composition, systemic inflammation levels, N-terminal pro brain natriuretic peptide (NTproBNP), and Kansas City Cardiomyopathy Questionnaire (KCCQ)].

Results

Patients with NYHA I and NYHA II/III spent a median (interquartile range) of only 225 (105;412) vs 167 (72;326) min per week with MPA, 17 (7;44) vs 9 (4;23) min per week with VPA, and 8021 (6474;9910) vs 6438 (5181;8627) steps per day, respectively. VPA was positively associated with maximum oxygen uptake (ml/kg/min) (VO_2max; β=3.815, 95% CI [1.430–10.186], P = .008), and MPA was inversely associated with NT-proBNP (pg/ml) (β=−0.189, 95% CI [−0.353 to −0.025], P = .024). For step count, a positive association with VO_2max (ml/kg/min) (β=0.545, 95% CI [0.295–0.795], P < .001), and a negative association with fat mass (%) (β=−1.058, 95% CI [−1.441 to −0.674], P < .001), C-reactive protein (mg/l) (β=−0.057, 95% CI [−0.097 to −0.017], P = .005), and interleukin-6 (pg/ml) (β=−0.022, 95% CI [−0.043 to −0.001], P = .041) was observed. Absolute changes (Δ) for improved VO_2max (ml/kg/min) (r = 0.197, P = .043) and reduced fat mass (%) (r = 0.205, P = .014) were associated with VPA between baseline and 12 months. No relationship was found with KCCQ.

Conclusions

Long-term device-based physical activity monitoring showed that patients with heart failure met recommendations for MPA, yet vigorous activity exceeding 60% of heart rate reserve was almost entirely absent. Higher aerobic physical activity was associated with greater VO_2max, lower fat mass, reduced systemic inflammation (C-reactive protein and interleukin-6), and lower NT-proBNP, but not with KCCQ, whereas only VPA, unlike MPA, was associated with higher VO_2max and lower fat mass.

Graphical Abstract

Graphical Abstract.

Introduction

Obesity, chronic systemic inflammation, inactivity, and decreased cardiorespiratory fitness are risk factors and consequences of the pathophysiology and progression of heart failure (HF).^1–4 The European Society of Cardiology (ESC) guidelines recommend physical activity as a potential corrective action for the development of HF.⁵ The recommendations for adults of all ages are to strive for at least 150–300 min/week of moderate physical activity (MPA) or 75–150 min/week of vigorous physical activity (VPA), or an equivalent combination.⁶ For patients with stable New York Heart Association (NYHA) I–III, aerobic exercise/physical activity is recommended in ESC sport cardiology guidelines with an optimal dose of 20–60 min (duration), 3–5 days/week (frequency) and 40–80% of maximum oxygen uptake (VO_2max) (intensity).⁷ However, current recommendations predominantly rely on self-reported information, which is limited by subjectivity, recall bias, and lower reproducibility, or short-term device-based accelerometry periods of four to seven days. In a direct comparison between self-reported (questionnaire) and device-based (accelerometry) measurement methods, self-reported measurement methods significantly overestimated physical activity, and the differences increased with higher intensity (VPA) in healthy adults and patients with HF.^8,9 Device-based measurements can assess more dimensions of physical activity (e.g. frequency, intensity, duration, and type) and are considered superior to self-reporting.

In recent years, wearable devices have been used more and more frequently in research, but there is limited evidence for observation periods extending beyond 1 week.¹⁰ Moreover, there is a large variety of devices with different sensors and algorithms, each with limited valid data and different reported values, which hinders the comparability of results.¹¹ Most activity monitors measure activity counts via acceleration signals that exceed a cut-off point and are therefore ‘counted’ as light, moderate or vigorous activity. However, device-based accelerometry to measure activity intensity is limited for health research purposes, so continuous heart rate measurement to quantify physical activity intensity during free-living is currently recommended.¹² In addition, a current position paper from ESC for measuring physical activity recommends long-term physical activity monitoring to explore the relation between physical activity and traditional measures of functional status (cardiorespiratory fitness).¹¹

This observational study aimed to capture and evaluate long-term physical activity of those patients with HF via acceleration and optical heart rate sensors. We used a device-based continuous heart rate measurement to quantify the intensity of physical activity in free-living conditions for 1-min intervals. Furthermore, the temporal relationship to HF-related health outcomes should be investigated. Therefore, two main research objectives were investigated: First, we analysed and characterised 12 months of continuously tracked aerobic physical activity using consumer wearable devices, differentiated by step count (volume) and time spent in 10% clusters of heart rate reserve (HRR) (intensity). Second, we evaluated the association between aerobic physical activity and cardiorespiratory fitness (VO_2max, peak power output), body composition (lean body mass and fat mass), biomarkers [C-reactive protein (CRP), interleukin-6 (IL-6), N-terminal pro brain natriuretic peptide (NT-proBNP)], and health status [Kansas City Cardiomyopathy Questionnaire (KCCQ)]. We hypothesised that higher moderate-to-vigorous physical activity (MVPA) or daily step count would be associated with improved HF-related health outcomes.

Methods

Study design and participants

This prospective observational analysis used longitudinal data (baseline, 6 months, and 12 months) from the HITS (Heart Failure, Individualised Training, Telemonitoring, and Self-management) project. It followed the STROBE guidelines and reanalysed the association between long-term physical activity and HF-related health outcomes.¹³ HITS was a multicenter randomised controlled trial with two study arms (A and B) that implemented a multimodal, online-based intervention for patients with different stages of HF (Study arm A with NYHA I, Study arm B with NYHA II/III) to improve physical activity, quality of life, and performance and reduce hospitalisation according to current ESC guidelines.⁵ All procedures described in this study were performed in accordance with the principles of the Declaration of Helsinki. The HITS study protocol was approved by the Ethics Committee of the University of Leipzig (reference number 479/19-ek) and by local ethics committees for the study centres. The study was registered in the German Clinical Trials Register under trial ID DRKS00019022 on 28 May 2020. All participants gave their written informed consent to participate. The evaluation report has previously been published.¹⁴

From September 2020 until June 2022, 341 participants were recruited and randomised to the original HITS trial intervention group (Figure 1). Inclusion criteria for the HITS main study were patients aged 18 years or older with existing or newly diagnosed HF at clinical symptom stages NYHA Class I, II, and III. For the latter, we followed the ESC diagnostic criteria based on evidence of ventricular overload, assessed by NT-proBNP (cut-off >125 pg/ml), and incorporated clinical and echocardiographic assessment.⁵ Classification into NYHA Stages I, II, or III was based on existing diagnoses or on the study physicians’ assessment of symptom severity during physical activity. 247 participants completed the 12-month study period, during which 94 dropouts were recorded. The influence of serious clinical events was excluded via chart review of the medical history.

Figure 1.

Flow diagram illustrating the composition of the observed population. Inclusion criteria were: successful participation in the activity-tracked intervention group of the Heart failure, Individual training, Telemonitoring and Self-management study (n = 247). Exclusion criteria were: No baseline test for maximum oxygen uptake (n = 62), pacemaker or cardiac assist device due to the bias of invalid heart-rate-based intensity assessments (n = 14), and less than 16 weeks of device-based activity-tracking data (n = 15). A total of n = 156 patients met the criteria for this analysis.

For the present secondary analysis, the inclusion criteria were successful participation in the activity-tracked intervention group of the HITS study. The exclusion criteria were no baseline VO_2max test (n = 62), pacemaker or cardiac assist device due to the bias of invalid heart-rate-based intensity assessments (n = 14), and <16 weeks of device-based activity-tracking data (n = 15). A total of n = 156 patients met the criteria for this analysis (Figure 1).

Physical activity tracking

All participants were given a consumer multisensory wearable device at baseline (vívoactive 4 and 4 s; Garmin, Olathe, Kansas, USA) to determine volume (daily step count) via acceleration sensors and intensity (time spent in 10% clusters of HRR) via wrist-based optical heart rate sensors (photoplethysmography [PPG]) of physical activity throughout the 52-week study period. The activity tracker continuously measures heart rate and should be worn 24 h a day. The acceleration sensor measures multidirectional movement and was used to determine daily step count.

Heart rate was used to assess activity intensity and, by extension, to calculate MPA and VPA by the HRR. The use of continuous heart rate allows analysis of intensity from 1-min daily activity-sampling intervals of awake time. HRR with 10% clusters was used to define intensity target zones (i.e. 10%, 20%, 30%, etc.). The average heart rate of every tracked minute was assigned to a target zone. The maximum heart rate (HR_max) was determined as the highest averaged over 30-s intervals during the cardiopulmonary exercise test (CPET). The highest HR_max value across the three measurement points, or the highest value observed in the free-living heart rate data, was used in the calculation. To determine the resting heart rate (RHR), the nocturnal heart rate was assessed to identify the lowest mean value within the reported sleep window. If no sleep data were available, 10-min awake intervals with the lowest heart rate and no movement (steps=0) were identified. For the calculation of MPA and VPA, the relative exercise intensities are 40–59% and 60–100% of the HRR, respectively.

The activity data were imported using the Garmin Health software development kit to take control of the wearables without requiring Garmin web service integration. The collected data were uploaded from the devices to a tablet-based native application (HITS-App) that every participant received at baseline, and from there to our dedicated server. So, we got access to the activity counts for 1-min sampling intervals over a day (up to 1440 min per day). This direct data transfer led to data loss. However, that lost data did not systematically bias physical activity behaviour, as participants were unaware of the loss and wore the devices continuously. We determined that each participant’s wearable device data was representative if it included at least 16 weeks of activity data. The minimum reporting thresholds for data generated by wearable activity trackers, which most studies require, were considered.¹⁰ A valid day was defined as having at least 80% of the awake period (1440 min minus sleep duration) data recorded. A valid week was defined as having at least four valid days. The data were then normalised to seven days, and weekly averages were calculated accordingly. We removed the first and last 2 weeks from the data generation to improve data quality, resulting in 48 weeks in this observational analysis. We aggregated the activity data to obtain mean values over the observational period, standardised to 1 week, to characterise the total observational period. These proxy values were used to assess the association between physical activity and health outcomes.

Outcomes and clinical assessments

The following health outcomes as dependent variables were considered:

• The cardiorespiratory fitness was assessed as VO_2max (ml/kg/min) and peak power output (Watt/kg): The testing procedure was reported in detail in a former study of our working group.¹⁵ We used Dynostics Ergo-Spirometry (Sicada GmbH, Germany) to assess VO_2max (ml/kg/min). The CPET was conducted on an electronically braked semi-recumbent ergometer.

• Body composition: body weight, fat mass, and lean body mass (LBM). We used the BIACORPUS RX 4004 M, MEDI CAL HealthCare GmbH, Germany, phase-sensitive impedance analysis to measure body composition.

• Laboratory markers were obtained from a venous blood sample. Longitudinally, we analysed the association between cardiac-specific biomarkers NT-proBNP, CRP, IL-6, and physical activity. For baseline description only, we presented the following parameters: creatine kinase (CK), isoenzyme creatine kinase-MB (CK-MB), haemoglobin, haematocrit, ferritin, glucose, haemoglobin A1c (HbA1c), cholesterol, low-density lipoprotein (LDL), high-density lipoprotein (HDL), triglycerides, aspartate aminotransferase (ASAT), alanine aminotransferase (ALAT), and alkaline phosphatase (AP). Blood parameters were analysed in the central core laboratory of the dedicated hospital study centres.

• Transthoracic echocardiography with colour flow Doppler and tissue Doppler was used to describe the indicated-specific clinical conditions: ejection fraction (%), left ventricular end diastolic volume index (ml/m²) and E/eʹ ratio. Due to missing data at baseline, 6 months, and 12 months, the average across the three measurement points was used.

• HF-related health status: The 23-item KCCQ was used to quantify physical limitations, symptoms, social interference, and quality of life in patients with chronic HF.¹⁶ The questionnaire includes multiple scales and summary scores to map different HF manifestations.

Independent variables and covariates

In addition to the physical activity as the independent variable, the following covariates were included as fixed effects (as dichotomous or metric variables) in our linear mixed-effects models:

• The two study arms of the original HITS trial as a dichotomous variable: NYHA Class I (Study arm A) or NYHA Class II/III (Study arm B).

• Age: metric variables calculated from the birth date until the date of study enrolment.

• Gender: dichotomous variable, male or female.

• Measurement time point: baseline, after 6 months, and after 12 months.

Other potential covariates were excluded from the analyses: Outcomes were related to body mass or directly assessed via body composition measurements. The NYHA classification already considers cardiovascular diseases and their impact on outcomes in the model.

Statistics

All analyses were performed using IBM SPSS Statistics (version 29.0, IBM Corp., Armonk, NY, USA), and Figure 2 and S1 were displayed using GraphPad Prism (Version 9; GraphPad Software Inc., California, USA). Statistical significance was P < .05 (two-tailed) for all tests.

Figure 2.

(A1–A3) Moderate physical activity and vigorous physical activity in min per week, and steps per day as median with 95% confidence interval during the observational period for New York Heart Association Class I (n = 66) and II/III (n = 90) with a mean (SD) of valid data recording 35 ± 8.5 weeks; min: 16; max: 48. (B) Steps per day as median with 95% confidence interval for every week during the observational period for the total group (n = 156) with included and excluded periods of strict lockdown during COVID-19 nationwide quarantine from November 2020 to May 2021. (C1, C2) Free-living heart rate with a 1-min sampling interval displayed in 10% clusters (Zones) and clustered according to the common cut-off points by the ACSM based on heart rate reserve as average for one day as bar charts with median and 95% confidence interval during the observational period. Numbers above the bars indicate the absolute (min) and relative time (%) spent in each zone per day

Baseline characteristics are presented as medians with interquartile range (IQR), with numbers and percentages (Table 1). The differences between the NYHA classes were analysed, depending on the distribution and the scale level, using a t-test for independent samples (Mann–Whitney U test) and a chi-square test. MVP, VPA, %HRR zones, and daily step count were presented descriptively as the medians and with 95% confidence intervals (CIs) for the entire 48-week study period (Figure 2).

Table 1.

Baseline clinical characteristics

	Baseline
	Total = 156	NYHA I = 66	NYHA II/III = 90	P-value
Age	69 (63;74)	68 (59;74)	69 (64;74)	.634
BMI, kg/m2	27.2 (24.5;30.6)	26.0 (22.8;29.1)	28.4 (24.8;31.5)	.015*
WtHR	0.58 (0.52;0.62) [148]	0.55 (0.48;0.60) [62]	0.59 (0.55;0.62) [86]	.009*
Sex, n (%)				.257
Female	78 (50)	29 (44)	49 (54)
Male	78 (50)	37 (56)	41 (46)
Echocardiography
LVEDVi (ml/m2)	40 (34;46) [142]	39 (31;44) [63]	42 (36;47) [79]	.030*
E/e´ ratio	9.0 (7.6;10.9) [132]	8.2 (7.0;9.8) [57]	9.5 (7.9;11.6) [75]	.020*
LVEF (%)	60 (52;68) [140]	62 (55;68) [61]	58 (49;68) [79]	.032*
CVRF, n (%)
Diabetes type 2	17 (11)	7 (11)	10 (11)	1.000
Hypertension	122 (78)	50 (76)	72 (80)	.560
Hyperlipiemia	61 (39)	27 (41)	34 (38)	.741
Alcohol consumption	28 (21) [133]	14 (25) [56]	14 (18) [77]	.575
Current Smoking	0 (0)	0 (0)	0 (0)	—
Medical history, n (%)
CAD	55 (35)	20 (30)	35 (39)	.311
Arrhythmia	52 (33)	16 (24)	36 (40)	.041*
PAD	4 (3)	1 (2)	3 (3)	.638
Obesity	22 (14)	7 (11)	15 (17)	.355
Cardiomyopathy	10 (9)	1 (2)	9 (10)	.045*
COPD	3 (2)	3 (5)	0 (0)	.074
Laboratory markers
CK, µkat/l	1.81 (1.37;2.36) [146]	1.97 (1.42;2.36) [62]	1.67 (1.26;2.43) [84]	.168
CK-MB, µg/l	0.24 (0.21;0.29) [145]	0.24 (0.21;0.30) [62]	0.24 (0.22;0.27) [83]	.783
Hematocrit, %	41.0 (38.9;43.4) [154]	41.1 (39.2;43.0) [65]	40.3 (38.1;43.6) [89]	.422
Haemoglobin, (g/dl)	13.9 (13.1;14.7) [150]	14.0 (13.2;14.7) [63]	13.7 (12.9;14.7) [89]	.274
Ferritin, ng/ml	0.91 (0.53;90.0) [147]	0.71 (0.47;67,0) [63]	1.27 (0.58;107) [84]	.105
Glucose, mmol/l	5.3 (4.9;5.8) [152]	5.2 (4.7;5.7) [64]	5.3 (4.9;5.9) [88]	.295
HbA1c, %	5.6 (5.4;5.9) [148]	5.6 (5.2;5.9) [62]	5.7 (5.4;6.0) [86]	.201
Cholesterol, mmol/l	4.60 (3.70;5.00) [151]	4.39 (3.67;5.40) [63]	4.75 (3.80;5.67) [88]	.292
HDL, mmol/l	1.44 (1.19;1.79) [151]	1.49 (1.26;1.84) [63]	1.38 (1.17;1.77) [88]	.185
LDL, mmol/l	2.72 (2.00;3.55) [151]	2.66 (1.90;3.52) [63]	2.95 (2.00;3.62) [88]	.364
Triglycerides, mmol/l	1.13 (0.92;1.56) [151]	1.03 (0.79;1.38) [63]	1.24 (0.93;1.80) [88]	.012*
ALAT, µkat/l	0.38 (0.29;0.49) [152]	0.38 (0.29;0.57) [64]	0.38 (0.25;0.48) [88]	.469
ASAT, µkat/l	0.40 (0.35;0.47) [152]	0.41 (0.37;0.48) [64]	0.38 (0.34;0.46) [88]	.080
AP, nkat/l	1.19 (0.99;1.41) [152]	1.16 (1.01;1.40) [64]	1.20 (0.98;1.43) [88]	.767
Medication, n (%)
ACE inhibitors	50 (32)	22 (33)	28 (31)	—
Beta blockers	105 (67)	35 (53)	70 (78)	—
Diuretic	36 (23)	11 (17)	35 (39)	—
MRA	18 (12)	3 (5)	15 (17)	—
Digitalis glycosides	7 (4)	3 (5)	4 (4)	—
Sartane	68 (44)	23 (35)	45 (50)	—
SGLT2 inhibitors	7 (4)	3 (5)	4 (4)	—
Cholesterol reducers	78 (50)	28 (42)	50 (56)	—
Platelet inhibitors	51 (33)	19 (29)	32 (36)	—
Antidiabetic drugs	14 (9)	6 (9)	8 (9)	—
Antidepressants	8 (5)	4 (6)	4 (4)	—

Values presented as the median and interquartile range, numbers with percentages, and P-values (group differences, t-test or Mann–Whitney test, and chi-square test), significant P-values (P < .05).

ACE, angiotensin-converting-enzyme; ALAT, Alanine aminotransferase; AP, alkaline phosphatase; ASAT, aspartate aminotransferase; BMI, body mass index; CAD, coronary artery disease; CK, creatine kinase; COPD, chronic obstructive pulmonary disease; CVRF, cardiovascular risk factor; HbA1c, glycated haemoglobin; HDL, high-density lipoprotein; LDL, low-density lipoprotein; LVEDVi, left ventricular end-diastolic volume index; LVEF, left ventricular ejection fraction; MRA, mineralocorticoid receptor antagonist; NYHA, New York Heart Association; PAD, peripheral artery disease; WHtR, waist-to-height ratio.

We fitted two linear mixed-effects model (LMM) (estimated using restricted maximum likelihood) with a random intercept for participants to predict the association between physical activity and each outcome (dependent variables), reporting marginal and conditional R² values, beta coefficients, 95% CIs, and P-values (Table 3). The variables MPA, VPA, NT-proBNP, IL-6, and CRP were logarithmized based on their distribution. After analysis, the beta coefficients (and 95% CIs) of the variables MPA and VPA, except for the models for NT-proBNP, IL-6, and CRP, were retransformed using the exponential function. Therefore, variables have to be interpreted as odds ratios. To improve data quality at CPET, outliers were identified and excluded using studentised residuals, with cases with absolute studentised residuals greater than 1.5 classified as outliers in the regression of P_max on VO_2max. The first model treated MPA and VPA as independent variables, included NYHA class, age, sex, and measurement time points as fixed effects, and patient-level as random intercepts. Model 1 was calculated for each outcome as the dependent variable. In the second model, steps were treated as an independent variable, with NYHA class, age, sex, and measurement time points as fixed effects, and patient-level as random intercepts. The variable steps have been transformed so that one unit change of the outcomes corresponds to 1000 steps. Model 2 was calculated for each outcome as the dependent variable.

Table 3.

Association of heart failure-related health outcomes with intensity (MPA, VPA) and volume (step count) of physical activity

		MPA and vPA			Step count
	Conditional/marginal R2		Beta (95% CI)	P-Value	Conditional/marginal R2	Beta (95% CI)	P-Value
Cardiorespiratory fitness
VO2max, ml/kg/min	0.89/0.31	MPA	0.678 (0.220 to 2.312)	0.571	0.89/0.34	0.545 (0.295 to 0.795)	<.001a
		VPA	3.815 (1.430 to 10.186)	0.008a
Peak power output, Watt/kg	0.89/0.27	MPA	0.978 (0.883 to 1.084)	0.679	0.88/0.33	0.044 (0.023 to 0.066)	<.001a
		VPA	1.067 (0.980 to 1.162)	0.132
Body composition
Fat mass, kg	0.97/0.12	MPA	1.426 (0.083 to 24.459)	0.805	0.97/0.23	−1.416 (−1.996 to −0.835)	<.001a
		VPA	0.377 (0.036 to 3.979)	0.415
Fat mass, %	0.94/0.31	MPA	1.279 (0.193 to 8.466)	0.798	0.94/0.40	−1.058 (−1.441 to −0.674)	<.001a
		VPA	0.309 (0.064 to 1.478)	0.140
Lean body mass, kg	0.98/0.64	MPA	0.349 (0.065 to 1.862)	0.216	0.98/0.64	−0.372 (−0.740 to −0.004)	.048a
		VPA	3.684 (0.919 to 14.761)	0.066
Waist circumference, cm	0.94/0.26	MPA	1.454 (0.068 to 31.249)	0.809	0.94/0.37	−1.682 (−2.302 to −1.063)	<0.001a
		VPA	0.404 (0.032 to 5.176)	0.484
Laboratory marker
C-reactive protein, mg/l	0.58/0.03	MPA	−0.006 (−0.192 to 0.180)	0.949	0.58/0.06	−0.057 (−0.097 to −0.017)	.005a
		VPA	−0.012 (−0.166 to 0.143)	0.882
Interleukin-6, pg/ml	0.524/0.061	MPA	0.013 (−0.084 to 0.110)	0.793	0.52/0.08	−0.022 (−0.043 to −0.001)	.041a
		VPA	−0.002 (−0.083 to 0.078)	0.956
NT-proBNP, pg/ml	0.78/0.08	MPA	−0.189 (−0.353 to −0.025)	0.024a	0.78/0.05	0.007 (−0.030 to 0.044)	.696
		VPA	0.098 (−0.039 to 0.235)	0.158
Quality of life
KCCQ overall score, points	0.73/0.09	MPA	0.839 (0.043 to 16 346)	0.907	0.73/0.09	0.246 (−0.406 to 0.889)	.457
		VPA	0.841 (0.071 to 9.924)	0.890

Values are presented as conditional and marginal R², beta coefficient with 95% confidence intervals, and P-values. MPA, VPA, C-reactive protein, interleukin-6, and NT-proBNP were log-transformed, so the coefficients must be interpreted as odds ratios. After analysis, the beta coefficients (and 95% CIs) of the variables MPA and VPA, except for the models for NT-proBNP, IL-6, and CRP, were retransformed using the exponential function. For steps, the changes in outcomes relate to an increase of 1000 steps per day.

KCCQ, Kansas City Cardiomyopathy Questionnaire; MPA, moderate physical activity (%HRR 40–59); NT-proBNP, N-terminal prohormone of brain natriuretic peptide; NYHA, New York Heart Association; VO_2max, maximum oxygen uptake; VPA, vigorous physical activity (%HRR 60–100).

^aLinear mixed-effects model (LMM), significant P-value <.05.

We descriptively presented all outcomes at baseline and after 12 months (Table 2). The results were modelled by group (NYHA class) and time (baseline and 12 months) to estimate expected contrasts between NYHA classes. Additionally, univariate regression models were employed to further support correlations between average VPA (min/week) or daily step count and absolute changes (Δ) in health-related outcomes from baseline to 12 months (Figure 3).

Table 2.

Health-related outcomes at baseline and after 12 months in NYHA classes, as well as the median with the 25th and 75th percentiles of device-based physical activity 12 months

	NYHA i Mean (SD)		NYHA iI/III Mean (SD)
	Baseline	After 12 months	Baseline	After 12 months	Time effectb  P-value	Group effectb  P-value	Interaction effectb  P-value
Cardiorespiratory fitness
VO2max, ml/kg/min	25.2 (5.9) [59]	25.3 (6.8) [47]	18.7 (4.1) [85]	20.1 (5.0) [65]	.001b	<.001b	.050
Peak power output, Watt/kg	1.69 (0.53) [59]	1.77 (0.60) [48]	1.25 (0.36) [85]	1.36 (0.42) [66]	.012b	<.001b	.468
Body composition
Fat mass, kg	22.4 (9.8) [61]	21.6 (10.3) [62]	28.6 (13.2) [87]	27.2 (12.7) [85]	.016b	.002b	.922
Fat mass, %	28.0 (9.2) [61]	27.0 (9.6) [62]	32.9 (8.9) [87]	32.2 (9.0) [85]	.039b	<.001b	.864
Lean body mass, kg	55.8 (10.4) [61]	56.1 (10.6) [62]	56.0 (12.4) [87]	55.2 (12.3) [85]	.222	.811	.443
Waist circumference, cm	95 (15) [62]	93 (14) [66]	100 (14) [86]	98 (15) [86]	<.001b	.035b	.349
Laboratory marker
C-reactive protein, mg/l*	1.0 (0.6;1.8) [66]	1.1 (0.6;2.2) [66]	1.6 (0.6;2.9) [90]	1.1 (0.6;2.6) [89]	.267	.047b	.738
Interleukin-6, pg/ml*	3.0 (2.5;3.5) [63]	3.5 (3.0;3.7) [66]	3.0 (2.5;4.1) [83]	3.5 (3.0;4.3) [88]	.020b	.042b	.352
NT-proBNP, pg/ml*	216 (156;298) [66]	213 (133;414) [66]	249 (162;495) [90]	273 (150;484) [89]	.327	.119	.756
KCCQ (points)
Overall	83 (14) [64]	86 (12) [58]	75 (16) [87]	82 (14) [84]	<.001b	.005b	.131
Symptom	86 (16) [64]	89 (14) [58]	78 (18) [87]	84 (16) [84]	<.001b	.005b	.299
Functional	87 (14) [64]	89 (12) [58]	79 (15) [87]	84 (15) [84]	<.001b	.002b	.122
Clinical	85 (14) [64]	88 (12) [58]	77 (16) [87]	83 (15) [84]	<.001b	.002b	.088
Physical activity
MPA (%HRR 40–59), min/week*	225 (105;412)		167 (72;326)		.260
VPA (%HRR 60–100), min/week*	17 (7;44)		9 (4;23)		.060
MVPA recommendation achieved, n (%)	42 (63,6)		56 (62,2)
Steps per day*	8021 (6474;9910)		6438 (5181;8627)		.005a

Values presented as mean and standard deviation or as median* and interquartile range*, and number.

KCCQ, Kansas City Cardiomyopathy Questionnaire; MPA, moderate physical activity; NT-proBNP, N-terminal prohormone of brain natriuretic peptide; NYHA, New York Heart Association; VO_2max, maximum oxygen uptake; VPA, vigorous physical activity.

^aMann-Whitney U test, significant difference (P < .05).

^bMain effects of mixed-effects model with repeated measurements, significant P-value (P < .05).

Figure 3.

Association of absolute changes (Δ) in heart failure-related health outcomes from baseline to 12 months with physical activity. Scatterplots show correlations between average vigorous physical activity (A1) or daily step count (A2) and VO_2max Diff (ml(kg/min), between average vigorous physical activity (B1) or daily step count (B2) and fat mass Diff (%), between average vigorous physical activity (C1) or daily step count (C2) and NT-proBNP Diff (pg/ml), and between average vigorous physical activity (D1) or daily step count (D2) and Kansas City Cardiomyopathy Questionnaire—overall score Diff (Points). The Pearson correlation coefficients (r), P-values and numbers are shown. The solid lines represent the linear regression lines with 95% confidence intervals. The discrepancy in numbers is due to missing values at baseline or after 12 months. *significant P-value < .05.

Results

The results are divided into two sections, depending on the research objectives. After presenting baseline clinical characteristics, the first section characterises the results of long-term device-based physical activity tracking. The second section provides the results for the temporal association between physical activity and HF-related health outcomes.

Baseline clinical characteristics

Table 1 presents the baseline and clinical characteristics for the total cohort (n = 156) of HF patients and for the subcohorts, NYHA Class I (n = 66) and II/III (n = 90). NYHA Class I and II/III patients differed significantly in left ventricular ejection fraction (LVEF%) and diastolic function (E/eʹ ratio), body mass index, waist-to-height ratio, medical history (arrhythmia and cardiomyopathy), and triglycerides.

Objective I: long-term device-based physical activity tracking

A1–A3 of Figure 2 illustrates the descriptive volume and intensity of physical activity [daily step count, MPA (%HRR 40–59) min/week, and VPA (%HRR 60–100) min/week] per week, presented as median with 95% confidence interval, during the entire study period in the NYHA I and NYHA II/III classes. On average, participants had 35 valid weeks of data (SD: 8.5; min: 16; max: 48). The longitudinal data demonstrate a decrease in activity levels for NYHA II/III class patients regarding activity intensity, particularly for vigorous activity. The large confidence intervals also show the interindividual differences for behaviour in MPA and VPA. C1 of Figure 2 demonstrates the average per-day time spent in physical activity intensity zones, as a 10% cluster of HRR over the entire study period. Only 7.5 min per day for NYHA Class I and 5.3 min per day for NYHA Class II/III were spent in HRR zones over 50%. Standardised for a week, the VPA (median [IQR] for NYHA I: 17 [7;44] min/week; NYHA II/III: 9 [4;23] min/week) consistently falls below the recommended minimum physical activity threshold of 75 min of vigorous intensity activity (Table 2, C2 of Figure 2). In the higher zones, between 70% and 100% of HRR, the proportion was <4 min per week. On the other hand, MPA (median [IQR]) for NYHA I: 225 [105;412] min/week and NYHA II/III: 167 [72;326] min/week) for the most part exceeded the recommended minimum physical activity threshold. So, 63.6% of NYHA I and 62.2% of NYHA II/III participants met the recommended thresholds of 150 min/week MPA, 75 min/week VPA, or an equivalent combination.

For steps, we found significant differences between groups: NYHA I: 8021 [6474;9910] and NYHA II/III: 6438 [5181;8627] steps per day (Table 2).

Objective II: temporal association between physical activity and health outcomes

The HF-related health outcomes are presented descriptively at baseline and after 12 months (Table 2). As hypothesised in the context of HF progression, there were group differences for the health-related outcomes between NYHA I and NYHA II/III, except for lean body mass and NT-pro BNP. However, the differences in time, group, and interaction were hypothesised, but were negligible in the context of this study's research objectives.

Association of health outcomes with physical activity

Table 3 presents the association between physical activity and health outcomes. The mixed-effect models revealed that only increased VPA was significantly associated with higher VO_2max (ml/kg/min) (β=3.815, 95% CI [1.430–10.186], P = .008), and higher MPA was significantly associated with lower NT-proBNP (pg/ml) (β=−0.189, 95% CI [−0.353 to −0.025], P = .024). A 10% increase in VPA is in relation to a higher VO_2max by approximately 0.38 ml/kg/min (3.815/10 = 0.382), and a 10% higher MPA is associated with an approximate 1.9% decrease in NT-proBNP.

For daily step count, the analysis showed also a significant positive association with VO_2max (ml/kg/min) (β=0.545, 95% CI [0.295–0.795], P < .001), and peak power output (Watt/kg) (β=0.044, 95% CI [0.023–0.066], P < .001), and a significant negative association with fat mass (kg) (β=−1.416, 95% CI [−1.996; −0.835], P < .001), fat mass (%) (β=−1.058, 95% CI [−1.441; −0.674], P < .001), LBM (kg) (β=−0.372, 95% CI [−0.740; −0.004], P = .048), waist circumference (cm) (β=−1.682, 95% CI [−2.302; −1.063], P < .001), CRP (mg/l) (β=−0.057, 95% CI [−0.097; −0.017], P = .005), and IL-6 (pg/ml) (β=−0.022, 95% CI [−0.043; −0.001], P = .041). The beta coefficient of IL-6 (exp(–0.022) = 0.978) and CRP (exp(–0.057) = 0.945) have to be interpreted as odds ratios and retransformed using the exponential function. An increase of 1000 steps per day was associated with higher VO_2max by 0.55 ml/kg/min, peak power output by 0.04 Watt/kg, and reduces fat mass by 1.4 kg or 1.1%, waist circumference by 1.7 cm, CRP by approximately 6% (1/0.945 = 1.059 or 5.9%), and IL-6 by approximately 2% (1/0.978 = 1.022 or 2.2%).

No significant associations were found between MPA, VPA, and steps per day and the KCCQ overall summary scales.

Association of absolute changes (Δ) in health outcomes with physical activity

Figure 3 shows the association of absolute changes (Δ) in different health-related outcomes with average VPA (min/week) and daily step count. When using changes (Δ) from baseline to 12 months, univariate correlations were significant between increased VPA and improvement ΔVO_2max (r = 0.197, P < .043), and reduced Δfat mass (%) (r = 0.205, P = .014), and between increased steps per day and reduced Δfat mass (%) (r = 0.166, P = .047). No significant associations were found between VPA and steps per day and absolute changes (Δ) in NT-proBNP and KCCQ overall summary scale. Other absolute change (Δ) outcomes were not significant and are provided in the Supplementary Figure S1.

Discussion

To our knowledge, this is the first prospective observational cohort study to measure device-based aerobic physical activity over 12 months in HF patients via an acceleration sensor and continuous optical heart rate measurements with a 1-min sampling interval. The activity data indicate that HF patients met current MPA recommendations despite not effectively engaging in vigorous activity. Our results establish that higher aerobic physical activity was associated with increased cardiorespiratory fitness (VO_2max and peak power output), lower fat mass, lower systemic inflammation (CRP and IL-6) and lower myocardial stress (NT-proBNP), but not with KCCQ.

Objective I: long-term device-based physical activity tracking

The long-term activity tracking highlighted the intraindividual differences for MVPA during the observation period. This suggests that longer observation periods than one week should be considered when assessing physical activity. The majority of included HF patients (NYHA I: 63,6%; NYHA II/III: 62,2%) met the recommended guidelines for MVPA (%HRR > 40%). However, minutes of vigorous activity—i.e. above 60% of HRR—were almost entirely missed during the entire observation period. Wang et al. (2025) reported that maintaining VPA at 30–60% of total MVPA appears to be associated with the greatest reduction in CVD mortality risk.¹⁷ MVPA appears to decrease with higher NYHA class or lower physical performance (VO_2max).¹⁸ Another study that measured MVPA device-based via accelerometry consecutively over 12 days also found this association.¹⁹ Our results were consistent with this relationship. In a large observational study, Ho et al. (2022) analysed acceleration data and found that most healthy individuals engaged in less than 25 min of VPA, and that even 25 to 50 min weekly could significantly reduce HF incidence.²⁰ Our results confirmed the low level of vigorous activity above 60% of the HRR. On average, vigorous activity lasted less than 19 min per week for NYHA Class I and less than 9 min per week for NYHA Class II/III. Above 70% of HRR, there was effectively no activity. A current meta-analysis, which included self-reported and, for the most part, device-based activity data, reported that patients with HF (80% were NYHA II and III) are typically low-active, with a mean step count per day of 5040 (95% CI: 4272 to 5807).²¹ The NYHA II/III results exceeded 5040 steps/day in this analysis. However, the optimal number and cut-off point of daily step count for disease-specific conditions are still unclear, but health benefits start with even 2500 steps per day.²²

The consumer wearable devices we used (Vivoactive 4 and 4 s) are reliable and valid for measuring daily step count for research and clinical purposes.^23,24 The accuracy decreases in patients with HF, especially at lower, non-self-selected walking speeds.²⁵ However, a standardised validation protocol should be used in future studies to validate consumer wearable step counters.²⁶ Consumer wearable devices with acceleration sensors are valid for measuring MVPA, whereby essential sources of bias are different cut-off points for light, moderate and vigorous activity intensity.²⁷ Additionally, cut-off points validated in a specific age group might not be valid for other groups due to differences in age, performance, or disease.¹¹ In recent years, optical heart rate sensors for measuring activity intensity have become increasingly common and demonstrate acceptable validity.²⁸ Moreover, Neudorfer et al. (2024) reported that in various laboratory­based and free­living activities, the validity and reliability of optical heart rate sensors were high enough to determine moderate or vigorous activity minutes.²⁹ The HRR we used, as a range-based exercise intensity, is a relative measure and is recommended over fixed HRmax or VO_2max zones.³⁰ Some authors advise a shift from a ‘range­based’ to a ‘ventilatory threshold–based’ exercise intensity determination for patients with cardiac disease.³¹ However, from a practical perspective, the comprehensive determination of ventilatory or metabolic thresholds would be more accurate, but more costly. Furthermore, these thresholds cannot be applied to other types of movement or sports.

ß-Blockers affect resting heart rate, peak heart rate, and consequently the HRR. However, the intake of ß-blockers has no or only a small influence on aerobic capacity in healthy adults.^32,33 The relationship between the percentage of oxygen consumption reserve and percentage of HRR in HF patients on optimised beta-blocker therapy appears reliable.³⁴ The ESC recommends using heart rate to prescribe and evaluate exercise intensity in persons taking ß-blockers, provided aerobic capacity is determined while taking the medication.⁷ We measured resting and peak heart rates while on β-blocker medication. The resting heart rate was measured dynamically as the average over four weeks and used to calculate the HRR. The sampling interval for heart rate was 1 min over the awake period of 1 day, and every minute was assigned to a 10% cluster of the HRR. So, the setup we used for device-based activity monitoring complies with current best-practice recommendations.¹²

Objective II: temporal association between physical activity and health outcomes

Our results underline an intensity (VPA) and volume-dependent significant positive association between cardiorespiratory fitness (VO_2max and peak power output) and aerobic physical activity. The results demonstrate that increased VPA was associated with improved maximum oxygen uptake after a 12-month observation period. Over the long-term observation period, we established a direct temporal association between activity and increased cardiorespiratory fitness. Another device-based (accelerometry) study with 108 HF patients supported the correlation between total MVPA and maximal oxygen uptake.¹⁹ The different amounts of moderate or vigorous intensities were not investigated. In another secondary analysis, Bobenko et al. (2018) analysed the amount and intensity of self-reported physical activity for HF patients with NYHA Class II and III. They demonstrated that only vigorous physical activity was significantly associated with maximal exercise capacity.³⁵ Interestingly, although their reported maximal oxygen uptake was slightly lower than in our study, self-reported vigorous physical activity was substantially higher. However, the measured maximum oxygen uptake in the HF patients was below the threshold of six METs, which is the cut-off point for absolute vigorous intensity activities, so the self-reported amount of VPA had to be clearly overestimated. This demonstrates the following problems with self-reported physical activity: It is based on absolute intensity measured in METs as a fixed value of 3,5 ml/kg/min oxygen uptake or 1 kcal/kg/hour, respectively, of young healthy adults, rather than on an individual measurement of basal metabolic rate. However, a fixed value of 3,5 ml/kg/min base metabolic rate is not transferable to HF patients. The basal metabolic rate per kg of body weight varies among individuals and depends, among other factors, on LBM.³⁶ Dibben et al. (2020) demonstrated that HF patients had basal metabolic rates 24% lower than the standardised value of 3.5 ml/kg/min.³⁷ This might lead patients with poor performance to overestimate their MVPA, overestimate the thresholds for improvement in cardiorespiratory fitness, and introduce significant bias in MVPA calculations. Additionally, the amount of self-reported physical activity, which is based on absolute intensity measured in METs, does not consider the individual´s maximum cardiorespiratory fitness (VO_2max), and therefore does not reflect the relative intensity (percentage of VO_2max or HRR) of their activity.

Physical training in patients with HF is associated with reduced NTproBNP and biomarkers of inflammation (CRP and IL6).³⁸ Our results showed an association between lower NT-proBNP higher MPA (%HRR 40–59). However, we did not observe any association between physical activity and the KCCQ overall summary scale in patients with HF. Another study also demonstrated this with device-based data collection and comparable or higher amounts of physical activity, in which aerobic capacity improved, but not the KCCQ.¹⁹ Kwast et al. (2025) showed no differences in arterio-venous oxygen difference, cardiac output, or VO_2max between patients with NYHA Class I and II, which were classified via KCCQ, compared to the NYHA classification via cardiopulmonary exercise test.¹⁵

Eisen et al. (2014) showed a positive relationship between increased CRP and HF risk in CAD patients.³⁹ Zheng et al. reported in a meta-analysis of RCTs that aerobic exercise (2–12 months, 2–5 sessions weekly, each lasting 20–90 min, HRR of 45–80%) reduced inflammation markers such as CRP and IL-6 in healthy middle-aged and older adults.⁴⁰ We observed an inverse association between IL-6, CRP, and daily step count, although the volume and intensity of our aerobic physical activity were substantially lower. However, we didn’t find an intensity-dependent correlation of inflammation reduction between baseline and 12 months.

Older HF patients have reduced LBM compared with healthy controls.⁴¹ Our findings underline the different relationships between the volume and intensity of physical activity and improvements in body composition. Higher daily step count was associated with lower absolute and relative fat mass and waist circumference. Increasing daily step count and VPA were associated with reduced relative fat mass (Δ) from baseline to 12 months.

Limitations

The present study has limitations that have to be acknowledged. First, this study was an observational analysis, so we cannot make any causal claims. Second, consumer wearable devices may be faulty in capturing certain types of physical activity, especially strength training. So, the possible response of resistance training on our results is unclear. Third, we did not analyse light activity in our model. Training studies suggest that a minimum effective intensity of 30% VO₂R increases cardiorespiratory fitness and is suitable for lower-fit subjects.⁴² Fourth, the quarantine due to COVID-19 from November 2020 to May 2021 may have reduced habitual physical activity in HF patients.⁴³ However, the main observation period and the highest proportion of participants in our analysis occurred after the strict COVID-19 lockdown, so COVID-19 was not a confounding factor (Figure 2B). Fifth, step-counting accuracy decreases at lower walking speeds. This technical bias might have underestimated the step count, especially in older participants of NYHA Class II/III.

Conclusion

Long-term device-based physical activity tracking showed that HF patients met MPA recommendations, but vigorous activity exceeding 60% of HRR was almost entirely missed. Higher aerobic physical activity was associated with increased cardiorespiratory fitness (VO_2max and peak power output), lower fat mass, lower systemic inflammation (CRP and IL-6) and lower myocardial stress (NT-proBNP), but not with KCCQ. Despite low levels of vigorous intensity, only VPA was associated with improved VO_2max and reduced fat mass, whereas MPA was not. Consumer activity trackers with continuous optical heart rate measurement will allow researchers to precisely analyse long-term physical activity intensity, thereby revising activity recommendations based on scientific evidence.

Author contributions

C.L. holds first authorship. S.K., C.P., J.V., M.K., and P.L. contributed to wearable device data acquisition and analysis, statistical analysis, literature research, hypothesis development, and study development. R.F. contributed to data acquisition, statistical analysis, and revising original draft. U.T. advised implementing the study design at the study sites and is involved in recruitment and data acquisition. M.B. conceived and designed the original HITS study. All authors have read and approved the submitted manuscript.

Supplementary data

Supplementary data are available at ESC Heart Failure online.

Declarations

Disclosure of Interest

All authors declare no disclosure of interest for this contribution.

Data Availability

No data were generated or analysed for this manuscript.

Funding

This study was funded by the Gemeinsame Bundesausschuss (G-BA) under 01NVF18037. The funder had no role in the study design, data collection, data analysis, data interpretation, report writing, or decision to submit for publication. All coauthors had access to the raw data if needed. The corresponding author had full access to all study data and was ultimately responsible for submitting it for publication.

Ethical Approval

Ethical Approval was not required.

Pre-registered Clinical Trial Number

None supplied.