Quantifying Exercise Intensity to Predict Changes in Walking Capacity in People with Chronic Stroke

Abstract

Objective:

To examine if exercise intensity, quantified as heart rate or training speed, predicts walking outcomes in people with chronic stroke.

Design:

This is a secondary analysis from a larger randomized clinical trial (“PROWALKS”; NIH1R01HD086362).

Setting:

Four, outpatient rehabilitation clinics.

Participants:

Participants with chronic stroke with a walking speed of 0.3–1.0m/s and step-activity of <8000 steps-per-day. This analysis included participants (n = 169; age: 63.1 ± 12.5, 46% female) with complete pre- and post-intervention data.

Interventions:

Participants were randomized into (1) fast-walking training or (2) fast-walking training and step-activity monitoring group. Of importance, participants received up to 36-sessions of 30-minute high-intensity treadmill walking training across 12-weeks.

Main Outcome Measure(s):

The primary outcomes were a pre-to-post intervention change in six-minute walk test distance and fastest walking speed. Exercise intensity was quantified as either a percentage of heart rate reserve or self-selected walking speed.

Results:

Two separate multiple linear regressions with robust errors analyzed the relationship of exercise intensity metrics (% heart rate, training speed) on pre-to-post intervention changes in two walking capacity outcomes (Six-Minute Walk Test, Fastest Walking Speed) after accounting for covariates.

Training speed was a significant predictor of both a change in Six-Minute Walk Test distance (b = 0.359 (95% CI [0.108 – 0.610]), p = 0.005) and Fastest Walking Speed (b = 0.001 (95% CI [0.001–0.002]), p = 0.003). Heart rate was not a significant predictor of either outcome (both p > 0.373).

Conclusions:

Training speed significantly predicts changes in walking capacity outcomes in people with chronic stroke following a fast-walking treadmill intervention. This suggests rehabilitation clinicians may use training speed as the metric of exercise intensity when prescribing walking interventions to people with chronic stroke.

More than 80% of people with chronic stroke experience a severe reduction in their ability to walk.¹ This includes notable decreases in their walking capacity, collectively represented by walking speed and walking endurance.^1–4 Moderate-to-high intensity walking interventions are currently one of the most effective interventions to improve walking capacity in individuals with chronic stroke.^5,6 However, following these interventions, there remains significant variability in walking capacity outcomes across participants.⁶

The aggregation of aerobic and/or walking intervention studies in people with chronic stroke suggests that higher levels of intensity may facilitate greater changes in cardiorespiratory fitness, walking speed, and walking endurance.⁷ Exercise intensity is most frequently quantified as a percentage of heart rate maximum or heart rate reserve (HRR) in moderate-to-high walking interventions for people with chronic stroke. Heart rate is used as both the intensity metric which guides the exercise intervention design (e.g., “training heart rates were established at 70–80% HRR”)⁸ and the “real-time” response when a participant is exercising (e.g., “Treadmill incline was adjusted or weight vests added as needed to keep heart rate in the target heart rate range during all walking”).⁹ While heart rate is a valid and frequently used measure of cardiovascular intensity in response to exercise, it is influenced by other factors, agnostic to exercise, including stress, anxiety, sleep quality, and dehydration.^10,11

The previous studies which have suggested higher intensity exercise predicts greater changes in walking outcomes in people with chronic stroke have had several challenges.^8,12–21 These include small sample sizes, not accounting for important co-variates, and/or only demonstrating a correlation between intensity and outcome.^8,12–21 Secondly, intensity has almost uniformly been quantified based on heart rate. This has limited our understanding to only one metric of exercise intensity.

While many individuals with chronic stroke have low cardiorespiratory fitness, they often also have significant neuromuscular deficits.^22,23 These neuromuscular deficits may limit the cardiovascular intensity people with stroke can reach in a walking intervention, as the cardiovascular system may not be what is primarily limiting the individual. Therefore, it has been suggested that walking speed may provide a more specific measure of exercise intensity in this population.²⁴ While it may be intuitive to assume that faster walking speeds are associated with or equated to higher heart rates, it is unknown if these metrics of exercise intensity are equally predictive of walking outcomes in people with chronic stroke. Probing which parameter of exercise intensity is most predictive of walking outcomes remains largely unexplored.^24,25 Understanding the impact of each exercise intensity metric on walking capacity outcomes is imperative to provide the optimal training parameters for clinicians to use when implementing walking exercise interventions in people with chronic stroke.

The purpose of this secondary analysis was to examine how exercise intensity, quantified as either training heart rate or training speed, would impact changes in walking capacity outcomes in people with chronic stroke after accounting for several covariates. We hypothesized after accounting for important covariates (such as age, baseline walking capacity, exercise volume, or balance self-efficacy), that exercise intensity quantified as either heart rate or training speed, would be a predictor of changes in walking endurance and fastest walking speed following a walking exercise intervention in people with chronic stroke.

METHODS

Participants

This is a secondary analysis from a larger randomized controlled trial “Promoting Recovery Optimization of Walking Activity in Stroke” (PROWALKS”; NIH 1R01HD086362). The full protocol and primary results have been published.^9,26 All study procedures were approved by the University of Delaware Institutional Review Board and participants signed an informed consent prior to enrollment. Participants were 21–85 years old, had their most recent stroke at least 6 months prior to enrollment, had self-selected walking speeds of 0.3–1.0m/s, and could ambulate without the assistance of another person. Of the total 250 participants randomized to PROWALKS, only the 169 participants randomized into the (1) fast-walking training (FAST, n = 89) or (2) fast-walking training and step activity monitoring combined intervention (FAST + SAM, n = 80) were included in this analysis.²⁷ The 81 participants randomized to the step activity monitoring intervention (SAM) were excluded as they did not undergo any treadmill training.

Clinical Evaluation

Prior to randomization, all participants underwent a baseline clinical evaluation which included collecting demographic and medical information. Demographic information included age and biological sex, and medical information included time since the most recent stroke (TSS) and the Charlson Comorbidity Index (CCI). The CCI is a 16-item questionnaire comorbidity scoring system which weights factors based on disease severity and quantifies comorbidity burden.^28,29 Participants also completed the Activities-specific Balance Confidence Scale (ABC), a reliable and valid 16-item self-report questionnaire that measures balance self-efficacy.^30,31

Participants also completed measures of walking capacity. Walking capacity is defined as what a person can do usually as assessed by standardized tests conducted in structured environments such as a clinic or laboratory.³² The six-minute walk test (6MWT) and 10-meter walk test (10mWT) are valid measures and are recommended to quantify walking capacity in the chronic stroke population.^33,34 The 6MWT may represent an individual’s walking endurance and be influenced by their cardiovascular fitness and/or neuromotor function.^33,35–37 The 10mWT can capture individuals self-selected (SSWS) and fastest walking speeds (FWS) over a short distance.^33,38 Participants were instructed to “walk at your normal pace” (SSWS) or “walk at your fastest possible pace” (FWS) and completed three trials at each walking speed, with the average of each trial recorded.²⁶ Walking capacity measures were completed at baseline and after completing the intervention. The previously published protocol outlines all measures collected in the parent trial.²⁶

Exercise Intervention

Participants randomized to the FAST or FAST+SAM intervention received up to 36, 30-minute sessions of treadmill training over 12 weeks.^9,26 The intervention protocol was designed to promote continuous walking at high cardiovascular intensities. The prescribed exercise intensity for all treadmill sessions was 70–80% of an individual’s heart rate reserve (HRR).²⁶ The goal of the FAST intervention was to accumulate as many minutes as possible (maximum 30 minutes) in each session in the target heart rate range (70–80% HRR).

Training physical therapists were first instructed to manipulate treadmill speed to achieve the target heart rate zone. If the participant was not able to attain the target heart rate with only increases in speed, training physical therapists could increase treadmill incline and/or use ankle weights or a weighted vest to achieve the target heart rate. All participants wore a safety harness which did not provide any body weight support. The exercise session was terminated if a participant’s response violated the guidelines set forth by the American College of Sports Medicine (ACSM) for individuals in phase III or IV cardiac rehabilitation.³⁹ Heart rate was monitored continuously (Polar H10 chest-straps), with heart rate and treadmill speed recorded each minute.

In addition to the fast-walking training, participants randomized to the FAST + SAM intervention simultaneously received the SAM intervention which used motivational interviewing and coaching techniques to target progressive increases in daily step-activity.^9,26

Exercise Intensity

For this analysis exercise intensity was quantified in two ways: (1) the average percent HRR and (2) the average training speed attained across the entire intervention. Using the maximum heart rate (MHR) attained on the cardiopulmonary exercise stress test prior to randomization and the individual’s daily resting heart rate (RHR), percent HRR was quantified using the Karvonen method.³⁹ Participants on beta-blocker medications were instructed to take their medications prior to the exercise test (which determined heart rate maximum) and for each training session, which accounted for the use of beta-blocker medications. For each minute of heart rate (HR) data, percent HRR was calculated as:

$$\%\text{HRR}=\left[\left(\text{HR}–\text{RHR}\right)/\left(\text{MHR}–\text{RHR}\right)\right]\times 100.$$

For each minute of treadmill speed data, training speed was calculated as a percentage of an individual’s baseline SSWS as follows:

$$\text{training speed}\left(\%\text{SSWS}\right)=\left[\text{treadmill speed}/\text{baseline SSWS}\right]\times 100.$$

Prior to training speed calculations treadmill speeds were converted from miles per hour to meters per second.

For both exercise intensity metrics, values were averaged across every minute of training data from the intervention.

Statistical Analyses

Descriptive statistics were used for all baseline characteristics and training fidelity metrics. Paired t-tests used for continuous variables and chi-square tests for categorical variables to participants randomized to the FAST and FAST + SAM interventions.

Two separate multiple linear regressions with robust errors were run to analyze the relationship of exercise intensity (% HRR, training speed) on a baseline to post-intervention change in walking capacity outcomes (outcome; 6MWT, FWS) after accounting for covariates. Covariates included age, sex, TSS, CCI, ABC, baseline 6MWT, and total exercise volume. These covariates were chosen based on prior literature which has demonstrated their impact on rehabilitation outcomes or relationship to walking capacity in people with chronic stroke.^13,40–42 If these covariates were not controlled for, it could signal a false relationship between exercise intensity and change in walking capacity outcome that is accounted for by other variables. Baseline 6MWT and baseline FWS were highly correlated in this sample (r = 0.903, p < .001), and both are measures of walking capacity, therefore, only baseline 6MWT was retained in the models. Exercise volume was quantified as the total number of minutes walked in the intervention (e.g., if a participant attended 24 sessions and walked 26 minutes in each session, their exercise volume was 624 minutes). This was included to account for differences between participants in the number or frequency of training sessions and/or number of minutes walked within a session. Statistical analyses were conducted using Statistical Package for the Social Sciences (SPSS) (Version 29.0; IBM, Armonk, New York).

RESULTS

One-hundred twenty-nine participants randomized to the FAST (n = 68) or FAST+SAM (n = 61) intervention completed both baseline and post-intervention walking capacity measures. Participants were (mean±SD) 63±13 years old, 46% female, and were approximately 3.75 years post-stroke (Table 1). Across the intervention, these participants attended 29 training sessions and walked for an average of 28 minutes per session at a heart rate intensity of 63.5% HRR (Table 2). There were no differences in baseline characteristics or training fidelity metrics between participants in the FAST or FAST+SAM interventions, so intervention group was not included as a covariate in the regression analyses (Tables 1, 2).

Table 1:

Participant Characteristics

Characteristic	FAST (n = 68)	FAST + SAM (n = 61)	P
Continuous
Age (years)	63.7 ± 11.3	63.2 ± 13.7	.829
Time Since Initial Stroke (months; median (IQR))	21 (10–65)	21 (13–58)	.524
Body mass index (kg/m2)	29.1 ± 7.1	29.6 ± 5.2	.415
Self-selected walking speed (m/s)	0.7 ± 0.2	0.7 ± 0.2	.538
Fast walking speed (m/s)	1.0 ± 0.3	1.0 ± 0.4	.374
6-Minute Walk Test (m)	299.2 ± 113.0	283.7 ± 122.2	.455
Average Baseline Steps/Day	3954 ± 1994	3479 ± 1686	.149
Activities-specific Balance Confidence Scale	76.3 ± 15.8	71.4 ±19.5	.119
Charlson Comorbidity Index	3.3 ± 1.8	3.3 ± 2.0	.985
Categorical
Sex	Male: n = 34 (50%) Female: n = 34 (50%)	Male: n = 35 (57%) Female: n = 26 (43%)	.402

Continuous data presented as mean ± SD, unless otherwise indicated; Categorical data presented as n (% sample)

Table 2:

Exercise Dose Parameters

Exercise Dose Parameter	FAST (n = 68)	FAST + SAM (n = 61)	P
Number of Sessions	26.1 ± 7.9 (21 – 36)	29.2 ± 4.4 (19 – 36)	0.624
Volume of Walking (minutes)	818.4 ± 124.9 (427 – 1050)	828.2 ± 136.7 (431 – 1050)	0.674
Average Heart Rate Reserve (%)	64.3 ± 12.1 (45.9 – 123.3)	62.7 ± 12.5 (19.6 – 120.7)	0.449
Average Training Speed (% of SSWS)	104.7 ± 26.1 (48.9 – 169.7)	109.7 ± 37.5 (36.3 – 260.6)	0.375

Data presented as mean ± SD (range).

If a participant walked for all 30 minutes at each of a maximum of 36 sessions, the highest possible volume of walking was 1080 minutes.

Outcome – Change in FWS

The model was significant (R² = 0.195, F(9,118) = 3.167, p = 0.002). After accounting for covariates, the average training speed was a significant predictor (b = 0.001 (95% CI: 0.001 – 0.002), p = 0.003) of a change in FWS, while average percent HRR was not a significant predictor (b = −0.001 (95% CI: −0.003 – 0.002), p = 0.571) of a change in FWS (Table 3a).

Table 3a:

Predictors of Change in FWS

Parameter	B	Std. Error	t	Sig.	95% CI
Intercept	0.140
Sex	−0.004	0.031	−0.139	0.890	−0.067 – 0.058
Age	0.000	0.001	−0.128	0.899	−0.002 – 0.002
TSS	0.000	0.000	−1.049	0.296	−0.001 – 0.000
Baseline 6MWT	0.000	0.000	2.784	0.006	0.000 – 0.001
ABC	0.001	0.001	0.908	0.366	−0.001 – 0.002
CCI	−0.006	0.007	−0.922	0.359	−0.019 – 0.007
Exercise Volume	0.000	0.000	−1.694	0.093	0.000 – 0.000
Heart Rate Reserve	−0.001	0.001	−0.569	0.571	−0.003 – 0.002
Training Speed	0.001	0.000	3.072	0.003	0.001 – 0.002

Robust Errors were used. Significance set at p < 0.05.

Outcome – Change in 6MWT

The model was significant (R² = 0.135, F(9,117) = 2.037, p = 0.041). After accounting for covariates, the average training speed was a significant predictor (b = 0.359 (95% CI: 0.108 – 0.610), p = 0.005) of a change in the 6MWT, while average percent HRR was not a significant predictor (b = −0.369 (95% CI: −1.187 – 0.449), p = 0.373) of a change in the 6MWT (Table 3b).

Table 3b:

Predictors of Change in 6MWT

Parameter	B	Std. Error	t	Sig.	95% CI
Intercept	33.872
Sex	2.821	8.834	0.319	0.750	−14.675 – 20.317
Age	−0.202	0.300	−0.675	0.501	−0.796 – 0.391
TSS	−0.046	0.051	−0.906	0.367	−0.146 – 0.054
Baseline 6MWT	−0.075	0.040	−1.857	0.066	−0.155 – 0.005
ABC	0.233	0.311	0.749	0.455	−0.382 – 0.848
CCI	−0.886	2.513	−0.353	0.725	−5.864 – 4.092
Exercise Volume	0.017	0.030	0.567	0.572	−0.042 – 0.076
Heart Rate Reserve	−0.369	0.413	−0.894	0.373	−1.187 – 0.449
Training Speed	0.359	0.127	2.832	0.005	0.108 – 0.610

Robust Errors were used. Significance set at p < 0.05.

Unadjusted relationships for these two outcomes (change in FWS, 6MWT) were also explored and there were no differences in the direction or significance of the results.

DISCUSSION

This study sought to understand if training intensity (quantified as average percent HRR or training speed) is predictive of changes in walking capacity outcomes after accounting for multiple known covariates following a moderate-to-high intensity intervention in people with chronic stroke. After accounting for covariates, training speed was a significant predictor of changes in both walking endurance (6MWT) and FWS. Conversely, training heart rate was not a significant predictor of change in either walking capacity outcome. With the largest known sample of individuals with chronic stroke to complete a fast-walking training intervention (n = 129), this analysis had the statistical power to account for covariates which may influence walking capacity outcomes. These results provide evidence of training speed as a predictor of changes in walking capacity outcomes after accounting for additional influential factors, such as age, baseline walking endurance, balance self-efficacy, and training heart rate reserve.^12,15,43 It was surprising that average training heart rate was not a predictor of walking capacity outcomes after accounting for known covariates.

These results build on prior evidence which has demonstrated training speed and/or step count (a proxy of exercise repetition in walking interventions) have a stronger association with walking capacity outcomes than training heart rate.^8,17,24 The HIT-Stroke randomized clinical trial found training speed and training step count were significant mediators of the change in 6MWT, regardless of if participants received a high-intensity interval walking intervention or moderate-intensity continuous walking intervention.^17,24 Similar to this current analyses, the prior analysis also found heart rate was not a significant mediator of 6MWT outcomes. As the present analysis incorporates a much larger (n = 129 vs. 55) sample, and accounts for various demographic variables, clinical factors, and exercise volume, it further emphasizes that training speed may be a more valuable metric of exercise intensity to improve walking capacity in people with chronic stroke. Step counts, which are correlated to walking capacity outcomes in people with chronic stroke,⁸ were not recorded during training sessions in this parent clinical trial.^9,26 However, as training time remained constant in this protocol, an increase in walking speed in the intervention would suggest a concurrent increase in training step counts.

Previous evidence indicates walking capacity may be most strongly influenced by neuromuscular deficits in people with chronic stroke.^36,37,44 If the neuromuscular system is the “weakest link”, walking interventions could target these deficits by focusing on challenging walking speed in training.^45–47 Our results support this theory by suggesting a slight change in training speed, from self-selected to fastest walking speed, may lead to greater improvements in walking capacity outcomes.^45–47 Furthermore, the strength of training speed as the metric of exercise intensity stands even while accounting for various other covariates known to impact walking capacity in people with chronic stroke. This is especially important as previous work which found associations between exercise intensity and walking outcomes, did not account for other variables.^8,12–21 Therefore, the previously found association between heart rate intensity and walking capacity outcomes may have been due to other factors.

Given that walking speed was found to be a significant predictor of change in walking capacity, it is important to put into a clinical perspective the potential impact of training at various speeds. Using the regression coefficients from the results, we can calculate the gains in 6MWT distance and FWS clinicians could expect if their patient walked on average at their self-selected walking speed or their fastest walking speed. If participants walked on average at their SSWS, there was a 41.3m increase in 6MWT and .157m/s increase in FWS following the intervention. If participants walked on average at their FWS (~133% of SSWS), there was a 58.0m increase in 6MWT and a 0.203m/s increase in FWS. On average, training at FWS led to greater changes in 6MWT distance (+16.7m) and FWS (+0.046m/s) compared to training at SSWS. This suggests challenging people with chronic stroke to walk at the upper limit of their walking speed during training may further improve walking outcomes. Future studies which directly compare the changes in walking capacity outcomes after participants complete a walking intervention at different training speeds are needed to confirm these results.

Surprisingly, training heart rate was not a significant predictor of walking capacity outcomes. The average training heart rate (64% HRR) exceeded the ACSM threshold of “vigorous intensity” exercise (> 60% HRR) and the recommended intensity of aerobic exercise for people with chronic stroke.^7,39,48,49 There is likely a relationship between training speed and training heart rate which may partially explain why previous walking exercise interventions which aimed for moderate-to-high heart rate intensities demonstrated significant changes in walking capacity.⁷ In this study sample, participant’s average training heart rate and training speed were not correlated (r = −0.029, p = 0.741). Recent work has questioned the use of heart rate to guide exercise intensity, noting it may not be specific enough to elicit desired outcomes.^25,50 This is because heart rate is an indirect metric of intensity, and can easily be impacted by factors unrelated to exercise.¹⁰ For example, anxiety, poor sleep, or caffeine intake could influence an individual’s training heart rate. Despite these influences on heart rate being completely agnostic to the exercise intervention, they would influence this as a metric being used to quantify training intensity. It is possible that heart rate is too indirect of a measurement of intensity, particularly in people with chronic stroke who may have concurrent cardiovascular comorbidities, use beta-blocker medications, or neuromuscular deficits which impact their capacity to walk. If in the context of walking the cardiovascular system is not the most limiting system, then heart rate may not be the optimal target of exercise intensity. As most people with stroke cite a desire to improve their walking, determining the most specific parameters of walking exercise dose to achieve this goal is paramount.⁵¹

Study Limitations

This walking exercise protocol was designed as a continuous walking model, so it cannot directly be compared to walking protocols which used intermittent, high-intensity (dictated by either speed or heart rate) intervals. As this protocol was directed by heart rate intensities, it is unknown if participants would be able to sustain higher training walking speeds (107% of SSWS) for the same training duration. Future work is needed to decipher the strength and relationship of each exercise dose parameter to optimize walking capacity outcomes.

As the cardiopulmonary exercise stress test which determined the maximal heart rate and training heart rate zones was not repeated during the intervention, the training heart rate was not adjusted throughout training. This may have failed to account for potential training effects, and limited heart rate as a metric of exercise intensity. While heart rate calculations did not adjust for beta-blockers, participants were instructed to take beta-blocker medications both at the exercise stress test and for each training session to account for beta-blocker use. Prior work has demonstrated individuals with chronic stroke may not reach a cardiovascular maximum during cardiopulmonary exercise stress tests,⁵² which could have limited the average heart rate intensities achieved by the sample. Of note, the average respiratory exchange ratio (RER) achieved on the cardiopulmonary exercise stress tests was 0.99 ± 0.13 (range 0.70–1.23), indicating a cardiovascular maximum may not have been met.

Lastly, while improvements in self-selected walking speed occurred from the baseline to post-intervention timepoint, it is unknown when these improvements occurred within the intervention. Therefore, in this secondary analysis, we were unable to adjust the training speed to reflect potential changes in walking speed that were occurring throughout the intervention. Future studies could assess walking speed throughout the intervention to probe when these changes are occurring. Probing when changes in the neuromuscular and cardiovascular system are occurring within a walking intervention may elucidate which (or when) each system is the primary driver of walking capacity changes. Lastly, while these results, along with recent results from an independent cohort of people with chronic stroke, suggest training speed may be a stronger measure of exercise intensity, this may not be true for all within the chronic stroke population.²⁴

CONCLUSIONS

Walking exercise interventions, primarily conducted at moderate-to-high heart rate intensities, have demonstrated significant increases in walking capacity in people with chronic stroke. Results from this large cohort of people with chronic stroke who underwent a walking exercise intervention indicate training speed, but not training heart rate, significantly predicts changes in walking outcomes, after accounting for other important factors. Using training speed to guide exercise intensity, and challenging people with stroke to train at their fastest walking speed, may increase walking capacity gains in rehabilitation.

LIST OF NON-STANDARD ABBREVIATIONS

10mWT

10-meter walk test

6MWT

Six-Minute Walk Test

ABC

Activities-specific Balance Confidence scale

ACSM

American College of Sports Medicine

CCI

Charlson Co-morbidity Index

FAST

high-intensity treadmill training

FWS

fastest walking speed

HRR

Heart Rate Reserve

MHR

maximum heart rate

PROWALKS

Promoting Recovery Optimization of Walking Activity in Stroke

RHR

resting heart rate

SAM

step-activity behavioral intervention

SSWS

self-selected walking speed

TSS

time since stroke