Effects of long-term, home-based, virtual cycling after COPD exacerbation – results from the randomised controlled COPDtoParis trial

ABSTRACT

Background

Acute exacerbations of chronic obstructive pulmonary disease (AECOPD) are associated with functional decline, reduced quality of life, and high rates of readmission and mortality. Although exercise is recommended following AECOPD, long-term adherence and maintenance of benefits remain challenging. Home-based, technology-supported exercise modalities may offer accessible alternatives, but evidence on long-term effectiveness remains limited.

Objective

To evaluate the effectiveness of a long-term, home-based, virtual cycling intervention on physical performance, physical activity, lung function, functional mobility, frailty, symptoms, health-related quality of life (HRQoL), and hospital readmission in people with COPD following AECOPD.

Methods

This single-centre, parallel-group, cluster randomised controlled trial recruited patients hospitalised with AECOPD. Participants were randomised in teams to 12 months of home-based virtual cycling or standard care. The intervention consisted of team-based cycling on pedal trainers guided by pre-recorded route videos. The primary outcome was the five-repetition sit-to-stand test (5RSTS). Secondary outcomes included physical performance, lung function, physical activity measured by leg-mounted triaxial accelerometers, dyspnoea, frailty, functional mobility, and HRQoL. Outcomes were assessed at baseline, 6 weeks, 6 months, and 12 months using linear mixed models.

Results

Forty participants were randomised (21 intervention, 19 control). Physical and respiratory outcomes remained largely stable over time with no significant between-group differences. At 12 months, the intervention group reported higher self-rated health and functional mobility compared with controls in secondary outcomes. Sedentary time remained high in both groups, and readmission and mortality rates were similar.

Conclusion

In older, multimorbid individuals recovering from AECOPD, long-term home-based virtual cycling did not improve physical performance, activity levels, or lung function, but was associated with higher self-perceived health and functional mobility at 12 months. However, given attrition and limited power, these findings should be cautiously interpreted. Low-threshold, technology-supported exercise may offer patient-centred benefits in populations with limited exercise reserve but research with sufficient power is needed.

Patient and public involvement

Persons living with COPD were extensively involved in the development of the intervention through qualitative interviews with test pilots, participants, and non-participants, informing bicycle selection, app functionality, video content, guidance needs, and practical implementation. This formative work directly shaped the design, delivery, and tailoring of the virtual home-based cycling intervention. The development process and patient involvement are described in detail in a previously published qualitative study. Patients were not involved in data analysis, interpretation of results, or manuscript writing for the present randomised controlled trial [1].

Introduction

Chronic Obstructive Pulmonary Disease (COPD) is the most prevalent respiratory disease globally and is associated with increased risk of early death, frequent hospitalisations, and increasing expenses for society and patients [2]. Expenses relate especially to acute exacerbation of COPD (AECOPD) events that can lead to decreased physical performance, hospitalisation, need for more health care contacts, medicine, rehabilitation and home care [3]. Furthermore, readmissions after an AECOPD are common, with approximately 17% of patients readmitted within 60 days [4].

Pulmonary rehabilitation (PR) following an exacerbation is widely recognised as an important intervention to counteract the effects that AECOPD can entail [5,6]. Participating in PR including exercise is an important element in the management of COPD, aiming at improving quality of life, symptoms such as dyspnoea, fatigue, cough and sputum, as well as reducing anxiety, depression, exacerbations, readmissions, and early death [7]. Furthermore, PR can increase physical activity, which is important for maintaining the effects of the rehabilitation [8]. The long-term maintenance of benefits derived from attending PR is crucial for sustaining improvements in physical and psychological health [9]. However, the challenge lies in ensuring ongoing engagement and adherence to health-enhancing behaviours post-PR [9,10]. The more frail and disabled patients [11], who refrain from leaving their home, are especially challenged by engaging in and adhering to physical activity [12].

Persons with more advanced COPD, who experience exacerbations, report several obstacles to engaging in regular exercise [13], e.g. debilitating symptoms, pain and discomfort, accessibility of exercise facilities, comorbidities, weather conditions, and costs [12,14]. Furthermore, lack of social support significantly affects adherence to exercise for persons living with COPD (plwCOPD) [15]. These barriers highlight the need for structure and accessible exercise formats, supporting continuity after PR.

The implementation of PR programmes has evolved to include various modalities, such as home-based exercise and telehealth approaches provide alternatives to traditional centre-based rehabilitation [16]. Although these approaches have demonstrated feasibility and acceptability, their effectiveness in maintaining long-term improvement is still an area of ongoing research [17–19]. The maintenance of benefits derived from PR in plwCOPD remains uncertain, particularly for persons recovering from AECOPD [20]. However, home-based exercise is a viable alternative exercise format to improve exercise tolerance, physical capability, HRQoL, disabilities, and improving COPD symptoms, but long-term adherence to participating in long-term, home-based exercise is sparsely studied as of yet [19].

Physical activity (PA) is increasingly recognised as an important outcome in recovery from AECOPD [21–23]. However, most existing studies have relied on self-report or wrist-worn step count, which struggle to capture the irregular gait patterns of plwCOPD precisely [24]. Recent work has demonstrated that leg-mounted triaxial accelerometers provide a more valid and reliable measure of PA in this population [25].

Cycling has been described as an effective part of PR, improving lower limb muscle function, exercise capacity and health related quality of life (HRQoL) for plwCOPD, including those recovering from AECOPD [26,27]. The cycle allows for controlled and adjustable exercise intensity and can be tailored to the individuals’ capabilities, making it a suitable option for those with varying levels of disability and disease severity. The use of cycle ergometers in home-based exercise has shown promising results, allowing patients to engage in easily accessible exercise without the barriers from centre-based exercise [28].

The aim of this study was to evaluate the effectiveness of a long-term, home-based virtual cycling intervention on physical performance, daily activity levels, lung function, mobility, frailty, symptom severity, health-related quality of life, and hospital readmission in plwCOPD following an AECOPD.

Methods

Study design

This study was designed as a single-centred, parallel-group cluster randomised controlled trial (RCT) conducted at Aalborg University Hospital in Denmark, as described in the published study protocol [29]. Patients diagnosed with AECOPD were recruited during hospitalisation. Following discharge, all participants underwent municipality-based acute rehabilitation, as mandated by Danish legislation, typically lasting around two times a week for 8 weeks. This acute rehabilitation phase, including detailed descriptions of physical activity patterns and clinical outcomes during the early post-exacerbation period, has been reported separately as part of the research project [30]. After completing this period, participants were group-randomised to either the intervention or the control group, allowing team-formations for simultaneous cycling in the virtual platform. Accordingly, the randomised population represents patients able to complete early rehabilitation after AECOPD and are willing to participate in exercise, rather than the full population of patients hospitalised with AECOPD.

The intervention group commenced 1 year of home-based, virtual cycling and the control group received standard care. Data were collected during home visits at baseline, after 6 weeks, 6 months, and after 12 months of the home-based, virtual cycling. The sample size followed the a priori specification described in the published protocol [29]. It was based on the five-repetition sit-to-stand test (5RSTS) as the primary outcome. Using the minimally clinically important difference (MCID) of 1.7 repetitions [31] and an assumed standard deviation of 2 repetitions, the standardised effect size was 0.85. With a two-sided alpha of 0.05, power of at least 80%, and an allocation rate of 1:1, this required 40 participants. Allowing for an expected dropout rate of 20%, the target sample size was 50 participants.

The study recruited patients from October 2023 until August 2024 after more than 50 participants consented to take part, thus reaching the target of 50 participants to allow for attrition. A substantial number, however, were unable to complete the preceding acute rehabilitation. Since recruitment had reached the planned date of inclusion period, and beyond, 40 participants were randomised, as shown in Figure 2. The final sample size was therefore below the planned target, and the trial should be considered exploratory and potentially underpowered for between-group comparisons. The study finalised in August 2025.

Figure 2.

Flowchart of participant screening, eligibility assessment, exclusion, and randomisation in the trial. Of 364 individuals screened, 197 were eligible and 40 were ultimately randomised. Reasons for ineligibility, exclusion, and loss prior to randomisation are shown.

Intervention

Participants in the intervention group engaged in home-based cycling in the teams they were clustered with in the randomisation process. On exercise bikes connected to a tablet, participants pushed forward cycling videos from Aalborg, Denmark to Paris, France in the 4MVideo app (4MVideo Aps, Kongens Lyngby, Denmark). The intervention is visualised in Figure 1. During cycling, participants were able to communicate with other team members in their group through a chat function in the app to recreate a team experience. Furthermore, they could track their progress, their own as well as team members’, on a virtual map with avatars showing their placement on the route of the day. The recordings of the route were captured by a local team, Team Rynkeby, who filmed with a camera mounted on their chest. The recordings were divided into 42 stages, one for each day of the 6 weeks, every video approximately 20 min long. The app and the cycle were adjusted by physiotherapists to fit individual exercise, enabling participants at different physical capacities to cycle at the same stages together. The participants were encouraged to cycle every day at a suitable timepoint. They were contacted semi-weekly by telephone by physiotherapists to assess whether they needed assistance. The progress of each cyclist was visible to the physiotherapists associated with the study. After the trip was completed, other cycling routes were made available for the participants of the intervention group to partake in the following 10.5 months. The cycle and tablet were available free of charge for a total of 12 months.

Figure 1.

Illustration of the virtual, home-based cycling intervention. Participants used a compact pedal-trainer while seated and followed pre-recorded cycling videos on a tablet to guide pace and maintain engagement throughout the home-based exercise sessions.

A pilot study conducted prior to trial initiation informed on the design of the virtual cycling, helped deliver a setup that was safe, feasible, and accessible for the participants. On the basis hereof, cycling was selected as the home-based exercise modality [1].

Control

Participants in the control group received standard care and an individual solution which included either continuing in exercise offered by the municipality or receiving a standardised home-based exercise programme aimed at plwCOPD, to their liking and the physiotherapeutic assessment.

Recruitment

Recruitment was carried out by trained physiotherapists, validated by doctors when there was doubt of the patients’ eligibility. Newly admitted patients with a diagnosis of AECOPD were identified daily by screening patient lists at the Emergency Department and the Department of Lung Diseases. Patients who met the inclusion criteria and none of the exclusion criteria were approached. Potential participants were provided with oral and written information and were allowed at least 24 h before consent was obtained. Participants could at all times withdraw consent and were discontinued in the event of medical or cognitive conditions that inhibited the participant from safely exercising. After consenting, participants were referred to acute rehabilitation, which commenced in discharge from the hospital.

Randomisation

Cluster randomisation was used for logistical reasons, since the intervention was delivered to existing rehabilitation groups, and all participants within a group had to receive the same allocation. Groups were randomised as units at the time they entered the trial. An automatic allocation procedure, as described in the protocol [29], adjusted the allocation probability of the next cluster based on the previous allocations to ensure that the two arms remained closely balanced throughout the process. The randomisation sequence was computer generated, concealed from investigators, and implemented centrally.

Clinical assistants were responsible for the installation of exercise equipment, mounting of activity measurement devices, and administration of questionnaires and physical testing, thus ensuring the blinding of the research team. The statistician analysing the data was blinded to the group assignments during the statistical analysis phase. This was ensured by coding the data in such a way that the group labels were not disclosed. The participants were not blinded to group allocation.

Eligibility criteria

In- and exclusion criteria have been described in the published protocol for the RCT [29] and included adults (>18 years old), who had been admitted to Aalborg University Hospital, Denmark, with an AECOPD with pre-existing COPD diagnosis. Participants had to be able to provide informed consent and agree to be referred to the acute rehabilitation programme offered by Aalborg Municipality [29].

Patients were not eligible if they had a terminal illness or presented with unstable cardiac disease, including ischaemic heart disease. They were also excluded for other congenital, medical, or physical conditions that prevented from engaging safely in home-based, virtual cycling. This could be lower-limb amputation, blindness, or restrictions related to previous or upcoming surgery that made a seated cycling position inadvisable. Finally, patients who were unable to understand basic written or spoken Danish were excluded.

Outcomes

For all participants, relevant background information was extracted from the digital patient record. This included sex, age, smoking status together with tobacco exposure history, and current occupational status. Body Mass Index (BMI) was recorded, along with documented comorbidities and hospital contacts. Mortality data were also retrieved from the patient record. Comorbidities were reported as the total number of conditions per participant. A complete list of comorbidities and their distribution between groups is provided in the supplementary material.

Primary outcomes

The primary outcome of this RCT was the five-repetition sit-to-stand test (5RSTS), which is a validated measure of physical performance, functional lower limb strength, and endurance in plwCOPD [31]. Participants unable to perform the 5-repetition sit-to-stand test were assigned a score of 0.

Secondary outcomes

Secondary outcomes covered physiological, functional, patient-perceived, and behavioural domains. Lung function was assessed as forced expiratory volume in the first second (FEV1), using a mobile, hand-held spirometer (CareFusion, Jaeger, San Diego, USA). Functional lower-limb performance was evaluated with the 30 Second Chair Stand Test (30s CST), and balance and lower-limb performance was evaluated with the Short Physical Performance Battery (SPPB). Dyspnoea impact was assessed using the modified Medical Research Council Dyspnea Scale (mMRC).

Disease specific health status was attained using the St. George’s Respiratory Questionnaire (SGRQ), and the COPD Assessment Test (CAT), while generic HRQoL was measured with the EuroQol 5-Dimensions 5-Level (EQ-5D-5 L) questionnaire. Frailty was assessed using the Tilburg Frailty Indicator (TFI) and the Clinical Frailty Scale (CFS). Functional mobility was measured with the Life-Space Assessment Test (LSA-DK) and the Clinical Frailty Scale (CFS). Functional mobility was measured with the Life-Space Assessment Test (LSA-DK). All instruments were used for assessment of both the intervention group and controls at baseline, 6 weeks, 6 months and 12 months, except Daily Activity Levels which were assessed at baseline and 12 months. All questionnaires were available in validated Danish versions for the use in plwCOPD or populations with chronic illness.

Activity levels were measured with triaxial accelerometers (SENS Innovation ApS, Copenhagen, Denmark). Sensors were mounted on lateral side of participants’ knees and recorded activity for 1 week before randomisation and 1 week after the intervention. Changes in activity are reported as changes in sedentary hours per day. All measurements are listed in Table 1.

Table 1.

Overview of measurements per home visit.

	Baseline	6 weeks	6 months	12 months
Daily Activity Levels	X			X
Lung function (Forced Expiratory Volume (FEV1))	X	X	X	X
Six Minutes Walking Test (6MWT)	X	X	X	X
30 seconds sit-to-stand test	X	X	X	X
Short Physical Performance Battery (SPPB)	X	X	X	X
Modified Medical Research Council Dyspnea Scale (mMRC)	X	X	X	X
St. George’s Respiratory Questionnaire (SGRQ)	X	X	X	X
COPD Assessment Test (CAT)	X	X	X	X
EuroQol 5-Dimensions 5-Level (EQ-5D-5 L)	X	X	X	X
Tilburg Frailty Indicator (TFI)	X	X	X	X
Clinical Frailty Scale (CFS)	X	X	X	X
Life-Space Assessment (LSA-DK)	X	X	X	X

Data management and statistical analysis

Data were stored electronically on a secured server in accordance with the General Data Protection Regulation (GDPR). Each participant was allotted a unique identification number.

Baseline characteristics were described through mean and SD for continuous variables (skew variables were supplemented with median and IQR) and percentages for categorical variables. Between-group differences at baseline were compared with unpaired t-tests and Fisher’s exact tests for continuous and categorical variables, respectively.

For data analysis STATA 18 (StataCorp. 2025. Stata Statistical Software: Release 18. College Station, TX: StataCorp LLC) was used. Primary analyses were conducted on available data at each time point using linear mixed models with patient as random intercept. Given the high mortality and withdrawal during follow-up, these analyses rely on the assumption that missing outcome data did not materially distort the observed between-group comparisons. However, this should be considered when interpreting them. Between-group comparisons at 6 weeks, 6 months, and 12 months were estimated for each outcome, and results are presented as summary outcome measures with 95% confidence intervals and two-sided p-values. Random intercept for clusters was not included due to the small dataset was likely not to allow separation between the between-patient (nested in clusters) variation and the between-cluster variation. Given the limited sample size, emphasis was placed on estimated between-group differences and 95% confidence intervals when interpreting the results. Analyses were performed at individual participant level. A random intercept for clusters was not included since the total sample and numbers of clusters were too small.

Profile plots depicting the development over time were made for selected outcomes.

Two sensitivity analyses were conducted. First, a baseline-restricted analysis addressing the statistically significant baseline imbalance was conducted by restricting the baseline values between groups to be equal. Second, an analysis adjusting for possible informative dropout due to death was used to assess the robustness on inference. This analysis was performed in a joint model with generalised structural equation modeling combining the linear mixed model and a parametric Weibull survival model of time to death. These sensitivity analyses are reported in the supplementary material.

Readmissions and mortality were analysed as exploratory safety outcomes. Rates per person-year were estimated with Poisson regression using a log link and an offset for log person-time and are reported as rate ratios with 95% confidence intervals. The comparisons of readmissions rates were not adjusted for the competing risk of death.

A priori, a two-sided level of significance was set at p < 0.05.

Ethics

This study and participants information materials, recruitment documents, and participant consent form were approved by a regional committee on health research ethics and in compliance with the Helsinki Declaration. Safety and progress reports reporting potential adverse events were made annually. The study was registered at a public clinical trial registry [details removed for peer review]. The study is reported in accordance with the Consolidated Standards of Reporting Trials (CONSORT) guidelines.

Results

Demographics

A total of 40 individuals were enrolled in the analysis, with 21 randomly allocated to the intervention group and 19 to the control group. Figure 2 provides an overview of screening, inclusion, and exclusion. Out of the eligible population, 47 entered municipal rehabilitation and 7 were lost in that period, before randomisation, including 3 who withdrew and 4 who died. Ten clusters were randomised. Cluster size was of three to five participants depending on the number of participants available for team allocation at the time of randomisation.

The baseline characteristics of the included participants who were able and consented to complete early rehabilitation after AECOPD are reported in Table 2. The two groups were predominantly older adults with a long history of smoking, multi morbidity, and moderate to severe airflow limitation at baseline distribution. Frailty scores indicated that most participants were moderately frail. The groups were comparable with respect to sex, age, BMI, smoking, comorbidities, lung function, and frailty status, except for sedentary behaviour. Participants in the intervention group spent significantly more time sedentary per day compared with the control group. To account for the baseline imbalance, additional sensitivity analyses were performed in which the models were adjusted for sedentary time at baseline.

Table 2.

Baseline characteristics of participants in the intervention and control groups.

		Intervention, n = 21	Control, n = 19	p-value	Missing, n
Sex	Female	15	13	0.84
	Male	6	6
Age, mean (SD)		74.3 (8)	75.1 (8.7)	0.79
Height, mean (SD)		165.1 (9.3)	163.2 (11.73)	0.56
Weight, mean (SD)		70.1 (17)	66.2 (19.1)	0.50
BMI, mean (SD)		25.8 (6.7)	25.0 (7.5)	0.70
Smoking status, category	Former	15	14	0.51
	Current	6	4
	Never	0	1
Pack years, mean (SD)		44.9 (13.21)	42.0 (10.65	0.51	7
Comorbidities, mean (SD)		3.9 (1.7)	3.34(2.79)	0.46
Lung function FEV1 pred, mean (SD)		44.1 (10.9)	45.6 (12.1)	0.69	3
COPD assessment scale, mean (SD)		11.3 (5.1)	13.7 (8.1)	0.29	5
Clinical Frailty Scale, mean (SD)		5.2 (1.1)	5.0 (1.5)	0.62	4
Tilburg Frailty Indicator, mean (SD)		6.2 (3.4)	5.6 (2.77)	0.52	4
Sedentary activity, hours per day, mean (SD)		21.59 (2.52)	19.16 (3.1)	0.01	1

Abbreviations: Standard Deviation (SD), Body Mass Index (BMI), Forced Expiratory Volume in the first second, percent predicted (FEV1, pred.), Chronic Obstructive Pulmonary Disease (COPD).

Clinical outcomes

The longitudinal development of clinical outcomes from baseline to 12 months is presented in Table 3. Outcomes at 6 weeks and 6 months are provided in the supplementary material (Table A2, Appendix). Across assessments, the performance-based outcomes 5RST, 30s STS, and SPPB showed similar trajectories in the two groups with no consistent between group differences at any point. Lung function, measured FEV1, % predicted and COPD impact measured with CAT remained stable in both groups with no between-group differences. Disease impact measured by SGRQ decreased clinically relevant for both groups [32] with no statistically significant difference between groups. Sedentary time remained high and with minimal change.

Table 3.

Outcomes at baseline and 12 months.

Outcome	Group (n)	Baseline [95% CI]	12 months [95% CI]	Mean differences between groups	p-value
5 RSTS (seconds)	Intervention (18)	1.6 [0.8;2.5]	1.8 [0.9;2.7]	−0.4 [−19.2;11.2]	0.5
	Control (19)	2.1 [1.4;2.8]	2.2 [1.4;3.0]
30s STS (repetitions)	Intervention (19)	10.5 [0.1;20.9]	11.7 [−0.1;23.4]	−3.5 [−18.2;11.2]	0.6
	Control (19)	13.2 [−0.1;26.6]	15.2 [−0.2;30.6]
Sedentary time (hours)	Intervention (19)	21.6 [20.5;22.7]	21.5 [19.5;23.4]	2.3 [−0.2;4.8]	0.08
	Control (18)	19.2 [17.9;20.4]	19.2 [17.6;20.7]
Lung function, FEV1% pred.	Intervention (19)	44.1 [38.7;49.6]	44.7 [38.5;50.9]	0.2 [−8.3;8.6]	0.9
	Control (19)	46.3 [40.8;51.8]	44.6 [38.8;50.4]
SPPB, points	Intervention (19)	5.1 [3.7;6.5]	5.3 [3.5;7.1]	−1.3 [−3.8;1.1]	0.3
	Control (19)	6.4 [4.9;7.8]	6.6 [5.0;8.3]
SGRQ, points	Intervention (18)	54.9 [47.5;62.3]	49.2 [40.8;57.6]	−3.2 [−14.6;8.2]	0.6
	Control (19)	57.4 [50.1;64.8]	52.4 [44.6;60.1]
CAT, points	Intervention (18)	11.3 [8.0;14.6]	10.6 [6.7;14.4]	−4.7 [−9.9;0.5]	0.08
	Control (19)	13.6 [10.4;16.9]	15.2 [11.7;18.7]
EQ-5D-5 L, index score	Intervention (18)	0.8 [0.6;0.9]	0.6 [0.5;0.8]	−0.1 [−0.3;0.2]	0.6
	Control (19)	0.7 [0.6;0.8]	0.7 [0.5;0.8]
EQ-VAS, 0–100	Intervention (18)	61.9 [51.8;72.1]	69.3 [57.0;81.7]	23.3 [6.7;39.9]	0.006
	Control (18)	58.3 [48.2;68.5]	46.1 [35.0;57.1]
LSA-DK, points	Intervention (18)	58.5 [45.4;71.6]	69.7 [54.9;84.6]	26.9 [6.6;47.2]	0.009
	Control (19)	56.0 [43.3;68.8]	42.8 [29.0;56.6]

Abbreviations: Confidence interval (CI), 5RSTS, five-repetition sit-to-stand test; 30s STS, 30-second sit-to-stand test; FEV1% pred., forced expiratory volume in 1 second (% predicted); SPPB, Short Physical Performance Battery; SGRQ, St George’s Respiratory Questionnaire; CAT, COPD Assessment Test; EQ-5D-5 L, EuroQol 5-Dimensions 5-Levels; EQ-VAS, EuroQol Visual Analogue Scale; LSA-DK, Life-Space Assessment, Danish version.

HRQoL measured with EQ-5D-5 L, followed a similar trajectory between groups until at 12 months, where the intervention group reported higher self-reported health compared with the control group, with a between-group difference of 23.3 points (95% CI 6.7 to 39.9) on the EQ-VAS (Figure 3). Furthermore, the intervention group experienced a higher functional mobility measured with the LSA-DK compared with the control group at 12 months, with a between-group difference of 26.9 points (95% CI 6.6 to 47.2) (Figure 4).

Figure 3.

Changes in EQ-VAS scores from baseline to 12 months for the intervention and control groups. Mean EQ-VAS scores with 95% confidence intervals are shown for both groups at baseline, 6 weeks, 6 months, and 12 months. Higher scores indicate better self-reported health status. While trajectories were similar during early follow-up, the intervention group demonstrated a higher EQ-VAS score at 12 months compared with the control group.

Figure 4.

Changes in life-Space Assessment (LSA-DK) scores from baseline to 6 weeks, 6 months, and 12 months for the intervention and control groups. Values represent means with 95% confidence intervals.

Ten participants died between baseline and 12 months, six in the intervention group and four in the control group. Eight participants withdrew their consent during the study, five in the intervention group and three in the control group. Sensitivity analyses did not alter the overall pattern of the results and are presented in the supplementary material (Table A1, Appendix). Readmission and mortality rates per person-year were similar between groups. These analyses supported the overall direction of findings, but the extent of death and withdrawal means that the 12-month estimates were vulnerable to imprecision and possible survivor bias.

Discussion

Key findings

This pragmatic randomised trial aimed to examine the effects of a long-term, home-based, virtual cycling intervention for plwCOPD after an AECOPD. There were no significant changes between groups in the primary outcome. Across the follow-up period, most clinical outcomes stayed stable in the two groups, except for higher self-perceived health on the EQ-VAS and higher functional mobility found in the intervention group at 12 months. Findings remained in sensitivity analyses that corrected for baseline imbalance in sedentary time and further adjusted for deaths and dropouts. This supports the trajectories of the overall pattern but does not eliminate uncertainty related to missing data in this small, selected and frail cohort.

The observed between-group differences in EQ-VAS and LSA-DK exceeded the MCID, suggesting that the results represent a meaningful patient-perceived improvement [33,34] and may indicate clinical relevance. These results, however, were only evident at 12 months, indicating that the effect on perceived health and functional mobility may require sustained exposure to the intervention before becoming measurable. Furthermore, these outcomes were secondary, the study was underpowered, and no adjustment for multiplicity was applied. Clinical significance should therefore be interpreted cautiously considering the statistical uncertainty and timing of these findings.

This pattern is partly consistent with previous work. The limited change across most objective measures is likely influenced by the cohort’s advanced disease severity, age, frailty, and high mortality, which together reduce exercise reserve and limit the capacity for measurable gains. Rehabilitation is well-known to improve HRQoL and physical capacity, but non-responders are commonly older, comorbid, and frail individuals with COPD [35–37]. Our findings may reflect the limited exercise reserve in the population rather than a lack of efficacy of exercise. By showing that long-term, home-based, virtual cycling may influence patient-reported outcomes even when objective performance measures remain stable, this study adds evidence on the potential of low-threshold, home-based modalities for maintaining or perceived health after an AECOPD. The interpretation of these findings should take the substantial attrition and mortality over the 12-month period into account. Participants who remained in the study may represent a subset with greater stability or capacity, therefore presenting a potential survivor bias that could influence the differences in observed changes in EQ-VAS and LSA-DK. In the context of this study, the findings may partly reflect selective retention of participants who were less affected generally by their condition and thus better able to participate, rather than a clear intervention effect. This effect should therefore be considered when interpreting the statistically significant changes at 12 months. No significant between-group differences were found in other objective physical measurements, which contrasts typical short-term benefits associated with PR [38]. It is, however, consistent with emerging studies of home-based or digitally supported rehabilitation [17]. Several factors might explain this. Compared with previous studies, this cohort was older, had multiple comorbidities, high disease impact, and severe airflow limitation. Furthermore, they recently had an AECOPD where previous clinical studies have been in stable-phase patients [17,39]. This specific cohort underwent structured acute rehabilitation before commencing on home-based cycling and showed minimal gains in this context [40]. This points towards limited exercise reserve and reduced responsiveness to exercise in this cohort. Previous studies have demonstrated the difficulty of changing activity patterns [8]. Our findings reinforce this challenge of largely unchanged outcomes despite feasibility and engagement [1]. Feasibility and patient-reported improvements do not counter the limited effects on other outcomes. This highlights the need for more real-world research into rehabilitation approaches tailored to individuals with low functional reserve.

Our findings suggest that prolonged access to easily accessible exercise modalities after an AECOPD may be associated with differences in participants’ perceived health-status and mobility range, more than it shifts objective performance or respiratory measures. These findings align with the studies showing that technology-supported and home-based programmes can sustain engagement and generate patient reported benefits, while effects on physical performance may be modest [41]. In this trial, the late on-set significant differences suggest that changes in perceived health and functional mobility may develop gradually over prolonged engagement, which is consistent with long-term behavioural patterns in older and multimorbid individuals. Such discrepancies between perceived and objective outcomes are consistent with recent movements to personalise rehabilitation [42]. Taken together, this underscores that patient-perceived benefit may depend on alignment between participants’ preferences and the design of the intervention. These aspects of engagement have also been explored from the patient perspective in a qualitative interview study nested in the clinical study [43]. While our results may indicate patient-perceived benefits from long-term, home-based cycling on HRQoL and functional mobility, they should not be taken as definitive evidence of intervention efficacy or lack thereof, given the reduced number who completed the 12-months assessment. Even so, the study contributes by demonstrating that, in a population with high comorbidity and mortality, and a limited exercise reserve, easy-accessible and low-threshold interventions could be relevant for a select population.

Strengths and limitations

A key strength of this study is its pragmatic randomised design, which reflects real-life rehabilitation following an AECOPD. The intervention followed a prespecified protocol, and outcomes were assessed repeatedly over 12 months using validated measures. The use of tri-axial accelerometry provided high-resolution activity data that are well-suited for older individuals with irregular walking patterns.

Several limitations must be considered. First, although the a priori calculation indicated that the design required 50 participants to detect the prespecified between-group difference in the primary outcome, only 40 participants could be randomised. More than 50 individuals consented to participate but were unable to complete the acute rehabilitation phase before randomisation. The study was therefore likely underpowered. The randomised sample further reflects a selected sub-group of patients able to complete early rehabilitation and thus represents persons who, for individual reasons, cannot. This increases the risk of type II error and may have contributed to the absence of significant between-group differences in several outcomes. Null findings may represent bias and imprecision as well as absence of effects from long-term, home-based cycling. This is particularly important for secondary outcomes and for interpretation of statistical and clinical significance. Even so, it limits the generalisability to real-world AECOPD populations.

Dropouts and mortality rates were higher than anticipated. Mortality and withdrawal are expected in a cohort of older, multimorbid, and frail persons but nevertheless reduce precision and increase the risk of type I and type II errors. Likewise, they increase the risk that missing data was informative. This is particularly relevant for the 12-month data where outcomes may be influenced by participants with more stable trajectories. These characteristics also limit the potential responsiveness to exercise-based interventions, and this may have contributed to the largely stable trajectories in objective performance outcomes. Thus, the intervention’s feasibility cannot be considered a substitute for clinical efficacy, and the predominantly neutral findings should be interpreted as they are. Third, despite algorithm-based randomisation, a significant baseline imbalance in sedentary behaviour was observed. As sedentary behaviour is a strong prognostic marker in person with COPD, this imbalance may have influenced outcomes. Sensitivity analyses adjusting for this imbalance did not change the results but cannot entirely adjust for confounding. Fourth, the significant between-group differences found in the EQ-VAS and LSA-DK were not corrected for multiple testing as these were secondary outcomes, and the results may therefore be subject to chance.

A further limitation relates to the study design, where participants had to complete the acute rehabilitation phase before randomisation. Several who had initially consented were unable to reach randomisation due to health complications, disinclination, or death. As a result, only those able or willing to complete the initial phase of exercise entered the clinical trial. This might have introduced a selection bias.

Cluster randomisation was used for logistical reasons, as participants had to start simultaneously in teams. However, the analyses were conducted at individual level and did not account for clustering. Given the small number of participants and clusters, it was not relevant or reliable to conduct cluster-level variation. This approach was pragmatic but may have contributed with bias and should be considered.

In addition, the 5RSTS assesses lower-limb strength, while the intervention focused on prolonged, low-intensity cycling and changes in daily activity. The primary outcome therefore may not capture all types of changes that long-term, home-based cycling can entail. Measures of walking distance or endurance might have reflected other aspects of the intervention, but several participants were unable to complete such tests due to disease severity.

Lastly, the study was not powered to detect differences in readmission or mortality. Therefore, competing risks are still possible. Because mortality can correlate with readmission in this population, models that account for competing risks are recommended, but were not attainable for this study due to lack of power. The direction of this potential bias remains unclear [44].

Conclusion

In this pragmatic RCT, the long-term, home-based, virtual cycling intervention participants did not improve significantly on physical performance, activity levels, lung function, or disease impact, but the intervention was associated with higher self-reported health measured with EQ-VAS and functional mobility measured with LSA-DK after 12 months. These findings should, however, be interpreted in light of the limited sample size, substantial attrition, and methodological restraints of the study.

Long-term, home-based, virtual cycling may represent a supportive exercise option for the frailer plwCOPD and with limited mobility, provided adequate support is available. Importantly, this study highlights the methodological challenges of conducting clinical trials in a frail population like post-AECOPD persons. This includes issues of recruitment, retention, and attrition.

Future trials should account for these constraints when designing the intervention and aim to improve uptake and retention, and thus secure sufficient power.

Appendix.

Table A1.

Readmission and mortality rates.

		N events	Rate	Rate Ratio	95% CI		p-value
Readmission	Intervention	23	1.401	1.193	0.655	2.173	0.563
	Control	20	1.176	1
Mortality	Intervention	6	0.366	1.557	0.439	5.516	0.493
	Control	4	0.235	1

Table A2.

Outcomes at 6 weeks and 6 months.

Outcome	Group (n)	6 weeks [Lower;upper]	Mean differences between groups	p-value	6 months [Lower;upper]	Mean differences between groups	p-value
5 RSTS (seconds)	Intervention (18)	1.8 [0.9;2.6]	−0.2 [−1.3;1.0]	0.8	1.5 [0.6;2.3]	−0.9 [−2.1;0.3]	0.1
	Control (19)	2.0 [1.2;2.7]			2.3 [1.6;3.1]
30s STS (repetitions)	Intervention (19)	10.8 [0.1;21.5]	−1.9 [−14.4;10.6]	0.8	8.6 [0.0;17.2]	−5.6 [−18.8;7.6]	0.4
	Control (19)	12.7 [−0.2;25.6]			14.3 [−0.3;28.8]
Lung function, FEV1% pred.	Intervention (19)	47.1 [41.4;52.9]	−0.1 [−0.3;0.3]	0.9	48.9 [42.8;54.9]	0.1 [−0.3;0.4]	0.5
	Control (19)	47.1 [41.4;52.8]			42.8 [39.9;51.6]
SPPB, points	Intervention (19)	4.5 [3.0;6.1]	−1.2 [−3.4;1.0]	0.3	4.2 [2.6;5.9]	−1.8 [−4.1;0.5]	0.1
	Control (19)	5.8 [4.2;7.3]			6.0 [4.4;7.7]
SGRQ, points	Intervention (18)	49.1 [41.4;56.8]	−6.6 [−17.3;4.2]	0.2	51.0 [43.0;59.0]	−2.8 [−13.9;8.3]	0.6
	Control (19)	55.6 [48.1;63.1]			53.8 [46.1;61.6]
CAT, points	Intervention (18)	11.5 [8.0;14.9]	−0.5 [−5.3;4.2]	0.8	10.1 [6.5;13.7]	−1.6 [−6.7;3.4]	0.5
	Control (19)	12.0 [8.6;15.4]			11.7 [8.2;15.2]
EQ-5D-5L, index score	Intervention (18)	0.7 [0.5;0.8]	−0.1 [−0.3;0.1]	0.2	0.7 [0.6;0.8]	0.1 [−0.1;0.3]	0.5
	Control (19)	0.8 [0.7;0.9]			0.6 [0.5;0.8]
EQ-VAS, 0–100	Intervention (18)	62.4 [51.5;73.3]	9.9 [−5.2;25.1]	0.2	59.1 [47.7;70.6]	7.7 [−8.2;23.6]	0.3
	Control (18)	51.4 [40.3;62.5]			51.4 [40.3;62.5]
LSA-DK, points	Intervention (18)	57.5 [43.9;70.9]	0.8 [−18.1;19.7]	0.9	48.8 [34.7;62.8]	−1.9 [−21.4;17.7]	0.9
	Control (19)	56.6 [43.4;69.9]			50.7 [37.1;64.2]

Abbreviations: 5RSTS, five-repetition sit-to-stand test; 30s STS, 30-second sit-to-stand test; FEV1% pred., forced expiratory volume in 1 second (% predicted); SPPB, Short Physical Performance Battery; SGRQ, St George’s Respiratory Questionnaire; CAT, COPD Assessment Test; EQ-5D-5L, EuroQol 5-Dimensions 5-Levels; EQ-VAS, EuroQol Visual Analogue Scale; LSA-DK, Life-Space Assessment, Danish version.

Funding Statement

This work was supported by the Karen Elise Jensens Foundation, Simon Fougners Hartmanns Family Foundation, Region Nordjylland Innovation Foundation, Skibsreder Per Henriksen, R og hustrus Foundation, The Beckett Foundation, Lungeforeningen and Danish Physiotherapists.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Generative AI (ChatGPT, OpenAI, GPT-5.2) was used for language editing and formatting support. The authors reviewed and edited all content and take full responsibility for the final manuscript.