This was a multicentre open-label pragmatic parallel-group randomized clinical trial involving 13 hospitals and was conducted between 2014 and 2017. Patient selection criteria were having a diagnosis of moderate-to-severe COPD (a Body mass index, Airflow obstruction, dyspnoea and Exercise capacity [BODE] index score of 3•7) according to international guidelines9 and having been clinically stable for the previous 4 weeks (see the online supplementary material).

The study was carried out in accordance with the principles of the Declaration of Helsinki and started after receiving authorization from the Spanish Agency for Medicines and Health Products and the approval of the Basque Country Ethics Committee, as well as the local Ethics Committee and Director at each institution. All patients gave written informed consent. ClinicalTrials.gov Identifier: NCT03247933.

Recruited patients (recruitment visit, ∧1) attended an initial 8-week outpatient-based PR programme conducted in each hospital rehabilitation department. Following completion of this programme, patients were randomly allocated (baseline visit, 0) to either the TelePR programme [intervention group (IG)] or the standard, usual care programme [control group (CG)]. Allocation was stratified by centre and based on randomly permuted blocks of variable size with a 1:1 allocation ratio. Clinical epidemiology staff kept the centralized allocation lists concealed (see online supplementary material).

Continuous variables are described using either the mean and standard deviation or the median and range, depending on the data distribution. Categorical variables are described using frequencies and percentages. The main efficacy analysis used analysis of covariance (ANCOVA) techniques to test for differences between groups in outcome variables at the 12-month visit, adjusted for baseline values. Least squares mean differences and 95% confidence intervals were calculated. Secondly, additional analyses were conducted using linear mixed models, adjusted for baseline values and including time-group interaction terms. Although P-values lower than .05 were a priori considered statistically significant in two-sided tests, the threshold was adjusted for multiple comparisons using the Holm method whenever a P-value lower than .05 was found in a main effect test (group variable or time effect). Missing values were not imputed. All patients were analyzed in the group to which they were randomized. Stata 15.1 for Windows was used for all the analyses (StataCorp. 2017.Stata Statistical Software: Release 15. College Station, TX: StataCorp LLC).

The sample size calculation was based on the effect of the intervention on exercise tolerance (6MWT distance) at 12 months as primary outcome: assuming superiority of the IG, we calculated that 70 patients (35 per group) would be required for a two-sided statistical comparison with a 5% type I error rate, to detect a clinically relevant difference of 35m between arms with a standard deviation of 50m and statistical power of 80%. The initial plan was to recruit 84 patients, allowing for a 20% dropout rate. This was subsequently altered, given recruitment pace and logistical capabilities. An extra 10 patients were recruited increasing the statistical power to 90%, while keeping all the other statistical considerations unchanged.

Rates of adherence to the scheduled appointments were 92.4% in the IG and 84.4% in the CG (Table 4). Four patients in each group were non-attenders according to our prespecified definition. Overall compliance in the IG was 60% of the maximum scheduled days (weekly compliance is reported in the online supplementary material). Twelve patients in the IG were classified as non-compliers.

There were a total of 56 registered episodes of acute exacerbations in the IG, affecting 26 patients versus 47 in the CG affecting 21 patients, with no statistically significant differences between groups (Table 5). Including the fact that the patient had had at least one exacerbation episode as a covariate, ANCOVA indicated a relative increase in 6MWT distance of 23.2m (∧1.4/47.7) (P=.064) in the IG. The occurrence of exacerbation was associated with an overall decrease of 14.9m (∧39.6/9.7) (P=.231). Mixed models showed that IG patients were able to walk 14.9m more (∧4.2/33.9) (P=.125) and the occurrence of exacerbation was associated with an overall decrease of 15.8m (∧32.3/0.8) (P=.062).

As illustrated in Table 6, only the subgroup of patients allocated to the IG who did not experience any exacerbation episode throughout the study period improved their exercise tolerance, although the difference did not reach significance.

Several studies have described difficulties in maintaining the benefits obtained after an intensive PR programme. Unfortunately, improvements achieved during intensive rehabilitation schedules tend to diminish over time.17•20 Possible contributing factors include programme duration, intensity, location, and psychosocial factors.21•23 Some research has suggested that supervised post-rehabilitation exercise programmes could be effective in preserving exercise capacity.24,25 In a 12-month study, Spencer26 did not find any difference between two groups doing once-a-week supervised outpatient-based exercise and unsupervised exercise, both maintaining the improvements obtained in the initial phase over the study duration.

Telemedicine-based interventions promoted through phone calls, websites, or mobile phones have shown promising results in COPD patients and several studies have demonstrated the feasibility of these programmes.26•28,4,29 In our study, all patients were able to use the technology and it was not necessary to withdraw any patients for difficulties managing the devices provided.

Most telehealth studies reported have major methodological limitations such as a short duration, small sample sizes and high dropout rates.30•34 There is also a paucity of evidence on the long-term benefits of telerehabilitation interventions conducted to help maintain benefits gained during initial rehabilitation. One of the two studies found is a small non-controlled study of 10 COPD patients attending PR and followed up with a telerehabilitation programme over 2 years by Zanaboni et al., 7 and they concluded that exercise maintenance via telerehabilitation was feasible and effective, achieving improvements in physical performance, health status and quality of life. The other is a randomized study of the same length as ours conducted by Vasilopoulou et al.,8 in which 6MWT distance (a secondary endpoint) was stable in both active interventions: multimodal home-based maintenance and hospital-based maintenance rehabilitation groups, with a similar pattern to that in our TelePR group. They only found significant deterioration in exercise tolerance in their control group, which did not receive any initial intensive PR intervention.

In our study, we have also observed that the occurrence of exacerbation episodes during the maintenance period can have a major influence on exercise tolerance, to such an extent that it may even counterbalance any potential benefit of a planned regular maintenance strategy (Tables 6 and 7).

Level of compliance with the planned maintenance exercises was not a significant predictor of final physical performance in our study, either alone or after accounting for the patient having experienced at least one acute exacerbation episode (Table 7). The deleterious influence of exacerbations on functional capacity is widely acknowledged. Symptoms and physiological performance usually return to baseline pre-exacerbation levels after 7 (range 4•14) days for dyspnoea, but full recovery is not achieved in all patients even more than 1 month after the start of the acute event.22 In relation to this, our estimates of around 16m of deterioration in 6MWT distance associated with at least one acute exacerbation episode are, to the best of our knowledge, the first data provided on the potential magnitude of this detrimental effect during a PR maintenance programme. Alahmari et al. state that having attended a PR programme in the past is associated with an attenuated short-term impact of exacerbation on exercise tolerance, but their patients were not undergoing an active maintenance intervention.35

Furthermore, the only subgroup in our trial that showed some improvement in exercise tolerance at the end of the maintenance period, although not statistically significant, was that composed of IG patients without any exacerbation event during the study (Table 5). We suggest that any clinical study aimed at assessing the effects of PR maintenance programmes on physical, emotional and HRQoL endpoints should also carefully register the occurrence, main features and clinical consequences of acute exacerbations as a way to gain deeper insight into the potential effect-modifying role played by these common intercurrent events. Frequency and/or intensity of exacerbations could also be used as an additional selection criterion for inclusion of a COPD patient on a maintenance programme or adaption of such a programme to specific clinical characteristics.

There is little experience on the issue of whether interventions using telerehabilitation programmes have a protective effect and reduce the rate of exacerbations. A recent study8 reports a protective effect of a telerehabilitation programme on the number of hospital admissions due to COPD acute exacerbations. In our study, as in others,26 the number of episodes of acute exacerbation was indeed higher in the IG (56 episodes versus 47 in the CG), but without statistical significance (Table 5). Although all patients were instructed to report any adverse events and medical records were systematically assessed to identify exacerbation events, it cannot be ruled out the possibility that the TelePR programme, with an accurate tally of any period of physical inactivity, could strengthen the detection and reporting of intercurrent episodes.

There is general agreement that compliance with long-term maintenance programmes is often low and decreases over time. In our study, the rate of compliance with the TelePR was 60% of the overall planned exercise days. We do not know of any other study that has given such a detailed account of the extent of accomplishment of planned exercises (see online supplementary material). If we focus on attendance to scheduled appointments, our rates are higher than those reported by most studies regardless of their design. Brooks19 in a study involving 109 patients followed over 1 year observed that 24 (22%) did not return for follow-up appointments after the baseline visit and less than half attended the last visit. Further, Spencer26 reported an adherence rate of 30% at 12 months. Zanaboni7 observed that rates of daily card registration decreased from 42.9% to 13.9% over the 2-year study, and Guell36 in a 3-year randomized trial reported adherence rates of 66% and 17% for the IG and CG respectively. In contrast, the multimodal home-based maintenance TelePR intervention by Vasilopoulou et al.8 showed notably high adherence rates (of over 90%) to each component of their telehealth strategy.