Characteristics of the included studies

Following the removal of duplicate articles and those with non-matching titles, 432 records were retrieved for screening. The selection criteria identified a total of 16 studies (see Fig. 1). The main and supplementary outcomes are usually presented as abbreviations; therefore, Table 2, which contains the abbreviations and acronyms used in this systematic review, has been included. Details of the chosen studies, also referred to as articles, were provided in Tables 3 and 4, numbered chronologically based on their publication dates.

Fig. 1: Flowchart and selection of studies to conduct the review based on the PRISMA guidelines.

Workflow diagram describing the process for selecting articles aligned with the systematic review, detailing the studies from identification, screening, and final selection.

Table 2 Abbreviations and acronyms used.Table 3 Set of studies reporting WBPTR, detailing the characteristics of the participants and the programme.Table 4 Set of studies reporting the content of web-based pulmonary telerehabilitation and the outcomes.

The total number of participants in these studies was 3190, with 53.2% (1697 participants) using WBPTR, while the remainder belonged to control groups. For the intervention group, participants had mean ages ranging from 57.7 to 69.2 years, with forced expiratory volume in one second (FEV1) ranging from 37.8% to 60.6% predicted. Meanwhile, for the control group, mean participant ages ranged from 57.6 to 70.4 years, with FEV1 ranging from 36% to 65.2% predicted (Table 3). Table 3 also includes the percentage of oxygen users and current smokers in both the intervention and control groups.

Among all studies, 81.3% reported the FEV1 % predicted (n = 13), 37.5% reported oxygen users (n = 6), 68.75% reported current smoking status (n = 11), and 93.75% reported gender (n = 15). Regarding the race of participants, two studies included percentages of various races (12.5%), four clarified the percentage of Caucasian participants (25%), and 62.5% did not include this information (n = 10).

The type of training varies across studies (Table 3: Training type). We categorise them into two main groups: Intensive training, which includes exercises and weight lifting in short sessions; extensive training, which involves activities like walking and cycling in longer sessions (n = 10, 62.5%); and a combination of both (n = 6, 37.5%). Session frequency was measured weekly (Table 3: Freq.), indicating the minimum number of days to be completed. 87.5% of studies reported this (n = 14). Regarding the duration lengths of the programmes (Table 4: Wks.), 18.75% (n = 3) of the articles specified trial lengths between 4 and 8 weeks, 37.5% (n = 6) between 12 and 24 weeks, and seven studies (43.75%) indicated a length longer than 24 weeks. Additionally, a large proportion of studies (75%) did not report the required equipment to perform the tasks (Table 3).

Web-based platforms enable digital reporting of participants’ progress. The supervision of goal achievement by therapists varies across studies, either using the web platform alone for feedback (n = 5, 31.25%) or enhancing it with telephone calls (n = 4, 25%), text messages (n = 2, 12.5%), or email (n = 4, 25%). Moreover, the web-based telerehabilitation feature allows for the inclusion of additional information accessible to users. Consequently, only 26.32% of articles lack any educational content regarding the disease. Conceivably, every study require a computer to access the web platform. However, access can also be done through devices such as tablets and smartphones (n = 4, 25%). Accessories such as spirometers (n = 2, 12.5%), pedometers (n = 7, 43.75%), oximeters (n = 1, 6.25%), and accelerometers (n = 2, 12.5%) were also used to collect physiological information.

All trials include at least some objective or subjective measurement of the progress in the patient’s recovery. We distinguish between two groups: those suitable for meta-analysis (primary outcomes) and those studied independently (secondary outcomes). For the meta-analysis, measurements from each article were extracted, normalising values when necessary (using SI units).

Primary outcomes were related to: a) Measurements of exercise performance (EP), where 62.5% of studies report the 6MWT (n = 10) and 37.5% study steps per day; b) Questionnaires to evaluate Health-Related Quality of Life (HRQL), where seven articles include the SGRQ (43.75%) and 18.75% report the CRQ-SAS (n = 3); c) Questionnaires about dyspnoea, where 25% of studies (n = 4) report the CRQ-D (independently or as part of CRQ-SAS), while three articles (18.75%) include the mMRC.

On the other hand, secondary outcomes consider: a) Arm lifts (n = 1, 6.25%) and the Endurance Shuttle Walk Test (n = 1, 6.25%), both related to EP; b) The CAT (n = 2, 12.5%) and the CCQ (n = 2, 12.5%), both related to HRQL; c) Measurements related to disease knowledge: the Bristol COPD Knowledge Questionnaire (n = 2, 12.5%); d) Fatigue-related health status: the Metabolic Equivalent (n = 1, 6.25%) and Multidimensional Fatigue Inventory (n = 2, 12.5%); e) Physical status of patients, the Body Mass Index, Airway Obstruction, Dyspnoea, and Exercise Tolerance (n = 2, 12.5%); and f) other measures such as the Hospital Anxiety and Depression Scale and COPD exacerbations, both included in at least one of the studies.

Risk of bias in the included studies

The risk of bias in each of the selected studies was depicted in Fig. 2, indicating a high likelihood of bias (due to the study nature) in blinding of participants and personnel, and blinding of outcome assessment, with 62.5% of articles showing high or unclear ROB.

Fig. 2: Risk of bias for the included studies.

On the right, the total percentage for the Risk of Bias (ROB) is displayed, while on the left, each risk of bias item is detailed for each author.

Primary outcomes: meta-analysis

A total of thirteen studies provide quantifiable information covering Exercise Performance. The analysis of the 6WDT and the SPD was discussed in this subsection. Studies reporting quantifiable information were included in the meta-analysis and those that did not were analysed independently.

With the information available from the 6WDT, ten studies (Fig. 3) involving a total of 1121 participants were included in the meta-analysis22,23,24,25,26,27,28,29,30,31. The observed mean differences ranged from −12.27 to 34.50, with more estimates favouring WBPTR (60%). The estimated average mean difference based on the random-effects model was \(\hat\) = 5.00 (95% CI: −5.19 to 15.21). Therefore, based on the values dimension, the average outcome did not differ significantly from zero (z = 0.96, p = 0.34). A 95% prediction interval for the true outcomes was given by −10.76 to 20.77. Hence, although the average outcome was estimated to favour WBPTR, in some studies the true outcome were in fact favouring traditional care30,31.

Fig. 3: Six-minute walking test forest plot.

On the left, a tabulation of the information extracted from each study is provided, covering both the WBPTR group (583 participants) and the control group (538 participants). On the right, a graphical representation shows the incidence of each study based on the number of patients.

Regarding the SPD, six articles (involving a total of 854 participants)23,27,30,32,33,34 provide enough information for a meta-analysis (Fig. 4), whereas the study of Wan and colleagues35 was analysed independently. For the meta-analysis, the observed mean differences ranged from −238.00 to 1312.28, with the majority of estimates favouring web-based telerehabilitation approaches (83%). The estimated average mean difference based on the random-effects model was \(\hat\) = 446.66 (95% CI: 96.47 to 796.86). Therefore, the average outcome differed significantly from zero (z = 2.5, p = 0.01), denoting better results when using WBPTR. A 95% prediction interval for the true outcomes was given by −71.78 to 965.09. Hence, although the average outcome was estimated to be positive, one study favours traditional care23. On the other hand, the results obtained by Wan et al.35 suggest that no significant statistical differences were reached when comparing both traditional and WBPTR programmes.

Fig. 4: Steps per day forest plot.

On the left, a tabulation of the information extracted from each study is provided, covering both the WBPTR group (496 participants) and the control group (358 participants). On the right, a graphical representation shows the incidence of each study based on the number of patients.

Considering the Health-Related Quality of Life, ten studies include either the St. George’s Respiratory Questionnaire or the Chronic Respiratory Disease Questionnaire Self-Administered Standardised, both of which were analysed subsequently.

For the SGRQ (Fig. 5), all articles mentioning its use provide sufficient information to be included in the meta-analysis. Therefore, a total of seven studies were considered24,25,29,30,33,34,36, involving a total of 1166 participants. The observed mean differences ranged from −4.70 to 4.60, with more of estimates favouring WBPTR (71%). The estimated average mean difference based on the random-effects model was \(\hat=-0.15\) (95% CI: −2.24 to 1.95). Therefore, the average outcome did not differ significantly from zero (z = −0.14, p = 0.89). A 95% prediction interval for the true outcomes was given by −3.12 to 2.83. Hence, although the average outcome was estimated to be negative, in two studies the true outcome was in fact positive (favours traditional care)26,30.

Fig. 5: St. George’s Respiratory Questionnaire forest plot.

On the left, a tabulation of the information extracted from each study is provided, covering both the WBPTR group (668 participants) and the control group (498 participants). On the right, a graphical representation shows the incidence of each study based on the number of patients.

In a case similar to the SGRQ, all trials that suggest measuring the CRQ-SAS provide sufficient information to analyse their results jointly (Fig. 6). Thus, data from three studies22,23,28 were included in the meta-analysis, involving 286 participants. The observed mean differences ranged from 0.05 to 0.31, with all the estimates favouring WBPTR. The estimated average mean difference based on the random-effects model was \(\hat=0.17\) (95% CI: −0.13 to 0.46). Therefore, the average outcome did not differ significantly from zero (z = 1.11, p = 0.27).

Fig. 6: Chronic Respiratory Disease Questionnaire Self-Administered Standardised (CRQ-SAS) forest plot.

On the left, a tabulation of the information extracted from each study is provided, covering both the WBPTR group (146 participants) and the control group (140 participants). On the right, a graphical representation shows the incidence of each study based on the number of patients.

This subsection analyses the seven studies that include the Chronic Respiratory Disease Questionnaire and the Modified Medical Research Council scale, both of which refer to dyspnoea.

Either independently or as part of the CRQ-SAS, four studies considers the dyspnoea22,23,28,36, all including sufficient information to analyse them together in the meta-analysis (Fig. 7) and involving a total of 348 participants. The observed mean differences ranged from −0.20 to 0.30, with the majority of estimates being negative (75%), i.e., favouring traditional care. However, the average outcome was practically zero (z = −0.08, p = 0.94). The estimated average mean difference based on the random-effects model was \(\hat=-0.01\) (95% CI: −0.29 to 0.27).

Fig. 7: Chronic Respiratory Disease Questionnaire - dyspnoea (CRQ-D) forest plot.

On the left, a tabulation of the information extracted from each study is provided, covering both the WBPTR group (168 participants) and the control group (180 participants). On the right, a graphical representation shows the incidence of each study based on the number of patients.

For the mMRC, three studies24,30,32 provide information to include them in the meta-analysis (Fig. 8), whereas the study of Bourne et al.25 was analysed independently. A total of 291 participants were involved. Thus, for the data analysed together, the observed mean differences ranged from −0.20 to −0.06, with all of them favouring WBPTR. The estimated average mean difference based on the RE model was \(\hat=-0.14\) (95% CI: − 0.43 to 0.14). Therefore, the average outcome differ from zero, but not significantly (z = −0.98, p = 0.33). Moreover, the independent results obtained from Bourne et al.25 show equivalent benefits between WBPTR and traditional care, suggesting an improvement in dyspnoea, although there were no significant differences between treatments.

Fig. 8: Modified Medical Research Council dyspnoea Scale forest plot.

On the left, a tabulation of the information extracted from each study is provided, covering both the WBPTR group (146 participants) and the control group (145 participants). On the right, a graphical representation shows the incidence of each study based on the number of patients.

Remark 1

According to Cook’s distances, the study by Kessler and colleagues26 could be considered overly influential for the 6WDT and the SGRQ. For the remaining outcomes, no study was considered to be overly influential.

Remark 2

According to the Q-test, there was no significant heterogeneity in the true outcomes for any of the studies.

Remark 3

Examination of the studentised residuals revealed that none of the meta-analyses developed presented outliers. None of the studies had a value greater than ± 2.81, ± 2.64, ± 2.69, ± 2.39, ± 2.49, and ± 2.39 for 6MWT, SPD, CRQ-SAS, SGRQ, CRQ-D, and mMRC, respectively.

Secondary outcomes

For exercise performance, arm lifts and the Endurance Shuttle Walk Test (ESWT) were also considered secondary outcomes. In this case, Nguyen and colleagues22 mention that at six and twelve months, WBPTR participants perform more arm lifts compared to the control group, but numerical data was not provided. On the other hand, the ESWT was analysed in the study by Chaplin et al.36, which shows a significant change at the end of treatment for both the control and intervention groups, but no significant difference between the groups.

Secondary outcomes for Health-Related Quality of Life include the COPD Assessment Test and the Clinical COPD Questionnaire. In the study by Bourne and colleagues25, the disparity in the CAT scores between groups favoured the WBPTR, with a difference of −1.0. Similarly, the mean evolution graph indicates a 3.6-point difference between the HappyAir group and the control group after 12 months of follow-up in the study by Jiménez et al.29. This suggests a better quality of life in the WBPTR group compared to the baseline evaluation, as a 2.5-point difference in the CAT questionnaire was considered clinically relevant. Moreover, two articles in the set of studies mention the CCQ31,32, both showing improvement within groups, although there were no significant changes between the treatment groups.

Although in a smaller set, several studies consider parameters such as COPD knowledge, fatigue, physical condition of the patients, anxiety, depression, among others. To encompass these subjective measurements, this section includes the results of these studies.

The Bristol COPD Knowledge Questionnaire (BCKQ) was measured and compared between the intervention group and the control group in studies by Wan et al.24 and Robinson et al.30. Both articles show a statistically significant change compared to baseline; however, no noticeable difference was observed when comparing treatments.

The Multidimensional Fatigue Inventory (MFI) was analysed in two studies31,32, examining all sub-items: general fatigue, physical fatigue, reduced activity, reduced motivation, and mental fatigue. Similar to other studies mentioned earlier, both studies show improvement in general fatigue from the end of treatment compared to the beginning. However, no significant difference was observed between the WBPTR group and the control group.

Additionally, the BODE (Body Mass Index, Airway Obstruction, Dyspnoea, and Exercise Tolerance) index was analysed in Kessler et al.26 and Galdiz et al.28. The results from these two trials differ significantly in both within-group and between-group comparisons. Galdiz and colleagues28 report non-significant differences between groups, while Kessler and colleagues26 indicate notable differences both within groups and between groups, showing superior results with WBPTR compared to traditional care and usual care methods.

Additionally, the Metabolic Equivalent of Task (MET) was analysed solely in the study by Vorrink et al.23, which indicates no statistically significant difference over the entire treatment duration, both within groups and in the comparison between groups.

Regarding smoking habits, Voncken et al.37 did not find significant differences in any comparison, whereas Kessler and colleagues26 show that more patients in the WBPTR group stopped smoking during the study.

When analysing parameters related to intervention from health centres and hospitals, several measures provide information on treatment comparisons. Specifically, the Hospital Anxiety and Depression Scale (HADS) was analysed by two studies25,26, with only Bourne and colleagues25 reporting a significant reduction favouring WBPTR. Lastly, COPD exacerbations were examined in one study26, noting a reduction for WBPTR that may potentially be influenced by external factors unrelated to the treatment.