Studies were included if: i) their sample was composed of adult (≥18 years) people with COPD in a stable phase of the disease (i.e., 4 weeks without hospital admissions or exacerbations, nor changes in medication, according to Global Initiative for Chronic Obstructive Lung Disease – GOLD report1); ii) included unsupervised PA interventions for people with COPD compared to usual care (i.e., had not received any PA intervention in the study period); iii) they were original randomised controlled trials (RCT) or quasi-experimental studies; iv) written in Portuguese, English, Spanish or French languages. Studies were excluded if they: i) involved proxy versions; ii) were qualitative studies; iii) included other treatments/activities as an intervention while performing PA; iv) included any directly supervised training (other than a single session), i.e., face-to-face or remote contact (e.g. video-conference); and, v) were performed in hospital-based settings.

For the purpose of this review, the following definition of PA was used: “any bodily movement produced by skeletal muscles that requires energy expenditure.”18 Unsupervised PA interventions were defined as any PA without any supervision, undertaken in any environment and/or at any time, which best suits the person.16 It could include a single supervised session to explain and/or demonstrate the activities; and, remote contact with healthcare professional using technologies, such as, telephone, mobile phone and/or tablet devices, to check patients’ health and monitor their evolution, without being used to interactively coach/instruct the patient (e.g., video-conference).

A systematic literature search was conducted in April 2020, on the following electronic databases: Cochrane Library, PubMed, Scopus, Web of science, EBSCOhost. Electronic search was supplemented by weekly automatic updates retrieved from the databases until September 2021, and hand-searches of references in key systematic reviews.17 The search strategy was performed by title, abstract, keywords/MESH term. The full search strategy is presented in Supplementary material - Appendix 1.

After removing duplicates, two reviewers (CP and VR) independently screened the potential studies (title and abstract), according to the eligibility criteria. The full-text of each potentially relevant study was then independently screened by the same reviewers to decide on its inclusion. Discrepancies were solved by consensus, and if agreement could not be reached, the other authors’ opinion was obtained. Primary outcomes were dyspnoea, exercise capacity and PA. Secondary outcomes included body composition, emotional function, fatigue, health behaviours, healthcare utilisation, HRQoL, mastery, muscle strength, self-efficacy, symptoms of anxiety and depression, adverse events and, dropouts and adherence to interventions.

One reviewer independently extracted the data from included studies, and the other authors checked the accuracy and completeness of information. Data extraction was performed using a pre-developed and structured table-format covering the following topics: characteristics of the study (first author, year of publication, country and study design); setting (i.e., home-based); population (number of participants, sex, age, forced expiratory volume in one second percentage of predicted [FEV1pp], severity of airway limitation [GOLD grades 1-4]1 and comorbidities [type and severity, classified with Charlson Comorbidity Index-CCI]; intervention (type, frequency and duration); outcome and outcome measures; and, results obtained in each outcome measure. Studies with multiple publications were identified to avoid duplicate reports (e.g., number of participants). Corresponding authors of the included studies were contacted via e-mail to request additional data (i.e., means and SD), whenever needed.

Two reviewers independently assessed the methodological quality of each study using the Quality Assessment Tool for Quantitative Studies, developed by the Effective Public Health Practice Project, Canada.23 This tool is comprised of six domains of methodological quality: 1) selection bias; 2) study design; 3) confounders; 4) blinding; 5) data collection methods; and, 6) withdrawals and dropouts.23 Each domain is rated as “strong”, “moderate” or “weak”, according to a standardized guide.24 The overall rating of each study is determined by the total number of “weak” scores, i.e., if the study presented: i) no weak scores, it was rated as strong quality; ii) one weak rating - moderate quality; and, iii) two or more weak ratings - weak quality.23

Inter-rater agreement analysis was assessed using Cohen's kappa to explore the consistency of the quality assessment performed by the two reviewers. The Cohen's kappa ranges from 0 to 1 and agreement was interpreted as: slight (⩽0.2), fair (0.21–0.4), moderate (0.41–0.6), substantial (0.61–0.8), or almost perfect (≥0.81).25

Studies were grouped according to the outcome measures reported. An effect direction plot was computed to deal with the diversity of outcome measures used in the included studies, following the SWiM recommendations.26 This plot considers the study design, effect estimates of each outcome (represented with arrows, i.e., upward arrow ▲= positive health impact, downward arrow ▼= negative health impact, sideways arrow ◄►= no change/mixed effects/conflicting findings), sample size and studies quality (using a traffic light system, i.e., green for studies of high quality, amber for moderate and red for weak quality of evidence).26 The effect estimates were analysed with the Cohen's d effect sizes (ES) based on the Pre/Post means and SD, according to the formula of Morris.27 The ES were interpreted as very small (≥0.01), small (≥0.20), medium (≥0.50), large (≥0.80), very large (≥1.20) and huge (≥2.0).28,29 Results were analysed by counting the effect direction and interpreted using the proportion of effects favouring the intervention 30 Proportions higher than 50% were considered as an improvement in the respective outcome measure.30

Meta-analysis, with forest plots, only included studies reporting the mean changes between the experimental (EG) and control (CG) groups and the respective SD or data allowing the calculation of these estimates. Between-study heterogeneity was quantified using I-squared (I2) statistic. Statistical homogeneity was defined as ≤40%.30

Some data transformation occurred to compute ES. Data presented as 95% of confidence intervals (95% CI) 31 were transformed into SD, using the formula: SD=n*(upperlimit−lowerlimit)/3.92, where n is the sample size.30 Additionally, data presented as median and interquartile range (IQR) 31 were converted into mean and SD using the summary table proposed by Wan and colleagues.30,32

Data analysis were performed using IBM SPSS 24.0 (IBM, Armonk, New York, USA) and RStudio, V1.2.5033 (RStudio, Inc; Boston, MA, USA).

The literature search provided 738 studies. After duplicates removed, 396 records were screened and 303 were excluded. The full-text of 93 articles was assessed and four studies were included. Seven additional studies were identified and retrieved, two from the databases weekly automatic updates and five from the reference list of a key systematic review 17 (Fig. 1). A total of 11 articles were included.

Fig. 1..

Flow diagram of the articles screened and included in the study (n = 11) according to the preferred reporting items for systematic reviews and meta-analyses (PRISMA).

Abbreviations: RCT – randomised controlled trial.

(0.39MB).

Four studies were rated as strong 33-36 (36%), three 31,37,38 (28%) as moderate and four 39-42 (36%) as weak quality. Inter-rater agreement was substantial (Cohen's Kappa=0.72; 95% CI=0.37-1.07; p = 0.003; percentage of agreement= 82%). Quality assessment details can be found in Supplementary material (Table S1).

Included studies were published between 197742 and 202031, and were mostly conducted in the United States of America,33,34 Australia,31,39 United Kingdom 36,38 and Taiwan.35,41 A total of 900 participants with COPD, 446 in the EG and 454 in the CG, were included. Sample sizes ranged between 2038 to 305 participants.33,34 Participants were on average 68±10 years old (n = 6), 541 (59%; n = 9) were male, presented a mean FEV1 of 63.7±15.8 %predicted (n = 3) and the majority had moderate to severe (GOLD grades 2-3) grades of the disease (n = 4). Comorbidities reported (n = 3) included: cardiovascular disease (hypertension, heart failure, myocardial infarction, and peripheral vascular diseases), metabolic syndrome (e.g., diabetes), depression, musculoskeletal disease (e.g., arthritis) and ulcer disease.33,34,36Table 1 presents detail characteristics of the included articles.

Interventions lasted from 6 36,38 to 66 weeks,33,34 being 8-12 weeks 31,35,37,39-42 the most common range duration, and were performed 3 days/week,37 4 days/week 38 or daily. 31,33-36,39-42 All interventions were designed by health professionals (i.e., general practioners,35,42 nurses35,37,39-41, physiotherapists31,36,38 and health coachs33,34) and performed at home.31,33-42

Interventions were single-component in seven 33-35,37,39,41,42 and multi-component in four 31,36,38,40 studies. Aerobic exercise 31,35,36,38,39,41,42 (e.g., walking) and muscle strength 31,36-38,40 were the main training components. Two studies 33,34 focused on promotion of lifestyle PA (i.e., promotion of activities of daily living). Interventions also included diaries;31,41,42 action plans;39,41 information about healthy behaviours;39 phone calls to support the intervention, promote healthy behaviours or deliver self-management training;31,33,34,36,39,40 distribution of handbook/manual;33,34,41 workbook activities;33,34 reading assignments;33,34 and nutritional and psychosocial support.40 More details are presented in Table 1.

A total of 14 outcomes, evaluated by 44 different measurement tools were found in the included studies (Table S1, Figs. 2 and 3).

Fig. 2..

Effect direction plot of unsupervised physical activity interventions in people with chronic obstructive pulmonary disease (n = 11).

(0.25MB).

Fig. 3..

Forest plots illustrating the effect of unsupervised PA interventions on: a) Chronic respiratory questionnaire – dyspnea domain (CRQ-D), b) 6-minute walk distance (6MWD), and c) incremental shuttle walk distance (ISWD), in comparison to usual care.

(0.58MB).

Four studies33,34,37,41 explored the adverse events of unsupervised PA interventions. Two33,34 of these studies found that 63% (n = 192) of participants had no adverse events and 37% (n = 106) had, at least, one adverse event. The most common adverse event was acute exacerbation of COPD, with a twice higher prevalence in the CG (15%, n = 47) than in the EG (9%, n = 28) (p < 0.01).33,34

Nine studies33-39,41,42 reported dropouts, ranging between 7.1%41 to 38.5%.39 Reasons to dropout included: abrupt dizziness,37 acute exacerbation of COPD,35,38 cataract surgery,37 comorbidities,36 death,42 failure to keep appointments,39 intercurrent depressive illness,42 knee pain,38 lack of enthusiasm,42 lost to follow-up,36,38,41 non-COPD related hospital admission,35,37 poor health,39,41 social reasons,36 programme was too easy or not so serious,36,37 time constraints,35,39 travel issues,37,39 too busy to participate 41 and work commitement.36

Only four studies reported adherence to the intervention,31,35,37,39 which varied between limited39 to 93%.31

This systematic review has several limitations that need to be acknowledged. Firstly, the small number of existing studies; their large diversity of designs, outcomes and outcome measures; lack of consensus on the definition of unsupervised PA; and, the high heterogeneity observed in the meta-analysis, limited our conclusions. Nevertheless, a synthesis of the results, using the effect direction plot was computed which provided a thorough synthesis of data. Secondly, the imbalance between participants, i.e., more males than females and moderate to severe participants, limited the generalisation of results. Further studies including more people with COPD in mild and very severe grades and females should be conducted. Thirdly, our search was limited to studies published in English, Portuguese, Spanish or French included in databases. Additional studies may exist in the unpublished grey literature and may have been missed. A thorough search in different databases and scanning the references of key articles and systematic reviews were however conducted to minimise this limitation. Finally, approximately one third of the included studies were of low quality, nevertheless most studies presented moderate to high quality.