Introduction

The epidemiological transition from communicable to non-communicable disease (NCD) imposes a ‘double burden’ on low- and middle-income countries (LMICs)1, which continue to combat infectious diseases but are typically not yet ready to manage NCDs including chronic respiratory diseases (CRDs)2. CRDs are common3,4 and disabling5,6,7 imposing a substantial burden in LMICs. Poor awareness and insufficient resources8,9,10 in terms of infrastructure for diagnosis, availability of essential drugs, skilled health professionals, and overall healthcare priorities5 limit management options11.

Pulmonary rehabilitation (PR) is an effective component of CRD care12. PR is a comprehensive, multidisciplinary, individually tailored intervention designed to overcome the deconditioning induced by CRDs13. The components of PR include, but are not limited to, exercise programmes, chest physiotherapy, education, and supporting self-management and lifestyle change, after optimising the recommended pharmacotherapy13,14,15. PR cost-effectively reduces symptoms, morbidity, hospital admission (and readmission), duration of hospital stay, and emergency medical help and improves functional exercise capacity and health-related quality of life (HRQoL)16,17,18,19,20.

However, most of the evidence is generated from high-income countries (HICs) and is disease specific21,22,23,24 (most commonly chronic obstructive pulmonary disease (COPD)), whereas respiratory disease is often much less differentiated in LMICs. In addition, PR services as developed in HICs may not be deliverable in the same format in LMICs25,26 with substantial differences in resources, awareness, culture, healthcare configuration, and profile of diseases27,28, which may affect overall management strategy. The potential gains to individuals and healthcare economies, however, are large given the burden of disease in LMICs29,30.

Despite well-established effectiveness19,23, PR services are often unavailable even in HICs31,32,33 and uptake (by clinicians and patients) is poor particularly in LMICs and especially in rural communities34. A strategy is needed to elaborate PR programmes that are deliverable and effective in LMICs. We therefore aimed to systematically search the literature to: (1) assess the impact of PR on HRQoL and exercise capacity, when delivered in low-resource settings for people with CRD, (2) identify the components used in effective interventions, and (3) describe the models of care deliverable in low-resource settings.

Results

Study selection

Our systematic review identified 8912 records. We also found an additional 82 records from forward citation. Following the removal of duplicates, 7437 titles and abstracts were screened (Fig. 1). Fifty-six articles were reviewed in full text, with 43 articles excluded. Thirteen articles met the review criteria and were included35,36,37,38,39,40,41,42,43,44,45,46,47. No additional papers were identified in the pre-publication update. Total recruitment for the study was 661 individuals with CRD. Attrition was reported in 9 studies; 96 (20%) of the 479 subjects dropped out.

Fig. 1: PRISMA flow diagram.
figure 1

Flowchart reporting the number of articles identified, screened, excluded and included.

Full size image

Study participants

Study participants were COPD patients35,37,38,39,40,41,42,43,44,45,46,47 of varying degree of severity in all the trials except one which recruited people with pulmonary impairment after TB (PIAT)36. Total number of enrolled participants was 661 of which COPD and PIAT were 83% and 17%, respectively.

Geographical area

The trials were conducted in Turkey (n = 4)35,39,40,43, Brazil (n = 3)37,41,46, India (n = 2)38,47, Egypt (n = 1)42, Iran (n = 1),44 South Africa (n = 1)36, and Venezuela (n = 1)45.

Study settings

Five studies were conducted at hospital outpatient departments37,38,39,43,45 with or without continuation of exercise at home, seven were home-based35,36,40,42,44,46,47 training with or without telephonic/face-to-face monitoring or supervision, and one trial was conducted in a community centre41. Wherever the PR was delivered, all baseline and follow-up data were collected in a hospital/centre setting.

Risk of bias (RoB) assessment

Overall RoB is shown in the first column of Table 1 and detailed in Supplementary Results 1. Almost all studies were at overall high RoB, with only two studies36,39, which concealed randomisation and took steps to avoid other biases, at moderate RoB. Due to the nature of the intervention, blinding of the patients or the personnel delivering the PR was not possible, but only one study explicitly stated that outcome assessment was blind to allocation36. Attrition was a problem or was not clear in all but three studies39,41,46. None of the studies had a published protocol, so selective reporting could not be assessed.

Table 1 Summary table of included trials with key characteristics, main findings, and interpretation.
Full size table

Effectiveness of intervention (Objective 1)

Although 6-min walking test (6-MWT), St George’s Respiratory Questionnaire (SGRQ), and modified Medical Research Council (mMRC) were widely used to assess functional exercise capacity, HRQoL, and breathlessness respectively, only six of the trials presented between-group comparisons36,39,40,42,44,46. The other seven provided within-group differences35,37,38,41,43,45,47. In addition, heterogeneity in terms of mode of intervention, duration, setting, comparator, and baseline measurements confirmed our decision that meta-analysis was not appropriate.

We therefore undertook a narrative synthesis and illustrated functional exercise capacity, HRQoL, and breathless in a harvest plot (Fig. 2). Our interpretation of the study findings and the structured process determining the decisions that underpinned the harvest plot are described in column 5 of Table 1.

Fig. 2: Harvest plot illustrating the impact of pulmonary rehabilitation on functional exercise capacity, health-related quality of life, and breathlessness.
figure 2

Each column represents an included study, shaded according to whether it is a RCT (solid shading) or within group comparison (hatched shading). The depth of shading represents study duration of 4-7 weeks (light shading); 8-11 weeks (moderate shading); 12 weeks or more (dark shading). The height of the bars represent the number of patients. The icon on the top of the bars represents the overall risk of bias as high risk of bias (red) or moderate risk of bias (yellow). Within the icon the mode of delivery of the PR is indicated as + (OPD-based); ^ (Home-based) or C (Community-based). The effectiveness of interventions is illustrated with respect to functional exercise capacity, health-related quality of life, and breathlessness in the three tiers of the graph. Studies are positioned according to whether overall the outcomes were positive (i.e., interventions were significantly beneficial), negative (i.e., interventions were significantly harmful), or had no effect. Table 1; Column 5 details how these decisions were reached.

Full size image

Changes in functional exercise capacity were measured in 11 studies35,36,37,38,39,40,41,42,43,46,47. Significant positive changes were found in 10 studies35,37,38,39,40,41,42,43,46,47; the exception being one of the two studies at moderate RoB53. HRQoL was measured in 12 studies35,37,38,39,40,41,42,43,44,45,46,47; all showing positive changes. Breathlessness was measured in 11 studies35,36,37,38,39,41,42,43,45,46,47 of which 9 studies35,37,38,39,41,42,43,45,47 showed significant positive changes and 2 studies (1 at moderate RoB)36,46 showed no changes after intervention. None of the studies reported negative effects after the intervention.

Components of the intervention (Objective 2)

All interventions included exercise and non-exercise components (as per inclusion criteria), though the approach, content, method of delivery, and duration varied. The components are described in Table 1 and their presence are indicated in a matrix in Table 2.

Table 2 Components of pulmonary rehabilitation from the selected papers.
Full size table

Endurance training was included in all 13 studies. Other common exercises were upper limb exercise35,36,37,39,45,46 and strength training in seven studies37,38,39,40,42,43,46 and stretching exercises in four studies39,42,43,45. Although not described in detail, the other common component was breathing exercises included in eight studies35,36,38,42,43,44,45,47. Along with the exercise, patient education was provided in ten studies35,36,38,39,40,41,42,43,44,46, and skills (such as inhaler technique and airway clearance) were included in seven studies35,36,39,40,42,43,47. Other components in a minority of studies were social support38, optimisation of pharmacotherapy35,37, nutrition40,42,43,44, coping strategies35,38,40,43,47, psychological intervention35,40,43,46, self-management42, and physical activity interventions43,44,46. Smoking cessation support was reported in only two studies35,44.

Models of care (Objective 3)

We identified three models of PR service in our included studies according to the settings in which they were delivered (see Table 3). Five were based in hospital or rehabilitation centres37,38,39,43,45, and one was based in a community health centre41. Only one was delivered completely at home35 while most home-based programmes36,40,42,44,46,47 provided initial training in the hospital or centre and maintained telephone40,44,46 or face-to-face supervision42,47. The programmes typically lasted 8 weeks (range 4–12), with supervised sessions lasting between 30 and 120 min provided 2 or 3 times per week. Home-based programmes promoted more frequent exercise sessions often supported by telephone or face-to-face contacts. Physiotherapists provided the sessions in six studies36,38,39,40,41,43, with nurses involved in four studies35,40,42,44. Adherence to the PR course was poorly reported with no details provided about reasons for non-completion.

Table 3 Models of pulmonary rehabilitation services.
Full size table

Inexpensive instruments were often used in the studies, which ensured the wide availability and acceptability to the consumers. Lower limb endurance exercise was conducted by walking as opposed to expensive stationary bicycle with upper limb resistance/strength training conducted using home-made weights, such as water bottles. Breathing exercises were done with similar devices that are used in higher resource setting (e.g. incentive spirometers, tri-flow).

Discussion

In summary, our systematic review identified and selected 13 heterogeneous studies from 7 different countries with a total study population of 661 patients. Overall, PR was reported as being effective in terms of improving functional exercise capacity, HRQoL, and breathlessness, though RoB was high in 11 studies. Of the two at moderate RoB, one showed no benefit in any of the outcomes reported36. The exercise programmes typically included endurance, interval, upper limb, and resistance/strength training. The commonest additional components were education to improve knowledge and skill acquisition (e.g. inhaler technique) and strategies for coping with breathlessness. Smoking cessation was provided in only two studies. Most PR services were provided in hospital settings or home based, with some describing adaptations to locally acceptable and deliverable approaches.

The strength of this systematic review is its broad literature search constructed with the help of a senior librarian and informed by Cochrane’s standard search terms for COPD and LMICs. Nevertheless, we may have missed important studies of PR conducted in low-resource settings. Although we did not specifically search for papers in other languages, we were open to including non-English language papers but none were identified in our searches, perhaps because locally conducted studies or articles in local languages are often not published in indexed journals48. We may have missed important information from these studies but lacked resources to extend the search to non-indexed publications and grey literature.

We followed rigorous Cochrane methodology duplicating the selection, data extraction, and quality assessment procedures, but confidence in our findings is limited by the high RoB in most of the studies included. We only included controlled trials because we wanted to assess effectiveness. We acknowledge, however, that in LMICs there are many challenges and barriers such as lack of infrastructure, heterogeneity of resources, and poor health literacy, which discourage clinical trials49,50. Reliable tools for measuring outcomes (e.g. validated questionnaires in local language, well-trained assessors, effective training facilities, etc.) may not be available in low-resource settings reducing accuracy of assessing effectiveness51,52. We did not search for health economic assessments.

All our included studies reported positive outcomes, but the high RoB limits interpretation of this finding. In contrast, the evidence from studies conducted in HICs are mostly at low-to-moderate RoB, so that the Cochrane review was able to conclude confidently that PR was an effective intervention for people with COPD23. It is likely that insufficient resources, training, and facilities in LMICs is responsible for the lack of high-quality trials. This is a gap that NIHR-funded initiatives, such as RESPIRE53, and RECHARGE54 aim to address.

Compared to high-resource settings, under-diagnosis due to lack of awareness of CRD compounded by limited access to diagnostic tools such as spirometry results in a minority of potentially eligible participants being approached to be enrolled in studies. Poor universal health coverage55 and ‘catastrophic’ costs of healthcare56 further limit participation in trials.

The lack of diagnostics means that patients recruited as COPD may in fact have a range of undifferentiated CRDs (e.g. pulmonary impairment after tuberculosis or combined obstructive and restrictive disorder57). While this lack of detailed characterisation may impact on findings, offering PR to people with CRD (regardless of specific diagnosis) may be a more appropriate strategy especially in resource-limited settings.

There was considerable variation in the clinical status of participants, which might affect outcomes. There was considerable range in severity of functional limitation (see Table 1). In addition, some of the patients were stable at enrolment37,39,40,43,45,47 while some had been hospitalised for a recent exacerbation38,42,44.

Exercise training is the cornerstone of PR58 and was an inclusion criterion for the studies in our review. Endurance training was included in all the studies in addition to a range of other modalities as per recognised guidelines. Behavioural changes and continuing physical activities are crucial for maintaining effectiveness of PR59, but these were not reported in any of the studies.

Education on CRD and its treatment was widely provided along with strategies on managing breathlessness, but other components such as self-management support and addressing social care needs were rarely reported, despite evidence of effectiveness in CRDs60. In HICs, smoking is the predominant risk factor and cessation support is seen as essential. Surprisingly, only two of the studies in our review reported a smoking cessation component and none reported avoidance of pollution and indoor biomass exposure, which are also important risk factors in LMICs61,62. The brief descriptions in the papers make it difficult to assess how these and other important educational topics (such as inhaler technique) were addressed.

Models of PR delivery depends on who, where, to whom, and how the service is delivered63. Different models of PR services were described in the included studies reflecting diversity in the healthcare context and access to PR services; individuals’ health literacy; and background beliefs, attitudes, and preferences, as well as practical factors such as availability of transport and capability of payment64. A home-based, inexpensively equipped PR service with minimal attendance at a potentially distant centre may be more suitable model in rural areas with limited resources and poor transport infrastructure65,66. In home-based models, the cost to the patient is minimised, and people have flexibility in how they invest their time67,68,69. Digital technology is a rising paradigm in LMICs, which may be considered in developing a remote model of PR service70.

Our findings have implications for clinical practice and research. Breathlessness is the principal symptom that drives the patients with CRDs to seek medical help71. In LMICs, diagnosis of chronic respiratory symptoms depends on clinical history and physical examination, with limited, or sometimes no, access to spirometry or other investigations72. Poor healthcare coverage may mean that tasks regarded as prerequisites to referral in HICs, such as identifying co-morbidities, optimising pharmacotherapy, and exclusion of contraindications, may need to be a component of PR in LMICs73. The studies included in this review identified some practical solutions to these challenges, but high-quality evidence of the clinical and cost effectiveness of these pragmatic approaches is urgently needed.

In conclusion, recommendations in PR guidelines typically reflect services delivered in high-income settings. Our literature review, although identifying studies with high-to-moderate RoB, highlighted the feasibility of conducting PR in LMICs with positive effects on outcomes such as exercise tolerance, HRQoL, and symptoms improvement. Our findings point to the need for PR services that are effective across a broad range of (potentially poorly differentiated) CRDs, overcoming barriers of cost, distance, and access to healthcare such that they are deliverable and sustainable in low-resource settings with minimal equipment. Only then will the known benefits of PR be available to address the increasing burden of CRDs in LMICs.

Methods

Published review protocol

The review is registered with PROSPERO [ID: CRD42019125326]. The detailed systematic review protocol is published74 with salient points described here. We followed the procedures described in the Cochrane Handbook for Systematic Reviews of Interventions75.

Deviation from published protocol

We planned to use Grading of Recommendations Assessment Development and Evaluation (GRADE76) approach to rate the quality of evidence for primary outcomes and the important secondary outcomes; however, there was substantial missing information in the papers, so we were unable to apply the GRADE approach (see Supplementary Results 2 for our limited GRADE exercise).

Search strategy

Table 4 gives details of the search strategy developed to detect randomised controlled trials (RCTs) and controlled clinical trials of ‘Pulmonary Rehabilitation’ AND ‘COPD or other CRD’ AND ‘LMIC or low-resource settings’ from 1990 (when global COPD guidelines first recommended PR77) to November 2018 with no language restrictions. We searched MEDLINE (Supplementary Methods 1) EMBASE, Global Health (CABI), AMED, PubMed, and the Cochrane Database of Controlled Trials (CENTRAL). We did not undertake hand searching as we found no journal that regularly published PR papers in LMICs. Additionally, we conducted forward citations of the included articles. We used EndNote for overall data management.

Table 4 PICOS search strategy.
Full size table

The searches were completed on 28 October 2018, with a pre-publication update on 8 March 2020 using the ‘efficient and effective’ approach78 of forward citation using Google Scholar, of all included papers, and the Cochrane review23.

Selection process

Details of inclusion and exclusion criteria and definitions used are in Table 4. In summary, we undertook a duplicate selection process using rules for operationalising the inclusion/exclusion criteria (see protocol for details74). Two trained reviewers (G.M.M.H. and M.N.U.) independently screened titles and abstracts, then full-text papers (G.M.M.H., M.N.U., and K.D.). Disagreements were resolved by discussion, involving H.P. and R.R. or the wider team as necessary. We reported the process in a PRISMA flow diagram (Fig. 1)79.

Outcome measurement

Our primary outcomes were between-group difference in functional exercise capacity (e.g. 6-MWT80,81,82) and HRQoL (e.g. SGRQ83,84). We also included breathlessness (e.g. mMRC Dyspnoea score85). These are defined, and secondary outcomes are described in Table 4.

Data extraction and RoB

Two reviewers (G.M.M.H. and M.N.U. and checked by H.P.) extracted data on a piloted data extraction form (Supplementary Methods 2) based on the Cochrane Effective Practice and Organisation of Care guidance86; G.M.M.H. and M.N.U. (checked by H.P.) independently assessed the methodological quality of all the included studies according to the Cochrane RoB tool75.

Data analysis

The analysis addressed our three objectives:

  1. 1.

    Effectiveness of PR in low-resource settings: On the basis of our initial scoping, we anticipated that our included studies would have substantial clinical, methodological, and statistical heterogeneity, and meta-analysis would not be appropriate. We, therefore, conducted a narrative synthesis illustrating the key outcomes on a harvest plot87,88. In order to ensure transparency of interpretation, the decisions that underpinned the harvest plot are described in Table 1: column 5.

  2. 2.

    Components used in effective studies: We identified the components that are described in internationally recognised guidelines13,15,89 using categories from the American Thoracic Society/European Respiratory Society task force report13, British Thoracic Society guidelines for PR15, and Lung Foundation of Australia90. We then constructed a matrix with the components used in the (effective and ineffective) studies.

  3. 3.

    Models of care used in the PR interventions: We described the models of care used, including PR providers and (if specified) their training, venue and equipment available, number and frequency of training sessions, use of telehealth, and strategies for sustainability.