Abstract

Background In adults with serious respiratory illness, fatigue is prevalent and under-recognised, with few treatment options. The aim of this review was to assess the impact of graded exercise therapy (GET) on fatigue in adults with serious respiratory illness.

Methods Electronic databases were searched to identify randomised controlled trials (RCTs) testing GET (involving incremental increases in exercise from an established baseline) in adults with serious respiratory illness. The primary outcome was fatigue and secondary outcomes were health-related quality of life (HRQoL) and adverse events. Two authors independently screened for inclusion, evaluated risk of bias and extracted data.

Results 76 RCTs were included with 3309 participants, most with a diagnosis of COPD or asthma. Reductions in fatigue measured by the Chronic Respiratory Disease Questionnaire fatigue domain score were demonstrated following GET consisting of aerobic with/without resistance training (mean difference (MD) 0.53 points, 95% CI 0.41–0.65, 11 RCTs, 624 participants) and GET using resistance training alone (MD 0.58 points, 95% CI 0.21–0.96, two RCTs, 82 participants) compared with usual care. Although the mean effect exceeded the minimal important difference, the lower end of the confidence intervals did not always exceed this threshold so the clinical significance could not be confirmed. GET consistently improved HRQoL in people with a range of chronic respiratory diseases on multiple HRQoL measures. No serious adverse events related to GET were reported.

Conclusion GET may improve fatigue alongside consistent improvements in HRQoL in people with serious respiratory illness. These findings support the use of GET in the care of people with serious respiratory illness.

Shareable abstract

Graded exercise therapy may reduce fatigue and consistently improves quality of life in people with serious respiratory illness. These findings support the use of graded exercise therapy in a range of chronic respiratory diseases in clinical practice. https://bit.ly/3wKr1kt

Introduction

People with serious respiratory illness experience a high symptom burden, including fatigue, which is highly prevalent and has a profound impact on daily life participation and health-related quality of life (HRQoL) [1–5]. Fatigue is reported by up to 95% of people with COPD or interstitial lung disease (ILD) and 60% of people with asthma or pulmonary hypertension [1–4]. There is emerging evidence that fatigue is associated with poor long-term outcomes in people with chronic respiratory disease, including exacerbations and mortality [6]. There are few specific treatments for fatigue available and the symptom of fatigue is under-recognised by health professionals and under-reported by patients [7]. Patients may believe it is a natural consequence of their respiratory illness, such that its presence is not expressed [7]. People living with chronic respiratory disease report unmet needs for effective treatments to reduce fatigue [8, 9].

Graded exercise therapy (GET) is commonly employed in the care of people with serious respiratory illness, with components including aerobic (endurance) and resistance (strength) training. Previous systematic reviews have noted the low-quality [10] and limited evidence [11] for the effect of exercise training on fatigue in people with COPD, with fatigue indirectly considered through the assessment of HRQoL in other chronic respiratory conditions [12–15]. Successfully delivered in a wide range of healthcare and nonhealthcare settings globally, GET has the potential for widespread implementation if proven effective.

The aim of this review was to assess the impact of GET on fatigue in adults with serious respiratory illness. Our research question was “should graded exercise therapy be used to reduce fatigue in people with serious illness related to lung disease?”. The review was conducted as part of the evidence synthesis for the European Respiratory Society (ERS) clinical practice guideline on symptom management for adults with serious respiratory illness.

Methods

The review protocol was developed a priori but was not published, due to the confidentiality requirements of the ERS clinical practice guideline development process. Instead, the protocol was submitted to European Respiratory Review editorial office in April 2023 to be held in confidence and made available to reviewers. The protocol can be found in the supplementary material.

Search strategy and study selection

Initial searches were conducted between August 2022 and November 2022 in Medline (OVID), Embase (OVID), the Cochrane Database of Systematic Reviews and CENTRAL (The Cochrane Library) to identify relevant systematic reviews. Systematic reviews providing evidence for at least one of the outcomes of interest were used to identify relevant studies. Subsequently, searches were conducted in order to identify randomised controlled trials (RCTs) that had been published since the search date of the most recent relevant systematic review. Studies were included irrespective of date or language of publication. Reference lists of all primary studies and review articles were manually checked for additional references and we searched for errata or retractions from included studies published in full-text on PubMed. A preferred reporting items for systematic reviews and meta-analyses (PRISMA) diagram documented the review process [16].

Types of studies

We included RCTs. Randomised crossover trials were only included if pre-crossover data were available, since the intervention includes behavioural components and therefore carryover of intervention effects to the second period may occur.

Participants

Aged ≥18 years with serious respiratory illness, defined as a respiratory condition that carries a high risk of mortality, negatively impacts quality of life and daily function, and/or is burdensome in terms of symptoms, treatments or caregiver stress [17]. There was no restriction according to the underlying diagnosis of lung disease, except that patients with cancer were excluded. Patients recovering from coronavirus disease 2019 infection were not included.

Types of interventions

GET is defined as establishing a baseline of achievable exercise or physical activity and then making fixed incremental increases in the time spent being physically active [18]. Because we were specifically interested in the effects of GET on fatigue, we did not include exercise therapy conducted in the context of pulmonary rehabilitation, which is a broader package of interventions for people with chronic lung disease that aims to address broader outcomes. This allowed us to evaluate GET as an intervention that could be delivered outside of a pulmonary rehabilitation context. We defined pulmonary rehabilitation as “a comprehensive intervention based on a thorough patient assessment followed by patient tailored therapies that include, but are not limited to, exercise training, education, and behaviour change, designed to improve the physical and psychological condition of people with chronic respiratory disease and to promote the long-term adherence to health-enhancing behaviours” [19].

Types of comparisons

We included studies that reported 1) the effects of the intervention compared with no intervention or a sham intervention and 2) where the intervention of interest was added to an intervention common to both groups (e.g., GET added to pedometer-based intervention versus pedometer-based intervention alone). We reported these two comparisons separately.

Outcomes

The primary outcome was fatigue assessed using any validated tool. This included measures at rest or during exercise, but exercise measures obtained before and after an intervention must have been recorded at iso-workload. Secondary outcomes were HRQoL (assessed using any validated tool) and adverse events (defined according to the investigators’ definition). Studies were not excluded if they did not report these outcomes.

Identification of studies

Citations were uploaded into Covidence (www.covidence.org) where duplicates were removed before titles and abstracts were screened for eligibility by two independent reviewers (A.T. Burge and A.M. Gadowski) with conflicts resolved by discussion and adjudication by a third reviewer (A.E. Holland) if required. Studies classified as “include” or “unsure” were obtained as full text. Full-text review was conducted by the same two reviewers to determine eligibility for inclusion.

Quality assessment

Risk of bias was assessed by two independent reviewers (A.T. Burge and A.M. Gadowski) with conflicts resolved by discussion and adjudication by a third reviewer (A.E. Holland) if required. Where included studies were identified from systematic reviews, we retrieved the published assessments of risk of bias. Methodological quality of the systematic reviews was assessed using the Assessing the Methodological Quality of Systematic Reviews (AMSTAR-II) checklist [20]. For individual studies in the updated search, risk of bias was assessed using the Cochrane Risk of Bias tool 1 [21].

Data extraction

Data extraction was undertaken by two independent reviewers (A.T. Burge and A.M. Gadowski) with conflicts resolved by discussion and adjudication by a third reviewer (A.E. Holland) if required. Custom-designed data collection forms were used to record study details, participant inclusion and exclusion criteria, demographic characteristics, intervention details (type, duration, frequency and dose), care received by the comparison group, and outcome data of interest.

Data analysis and synthesis

We analysed dichotomous data as odds ratios and continuous data using mean differences (MDs). Where available, we used the change from baseline, otherwise the adjusted results or final score were used. Skewed data were described using medians and interquartile ranges (IQRs). Data were presented as a scale with a consistent direction of effect (e.g. HRQoL data). Attempts were made to contact study authors to clarify key study characteristics and obtain missing outcome data where possible.

Where trials were clinically heterogeneous, a narrative synthesis was performed. Meta-analyses (random-effects model) were undertaken only if interventions and/or participant features were clinically homogeneous. Where multiple arms were reported in a single trial, we included only the relevant arms. If two comparisons were combined in the same meta-analysis, we halved the control group to avoid double-counting. We used the I2 statistic to measure heterogeneity. We performed sensitivity analyses to examine the effects of methodological quality on the pooled estimate by removing studies that were at high or unclear risk of bias for the domains of blinding of outcome assessors and incomplete outcome data. Statistical analysis was conducted using RevMan version 5.4 (Cochrane Collaboration).

Results

The search for systematic reviews identified 1567 records, of which 38 were screened in full text (figure 1). 12 relevant systematic reviews were identified (table S1) containing 74 eligible RCTs. The search for additional RCTs identified 6750 records, of which 203 were screened in full text, with an additional two eligible RCTs identified.

FIGURE 1

Preferred reporting items for systematic reviews and meta-analyses flow of studies through the systematic review.

We included 76 RCTs (3309 participants) including participants with COPD (41 RCTs), asthma (10 RCTs), ILD (seven RCTs), pulmonary hypertension (seven RCTs), cystic fibrosis (CF) (four RCTs), bronchiectasis (two RCTs) and other chronic respiratory conditions (five RCTs) (tables S2 and S3). The average number of participants in each RCT was 44 (range 10–145). Most participants had moderate-to-severe chronic lung disease. One study assessed an intervention following hospital admission for an exacerbation [22]. Studies were undertaken in 25 countries, most commonly Australia (12 studies), Brazil (10 studies), the United Kingdom (eight studies), Turkey (seven studies), Spain and the United States (five studies each).

The most common intervention was GET that included aerobic training with/without resistance training (71 studies), followed by GET that included resistance training alone (eight studies). Two studies had two intervention groups that were combined to compare with usual care [23, 24] and five studies had two intervention groups compared with usual care [25–29]. One study compared GET to sham/placebo [30]. Four studies added GET or resistance training to another intervention common to both groups (pedometer [31]; placebo/sham [32–35]). Most interventions were 4–12 weeks in duration, but a few were as long as 12 months (table S2). To ensure clinical homogeneity of interventions, only studies up to 12 weeks in duration were combined in meta-analysis.

Quality assessment using the AMSTAR-II checklist demonstrated that systematic reviews scored well for formulation of PICO (patient, intervention, comparison, outcome) questions, comprehensive searches, a priori registration of a review protocol and assessment of risk of bias (table S1). However, few reported sources of funding for included studies (three of 12 reviews) or explained the selection of included study designs (four of 12). Risk of bias was low or unclear for the majority of RCTs in the domains of random sequence generation, allocation concealment, blinding of outcome assessment, incomplete outcome data and selective reporting (figures 2 and S1). As expected, due to the nature of the intervention, few trials blinded participants or personnel. The overall certainty of evidence was low.

FIGURE 2

Risk of bias graph: judgements about each risk of bias item presented as percentages across all included studies.

Primary outcome: fatigueComparison: GET (aerobic with/without resistance training) versus usual care

GET reduced fatigue as measured by the Chronic Respiratory Disease Questionnaire (CRQ) fatigue domain score (MD 0.53 points, 95% CI 0.41–0.65, 11 RCTs, 624 participants, I2=0%) (figure 3a). The mean effect exceeded the minimal important difference (MID) (0.5 points) [44, 45]; however, the lower end of the confidence interval did not reach the MID and clinical significance is unclear. Sensitivity analysis reduced the mean effect (MD 0.46 points, 95% CI 0.31–0.61, four RCTs, 274 participants, I2=0%) (table S7). Effects were similar when participants with COPD and ILD were considered separately (figures 3b and c). Improvements in fatigue were demonstrated using a range of other measures that could not be combined for meta-analysis (table S4) [25, 43, 46–49].

FIGURE 3

Forest plot: fatigue (Chronic Respiratory Disease Questionnaire (CRQ) fatigue domain score) for graded exercise therapy (GET) versus usual care at end of intervention (8–12 weeks). a) CRQ fatigue domain score. b) CRQ fatigue domain score in participants with COPD. c) CRQ fatigue domain score in participants with interstitial lung disease. IV: inverse variance.

For studies that included longer-term follow-up, improvements in CRQ fatigue domain score were maintained at 6 months in participants with ILD (MD 0.56 points, 95% CI 0.07–1.04, three RCTs, 209 participants, I2=34%) (figure S2) but not maintained at 12-month follow-up in participants with bronchiectasis (MD 0.12 points, 95% CI −0.45–0.70, one RCT, 55 participants) (table S4) [43].

Following hospitalisation for a COPD-related exacerbation, 6 months of GET improved CRQ fatigue domain score at the end of the intervention (MD 1.9 points, 95% CI 1.0–2.8), 12-month follow-up (MD 2.2 points, 95% CI 1.3–3.1) and 18-month follow-up (MD 2.7 points, 95% CI 1.9–3.6) (table S4) [22]. The mean effect and confidence interval exceeded the MID, but dropout was substantial.

Comparison: GET (tai chi) versus usual care

Tai chi in participants with COPD improved rating of perceived exertion (RPE) at isotime during an endurance shuttle walk test (MD −1, 95% CI −3–−0.2, one RCT, 38 participants) and CRQ fatigue domain score (MD 0.5 points, 95% CI 0.1–1.0, one RCT, 38 participants) (table S4) [50]. The mean effect met the MID, but the lower end of the confidence interval included the MID and clinical significance is unclear.

Comparison: GET (resistance training) versus usual care

Resistance training in participants with COPD improved CRQ fatigue domain score (MD 0.58 points, 95% CI 0.21–0.96, two RCTs, 76 participants, I2=0%) (figure S3). The mean effect exceeded the MID, but the lower end of the confidence interval did not reach the MID and clinical significance is unclear. Improvements were not maintained at 6-month follow-up (table S4) [51]. Resistance training did not report improvement in RPE at isotime on a constant work rate test (table S4) [52].

Secondary outcome: HRQoLComparison: GET versus usual care

GET improved HRQoL measured by St George's Respiratory Questionnaire (SGRQ) total score (MD −14.07 points, 95% CI −18.85–−9.30, 15 RCTs, 627 participants, I2=89%) (figure 4a). Improvements were also demonstrated in SGRQ symptoms (MD −17.00 points, 95% CI −23.44–−10.56, 14 RCTs, 598 participants, I2=82%) (figure 4b), activity (MD −14.48 points, 95% CI −21.48–−7.47, 14 RCTs, 598 participants, I2=91%) (figure 4c) and impact domain scores (MD −15.53 points, 95% CI −20.74–−10.32, 14 RCTs, 598 participants, I2=78%) (figure 4d). For all estimates, the mean effect and confidence interval exceeded the MID (four points) [5], but heterogeneity was high (I2=78–91%). Sensitivity analyses did not substantially alter the size of the effect and high heterogeneity persisted (table S7). Effects were similar when participants with COPD and ILD were considered separately (figures S4 and S5). Variable results in participants with COPD were reported in studies that could not be combined in meta-analysis (table S5) [62, 63].

FIGURE 4

Forest plot: health-related quality of life (St George's Respiratory Questionnaire (SGRQ) total, symptoms, activity and impact domain scores) for graded exercise therapy (GET) versus usual care at end of intervention (6–12 weeks). a) SGRQ total score. b) SGRQ symptoms domain score. c) SGRQ activity domain score. d) SGRQ impact domain score. IV: inverse variance.

At 6–11-month follow-up, participants with ILD demonstrated maintained improvement in SGRQ symptoms domain (MD −11.50 points, 95% CI −18.57–−4.43, three RCTs, 180 participants, I2=9%) (figure S6.2) where both the mean effect and confidence interval still exceeded the MID, indicating clinical significance. The mean effect of SGRQ impact domain score exceeded the MID at follow-up but did not reach statistical significance (MD −4.81 points, 95% CI −11.88–2.26, three RCTs, 180 participants, I2=68%) (figure S6.4). The mean effects for SGRQ total (MD −3.83 points, 95% CI −8.12–0.47, three RCTs, 180 participants, I2=29%) (figure S6.1) and activities domain scores at follow-up (MD −1.25 points, 95% CI −2.83–0.33, three RCTs, 180 participants, I2=0%) (figure S6.3) did not meet the MID and clinical significance is unclear.

A number of other HRQoL measures were reported in smaller numbers of studies, with a similar pattern of results (see extended results in the supplementary material).

Comparison: GET (tai chi) versus usual care

The mean effect and confidence interval for CRQ dyspnoea domain score (MD 0.9 points, 95% CI 0.5–1.3, one RCT, 38 participants) exceeded the MID, indicating clinical significance. The mean effect for CRQ total (MD 0.7 points, 95% CI 0.3–1.0, one RCT, 38 participants) and mastery domain scores (MD 0.5 points, 95% CI 0.1–0.9, one RCT, 38 participants) exceeded/met the MID, but the lower end of the confidence interval included the MID and clinical significance is unclear. The mean effect for CRQ emotional function domain score (MD 0.4 points, 95% CI 0.02–0.8, one RCT, 38 participants) did not meet the MID and clinical significance is unclear (table S5) [50].

Comparison: GET (resistance training) versus usual care

The mean effect for CRQ dyspnoea domain score (MD 0.71 points, 95% CI 0.23–1.20, two RCTs, 76 participants, I2=0%) (figure S15.1) exceeded the MID, but the lower end of the confidence interval did not reach the MID and clinical significance is unclear. The mean effects for CRQ emotional function (MD 0.33 points, 95% CI −0.01–0.68, two RCTs, 76 participants, I2=11%) (figure S15.2) and mastery domain scores (MD 0.39 points, 95% CI −0.27–1.05, two RCTs, 76 participants, I2=64%) (figure S15.3) did not meet the MID and did not reach statistical significance. At 6-month follow-up, the mean effect for CRQ dyspnoea (MD 0.3 points, 95% CI −0.2–0.8, one RCT, 54 participants), emotional function (MD 0 points, 95% CI −0.5–0.4, one RCT, 54 participants) and mastery domain scores (MD −0.1 points, 95% CI −0.6–0.4, one RCT, 54 participants) did not meet the MID and did not reach statistical significance (table S5) [51].

Comparison: GET with pedometer versus pedometer alone

After GET with a pedometer compared with a pedometer alone, the mean effect for SGRQ total (MD −5.5 points, 95% CI −8.4–−2.6, one RCT, 33 participants), symptoms (MD −7.8 points, 95% CI −14.5–−1.1, one RCT, 33 participants) and impacts domain scores (MD −7.3 points, 95% CI −11.1–−3.4, one RCT, 33 participants) exceeded the MID, but the upper end of the confidence interval did not reach the MID and clinical significance is unclear. The mean effect of SGRQ activity domain score (MD −3.9 points, 95% CI −8.9–1.2, one RCT, 33 participants) did not meet the MID (table S5) [31]. At follow-up, the mean effect for SGRQ total and impacts domain scores exceeded the MID. The mean effect for SGRQ symptoms domain score exceeded the MID but the upper end of the confidence interval did not reach the MID and clinical significance is unclear. The mean effect for SGRQ activity domain score did not meet the MID at 3 months but did meet the MID at 12 months. However, neither effect reached statistical significance (table S5) [31].

Comparison: GET with sham versus sham only

In a study in participants with asthma that could not be meta-analysed, GET with sham improved Asthma Quality of Life Questionnaire (AQLQ) total (MD 0.9 points, 95% CI −0.1–1.7, one RCT, 43 participants) and activity limitation domain scores (MD 1.1 points, 95% CI −0.3–1.8, one RCT, 43 participants) with the mean effect exceeding the MID; however, the lower end of the confidence interval included the MID and clinical significance is unclear (table S7) [33]. Mean effects for the remaining AQLQ domains met the MID but statistical significance was not reached (symptoms MD 0.7 points, 95% CI −0.1–1.5; emotion MD 0.9 points, 95% CI −0.1–2.0; environmental stimuli MD 0.9 points, 95% CI −0.3–2.0, one RCT, 43 participants) (table S5) [33]. Another reported greater median improvements in AQLQ domain scores with only the activity limitation domain score reaching statistical significance (table S5) [35].

Secondary outcome: adverse events

46 RCTs (2030 participants) reported on this outcome.

Comparison: GET versus usual care

Five studies reported no adverse events related to GET in 181 participants with COPD [26, 32, 43, 50, 52]. In one study, eight of the 22 participants with COPD in the GET (high-intensity single-limb training) group reported 28 minor adverse events (64% musculoskeletal, 32% related to the elastic resistance bands, dizziness 4%) [64]. There was no difference in the number of participants with COPD who developed an exacerbation (OR 0.96, 95% CI 0.47–1.96, 12 RCTs, 425 participants, I2=0%) (figure 5a) during the intervention period; sensitivity analyses demonstrated a larger odds ratio for exacerbations but wider confidence intervals (OR 2.02, 95% CI 0.44–9.20, three RCTs, 118 participants, I2=0%) (table S7). There was no difference in the number of participants with COPD who were hospitalised (OR 1.62, 95% CI 0.36–7.28, three RCTs, 234 participants, I2=0%) (figure S16.1) or died during the intervention period (OR 0.33, 95% CI 0.10–1.05, four RCTs, 180 participants, I2=0%) (figure S16.2) (sensitivity analysis not possible). A number of other measures of adverse events similarly showed no difference between groups (see extended results in the supplementary material).

FIGURE 5

Forest plot: number of participants with exacerbations during the intervention period for graded exercise therapy (GET) versus usual care. a) Participants with COPD. b) Participants with asthma. c) Participants with interstitial lung disease. M-H: Mantel–Haenszel.

Two studies reported no adverse events related to GET in 48 participants with asthma [68, 69]. There was no difference in the number of participants with asthma who developed an exacerbation (OR 2.49, 95% CI 0.51–12.10, two RCTs, 74 participants, I2=0%) (figure S5.2). One study reported one hospitalisation in 40 participants during the intervention period (table S6) [67].

Four studies reported no adverse events related to GET in 126 participants with ILD [36, 38, 41, 61]. There was no difference in the number of participants with ILD who developed an exacerbation (OR 4.65, 95% CI 0.76–28.52, three RCTs, 233 participants, I2=0%) (figure 5c) or number of deaths (OR 0.22, 95% CI 0.02–2.01, two RCTs, 199 participants, I2=0%) (figure S16.3) during the intervention period (sensitivity analysis not possible). A number of other measures of adverse events similarly showed no difference between groups (see extended results in the supplementary material).

Four studies reported no serious adverse events related to GET in 95 participants with pulmonary hypertension [48, 70–73]. One study reported one of five participants with pulmonary hypertension in the GET group had to stop exercise training for a single session due to extreme light-headedness [74] and another study reported one death in 39 participants during the intervention period (table S6) [70].

One study reported no adverse events related to GET in 27 participants with bronchiectasis and one of 55 participants with exacerbations during the intervention period (table S6) [49]. There was a reduction in the number of participants with exacerbations (OR 0.27, 95% CI 0.09–0.78, two RCTs, 110 participants, I2=0%) (figure S18) but no difference in the number of deaths (OR 0.27, 95% CI 0.01–6.87, one RCT, 55 participants) (table S6) [41] during the 6–12-month follow-up period. There was a longer time to first exacerbation in the GET group (log rank 0.49, 95% CI 0.01–0.97, one RCT, 55 participants) (table S6) [41] over a 12-month follow-up period.

One study reported no adverse events related to GET in eight participants with CF [75]. One study reported mild post-exercise muscle soreness (visual analogue scale 0–100) on a single occasion in four of seven participants during the intervention period (median 8 mm, IQR 5–8) [76]. One study reported two of 43 participants with exacerbations during the intervention period (table S6) [27].

Comparison: GET (resistance training) versus usual care

One study reported one episode of minor muscle soreness related to resistance training in 27 participants with COPD [51].

Comparison: GET (resistance training) with breathing exercises versus breathing exercises alone

One study reported no adverse events related to resistance training with breathing exercises in 21 participants with COPD and one participant with an exacerbation during the intervention period (42 participants) (table S6) [77].

Comparison: GET with sham versus sham alone

In participants with asthma, one study reported fewer participants with exacerbations (53% versus 20%, p=0.03, one RCT, 51 participants) (table S6) [35] and one study reported fewer exacerbations per participant in the GET with sham group (0.6 versus 1.5, p=0.021, one RCT, 43 participants) (table S6) [34].

Comparison: GET with pedometer versus pedometer alone

One study reported fewer exacerbations of COPD in the 12 months following GET with a pedometer relative to the previous 12 months (MD −0.03 exacerbations, 95% CI −1.3–0.7, one RCT, 33 participants) (table S6) [31] compared with a pedometer alone.

Discussion

This systematic review demonstrates that GET, compared with usual care, improves fatigue, based on trials using a variety of measures in people with serious respiratory illness. However, there is uncertainty regarding the magnitude of the effect and clinical significance. Additionally, clinically important improvements in HRQoL were demonstrated following GET. There were few reports of minor adverse events associated with the delivery of high-intensity GET [64, 76] or resistance training interventions [51] with the exception of a single event in a study of five participants with pulmonary hypertension [74]. Limited data demonstrated reductions in exacerbations following GET which warrant further exploration, along with maintenance of benefits for fatigue and HRQoL, where the limited evidence is restricted to people with COPD, ILD or bronchiectasis.

Implications

Based on the evidence presented in this systematic review, the recent ERS guideline on symptom management has made a conditional recommendation that GET be utilised to reduce fatigue in people with serious respiratory illness. This recommendation places a high value on consistent improvements in fatigue and HRQoL for people who undertook GET. GET is an acceptable intervention to address fatigue in people living with serious respiratory illness [8] and this review demonstrated a low likelihood of undesirable effects, with the caveat that included studies utilised trained staff to deliver supervised GET to participants. Delivery of GET does not require specialised equipment and is already a component of pulmonary rehabilitation programs. Whilst well established in many countries, there are well-documented patient-related barriers to uptake [7, 8] and disparities in access to pulmonary rehabilitation that are likely to be relevant to the feasibility of delivering GET [6]. GET is within the scope of practice for many health professionals, including physiotherapists, physical therapists and exercise physiologists, but certain cohorts (e.g., patients with severe pulmonary hypertension, history of arrhythmia, syncope or pre-syncope during exercise) may require additional monitoring from health professionals with relevant expertise in a clinical setting.

Our conclusions were limited by the low certainty of evidence. Most commonly, studies did not incorporate participant and personnel blinding, intention-to-treat analysis and prospectively registered protocols, thus affecting assessment of risks of detection, attrition and reporting bias. Despite substantial evolution of clinical trial conduct and reporting since the earliest included study was published in 1977 [78], high risk of bias was also evident in more recent studies. There was considerable variation in the intervention components (e.g., frequency, intensity, progression and duration), which may have influenced the magnitude and maintenance of the observed effects. Future trial reporting for RCTs of nonpharmacological interventions should employ the TIDieR (template for intervention description and replication) checklist [79] to report intervention components and adhere to the CONSORT statement [80].

Although current data support the implementation of GET in clinical practice, there are knowledge gaps that should be addressed in future research. There are limited data examining the use of GET modalities outside of aerobic and resistance training, and future studies should evaluate other modalities that may be preferred by some patients, such as water-based exercise or tai chi. One small included study targeted participants following a hospital admission for COPD exacerbation, with promising results [23]. The role of GET in people who are towards the end of life, when fatigue may be particularly problematic, has not been explored. Similarly, it is not clear whether the tested interventions would be useful in people with severe fatigue or post-exertional malaise. Future studies should use fatigue-specific outcome measures.

Strengths and limitations

Strengths of this review are the inclusion of RCTs from 25 countries, use of two independent reviewers to screen citations and extract data, and the application of a rigorous definition of GET. Whilst our strategy of using existing systematic reviews to identify earlier trials increased research efficiency, it is possible that some earlier trials could have been missed using this strategy, although the use of 12 systematic reviews (table S1) reduces this risk. Some studies were excluded (figure 1) because they contained insufficient detail to confirm eligibility, which may have excluded relevant data. We only included RCTs, which means other types of evidence did not contribute to our results. There were few RCTs including participants with CF or bronchiectasis. The lack of fatigue-specific outcome measures may have impacted on our estimates of effect and the heterogeneity of outcomes and timepoints of measurement may also reduce certainty in our estimates.

Conclusion

In conclusion, GET may improve fatigue in people with serious respiratory illness, alongside consistent improvements in HRQoL. Evidence is strongest for GET including aerobic training with/without resistance training in people with COPD, asthma, pulmonary hypertension and ILD, with limited data in CF and bronchiectasis. These findings support the safe use of GET to reduce symptoms in people with serious respiratory illness.

Points for clinical practice

• GET (aerobic with/without resistance training) may be useful to improve fatigue and enhance quality of life in people with serious respiratory illness.

• GET can be implemented by a variety of health professionals in many settings; however, certain cohorts (e.g., patients with severe pulmonary hypertension, history of arrhythmia, syncope or pre-syncope during exercise) may require additional monitoring from health professionals with relevant expertise in a clinical setting.

Questions for future research

• Do the benefits of GET persist over time in people with serious respiratory illness?

• What are the effects of GET in people with serious respiratory illness and severe fatigue?

• Can GET improve fatigue in people with serious respiratory illness towards the end of life?

Supplementary materialSupplementary Material

Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.

Systematic review protocol and search strategies ERR-0027-2024.SUPPLEMENT

Extended results, supplementary figures and tables ERR-0027-2024.SUPPLEMENT

Acknowledgements

We acknowledge the contribution of Jeanette Boyd from the European Lung Foundation, and extend our sincere thanks to the consumer partners who participated in the ERS Symptom Management Guideline Task Force – Phil Collis (CPROR Birmingham University, UK; Patient Advisory Group, European Lung Foundation, Sheffield, UK), Tessa Jelen (Patient Advisory Group, European Lung Foundation, Sheffield, UK), John Solheim (EU-PFF – European Pulmonary Fibrosis Federation, Overijse, Belgium; LHL-IPF, Jessheim, Norway) and Chantal Vandendungen (EU-PFF – European Pulmonary Fibrosis Federation, Overijse, Belgium; ABFFP – Association Belge Francophone contre la Fibrose Pulmonaire, Rebecq, Belgium).

Footnotes

Provenance: Commissioned article, peer reviewed.

Previous articles in this series: No. 1: Smallwood NE, Pascoe A, Wijsenbeek M, et al. Opioids for the palliation of symptoms in people with serious respiratory illness: a systematic review and meta-analysis. Eur Respir Rev 2024; 33: 230265.

Number 2 in the Series “Symptom management for advanced lung disease” Edited by Anne E. Holland, Magnus Ekström and Natasha E. Smallwood

Conflicts of interest: L. Romero declares funding from the European Respiratory Society to design search strategies for this review. A.E. Holland declares authorship on four of the systematic reviews included in this study. A.E. Holland declares no other conflicts of interest. All other authors have nothing to disclose.

Support statement: The European Respiratory Society funded a medical librarian (L. Romero) to design the search strategies for this review. Funding information for this article has been deposited with the Crossref Funder Registry.

Received February 15, 2024.Accepted May 15, 2024.Copyright ©The authors 2024