Article Text

Understanding the effectiveness of different exercise training programme designs on V̇O2peak in COPD: a component network meta-analysis
  1. Thomas JC Ward1,2,
  2. Charles D Plumptre3,
  3. Alessandra V Fraser-Pye4,
  4. Thomas E Dolmage5,
  5. Amy V Jones6,
  6. Ruth Trethewey6,
  7. Lorna Latimer1,2,
  8. Sally J Singh1,2,
  9. Martin R Lindley6,7,
  10. Michael C Steiner1,2,
  11. Rachael A Evans1,2
  1. 1 Respiratory Sciences, University of Leicester, Leicester, Leicestershire, UK
  2. 2 Centre for Exercise & Rehabilitation Science, NIHR Leicester Biomedical Research Centre, Leicester, UK
  3. 3 North Bristol NHS Trust, Westbury on Trym, Bristol, UK
  4. 4 Department of Medicine, University of Leicester, Leicester, UK
  5. 5 West Park Healthcare Centre, Toronto, Ontario, Canada
  6. 6 School of Sport, Exercise and Health Science, Loughborough University, Loughborough, UK
  7. 7 Institute of Clinical Sciences, University of Birmingham, Birmingham, UK
  1. Correspondence to Dr Thomas JC Ward, Respiratory Sciences, University of Leicester, Leicester, UK; tom.ward{at}


Pulmonary rehabilitation programmes including aerobic training improve cardiorespiratory fitness in patients with COPD, but the optimal programme design is unclear. We used random effects additive component network meta-analysis to investigate the relative effectiveness of different programme components on fitness measured by V̇O2peak in COPD. The included 59 studies involving 2191 participants demonstrated that V̇O2peak increased after aerobic training of at least moderate intensity with the greatest improvement seen following high intensity training. Lower limb aerobic training (SMD 0.56 95% CI 0.32;0.81, intervention arms=86) and the addition of non-invasive ventilation (SMD 0.55 95% CI 0.04;1.06, intervention arms=4) appeared to offer additional benefit but there was limited evidence for effectiveness of other exercise and non-exercise components.

  • Pulmonary Rehabilitation
  • Exercise
  • COPD Pathology

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Pulmonary rehabilitation (PR) is an accepted treatment for COPD demonstrating improvements in peak oxygen uptake (V̇O2peak), the gold standard measure of cardiorespiratory fitness.1 However, PR is a complex intervention with significant variation in design and delivery across settings and the optimal programme design remains uncertain.2 Furthermore, there is substantial heterogeneity in response to PR in both clinical trials3 and real world clinical practice4 which is only partially explained by differences in disease severity.1 It is possible that some of the heterogeneity in response could be explained by variation in PR design.

Previous pairwise meta-analyses have examined individual elements of PR in isolation including the volume and modality of training as well as the addition of non-exercise components.5–7 While this approach may give useful information, a pairwise meta-analysis can only include homogeneous trials in which a single component of PR programme design varies between treatment arms, and therefore ignores the wider complexity of PR. Component network meta-analysis (CNMA) is a novel statistical technique which allows inclusion of differing intervention designs in a single analysis to increase the power to detect differences in the effectiveness of individual components.8 CNMA assumes the effects of individual components within complex interventions are additive and therefore allows estimation of the individual effect of each component in isolation.

Our previous systematic review demonstrated a moderate improvement in V̇O2peak for people with COPD following a programme that included aerobic exercise training.1 We performed a secondary analysis on this data using CNMA to investigate the effect of specific exercise and non-exercise components on improvements in V̇O2peak.


This is an extended analysis of data from a previous systematic review, full methods of which are detailed elsewhere1 including prospective registration on PROSPERO (CRD42018099300). A comprehensive database search was performed based on ‘COPD’, ‘exercise training’ and ‘V̇O2peak’ from inception to April 2018. The eligibility criteria for inclusion were a diagnosis of COPD in≥90% of the population, an intervention group performing directly supervised lower limb aerobic training, at a minimum frequency of once a week for at least 3 weeks and a direct measure of V̇O2peak using any exercise modality. The current analysis was limited to both controlled trials comparing an intervention incorporating supervised lower limb aerobic training with no training, and controlled trials comparing different interventions for which supervised lower limb aerobic training was a component.

A frequentist random effects additive CNMA was performed using the netmeta package in R (version 4.1.1). We created two models: Model 1, a model with traditional lower limb aerobic training modalities combined in a single component to investigate the effects of non-aerobic exercise components and Model 2, a split model comparing different lower limb aerobic training components.

We performed multiple sensitivity analyses by excluding studies at high risk of bias and studies of low quality, and with imputation of data using a conservative estimate of correlation coefficient (0.5). To examine inconsistency, unconnected studies were excluded to create connected networks and results from net heat plots9 and node splitting were assessed.


Fifty-nine studies involving 2191 participants were included (eTable 1). These included 15 trials of aerobic training vs “usual care” or structured education in isolation, 37 controlled trials in which both arms performed aerobic training in different forms or with different “add-ons”, and seven 3-arm studies with three interventions or two interventions vs “usual care”. Of these, 50 studies were included in Model 1 (eFigure 1) and 30 studies were included in Model 2 (figure 1).

Figure 1

Network graph showing comparisons made in studies included in the component network metaanalysis for type of aerobic training (model 2). The size of each node is proportional to the number of studies including that intervention and the thickness of the lines is proportional to the number of times each comparison was made. Embedded Image: interval training, Embedded Image: continuous training, Embedded Image: cycling, Embedded Image: walking, High: high intensity, Mod: Moderate intensity, UC: Usual care.

Lower limb aerobic training, non-invasive ventilation (NIV) during exercise and administration of ghrelin alongside exercise training resulted in significant improvements in V̇O2peak (figure 2). No other components provided additional benefit when performed alongside lower limb aerobic training. Both moderate to high intensity continuous cycling and walking training modalities, and high intensity interval cycling and walking resulted in improvements in V̇O2peak (all p<0.05, figure 3). Addition of stair climbing (SMD −0.79 95% CI −1.28 to −0.31 k=4) and low intensity walking (SMD −0.43 95% CI −0.83 to −0.03 k=2) appeared to reduce improvements in V̇O2peak. While high intensity training modalities resulted in the greatest increase in V̇O2peak, the differences compared with moderate intensity training were not significant (p>0.05 for all pairwise comparisons). Results of multiple sensitivity analyses, bias assessment and validity assessment can be found in the supplement (eTable 2 and eFigures 2–3).

Figure 2

Forest plot showing the individual effect of programme components on oxygen uptake at peak exercise within an additive component network meta-analysis (model 1). The results represent the unique contribution of each individual component when performed as part of a multi-component intervention. All intervention groups included supervised lower limb aerobic training. Reference group is “usual care”.

Figure 3

Forest plot showing the individual effect of different components of the aerobic training aspect of the intervention on oxygen uptake at peak exercise within an additive component network meta-analysis (model 2). The results represent the unique contribution of each individual component when performed as part of a multi-component intervention. Reference group is “usual care”.


This study used novel methodology to investigate the effect of different potential components of PR on change in V̇O2peak for people with COPD, demonstrating robust evidence that lower limb aerobic training is the core component for improvements in maximal exercise capacity. Although most moderate to high intensity aerobic exercise components resulted in improvement in V̇O2peak, the largest numeric improvement was seen with high intensity training modalities. The addition of NIV to aerobic exercise appeared to enhance outcomes but there was limited evidence for the effectiveness of other components in the included trials.

Previous evidence synthesis assessing comparative effectiveness of PR components has been limited to analysis of trials which performed the same head-to-head comparison of different programme designs.6 10 Our study takes a different approach, allowing multiple programme designs to be compared in a single analysis, providing novel insights into the effectiveness of individual components of complex interventions such as PR. In contrast to previous analyses,11 our results suggest that high intensity exercise training may be more effective than moderate intensity training for people with COPD. The similarity between interval and continuous training is similar to previous work.12

This analysis had several limitations. Only one outcome was investigated and repeating this analysis with a wider range of outcome measures might allow conclusions to be made on the effectiveness of PR on other patient related outcomes. We excluded studies in which the intervention arm performed exercise for less than six sessions in 3 weeks or exercise that was not considered to be aerobically demanding, and we cannot therefore comment on the effect of very low intensity exercise or resistance training in isolation. Our analysis investigated the effects of components and did not consider confounding from factors such as programme duration and frequency, or disease severity. Participants in most studies were not blinded to their intervention introducing the possibility of performance bias however high intensity continuous aerobic training remained effective in those studies at low risk of performance bias. Finally, some of the components included were investigated in a small number of studies and we have therefore avoided making definitive conclusions for components with fewer than three comparisons including the potentially promising effect of ghrelin supplementation alongside exercise training.

In conclusion, lower limb aerobic training is the core effective component of PR to improve V̇O2peak and there is evidence that NIV during exercise may enhance outcomes. We found limited evidence of the effectiveness of other programme components for additional gains in cardiorespiratory fitness. These results demonstrate the significant heterogeneity that exists in the current delivery of PR programmes and highlights the need for a standardised approach. Although high intensity training may be challenging for patients, our data supports this as the optimal target, and this should be considered in future PR guidelines and standards. Further research is needed to understand the impact of PR components on other patient reported outcomes and to understand the interaction of PR components, training dose and patient demographics.

Supplemental material

Ethics statements

Patient consent for publication

Ethics approval

Not applicable.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Twitter @tom_jc_ward, @REvans_Breathe

  • Contributors TJCW, MCS and RAE developed the concept with input from TED. TJCW and CDP screened abstracts. TJCW, CDP, TED, LL, SJS, MRL, MCS, RF, AVJ and RAE screened full papers. TJCW and RAE resolved conflicts. AVJ, RT, AVF and TJCW assessed bias. TJCW and LL extracted data. TJCW performed the analysis and drafted the manuscript. All authors revised the final manuscript.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Dr Ward is funded by the NIHR [Academic Clinical Lectureship CL-2020-11-004]. The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

Linked Articles