Effect of ventilatory support pressure during exercise on endurance time in patients with COPD: a systematic review

Abstract

Rationale

Noninvasive ventilation during exercise has been proposed to improve exercise tolerance in patients with COPD. However, the influence of fixed pressure support intensity on endurance performance and exercise-related physiological responses remains insufficiently clarified.

Materials and methods

A systematic review and meta-analysis were conducted according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (PROSPERO CRD420251018584). Randomised controlled and crossover trials evaluating noninvasive ventilation delivered in fixed pressure support mode during exercise were included. Random-effects models were applied to pool standardised mean differences (SMD). Mixed-effects meta-regression analyses assessed the association between pressure support level and outcomes. The primary outcome was endurance time. Secondary outcomes included dyspnoea (modified Borg scale) and hypercapnia.

Results

A total of 15 studies (285 participants) were included. Noninvasive ventilation significantly improved endurance time compared with control conditions (SMD 0.47, 95% CI 0.20 to 0.73). Pressure support intensity was not significantly associated with endurance time (β=0.0345 per cmH₂O, p=0.25). Dyspnoea was significantly reduced in pooled analysis (SMD −0.50, 95% CI −0.92 to −0.08) and showed a significant dose–response relationship with pressure support (β= −0.0880 per cmH₂O, p=0.019). Carbon dioxide tension was significantly reduced with noninvasive ventilation (SMD −0.48, 95% CI −0.92 to −0.04, I²=68.2%), with a pressure-dependent reduction observed when post-exercise measurements were excluded (β= −0.1055 per cmH₂O, p=0.033).

Conclusion

Noninvasive ventilation during exercise improves endurance and dyspnoea in COPD. Pressure support intensity influences ventilatory unloading, but endurance performance appears multifactorial. Further standardised studies are needed to optimise pressure titration during exercise.

Shareable abstract

Noninvasive ventilation during exercise improves endurance and reduces dyspnoea (showing a significant dose–response relationship) and P_CO2 in COPD. Evidence supports its targeted use during pulmonary rehabilitation, particularly in severe disease. https://bit.ly/48EElXb

Introduction

COPD is a major public health concern and remains one of the leading causes of morbidity and mortality worldwide [1]. Its global prevalence is estimated at 10.3% of the adult population, representing nearly 392 million affected individuals [2], and it generates substantial healthcare costs, particularly in Europe [3]. Beyond acute exacerbations, COPD is characterised by progressive functional decline, reduced autonomy and impaired quality of life.

COPD is defined by persistent respiratory symptoms and incompletely reversible airflow limitation. In advanced stages (Global Initiative for Chronic Obstructive Lung Disease (GOLD) III–IV), patients experience severe exertional dyspnoea and marked exercise intolerance, leading to progressive physical inactivity and deconditioning [4, 5]. A key mechanism underlying exercise limitation is dynamic hyperinflation, which results from incomplete lung emptying during increased ventilatory demand. This leads to elevated end-expiratory lung volume, reduced inspiratory capacity, increased work of breathing and heightened dyspnoea [6, 7]. Targeting respiratory muscle overload and ventilatory constraints is therefore central to improving exercise tolerance.

Pulmonary rehabilitation is a cornerstone of COPD management and has consistently demonstrated improvements in quality of life and exercise capacity [8]. Among adjunctive strategies, noninvasive ventilation (NIV) during exercise has been proposed to enhance training tolerance [9]. NIV delivers inspiratory and expiratory positive airway pressures and is well established in the management of acute and chronic hypercapnic respiratory failure [10–15]. During exercise, NIV may reduce inspiratory muscle workload, improve ventilatory mechanics and gas exchange, decrease dyspnoea and potentially enhance endurance performance [16]. Peripheral muscle abnormalities observed in COPD further contribute to exercise limitation and provide additional rationale for ventilatory support [17].

Several studies have reported improvements in endurance time and dyspnoea when NIV is applied during exercise. However, results remain heterogeneous, partly due to variability in study design, patient selection and ventilatory settings. In particular, the level of pressure support, calculated as inspiratory positive airway pressure minus the expiratory positive airway pressure, applied during exercise is not standardised, despite suggestions that low pressure support levels (<6.0 cmH₂O) may be insufficient to produce meaningful physiological effects [18].

Previous systematic reviews have evaluated NIV during exercise in COPD. A meta-analysis published in 2021 reported that NIV during exercise training improved exercise capacity and dyspnoea in patients with COPD; however, this analysis primarily assessed the overall effect of NIV compared with control conditions and did not explore the influence of different fixed pressure support levels or potential dose–response relationships [19]. The Cochrane review by Menadue et al. [20] also reported potential benefits on exercise capacity but was limited by small sample sizes and heterogeneity, without specifically analysing the influence of pressure support levels. More recently, a network meta-analysis published in BMC Medicine [21] suggested a benefit of noninvasive positive-pressure ventilation during pulmonary rehabilitation in severe COPD; however, this analysis focused on comparisons between ventilatory modalities rather than on the impact of different fixed pressure support levels within comparable modes. To date, the specific influence of the level of fixed ventilatory pressure support applied during exercise on endurance time and exercise tolerance in COPD remains insufficiently clarified.

The primary objective of this systematic review is to evaluate the influence of the level of ventilatory pressure support applied during exercise on endurance time in patients with COPD. The secondary objectives are to assess its impact on exercise tolerance, particularly perceived dyspnoea (measured using the modified Borg scale) and changes in carbon dioxide tension (P_CO2) during ventilation. The research question is as follows: In patients with COPD, how does the level of fixed ventilatory pressure support applied during exercise influence endurance time and exercise tolerance, and what implications might these findings have for optimising ventilatory settings during exercise?

Materials and methods

This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and was prospectively registered in the PROSPERO database (CRD420251018584). A comprehensive literature search was performed in PubMed and the Cochrane Library from database inception to March 2025, with no lower date limit. The search strategy combined controlled vocabulary and free-text terms related to NIV and exercise in COPD, including [“non-invasive ventilation” OR “NIV”] AND [“chronic obstructive pulmonary disease” OR “COPD”] AND [“exercise” OR “endurance” OR “rehabilitation”]. To ensure comprehensive identification of studies specifically using fixed pressure support modes, additional terms were incorporated into the search strategy, including [“inspiratory pressure support” OR “pressure support ventilation” OR “pressure support”]. The search was updated prior to the final inclusion of studies, and the selection process was repeated according to the predefined eligibility criteria. The reference lists of eligible articles were also screened to identify any additional relevant studies.

Eligible studies were randomised controlled trials (RCTs) or randomised crossover trials published in English or French, including adults (≥18 years) with COPD confirmed by spirometry. Studies were required to evaluate the application of NIV delivered in a fixed pressure support mode during an exercise test or as part of a structured exercise training programme. Studies were only included if the exercise test used as the outcome measure was performed under NIV during the intervention and directly compared, within the same session, to a control condition without NIV. Trials assessing only the effects of a training programme conducted with NIV but performing outcome exercise tests without NIV were excluded. Comparative conditions included spontaneous breathing or sham ventilation. To enable meaningful comparison of pressure levels and exploration of a potential dose–response relationship, studies using proportional assist ventilation or other proportional modes were excluded a priori, because these modes do not provide fixed pressure levels expressed in cmH₂O and are therefore not directly comparable with fixed pressure support settings across studies.

Study selection and data extraction were performed by a single reviewer according to a predefined standardised protocol. Uncertainties were resolved through discussion with a second investigator when necessary.

The primary outcome was exercise endurance performance. Incremental exercise tests reporting peak oxygen uptake (V′_O2) or maximal oxygen uptake (V′_O2max) were not considered to be primary endurance outcomes and were therefore excluded. For studies using cycle ergometry, endurance time during constant-load exercise was considered the primary end-point. Results from the 6-min walk test and treadmill tests were included in the overall analysis; however, a prespecified subgroup analysis was performed restricting the analysis to studies using cycle ergometry, given the biomechanical and pathophysiological differences between cycling and walking.

Secondary outcomes included dyspnoea assessed using the modified Borg scale and measurements of P_CO2 (arterial, capillary or transcutaneous). When both isotime and end-exercise data were available, isotime values were preferentially extracted. End-exercise values were analysed separately when necessary. Mean±sds were extracted when directly reported. When studies reported standard errors, confidence intervals (CIs), ranges or interquartile ranges, sds were calculated or estimated using established statistical methods. When results were expressed as medians with interquartile ranges, mean±sds were approximated using validated conversion formulas.

The risk of bias of the included studies was assessed using the Cochrane Risk of Bias 2 (RoB 2) tool [22], which evaluates domains including randomisation process, deviations from intended interventions, missing outcome data, measurement of the outcome and selective reporting. Publication bias was assessed through visual inspection of funnel plots when sufficient studies were available.

Statistical analyses were performed using R software (metafor package; www.r-project.org). Random-effects models were applied to account for expected between-study heterogeneity. Effect sizes were expressed as standardised mean differences (SMDs) with 95% CIs, due to differences in exercise modalities and measurement units across studies. Heterogeneity was assessed using the I² statistic and Cochran's Q test. Mixed-effects meta-regression analyses were conducted, introducing pressure support as a continuous moderator. Statistical significance was set at p<0.05.

Results

Study selection

The study selection process involved screening of titles and abstracts followed by full-text assessment according to predefined inclusion and exclusion criteria. This process is summarised in the PRISMA flow diagram presented in figure 1.

FIGURE 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analysis flow diagram of the study selection process. NIV: noninvasive ventilation.

A total of 226 records were initially identified through database searches. After removal of duplicates, 199 records remained for title and abstract screening. Following this step, 30 full-text articles were assessed for eligibility. Ultimately, 15 studies met the predefined inclusion criteria and were included in the qualitative and quantitative synthesis. The main reasons for exclusion during full-text assessment included irrelevant outcome measures, insufficient quantitative data for analysis, exercise tests not performed under NIV and the use of proportional assist ventilation.

Population and ventilatory characteristics

A total of 15 studies were included in the analysis, comprising 285 participants. The selected publications were published between 1994 and 2022, with study designs predominantly consisting of crossover trials (n=12) and RCTs (n=3).

Population characteristics are summarised in table 1. The number of participants per study ranged from 8 to 45, with balanced intervention and control conditions. The mean age across the 15 studies was 65.4±7.2 years, with study-level mean ages ranging from 58.9 to 71 years. Among the 270 participants for whom sex was reported (data were missing in two studies), 192 were men and 78 were women, corresponding to a male-to-female ratio of approximately 2.5:1.

TABLE 1.

Population characteristics

Study	Design	Participants (n)	Age (years)	FEV1 (% predicted)
Keilty et al. 1994 [23]	Crossover	8	68.0±6.0	NR
Bianchi et al. 1998 [24]	Crossover	15	64.0±8.0	32.0±10.0
Johnson et al. 2002 [25]	RCT	22	G1: 67.0±8.0 G2: 69.0±9.0	G1: 30.7±11.3 G2: 31.6±9.3
Highcock et al. 2003 [26]	Crossover	8	66.0±8.0	36.0±15.0
Van 't Hul et al. 2004 [18]	Crossover	45	67.0±7.0	39.0±14.0
Van 't Hul et al. 2006 [27]	RCT	29	G1: 71.0±4.0. G2: 70.0±5.0	G1: 38.0±9.0 G2: 41.0±10.0
Barakat et al. 2007 [28]	Crossover	12	63.3±8.2	34.0±0.0
Dreher et al. 2007 [29]	Crossover	20	65.1±8.7	27.0±7.5
Allan et al. 2009 [30]	Crossover	12	70.0±3.0	42.0±10.0
Oliveira et al. 2010 [31]	RCT	21	63.0±7.8	40.1±10.7
Gloeckl et al. 2019 [32]	Crossover	20	60.0±6.0	19.0±4.0
Koopman et al. 2019 [33]	Crossover	13	61.5±6.8	34.0±10.2
Bonnevie et al. 2020 [34]	Crossover	21	58.9±10.7	35.3±8.3
Dennis et al. 2021 [35]	Crossover	19	69.0±7.0	39.0±6.0
Labeix et al. 2022 [36]	Crossover	20	64.0±7.0	29.0±7.0

Data presented as mean±sd. FEV₁: forced expiratory volume in 1 s; G1: control group; G2: intervention group; NR: not reported; RCT: randomised controlled trial.

The severity of airflow obstruction, assessed by forced expiratory volume in 1 s, showed a mean value of 33.4±8.9% of the predicted value. This indicates that the study population predominantly consisted of patients with severe to very severe COPD (GOLD stages III–IV).

Only the 13 studies that reported ventilation characteristics in sufficient detail, explicitly including inspiratory and expiratory pressure levels or pressure support, were included in the meta-regression analysis assessing the effect of ventilatory assistance intensity on endurance. The pressure support observed ranged from 8.0 to 21.0 cmH₂O, with a mean value of 12.2 cmH₂O (appendix C).

Risk of bias

Overall, most studies were judged to have some concerns regarding risk of bias. Visual inspection of funnel plots did not suggest substantial publication bias (appendices A and B).

Effect of NIV on endurance

The primary meta-analysis assessing the effect of pressure support ventilation on exercise endurance time in patients with COPD included 15 studies (appendix D). Using a random-effects model, pressure support ventilation was associated with a significant improvement in endurance time compared with control conditions (SMD 0.47, 95% CI 0.20 to 0.73), with moderate heterogeneity (I²=52.8%, p=0.0085).

In subgroup analyses according to exercise modality, the effect remained significant for cycle ergometer tests (SMD 0.55, 95% CI 0.22 to 0.88, I²=58.6%), whereas no statistically significant effect was observed for treadmill tests (SMD 0.16, 95% CI −0.53 to 0.86, I²=55.2%) or walking tests (SMD 0.50, 95% CI −0.13 to 1.13). No significant differences were detected between subgroups under either the common-effect model (χ²=1.22, p=0.54) or the random-effects model (χ²=1.00, p=0.61) (figure 2).

FIGURE 2.

a) Meta-analysis of the effect of noninvasive ventilation on endurance and b) meta-regression of the effect of pressure support on endurance. SMD: standardised mean difference.

A mixed-effects meta-regression model (k=13) was conducted to assess the relationship between pressure support level (cmH₂O) and endurance time. Although the regression slope was positive, pressure support intensity was not significantly associated with endurance time (β=0.0345 per cmH₂O, p=0.25). Residual heterogeneity remained moderate (τ²=0.0797, residual I²=39.97%), and the model explained 12.11% of between-study heterogeneity (QM (df=1)=1.32, p=0.25).

Effect of NIV on dyspnoea

A random-effects meta-analysis showed a significant reduction in dyspnoea with pressure support ventilation compared with control conditions (SMD −0.50, 95% CI −0.92 to −0.08), with moderate heterogeneity (I²=55.1%, p=0.0374) (appendix E, figure 3). Meta-regression analysis (k=6) demonstrated a significant association between pressure support intensity and dyspnoea (β= −0.0880 per cmH₂O, 95% CI −0.1619 to −0.0142, p=0.019), with higher levels of pressure support associated with greater reductions in dyspnoea. Residual heterogeneity was fully accounted for (I²=0%). Sensitivity analyses excluding post-exercise measurements yielded similar results (β= −0.1730 per cmH₂O, p=0.017).

FIGURE 3.

a) Meta-analysis of the effect of noninvasive ventilation on dyspnoea and b) meta-regression of the effect of pressure support on dyspnoea. SMD: standardised mean difference.

Based on this coefficient, an increase in pressure support from 6.0 to 16.0 cmH₂O would correspond to a predicted reduction in dyspnoea of approximately −0.88 SMD units at the study level. However, given the limited number of studies included in the meta-regression, these findings should be interpreted cautiously and considered hypothesis-generating.

Effect of NIV on P_CO2

A random-effects meta-analysis demonstrated a significant reduction in P_CO2 with pressure support ventilation compared with control conditions (SMD −0.48, 95% CI −0.92 to −0.04), with substantial heterogeneity (I²=68.2%, p=0.0025). Meta-regression including all available time points (k=8) showed a negative but nonsignificant association between pressure support intensity and P_CO2 (β= −0.0673 per cmH₂O, 95% CI −0.1469 to 0.0123, p=0.098).

When post-exercise measurements were excluded (k=6), the association became statistically significant (β= −0.1055 per cmH₂O, 95% CI −0.2026 to −0.0084, p=0.033), suggesting a greater reduction in P_CO2 with higher levels of pressure support during exercise (appendix F, figure 4). However, given the limited number of studies and the substantial heterogeneity observed in the primary analysis, these findings should be interpreted cautiously.

FIGURE 4.

a) Meta-analysis of the effect of noninvasive ventilation on carbon dioxide tension and b) meta-regression of the effect of pressure support on carbon dioxide tension. SMD: standardised mean difference.

Discussion

The present systematic review and meta-analysis demonstrates that NIV applied during exercise significantly improves endurance time and dyspnoea in patients with COPD. The magnitude of this effect (0.47, 95% CI 0.20 to 0.73) corresponds to a small-to-moderate improvement, which is consistent with clinically relevant gains reported in individual trials. However, the use of SMDs limits direct comparison with established minimal clinically important differences for specific exercise tests. The relatively wide confidence intervals observed across several analyses indicate some uncertainty regarding the precise magnitude of these effects, which may range from small to more substantial.

However, no clear dose–response relationship was observed between pressure support intensity and endurance performance. By contrast, pressure support intensity was significantly associated with reductions in dyspnoea and P_CO2 during exercise, suggesting that ventilatory unloading is pressure-dependent, whereas endurance capacity remains influenced by additional non-ventilatory factors. From a physiological perspective, NIV reduces inspiratory muscle workload and mitigates dynamic hyperinflation, thereby decreasing the mechanical constraints imposed on the respiratory system. This unloading effect is closely linked to a reduction in neural respiratory drive (NRD), which reflects the central motor output required to sustain ventilation [37]. In COPD, increased NRD results from airflow limitation, intrinsic positive end-expiratory pressure and impaired ventilatory efficiency. By assisting inspiratory effort, NIV decreases the pressure-generating demand placed on respiratory muscles, leading to reduced NRD and, consequently, lower perception of dyspnoea [38]. The significant dose–response relationship observed for dyspnoea and P_CO2 during exercise reinforces this mechanism. Higher pressure levels likely enhance alveolar ventilation and reduce CO₂ retention under conditions of increased ventilatory demand. The strengthening of this association after exclusion of post-exercise measurements further supports the importance of assessing physiological variables at isotime or peak exercise, when ventilatory load is maximal and the unloading effect of NIV is most relevant.

Because this meta-regression was conducted at the study level, the observed association reflects between-study differences in mean pressure support and should not be interpreted as defining an individual titration target in clinical practice. The absence of a dose–response relationship for endurance time suggests that exercise performance in COPD is not determined solely by ventilatory limitation. Peripheral muscle dysfunction, cardiovascular constraints, deconditioning and impaired oxidative capacity are well-recognised contributors to exercise intolerance [39]. Thus, while increasing pressure support may further alleviate respiratory effort, endurance gains may plateau once non-ventilatory limitations become predominant. Rather than indicating a true ceiling effect of NIV, our findings suggest that while symptom relief may continue to improve with higher levels of assistance, endurance performance remains constrained by multifactorial systemic limitations.

Another factor likely contributing to heterogeneity is the type of interface used to deliver NIV. The included studies employed nasal and oronasal masks, which differ in terms of leakage, patient–ventilator synchrony, and ability to maintain effective support during high-intensity exercise. Data from studies evaluating interface-related asynchronies indicate that while oronasal masks may reduce patient–ventilator asynchronies, ventilator flow limitations can occur at high exercise intensities, leading to reductions in delivered inspiratory pressure. Conversely, nasal interfaces may be associated with a higher frequency of ineffective efforts [34]. These findings suggest that the effectiveness of ventilatory support depends not only on the programmed pressure level but also on the quality of synchrony and the technical capacity of the ventilatory system, which may partly explain variability in endurance outcomes independent of the nominal increase in ventilation achieved. In line with this concept, Highcock et al. [26] reported that although noninvasive intermittent positive-pressure ventilation increased ventilation during exercise, it did not improve exercise capacity, and no significant differences were observed between ventilator types. These results further support the notion that augmenting ventilation alone may be insufficient to translate into functional performance gains, and that the optimal method of assistance remains to be clarified.

Previous meta-analyses have reported improvements in exercise tolerance with NIV in COPD, but were based on a limited number of studies and did not evaluate the influence of pressure support intensity or measurement timing. The present analysis extends these findings by incorporating a larger body of evidence, exploring dose–response relationships, and distinguishing between exercise and post-exercise measurements. Our results indicate that pressure-dependent effects are more consistently observed on physiological markers of ventilatory load than on endurance performance itself. Other strategies have also been investigated to improve exercise tolerance in COPD, including supplemental oxygen, high-flow nasal oxygen and heliox. Oxygen supplementation primarily enhances tissue oxygen delivery and may reduce ventilatory requirements for a given workload [40]. High-flow nasal oxygen can decrease anatomical dead space and improve comfort, while heliox reduces airway resistance and the work of breathing [41]. By contrast, NIV uniquely provides active inspiratory muscle unloading. These mechanistic differences may explain why increasing pressure support predominantly influences dyspnoea and P_CO2, whereas endurance performance remains influenced by broader systemic determinants.

These findings have relevant clinical implications. NIV during exercise may be particularly beneficial in patients limited by ventilatory constraints, especially those experiencing marked dyspnoea or exercise-induced hypercapnia. Pressure titration strategies might therefore aim primarily to optimise symptom relief and ventilatory unloading rather than to seek linear improvements in endurance performance. Several limitations should be acknowledged. Heterogeneity remained moderate to substantial across analyses, reflecting variability in study design, exercise protocols, ventilatory modes and patient characteristics. The number of studies included in meta-regression analyses was limited, reducing statistical power for detecting dose–response effects. In addition, measurement timing differed across studies, and SMDs were used to pool outcomes, which may limit clinical interpretability. Variability in pressure support strategies (fixed versus individualised titration) may also have introduced residual heterogeneity, because study-level mean values may not fully capture intra-study variability. Baseline P_CO2 levels varied across studies, with three studies including predominantly normocapnic participants. This may have contributed to heterogeneity and limits definitive conclusions regarding differential effects in hypercapnic versus normocapnic populations. Finally, study selection was performed by a single reviewer, and reporting of tolerance and adverse effects was limited in several trials.

Conclusion

This systematic review and meta-analysis shows that NIV during exercise improves endurance time and dyspnoea in patients with COPD. Pressure support intensity was associated with reductions in dyspnoea and P_CO2 during exercise, supporting a pressure-dependent ventilatory unloading effect. However, no dose–response relationship was observed for endurance performance, underscoring the multifactorial nature of exercise limitation in COPD. Further standardised studies are needed to optimise pressure titration strategies during exercise.

Proposed opening for future research

Ultimately, the clinical objective is to optimise and personalise exercise-based rehabilitation in patients with severe COPD. Our findings indicate that while pressure support intensity modulates ventilatory unloading during exercise, endurance performance remains multifactorial and not strictly pressure-dependent. This raises an important research question: how should pressure support be titrated during exercise to achieve optimal symptom relief and physiological unloading, while accounting for individual patient characteristics, ventilatory constraints and interface-related factors?

Future studies should prioritise well-designed, standardised protocols directly comparing different pressure levels during exercise, with careful assessment at isotime and peak effort. Such investigations are needed to better define dose–response relationships for physiological outcomes, identify patient subgroups most likely to benefit, and translate mechanistic improvements into meaningful and individualised clinical strategies.

Supplementary material

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

Supplementary material

00049-2026.SUPPLEMENT

Data availability

All data generated or analysed during this study are included in this published article and its supplementary material.