Nasal High Flow for Stable Patients with Chronic Obstructive Pulmonary Disease: A Systematic Review and Meta-Analysis

Abstract

There is a growing body of evidence supporting the use of nasal high flow (NHF) to treat acute respiratory failure, particularly in Chronic Obstructive Pulmonary Disease (COPD) patients. Conversely, there are sparse data evaluating its effects in stable COPD patients.

We identified randomized controlled trial comparing the effects of delivering air or oxygen via NHF, compared with delivering the same gas without NHF, in stable COPD patients through a systematic search using MEDLINE, CENTRAL, Science Direct, and others sources until January 2019. Study selection, data extraction and assessment of the risk of bias (using the Cochrane Risk of Bias tool) was performed by two independent authors.

We included 6 studies (339 participants). Our meta-analysis showed a significant reduction of arterial carbon dioxide pressure (PaCO₂) at long (two studies, MD −3 mmHg, [95% Confidence interval (CI) −4 to −2]) and short-term (two studies, MD -3 mmHg [95% CI −4 to −2]). NHF significantly improved quality of life on the St George’s Respiratory Questionnaire (two studies, MD −5 out of 100, [95% CI −8 to −2]). NHF significantly reduced the rate of acute exacerbation at 1 year (one study, rate ratio: 0.6, [95% CI 0.6 to 0.7]). NHF did not significantly improve exercise capacity, hospitalization rate or mortality, but improved breathing pattern.

NHF reduced PaCO₂, acute exacerbation and improved quality of life in stable COPD patients. Further long-term studies are needed to confirm the present results and provide more data on patient-centered outcome such as quality of life, exacerbation, hospitalization and mortality.

Keywords:

• Pulmonary disease, chronic obstructive
• nasal high-flow
• high-flow cannula
• meta-analysis

List of abbreviations
COPD	=	chronic obstructive pulmonary disease;
HR	=	heart rate
FEV1	=	forced expiratory volume in 1s;
FVC	=	forced vital capacity;
NHF	=	nasal high-flow;
NIV	=	noninvasive ventilation;
mMRC	=	modified Medical Research Council dyspnea scale;
MV	=	minute ventilation;
PaCO2	=	arterial carbon dioxide partial pressure;
Pga	=	gastric pressure;
Poes	=	esophageal pressure;
PtcCO2	=	transcutaneous arterial carbon dioxide partial pressure;
Ptdia	=	transdiaphragmatic pressure;
REM	=	rapid eye movement;
RR	=	respiratory rate;
RSBI	=	rapid shallow breathing index;
SpO2	=	pulsed oxygen saturation;
StO2	=	transcutaneous oxygen saturation;
SaO2	=	arterial oxygen saturation;
TcO2	=	transcutaneous oxygen;
Vt	=	tidal volume.

Introduction

COPD is a worldwide cause of morbi-mortality with a growing burden [1, 2]. This disease progressively leads to chronic respiratory insufficiency which can lead to hypoxia and hypercapnia [3], each of which is associated with poor outcomes [3, 4].

Long-term oxygen therapy reduces both hypoxia and mortality significantly in severely hypoxic patient with COPD [5] whereas intermittent noninvasive ventilation (NIV) with high pressure has been used to reduce arterial carbon dioxide and improve mortality in chronic hypercapnic COPD patients (>55 mmHg) [6, 7].

Unfortunately, there are important drawbacks associated with the use of NIV, including interface discomfort, excessive high air pressure, sleep disturbance and intolerability due to patient-ventilator asynchrony, each of which can lead to poor compliance or treatment failure [8–11]. Therefore, alternative strategies are warranted.

Nasal high flow (NHF) delivers heated and humidified high flow air (up to 60 L/min) through nasal canula, providing promising physiological benefits such as positive airway pressure [12] or upper airway CO₂ washout [13]. It can be used in association with oxygen and offers in this situation the advantage to match the patient’s inspiratory flow, preventing any dilution of the inspired fraction of oxygen [14]. NHF has widely been studied in adult intensive care units and seems better than conventional oxygen therapy and as effective as NIV with regards to mortality to treat hypoxemic acute respiratory failure [15–19]. Moreover, it may be more comfortable than conventional treatments [20] and could help to control hypercapnia in patients with COPD [21].

Despite this promising background evidence about NHF, there are sparse data evaluating its effects on clinically important outcome in stable patients with COPD.

Therefore, the overall aim of this review was to summarize the available evidence assessing the effects of delivering air or oxygen via NHF, compared with delivering the same gas without NHF, in people with stable COPD.

Method

Study registration and methodology

The protocol of this systematic review and meta-analysis was prospectively registered at PROSPERO (www.crd.york.ac.uk/prospero; CRD: 42018103358). It has been designed according to the Cochrane Handbook for Systematic Reviews of Interventions [22] and reported according to the PRISMA statement.

Criteria for considering studies for this review

Types of studies

Parallel and cross-over randomized trials, included those in the format of an abstract, assessing one or more of the considered outcomes.

Type of participants

Adult patients with stable COPD (no acute exacerbation in the previous 3 weeks), of any age, diagnosed based on the individual study’s criteria. Studies with a mixed population of COPD and other respiratory disease(s) could be included if the data for the participants with COPD could be extracted separately.

Type of intervention

NHF with air, or supplemental oxygen if indicated by the patient’s clinical status, compared with the same gas delivered without NHF.

Type of outcome measures

Primary outcomes

1. Arterial carbon dioxide partial pressure measured transcutaneously (PtcCO₂) or by arterial blood gases (PaCO₂);

2. Arterial oxygen partial pressure measured transcutaneously (TcO₂) or by arterial blood gases (PaO₂);

3. pH measured by arterial blood gases;

4. Quality of life (assessed either with a general or a disease-specific questionnaire);

5. Deaths;

6. Number of acute exacerbations per year;

7. Number of hospitalizations per year;

Secondary outcomes

1. Oxygen saturation measured by pulse or transcutaneous oximetry (SpO₂ and StO₂ respectively) or by arterial blood gases (SaO₂);

2. Cardiorespiratory function (heart rate (HR); forced expired volume in 1s (FEV₁); forced vital capacity (FVC);

3. Breathing pattern (volume tidal (Vt); respiratory rate (RR); minute ventilation (MV); rapid shallow breathing index (RSBI));

4. Respiratory mechanics (gastric pressure (Pga); esophageal pressure (Poes); transdiaphragmatic pressure (Ptdia); end-expiratory lung impedance);

5. Dyspnea assessed either during or after an exercise or at rest;

6. Exercise capacity measured either by a maximal cardiopulmonary exercise testing or field tests;

7. Objective or self-reported physical activity;

8. Comfort and patient preference;

9. Adverse events.

Search methods for identification of studies

Electronic searches

MEDLINE, CENTRAL, Science Direct, Scopus, PEDro, OpenGrey and Greylit were searched from inception up to January 2019 for relevant studies in English or French. Reference list of the included studies were also checked for eligible studies.

Detailed searching strategy is shown in eAppendix 1.

Data collection and analysis: see eAppendix 1.

Study selection, data extraction and assessment of the risk of bias (using the Cochrane Risk of Bias tool[22]) was performed by two independent authors. Any disagreement was resolved by discussion or the intervention of a third author.

Data synthesis

Heterogeneity was assessed using the I² and Chi² statistic, considering values I² ≥ 50% as a sign of moderate to high heterogeneity. Meta-analysis was performed with a fixed or a random-effect model according to heterogeneity. RevMan 5.3.5 was used for analysis. The quality of evidence was rated using the GRADE system.

Results

Descriptions of studies and participants

188 records were retrieved from database searching after removal of duplicates and 6 studies [23–28] were included giving a total of 339 participants. See Figure 1.

Figure 1. Study flow diagram.

There was a good agreement between authors for the study selection (kappa score: 0.71).

Characteristics (methods, participants, intervention and outcomes) of the included studies are shown in Table 1.

Table 1. Characteristics of included studies.

First author (date)	Methods	Participants	Intervention	Outcomes	Notes
Long-term
Nagata (2017)	Randomized cross-over study Data collection at 6 weeks for each intervention	29 patients with stable hypoxic and hypercapnic COPD (mean age 75 years, 90% males, mean FEV1 29% predicted)	NHF group: AIRVO 20 to 40 L/min with O2 adjusted to maintain SpO2>88% 7h/day   LTOT group: O2 1 to 2 L/min or usual care	Primary: QOL Secondary: PCO2, nocturnal TcCO2, PaO2, pH, SaO2, FEV1, FVC, dyspnea, physical activity, exercise capacity, adverse events and acute exacerbation	No washout period
Storgaard (2018)	Randomized parallel study Data collection every 3 months during 12 months	200 COPD patients with chronic hypoxemic respiratory failure (mean age 71 years, 40% males, mean FEV1 31% predicted)	NHF group: AIRVO 20 L/min with O2 adjusted to maintain SpO2>88% 6-7h/day (>8h recommended) LTOT group: conventional oxygen nasal cannula 1- to 2 L/min	Primary: acute exacerbation Secondary: PCO2, FEV1, dyspnea, exercise capacity, QOL, hospitalization and mortality
Short-term
Fraser (2016)	Randomized cross-over study 2.5 to 3 hours data collection period	30 patients with stable hypoxic COPD (mean age 74 years, 100% males, FEV1% not reported)	NHF group: AIRVO 30 L/min with O2 to reach the same FiO2 than with conventional oxygen LTOT group: conventional oxygen nasal cannula 2 to 4 L/min 20min (×2)	PCO2, PO2, SaO2, Vt, RR, HR, MV, EELI, dyspnea and comfort	Only males Primary and secondary outcomes not differentiated
Mckinstry (2018)	Randomized cross-over study >2.5 hours data collection period	48 patients with stable COPD (mean age 69 years, 60% males, FEV1 53% predicted)	NHF group: AIRVO 15 L/min, 30 L/min and 45 L/min of air Control group: Room air only, without NHF 20min (×2) 15min washout	Primary: PCO2 Secondary: SaO2, RR, HR
Nilius (2013)	Randomized cross-over study >6 hours data collection period	20 patients with stable hypoxic and hypercapnic COPD (mean age 66 years, % males not reported, FEV1 28% predicted)	NHF group: TNI20oxy, 20  L/min with O2 (2 to 3 L/min) LTOT group: conventional oxygen nasal cannula 2 to 3 L/min	Nocturnal TcCO2, arterial blood gases and sleep indices	Primary and secondary outcomes not differentiated
During exercise
Cirio (2016)	Randomized cross-over study Data collection period: 3 separate days	12 patients with stable COPD Ventilatory limitation (MVV-peakVe < 11  L/min) Limited to effort (6MWT <75%) (mean age 70 years, 83% males, mean FEV1 35% predicted)	NHF group: AIRVO 55 to 60L/min Control group: Venturi mask (with or without O2) Material: Cycle-ergometer (44 ± 17W) Process: Incremental test (day 1) In a random order, two constant-load exercise tests (day 2 and 3), at 75% of the maximal workload achieved during the incremental test (day 1) with HFNC or Venturi mask	Primary: exercise capacity Secondary: SaO2, dyspnea and adverse events	Venturi mask: -connected to compressed air in place of oxygen -additional oxygen to maintain SaO2 >88%

Five studies were reported as full-text publications [23–25, 27, 28] while one was available as an abstract only [26]. Three assessed the short-term effects of NHF [24–26], two assessed the long-term effects [27, 28] and the remaining study assessed NHF during exercise [23].

Patients with chronic hypoxemia [24, 28], or both chronic hypoxemia and hypercapnia [26, 27] were specifically included in four studies. When patients had oxygen during the control intervention, the oxygen flow during NHF was adjusted as necessary to maintain the same baseline transcutaneous oxygen saturation (SpO₂) [23, 24, 27, 28].

Risk of bias in included studies

The risk of bias among the included studies is depicted in Figure 2A and Figure 2B. The detailed evaluation is available in eAppendix 2.

Figure 2. A/Risk of bias among the included studies. Review authors' judgements about each risk of bias item presented as percentages across all included studies; B/Risk of bias summary. Review authors' judgements about each risk of bias item for each included study.

Effects of interventions

The results for the three categories of trials (long-term, short-term, and during exercise) are detailed below. The overall findings are also summarized in Table 2.

Table 2. Summary of findings.

Nasal high-flow compared to usual care for stable patients with chronic obstructive pulmonary disease
Patient or population: stable patients with chronic obstructive pulmonary disease Setting: Intervention: Nasal high-flow Comparison: usual care
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with usual care	Risk with Nasal high-flow
Arterial carbon dioxode partial pressure (mmHg); follow up: 1.5-12 months		The mean arterial carbon dioxode partial pressure (mmHg) in the intervention group was 2,95 lower (4,27 lower to 1,64 lower)	–	229 (2 RCTs)	⨁⨁⨁◯ MODERATE^a	Long-term nasal high-flow likely reduces arterial carbon dioxode partial pressure (mmHg) slightly.
Arterial carbon dioxide partial pressure (mmHg): acute effects with NHF		The mean arterial carbon dioxide partial pressure (mmHg): acute effects with NHF in the intervention group was 2,73 lower (3,8 lower to 1,65 lower)	–	78 (2 RCTs)	⨁⨁⨁◯ MODERATE^a	Acute session of nasal high-flow likely reduces arterial carbon dioxide partial pressure (mmHg) slightly.
Quality of life assessed with: Saint George's Respiratory Questionnaire (%); Scale from: 0 to 100; follow up: 1.5-12 months		The mean quality of life in the intervention group was 5,29 lower (8,3 lower to 2,28 lower)	–	229 (2 RCTs)	⨁⨁⨁◯ MODERATE^a	Lower value indicates a better quality of life. Long-term nasal high-flow likely results in a large improvement in quality of life, higher than the defined minimum important difference for this outcome (4 points).^1
Mortality; follow up: 1 years	120 per 1 000	150 per 1 000 (74 to 304)	RR 1.25 (0.62 to 2.53)	200 (1 RCT)	⨁⨁◯◯ LOW^b	The evidence suggests that long-term nasal high-flow does not improve mortality.
Rate of acute exacerbation of COPD; follow up: 100 patient years	50 per 1 000	31 per 1 000 (27 to 36)	Rate ratio 0.63 (0.55 to 0.72)	200 (1 RCT)	⨁⨁◯◯ LOW^b	Long-term nasal high-flow may result in a reduction in rate of acute exacerbation of COPD.
Rate of hospitalization; follow up: 100 patient years	122 per 1 000	109 per 1 000 (84 to 140)	Rate ratio 0.89 (0.69 to 1.15)	200 (1 RCT)	⨁⨁◯◯ LOW^b	The evidence suggests that nasal long-term high-flow does not reduce rate of hospitalization.
Dyspnea assessed with: mMRC dyspnea scale; Scale from: 0 to 4; follow up: 1-1.5 months		The mean dyspnea in the intervention group was 0,27 lower (0,41 lower to 0,13 lower)	–	229 (2 RCTs)	⨁⨁⨁◯ MODERATE^a	Lower value indicates an improvement in dyspnea. Long-term nasal high-flow probably results in little to no difference in dyspnea.
Exercice capacity assessed with: six-minute walk test (m); follow up: 1.5-12 months		The mean exercice capacity in the intervention group was 18,19 higher (9,18 lower to 45,55 higher)	–	229 (2 RCTs)	⨁◯◯◯ VERY LOW^c,d,e	The evidence is very uncertain about the effect of nasal high-flow on exercice capacity.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; MD: Mean difference; RR: Risk ratio
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect Moderate certainty: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low certainty: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low certainty: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

^a The quality of evidence was downgraded to moderate due to the risk of bias of the included studies

^b The quality of evidence was downgraded to moderate due to the risk of bias of the included studies.

^c The quality of evidence was downgraded due to the risk of biais of the included studies.

^d Included studies showed divergent results and high heterogeneity.

^e Wide confidence interval.

Long-term effects

Two studies including 229 participants contributed data on the long-term effects of NHF [27, 28]. The follow-up period for these studies ranged from 6 weeks to 12 months.

Primary outcomes

Arterial carbon dioxide pressure

Meta-analysis of two studies showed a statistically significant benefit for NHF on PaCO₂ (pooled MD −3 mmHg, 95% CI −4 to −2; I²=0%; Figure 3A).

Figure 3. Long-term (A) and short-term (B) effects of nasal high-flow on arterial carbon dioxide pressure (mmHg).

Nagata et al. also assessed single night PtcCO2 (without NHF) [27]. They found a significant improvement in the mean median PtcCO₂ (MD −5mmHg, 95% CI −8 to −2) in favor of NHF.

Arterial oxygen pressure

One study was available for this outcome (total 29 participants) [27]. There was no difference between NHF and control in PaO₂ (MD −3 mmHg, 95% CI −9 to 3).

pH

One cross-over study (total 29 participants) found a significant increase in pH in favor of NHF (MD 0.02, 95% CI 0 to 0.04) [27]. Conversely, Storgaard et al. (total 200 participants) found no significant difference in baseline-adjusted changes in pH (detailed data not available on the original report and we had no answer from the author). The quality of evidence was “very low”.

Quality of life

The two studies that measured quality of life both used the Saint George’s Respiratory Questionnaire, which is measured from 0 (no impairment) to 100 (maximum impairment). The result showed a statistically significant effect in favor of NHF (pooled MD −5, 95% CI −8 to −2; I²=11%; Figure 4).

Figure 4. Long-term effects of nasal high-flow on quality of life (Saint George’s Respiratory Questionnaire, %).

Mortality

One study was available for this outcome (total 200 participants) [28]. There was no difference between NHF and control in death events (risk ratio = 1.25, 95% CI 0.62 to 2.53).

Acute exacerbation

Because of the relatively short-term evaluation for this particular outcome in one study (6 weeks) and its cross-over design [27], meta-analysis was deemed inappropriate and results are reported narratively instead.

One cross-over study (total 29 participants) [27] reported three exacerbation’s of COPD with long-term oxygen (10.3%) and none with NHF over a 6-week period of treatment. One parallel study (total 200 participants) reported a significant reduction in acute exacerbations of COPD (rate ratio: 0.63, 95% CI 0.55 to 0.72) at 1 year.

Hospitalization

One study was available for this outcome (total 200 participants) [28]. There was no difference between NHF and control in patient-year hospitalization rate (rate ratio: 0.89, 95% CI 0.69 to 1.15).

Secondary outcomes

Results for the secondary outcomes are shown in eAppendix 3. The quality of evidence ranged from “moderate” (significant improvement in the mMRC dyspnea scale with NHF) to “low” (no significant effects FEV₁(%)) and “very low” (no significant effects on SpO₂, exercise capacity, and physical activity).

Short-term effects

Three cross-over studies including 98 participants contributed data on the short-term effects of NHF [24–26].

Primary outcomes

Arterial carbon dioxide pressure

Two studies including 78 participants contributed data on arterial carbon dioxide pressure [24, 25]. Meta-analysis showed statistically significant effects of nasal high flow on arterial carbon dioxide pressure (pooled MD −3 mmHg, 95% CI −4 to −2; I²=0%; Figure 3B).

One study including 20 participants assessed single-night nocturnal PtcCO2 [26]. Nilius et al. found a significant improvement both in the non-REM PtcCO2 (MD −1 mmHg, 95% CI −3 to −0) and in the REM PtcCO2 (MD −2 mmHg, 95% CI −3 to −1) with NHF.

Arterial oxygen pressure

One study was available for this outcome (total 30 patients) [24]. There was a significant difference between NHF and control in TcO₂ in defavor of the COPD patients randomized to the NHF group (MD −4 mmHg, 95% CI −7 to −1). The quality of evidence was “very low”.

Secondary outcomes

Results for the secondary outcomes are shown in eAppendix 3. The quality of evidence ranged from “very low” (no significant effects on SpO₂, HR and MV, significant improvement in Vt and end-expiratory lung impedance and significant worsening in dyspnea and comfort with NHF) to “low” (significant improvement in RR with NHF).

During exercise

Only one cross-over study (12 participants) assessed the effects of during exercise NHF (constant workload exercise testing) compared with oxygen [23]. The overall quality of evidence was “very low” (significant improvement in endurance time, isotime dyspnea and isotime SaO₂).

Detailed results are shown in eAppendix 4.

Discussion

The main findings of this meta-analysis were that NHF significantly reduced PaCO₂ (about -3mmHg) compared with usual care. Moreover, it significantly increased health-related quality of life and breathing pattern, and reduced acute exacerbations of COPD. NHF had no significant effects on mortality, hospitalization rate, pulmonary function and exercise capacity. One short-term study revealed that NHF could reduce TcO₂ and worsen dyspnea and comfort. Finally, the use of NHF during exercise might improve endurance capacity and dyspnea.

The improvement in PaCO₂ was found both at short and long term. The main explanation for the short-term improvement lies in the improvement of the breathing pattern which is supported by our results suggesting an increase in Vt and a decrease RR. Since minute ventilation was not significantly different between conditions, the improvement in RR and Vt contributed to improve alveolar ventilation. This deeper breathing pattern likely arises from a combination of several factors such as a reduction of inspiratory resistance, improved respiratory mechanics [29], humidification [30] and a positive expiratory pressure effect [14, 29, 31]. Moreover, the flow and leakage dependent washout of the upper airway has been highlighted as an important factor to reduce PaCO₂ [13]. The mechanisms of long-term improvement are unclear and remain to be elucidated. It could be hypothesized that a better NHF-mediated humidification improved airway clearance and inflammatory status [14, 30] which in turn decreased dynamic hyperinflation and inspiratory load [32], while improving breathing pattern (which was also observed in this review) [32], and finally improved ventilation-perfusion matching [33]. The magnitude of improvement of PaCO₂ with NHF was relatively modest (-3 mmHg, [95% CI -4to −2]) compared to that found with high pressure and rather high backup rate NIV (−7 mmHg [CI 95% −9 to −4], recalculated data from Tables 3 and 4) [34]. This difference could in part be explained by the difference in baseline PaCO₂ in the study by Kohnlein et al. (about 58 mmHg) compared to that of the studies included in the present meta-analysis (< 50 mmHg). Despite this, the indirect comparison of these confidence intervals shows some degree of overlapping so that an equivalent effect cannot totally be ruled out. This is strengthened by a recent study which compared the effects of 6 weeks of NHF with 6 weeks of NIV on PaCO₂ and showed substantial improvement with both treatments without any significant difference between them [35]. Therefore, NHF could be relevant for hypercapnic COPD patients who do not tolerate high-pressure NIV and long-term studies comparing the effects of NHF and NIV are now warranted.

The present results also suggest that NHF did not improve SpO₂ both at short and long-term. This was also the case for long-term PaO₂. This is strengthened by a recent study which did not support the use of NHF with room air as a stand-alone therapy to oxygenate hypoxemic COPD patients who already benefit from long-term oxygen therapy [36]. Surprisingly, one short-term study reported a significant fall in TcO₂ when using NHF supplemented with an equivalent fraction of inspired oxygen compared with 2 to 4 L/min of oxygen through nasal cannula (about -4mmHg) [24]. Reassuringly, the mean TcO₂ in the NHF group remained largely within a normal range (97 mmHg (SD 24)). Nonetheless, this result raises concern about the necessity to titrate oxygen during NHF to avoid any flushing effect of the usual oxygen prescription as previously described [29].

Interestingly, long-term NHF was associated with a decrease in the rate of acute exacerbation of COPD of about 40% in the only long-term study, which assessed this outcome at 1 year. Considering the burden of exacerbations on disease progression and functional outcome [37, 38], even a 25% reduction in this relative risk, using a conservative approach based on the lower bound of the 95% CI, would still be particularly relevant. The underlying mechanisms remain to be elucidated but the implication of humidification in the management of airway secretions retention is a plausible explanation [30]. Since exacerbation and quality of life are closely related, the improvement of the former can explain the improvement of health-related quality of life with NHF (-5% [95% CI: -8 to -2]) which exceeded the minimum clinically important difference (MCID) of -4% for the SGRQ. These results are in line with those of Rea et al., who found a significant reduction in the rate of acute exacerbation and improvement in quality of life with NHF in a mixed population of COPD and bronchiectasis patients [39]. However, the 95% CI of the present estimate is relatively wide and its lower bound falls below this MCID, so a clinically trivial effect cannot be excluded.

In the context of intensive care unit, NHF is frequently reported as better tolerated (comfort and dyspnea) than the treatment it is compared with (oxygen or NIV) [15, 16, 40]. Conversely, the results presented here, from a single study, highlight that it may be less comfortable and elicit more dyspnea than conventional oxygen in stable COPD patients [24]. This may be attributable to the initiation of a new treatment and seems to be reversible with time because long-term dyspnea was improved. Also, the choice of the NHF flow and temperature is of real concern with regards to comfort. 30 L/min and temperature below 37 °C seems to be a good compromise between efficiency in reducing PaCO₂ and comfort [25, 41]. Despite this, NHF was not associated with an increase in adverse event.

Finally, while the improvement in exercise capacity associated with the use of long-term NHF was not significant, its 95% CI encompassed the MCID (25 to 33 m [42]). Because of a high level of heterogeneity, the quality of evidence was downgraded to “very low” and further studies are warranted to refine this estimate.

Limitations: The main limitation of this study was the number of included studies, reducing the meta-analysis to the combination of a maximum of two studies. Major outcomes, including hospitalization rate and mortality were only assessed in one study. Moreover, the confidence in the estimates was restricted due to methodological issues, particularly because of inadequate blinding of the participants and outcome assessors. Although some outcomes such are exacerbation are most likely not influenced by the lack of blinding, the overall quality of evidence ranged from moderate to very low.

Conclusion

NHF has the potential to reduce PaCO₂, acute exacerbation and improve long-term health related quality of life in stable COPD patients. Although these estimates are promising, further evidence could refine them. Caution should be taken when setting up NHF in those patients with oxygen to avoid any flushing effect of oxygen. Further long-term studies are needed to confirm the present results and should now compare the effects of NHF with those from NIV.

Declaration of interest

TB reports grant from Fisher & Paykel. Dr MP reports grants from B&D Electromedical, personal fees from ResMed and Philips Respironics, grants and nonfinancial support from Fisher & Paykel, nonfinancial support from MSD, nonfinancial support from Asten, and grants from ADIR Association, outside the submitted work. The other authors report no conflicts of interest in this work.

Ethical approval and consent to participate

Written informed consent was not required for this study.

Acknowledgments

We thank Gregory Reychler for his support during data analysis.

Additional information

Funding

None

Notes on contributors

Tristan Bonnevie

TB, ME, CP, FEG, and GP have made substantial contributions to conception and design, or acquisition of data, or analysis and interpretation of data; have drafted the submitted article or revised it critically for important intellectual content; have provided final approval of the version to be published; and have agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. CM, YC, MP, JFM, and AC have drafted the submitted article or revised it critically for important intellectual content; have provided final approval of the version to be published; and have agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.