X3.2 Non-invasive positive pressure ventilation

NIV is an effective and safe means of treatment of ventilatory failure. Its use allows preservation of cough, physiological air warming and humidification, and normal swallowing, feeding and speech. Applying NIV in addition to conven­tional therapy reduces the risk of mortality by 46% (risk ratio (RR) 0.54, 95% confidence interval (CI) 0.38 to 0.76); NNT 12 and decreases the risk of needing endotracheal intubation by 65% (RR 0.36, 95% CI 0.28 to 0.46; NNT 5) (Osadnik 2017). This benefit is similar for patients with mild acidosis (pH 7.30 to 7.35) versus a more severe nature (pH < 7.30), and when NIV is applied in a ward or intensive care unit (Osadnik 2017). The use of NIV reduces hospital length of stay mean difference -3.39 days (95% CI -5.93 to -0.85) (Osadnik 2017).

A local prospective observational cohort study demonstrated that ward-based NIV (managed by respiratory medical and nursing staff) compared with high dependency unit (HDU) and ICU-based NIV achieved equivalent clinical outcomes and was substantially more cost-effective (Parker 2018). These findings were replicated in a similar but retrospective study based in a teaching hospital in China (Hong 2020). The optimal location for provision of NIV should be determined by local experience and availability of expertise.

Hartley et al used a derivation cohort of 489 patients to derive a mortality prediction score for patients with with an exacerbation of COPD and hypercapnic respiratory failure receiving NIV. The NIVO score was then validated in a group of 733 patients from across 10 hospitals in England and Wales. The NIVO score consisted of 6 measures that should be available at the bedside (see below). The area under the curve form predicting mortality was 0.79.  The score also allowed for mortality risk stratification – see table below.  The NIVO score performed better in this patient group than all other mortality prediction scores tested.  Use of this score may assist clinicians, patients and their carers in making decisions regarding acute non-invasive ventilation (Hartley 2021) [evidence level III].

X3.2.1 High flow versus non-invasive ventilation for acute exacerbations

Non-invasive ventilation (NIV) is the established first-line ventilatory support for patients with acute exacerbations of COPD (AECOPD) complicated by moderate-to-severe hypercapnic respiratory failure, reducing the need for intubation and improving survival. However, NIV can be poorly tolerated in some patients due to discomfort, claustrophobia, and mask-related complications (Papachatzakis 2020). Humidified nasal high flow (hNHF), also called high-flow nasal oxygen (HFNO), has been proposed as a more comfortable alternative that may improve dyspnoea relief and patient tolerance. In recent years, several randomised controlled trials have compared hNHF with NIV, but their findings are mixed and limited by small sample sizes, heterogeneous populations, and variability in outcome definitions.

Five randomised controlled trials (Xia 2022; Cortegiani 2020; Tan 2024; Pantazopoulos 2024; Papachatzakis 2020) and one systematic review/meta-analysis (Fahey 2023) have evaluated hNHF as an alternative to NIV in hypercapnic AECOPD.

Study populations across these trials were highly heterogeneous. None of the studies evaluated patients with severe hypercapnia, defined as PaCO₂ >70 mmHg or pH <7.25, who are often the group most likely to require NIV in real-world practice. Two of the studies enrolled patients with moderate hypercapnia, the Italian study with a threshold (PaCO₂ ≥55 mmHg) slightly higher than that seen in a non-inferiority study in China (≥50 mmHg) (Cortegiani 2020, Tan 2024). Another trial in China enrolled patients with only mild hypercapnia, where inclusion criteria were pH ≥7.35 and PaCO₂ >45 mmHg (Xia 2022).

Similarly, the severity of acidosis varied between trials. For example, one study enrolled participants with mild acidosis pH ≥7.30) (Pantazopoulos 2024), while other studies included patients with pH down to 7.25, capturing a moderately acidotic cohort but still not reflecting the most severe cases typically seen in clinical settings (Cortegiani 2020, Tan 2024).

Another important source of heterogeneity was baseline BMI differences, which may have subtly influenced outcomes. Median baseline BMI skewed higher in the intervention group approaching significance for the Italian study (HFNC 30.5 ± 8.7, n=40 versus NIV control 26.7 ± 5.5, n=39; p=0.0622) (Cortegiani 2020), while the opposite was observed at baseline for a Greek study (HFNC 25.9 ± 8.0, n=20 versus NIV 30.9 ± 8.5, n=20; p=0.06) (Papachatzakis 2020). These imbalances were unlikely to affect early gas exchange but could have contributed to longer-term treatment failure differences.

The timing of treatment failure also followed a consistent pattern across studies. Early gas exchange improvements within the first two hours were generally similar between HFNO and NIV (PaCO2 reduction at 2 hours: hNHF −6.8 ± 8.7; NIV −9.5 ± 8.5 mmHg, p=0.404), suggesting initial physiological equivalence (Cortegiani 2020). However, by six hours, 32% of patients initially treated with HFNO required transition to NIV (Cortegiani 2020).

When examining the individual trials, one study in patients with only mild hypercapnia (pH ≥7.35, PaCO₂ >45 mmHg) demonstrated very low intubation rates with no significant difference between HFNO (2.5%) and conventional oxygen therapy (0.6%, p=0.198). Treatment failure rates were similar (9.5% vs 12.8%, p=0.343), but HFNO was associated with a 14.6% increase in treatment costs (Xia 2022).

In moderate hypercapnia (PaCO₂ ≥50–55 mmHg, pH 7.25–7.35), the findings were less favourable. In one multicentre non-inferiority trial, HFNO achieved comparable short-term PaCO₂ reduction at two hours (−6.8 vs −9.5 mmHg, p=0.404; non-inferiority margin 10 mmHg, one-sided 95% CI up to 6.1 mmHg, p=0.0003), yet 32% of patients required escalation to NIV within six hours, whereas only a single NIV-treated patient progressed to intubation (Cortegiani 2020). Another trial in a similar population found higher treatment failure with HFNO compared with NIV (25.7% vs 14.3%, p=0.033) and higher intubation rates (14.2% vs 5.4%, p=0.026), although no significant differences were seen in hospital length of stay or 28 day mortality (Tan 2024).

In a mild-to-moderate hypercapnic cohort (pH ≥7.30, PaCO₂ >45 mmHg), treatment failure remained numerically higher with HFNO (19.6% vs 14.8%, not significant), but dyspnoea and discomfort scores favoured HFNO, suggesting potential tolerance advantages despite similar gas exchange outcomes (Pantazopoulos 2024). Another smaller trial also observed equivalent early PaCO₂ reductions between HFNO and NIV but reported that failure rates became more pronounced in the HFNO group by six hours, particularly in higher-risk patients (Papachatzakis 2020).
The Fahey meta-analysis pooled seven RCTs (n=30 – 168) and found trends slightly favouring HFNO for pCO₂, pH, pO₂, mortality, and risk of intubation, but none of these reached statistical significance. Switching to an alternate modality was higher with HFNO, although not significantly (Fahey 2023).

Overall, taking together the results of these studies, there was no clear benefit overall from the use of hNHF in patients hospitalised with mild-moderate hypercapnic exacerbations of COPD. However, these findings suggest that HFNO may be considered for patients with mild hypercapnia (PaCO₂ 45–55 mmHg) or in situations where NIV is poorly tolerated, as it offers better comfort and dyspnoea relief. However, in moderate hypercapnia (PaCO₂ ≥50–55 mmHg, pH 7.25–7.35), HFNO is associated with higher treatment failure and intubation rates compared with NIV (Tan 2024; Cortegiani 2020). BMI imbalances across study arms, while likely not influencing short-term gas exchange, may have contributed to longer-term treatment failure signals.

Severe hypercapnia, defined as pH <7.25 or PaCO₂ >70 mmHg, has not been studied in any of these trials, and NIV remains the standard of care for this population. The time-dependent pattern across trials indicates that HFNO may appear comparable initially (≤2 h) but loses efficacy by six hours, necessitating early reassessment and escalation if no improvement is observed (Papachatzakis 2020; Cortegiani 2020).