Of 13,689 abstracts screened, 114 abstracts from studies of patients with acute respiratory illnesses and 63 abstracts from studies of patients with chronic respiratory illnesses progressed to full-text assessment. 52 studies (18 RCTs, 13 crossover studies, 21 observational studies) described in 63 articles for acute respiratory illnesses(3–9, 28, 30–73), and 11 studies (5 RCTs, 3 crossover studies, 3 observational studies) for patients with acute on chronic respiratory illnesses were included(63–73) (Fig. 1).
Primary outcome
Invasive mechanical ventilation escalation
Need for IMV escalation was reported at variable illness time points (i.e. ranging from within 2 hours to the end of hospital admission) in 29 studies (15 RCTs and 14 observational studies, n = 4778 participants)(6, 8, 28, 30, 31, 34–38, 41–49, 52, 57, 61–63, 66, 67, 71–73) and 6 studies (4 RCTs, 2 observational studies, n = 548 participants)(63, 66, 67, 71–73) for acute and chronic respiratory illnesses respectively (Table 2). Overall need for IMV was defined in this systematic review as any need for IMV escalation, irrespective of the reported time point when IMV was considered.
Table 2
Timepoint at which need for IMV escalation was reported.
Timepoint at which need for IMV escalation was measured | Acute respiratory illnesses 29 studies (15 RCTs, 14 observational studies) | Chronic respiratory illnesses 6 studies (4 RCTs, 2 observational studies) |
2 hours | Total: 1 RCT (48) | - |
24 hours | Total: 4 studies 3 RCTs (8, 57, 71) 1 Observational study* (62) | Total: 1 RCT (71) |
48 hours | Total: 1 RCT (6) | - |
72 hours | Total: 4 studies 2 RCTs (36, 44) 2 Observational studies* (45, 62) | |
168 hours | Total: 1 Observational study* (45) | - |
28 days | Total: 5 studies 3 RCTs (34, 37, 42) 2 Observational studies (30, 47) | - |
At the end of hospital stay | Total: 15 studies 5 RCTs (35, 52, 66, 67, 73) 10 Observational studies (28, 31, 38, 41, 43, 46, 49, 61, 63, 72) | Total: 5 studies 3 RCTs (66, 67, 73) 2 Observational studies (63, 72) |
*Studies with IMV escalation reported on multiple timepoints: Xing 2021(62): 24 hours, 72 hours, Lee 2019(45): 72 hours, 168 hours. RCT: Randomised controlled trial. |
For acute respiratory illnesses, pooled data from 15 RCTs (n = 2239 participants)(6, 8, 34–37, 42, 44, 48, 52, 57, 66, 67, 71, 73) did not demonstrate a significant reduction in overall need for IMV escalation for HFNO when compared to COT (RR = 0.82, 95% CI = 0.65–1.05 p = 0.11, I2 = 23%) (Fig. 2). Similarly, pooled RCT data for IMV escalation at specified time points did not reveal any significant reductions for HFNO compared to COT (Supplementary Figs. 6–8).
Secondary analysis of combined data from all studies (RCTs and observational studies) demonstrated a significant reduction in need for IMV escalation in favour of HFNO compared to COT at 24 hours (RR = 0.32, 95% CI = 0.13–0.79; p = 0.01; 4 studies, n = 487 participants, I2 = 0%)(8, 57, 62, 71) and at 28 days (RR = 0.84, 95% CI = 0.73–0.96; p = 0.01; 5 studies, n = 1635 participants, I2 = 35%)(30, 34, 37, 42, 47) (Supplementary Figs. 6–7). However, overall need for IMV escalation remained unchanged with secondary analysis (RR = 0.86, 95% CI = 0.73–1.02; p = 0.08; 29 studies, n = 4778 participants, I2 = 64%)(6, 8, 28, 30, 31, 34–38, 41–49, 52, 57, 61–63, 66, 67, 71–73) (Supplementary Fig. 5a).
For chronic respiratory illnesses, pooled RCT data similarly did not demonstrate a significant reduction in the overall need to escalate to IMV for participants receiving HFNO compared to COT (RR = 0.86, 95% CI = 0.33–2.23; p = 0.76; 4 studies, n = 438 participants, I2 = 0%)(66, 67, 71, 73)(Fig. 2). Only one study on chronic respiratory illness reported need for IMV escalation at 24 hours (n = 37 participants)(71) with no between-group difference (HFNO: 5.6%, COT: 5.6%; p = 0.97). Notably, secondary analysis of combined data from all studies demonstrated a significant reduction in the overall need for IMV escalation for HFNO compared to COT (RR = 0.68, 95% CI = 0.48–0.97; p = 0.03; 6 studies, n = 548 participants, I2 = 0%)(63, 66, 67, 71–73) for patients with chronic respiratory illnesses (Supplementary Fig. 5b).
Secondary outcomes
Mortality
Pooled RCT data for acute respiratory illnesses reported no significant difference between in-hospital mortality for patients receiving HFNO compared to COT (RR = 1.00, 95% CI 0.85–1.17; p = 1.00; 5 studies, n = 1226 participants, I2 = 0%)(8, 35, 37, 44, 73) (Fig. 3). Similarly, there were no significant differences in short-term mortality at ≤ 30 days (RR = 1.02, 95% CI = 0.85–1.21; p = 0.86; 3 studies, n = 891 participants, I2 = 0%)(34, 36, 37) and long-term mortality at > 30 days (RR = 0.92, 95% CI = 0.65–1.29; p = 0.62; 3 studies, n = 1279 participants, I2 = 57%)(8, 37, 42) between patients receiving HFNO or COT for acute respiratory illnesses (Supplementary Figs. 10–11). Secondary analysis of pooled data from all study designs did not alter any of the mortality outcomes for patients with acute respiratory illnesses (Supplementary Figs. 9a, 10, 11).
For patients with chronic respiratory illnesses, only one RCT reported on in-hospital mortality which showed no significant difference between HFNO compared to COT (RR = 0.40, 95% CI = 0.04–4.10; p = 0.44; n = 45 participants)(73) (Fig. 3). No studies reported on short-term mortality or long-term mortality for chronic respiratory illnesses. Of note, secondary analysis of pooled data from all study designs for chronic respiratory illnesses demonstrated a significant reduction of in-hospital mortality (RR = 0.58, 95% CI = 0.37–0.92; p = 0.02; 3 studies, n = 155 participants, I2 = 0%)(64, 74, 75) for participants receiving HFNO compared to COT (Supplementary Fig. 9b).
Non-Invasive Ventilation escalation
Need for NIV escalation was reported in 10 studies(6, 8, 34, 42, 44, 48, 49, 57, 67, 73) for acute respiratory illnesses (9 RCTs)(6, 8, 34, 42, 44, 48, 57, 67, 73) and 2 studies for chronic respiratory illnesses (2 RCTs)(67, 73). Overall need for NIV escalation was pooled from need for NIV escalation measured at different timepoints for individual studies ranging from 24 hours to end of hospital stay.
Pooled RCT data for patients with acute respiratory illnesses revealed a statistically significant reduction in overall need for NIV escalation for HFNO compared to COT (RR = 0.62, 95% CI = 0.41–0.94; p = 0.02; 9 studies, n = 1289 participants, I2 = 40) (Supplementary Fig. 12a). Only one RCT reported on need for NIV escalation at 28 days. Need for NIV escalation at 24 hours was not statistically different compared to COT in patients with acute respiratory illnesses (Supplementary Fig. 13). Secondary analysis of combined data from all study designs showed similar results for overall need for escalation to NIV, with HFNO significantly reducing NIV escalation need compared to COT (RR = 0.63, 95% CI = 0.43–0.92; p = 0.02; 10 studies, n = 1391 participants, I2 = 33 ) in patients with acute respiratory illnesses (Supplementary Fig. 12a).
Pooled RCT data for patients with chronic respiratory illnesses did not reveal any significant differences in overall need for NIV escalation between HFNO when compared to COT (RR = 0.96, 95% CI = 0.17–5.26; p = 0.96; 2 studies, n = 365 participants, I2 = 63) (Supplementary Fig. 12b). No observational or crossover studies reported on need for NIV escalation for patients with chronic respiratory illnesses.
Inpatient admission rates
Of two RCTs examining patients with acute respiratory illnesses within an ED setting, there was no significant risk reduction in the need for inpatient admission for patients receiving HFNO compared to COT (RR = 0.88, 95% CI = 0.53–1.45; p = 0.61; n = 77 participants; I2 = 4%)(52, 71). However, secondary analysis of pooled data for all study designs revealed a small but significant risk reduction in inpatient admission for patients with acute respiratory illnesses receiving HFNO compared to COT (RR = 0.92, 95% CI = 0.85–0.99; p = 0.03; 3 studies, n = 198 participants; I2 = 0%)(43, 52, 71) (Supplementary Fig. 14).
For patients with chronic respiratory illnesses, hospital admission rates were only reported in one RCT which showed no significant differences between both groups (RR = 1.42, 95% CI = 0.48–4.22; p = 0.53; 1 study, n = 37 participants)(71).
Other hospitalisation-related outcomes
Only one small RCT reported on ICU LOS in patients with chronic respiratory illnesses which showed a significant reduction in favour of HFNO (mean difference= -1.58 days, 95% CI= -2.28 days to -0.88 days; p < 0.001; n = 45 participants)(73). This significant reduction remained with the inclusion of an additional observational study in secondary analysis(72) (Supplementary Fig. 18b).
Meta-analyses of RCTs for all other outcomes, including ICU admission rates, ICU mortality, hospital LOS, and ED LOS, were not significantly different between patients who received either HFNO or COT for both acute and chronic respiratory illnesses. The statistical significance of these outcomes remained unchanged for secondary analysis of combined data from all study designs (Supplementary Figs. 15–19).
Physiological and disability outcomes
Meta-analysis was not performed for change in PaCO2 on arterial blood gases or disability scores (including dyspnoea, comfort, and dryness scores) due to high methodological heterogeneity between studies (Supplementary Tables 11–14). Of the 10 RCTs, 6 (60%) reported no significant difference in change in PaCO2 when comparing HFNO to COT use in patients with acute respiratory illnesses(42, 44, 57, 66, 68, 71) (Supplementary Table 11a). All 4 RCTs that reported a significant difference in change in PaCO2 at one or more timepoints during the study were in favour of HFNO(6, 35, 67, 73). In patients with chronic respiratory illnesses, 3 of 5 RCTs (60%) reported no significant reduction in PaCO2 levels in patients receiving HFNO compared to COT(66, 68, 71) (Supplementary Table 11b). Both RCTs that reported a significant difference in change in PaCO2 at one or more timepoints were in favour of HFNO(67, 73). Secondary analyses of pooled data from all studies identified similar findings for patients with acute respiratory illnesses with most studies (20 of 30, 67%) reporting no significant difference in change in PaCO2 levels when comparing HFNO to COT(3, 5, 7, 31, 42–45, 49–51, 54, 56, 57, 60, 61, 65, 66, 68, 71). In contrast, only 4 of the 9 studies (44%) included in secondary analysis for chronic respiratory illnesses found no significant differences in change in PCO2 between HFNO and COT (65, 66, 68, 71). For all study designs, all of the studies which found a significant difference between HFNO and COT (10 studies for acute respiratory illnesses(6, 9, 35, 58, 62, 64, 67, 69, 70, 73), and 5 studies for chronic respiratory illnesses(64, 67, 69, 70, 73)), were in favour of HFNO as well.
Dyspnoea scores were most commonly measured using the Borg scale and visual analogue scale. While half of RCTs (5 of 10, 50%) reported statistically significant lower dyspnoea scores in patients receiving HFNO compared to COT in patients with acute respiratory illnesses(35, 42, 52, 68, 71), both RCTs for chronic respiratory illnesses (2 of 2, 100%) reported significantly lower dyspnoea scores in patients receiving HFNO compared to COT(68, 71). Results from secondary analysis of combined data from all studies differed slightly with 12 of 23 (52%) studies(7, 35, 42, 49, 52–55, 58, 64, 68, 71) and 3 of 5 (60%) studies(64, 68, 71) reporting significantly lower dyspnoea scores in patients receiving HFNO compared to COT for acute and chronic respiratory illnesses respectively (Supplementary Tables 12a-b).
Subgroup analyses: Types of respiratory illness
Only secondary analyses of pooled data from RCTs and observational studies were performed for patients with COVID-19 (1 RCT, 4 observational studies), as only one RCT (n = 22 participants) was available (33) reporting ICU LOS. Patients with COVID-19 receiving HFNO compared to COT had significantly reduced short-term mortality at ≤ 30 days (RR = 0.62, 95% CI = 0.48–0.79; 4 observational studies, n = 652 participants; p < 0.001; I2 = 0%)(28, 30–32), long-term mortality at > 30 days (RR = 0.67, 95% CI = 0.48–0.92; 2 observational studies, n = 517 participants; p = 0.01; I2 = 0%)(28, 30), ICU LOS (mean difference=-0.86 days, 95% CI=-1.59 days to -0.13 days; 2 observational studies and 1 RCT study, n = 203 participants; p = 0.02; I2 = 0%)(28, 31, 33), and need for IMV (RR = 0.72, 95% CI = 0.63–0.82; p < 0.001; 3 observational studies, n = 560 participants; I2 = 0%)(28, 30, 31) (Supplementary Figs. 20a-d).
There were no significant differences in outcomes between HFNO and COT in other sub-group analyses by respiratory illness, including COPD exacerbations, asthma exacerbations, and patients with type 2 respiratory failure (Supplementary Figs. 21–23).
Sensitivity analyses: Studies of low risks of bias
Sensitivity analyses were performed excluding studies with an overall high risk of bias, defined as: RCTs and randomised crossover studies with a “high” risk of bias using the ROB2 framework, non-randomised interventional studies with a serious risk of bias using the ROBINS-1 framework, or observational studies with scores < 6 using the Newcastle-Ottawa Scale.
In patients with acute respiratory illnesses, utilising only pooled RCT data from studies not at high risk of bias yielded the same findings for primary and secondary outcomes as reported with all RCTs. Overall need for IMV escalation remained similar for HFNO compared to COT (RR = 0.82, 95% CI = 0.65–1.05; p = 0.11; 14 RCTs, n = 2224 participants; I2 = 23%)(6, 8, 34, 35, 37, 42, 44, 48, 52, 57, 66, 67, 71, 73).
In this systematic review, HFNO conferred no significant reduction in overall need for IMV escalation or in-hospital mortality when compared to COT for both patients with acute and chronic respiratory illnesses in the primary analyses of RCTs only. However, there was a significant reduction in the overall need for NIV escalation for patients with acute respiratory illnesses compared to COT for patients with acute respiratory illnesses. Additionally, patients with COVID-19 who received HFNO experienced a significant reduction in need for IMV, mortality and ICU LOS, compared to COT.
Acute respiratory illnesses
Recent meta-analyses comparing IMV escalation need in patients with acute respiratory failure have generated mixed results with no significant differences between HFNO and COT in some reviews(10, 25, 26) and significant reductions in IMV escalation for HFNO in others(11, 21, 24). The 2021 European Respiratory Society clinical practice guidelines recommends HFNO over COT in patients with acute hypoxemic respiratory failure, based on pooled data from 12 RCTs and 4 cross-over studies showing a trend but not statistically significant reduction in escalation to IMV with the use of HFNO(26). Our meta-analysis adds to the mixed findings to date, with a reduction in overall IMV requirement that was not statistically significant when considering RCTs only but yielded a statistically significant reduction in IMV escalation at 24 hours and at 28 days when pooling RCTs and observational studies together in patients with acute respiratory illnesses receiving HFNO compared to COT. Furthermore, HFNO significantly reduced need for NIV escalation compared to COT in patients with acute respiratory illnesses. While this inconclusive result may relate to heterogeneity of study populations, methodologies and variation in outcome measurements and timing, there does appear to be a signal that HFNO is superior to COT regarding the need to escalate to IMV, and certain sub-populations of patients with acute respiratory illnesses, such as those with COVID-19, respond better to HFNO than others.
Importantly, similar to previous systematic reviews, we found no significant differences between HFNO and COT in terms of mortality benefit(11, 12, 16, 18–25), ICU admission rates(12), or ICU LOS(11, 17–19, 21, 24, 25) in patients with acute respiratory illnesses. The findings regarding mortality are likely related to the lack of effect of HFNO on overall need to escalate to IMV, as it is well recognised that once someone requires IMV for acute respiratory failure they are at increased risk of dying(38, 74, 75).
This is the first meta-analysis to compare the impacts on mortality and morbidity between HFNO and COT in patients with COVID-19, with substantial outcome benefits identified with HFNO. This evidence supports the increasing adoption globally of HFNO for patients with COVID-19. The pathological processes driving hypoxemia in people with COVID-19 and acute respiratory failure are complex and may involve diffuse interstitial and alveolar oedema, pulmonary endothelial injury, impaired lung diffusion capacity, and pulmonary perfusion abnormalities including impairment of the hypoxic pulmonary vasoconstriction response(76–79). Consequently, large ventilation perfusion mismatches may be seen in acute respiratory failure due to COVID-19 that necessitate higher oxygen flows and fractions of inspired oxygen delivery to the alveoli(76, 78). Compared to COT, HFNO provides higher oxygen flows to match COVID-19 patients increased respiratory flow demands(27). Moreover, HFNO provides greater reduction in inspiratory effort than COT(3, 27), which could reduce the extent of patient self-inflicted lung injury seen in COVID-19 ARF(27, 80). Importantly, with the advent of auto-titration nasal high flow systems that can titrate FiO2 to oxygen saturation targets(81) and thus reduce the need for healthcare workers to enter patients’ rooms to adjust settings, the delivery of HFNO to patients with highly infectious diseases such as COVID-19 may become safer. Notwithstanding the aforementioned, the findings for patients with COVID-19 must be interpretated with caution as pooled data were predominantly from observational studies and therefore subject to the inherent weaknesses of non-interventional studies.
Chronic respiratory illnesses
Although pooled RCT data for chronic respiratory illnesses did not reveal a significant reduction in need for IMV escalation or in-hospital mortality for patients receiving HFNO compared to COT, there was a significant reduction for both of these outcomes, in favour of HFNO, when including both RCTs and observational studies. This was likely due to the fact that 64% of escalations to IMV and 87% of reported in-hospital mortalities occurred in a single small observational study of lung transplant patients with ARF (n = 40 participants)(72). Importantly, findings for chronic respiratory illnesses were not generalisable to sub-populations of restrictive lung diseases like idiopathic pulmonary fibrosis (IPF), albeit patients with those illnesses were under-represented in this study. Nonetheless, a 2019 retrospective study has suggested mortality benefits from HFNO therapy in patients with acute exacerbations of IPF who remain hypoxemic despite COT(82). At the time of writing, no meta-analysis had reported mortality outcomes for HFNO compared to COT in patients with acute exacerbations of COPD, asthma, or interstitial lung disease.
Strengths and limitations
This is the first and largest systematic review to explore the impacts of HFNO compared to COT on morbidity and mortality in patients with a broad range of acute (including COVID-19) and chronic respiratory (including asthma, COPD, and lung cancer) conditions.
Limitations of this review include heterogenous timing of measurements for overall need for IMV escalation as this varied depending on time of treatment failure. Additionally, it is very likely that significant real-world heterogeneity existed between included trials as oxygen therapies were delivered under clinical settings and protocols that varied between studies. Secondly, the overall pooled population was small for certain outcomes, such as hospital admission rates for patients with respiratory illnesses. Thirdly, the risk of bias was moderate for most studies and there was high heterogeneity for some outcomes such as hospital LOS. Thus, large, high-quality RCTs are required to further increase the power of meta-analyses to address several key clinical questions that remain unanswered. Fourthly, it is challenging to blind participants and outcome assessors to the intervention (given the different sensation and delivery devices required for HFNO and COT), thus introducing bias, particularly for more subjective outcomes such as dyspnoea. Importantly, reporting regarding the causes of acute respiratory failure in individual studies was poor, with half of all patients (3035 of 6109, 50%) having no specified acute respiratory illness reported. Therefore, we were unable to further characterise the results by specific causes of ARF, which is crucial for understanding which patient populations with ARF are most likely to benefit from HFNO. It must also be highlighted that some of the meta-analyses were dominated by a few studies, including those performed for hospital mortality and need for IMV escalation, and the weaknesses of these dominant studies should be considered when interpreting the overall pooled results. Lastly, data from patients with a broad spectrum of respiratory conditions were pooled together to reflect real-world hospital setting for generalisability. While this introduces heterogeneity within the study population, it is important to examine HFNO’s clinical benefits compared to COT for patients with ARF secondary to heterogenous causes as HFNO becomes more universally adopted.