Early non-invasive ventilation for the management of patients with acute respiratory failure: a systematic review and meta-analysis

2.1 Study design and review question

We performed a systematic review and meta-analysis and reported the findings in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) guidelines (Supplementary Table 1) [14]. The Population, Intervention, Control, Outcome (PICO) framework was used to develop the review question: among patients with ARF (P), does early application of NIV (I), compared to standard treatment (C), reduce the intubation rate (O)?

The review protocol was registered in the prospective international register of systematic reviews (PROSPERO) (CRD420251153734).

2.2 Search strategy, study selection, and data extraction

Two trained investigators independently searched for pertinent peer-reviewed journal articles through web databases (PubMed, EMBASE via OVID, and the Cochrane Controlled Trials Register), from inception up to 04 September 2025. We included randomized and quasi-randomized studies (e.g., randomized by ambulance) published in peer-reviewed journals that compared the early application of NIV, defined as its application before or soon after hospital admission (including out-of-hospital, prehospital, and out of ICU scenarios), versus standard of care, used in adult patients with ARF. We considered NIV modalities including both Continuous Positive Airway Pressure (CPAP) and Bilevel Positive Airway Pressure (BIPAP) when delivered noninvasively; high-flow nasal cannula (HFNC) was not considered a NIV technique. Studies were eligible for inclusion only if they reported data on the primary outcome. Corresponding authors of selected publications were contacted by email up to three times to retrieve missing information. Since NIV is a well-established treatment for COPD exacerbations, we excluded studies only related to COPD exacerbation as the main cause of ARF. Studies with HFNC as comparator were also excluded.

After duplicates removal, the initial screening was conducted at the title and abstract level. The final selection of pertinent articles was conducted considering the full-text evaluation. In each phase, disagreements were solved by a third senior investigator.

2.3 Primary and secondary outcomes

The primary outcome of our systematic review and meta-analysis was the number of patients requiring endotracheal intubation during the index hospitalization. Secondary outcomes were:

- Mortality rate evaluated at multiple timepoints: short-term (≤30 days); medium-term (6–12 months); and at the longest follow-up available, as reported by each manuscript (Table 1, Ref. [13, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26]);

- Number of patients admitted to the ICU during the index hospitalization;

- Hospital length of stay.

We performed sensitivity analyses excluding quasi-randomized studies, excluding studies assessed as high risk of bias, and excluding studies that did not report the use of face mask interfaces. Finally, we performed planned subgroup analyses based on different NIV modalities (e.g., CPAP, BIPAP), type of provider (lay responders, physicians, nurses, paramedics), and etiology of ARF.

2.4 Risk of bias and certainty of evidence assessment

For each included trial, two authors independently assessed the risk of bias using the revised Cochrane risk of bias tool for randomized trials (RoB2), accessible online at https://www.riskofbias.info [27]. Disagreements were resolved through discussion with a third senior author to reach consensus. Risk levels were classified as “high risk of bias”, “some concerns”, or “low risk of bias”. Only if all domains were assessed as low risk of bias, we considered a trial as low risk of bias. We used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach to assess the overall quality of evidence for our primary and secondary outcomes [28]. Overall certainty of evidence was assessed as “high”, “moderate”, “low”, or “very low”. This information was summarized in the summary of findings table, which was created using the GRADEpro online tool (available at https://www.gradepro.org/).

2.5 Statistical analysis

Quantitative analyses were performed using the Cochrane Review Manager online version (https://revman.cochrane.org/). For dichotomous outcomes, relative risk (RR) and 95% confidence intervals (CIs) were calculated, whereas mean differences (MDs) with 95% CIs were used for continuous outcomes. We used an inverse-variance method with a random effect model, and results were summarized using forest plots. Heterogeneity analysis was performed with Cochran Q statistic and quantified with I² statistic. We interpreted the heterogeneity as outlined: 0% to 25% (might not be important); 25% to 50% (may represent moderate heterogeneity); 50% to 75% (may represent substantial heterogeneity); and 75% to 100% (would attest considerable heterogeneity). An unadjusted p-value of 0.05 was considered statistically significant. We calculated the number needed to treat (NNT) for primary and secondary outcomes with statistically significant differences between the two groups (intervention vs. usual care). Publication bias was evaluated using funnel plots for both primary and secondary outcomes.

Finally, we performed a trial sequential analysis (TSA) to assess the robustness of our findings, using TSA Viewer (Version 0.9.5.10 Beta. Copenhagen Trial Unit, Centre for Clinical Intervention Research, Rigshospitalet, Copenhagen, Denmark) and a two-sided alpha of 0.05, a power of 80%, an anticipated relative risk reduction of 40%, and a control event rate of 15%.

3.1 Study characteristics

Our search strategy (Supplementary material 1) identified a total of 782 records. After removing duplicates and conducting the initial title and abstract screening, 27 documents were selected and underwent the full-text assessment. Finally, 13 trials [13, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26] including a total of 2172 patients (1109 in the NIV group vs. 1063 in the usual care group) met the inclusion and exclusion criteria and were included in this analysis (Fig. 1).

Fig. 1.PRISMA flowchart of the literature search process. COPD: chronic obstructive pulmonary disease.

All included studies were published between 2000 and 2025; nine were conducted in Europe [13, 17, 19, 20, 21, 22, 23, 24, 26], two in North America [16, 25], and two in Australia [15, 18]. Twelve studies were randomized [13, 15, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26], and one was quasi-randomized (randomized by ambulance unit) [16].

Five studies investigated prehospital NIV in patients with acute cardiogenic pulmonary edema [15, 16, 17, 19, 22], while the remaining studies included patients with ARF due to any cause, including high-risk populations such as those with hematologic malignancies or following hematopoietic stem cell transplantation [24, 26]. Eight studies used CPAP as the only mode of NIV [15, 17, 18, 19, 20, 22, 24, 25]. All studies compared NIV (initiated prehospital or early outside the ICU) with standard oxygen therapy. Six studies were assessed at high risk of bias [15, 16, 20, 21, 22, 23], two as having some concerns [17, 18], and five as low risk of bias [13, 19, 24, 25, 26] (Supplementary Fig. 1). The characteristics of included studies are summarized in Table 1 and Supplementary Table 2, while Supplementary Table 3 reports general characteristics of the major excluded studies.

3.2 Primary and secondary outcomes

We found a significant reduction in rate of endotracheal intubation among patients receiving early NIV (prehospital or early ward-based) compared with standard oxygen therapy (56/1085 [5.2%] vs. 113/1047 [10.8%]; RR = 0.49; 95% CI, 0.36–0.66; p < 0.001; I² = 0%; 13 studies included) (Fig. 2). The NNT was 18.

Fig. 2.Effect of non-invasive ventilation on the rate of endotracheal intubation. CI: confidence interval; IV: inverse variance.

These findings were confirmed in all the subgroup and sensitivity analyses performed (Supplementary Figs. 2,3,4,5,6,7), with no significant difference between CPAP and BIPAP (interaction p = 0.73) (Supplementary Fig. 2) (Table 2). The funnel plot did not suggest the presence of small studies publication bias for the primary outcome (Supplementary Fig. 8). The TSA confirmed a conclusive benefit, as the total number of patients included in this meta-analysis (2132 for the primary outcome) exceeded the required information size (1152 patients) (RR = 0.49; alpha-spending adjusted 95% CI, 0.34–0.70) (Supplementary Fig. 9). The overall certainty of evidence for the rate of endotracheal intubation was assessed as moderate, specifically downgraded due to risk of bias (Supplementary Table 4).

Furthermore, our findings showed that early NIV reduced the ICU admission rate compared to the controls (129/831 [15.5%] vs. 153/798 [19.2%]; RR = 0.78; 95% CI, 0.62–0.97; p = 0.02; I² = 9%; with nine studies included) (Fig. 3) (Supplementary Fig. 10) (Table 2), with a NNT of 27.

Fig. 3.Effect of non-invasive ventilation on the rate of intensive care unit admission. CI: confidence interval; IV: inverse variance.

Early treatment with NIV did not significantly reduce mortality at the longest follow-up available compared with standard oxygen therapy (166/1109 [15.0%] vs. 182/1063 [17.1%]; RR = 0.80; 95% CI, 0.57–1.12; p = 0.19; I² = 51%; with 13 studies included) (Supplementary Figs. 11,12). The longest follow-up available for each study is reported in Table 1 (Table 1). Short- and medium-term mortality were also similar between the two groups (Supplementary Figs. 13,14) (Table 2).

There was no significant difference in hospital length of stay between groups (MD = −0.59 days; 95% CI, −2.18 to 1.00; p = 0.46; I² = 74%; with eight studies included) (Supplementary Fig. 15) (Table 2).

4.1 Key findings

Our systematic review and meta-analysis documented a substantial reduction in intubation rate in patients with ARF who were randomized to receive NIV either prehospital or out of the ICU, when compared to standard care. These results came from over 2000 patients randomized into 13 trials, with low heterogeneity observed. The TSA showed that we reached the required information size, confirming a conclusive benefit of NIV on endotracheal intubation and strengthening the robustness of our findings. Furthermore, early NIV was associated with a significant reduction in ICU admissions by more than 20%.

4.2 Relation to previous studies

Previous meta-analyses provided complementary insights that support our findings. In evaluating early NIV versus conventional oxygen therapy in immunocompromised patients with respiratory failure, authors found that early NIV significantly reduced intubation and ICU mortality rates, as well as ICU length of stay, although no clear benefit on all-cause mortality at the longest follow-up was observed [29, 30]. Another meta-analysis compared HFNC with NIV in patients with COPD exacerbation and hypercapnic ARF [31], demonstrating that HFNC was associated with a higher rate of treatment failure and treatment switch rates.

Our findings expand upon those by Scquizzato et al. [10]. First, we broadened the inclusion criteria, involving hospital general ward settings (out of the ICU). Our meta-analysis includes three additional randomized trials [13, 26, 28], and excludes one study focused on COPD exacerbation as the main cause of ARF [32], since the role of NIV in this population is well-established. Overall, the updated evidence comprises more than 500 additional patients. We also performed a TSA for the rate of endotracheal intubation, which confirmed the robustness of our results and a conclusive benefit. Finally, among secondary outcomes, we found a significant reduction in the rate of ICU admission, a finding not previously identified.

A recent position paper strongly recommended the application of early NIV by prehospital providers as an effective intervention to reduce intubation and mortality rates [9]. However, despite this evidence and formal endorsement, the actual implementation of NIV in the prehospital setting remains limited worldwide [9], likely due to the variability in training, protocols, and logistical barriers [33].

The exact timing of early NIV remains a critical and unresolved issue. Most trials did not precisely capture the interval between ARF onset and NIV initiation, representing a key limitation. Nevertheless, in patients with acute cardiogenic pulmonary edema, even a brief 15-minute delay in CPAP initiation worsened short-term mortality, emphasizing the potential importance of very early application [24]. Considering that the median time from EMS call to ambulance arrival is approximately 10–15 minutes, and transport to the hospital can take up to one hour (shorter in urban areas), the prehospital phase represents a valuable window of opportunity to initiate NIV and potentially influence outcomes [34].

Complementing these findings, the NAVIGATE trial [13] evaluated early NIV in patients with mild ARF admitted to non-ICU wards. This is the largest multicenter trial to date, demonstrating that early NIV reduced progression from mild to severe ARF, primarily by preventing severe respiratory distress and hypoxemia. The safety of early NIV was confirmed, as it did not increase complications such as pneumothorax, vomiting, or excessive airway secretions. Importantly, these benefits were consistent across prespecified subgroups, and early NIV did not result in excess adverse events.

Prehospital NIV was feasible across multiple randomized and pilot studies. CPAP was well tolerated in patients with acute cardiogenic pulmonary edema and ARF, with successful application in the majority of cases [25, 35]. Finally, a new CPAP device optimized for prehospital use was recently reported and may support broader implementation of early NIV strategies [36].

4.3 Significance of the study findings and contribution to existing knowledge

ARF remains one of the most frequent causes of critical illness and is associated with substantial morbidity and mortality, with in-hospital mortality rates up to 30% [4]. It frequently presents as a clinical emergency, requiring rapid intervention by EMS and subsequent advanced hospital care. Previous evidence from small randomized trials suggested that early initiation of NIV in the prehospital setting was effective in reducing dyspnea, improving oxygenation, and safe in terms of hemodynamic stability.

Our systematic review and meta-analysis confirms that the application of early NIV is generally safe and effective. Early NIV was associated with a substantial reduction in the risk of endotracheal intubation, with a 51% relative risk reduction and a NNT of 18. Avoiding invasive ventilation is a key goal in ARF management, as intubation is strongly associated with complications, especially in the prehospital setting, as well as prolonged hospitalization and adverse outcomes [2].

In our meta-analysis, NIV was safe with no significant differences in mortality (15.0% in the NIV group vs. 17.1% in controls). The frequent crossover to NIV in patients in control groups likely dilutes differences between treatment arms, potentially underestimating the true effect of early NIV.

4.4 Strengths and limitations

This systematic review and meta-analysis represents the most up-to-date synthesis of evidence on the use of early NIV outside of the ICU. The research question was well defined, following the established PICO framework. We reported our findings in accordance with the PRISMA guidelines, ensuring methodological transparency and reproducibility. The meta-analysis was conducted using a random-effects model, providing conservative estimates. We included more than 2000 patients and performed a TSA, which showed conclusive beneficial results for the primary outcome. Overall, the observed effects are physiologically plausible and in line with previous findings in ICU and other settings, reinforcing their external validity.

However, some limitations should be acknowledged. One of the included studies was quasi-randomized, potentially introducing selection bias, although it was excluded in a sensitivity analysis without affecting the main findings. Furthermore, several studies were assessed as having a high risk of bias, which is common in trials conducted in prehospital settings. Furthermore, although none of the patients included in these trials were in the ICU, there was heterogeneity between in-hospital and out-of-hospital settings, which may affect generalizability.

4.5 Future perspective

The heterogeneity of trial designs, EMS systems, NIV modalities, and patient populations contributes to ongoing uncertainty and underscores the need for further high-quality research. A RAND/UCLA consensus across experts could help define which patients, clinical indications, and NIV modalities would provide the greatest benefit, while large multicenter trials remain essential to establish definitive evidence.

However, our meta-analysis suggests that all out-of-hospital emergency services should be equipped and have the expertise to deliver NIV, thus implementing its routine use in the early management of patients with ARF with the goal of reducing the need for endotracheal intubation.