Article Text

Effect of high-flow nasal cannula versus non-invasive ventilation after extubation on successful extubation in obese patients: a retrospective analysis of the MIMIC-IV database
  1. Yun Ge1,
  2. Zhenxuan Li2,
  3. Ao Xia1,
  4. Jingyuan Liu1 and
  5. Dongmin Zhou1
  1. 1Department of Critical Care Medicine, The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital, Zhengzhou, Henan, China
  2. 2Department of Thoracic Surgery, The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital, Zhengzhou, Henan, China
  1. Correspondence to Dongmin Zhou; 13937120236{at}


Background The pathophysiological characteristics of the respiratory system of obese patients differ from those of non-obese patients. Few studies have evaluated the effects of high-flow nasal cannula (HFNC) and non-invasive ventilation (NIV) on the prognosis of obese patients. We here compared the effects of these two techniques on the prevention of reintubation after extubation for obese patients.

Methods Data were extracted from the Medical Information Mart for Intensive Care database. Patients who underwent HFNC or NIV treatment after extubation were assigned to the HFNC or NIV group, respectively. The reintubation risk within 96 hours postextubation was compared between the two groups using a doubly robust estimation method. Propensity score matching was performed for both groups.

Results This study included 757 patients (HFNC group: n=282; NIV group: n=475). There was no significant difference in the risk of reintubation within 96 hours after extubation for the HFNC group compared with the NIV group (OR 1.50, p=0.127). Among patients with body mass index ≥40 kg/m2, the HFNC group had a significantly lower risk of reintubation within 96 hours after extubation (OR 0.06, p=0.016). No significant differences were found in reintubation rates within 48 hours (15.6% vs 11.0%, p=0.314) and 72 hours (16.9% vs 13.0%, p=0.424), as well as in hospital mortality (3.2% vs 5.2%, p=0.571) and intensive care unit (ICU) mortality (1.3% vs 5.2%, p=0.108) between the two groups. However, the HFNC group had significantly longer hospital stays (14 days vs 9 days, p=0.005) and ICU (7 days vs 5 days, p=0.001) stays.

Conclusions This study suggests that HFNC therapy is not inferior to NIV in preventing reintubation in obese patients and appears to be advantageous in severely obese patients. However, HFNC is associated with significantly longer hospital stays and ICU stays.

  • Assisted Ventilation
  • Critical Care
  • Non invasive ventilation

Data availability statement

Data are available in a public, open access repository. The datasets analysed during the current study are available in the MIMIC-IV database (

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Few studies have evaluated the impact of high-flow nasal cannula (HFNC) therapy and non-invasive ventilation (NIV) on the short-term prognosis of obese patients after extubation.


  • In an obese population, the effects of HFNC therapy on the prevention of reintubation after extubation were not inferior to those of NIV. HFNC therapy demonstrated a significant advantage for severely obese patients.


  • Our study suggests that HFNC therapy is an effective treatment to prevent reintubation in the obese population. It provides further information for clinicians in making decisions about oxygen therapy modalities.


The main forms of non-invasive respiratory support after extubation include routine oxygen therapy, non-invasive ventilation (NIV) and high-flow nasal cannula (HFNC) therapy. In HFNC therapy, oxygen is delivered through a nasal cannula. This therapeutic modality not only ensures a constant and adjustable concentration of inhaled oxygen delivered at high flow rates but also provides low-level continuous positive airway pressure,1 increases end-expiratory lung volume2 and decreases the work of breathing. Additionally, the high gas flow velocity has a flushing effect, reducing the nasopharyngeal dead space.3 The inhaled gas is also humidified and heated, which increases the patient’s comfort levels. This may attenuate the airway inflammatory response4 and promote clearance of respiratory tract secretions.5 Therefore, HFNC therapy may serve as an alternative to non-invasive positive-pressure ventilation.

Previous studies have shown that, compared with routine oxygen therapy, HFNC therapy effectively enhances postextubation oxygenation, thereby reducing the incidence of respiratory failure recurrence and reintubation rates of critically ill patients and patients with low reintubation risk.6 However, its use in patients with high reintubation risk, such as obese patients, remains unsupported.

The rate of obesity among patients in the intensive care unit (ICU) is as high as 20%.7 The increased oxygen consumption and work of breathing in obese patients result in oxygenation reduction with a body weight increase. Under resting-state conditions, oxygen consumption is 1.5 times higher in obese than in non-obese individuals.8

Obese individuals are also prone to atelectasis, as the weight of the chest wall and a large amount of abdominal fat compress the lungs and reduce pulmonary compliance, thereby reducing functional residual capacity and arterial oxygenation. The supine position, which is usually adopted by patients during mechanical ventilation, also further aggravates atelectasis.

A case–control study of obese and non-obese patients receiving mechanical ventilation revealed that obese patients were at higher risk for difficult tracheal intubation (15% (obese) vs 6% (non-obese)) and stridor after extubation (15% (obese) vs 3% (non-obese)).9

The main challenge faced by clinicians after extubation of mechanically ventilated obese patients is the optimisation of postextubation oxygen therapy regimens, based on the pathophysiological characteristics of the respiratory system of obese patients, in order to reduce the risk of reintubation and improve patient outcomes. In recent years, HFNC therapy has received widespread attention due to its higher level of comfort than that of NIV.10 Some studies have found no significant differences in effectiveness between HFNC therapy and NIV.11 However, the effectiveness of HFNC therapy compared with NIV in obese patients remains unclear.

Therefore, we here compared the effects of HFNC therapy and NIV in preventing the need for reintubation after extubation in obese patients.


Study population and data extraction

Data were obtained from the Medical Information Mart for Intensive Care (MIMIC-IV) database,12 which is a critical care database developed and maintained by the Massachusetts Institute of Technology Laboratory for Computational Physiology. MIMIC-IV is a large, freely accessible single-centre critical care database containing comprehensive and high-quality data of patients hospitalised at the ICUs of Beth Israel Deaconess Medical Center between 2008 and 2019. One of the researchers was allowed to access and download data from the database after completing the Protecting Human Research Participants training. Data were extracted from the MIMIC-IV 1.0 database using PostgreSQL.

Study design

The data of patients who received invasive mechanical ventilation and for whom extubation had been planned were extracted and screened from the downloaded data. The inclusion criteria were as follows: obese patients (body mass index (BMI) ≥30 kg/m2 or patients diagnosed with obesity in accordance with the 9th and 10th revisions of the International Classification of Diseases (ICD-9, ICD-10)); aged ≥18 years; who underwent HFNC therapy or NIV within 6 hours after extubation. Given that clinicians always used to give conventional oxygen therapy first after extubation, and then switch to HFNC or NIV immediately after finding that it was not applicable. If only these patients were recruited, a large portion of the population would be missed. Therefore, for the real-world data, we set relatively broad inclusion criteria and chose to include patients who received HFNC or NIV within 6 hours rather than just immediately after extubation. For patients with multiple ICU stays, only the data of the first hospital admission and first ICU stay were included. For patients with multiple extubations, only the data related to the first extubation were included. We excluded patients with a length of ICU stay <24 hours; tracheotomy and death occurred within 1 hour after extubation without reintubation (ie, patients in whom treatment had been abandoned).

Patients were grouped as follows: The HFNC group included patients who received postextubation continuous or intermittent HFNC therapy until improvement or deterioration (following which conversion to NIV or reintubation was performed) occurred. The NIV group included patients who underwent continuous or intermittent NIV postextubation until improvement or deterioration (following which reintubation was performed) of the patient’s condition.

Baseline data

The following baseline patient data were collected: demographic data, anthropometric data, comorbidities, Simplified Acute Physiology Score (SAPS) II on ICU admission and diagnostic information. Other data collected include vital signs, inhaled oxygen concentration setting of the ventilator, tidal volume and arterial blood gases prior to removal of the tracheal catheter. The arterial oxygen tension/fractional inspired oxygen (FiO2) ratio was calculated. The use of vasopressors, such as epinephrine, norepinephrine, phenylephrine, vasopressin and dopamine, prior to extubation, was also recorded.


The primary outcome was reintubation within 96 hours after extubation. Secondary outcomes were reintubation within 48 hours, reintubation within 72 hours, in-hospital reintubation, length of stay in ICU, length of hospitalisation, in-ICU mortality and in-hospital mortality. Other outcomes included the time from extubation to reintubation, the number of patients in the HFNC group converted to NIV after deterioration and the number of HFNC patients subsequently reintubated.

Statistical analysis

Non-inferiority analyses were performed in this study. According to previous studies, the reintubation rate was about 10%–15% in obese patients treated with NIV after extubation,13 14 and about 10%–20% treated with HFNC.11 15 16 We hypothesised that the preset non-inferiority margin for the HFNC group was 10% based on considerations of clinical relevance and previous studies.6 17 The sample size of 757 patients was determined sufficient to support the unilateral 95% CI analysis with a statistical power of 80%, and a maximum tolerated patient loss rate of 15%.

Statistical analysis was performed using R Studio (R V.4.2.1). Categorical data were expressed as number and percentage (no (%)), normally distributed measurement data were expressed as mean±SD, and non-normally distributed data were expressed as median (Q1, Q3). Multiple interpolation was used to fill in the missing values. Intergroup comparisons were performed using the t-test, χ2 test or Mann-Whitney U test, as appropriate. Two-tailed tests were adopted, with differences considered statistically significant when p<0.05.

The doubly robust estimation model was used to explore the relationships between exposure and the primary outcome variable. Doubly robust estimation combines the inverse probability of treatment weighting (IPTW) model targeted towards exposure and a regression model targeted towards outcomes to analyse causal effects. When the two models are used singly, the results are unbiased only when statistical models are correctly specified. However, the doubly robust estimation model only requires one of the two models to be correct for the generation of unbiased results. A series of sensitivity analyses were performed to assess the robustness of the findings. We also tested the interaction between treatment and the degree of obesity of the patient and so on.

Propensity scores were calculated using the multivariate logistic regression model, with a matching ratio of 1:1 and a calliper of 0.01. The IPTW model was constructed using weights based on the propensity scores and was used for further multivariate logistic regression analysis of the weighted cohort. Covariates included in the model were variables with residual imbalance after IPTW. Standardised mean differences were calculated to evaluate the validity of IPTW. Propensity score matching (PSM) was also adopted to balance baseline characteristics between the two groups.


Baseline characteristics

A total of 23 337 planned extubation records from the MIMIC-IV database were screened. Based on the inclusion and exclusion criteria stated above, 757 cases were ultimately included in this study (figure 1). The included patients had a median age of 65 (57, 73) years. Overall, 34.4% of the patients were women, and the median BMI of the patients was 36.8 (33.2, 43.2) kg/m2. There were 282 patients in the HFNC group and 475 in the NIV group. The median SAPS II at ICU admission was 39 (32, 47) points. A total of 97 people were eventually reintubated, 52 in the HFNC group and 45 in the NIV group.

Figure 1

Study flow chart. BMI, body mass index; HFNC, high-flow nasal cannula oxygen; NIV, non-invasive ventilation; PSM, propensity score matching.

Table 1 shows the baseline characteristics of the two groups of patients. There were statistically significant differences in BMI, certain comorbidities (hypertension, coronary heart disease, chronic obstructive pulmonary disease (COPD), diabetes, cancer) and SpO2, tidal volume, arterial blood gases and PO2/FiO2 ratio prior to extubation. After propensity score-based IPTW.18 19 The majority of covariates were balanced between the two groups, but age remained significantly different (table 2). Figure 2 shows the comparison of covariate differences before and after adjustment using IPTW.

Table 1

Baseline characteristics of the original cohort

Table 2

Baseline characteristics of the cohort after IPTW based on propensity scores*

Figure 2

Standardised mean difference before and after IPTW. IPTW represents the variables of the cohort weighted by the inverse probability. Original represents variables of the original cohort. BMI, body mass index; COPD, chronic obstructive pulmonary disease; FiO2, fractional inspired oxygen; IPTW, inverse probability of treatment weighting; PCO2, partial pressure of carbon dioxide; PO2, partial pressure of oxygen.

Primary outcome

Outcome regression was combined with IPTW using the doubly robust estimation method20 for the construction of a regression model. Doubly robust estimation indicated that the risk of reintubation within 96 hours after extubation did not differ significantly between the HFNC and NIV groups (table 3) (OR 1.50, 95% CI 0.88 to 2.55, p=0.127 after adjustment for unbalanced covariates; OR 1.45, 95% CI 0.84 to 2.50, p=0.180 after adjustment for all covariates). To ensure the robustness of results, we performed a logistic regression analysis of multiple models, including post-PSM logistic regression analysis as well as logistic regression of the original cohort. The same conclusion was obtained with all models: compared with the NIV group, the HFNC group did not have a significantly elevated risk of reintubation within 96 hours after extubation (table 3). After PSM, the difference in rates of reintubation within 96 hours after extubation of the HFNC and NIV groups was not statistically significant (17.5% vs 14.3%, p=0.533; table 4).

Table 3

Analysis using different models of the primary outcome

Table 4

Secondary outcome analysis after PSM

Secondary outcomes and other outcomes

The outcomes of the two groups were compared after PSM had been performed, which yielded 154 matched patient pairs. The baseline characteristics of the two groups after PSM were shown in online supplemental table S1. The rates of reintubation within the various time frames after extubation did not differ significantly between the matched HFNC and NIV groups. Rates of reintubation within 48 hours after extubation were 15.6% vs 11.0% (p=0.314), that within 72 hours after extubation were 16.9% vs 13.0% (p=0.424), while rates of in-hospital reintubation were 20.1% vs 15.6% (p=0.372). The time from extubation to reintubation of the HFNC group was longer than that of the NIV group (13.7 hours vs 9.4 hours, p=0.002). The length of hospital stay and length of ICU stay were significantly longer in the HFNC than in the NIV group. The lengths of hospital stay and ICU stay were 14 days vs 9 days (p=0.005), and 7 days vs 5 days (p=0.001), respectively. However, in-hospital mortality (3.2% vs 5.2%, p=0.571) and ICU mortality (1.3% vs 5.2%, p=0.108) did not differ significantly between the two groups (table 4). Forty-two patients of the HFNC group of the original cohort were converted to NIV. Among these patients, 10 were reintubated, resulting in an in-hospital reintubation rate of 23.8%.

Subgroup analysis

Figure 3 shows that the presence or absence of COPD (p for interaction=0.218) and patient age (p for interaction=0.233) did not interact significantly with the primary outcome in either group. However, significant interactions were found between BMI and the primary outcome (p for interaction=0.009). Patients with more severe obesity tended to benefit more from postextubation HFNC therapy and had a lower rate of reintubation within 96 hours after extubation. For patients with BMI ≥40 kg/m2, the risk of reintubation within 96 hours after extubation was significantly lower in the HFNC group than in the NIV group (OR 0.06, 95% CI 0 to 0.39, p=0.016).

Figure 3

Logistic regression analysis of the primary outcome according to the predefined strata. BMI, body mass index; COPD, chronic obstructive pulmonary disease; HFNC, high-flow nasal cannula; NIV, non-invasive ventilation.


Critically ill patients usually have complex and varying health conditions that necessitate the consideration of many factors during treatment. The same diagnostic and treatment techniques may not apply equally in different populations.21 For instance, diagnostic and treatment techniques that bring benefit to the general population may not be equally beneficial in specific populations, such as those with obesity. In this comparison between the effects of HFNC therapy and those of NIV on the need for reintubation of obese patients after weaning from invasive mechanical intubation, the doubly robust analysis found no significant differences in the need for reintubation within 96 hours postextubation between the two groups. However, among patients with BMI ≥40 kg/m2, those who received HFNC therapy had a significantly lower risk of reintubation within 96 hours postextubation than those who received NIV.

We defined reintubation within 96 hours after extubation as treatment failure because of the following reasons: if the observation time is excessively short, then extubation failure is affected to a greater extent by factors such as erroneous indications for extubation and associated to a lesser extent with the selection of oxygen therapy measures after extubation, and more than 80% of reintubations occur within 96 hours after extubation.16 22 23 A multicentre study conducted in the USA showed that the choice of a time cut-off of 96 hours consistently caught about 90% of reintubation events in the ICU over a 12-year period, indicating that it was a stable indicator, and the researchers recommended that the 96 hours time cut-off be used as the criteria for reporting reintubation rates.22

Previous research data have shown that the rate of reintubation in ICU patients after weaning from mechanical ventilation is approximately 10%,22 while the reintubation rates of patients with a low risk of extubation failure are within the range of 5%–13%.24 25 Certain studies have reported that patients with a high risk of reintubation, who receive routine oxygen therapy, have reintubation rates of up to 22%–24%.17 26 Our results indicated that the overall in-hospital reintubation rate and reintubation rate within 96 hours after extubation in patients in the HFNC and NIV groups were 12.8% and 11.6%, respectively, in agreement with previous studies.

Our results demonstrated that in an obese population, the effect of HFNC therapy in preventing reintubation following extubation was not inferior to that of NIV. The prophylactic use of NIV after extubation in obese patients was generally considered beneficial because positive pressure ventilation can prevent alveolar collapse, improve lung inflation and reduce the inspiratory threshold load in obese patients.13 However, poor patient tolerance has always been a major challenge in NIV treatment. A previous study found that 29% of NIV failures were due to treatment intolerance, which was significantly higher than the 4% rate of HFNC.27 Patients often remove the mask or interrupt respiratory support due to claustrophobia, coughing up sputum, drinking or eating when treated by NIV.28 However, HFNC can generally be used continuously for several days because of its comfort.29 Compared with NIV, HFNC only creates a lower level of positive airway pressure through nasopharyngeal and airway resistance to high flow gases.29 30 Nevertheless, it provides appropriate temperature and high humidity gas, which can promote mucociliary clearance, thus improving small airway function and decreasing airway resistance in obese patients.31 32 Moreover, Corley et al33 found that HFNC increased end-expiratory volume in subjects, especially in obese subjects. A post hoc analysis of a large trial showed that NIV was not superior to HFNC in obese patients after thoracic surgery (mean BMI of 34 kg/m2), with 15% and 13% treatment failure in the NIV and HFNC groups, respectively.32

This study also suggests that HFNC therapy results in a lower risk of reintubation in severely obese patients compared with NIV. Previous studies on the application of high-flow oxygen therapy in severely obese patients were limited. A randomised controlled trial indicated that HFNC improves the efficiency of apnea oxygenation in morbid obesity (BMI ≥40 kg/m2),34 in the case of apnea, the flow rate and the proximity of fresh gas to the airway epithelium both affect the efficiency of respiratory oxygenation.35 In addition, we speculated that severe obesity might be associated with limited treatment adherence and a higher propensity for NIV intolerance based on clinical experience. In contrast, HFNC treatment provides a higher level of comfort. However, more studies are needed to support this idea.

Oxygen supplementation through HFNC or NIV may significantly improve the oxygen saturation of patients compared with conventional oxygen therapy, as a result, the time from extubation to reintubation will be prolonged compared with conventional oxygen therapy. HFNC, in particular, improves patient tolerance to respiratory failure due to its comfort. One study by Sztrymf et al has stated that patients who are not intubated tolerate HFNC for long periods.36 A randomised controlled trial showed that the time from extubation to reintubation was longer in patients receiving HFNC than in those receiving NIV,15 which is consistent with our study.

The duration before reintubation of the HFNC group was significantly longer than that of the NIV group. This may be due to the higher level of comfort and improvement in blood gas indicators associated with HFNC therapy, which may have masked the deterioration in patient condition to a certain extent and thus led to a delay in reintubation. Previous studies have demonstrated that delayed intubation may increase the risk of adverse outcomes, including death,37 but this was inconsistent with the results of this study.

The lengths of hospital stay and ICU stay of the HFNC group were significantly longer than those of the NIV group in this study. This may be partly attributed to delayed reintubation in the HFNC group. However, prolongation of hospital and ICU stay inevitably increases the costs of hospitalisation, which is an issue that requires consideration.

This study had certain limitations. The data used in this study were obtained from electronic health records. Although the majority of high-risk factors for reintubation have been included in our analysis, some factors were excluded due to missing or omitted data. Moreover, adjustments to the data based on the year of hospitalisation have not been made. Since the database spans a relatively long period, treatment methods may have evolved, but this was not taken into consideration. Furthermore, this was a single-centre retrospective study. Although biases have been reduced to a certain extent by doubly robust estimation, future multicentre, prospective studies with a large sample size are still required to validate our findings. Also noteworthy, we chose to include patients who received HFNC or NIV within 6 hours rather than just immediately after extubation, which means that a portion of patients might receive the two treatments because of developing respiratory distress after extubation, therefore, this study only examined the effects of HFNC and NIV on the prevention of extubation failure, but cannot represent the effects of the two treatments on the prevention of respiratory failure.

This retrospective observational study showed that, in an obese population, the effect of HFNC therapy in preventing reintubation after extubation was not inferior to that of NIV. Additionally, HFNC therapy demonstrated an advantage in severely obese patients. However, it is noteworthy that HFNC led to significantly longer hospital stay and ICU stay than NIV in obese subjects.

Data availability statement

Data are available in a public, open access repository. The datasets analysed during the current study are available in the MIMIC-IV database (

Ethics statements

Patient consent for publication

Ethics approval

The MIMIC-IV database is an anonymous public database that was created with the approval of the Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center. The investigators were granted access to the original database and data acquisition. The study protocol was approved by the Ethics Committee of Henan Cancer Hospital (approval number: 2022-475-001).


The authors would like to thank Changpeng Liu from the Medical Records Office of Henan Cancer Hospital for advice on the statistical analysis of this study.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Contributors YG was responsible for the conception and design of the study, as well as data extraction and drafting and revision of the manuscript. AX and JL were responsible for the collection of the literature and its systematic analysis. ZL performed data processing and statistical analyses. DZ was involved in the conception and design of the study and critically revising important elements of the manuscript. All authors read and approved the final manuscript. YG is responsible for the overall content of the manuscript as the guarantor.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.