Bronchiolitis is the leading cause of hospital admission in infants aged less than 1 year. Approximately 10–15% of hospitalised patients eventually require admission to paediatric intensive care units (PICUs). Supportive care is an essential part in the management of patients with bronchiolitis, including maintenance of appropriate nutrition and hydration and an adequate oxygen saturation.1 In this context, high-flow oxygen therapy delivered through nasal cannula (HFNC) was a modality initially used in the PICU setting that, with the development of inpatient paediatrics as a specialty, has started to be used in paediatric wards with positive outcomes. At present, HFNC is considered a safe and effective therapy for the management of bronchiolitis in inpatient wards.2

High-flow oxygen therapy acts by delivering an increased volume of air and oxygen to the airways using a higher flow of humidified and heated gas, which increases the inspiratory flow and delivers oxygen more efficiently to the alveoli, thus decreasing the work of breathing and improving gas exchange.3–5 Although there are standardised protocols for its administration, some aspects continue to be debated, such as the optimal initial flow of oxygen or the best timing for initiating HFNC.

There is considerable variability in the flow rate applied at initiation of HFNC: in most studies, it has been determined based on body weight, but other authors report its determination based on clinical severity or the judgment of the clinician.6 Some studies suggest that a flow rate of 2L/kg/min can achieve a pharyngeal pressure of 4cmH2O or greater, but the maximum flow that can be used in the management of infants with HFNC has not been established.5 After 7 years of clinical experience using this treatment in the paediatric ward, we consider that many infants and toddlers could benefit from the use of higher flow rates from the beginning of treatment.

The aim of our study was to assess the safety of HFNC for management of bronchiolitis in the paediatric ward and to analyse whether different initial flow rates (10 vs 15L/min) were associated with differences in the effectiveness of HFNC in terms of clinical severity and the probability of treatment failure.

During the 2014–2016 period, we performed a prospective observational and analytical clinical cohort study in children aged less than 24 months hospitalised due to bronchiolitis who required HFNC. The setting of the study was a tertiary hospital with a general paediatrics ward with 38 beds that manages 1750 admissions a year. We defined bronchiolitis as a first or second acute episode of wheezing preceded by manifestations of viral respiratory illness in children aged less than 2 years.7 Bronchiolitis was the leading cause of admission in this age group during the period under study.

We used consecutive sampling to include patients aged 1–24 months admitted with bronchiolitis to the paediatric ward that required HFNC and for whose participation we obtained a signed written consent from the parents. We excluded newborns, children admitted directly to the PICU, children in whom HFNC was prescribed in the emergency department of our hospital and were admitted with the aforementioned inclusion criteria, children in whom HFNC was initiated at a different hospital, and children in whom HFNC was contraindicated due to radiographic evidence of pneumothorax, facial malformations or a tracheostomy tube.

Patients admitted to the paediatric ward were evaluated by general hospital paediatricians, and if HFNC was indicated, were included in the HFNC protocol. We assessed the severity of illness by means of the Wood-Downes-Ferres (WDF) score8 (Table 1). The criteria for initiation of HFNC in the paediatric ward were: worsening of breathing difficulty as established by an increase in the WDF score to 8 points or higher, a decrease in oxygen saturation (SatO2) to less than 92% despite use of conventional oxygen therapy, or the presence of 2 or more apnoeic pauses per hour.

In our usual clinical practice, some patients start HFNC with a flow rate of 10L/min and others with a flow rate of 15L/min, depending on the prescribing physician, so patients were assigned to one of 2 initial flow rate groups based on the judgment of the physician that was in charge at the ward at the time of prescription. In both groups, the fraction of inspired oxygen (FiO2) was titrated to achieve a SatO2 of 92% or greater with humidified air at a temperature of 35–37°C. Thus, 2 groups were defined for analysis based on the initial flow rate: the HFNC-15 group, with an initial flow of 15L/min, and the HFNC-10 group, with an initial flow of 10L/min. High-flow oxygen therapy was delivered with the Fisher & Paykel® MR850® humidification system and Fisher & Paykel Optiflow Junior nasal cannulas in infant size (OPT-316).

We defined treatment failure as:

• -

  Presence of apnoeic pauses after initiation of HFNC, or

• -

  Increase in RR, HR or WDF score or decrease in SatO2 in the first 12–24h, or

• -

  Plateau in RR, HR, SatO2 or WDF score values (or improvement by less than 15% the baseline value).

We collected clinical and epidemiological data, including age, weight, gestational age at birth, personal medical history and results of the rapid respiratory syncytial virus (RSV) test. We designed a data collection form for the documentation during treatment with HFNC of the following variables: body temperature, respiratory rate (RR), heart rate (HR) and SatO2. These parameters were recorded at the time HFNC was prescribed, at 1, 3, 6, 12 and 24h of treatment, and thereafter once a day until treatment was discontinued. The physicians conducting the study administered the WDF scale on a daily basis. We collected data on all associated treatments, admission to the PICU, complications and tolerability of HFNC.

We performed the statistical analyses with the statistics software SPSS version 21.0 (IBM SPSS Statistics; Armonk, NY, USA) and GraphPad Prism version 7 (La Jolla, CA, USA). We have expressed quantitative variables as median and interquartile range, and qualitative variables as percentages. In the comparative analysis, we used the Fisher test to compare qualitative variables, and for quantitative variables we used the Mann–Whitney U and the Kruskal–Wallis tests for unpaired samples and the Friedman test for paired samples. To analyse the time elapsed to treatment failure, we used Kaplan–Meier survival curves and the Mantel–Cox test.

To analyse the factors that could have an influence on treatment failure and admission to the PICU, we fit 2 logistic regression models where we included all variables that were statistically significant in previous analysis or that are clinically relevant, including age, underlying disease, weight, WDF score and initial flow (HFNC-10 vs HFNC-15).

We present the results of the logistic regression analysis in terms of odds ratios (OR) and 95% confidence intervals (CIs). To select and compare regression models, we used the Akaike informative criterion (AIC) and the Bayesian information criterion (BIC). We chose the model with the lowest AIC and BIC values, which are indicative of a better fit. We ruled out the possibility of colinearity in the models. Statistical significance was defined as a P-value of less than .05. The study was approved by the Clinical Research Ethics Committee of the Hospital Gregorio Marañón (File 57/18).

The main objective of this study was to identify the optimal initial flow rate for HFNC for management of bronchiolitis in the paediatric ward, and to assess the safety of this oxygen therapy modality in this setting. First of all, we confirmed that the use of HFNC for bronchiolitis in the paediatric ward is an effective measure that improves severity parameters in the first hours of treatment and is also safe, as we observed no relevant side effects; our findings also suggest that an initial flow rate of 15L/min is associated with a more rapid improvement of the RR and a lower proportion of treatment failure compared to an initial flow rate of 10L/min (Table 4).

A growing number of articles confirm the benefits of HFNC in the paediatric ward setting for management of different respiratory illnesses, such as bronchiolitis, asthma or laryngitis, and respiratory failure from other causes.2,9–13 However, some aspects are still under debate, such as the optimal initial flow rate or the best time to initiate treatment, which is evinced by the wide variability found in the HFNC protocols.8

Due to the lack of consensus on the calculation of the initial flow rate, some authors calculate it based on clinical criteria such as age, weight or a clinical severity score. Some studies have used flows between 2 and 9L/min that were subsequently adjusted based on the work of breathing and SatO2. In many studies, the initial flow was calculated based on body weight, usually for delivery of 2L/kg/min, which appears to achieve adequate alveolar recruitment and reduce work of breathing,8,14–21 to a maximum of 8–12L/min, and some studies have applied initial flows of 1–3L/kg/min but to a maximum of 8L/min, and found no adverse events.17 Most of these studies, some of which were conducted in paediatric wards, used a maximum flow rate of 10L/min, and found this treatment to be safe.8,15,22–24 We did not find any studies in the literature reporting the use of 15L/min in the paediatric ward, although some authors have suggested that flow rates higher than those usually applied (10L/min) could be useful.24

In our study, in patients treated from the beginning with 15L/min, the RR started to decrease significantly from the first hour of treatment, whereas in the group of children initially treated with 10L/min this decrease happened from hour 6 of treatment, which delayed improvement in the patient. Furthermore, the decrease in RR was greater in the HFNC-15 group compared to the HFNC-10 group. We observed significant improvements in HR and the WDF score, with no difference between groups. We ought to underscore that we did not observe significant side effects in either group, and thus our findings suggest that initiating HFNC with a flow rate of 15L/min in children aged less than 2 years with bronchiolitis (with the exception of newborns, and, until data from clinical trials become available, with caution until the patient's weight reaches 5–7kg) would lead to quicker improvement and a higher probability of success.

In some studies conducted in the PICU setting, up to 25% of patients treated with HFNC experienced treatment failure and required a higher level of respiratory care.16,25 The age and weight of the patient are factors currently debated in the literature, as some authors have found an association with treatment failure and others have not.26 Different authors have analysed the association with treatment failure of other clinical factors, such as the SatO2, pCO2 and severity score, with inconclusive results.26–28 In our study, neither the WDF score nor the body weight of patients were associated with treatment failure, but age was.

In our study, treatment failure occurred in the first 12–14h, and in the HFNC-10 group it failed within the first 12h in as many as 50% of the patients. Other authors have reported treatment failure in the first 6–12h,15,26,29 which suggests that optimising treatment in the first few hours can have a direct impact on the effectiveness of this approach.

There were a series of limitations in our study. It was a pilot study conducted in a single centre with a relatively small sample. It was not a randomised clinical trial, and patients were assigned to groups based on the usual practices of the clinician in charge of their case.

Another limitation was that we did not monitor the partial pressure of carbon dioxide. Nevertheless, we obtained relevant data that suggest that it would be worth it to conduct a multicentric, randomised, blind clinical trial to confirm the findings of this study.

In conclusion, the use of HFNC for management of bronchiolitis in the paediatric ward setting was safe and effective, and our findings suggest that initiating treatment with a high-flow rate of 15L/min achieves a quicker decrease in clinical severity parameters and is associated with a lower frequency of treatment failure.

The authors have no conflicts of interest to declare.