Severe acute asthma exacerbation (SA) can be a potentially life-threatening condition1 and is a frequent cause of admission to a pediatric intensive care unit (PICU).2 Endotracheal intubation and invasive mechanical ventilation (invasive MV) are associated with a substantial risk of complications.3–6 Therefore, the use of other forms of respiratory support such as non-invasive ventilation (NIV)7 and high flow nasal cannula (HFNC) have been considered in an attempt to avoid mechanical ventilation in SA subjects.

Despite the fact that NIV is widely used in asthma subjects admitted in ICU7–13 only few randomized controlled studies (RCT) have been published on pediatric population.10,13 Nevertheless, there is a strong physiological basis behind the use of NIV in asthma.7

On the other hand, HFNC has become popular as it is easy to use and very well tolerated.14–17 Since last decade, HFNC has been introduced in pediatrics as respiratory support, mainly during seasonal bronchiolitis.18 Currently, HFNC has spread to hospital wards, emergency rooms19–21 and transport services.22

There is an ongoing discussion on the indications of high flow nasal cannula therapy or non-invasive ventilation (NIV) in subjects with acute respiratory failure. The clinical advantages of HFNC have not been established yet and there is a lack of published evidence comparing NIV and HFNC during SA.

The aim of this study is to describe our experience with HFNC and NIV in children with SA admitted to the PICU.

Forty-two children met the inclusion criteria and were included in the analysis. Twenty were in the HFNC cohort and 22 were in the NIV cohort. The baseline characteristics of the total sample, as well as the two cohorts are shown in Table 1. Age, gender, weight, PRISM III score, Wood–Downes score, respiratory rate, pH, PCO2, and oxygen saturation at admission, were similar between the two groups. The only significant difference was heart rate with a median (p25–p75) of 164 (141–167) in the HFNC cohort versus 146 (136–156) in the NIV cohort (p=0.009). No children had comorbidities. Apart from an episode of subcutaneous emphysema, no major complications were reported. None of the subjects required intubation. All patients survive.

There were 22 subjects in the initial mode of NIV. The respiratory support mode in this group was bi-level positive pressure in 93.5% and in one subject CPAP. The mean EPAP was 5 (4–7) and the mean IPAP was 12 (8–17). There was no treatment failure in this group. In the HFNC group, treatment failure occurred in 8 subjects. These subjects were changed to NIV with clinical improvement observed in all cases. Therefore, the need to change to a higher respiratory support in HFNC group was 40% (8/20), whereas no subject in the group NIV required higher respiratory support (p=0.001). Fig. 1 shows the probability of remaining free from treatment failure.

Figure 1.

Kaplan–Meier analysis of the probability of remaining free from treatment failure. Log-rank test: p-value=0.013. Censoring marks identify children weaned off the ventilatory support. The figure shows 95% confidence interval for the median.

(0.11MB).

In the subgroup analysis, the mean length of ventilator support and PICU LOS were similar in the groups with no failure. However, the Cox model shows that overall length of ventilatory support was three fold higher in HFNC-failure group as compared to NIV group. Figs. 2 and 3 show the length of ventilatory support (hours) and PICU LOS (days) for the three final (per protocol) groups.

Figure 2.

Kaplan–Meier analysis of the probability of remaining in ventilatory support. Per protocol sub-group analysis.

(0.19MB).

Figure 3.

Kaplan–Meier analysis of PICU LOS (days). Per protocol sub-group analysis.

(0.19MB).

A multivariate logistic regression analysis was performed to identify factors associated with treatment failure in children receiving respiratory support. Heart rate was introduced into the model, in order to avoid confusion because at first both groups were different in relation to this variable. Independent contribution of other variables was tested, and only the heart rate and respiratory rate measured prior to ventilatory support remained in the final model, as shown in Table 2.

The model introduces a highly good discriminatory power: Area under ROC curve (AUC)=0.90 [95% CI: 0.81–0.99 (DeLong)]. And for an optimal threshold of p>0.22, its diagnostic accuracy is excellent: Sensitivity=0.87 [95% CI=0.47–0.99], Specificity=0.82 [95% CI=0.65–0.93], likelihood ratio of a positive test=4.95 [95% CI=2.29–10.73] and likelihood ratio of a negative test=0.15 [95% CI=0.02–0.95]. In our population, this logistic model shows that a HR below 146bpm and a RR below 55bpm would predict HFNC success whereas in the face of a HR above 164bpm, only a RR below 37 would predict HFNC success.

In this study we present our experience with HFNC and NIV for SA in PICU. Even though both therapies are broadly used in asthmatic children all over the world, the lack of literature comparing them is significant. We have not found any randomized controlled trials on this topic and, to our knowledge; this is the first retrospective study looking at both ways of respiratory support in pediatric asthma.

Our hospital covers a population of 200,000 children within 0–14 years with a prevalence of asthma of 10%. Being asthma such a common disease in children, it is mostly managed by general pediatricians in the communities with very good rates of treatment compliance and success. However, SA is still a common presentation to the ED. Only a small percentage of these children fail to respond to medical treatment and require admission to PICU for further management. Increased use of NIV has been associated with less invasive MV and shorter hospital LOS.2,8,10,12,24 Several RCTs have proven NIV efficacy in decreasing work of breathing when used as an adjuvant to medical therapy with nebulized bronchodilator and anti-inflammatory therapy.2,25,26 In Spain, despite the lack of published data, there is a long standing tradition on the use of NIV in pediatric asthmatic subjects with good clinical results and very low rates of intubation.11 In our study 30 subjects were supported by NIV and no one was intubated. In terms of HFNC in asthma, there is also significant paucity of evidence. Kelly et al. described the largest observational study to date, which included 38 children younger than 2 years admitted with status asthmaticus.20 In this study the authors identified three variables associated with increased risk for intubation following HFNC trial: triage RR greater than 90th percentile for age (OR, 2.11; 95% CI, 1.01–4.43), initial venous PCO2 greater than 50mmHg (OR, 2.51; 95% CI, 1.06–5.98), and initial venous pH less than 7.30 (OR, 2.53; 95% CI, 1.12–5.74). We identify heart rate and respiratory rate prior to respiratory support as factors related to failure of HFNC. Also Abboud et al. identified the absence of a reduction in PCO2 as a predictive factor of HFNC failure in children affected by bronchiolitis.27 On the other hand, HFNC has been shown useful in hypoxemic, not hypercapnic subjects (PaCO2<45mmHg) requiring FIO2>0.4.28

We compared two groups of children treated with HFNC or NIV for SA in the PICU. Except for the heart rate, which was higher in the group of HFNC, the two groups in our study were similar at the time of PICU admission in terms of age, weight, PRISM III, Wood–Downes score and FIO2. The reason why HR was higher in the HFNC cannot be determined retrospectively.

We found favorable outcomes in the initial cohort of NIV. The absence of children requiring intubation could be interpreted as less severity of the illness. If we consider that most children with asthma are intubated for “severe work of breathing” or “exhaustion”, our subjects had a mean Wood–Downes score of 8 which is an indicator of severity. Also, we think that the fact that NIV is initiated relatively early in the course of SA management may play a role in preventing the need of invasive mechanical ventilation.

An interesting finding in our study is that in the HFNC group, 8 subjects (40%) needed to be escalated to NIV. Subjects in the HFNC-success group showed similar results to those in the NIV group but those in the HFNC-failure group had a three-fold longer time of respiratory support (p=0.05), compared with the initial NIV group. In addition, the PICU LOS was higher compared with the same group.

Even though causality cannot be established with our study, it is not unreasonable to think that HFNC, which is a lower level of support, could only be beneficial in selected subjects and that it could delay the initiation of NIV and therefore prolong PICU and hospital LOS.

We performed a multivariate logistic regression analysis to identify factors associated with treatment failure in children receiving respiratory support. Only heart rate and respiratory rate measured prior to ventilatory support remained in the final model. According to it, we noticed that children with heart rate less than 146bpm and respiratory rate less than 55bpm, HFNC would probably not fail. When heart rate is greater than 164bpm, HFNC would only be potentially successful in those children with RR less than 37rpm. In other words, in the most severely affected pediatric asthma subjects, it is highly likely that HFNC will fail if it is the initial mode of respiratory support applied. Validation of this study, in a randomized controlled trial with a larger sample, would help to identify a more reliable cut-off value which could guide clinicians on an appropriate algorithm to choose one or the other treatment.

Failure criteria have been described in both groups (NIV and HFNC).29 The respiratory support failure is associated with increased mortality. NIV is more likely to fail in hypoxemic subject30 and HFNC in hypercapnic subject.20 It has also been pointed out that their use may delay intubation.31 Because of low sample size and absence of intubation rate, we were not able to compare the differences in mortality or need for intubation.

Several limitations to our study must be mentioned. On the one hand, the sample size is small and there are no subjects in the NIV failure group limiting the value of our findings. On the other hand, this is a retrospective study with lack of randomization. Also, both initial therapy as well as treatment failure were determined by the attending physician preference. Therefore, the potential for bias is significant in terms of physician preference of one technique over the other or when to consider treatment failure. We tried to manage these confounding factors by using multivariate analysis. However, sample size and the fact that one of the groups had different initial HR limits possible comparisons between groups.

Since this audit, our current therapeutic approach, as other researchers reported,17 is to use HFNC in the Emergency Department in children with mild to moderate SA. Once decision is made to admit the child to PICU, NIV is initiated.

In this study comparing NIV with HFNC for SA in children, we observed that early initiation of NIV in association with bronchodilators and systemic steroids is a safe and feasible initial alternative for the treatment of SA. However, more severe cases HFNC could potentially delay the initiation of NIV resulting in a longer stay in PICU, and the consequent increase in morbidity and costs.

Javier Pilar: Study design and writing.

Vicente Modesto i Alapont: Analysis of data.

Yolanda M. López-Fernández: Manuscript preparation.

Olaia López-Macias: Data collection.

Diego Garcia Urabayen: Literature search.

Irene Amores Hernández: Review of manuscript.

The authors have not disclosed any potential conflicts of interest.