Randomized Trial of Weight-Based Versus Fixed Limit High-Flow Nasal Cannula in Bronchiolitis

We conducted a prospective, randomized open-label clinical trial (ClinicalTrials.gov number, NCT03492307) to compare the effectiveness of weight-based versus fixed-liter limit HFNC on key clinical outcomes from May 1, 2018 to March 30, 2020 in the inpatient pediatric ward at South Shore Hospital, a community hospital. The inpatient ward at South Shore Hospital is a 16 bed unit, with an average annual census of approximately 1500 patients. The hospital has provided HFNC to pediatric patients since 2013. From 2013 to the start of the trial, the maximum HFNC flow provided was 8 LPM. South Shore Hospital has an academic affiliation with Boston Children’s Hospital, a large referral center located approximately 20 miles away. The Institutional Board Review at South Shore Hospital approved this study, as did the Institutional Board Review at Boston Children’s Hospital, with reliance on the South Shore Hospital review. The study team at South Shore Hospital was comprised of physicians, nurses, and respiratory therapists, and was led by a physician with a joint affiliation at South Shore Hospital and Boston Children’s Hospital. The investigators were not aware of the trial outcome until the posttrial analysis was completed. The Optiflow high-flow equipment was donated by Fisher and Paykel Healthcare (Auckland, New Zealand), which was not involved in the design and conduct of the trial, the analysis of the data, or the preparation of the manuscript.

Children younger than 2 years old were recruited for the study if they were admitted to the inpatient ward with a primary diagnosis of bronchiolitis, and the medical team made the decision to initiate HFNC. We did not have any exclusions for chronic medical conditions or comorbidities. Children who were started on HFNC in the emergency department (and transferred to the inpatient ward on LFNC) could be included if the medical team made the decision to reinitiate HFNC at any point during the inpatient stay. We used a Bronchiolitis Assessment Severity Score (BASS) to assist in determining who may benefit from HFNC (Fig 1). The BASS score was based on respiratory rate, heart rate, degree of dyspnea and degree of hypoxia. The score has not been validated, but is derived from validated bronchiolitis scores,^21,22  and was used to quantify bronchiolitis severity at our institution before the initiation of this study. We recommended that the team consider HFNC initiation with a moderate or severe BASS score. Patients placed on HFNC were assessed hourly. We excluded patients whose parents were non-English speakers as we did not have translation services for the consent. Written consent was obtained from the parents of all the children before randomization. The risk of barotrauma with HFNC was disclosed to participants.

We randomized children in a 1:1 ratio to the 8 LPM fixed-liter limit arm (standard arm) versus weight-based HFNC arm (weight-based arm). We used sealed, numbered, opaque envelopes for the treatment assignment. As a single respiratory therapist was responsible for assessing the patient and adjusting the flow as required, we were unable to blind the medical team to the assigned group.

Patients in both the standard and weight-based arms received HFNC using identical devices that provided heated, humidified air mixed with oxygen. Nonocclusive nasal cannula selection was determined by the respiratory therapist. Patients in the standard arm had HFNC initiation, escalation, and weaning guided by a historic hospital pathway that included a maximum flow of 8 LPM, regardless of weight. The timing and extent of flow weaning was left to provider discretion. Patients in the weight-based arm were placed on a novel pathway. In this arm patients were started on 1 L/kg per minute and could escalate to 2 L/kg per minute (maximum flow 40 L per minute) if their BASS score did not improve to “mild.” Oxygen was titrated to maintain saturations >90%. Once captured, a period of 4 hours of observation was recommended. Following that period of stability, Fio₂ was first weaned to 21%. Once stable on 21% oxygen, flows were cut in half, then discontinued. For both groups, the study team suggested that the medical team consider interhospital transfer to a PICU if the patient met 3 of the 4 transfer criteria at 4 hours from time of initiation of HFNC: increased or unchanged heart rate, increased or unchanged respiratory rate, increased or unchanged work of breathing, or an Fio₂ greater than 40%. The medical team was encouraged to consider earlier transfer in the setting of sustained clinical worsening on the maximum HFNC allowed in whichever arm the child had been assigned.

We did not standardize the use of adjunctive therapies, including nebulized medications, suctioning frequency, or chest physiotherapy. Additionally, the decision to allow the patient to feed via mouth or nasoenteric tube was left to the discretion of the medical team. Children who were not enrolled in the study were managed using the standard arm pathway, which was the historic HFNC pathway at this hospital.

Our primary outcome was interhospital transfer rate. Our secondary outcomes were hospital LOS and time on HFNC. Transferred patients were excluded from the LOS and time on HFNC analyses. We also tracked maximum flow reached, maximum Fio₂ required, duration of wean (defined as time from first wean attempt to time off HFNC), whether the patient “back-tracked” and required re-escalation of flow after a wean attempt, and whether the patient had a nasoenteric tube placed.

A 2-sided Pearson χ² power analysis was run to determine the minimum sample size needed to detect at least a 30-percentage point difference in transfer rates between the groups with a power of 0.8 and α = 0.05. The 30% difference was chosen because of limitations in patient volume and staff availability at this community hospital site. For the reference proportion, we were assuming a 50% transfer rate (the historic average for patients with bronchiolitis on HFNC over the 5 years before the trial). Based on these assumptions, a sample size of 30 patients per group would give our study adequate power to detect a reduction of 30 percentage points or more in transfer rates.

Patient demographic and clinical characteristics were described using medians (interquartile range) for continuous variables and frequencies (SD) for categorical variables.

We compared the weight-based and standard care groups on demographic factors, clinical characteristics, and hospital utilization (transfer rate, LOS and HFNC hours among nontransferred cases). χ² or Fisher’s exact tests were used to assess for differences in categorical variables; Wilcoxon rank-sum tests were used for continuous variables.

A multivariable logistic regression model was used to assess associations between the standard and weight-based HFNC groups with the likelihood of interhospital PICU transfer, controlling for patient age at admission and presence of chronic medical condition (Y or N). For our secondary outcomes of nontransfer LOS hours and HFNC hours, we used multivariable Poisson regression models to assess associations of the 2 groups with these outcomes, while taking into account patient age at admission and presence of chronic medical condition (Y or N). We report adjusted odds ratios and 95% confidence limits for the logistic regression model predicting interhospital transfer to PICU, whereas adjusted risk ratios and 95% confidence limits are reported for Poisson regression models predicting LOS and HFNC hours. All analyses were performed using SAS version 9.4 (SAS Institute, Inc.; Cary, NC) and P < .05 were considered statistically significant.

The study ran from May 1, 2018, to March 30, 2020. It was discontinued before meeting our enrollment goal because of the burgeoning coronavirus disease 2019 (COVID-19) pandemic. During this period, 255 patients <2 years old with bronchiolitis were admitted to the inpatient program at South Shore Hospital. Of these, 94 escalated to HFNC and were eligible for inclusion in the study. Fifty eight families were approached by the research team; the parents of 6 children declined consent. Fifty two children were randomly assigned to the standard or weight-based arm. One patient was excluded because of the medical team’s decision to defer HFNC following randomization but before treatment initiation. The remaining 51 were included in the analysis. Twenty four were assigned to the standard Arm, and 27 were assigned to the weight-based arm (Fig 2).

There were no significant differences in the demographic or physiologic features of the 2 groups, including admission BASS and maximum BASS, a marker of illness severity (Table 1). The average age of the patients in the weight-based arm was 3 months older, but the difference was not significant (6.4 mo [1.9–11.9] versus 3.7 mo [1.8–7.2]) P = .177.

Give our reduced sample size because of the COVID-19 pandemic, we were ultimately powered to detect a 35% difference in transfer rates between the 2 arms. Transfer rate did not differ significantly between the standard (10 patients, 41.7%) and weight-based arms (14 patients, 51.9%) P = .47. All patients enrolled in the study whose medical team requested a transfer were successfully transferred.

Hospital length of stay was significantly shorter in the weight-based arm (45 h [interquartile range (IQR) 42.1–63.3]) when compared with the standard arm (77.6 h [IQR 47.3–113.4]), with a P < .001 in the multivariate analysis. Time on HFNC was significantly shorter in the weight-based arm (20.4 h [15.7–30.8]) when compared with the standard arm (42.1 h [22.4–51.2]), with a P < .001 in the multivariate analysis (Table 2 and 3; Fig 3).

Patients in the weight-based arm experienced significantly higher flow rates (12 L [IQR 8–18 L]), when compared with patients in the standard arm (8 L [IQR 7–8 L]), P < .01. There was no significant difference in the maximum Fio₂ between the 2 groups. There was also no significant difference in the number of patients who “back-tracked,” requiring re-escalation of their flow following a wean. The median wean time was shorter in the weight-based arm (18 hours [IQR 13–32 h] in the weight-based arm, versus 34.7 hours [IQR 22–48 h] in the standard arm), but this finding did not meet statistical significance, P = .09. Finally, only 1 patient in each group received nasoenteric feeds. The remaining patients received oral feeds once they were stabilized on a flow rate. Patients who required PICU transfer were placed on noninvasive ventilation while awaiting transfer per hospital protocol. No significant adverse events occurred in either group, including pneumothorax, need for intubation, or clinically significant aspiration.

In this single-center randomized clinical trial, we found that using weight-based HFNC for children with bronchiolitis did not significantly impact the rate of interhospital PICU transfer when compared with the standard fixed ≤8 LPM approach. Given early termination of the study, we were powered to detect a 35% difference in transfer rates in the 2 arms. Patients in the weight-based HFNC arm, which included an aggressive prescribed flow weaning pathway, had a shorter hospital LOS compared with those in the standard arm.

Despite evidence that 2 L/kg per minute of HFNC optimally alleviates work of breathing,¹⁷  our study did not demonstrate a major impact on a key clinical outcome: need for escalation to PICU care. This finding should be interpreted in context of the fact that we were forced to halt recruitment early because of COVID-19 and were only powered to detect a 35% difference in need for PICU transfer between the 2 arms. Therefore, there may be a clinically meaningful impact of weight based HFNC on need for ICU care that we were underpowered to detect. Our small sample size also precluded subgroup analysis, and so we cannot exclude the possibility that weight-based HFNC impacted illness trajectory for a subgroup of our patients.

The fact that we did not see a reduction in PICU transfer for patients with bronchiolitis on weight-based HFNC contrasts with a recent multicenter prospective (nonrandomized) study that described a significant drop in PICU need with the introduction of weight-based HFNC.²⁰  Our overall PICU transfer rate of 47% is above the rates described in several other randomized trials of patients with bronchiolitis on weight-based HFNC, which range from 13% to 33%,^11–13  but is consistent with our historic institutional PICU transfer rate for this patient population.

It is difficult to interpret whether the shorter hospital LOS in our weight-based arm was attributable to the differences in HFNC support or in the weaning protocols. Although the standard arm allowed weaning at clinician discretion, the weight-based flow arm had a weaning protocol that was more directive. Previous research has demonstrated a shorter LOS with regimented weaning protocols as the only intervention.⁴  In our study, the weight-based arm did have a shorter median weaning time, but the difference was not statistically significant.

Our weaning protocol in the weight-based arm was aggressive- recommending a decrease from 2 L/kg per minute directly to 1 L/kg per minute and then off. In addition to eliminating the gradual weaning of flows, we recommended monitoring for only 4 hours before initiating a wean. Higher HFNC flows have been associated with a longer LOS,¹⁷  but with this weaning protocol we saw a shorter LOS with the group on higher flows.

Concerns about the risk of rare adverse events associated with higher HFNC flows may have contributed to the more widespread use of lower flow limits in inpatient pediatrics wards that offer HFNC.¹⁸  Our study adds to a growing body of literature that suggests that, even at higher flows, the risk of significant adverse events, such as air leak from HFNC, is very low. This is an important consideration given that HFNC may be the only available form of noninvasive ventilation available at community hospitals without a PICU, where the majority of inpatient care occurs.²³ 

Our study had several limitations. It was impractical to blind the medical team, patients, and families to the study arm, as there was a single respiratory therapist charged with both assessing the patient and adjusting the flow. To minimize bias, we standardized criteria for escalation of flow and interhospital transfer. Adjunctive therapies, including albuterol, suctioning, and chest physiotherapy were not standardized or tracked. However, these therapies have been found to be ineffective in altering major outcomes in bronchiolitis and should not confound our findings. Finally, our study is limited by its small enrollment at a single center, which may reduce the generalizability of our findings and limit our ability to detect clinically significant differences between the groups.

In conclusion, weight-based HFNC support for infants with bronchiolitis did not reduce transfers to a PICU. Larger, multicenter studies are necessary to determine the appropriate role and administration of HFNC in patients with bronchiolitis.

We thank Jesslyn Lenox, Breda Devlin, Kerry Schunder, Elizabeth Dano, Christine Rindini, Samantha Cafaro, Kerry Coughlin-Wells, as well as Drs. Karen Gruskin, Kim Bergner, Alison Daigneault, Gregory Cardello, Rubina Pothiawala and Jamie Gruver for their assistance with the study.