Continuous positive airway pressure vs. high velocity nasal cannula for weaning respiratory support of preterm infants

Preterm infants are at risk of developing chronic lung disease (CLD) and ventilator induced lung injury [1], [2], [3]. Varying noninvasive respiratory devices have been used to provide these infants with support as their lungs grow and develop. Noninvasive respiratory support has been associated with reduced rates of CLD and death in preterm infants [4]. Continuous positive airway pressure (CPAP) has been the mainstay of noninvasive support in preterm infants, though high velocity nasal cannula (HVNC) has become a popular alternative [5]. CPAP provides positive end-expiratory pressure (PEEP) to maintain airway pressure, prevent alveolar collapse, and reduce atelectotrauma common in preterm infants with respiratory distress [6]. High flow nasal cannula (HFNC) is an open system, and therefore does not actively enhance tidal volume or provide PEEP, but rather functions by improving alveolar ventilation through anatomic dead space washout [4], 7]. HVNC provides respiratory support through mechanisms similar to HFNC, such as decreasing anatomic dead space through pharyngeal space washout. Preterm infants particularly benefit from the dead space washout provided by both modalities because they generally have greater volumes of anatomic dead space [8]. There are inconsistencies regarding the use and efficacy of HVNC as a primary mode of support, demonstrating need for further evaluation of long-term outcomes of HVNC use [9], 10]. To our knowledge, CPAP and HVNC have not been compared as primary weaning most of respiratory support in preterm infants.

This was an IRB exempt, retrospective, observational cohort study completed at New York University Langone Health of infants born less than 34 weeks’ gestation between April 2017 and April 2022, who were diagnosed with respiratory distress syndrome and required noninvasive support greater than 2 L per minute.

Infants were assigned to the CPAP (n=175) or to the HVNC (n=48) group if greater than 50 % of their noninvasive support days were spent on that mode of support. Infants were excluded if they required noninvasive support for less than 5 days, required respiratory support due to surgery or a procedure, or were discharged on respiratory support exceeding low-flow nasal cannula.

The primary outcome was the incidence of development of CLD, as defined by Jensen et al. as severity of CLD according to the mode of respiratory support administered at 36 weeks’ postmenstrual age. Within this diagnostic criteria, infants can be further classified into CLD grades based on severity; Grade 1 (mild CLD) requiring ≤2 LPM via nasal cannula, Grade 2 (moderate CLD) requiring >2 LPM via nasal cannula or other forms of noninvasive ventilation, and Grade 3 (severe CLD) requiring invasive mechanical ventilation [11]. Secondary outcomes included length of stay, time on noninvasive support >2 LPM, time on 2 LPM or less, time to wean to room air, time to full enteral and full oral feeds, the occurrence of pressure injuries or other skin complications, and other comorbidities of prematurity such as necrotizing enterocolitis (NEC) and retinopathy of prematurity (ROP).

Demographic information and relevant clinical information were collected and evaluated for all subjects. (Table 1) At baseline, the HVNC group had a slightly lower GA at birth, 29 vs. 30 weeks. Because of this difference, the birth weight, length and HC in the HVNC group were also slightly lower, with length being the only measurement that was statistically significant between the 2 groups. The HVNC group was also more likely to have been intubated, received surfactant, and on HFV at birth.

We found that a significantly higher proportion of infants in the HVNC group developed CLD as compared to the CPAP group (58 vs. 25 %, p<0.05). Furthermore, a binary logistic regression model was performed to investigate the relationship between respiratory support mechanism and development of CLD. Gestational age, sepsis, invasive respiratory support use, and antenatal steroid use were included as control variables. This model indicated that infants who received HVNC support were nearly four times (OR=3.941, CI: [1.79, 8.66]) more likely to develop CLD than those who were on CPAP (p<0.001). An alternative logistic regression analysis was also performed, including duration of invasive support. Based on this model, infants who received HVNC support were still more than three times (OR=3.18, CI: [1.38, 7.29]) more likely to develop CLD than those who were on CPAP (p=0.02). (Table 2)

Due to differences in baseline characteristics and severity of respiratory illness between our two groups, propensity score matching was also performed, resulting in formation of two groups containing 48 infants each. Within these groups, the HVNC group had a higher incidence of chorioamnionitis (37.5 vs. 10.5 %, p=0.02) and an increased need for high frequency ventilation (HFV) (33.3 vs. 14.6 %, p=0.031), although duration of HFV was not significantly different. In the HVNC group, 58 % developed CLD, compared to 27 % in the CPAP group (p=0.002), although the severity of CLD did not differ significantly. Further analysis using a logistic regression model, including the same control variables, revealed that infants in the matched HVNC group were almost five times (OR=4.975, CI: [1.38, 17.85]) more likely to develop CLD (p=0.014). (Table 2)

Further, because of the differences in known factors that affect likelihood of development of CLD, we performed a binary logistic regression to control for these clinically relevant variables. We found that, even when controlling for GA, need for and duration of invasive support, administration of antenatal steroids, and the presence or absence of sepsis, the difference in rates of CLD persisted. Weight was not included as a factor in the logistic regression due to multicollinearity with GA. Using these control variables, the HVNC group was still between 3 and 4 times as likely to develop CLD than the CPAP group. In addition, propensity score matching was performed, resulting in the formation of two similar groups of 48 infants, in which the risk of developing CLD when weaned primarily on HVNC was even higher.

In analyzing our secondary outcomes, those in the HVNC group experienced longer duration (in days) of noninvasive support above two L, longer duration of time on two L, and longer duration on any support until weaned to room air. Infants in this group also experienced significantly longer hospital says (p=0.01), longer wait before initiating oral feeds (p=0.005), and a longer time before achieving full oral feeds (p=0.02). This remained true after propensity score matching was performed. (Table 2)

We believe the difference seen in incidence of CLD is likely because CPAP delivers a reliable, constant PEEP whereas HVNC delivers a constant velocity, but the PEEP is variable and unknown. PEEP is important as premature lungs are developing [12]. The criteria initially used in our NICU for when to switch infants over from CPAP to HVNC (>1 week of age, >28 weeks cGA, >1 kg) may have been too permissive, allowing for transition of small babies to HVNC when they still required consistent PEEP for alveolar distention and development.

When evaluating our secondary outcomes, the HVNC group had longer hospitalizations, longer duration of noninvasive support and total respiratory support, longer time to start oral feeding and to reach full oral feeding. In other words, the HVNC group was “stuck” on noninvasive support longer than the CPAP group. The discharge growth parameters of the HVNC group were larger than the CPAP group, most likely because of the longer hospitalizations. The HVNC group also had a higher likelihood of having ROP requiring intervention, perhaps in part because of longer exposure to oxygen.

Limitations of our study include that it was a single-center retrospective cohort analysis. Additionally, some baseline characteristics of the 2 groups differed, which we addressed with a logistic regression model and propensity score matching. We chose the variables in the logistic regression model because they were either statistically different in our groups, or known factors associated with chronic lung disease. When these demographic differences were accounted for, the outcomes between the 2 groups remained statistically and clinically significant.

In our Level IV NICU, preterm infants born at less than 34 weeks’ gestation whose respiratory support were weaned primarily on HVNC were significantly more likely to develop CLD than those weaned primarily on CPAP. We believe the reason for this difference is that consistent delivery of PEEP is important to developing lungs. Given our data, preterm infants would benefit from waiting to switch from CPAP to HVNC until they reach a later corrected gestational age and greater weight.