Non-invasive Respiratory Support Is Associated with Increased Incidence of Occult Cerebral Hypoxia

Background: Impaired cerebral oxygen delivery is a known risk factor for preterm brain injury. Monitoring of cerebral oxygenation with near-infrared spectroscopy (NIRS) may offer more information on “hidden” cerebral hypoxia.

Objectives: To identify the incidence of occult cerebral hypoxia across respiratory support modes in preterm infants.

Design/Methods: Preterm infants born < 32 weeks gestation were enrolled in a prospective, longitudinal NIRS monitoring study. After recruitment, infants underwent daily cerebral NIRS and SpO2 monitoring for 2 weeks, every other day monitoring for 2 weeks, and then weekly monitoring until 35 weeks corrected age. The goal monitoring time was at least 4-6h but was held in cases of patient instability or parent/medical team request.

Respiratory support mode and average fraction of inspired oxygen (FiO2) were noted for each recording.

Modes were classified as:
-Invasive (conventional and high-frequency)
-Non-invasive positive pressure ventilation (NIPPV)
-High flow nasal cannula (HFNC, ≥ 2 LPM)
-Low-flow nasal cannula (LFNC, < 2 LPM)
-Room air

Four calculations were made for each recording: mean cerebral saturation (cSat), mean systemic saturation (SpO2), the proportion of recorded cSat samples below hypoxia threshold ( < 67%), and the proportion of SpO2 samples below desaturation threshold ( < 85%).

Results: 174 infants were included with mean GA of 26.3 weeks, BW of 965g, 47% female (Table 1). 991 recordings were made with a mean length of 6.9 hours. Invasive ventilation and CPAP were the most common modes (27 and 26% of recordings) and most recordings (59%) were made on supplemental oxygen (Table 2)

Mean SpO2 values by respiratory support type were roughly similar (range: 92-96%) and mean cSat values were within the normal range (69-77%). Although there was a minimal and equal SpO2 hypoxia burden between respiratory support types (0-2% of recording), cerebral hypoxia burden was significantly greater in the lowest support groups (29% of recording for RA; 8% of recording for LFNC) compared to other modes (p < 0.01).

Conclusion: Although infants on higher levels of respiratory support are exposed to greater FiO2 and have more clinically severe lung disease, they have significantly less cerebral hypoxia compared to infants on less support. These results raise significant concern about occult cerebral hypoxia during lower acuity periods of NICU hospitalization. Careful selection of respiratory support in conjunction with cerebral monitoring should be considered.