652 - Cerebral Hypoxia and Corrective Actions in the Multicenter Brain Oxygenation-II (BOx-II) Clinical Trial

Publication Number: 2635.652

Callie Marshall, Washington University in St. Louis School of Medicine, SAINT LOUIS, MO, United States; May Yee Szeto, Stanford University School of Medicine, Palo Alto, CA, United States; Jessica Liu, Stanford University School of Medicine, Palo Alto, CA, United States; Andrew Hopper, Loma Linda University Children's Hospital, Loma Linda, CA, United States; Nicole Kraus, Loma Linda University Children's Hospital, Loma linda, CA, United States; Santina Zanelli, University of Virginia, Charlottesville, VA, United States; Karen Fairchild, University of Virginia School of Medicine, Charlottesville, VA, United States; Lina Chalak, Childrens Health Dallas UT Southwestern, DALLAS, TX, United States; Zachary A. Vesoulis, Washington University School of Medicine, St. Louis, MO, United States; Valerie Y.. Chock, Stanford University School of Medicine, Palo Alto, CA, United States

Background: Among extremely low gestational age neonates (ELGANs), cerebral hypoxia is associated with an increased risk of intraventricular hemorrhage and neurodevelopmental impairment. Brain-specific oxygen monitoring may reduce hypoxia burden by alerting clinicians and prompting intervention.

Objective: To quantify cerebral hypoxia events and describe the frequency and effectiveness of corrective interventions in the BOx-II clinical trial.

Design/Methods: The BOx-II study recruited ELGANs (born < 28 weeks gestation) at five US academic centers within 6 hours of birth to undergo cerebral NIRS monitoring for the first 72 hours of life. Monitoring was paired with a standardized, hierarchical treatment algorithm to guide interventions with an initial focus on respiratory interventions to maintain cerebral saturations above device-specific thresholds (> 63% INVOS or > 67% ForeSight Elite; Figure 1). Time-synchronized NIRS and SpO2 recordings were made at all sites. Clinical data, adverse events, and patient intervention logs (including timing and type of intervention) were captured in a REDCap database.

A custom analytics pipeline in R (R Foundation for Statistical Computing, Vienna, Austria) was written to identify cerebral hypoxia events (≥10 minutes cerebral rSO2 below threshold). The number and severity of adverse events as well as the number and type of interventions were extracted from the REDCap database and matched against recording timestamps.

Results: Among 81 infants (GA 25±1.6 weeks, birthweight 872±204g), 229 cerebral hypoxia events occurred in 55 patients; interventions were performed in 20/229 events (8.7%). The most common interventions were increased FiO2 (9/20) and ventilator changes for hypocarbia (8/20, Figure 2). Events with interventions were 35.4 minutes longer (66.5 vs 31.1 min, p=0.01), and each 1-minute increase in event duration increased intervention odds by 1.2% (p=0.007, Figure 3). Mean nadir rSO2 was 2.3% lower in the intervention group (48.1 vs 50.4%) but was not statistically significant (p=0.41). Three device-related adverse events (skin breakdown) occurred, all resolving with sensor removal.

Conclusion(s): In this multicenter trial using algorithm-guided NIRS monitoring, interventions occurred in fewer than 9% of prolonged cerebral hypoxia events, with longer event duration predicting intervention. These findings suggest opportunity for standardized protocols to improve recognition and management of cerebral hypoxia in ELGANs.

Figure 1. Study Intervention Algorithm
The stepwise intervention algorithm focused on respiratory interventions for hypocarbia (Step 1), followed by treatment of hypoxia, hypotension, or anemia (Step 2), with evaluation for hemodynamically significant PDA or intraventricular hemorrhage if cerebral hypoxia persisted (Step 3).

Figure 2. Intervention Types by Frequency
Bar plot showing frequency of recorded interventions. FiO2 increase and ventilator adjustments were the most common interventions.

Figure 3. Event duration and probability of intervention
Caption: (A) Events with interventions had significantly longer duration than those without (66.5 vs 31.1 min, p=0.01). (B) Probability of intervention increased with event duration, from 3% for events 10-15 minutes to 25% for events >60 minutes, representing a 1.2% increase in odds per minute (p=0.007).