A. Patient Population
1. Neonates with moderate to severe NE
This was a single center, retrospective, observational cohort study approved by the University of Florida Institutional Review Board (IRB) prior to the collection of any data. Following IRB approval, an integrated data repository (IDR) of all neonates who underwent therapeutic hypothermia at UF Health, level IV NICU between 1/2012 and 1/2020 was created. All clinical data in the electronic health record were extracted and used for this work.
Entry criteria for hypothermia included a gestational age of 35 weeks or greater, a birth weight of 1.8 kg or greater, and less than or equal to 6 h of age. Enrolled neonates had evidence of encephalopathy as defined by seizures or abnormalities on a modified Sarnat exam (level of consciousness, spontaneous activity, posture, tone, primitive reflexes including suck and Moro, and autonomic system findings including pupils, heart rate, and respirations) (6). Evidence of hypoxic–ischemic injury was defined by 1) a pH of 7.0 or less and/or a base deficit of greater than 16, or 2) a pH between 7.01 and 7.15 and/or a base deficit between 10 and 15.9, or 3) no blood gas available and neonate had an acute perinatal event (cord prolapsed, heart rate decelerations, uterine rupture) (6).
B. Determination of timing of Hypoxic-Ischemic Event
Maternal and neonatal dyad chart review was done to segregate neonates into 3 groups. The definition for each of the 3 groups were as follows:
Group 1: Neonates with history of definite sentinel events like cord prolapse, uterine rupture, abruptio placenta, placenta previa, maternal hypotension following trauma or epidural analgesia and shoulder dystocia. History of these events during peripartum period would indicate that hypoxic-ischemic injury occurred at time of birth.
Group 2: Neonates with history of probable sentinel events like prolonged deceleration without clear etiology, cord compression and true knot indicating that hypoxic-ischemic injury to neonate occurred within few hours of time of birth.
Group 3: Neonates with no known history of sentinel events who suffered hypoxia-ischemia and underwent hypothermia treatment. Hence, these neonates could have chronic or acute on chronic hypoxia-ischemia.
Two physicians were involved in the chart review to classify neonates into each of the above 3 groups. When prenatal monitoring strips were available, the strips were analyzed by a Board Certified Obstetrician to confirm the timing of injury. Fetal monitoring strips were available only for inborn neonates (n = 13). Time of onset of abnormal fetal heart rate tracing was used as time of hypoxic-ischemic insult.
C. Nucleated Red Blood Cells
Neonates admitted for moderate-severe NE undergo serial labs for monitoring during therapeutic hypothermia process. Complete blood cell count (CBC) with differential is monitored every 12 hours during hypothermia treatment up to 24 hours after rewarming in our center’s hypothermia management protocol. NRBC are one of the components of this test. The CBC provides White Blood Cell (WBC) count per mm3 of blood and manual differential provides percentage of NRBC or NRBC/100WBC using wright stain of blood smear. For calculation of absolute NRBC, first a corrected WBC count was obtained, followed by absolute NRBC count using the following formula (15):
Absolute NRBC count (/mm3) = corrected WBC count* NRBC%(NRBC/100WBC)/100
NRBC per 100 WBCs and absolute NRBC counts were obtained from the time of admission, then every 12 hours up to 120 hours of life and trends were compared among the 3 groups as described above.
D. MRI scoring
MRI were performed at either 3–5 days of age following rewarming (n = 97) or 7–12 days of age (n = 32, not clinically stable for imaging at 3–5 days of age). Six neonates had MRIs at greater than 12 days of life (range 14–41 days) due to comorbidities and clinical instability. Neonates were imaged on a Siemens Magnetom Verio 3T scanner (n = 82) or 1.5 T Siemens Avanto (n = 50) (Siemens, Malvern, PA) at UF Health Gainesville. Analysis focused on the T1-weighted, T2-weighted, and diffusion-weighted imaging (DWI) abnormalities. A single blinded subspecialty board-certified neuroradiologist with over 10 years of experience in neonatal imaging interpreted all the MRI images using the Weeke scoring system (16). The Weeke scoring system scores injury in different brain regions using a scale with increasing values representing worsening injury. Individual brain regions scored included the total grey matter injury, total white matter injury, total cerebellar injury, and a total brain injury score. Each individual score was compared to both the absolute and corrected NRBC in each of the three groups to examine for correlations to the extent of injury to each of the scored brain regions.
E. Statistical analysis:
For each subject, peak NRBC values (absolute and normalized) were obtained, as well as the corresponding hours of life (i.e., hours of life of peak absolute NRBC and peak normalized NRBC values). A major hypothesis of whether the hours of life for each peak was within the 28–29 hours following birth was investigated using normal-theory (approximate) confidence intervals that were computed both for the entire sample, as well as for each of the 3 groups. Lack of significance of the group effect on the hours of life of the peak NRBC was ascertained using conventional normal theory one-way ANOVA model and Welch test (when the group variance homogeneity was violated) and validated nonparametrically using Kruskal-Wallis test (17).
Nonlinear associations between the NRBC values (response/outcome) and the corresponding hours of life were investigated using generalized additive (mixed) models (18). Likelihood ratio test (chi-squared) was used to test the full model with group-specific nonlinear temporal trends versus the null model that has the same nonlinear trend shared by the three sentinel event groups.
Model search approach emphasizing well-formulated linear model selection (19) was used to explore the associations between peak NRBC (response) and each individual explanatory variable that included age, severity of MRI brain injury as measured by Weeke scores of brain regions, Sarnat score, cord pH, cord base deficit, arterial blood gas (ABG) pH and ABG base deficit, while controlling for any potential group effects. Specifically, let Y denote the response and X denote one of the above covariates; for each choice of X, we fitted a sequence of nested models that included the interaction between X and the group, additive model with X and the group (no interaction) and the simple linear regression of Y on X. For each covariate, the model with the interaction term statistically significant (p-value < 0.05) or (else) the model with statistically significant main effects was used for inference. Additionally, we explored the group-specific regressions of Y on X. These analyses were carried out for peak absolute and peak normalized NRBC as the response variable.