Cohort data were extracted from the National Institute of Child Health and Human Development (NICHD) Neonatal Research Network (NRN) Generic Database4
which includes data from 15 different academic institutions on infants ≤ 1500g birth weight admitted to a network center within 14 days of birth or who were live-born but died in a network center delivery room. Follow up data were similarly obtained from the NICHD NRN ELBW Follow-Up Study at 18-22 mo CGA. Our cohort included infants born July 2000 - December 2006 at two participating hospitals at the Cincinnati site of the NICHD NRN, which were selected because the intimate knowledge of their neonatal intensive care unit (NICU) management strategies allowed the authors to determine with confidence that no unique institutional changes in management strategy for major neonatal morbidities occurred during the study period. The lack of definitive, evidence-based algorithm regarding the best mode of early respiratory support in the study cohort, meant the decision to place an infant on CPAP or MV was at the discretion of the attending neonatologist.
Infants included in the study: 1) Weighed ≤ 1000g at birth (e.g ELBW infants), 2) Were alive at 24 h of age, 3) Required either CPAP or MV at 24 h of age, 4) Had follow up ND assessment and Bayley Scale of Infant Development Version II (BSID-II)5
examinations recorded as part of the 18-22 mo CGA follow up visit. Infants placed on CPAP in the delivery room were eligible. Infants requiring immediate delivery room intubation or infants who did not require ventilatory support at 24 h of age were excluded.
During the study period 916 infants ≤1000g were delivered. One hundred fifty six died in the first 12 h of life, 393 were intubated during initial resuscitation and 60 required no ventilatory support at 24 h. Of the 307 infants meeting inclusion criteria at 24 h of life, 45 subsequently died in the NICU, 24 had no NRN follow up visit data recorded and 30 were missing BSID-II scores. 208 infants met all inclusion and exclusion criteria and were considered in the final assessment ().
Flow diagram illustrating categorization of ventilatory groups
Primary ND outcomes included: BSID-II Mental Development Index (MDI) and Psychomotor Development Index (PDI), rates of CP, deafness, blindness, BPD and death. BSID-II examination was administered at the 18-22 mo CGA follow up visit by a nationally recognized gold standard examiner, and in the presence of the patients’ primary caretaker to ensure optimal performance. Patients who could not complete the BSID-II examination due to ND deficits received a score of 49. MDI or PDI scores > 85 were defined as no ND disability. MDI or PDI scores < 70 were defined as significant ND disability. Other primary outcomes were defined as follows: Blindness - bilateral lack of functional vision. Deafness - hearing loss requiring amplification in both ears. Cerebral palsy - static encephalopathy with delayed achievement of motor milestones, abnormalities in muscle tone in at least one extremity, and dysfunctional control of movement or posture. Cerebral palsy was further categorized as moderate or severe using the Amiel-Tison criteria6
. BPD - need for supplemental oxygen when ≥36 wk CGA.
Secondary outcomes used to assess potential differences in common neonatal morbidities between groups included growth data at 36 wk CGA and 18-22 mo CGA, occurrence of seizure or use of seizure medications during initial neonatal hospitalization, length of NICU stay, rates of necrotizing enterocolitis (NEC), retinopathy of prematurity (ROP), Stage 3 or 4 ROP, patent ductus arteriosus (PDA) treated with indomethacin or by surgical ligation, grade III-IV intraventricular hemorrhage (IVH), periventricular leukomalacia (PVL) and late onset sepsis. Secondary outcomes were defined as: NEC - Bell's classification ≥ IIa.7
ROP - international classification of ROP8
, and ROP Stage 3 or 4 in either eye as documented by clinical ophthalmological examination. Late onset sepsis - occurring ≥ 72 h of life. IVH - Papile's criteria for grade ≥ III. PVL - determined by head ultrasound performed after 2 wk of age (If no head ultrasound was performed then data regarding PVL was not recorded). PDA - clinical or echocardiographic evidence for flow through the ductus.
Antenatal risk factors () were defined as: Prolonged preterm rupture of membranes (PPROM) - rupture occurring > 24 h prior to delivery. Antenatal hemorrhage - maternal bleeding documented after 20 wk gestation in a pregnant woman with placenta previa, abruption or threatened abortion which resulted in vaginal bleeding or occult retro-placental clot. Antenatal antibiotics - antibiotics administered to the mother during the hospitalization that resulted in the delivery of the infant. Antenatal steroids - steroids given to the mother in an attempt to accelerate development of the fetal lungs in preparation for preterm delivery and extrauterine life. Tocolytics - medication given to the mother in an attempt to stop preterm labor. Gestational age was determined by best obstetric dating methods (last menstrual period or first trimester ultrasonography). Race was categorized as Caucasian or non-Caucasian due to the demographic composition of the referral area for the study hospitals. Trained research nurses obtained all growth measurements at pre-specified intervals.
Antenatal Risk Factors and Demographics
Unadjusted bi-variable analyses compared mechanical ventilation and CPAP groups to assess confounders, covariates and outcomes. Categorical variables were analyzed by χ2 and Fisher's Exact test. Continuous variables were analyzed with Student's t test. Multivariable logistic regression analyses assessed independent risk factors for outcomes of interest while controlling for any potential confounders. A stepwise approach to model selection included variables in the models only if they were significantly related to the outcome or the risk factor at a significance level of p < 0.20 in the bi-variable analyses (, and ). Variables were retained in the model if they maintained a significance level of p<0.10. Because the number of events was limited for some outcomes, it was important to pick the most parsimonious regression model. The best fit models were confirmed using a combination of the Hosmer-Lemeshow test statistic, as well as the Deviance and Pearson's chi-square statistic. A variable for NICU site remained in all models to control for potential site/practice variability. Gestational age, birth weight or both remained in all models to control for level of prematurity. Final results were statistically significant if p <0.05.
Primary Outcomes by Mode of Ventilatory Support
Secondary Outcome Measures by Mode of Ventilatory Support