Assessment of Neonatal Intensive Care Unit Practices and Preterm Newborn Gut Microbiota and 2-Year Neurodevelopmental Outcomes

Key Points Question What are the long-term outcomes associated with dysbiosis of gut microbiota in very preterm newborns? Findings In this cohort study of 577 very preterm newborns across 24 neonatal intensive care units from a French nationwide cohort, gut microbiota at week 4 after birth showed 6 bacterial patterns that varied according to gestational age, perinatal characteristics, individual treatments, and neonatal intensive care unit strategies. Three clusters were associated with 2-year outcomes after adjustment for these confounders. Meaning Modifying strategies associated with alterations in microbiota, such as promoting enteral nutrition, reducing sedation use, promoting early extubation, or skin-to-skin practice, may be correlated with outcomes in preterm newborns.


eMethod  Nutrition strategies of NICUs
We characterized 8 strategies for each neonatal intensive care units (NICUs) as previously described (1), no intubation or extubation during the first 24 hours after birth, use of sedation during the first week after birth, direct breastfeeding during the first week after birth, skin to skin during the first week after birth, and treatment of ductus arteriosus during the first ten days after birth, speed of progression of enteral feeding during the first week after birth, duration of primary antibiotherapy started during the first 48 hours after birth and duration of secondary antibiotherapy.

 Strategies concerning received or not received treatments
For the first 5 strategies concerning received or not received treatments, a probability to receive each treatment for each preterm infant was calculated by logistic regression, according to gestational age, birth weight Z-score, sex, inborn or outborn, mode of birth, type of prematurity, single or multiple pregnancy, antenatal corticosteroid administration, Apgar score, number of administered doses of surfactant, birth nationality of mother and socioeconomic level of the parents using EPIPAGE2 database. From the average of the probabilities of receiving the treatment for children from the same NICU, we calculated an expected percentage to have this strategy applied in this NICU. If the observed difference in percentage was zero or greater than the expected percentage, the NICU strategy was considered favourable to the application of this strategy. If the difference was negative, the strategy was considered unfavourable.

No intubation or extubation during the first 24 hours after birth
For each infant i a probability, p(Infant i ), to be not intubated or extubated during the first 24 hours after birth was calculated by logistic regression according the characteristics of the preterm infant and his parents. Then we calculated an expected percentage of to be not intubated or extubated during the first 24 hours after birth for each NICU, expressed as %, according to the following formula: Expected percentage of infants not intubated or extubated during the first 24 hours after birth of one NICU = 100 x (  0 N (p(Infant i )) / Number of infants of this NICU Then we calculated the difference between the observed and the expected percentage of infants not intubated or extubated during the first 24 hours after birth for each unit. If the difference > or < zero, the NICU was considered favourable or not favourable to limit intubation, respectively.

Use of sedation during the first week after birth
For each infant i a probability, p(Infant i ), to receive sedation during the first week after birth was calculated by logistic regression according the characteristics of the preterm infant and his parents. Then we calculated an expected percentage of to be sedated during the first week after birth for each NICU, expressed as %, according to the following formula : Expected percentage of infants sedated during the first week after birth of one NICU = 100 x (  0 N (p(Infant i )) / Number of infants of this NICU Then we calculated the difference between the observed and the expected percentage of infants receiving a sedation during the first week after birth for each unit. If the difference > or < zero, the NICU was considered favourable or not favourable to use of sedation, respectively.

Direct-Breastfeeding policy
For each infant i a probability, p(Infant i ), to be partially direct-breastfed during the first week after birth was calculated by logistic regression according to gestational age, birth weight Z-score, sex, birth Nationality of mother, level of education. Then we calculated an expected percentage of partial direct-breastfeeding during the first week for each NICU, expressed as %, according to the following formula : Expected percentage of infants partially direct-breastfed during the first week after birth of one NICU =100 x (  0 N (p(Infant i )) / Number of infants of this NICU Then we calculated the difference between the observed and the expected percentage of infants direct breastfed during the first week after birth for each unit. If the difference > or < zero, the NICU was considered as favourable to direct breastfeeding or not favourable, respectively.

Skin to skin during the first week after birth
For each infant i a probability, p(Infant i ), to be in skin to skin contact with at least one of the parents during the first week of life after birth was calculated by logistic regression according the characteristics of the preterm infant. Then we calculated an expected percentage of to be in skin-to-skin contact during the first week after birth for each NICU, expressed as %, according to the following formula : Expected percentage of infants in skin to skin contact during the first week after birth of one NICU = 100 x (  0 N (p(Infant i )) / Number of infants of this NICU Then we calculated the difference between the observed and the expected percentage of infants being in skin to skin contact during the first week after birth for each unit. If the difference > or < zero, the NICU was considered as favourable to skin to skin contact or not favourable, respectively.

Treatment of ductus arteriosus during the first ten days after birth
For each infant i a probability, p(Infant i ), to be treated by Ibuprofen during the first 10 days after birth was calculated by logistic regression according the characteristics of the preterm infant. Then we calculated an expected percentage of to be to be treated by Ibuprofen during the first 10 days after birth for each NICU, expressed as %, according to the following formula : Expected percentage of infants to be treated by Ibuprofen during the first 10 days after birth of one NICU = 100 x (  0 N (p(Infant i ) / Number of infants of this NICU Then we calculated the difference between the observed and the expected percentage of infants receiving Ibuprofen during the first 10 days after birth for each unit. If the difference > or < zero, the NICU was considered as favourable to close ductus arteriosus during the first 10 days after birth or not favourable, respectively.

Longer duration of primary antibiotic treatments
Concerning primary antibiotic treatment started during the first 48 hours of life, an expected duration was calculated with an individual 95% confidence interval (CI), using linear regression according to explicative variables: type of prematurity, confirmed early neonatal infection and neonatal characteristics. If the observed duration was more than upper limit of the 95% CI, the duration of antibiotic treatment was considered as longer than the expected duration. As for other strategies, an expected probability for each infant was calculated by logistic regression. Expected percentage of infants with a longer duration of primary antibiotherapy of one NICU = 100 x (  0 N (p(Infant i )) / Number of infants of this NICU Then we calculated the difference between the observed and the expected percentage of infants with longer duration of primary antibiotic treatment for each unit. If the difference > or < zero, the NICU was considered as favourable to longer duration of primary antibiotic therapy or not favourable, respectively.

Longer duration of secondary antibiotic treatments
Concerning secondary antibiotic treatment administered, an expected duration was calculated with an individual 95% confidence interval (CI), using linear regression according to explicative variables: type of prematurity, confirmed secondary infection and neonatal characteristics. If the observed duration was more than upper limit of the 95% CI, the duration of antibiotic treatment was considered as longer than the expected duration. As for other strategies, an expected probability for each infant was calculated by logistic regression. Expected percentage of infants with a longer duration of secondary antibiotic therapy of one NICU = 100 x (  0 N (p(Infant i )) / Number of infants of this NICU Then we calculated the difference between the observed and the expected percentage of infants with longer duration of secondary antibiotic treatment for each unit. If the difference > or < zero, the NICU was considered as favourable to longer duration of secondary antibiotic therapy or not favourable, respectively.

Speed of progression of enteral feeding
We first calculated an expected enteral volume at day 7 with a 95% confidence interval (95%CI) for each infant, according to gestational age, birth weight Z-score and regularity of intestinal transit during the first 7 days. A low volume of enteral feeding was defined as a volume less than the lower limit of the 95%CI. As for other strategies, an expected probability to receive a low enteral volume at day 7 was calculated for each infant by logistic regression. Expected percentage of infants with a low enteral volume at day 7 of one NICU = 100 x (  0 N (p(Infant i )) / Number of infants of this NICU Then we calculated the difference between the observed and the expected percentage of infants with low enteral volume at day 7 for each unit. If the difference > or < zero, the NICU was considered as favourable to provide low enteral feeding during the first 7 days after birth or not favourable, respectively. eTable 1. Comparison between preterm infants with and without faecal collect among eligible preterm infants. Complete cases analysis. Data are number of events/number in group (percentages). Percentages are weighted to take into account the differences in survey design between gestational age groups. Denominators vary according to the number of missing data for each variable. * Eligible survivors at J21 non included in EPIFLORE project n=386 or included without fecal collect 1 month old n=143). † Score based on Olsen curves. ‡ Severe neonatal morbidity was defined as severe bronchopulmonary dysplasia or necrotizing enterocolitis stage 2-3 or severe retinopathy of prematurity stage >3 or any of the following severe cerebral abnormalities on cranial ultrasonography: intraventricular haemorrhage grade III or IV or cystic periventricular leukomalacia (Ancel 2015).

eFigure 6. Repartition of microbiota clusters according to gestational age and 2 year-outcome (complete cases).
Two year-outcome is defined by death or Age and Stages Questionaries' score less than 185 at 2 years on the part A of the figure, and by death or cerebral palsy on the part B of the figure.