568 - Risk of Late-Onset Sepsis Associated with Short Initial Empirical Antibiotic Treatment in Infants at Low Risk for Early-Onset Sepsis

Publication Number: 568.6462

Joseph Ting, University of Alberta, Edmonton, AB, Canada; Eugene Ng, University of Toronto Temerty Faculty of Medicine, Toronto, ON, Canada; Jonathan Wong, BC Children’s and Women’s Hospital, Vancouver, BC, Canada; Kyong-soon Lee, The Hospital for Sick Children, Toronto, ON, Canada; Prakesh S. Shah, Mount Sinai Hospital, Toronto, ON, Canada; Marc Beltempo, McGill University Faculty of Medicine and Health Sciences, Monreal, PQ, Canada

Background: Excessive antibiotic use in neonatal intensive care units (NICUs) is associated with increased antibiotic resistance, and other adverse neonatal outcomes.

Objective: To assess the risk of late-onset sepsis (LOS) in infants at low risk for early-onset sepsis (EOS) in relation to the duration of initial empirical antibiotic treatment.

Design/Methods: We conducted a retrospective cohort study using data from the Canadian Neonatal Network. Infants born at < 33 weeks’ gestational age (GA) were included if there was no history of chorioamnionitis, prolonged membrane rupture (>24 hours), or labor onset, and if they were delivered via Cesarean section between January 2010 and December 2023. Based on antibiotic exposure within the first week, infants were categorized into three groups: no antibiotics, 1–3 days, and 4–7 days. Exclusion criteria included culture-proven sepsis (positive growth in blood and/or cerebrospinal fluid) and/or bowel perforation within the first week, major congenital anomalies, death or discharge within 7 days, or non-consecutive antibiotic use. The primary outcome was the incidence of late-onset sepsis (LOS - culture-proven sepsis after 7 calendar days ) within the first 14 days of life (DOL). To assess associations between early antibiotic exposure and outcomes, we employed a Cox hazard model, adjusting for GA, small for gestational age (SGA), SNAP-II score >20, time period, multiple births, and invasive mechanical ventilation within the first three DOL.

Results: We identified 9749 infants, with 5709 (58.6%), 2364 (24.2%), and 1674 (17.2%) receiving zero, 1-3 and 4–7 days of antibiotics within the first week, respectively. Infants in the 4–7 day group were more likely to be male, singleton pregnancies, with lower birth weight and GA, and other parameters indicating higher illness severities [Table 1]. Within the first 14 DOL, 3.9% of infants developed LOS, with adjusted hazard ratios (aHR) of 0.62 (95% CI [0.44, 0.87]) and 0.73 (95% CI [0.57, 0.94]) when comparing LOS incidence among infants receiving 4–7 days of antibiotics to those receiving none or 1–3 days, respectively [Table 2]. There were no significant differences in aHR for LOS within the first 28 days or during NICU stay across the three groups. Subgroup analysis for infants born at < 29 weeks GA produced similar findings. [Table 3]

Conclusion(s): Early empirical antibiotic of 4-7 days in the first week may be associated with a lower incidence of LOS within the first 14 DOL in infants born at < 33 weeks GA and at low risk for EOS. However, there was no difference in LOS incidence by 28 DOL or during NICU stay.

Perinatal characteristics among infants receiving initial empirical antibiotics of various durations for EOS low risk group
Abbreviations: PROM: prolonged rupture of membrane; SD: standard deviation; SGA: small-for-gestational age; CPR: cardiopulmonary resuscitation; SNAP: Scores for Neonatal Acute Physiology; IPPV: intermittent positive pressure ventilation; HFV: high frequency ventilation; iNO: inhaled nitric oxide
p-value1 comparisons of three groups (none vs. 1-3 days vs. 4-7 days)
p-value2 comparisons of two groups (1-3 days vs. 4-7 days)


Distribution of late onset sepsis outcomes among infants born at <33 weeks GA with very low EOS risk at birth
*Adjusted for GA, SGA, SNAPII>20, time period (2010-2014 vs 2015-2019 vs 2020-2023), multiple births, IPPV/HFV for all first 3 days in a Cox hazard model and a logistic regression model. For both models, correlation within-site was accounting for (robust sandwich covariance matrix estimate for Cox model, and generalized estimating equation for logistic model).
Abbreviations: OR: odds ratio; CI: confidence interval

Distribution of late onset sepsis outcomes among infants born at <29 weeks GA with very low EOS risk at birth
*Adjusted for GA, SGA, SNAPII>20, time period (2010-2014 vs 2015-2019 vs 2020-2023), multiple births, IPPV/HFV for all first 3 days in a Cox hazard model and a logistic regression model. For both models, correlation within-site was accounting for (robust sandwich covariance matrix estimate for Cox model, and generalized estimating equation for logistic model).
Abbreviations: OR: odds ratio; CI: confidence interval