EBNEO Commentary: Duration of Antibiotic Therapy for Gram-Negative Bloodstream Infections in the Neonatal Intensive Care Unit

Fabian J. S. van der Velden and Prakash Kannan Loganathan from James Cook University Hospital, South Tees Hospitals NHS Foundation Trust, Middlesbrough, UK, review the article "Djordjevich CJ, Magers J, Cantey JB, Prusakov P, Sánchez PJ, Duration of Antibiotic Therapy for Gram-Negative Bloodstream Infections in the Neonatal Intensive Care Unit, The Journal of Pediatrics (2026), doi: https://doi.org/10.1016/j.jpeds.2026.114993. PMID: 41554433" for for ebneo.org.

READ HERE!

ACTA PAEDIATRICA COMMENTARY:

Acta Paediatrica - 2026 - Velden - EBNEO Commentary Duration of Antibiotic Therapy for Gram‐Negative Bloodstream.pdf

"Gram-negative bloodstream infections (GNBSIs) are an important cause of morbidity and mortality in neonates(1, 2). Recognition and early instigation of appropriate antibiotic treatment is paramount to improve outcomes. The optimal duration of treatment for GNBSI remains uncertain, and prolonged courses may not be necessary for all infants. The aim should be to use the shortest effective course, balancing clinical recovery with the risks of antimicrobial resistance and the potential longer-term harms of unnecessary antibiotic exposure. Prolonged courses of antibiotics have been associated with increased risk of necrotising enterocolitis in premature neonates(3), an increased risk for allergic disease, and inflammatory bowel disease later in life(4). In parallel, antimicrobial resistance remains a growing concern, with reported resistance among neonatal E. coli isolates reaching 50–75% for amoxicillin, 15–30% for gentamicin, and 46% for cefotaxime(1, 5).

 Although defined as ≤8 versus ≥9 days, the effective comparison was approximately 7 versus 14 days (short: median 7 days, 67% at 7 days; long: median 14 days; only 9 and 1 infant at the 8- and 9-day boundary). Within this contrast outcomes were broadly similar; treatment failure (same-organism recurrence within 14 days; 5% vs 0%) and subsequent multidrug-resistant gram-negative (MDRO) infection occurred only in the longer-course group. The design cannot recover a dose-response or inform intermediate durations. 

An important limitation is confounding by indication. Infants in the longer-course group appeared clinically more complex, with higher central venous catheter use (68% vs 38%), all cases with multiple positive blood cultures (19% vs 0%), and greater exposure to broader-spectrum monotherapy such as cefepime or piperacillin-tazobactam (49% vs 10%), whereas gentamicin monotherapy was more common in the shorter-course group (36% vs 11%). These findings suggest that clinicians were already tailoring both antibiotic duration and spectrum according to perceived illness severity or complexity. Therefore, the higher MDRO rate in the longer-course group may reflect broader-spectrum exposure and underlying case mix rather than duration alone. This is reinforced by the MDRO definition itself—any-site detection within 90 days—with events occurring 10–82 days later and predominantly new E. coli ESBL isolates that are plausibly independent acquisitions.

Blood culture volume, sampling technique, and repeat-culture practice were not reported, and the methods do not explain how metastatic foci (bone, abscess) were excluded. Because lumbar puncture was not always performed at presentation, post-antibiotic or traumatic CSF sampling may have under-recognised meningitis. The study is underpowered, with no multivariable or propensity adjustment and no confidence intervals. Survivorship bias is also a concern: the 27 of 67 excluded infants who died before day 7 could not experience the outcome, and reaching ≥9 days itself required survival, favouring the longer-duration group on mortality. Importantly, the 7 NICUs sit within only 2 healthcare systems under shared stewardship, so this reflects closer to two practice cultures than seven independent centres; gentamicin monotherapy was viable only because local E. coli and Klebsiella susceptibility was 92–95%, against the 15–30% resistance cited in the literature, and centres with higher resistance could replicate neither strategy—narrowing external validity. 

Taken together, these data support a hypothesis-generating comparison of approximately 7 versus 14 days in a selected, lower-acuity, vasopressor-excluded population managed within two health systems. Given the small sample, methodological limitations, and risk of bias, the findings justify a prospective trial rather than a change in practice, consistent with the authors’ own conclusion.

Artificial intelligence disclosure: ChatGPT (OpenAI; GPT-5.4 Thinking) was used to support drafting, language editing, and improvement of clarity in this commentary. The authors independently reviewed, critically appraised, revised, and approved the final content, and take full responsibility for the accuracy and interpretation of the work."