This Week in Pediatrics — Sep 28, 2026

This AudioScholar briefing is generated by artificial intelligence for healthcare professionals and trainees. It is not medical advice.

Welcome to This Week in Pediatrics. This week we're covering 10 notable papers spanning antimicrobial and drug stewardship in our most fragile patients, decision-making that avoids over-medicalizing newborns, and the behavioural and developmental side of chronic childhood disease. Let's dive in.

We'll start with stewardship, because two papers converge on the same uncomfortable truth, that a lot of antibiotic decision-making in children is not driven by anything we can measure. In the Journal of Pediatrics, Liu and colleagues looked retrospectively at very low birth weight infants born before 33 weeks across three neonatal intensive care units, focusing on 202 late-onset sepsis evaluations in which every culture came back negative [1]. The median antibiotic course was 47 hours, which sounds reassuringly disciplined, but 80 of those courses ran beyond 48 hours. The interesting part is what predicted the long courses, which is essentially nothing. Supervised machine-learning models built on structured electronic health record signals discriminated barely better than chance, and unsupervised clustering did identify two physiologic phenotypes based on neonatal organ dysfunction scoring, but those phenotypes did not map onto who got the prolonged course. In other words, continuing antibiotics past two days was not associated with life-threatening organ dysfunction. This is single-health-system and retrospective, so it describes practice rather than testing an alternative, but the authors argue that when cultures stay negative and organ dysfunction is absent or improving, continued treatment needs explicit objective criteria, both for stewardship today and for any credible artificial-intelligence decision support tomorrow. The companion piece comes from the European Journal of Pediatrics, where Echávarri de Miguel and colleagues described five years of World Health Organization Reserve antibiotic use at a tertiary paediatric hospital, 96 treatment episodes in 76 children [2]. Almost all of these patients had serious comorbidity, with haematological malignancy in about four out of five, and bloodstream infection was the leading indication. Prescribing of cefiderocol tracked closely with genuine resistance-driven indications and ceftazidime-avibactam was largely consistent with microbiological criteria, but ceftolozane-tazobactam and ceftaroline were more often used empirically in broader settings. The team then built a structured institutional framework around this, with formulary restriction, indication-based electronic prescribing prompts, and the AWaRe classification embedded in the hospital prescribing system. It is descriptive single-centre experience without outcome data, so it offers a governance model rather than proof of benefit.

The second theme is about doing less to newborns, and doing it deliberately. Also in the Journal of Pediatrics, Virk and colleagues report a before-and-after single-centre study of a protocol for prenatally suspected coarctation of the aorta, comparing 106 neonates managed before implementation with 42 afterwards [3]. Under the protocol, fetuses were stratified into low, moderate and high risk using validated fetal echocardiographic measures, and critically, no prostaglandin was started in any risk category until coarctation was actually demonstrated after birth. The stratification performed reasonably well, with a 7 percent probability of surgical coarctation in the low-risk group against 68 percent in the high-risk group. Prostaglandin exposure fell significantly and was eliminated in the false-positive cases, and among babies whose prenatal diagnosis proved wrong there were significant reductions in transfer to a tertiary centre, in delayed enteral feeding, and in central line placement. The safety signal deserves equal airtime, though. Two infants in the protocol era were diagnosed with coarctation only after hospital discharge, and one developed ventricular dysfunction before prostaglandin was started. This is a single centre with a modest post-protocol cohort, so the evidence supports the feasibility of watchful non-invasive monitoring in expert hands rather than establishing it as a standard. Staying with preterm infants, Pediatric Research published a retrospective cohort from Chen and colleagues of 134 preterm infants having intravitreal ranibizumab for retinopathy of prematurity under general anaesthesia, managed either with a laryngeal mask airway or endotracheal intubation [4]. Escalation of respiratory support within 24 hours occurred in about 7 percent of the laryngeal mask group compared with about 39 percent of the intubated group, and after inverse probability of treatment weighting the laryngeal mask was associated with roughly a two-thirds lower risk. In unadjusted secondary analyses those infants were also more likely to be back to baseline support by 24 hours and less likely to need unplanned reintubation. The authors are careful to call this associative and hypothesis-generating, applicable only where laryngeal mask placement is technically feasible and clinically appropriate, and they call for prospective multicentre work.

Our third theme is comparative therapeutics, starting with a trial that came out neutral. In the European Journal of Pediatrics, Acharya and colleagues randomised 50 children aged one to fourteen with nephrotic syndrome and moderate to severe oedema to furosemide plus metolazone or furosemide alone, with cumulative urine output over 48 hours as the primary outcome [5]. There was no significant difference, roughly three millilitres per kilogram per hour in both arms. The change from baseline looked better in the combination arm in the unadjusted comparison, but that advantage disappeared once baseline urine output was accounted for, and weight loss, electrolyte disturbance and length of stay were all similar. Meanwhile 16 percent of children in the combination arm had weight loss significant enough that therapy had to be stopped. The authors state plainly that upfront combination therapy is not supported by these findings in unselected children, and that the result should be treated as hypothesis-generating for selected patients. It is a small single trial, but the direction is negative and worth reporting as such. Alongside that, the Journal of Pediatrics carried a retrospective comparison from Raj and colleagues of intravenous iron formulations in 342 children and young adults, covering 659 treatment courses [6]. Inflammatory bowel disease and heavy menstrual bleeding were the dominant indications, and two thirds of patients were female. Haemoglobin and ferritin rose with every formulation, with the largest haemoglobin rise seen with low molecular weight iron dextran and the largest ferritin rise with ferric carboxymaltose. Adverse effects were uncommon overall, occurring in about 5 percent of doses, but iron dextran accounted for more reactions requiring medication, and among the twenty ferric carboxymaltose courses with a post-treatment phosphorus level, nine showed hypophosphataemia. That monitoring denominator is small and the design observational, so the message is that formulation-specific advantages sit alongside formulation-specific risks.

Finally, the developmental and behavioural burden of chronic paediatric disease, which four papers illuminate from different angles. In Pediatrics, Jimenez and colleagues report a three-arm randomised trial in 630 Latino caregiver-infant dyads testing whether primary care literacy promotion can be amplified, either by text message reminders or by texts plus referral to a community resource navigation programme [7]. This was a genuinely under-served sample, with about three quarters of caregivers having a high school education or less and most reporting limited English proficiency. At 18 months, the texting arm showed higher child language scores than literacy promotion alone, with a moderate effect size, while the arm that added community referral did not improve language but did show higher social-emotional development scores at both 9 and 18 months. Reading behaviour improved in the texting arm and verbal responsiveness in the referral arm. So both add-ons worked, but on different outcomes, and the authors frame these as scalable enhancements rather than interchangeable ones. Also in Pediatrics, Monzon and colleagues provide the first meta-analytic update in over a decade on behavioural interventions in paediatric type 1 diabetes, pooling 46 randomised trials [8]. The overall effect on haemoglobin A1C was small, and notably, pilot trials showed substantially larger effects than full trials, which is a familiar warning about early enthusiasm. Multicomponent interventions outperformed single-component ones, and family-focused programmes outperformed those aimed at youth alone or parents alone. The honest summary is that despite a decade of technology advances, behavioural work still adds a modest glycaemic benefit, with family-focused multicomponent designs the most promising. Two observational studies round out the burden picture. In the Journal of Pediatrics, Fawcett and colleagues pooled four years of the National Survey of Children's Health, nearly 170,000 children, and found parent-reported attention-deficit hyperactivity disorder in about 18 percent of children with congenital heart disease compared with about 11 percent of those without, a difference that persisted after multivariable adjustment [9]. Prevalence was around a quarter in school-age children and adolescents, and treatment rates with medication and behavioural intervention were comparable to peers without heart disease. It rests on parent report, but the authors argue it supports strengthened neurodevelopmental surveillance. And in Acta Paediatrica, Shoham and colleagues matched more than twenty thousand children with food allergy to non-allergic controls in a national health maintenance organisation [10]. Allergic rhinitis and atopic dermatitis were each roughly twice as common, eosinophilic oesophagitis was dramatically over-represented with 104 events against 5, and specialist visits, emergency department attendance and hospitalisation were all elevated.

If you only have time for one paper this week, make it the literacy promotion trial in Pediatrics [7]. It takes an intervention most of us already deliver and shows, in a large randomised sample of families who are usually under-represented in trials, that low-cost scalable additions shift child language and social-emotional outcomes, which reopens the question of what the minimum effective version of primary care literacy promotion actually is.

Here is what this week's evidence adds up to in Pediatrics. First, two independent stewardship papers suggest that prolonged or last-resort antibiotic use in our sickest children is often not explained by objective illness severity, and in the neonatal data it was not predictable from the record at all, which sets up the case for explicit criteria but does not yet tell us what those criteria should be. Second, protocolised restraint is showing promise at the start of life, both in prenatally suspected coarctation and in airway choice for eye surgery in preterm infants, though both signals come from single centres with small numbers and, in the coarctation study, from two late diagnoses that keep the safety question open. Third, a negative diuretic trial reminds us that combination therapy borrowed from other oedematous conditions did not improve urine output in childhood nephrotic syndrome and did require discontinuation in a subset, while the intravenous iron data show that all formulations work and the real choice is among their distinct adverse effect profiles. And fourth, behavioural and developmental care remains modest in effect but consistent in direction, with family-focused multicomponent strategies leading in type 1 diabetes, and the congenital heart disease and food allergy cohorts quantifying a comorbidity burden that current surveillance guidance has not fully caught up with.

That's your roundup for This Week in Pediatrics. The full transcript and references are available on the episode page in your AudioScholar library. This is an AI-curated summary — for clinical decisions, always consult primary sources and current guidelines. See you next week.

If this weekly briefing is useful, follow the show in your podcast app so new episodes arrive automatically. And if it saves you time, a quick rating in your podcast app helps other physicians find the show.

This is an automated summary generated by artificial intelligence, which can make mistakes. Always review the original source materials.