DURATION cut four antibiotic days; noninferiority rested on three readmissions · NICU360
Research
DURATION cut four antibiotic days; noninferiority rested on three readmissions
NICU360 Evidence Desk·8 min read·00
In 493 Danish newborns with probable culture-negative early-onset sepsis, a recovery- and CRP-guided strategy reduced median antibiotic exposure from 6.8 to 2.8 days. Infection-related readmission was 2/246 versus 1/247, so the prespecified 4-percentage-point noninferiority margin was met—but the safety estimate rests on only three events in a selected, closely followed population.
Hot take: the exposure result is robust; the safety conclusion is narrow
DURATION tested a question neonatal stewardship has usually answered without randomized evidence. Across all 17 Danish neonatal departments, 493 infants were randomized after a negative culture had not settled the decision to continue treatment. The trial makes a persuasive case that selected, recovered infants can receive much less antibiotic. It does not create a universal three-day EOS course.
Eligible infants were 0–3 days old, born at 35 weeks or later, weighed at least 2000 g, had antibiotics started within 72 hours for possible or probable early-onset sepsis, and remained culture-negative after 36–48 hours. Infants already meeting criteria to stop at 36–48 hours, those with culture-positive EOS or focal infection, and those whose caregivers could not participate in home observation were excluded.
What randomization actually compared
1
Screen after 36–48 hours
The culture was negative, but clinicians still planned prolonged therapy because of clinical signs or maternal risk factors combined with raised CRP.
2
Individualised strategy: 246 infants
Stop after 24 hours without clinical signs, provided CRP was falling to 30 mg/L or lower.
3
Standard strategy: 247 infants
Continue for 5–7 days; 15 of the 17 departments normally used 7 days.
4
Was this worth carrying into your next shift?
Your reaction stays connected to your NICU360 profile.
Comments
0
Join the conversation
Sign in to reply, vote, and keep your participation with your NICU360 profile.
Day-2 visit, day-21 call, and nationwide record review to day 100; coprimary outcomes were assessed within 28 days of treatment start.
Four days came off the median course
Median total antibiotic exposure within 28 days was 2.8 days (IQR 2.4–3.4) with the individualised strategy and 6.8 days (6.6–7.0) with standard duration. The between-group difference was 4.0 days (95% CI 3.8–4.1; p<0.0001). This is a large, precisely estimated treatment-exposure effect.
Median total antibiotic exposure within 28 days
Individualised strategy
Standard duration
Median days
In the individualised group, 185/246 infants (75%) received 3.4 days or less and 227/246 (92%) received fewer than 5 days. Yet exposure is not the same as clinical benefit: the trial did not directly measure antimicrobial resistance, microbiome effects, or long-term child outcomes.
Noninferiority needs the denominator, margin, and event count
Within 28 days, infection-related readmission occurred in 2/246 infants in the individualised group and 1/247 in the standard group. The risk difference was +0.4 percentage points (95% CI −1.5 to +2.6), meeting the prespecified 4-point noninferiority margin (p for noninferiority=0.0032). The per-protocol result was consistent: 2/223 versus 1/228, risk difference +0.5 points (95% CI −1.6 to +2.8; p for noninferiority=0.0054).
Outcome estimates in the intention-to-treat population
Outcome
Individualised
Standard
Between-group estimate
Infection readmission, 28 days
2/246 (1%)
1/247 (<1%)
RD +0.4 points (95% CI −1.5 to +2.6); pNI=0.0032
Infection readmission, 100 days
5/246 (2%)
4/247 (2%)
RD +0.4 points (95% CI −2.3 to +3.3)
Serious adverse event, 100 days
0/246
1/247 (<1%)
RD −0.4 points (95% CI −2.3 to +1.1)
Initial hospitalisation
69 h (IQR 52–95)
72 h (IQR 54–97)
Median difference 3 h (95% CI −3 to +9)
The primary outcome of infection-related readmission was very rare.
The upper confidence limit was below the 4-point margin in both analysis populations, which supports the trial's statistical conclusion. It does not prove equal safety. With only three 28-day events, a clinically relevant absolute excess remains compatible with the data; whether a 4-percentage-point margin is acceptable is a clinical judgment, not an algebraic fact.
A closely watched, mostly term population
Only 12/493 infants were born before 37 weeks. Median gestational age was 40.9 weeks in the individualised group and 41.0 weeks in the standard group; median birthweights were 3685 g and 3660 g. The late-preterm label should not be mistaken for robust subgroup evidence. These results do not extend to infants below 35 weeks or 2000 g, culture-positive infection, meningitis or osteomyelitis, persistent illness, or infants already safe to stop by 36–48 hours.
Coprimary and secondary outcome follow-up was complete, with 96% completing the day-2 visit and day-21 call on time. Families had direct access to neonatal departments, and nationwide records captured hospital contacts. That is a strong safety net. It is also part of the intervention's context, not background scenery that can be assumed everywhere.
Limits clinicians should not smooth over
“Probable EOS” is not a formally validated infection diagnosis, so some enrolled infants might not have had bacterial infection. Blood-culture volume was not systematically recorded: 1.0 mL was targeted, but lower volumes were accepted and 0.2 mL was allowed at one site. False-negative bacteraemia therefore cannot be excluded. A CRP-based stopping rule will also travel poorly where CRP is not used in this way.
The trial was open-label, and 493/811 eligible infants were enrolled. ClinicalTrials.gov still lists actual enrollment as 488 and frames readmission as 1–21 days after the first course; the paper reports 493 randomized and uses 28 days from treatment start. The January 2025 statistical analysis plan prespecified the 28-day analysis before database lock, and the paper explains the window change, but the public registry remains misaligned.
How to read this result without turning it into a protocol
DURATION is credible evidence for reviewing prolonged-course pathways when the population and safety infrastructure match: a negative culture obtained before antibiotics, sustained clinical recovery, falling CRP, and structured follow-up. It is not evidence to delay initial empiric therapy, accept poor culture technique, or abbreviate treatment for culture-positive, focal, persistent, or very preterm infection.
The stewardship target is large. In an international original-research cohort of 757,979 late-preterm and term infants, 7,996 culture-negative cases received at least 5 days of antibiotics, compared with 375 culture-proven EOS cases—about 21-fold more. DURATION supplies a randomized exposure-reduction signal; its rare events and selected setting still bound the safety claim.
