Publication

• Title: Oral step-down, optimal drug, and total duration of antibiotic treatment in African children hospitalised with severe community-acquired pneumonia (PediCAP): a factorial randomised controlled trial
• Acronym: PediCAP
• Year: 2026
• Journal published in: The Lancet
• Citation: Bielicki JA, Clements M, Musiime V, Moore DP, Sidat M, Mulenga V, et al. Oral step-down, optimal drug, and total duration of antibiotic treatment in African children hospitalised with severe community-acquired pneumonia (PediCAP): a factorial randomised controlled trial. Lancet. 2026;408(10554):545–557.

Context & Rationale

• Background

  World Health Organization guidance recommends at least 5 days of injectable antibiotics for children hospitalised with severe community-acquired pneumonia. However, prolonged intravenous treatment requires continued vascular access and often prolongs hospitalisation, while robust contemporary evidence supporting either the route or duration of treatment in sub-Saharan African children has been limited.
• Research Question/Hypothesis

  PediCAP tested three linked questions: whether clinically improving children could safely step down from intravenous to oral treatment; whether broader-spectrum co-amoxiclav was superior to amoxicillin; and whether total treatment could be shortened from 8 days to 4–5 days without an unacceptable increase in 28-day readmission or death.
• Why This Matters

  An effective early oral step-down strategy could reduce hospital stay, vascular-access complications, hospital-acquired infection, treatment costs and antimicrobial exposure. Establishing that amoxicillin is sufficient would also avoid unnecessary broader-spectrum co-amoxiclav use.

Design & Methods

• Research Question: In young children hospitalised with uncomplicated severe community-acquired pneumonia, is oral step-down treatment after initial intravenous therapy effective and safe, which oral agent should be used, and what is the shortest effective total antibiotic duration?
• Study Type: International, multicentre, open-label, parallel-group, 2×5 factorial randomised controlled trial with an additional intravenous-only benchmark group. The duration component used a response-over-continuous-intervention design intended to estimate the relationship between treatment duration and outcome rather than test only one short course against one long course.²
• Setting: Thirteen hospitals in Mozambique, South Africa, Uganda, Zambia and Zimbabwe.
• Recruitment: December 7th, 2020, to August 14th, 2023.
• Population:
  • Children aged 2 months to 6 years, weighing at least 3 kg but less than 30 kg.
  • Hospitalised with severe community-acquired pneumonia judged to require at least 24 hours of intravenous antibiotic treatment.
  • Clinical eligibility required difficulty breathing, with or without cough, plus hypoxaemia or central cyanosis, severe respiratory distress, or pneumonia accompanied by a general danger sign.
  • Point-of-care C-reactive protein had to be greater than 10 mg/L.
  • Children had received no more than 24 hours of intravenous antibiotics before randomisation.
  • Important exclusions included known or anticipated need for invasive ventilation or intensive care, complicated or nosocomial pneumonia, recent hospital admission, isolated viral bronchiolitis, predominantly reactive airways disease, penicillin allergy, long-term antibiotic treatment or an anticipated requirement for antibiotics outside the trial regimens.
• Randomisation:
  • Participants were allocated in a 5:5:1 ratio to oral step-down with amoxicillin, oral step-down with 7:1 co-amoxiclav, or continued intravenous-only treatment.
  • Children allocated to either oral drug were simultaneously assigned in equal proportions to total treatment durations of 4, 5, 6, 7 or 8 days.
  • Randomisation used a secure online system, was stratified by site and used randomly permuted blocks of 11 or 22.
• Intervention:
  • All children received at least 24 hours of an approved intravenous regimen.
  • Once clinically improved and able to tolerate oral medicines, children stepped down to weight-band dosing with either amoxicillin 250 mg dispersible tablets or co-amoxiclav dispersible tablets containing 200 mg amoxicillin and 28.5 mg clavulanate, given twice daily.
  • Intravenous and oral treatment together were intended to total the allocated 4, 5, 6, 7 or 8 days.
• Comparison:
  • A 5-day intravenous-only strategy using a locally selected World Health Organization-compatible regimen: penicillin or ampicillin, with or without gentamicin, or ceftriaxone or cefotaxime.
  • Clinicians could delay oral step-down, broaden treatment, restart intravenous therapy or extend treatment when clinically indicated; the trial therefore compared pragmatic treatment strategies rather than inflexible courses.
• Primary Outcome: All-cause hospital readmission or death within 28 days after randomisation.
• Secondary Outcomes: Pneumonia-related readmission or death; all-cause mortality; index and total hospital length of stay; supplemental oxygen duration; total antibiotic exposure; treatment modification or additional antibiotic courses; serious or grade 3–4 adverse events; antibiotic-related adverse events; clinical complications; and intravenous-line complications.
• Blinding: Participants, carers and treating clinicians were not blinded. An independent endpoint review committee blinded to allocation adjudicated whether readmissions and deaths were pneumonia-related.
• Statistics:
  • One thousand participants in the oral groups provided 81% power to detect an absolute 6 percentage point superiority difference between co-amoxiclav and amoxicillin, assuming event rates of 9% versus 15%, a two-sided α of 0.05 and 5% loss to follow-up.
  • Five hundred participants receiving co-amoxiclav versus 100 receiving intravenous-only treatment provided 89% power for non-inferiority, assuming a 9% event rate and an absolute non-inferiority margin of +10 percentage points.
  • The same +10 percentage point margin was applied to the duration comparisons.
  • The target sample size was 1,100.
  • The primary analysis used participants with ascertainable 28-day outcomes according to randomised allocation, with binomial regression adjusted for site and the factorial allocation. Results were presented as adjusted marginal risk differences.
  • Fractional-polynomial models were prespecified for duration-response analyses. Exploratory multiple imputation and time-to-event analyses assessed the effect of missing primary outcome data.
• Follow-Up Period: Twenty-eight days, comprising inpatient assessment, telephone follow-up during weeks 1–3 and a week 4 face-to-face visit, with telephone assessment permitted when attendance was not possible.
• Funding: The trial was supported through the EDCTP2 programme of the European Union. Sandoz donated the oral trial antibiotics but had no role in trial design, conduct, analysis or reporting.

Key Results

This trial was not stopped early. An independent data monitoring committee reviewed interim data on four occasions; no treatment group was discontinued.

Of 2,248 children screened, 1,101 were randomised. The primary outcome was ascertainable in 1,055 participants (95.8%).

• Baseline severity: Median age was 13 months; 45% were younger than 1 year, 62% were receiving oxygen at randomisation, 58% had hypoxaemia or cyanosis, 90% had severe respiratory distress, 81% had a C-reactive protein concentration above 40 mg/L, 12% had severe malnutrition and 1% had HIV.
• Implementation of the strategy: The median intravenous duration in children who stepped down was 2 days. Step-down occurred within the allocated total duration in 68.6%, 76.3%, 85.2%, 89.5% and 91.6% of the 4-, 5-, 6-, 7- and 8-day groups, respectively. Of 823 children who stepped down, 611 (74%) were discharged before completing their oral course.
• Treatment deviations: Fifty-five of 823 children (7%) received further intravenous treatment after oral step-down; 25 of 100 children assigned to intravenous-only treatment received an unplanned oral step-down; and 232 of 1,101 participants (21%) received an antibiotic outside their allocated and permitted regimens at some point.

Internal Validity

• Randomisation and allocation concealment: Central online randomisation, site stratification and varying block sizes should have provided secure allocation concealment. Baseline characteristics were well balanced across the drug and duration groups.
• Sample size and follow-up: The recruitment target was achieved, with 1,101 children randomised. The primary outcome was available for 95.8%, missingness was not associated with treatment allocation, and exploratory multiple-imputation results were consistent with the primary analysis.
• Outcome selection: The all-cause readmission-or-death primary outcome was clinically important and less subjective than a symptom-based cure outcome. Pneumonia-related events underwent independent blinded adjudication and gave results consistent with the primary analysis.
• Prespecification: The protocol and detailed statistical analysis plan were completed before the final analysis. The factorial and duration-response methods, non-inferiority margin, subgroup analyses and sensitivity analyses were prespecified.
• Open-label treatment: Blinding was impractical, but knowledge of allocation could influence the timing of oral step-down, antibiotic modification, adverse-event attribution and discharge. The approximately 1-day reduction in length of stay is clinically plausible but more vulnerable to this bias than the mortality component of the primary outcome.
• Pragmatic strategy rather than pure treatment exposure: Randomisation determined an intended management strategy, while clinicians retained discretion to delay step-down or extend treatment. This appropriately reflects clinical practice, but it means the trial does not establish that every eligible child can safely receive exactly 4 days of antibiotics.
• Imperfect separation: Only 68.6% of children allocated to 4 days stepped down within that duration, and their mean total exposure was 6.0 days. Conversely, 25% of the intravenous-only group received oral antibiotics. Such convergence between groups tends to make non-inferiority easier to demonstrate.
• No adherence-adjusted corroboration: The statistical analysis plan prespecified inverse-probability weighting for deviations from the randomised strategy, but this was not undertaken because adherence was judged sufficiently high. Given the 4-day and intravenous-group deviations, a conventional per-protocol or prespecified adherence-adjusted analysis would have provided useful additional reassurance for a non-inferiority conclusion.
• Non-inferiority margin: The +10 percentage point absolute margin was generous relative to the observed 6.3% primary-event rate in the intravenous group and was not reduced after the event rate proved lower than anticipated. Reassuringly, the upper confidence limits for the oral comparisons were +5.7 and +6.0 percentage points rather than approaching +10.
• Small intravenous benchmark group: Only 96 intravenous-only participants had an ascertainable primary outcome. Consequently, oral-versus-intravenous estimates were less precise than the comparison between the two oral drugs.
• Duration modelling: Primary-event rates were non-monotonic across duration groups—4.1%, 5.3%, 8.5%, 8.3% and 5.2%—and did not display a simple biological dose-response pattern. This supports the absence of benefit from longer treatment, but it limits confidence in defining an exact optimal stopping day from the fitted duration curve.
• Secondary outcomes: Numerous secondary and subgroup analyses were undertaken without adjustment for multiplicity. Isolated findings, particularly duration-specific mortality differences, should therefore be interpreted as exploratory.

Conclusion on Internal Validity: Internal validity was moderate to strong for the pragmatic treatment-policy conclusion that eligible, clinically improving children can be managed with early oral step-down without a clear increase in 28-day readmission or death. Confidence is lower that PediCAP identifies an exact pharmacological minimum of 4 days, because actual exposure overlapped, the intravenous benchmark group was small, and the prespecified adherence-adjusted analysis was not performed.

External Validity

• Sub-Saharan African relevance: Recruitment from 13 district and tertiary hospitals across five countries provides strong applicability to similar African hospital settings. No convincing heterogeneity in treatment effect was identified between countries or centres.
• Clinically severe cohort: The population was genuinely unwell: most had severe respiratory distress, nearly two-thirds required oxygen and more than half had hypoxaemia or cyanosis. The results are therefore more informative for current severe-pneumonia definitions than earlier trials enrolling children with chest indrawing alone.
• Critically ill children excluded: Children requiring or expected to require invasive ventilation or intensive care were ineligible. PediCAP should not be extrapolated to shock, respiratory failure requiring invasive ventilation, empyema, lung abscess, complicated pneumonia or other critical illness.
• Clinical improvement was essential: This was not a trial of oral therapy from presentation. Every child received at least 24 hours of intravenous antibiotics, and oral treatment began only after clinical improvement and demonstration that oral medicines could be tolerated.
• Selected host population: Only 1% of participants had HIV and 12% had severe malnutrition. Children receiving long-term tuberculosis or HIV-related antimicrobial treatment were excluded, limiting application to substantially immunocompromised or highly malnourished populations.
• Diagnostic approach: Pneumonia was diagnosed pragmatically, without mandatory chest radiography or microbiological confirmation, but enrolment required C-reactive protein greater than 10 mg/L. This resembles real-world care in resource-constrained settings while excluding children with lower inflammatory-marker values who might otherwise receive a clinical diagnosis of severe pneumonia.
• Applicability beyond Africa: The biological findings probably support early oral step-down in other settings, but pathogen prevalence, vaccination, antimicrobial resistance, admission thresholds, discharge systems and access to follow-up differ. Direct generalisation to high-income health systems should therefore be cautious.
• Implementation requirements: Safe adoption requires reliable assessment of clinical improvement, access to suitable dispersible amoxicillin formulations, carer ability to administer treatment after discharge and a pathway for prompt review if the child deteriorates.

Conclusion on External Validity: External validity is high for young children with uncomplicated severe community-acquired pneumonia who improve after initial intravenous treatment in similar sub-Saharan African hospitals. It is limited for critically ill children, complicated infection, substantial immunocompromise or malnutrition, and circumstances in which oral administration or post-discharge follow-up cannot be assured.

Strengths & Limitations

• Strengths:
  • Large, investigator-led randomised trial conducted in the population and health systems most affected by severe childhood pneumonia.
  • Efficient factorial design simultaneously addressed route, oral-drug choice and duration.
  • Central concealed randomisation, achievement of the recruitment target and 95.8% primary-outcome ascertainment.
  • Clinically important primary outcome and blinded adjudication of pneumonia-related events.
  • Pragmatic intervention permitting clinically necessary treatment modification.
  • Clear separation in average antibiotic exposure across the allocated duration groups.
  • Consistent findings across the primary outcome, adjudicated pneumonia-related outcome, sensitivity analyses and prespecified subgroups.
  • Direct assessment of a low-cost, scalable strategy with major implications for antimicrobial stewardship and hospital capacity.
• Limitations:
  • Open-label treatment, with potentially allocation-sensitive decisions regarding step-down, discharge and adverse-event attribution.
  • A generous +10 percentage point non-inferiority margin against an observed intravenous-group event rate of only 6.3%.
  • Only 100 children were allocated to intravenous-only treatment and 96 contributed primary-outcome data.
  • Substantial treatment overlap: 31.4% of the 4-day group did not step down within the assigned duration, while 25% of the intravenous group received oral treatment.
  • The 4-day group received a mean of 6.0 antibiotic days, so assignment to the shortest strategy should not be equated with universal receipt of exactly 4 days.
  • The prespecified inverse-probability-weighted analysis of treatment deviations was not performed.
  • Only 28 days of follow-up, without assessment of persistent respiratory symptoms or long-term pulmonary outcomes.
  • No mandatory radiographic or microbiological confirmation of bacterial pneumonia.
  • Critically ill children and several important high-risk populations were excluded.
  • Most recruitment occurred during the COVID-19 pandemic, which might have altered respiratory-pathogen epidemiology.

Interpretation & Why It Matters

• Route of treatment

  For children matching the trial population who have received at least 24 hours of intravenous antibiotics, are clinically improving and can tolerate oral medicines, continued intravenous treatment is usually unnecessary. The accompanying editorial similarly emphasised that PediCAP addresses step-down after early improvement, not automatic oral treatment from initial presentation.¹
• Choice of oral drug

  Co-amoxiclav did not improve readmission or survival compared with amoxicillin. In the absence of suspected resistant or beta-lactamase-producing infection, plain amoxicillin is therefore the more proportionate step-down agent because it is narrower spectrum, widely available and less expensive.
• Treatment duration

  A response-guided strategy targeting 4–5 total days appears sufficient for most eligible children. This should not be interpreted as a requirement to stop treatment on day 4 regardless of clinical progress: nearly one-third of those allocated to 4 days had not stepped down within that period and were appropriately treated for longer.
• Hospital resources

  Oral step-down shortened hospitalisation by approximately 1 day without a subsequent increase in total hospital days. At scale, this could release considerable paediatric bed capacity and reduce costs to hospitals and families.
• Antimicrobial stewardship

  The trial supports both narrower-spectrum therapy and shorter exposure. Antibiotic-related adverse events increased by 0.9 percentage points for every additional allocated treatment day, providing a measurable reason not to continue treatment without clinical indication.
• Bedside application

  The practical intervention is a monitored clinical pathway: give initial intravenous treatment, reassess for improvement and oral tolerance, step down preferentially to amoxicillin, target approximately 4–5 total days, and extend or broaden treatment only when the clinical response or complications justify doing so.

Controversies & Other Evidence

• Was the non-inferiority margin too wide? The +10 percentage point margin exceeded the 6.3% event rate observed with intravenous-only treatment. In principle, the design could therefore have labelled a clinically important increase in failure as non-inferior. The actual data are more reassuring than the margin: the upper confidence limits were +5.7 percentage points for amoxicillin and +6.0 for co-amoxiclav, although even these limits do not exclude a meaningful absolute increase.
• Absence of superiority is not proof of equivalence: Co-amoxiclav did not demonstrate superiority over amoxicillin, but the 95% confidence interval ranged from a 1.8 percentage point benefit to a 4.4 percentage point increase in events. Formal equivalence was not tested. Nevertheless, there is no demonstrated clinical advantage to justify routine use of the broader-spectrum drug.
• A strategy trial rather than a fixed-course efficacy trial: PediCAP primarily answers whether clinicians can safely pursue early step-down and shorter intended treatment while retaining discretion to continue therapy. It does not prove that every child with severe pneumonia should receive precisely 4 days, nor that outcomes would remain unchanged if extension were prohibited.
• Potential bias towards non-inferiority: Treatment crossover and extension reduced the biological contrast between groups. This is particularly important because the prespecified adherence-adjusted analysis was omitted. The intention-to-treat strategy estimate remains clinically relevant, but it offers less certainty about the effect of the treatments actually received.
• The minimum effective duration remains unresolved: The response-over-continuous-intervention design was intended to identify a duration threshold, but event rates were no higher with 4 days than with 8 days and showed no monotonic relationship with duration. PediCAP therefore supports a 4–5-day strategy but cannot exclude the possibility that an even shorter course would be sufficient in selected improving children.
• Apparent mortality signal at 6–7 days: Mortality was 9/188 (4.8%) in the 6-day group and 8/193 (4.1%) in the 7-day group, compared with 2/192 (1.0%) in the 8-day group. This isolated, non-monotonic finding is unlikely to indicate harm from those assigned durations: 21 of the 23 deaths in the oral-duration groups occurred before oral step-down, 14 occurred before the allocated stopping time could influence treatment and 11 occurred by day 4. Multiplicity and chance are the more credible explanations.
• Earlier route and duration evidence: APPIS previously found oral amoxicillin comparable with injectable penicillin, but used an older definition under which chest indrawing alone qualified as severe pneumonia. CAP-IT found 3 days of amoxicillin non-inferior to 7 days in children treated after discharge in the UK and Ireland, but did not study the contemporary severe, initially hospitalised population enrolled in PediCAP.³⁴
• Diagnostic uncertainty: Only 11 of 258 baseline blood cultures yielded a pathogen, and chest radiography was not mandatory. PediCAP is therefore a trial of clinically diagnosed severe pneumonia rather than microbiologically proven bacterial pneumonia. Some participants probably had viral or mixed infections, which might reduce the apparent importance of antibiotic duration, although this reflects the diagnostic reality of paediatric pneumonia care.
• Microbiome and resistance findings: A companion analysis of 149 children did not identify a clear differential effect of oral step-down, amoxicillin versus co-amoxiclav, or allocated duration on gastrointestinal microbiota or resistome outcomes; age and country were stronger determinants. The small substudy provides some reassurance but is not sufficiently precise to exclude clinically relevant ecological differences between regimens.⁵

Summary

• PediCAP randomised 1,101 young African children hospitalised with uncomplicated severe community-acquired pneumonia to oral step-down with amoxicillin or co-amoxiclav across five intended total durations, or to 5 days of intravenous-only treatment.
• After at least 24 hours of intravenous treatment and clinical improvement, oral amoxicillin met the prespecified non-inferiority criterion versus continued intravenous treatment for 28-day readmission or death.
• Co-amoxiclav did not demonstrate superiority over amoxicillin, supporting the narrower-spectrum agent as the preferred routine oral step-down treatment.
• Assignment to 4–5 total days was as effective as assignment to 8 days, but the 4-day strategy frequently involved clinically appropriate extension and produced a mean actual exposure of 6.0 days.
• Oral step-down reduced hospital stay by approximately 1 day, while longer treatment increased antibiotic-related adverse events.
• The findings apply to clinically improving children without critical illness or complicating factors and should not be extrapolated automatically to children requiring intensive care, invasive ventilation or treatment for complicated pneumonia.

Overall Takeaway

PediCAP provides persuasive evidence that most young children with uncomplicated severe community-acquired pneumonia who improve after at least 24 hours of intravenous antibiotics can step down to oral amoxicillin and complete a short, approximately 4–5-day total treatment strategy. The conclusion is strongest as a pragmatic, response-guided management pathway rather than proof that every child should receive exactly 4 days: the generous non-inferiority margin, small intravenous group and treatment overlap warrant some caution, but there is no demonstrated justification for routinely continuing intravenous therapy, extending treatment to 7–8 days or using co-amoxiclav in eligible improving children.

Bibliography

• 1.Gallagher JC. How long must we treat? A question as old as penicillin. Lancet. 2026;408(10554):496–497.
• 2.Quartagno M, Walker AS, Carpenter JR, Phillips PPJ, Parmar MKB. Rethinking non-inferiority: a practical trial design for optimising treatment duration. Clin Trials. 2018;15(5):477–488.
• 3.Addo-Yobo E, Chisaka N, Hassan M, Hibberd P, Lozano JM, Jeena P, et al. Oral amoxicillin versus injectable penicillin for severe pneumonia in children aged 3 to 59 months: a randomised multicentre equivalency study. Lancet. 2004;364(9440):1141–1148.
• 4.Bielicki JA, Stöhr W, Barratt S, et al. Effect of amoxicillin dose and treatment duration on the need for antibiotic re-treatment in children with community-acquired pneumonia: the CAP-IT randomized clinical trial. JAMA. 2021;326(17):1713–1724.
• 5.Rodriguez-Ruiz JP, Hiel S, Musiime V, Mulenga V, Mujuru HA, Madhi SA, et al. Impact of co-amoxiclav versus amoxicillin on the gastrointestinal microbiota in sub-Saharan children hospitalized with pneumonia. npj Antimicrob Resist. 2026;epublished May 22.