Publication

• Title: Prone Positioning in Infants With Acute Bronchiolitis: The PROPOSITIS Randomized Clinical Trial
• Acronym: PROPOSITIS
• Year: 2026
• Journal published in: JAMA
• Citation: Baudin F, Pouyau R, Subtil F, et al; PROPOSITIS Investigators. Prone positioning in infants with acute bronchiolitis: the PROPOSITIS randomized clinical trial. JAMA. 2026;336(4):315-322.

Context & Rationale

• Background

  • Acute viral bronchiolitis is a major cause of hospital admission and acute respiratory failure in young infants. Management is predominantly supportive, with high-flow nasal cannula (HFNC), continuous positive airway pressure and other non-invasive support used when standard oxygen therapy is inadequate.
  • HFNC is easy to deploy and generally well tolerated, but a clinically important proportion of infants still deteriorate and require escalation to positive-pressure ventilation.
  • Prone positioning has plausible physiological effects: redistribution of ventilation, reduced dorsal lung compression, improved thoracoabdominal synchrony and reduced respiratory effort.
  • In the preceding BRONCHIO-DV randomised crossover study of 14 infants with severe bronchiolitis receiving continuous positive airway pressure, prone positioning reduced the median oesophageal pressure-time product from 353 to 227 cm H₂O·s/min, although only 8 of 14 infants demonstrated a physiological response.¹
  • Before PROPOSITIS, it remained unknown whether these short-term physiological effects translated into fewer clinically important escalations of respiratory support.
• Research Question/Hypothesis

  • Among infants younger than 6 months with moderate to severe bronchiolitis receiving HFNC, would a strategy of early prone positioning for at least 24 cumulative hours during the first 48 hours reduce escalation to non-invasive or invasive ventilation within 72 hours compared with supine positioning?
  • The investigators hypothesised that modifying respiratory mechanics early in the illness would prevent respiratory exhaustion and reduce HFNC failure.
• Why This Matters

  • If effective, prone positioning would be a simple, inexpensive and rapidly deployable intervention capable of avoiding non-invasive ventilation, intubation, transfer and intensive care bed use.
  • Its apparent simplicity is deceptive: awake infants may become agitated, dislodge the nasal cannula, feed less effectively or develop pressure-related skin injury, and the intervention must not blur safe-sleep advice after discharge.
  • A clinical-outcome trial was therefore necessary before physiological plausibility could justify routine, prolonged prone positioning.

Design & Methods

• Research Question: Does assigning young infants with moderate to severe bronchiolitis receiving standardised HFNC to early prolonged prone positioning, rather than supine positioning, reduce actual escalation to non-invasive or invasive ventilation during the first 72 hours?
• Study Type: Investigator-initiated, publicly funded, pragmatic, multicentre, parallel-group, open-label superiority randomised clinical trial conducted in 15 paediatric intermediate care or intensive care units in tertiary and secondary hospitals in France. Recruitment ran from 13 January 2021 to 30 November 2023. Centralised 1:1 computer-generated randomisation used variable block sizes of 4 and 6 and was stratified by centre and age (≤3 months versus >3 months).
• Population:
  • Inclusion: infants aged younger than 6 months, admitted to a participating paediatric intermediate care or intensive care unit for less than 24 hours, with acute viral bronchiolitis requiring HFNC and persistent moderate to severe respiratory distress.
  • Severity threshold: modified Wood Clinical Asthma Score (mWCAS) ≥3 and/or hypercapnic acidosis during the preceding 6 hours, defined as pH <7.35 with Pco₂ >50 mm Hg (>6.7 kPa).
  • Principal exclusions: an immediate indication for non-invasive or invasive ventilation; pH <7.25; inability to maintain SpO₂ >92% regardless of FiO₂; more than 3 significant apnoeas per hour; severe impairment of consciousness; more than 3 hours of prone positioning before randomisation; significant respiratory, upper-airway, neuromuscular or haemodynamically important congenital cardiac disease; or a contraindication to prone positioning such as recent abdominal surgery or sternotomy.
  • Written parental or guardian consent was required; deferred completion of the second parent’s written consent was permitted after documented telephone agreement.
• Intervention:
  • Infants were placed prone immediately after randomisation, with the chest supported on the bed and the abdomen free, for a target of at least 24 cumulative hours during the first 48 hours.
  • Supine or lateral intervals were permitted between prone sessions according to care needs and infant tolerance; subsequent positioning after 48 hours was left to the treating team.
  • All infants received cardiorespiratory monitoring and standardised HFNC at 2 L/kg/min; FiO₂ was titrated to an SpO₂ of 92% to 97%, and humidification was set at 37°C.
  • Enteral feeding was encouraged according to local unit practice. Sedative or anxiolytic use was protocol-restricted and uncommon.
• Comparison:
  • Infants were managed in the supine position, with lateral positioning permitted for routine care.
  • They received the same HFNC flow, oxygen saturation target, humidification, monitoring, feeding approach and general supportive care as the intervention group.
• Blinding: Infants, parents and treating clinicians could not be blinded to body position. The primary outcome was the clinician’s actual decision to escalate to non-invasive or invasive ventilation within 72 hours. Prespecified escalation criteria included an mWCAS increase of ≥1 point, pH ≤7.30 with Pco₂ ≥8 kPa, FiO₂ >60% to maintain SpO₂ >92%, more than 3 significant apnoeas per hour, or a Paediatric Glasgow Coma Scale score <12. A separate three-member paediatric committee, blinded to group allocation, adjudicated the secondary treatment-failure outcome.
• Statistics: Assuming escalation rates of 39.0% with supine positioning and 25.4% with prone positioning—a 35% relative reduction—183 infants per group were required for 80% power at a two-sided 5% significance level. Allowing for a planned futility analysis and 15% protocol deviation increased the target to 226 per group, or 452 in total. The primary analysis was intention-to-treat using mixed-effects logistic regression, adjusted for age stratum with centre as a random intercept; a secondary per-protocol analysis and an age-interaction analysis were planned.
• Follow-Up Period: The primary outcome window was 72 hours after randomisation. Clinical outcomes and adverse events were followed to paediatric intermediate/intensive care and hospital discharge; the last participant was discharged on 11 December 2023.

Key Results

This trial continued to completion and was not stopped early. It randomised 451 of the planned 452 infants; 446 with valid completed consent were included in the primary intention-to-treat analysis (220 prone; 226 supine).

• The primary estimate favoured prone positioning, but the confidence interval crossed the null and ranged from a clinically important reduction in escalation to a small possible increase; the trial therefore did not establish efficacy.
• The absence of benefit across blinded treatment failure, oxygenation, support duration and length-of-stay outcomes weakens reliance on the primary point estimate alone.
• The apparently favourable per-protocol result cannot identify a true responder group because tolerance and exposure duration were post-randomisation variables strongly influenced by early deterioration.

Internal Validity

• Randomisation and Allocation: The computer-generated sequence used blocks of 4 and 6, with stratification by centre and age and central implementation through the electronic case-report system. This made foreknowledge of the next assignment unlikely and produced good pre-randomisation balance, although the manuscript did not detail the technical sequence-access safeguards.
• Post-randomisation Exclusions: Five of 451 randomised infants were excluded because complete written consent could not be obtained: 1/221 in the prone group and 4/230 in the supine group. The reported intention-to-treat population was therefore technically a consent-qualified modified intention-to-treat population. No infant in the 446-patient primary population was lost for the primary outcome, making material attrition bias unlikely.
• Performance and Detection Bias: Blinding of positioning was impossible. The actual escalation outcome was enacted by unblinded clinicians and therefore remained vulnerable to clinician expectation, comfort with HFNC and interpretation of partly subjective criteria. Blinded independent adjudication of treatment failure was a valuable safeguard, but it applied to a secondary outcome rather than the primary outcome.
• Protocol Adherence: Delivery was incomplete. Among 218 prone-assigned infants with duration data, median prone exposure during the first 48 hours was 25 hours (IQR 11 to 33), but 44/218 (20.2%) received less than 8 hours. Among 210 with switch-status data, 89 (42.4%) were definitively returned to supine before completing 24 hours.
• Baseline Characteristics: Groups were clinically similar at baseline: median age 40 versus 43 days; median mWCAS 4.0 in both groups; median HFNC FiO₂ 30% in both groups; and median Pco₂ 52 mm Hg in both groups. RSV was somewhat more frequent in the prone group (73.4% versus 65.9%), but no major prognostic imbalance was evident.
• Illness Severity: The enrolled infants were less likely to escalate than anticipated. The overall escalation rate was 18%, and the supine-group rate was 20.8%, compared with 39% assumed in the power calculation. Median FiO₂ was only 30%, hypoxaemia was uncommon and only 3 infants were intubated, limiting the number of hard events that the intervention could prevent.
• Heterogeneity: Bronchiolitis is a syndrome encompassing different viruses, obstructive and restrictive mechanics, respiratory-effort phenotypes and illness trajectories. PROPOSITIS stratified by age and centre but did not enrich or stratify by measured respiratory effort or lung-mechanical response. There was no age-treatment interaction (P=0.61), but the >3-month subgroup was small and the estimate was imprecise.
• Timing: Randomisation occurred within 24 hours of unit admission, at a median of 2 days from symptom onset, and prone positioning began immediately. This was a credible early-intervention window, before most escalation events.
• Dose: The selected dose was at least 24 cumulative prone hours in 48 hours, but the optimal duration and continuity of prone sessions were unknown. Although the median exposure met the target, the wide IQR and high early-switch rate show that many infants did not receive the intended dose.
• Separation of the Variable of Interest: The intervention group achieved a median 25 prone hours (IQR 11 to 33) in the first 48 hours and 27 hours (IQR 14 to 39) over the entire paediatric intermediate/intensive care stay. In the control group, only 6/226 infants received at least 8 prone hours. Thus, overall positional separation existed, but it was markedly diluted in a substantial minority of intervention patients.
• Crossover: Six supine-assigned infants received at least 8 hours prone, while 44 prone-assigned infants received less than 8 hours prone and 2 had missing exposure data. Crossover and non-adherence appropriately remained within the intention-to-treat strategy estimate, but they substantially compromised any attempt to estimate the biological effect of sustained proning.
• Adjunctive Therapy: Co-interventions were generally low and not grossly imbalanced: anxiolytic/sedative drugs were used in 6.8% versus 4.4%, antibiotics in 29.8% versus 33.2%, corticosteroids in 1.8% versus 0.4%, salbutamol in 1.8% versus 2.2%, and physiotherapy in 6.4% versus 11.1% in prone and supine groups, respectively.
• Outcome Assessment: Escalation to positive-pressure support is clinically meaningful but is not a fully objective biological endpoint. The discordance between 112 adjudicated failures and 80 actual escalations demonstrates that clinicians did not uniformly escalate when criteria were met. The very low intubation rate means the primary endpoint mainly measured the threshold for starting non-invasive ventilation.
• Statistical Rigour: The primary intention-to-treat mixed-effects model matched the prespecified design and adjusted for stratification variables. The protocol and statistical analysis plan were finalised before database lock, and the recruitment target was essentially achieved. However, the unexpectedly low event rate materially reduced power; most secondary comparisons were not multiplicity-adjusted; and the published per-protocol definition was changed post hoc and excluded unequal proportions from the two groups.

Conclusion on Internal Validity: Internal validity is moderate for the primary intention-to-treat strategy estimate: randomisation, baseline balance and near-complete follow-up were strong, but the unblinded clinician-driven endpoint, substantial non-adherence and lower-than-expected event rate limit certainty. Internal validity is limited for the favourable per-protocol estimate because it conditions on post-randomisation tolerance and exposure.

External Validity

• Population Representativeness: The cohort was highly relevant to paediatric critical care—median age 41 days, 84.3% aged ≤3 months and all receiving HFNC for moderate to severe bronchiolitis—but it represented a selected subset. Of 3268 bronchiolitis admissions, only 829 were assessed and 451 were randomised.
• Important Exclusions: Results do not directly apply to older infants or toddlers, children with important chronic respiratory, upper-airway, neuromuscular or cardiac disease, infants already requiring non-invasive or invasive ventilation, postoperative patients, or those with a contraindication to proning.
• Setting: All sites were French paediatric intermediate or intensive care units with continuous monitoring, ready access to escalation and staff able to reposition infants. Applicability to general wards, emergency departments without equivalent monitoring, transport settings and resource-limited environments is uncertain.
• Practice Dependence: The primary endpoint depends on local thresholds for HFNC failure and initiation of positive-pressure support. Clinician experience with HFNC, nurse-to-patient ratios and access to continuous positive airway pressure may differ materially between healthcare systems.
• Changing RSV Epidemiology: Recruitment occurred from 2021 to 2023, before widespread contemporary RSV monoclonal-antibody prevention. Nirsevimab and other preventive strategies may reduce severe RSV disease and further lower future escalation rates, altering both the case mix and the absolute benefit attainable from prone positioning.
• Safety Context: These data concern supervised prone positioning in monitored hospital care. They must not be extrapolated to unsupervised sleep at home; parents were specifically counselled about sudden infant death risk and supine sleeping after discharge.

Conclusion on External Validity: Generalisability is moderate for very young, otherwise relatively healthy infants receiving HFNC in closely monitored paediatric critical care environments. It is substantially more limited for older children, important comorbidity, ward-based care, patients already on positive-pressure support and healthcare systems with different escalation practices.

Strengths & Limitations

• Strengths:
  • The largest randomised clinical-outcome trial of prone positioning in bronchiolitis and one of the largest trials in critical bronchiolitis.
  • Multicentre participation across 15 tertiary and secondary hospitals, central stratified randomisation and an essentially complete primary-outcome dataset.
  • Standardised HFNC flow and oxygen targets in both groups, reducing respiratory-support co-intervention variability.
  • Clinically relevant escalation outcome with prespecified criteria and a complementary blinded adjudication committee for treatment failure.
  • Detailed measurement of actual prone exposure, crossover, tolerance, comfort, feeding and pressure-related harm, allowing feasibility to be judged rather than assumed.
  • Protocol and statistical analysis plan finalised before database lock, with transparent reporting of the post hoc per-protocol change.
• Limitations:
  • Open-label care and a primary outcome that depended on clinician behaviour rather than blinded adjudication.
  • Control-group escalation was 20.8%, far below the 39% assumed, materially reducing statistical power despite achieving the recruitment target.
  • Substantial intervention non-adherence: 42.4% were switched permanently supine before 24 hours and 20.2% received less than 8 prone hours.
  • The favourable per-protocol analysis used a post hoc definition and excluded 25% of prone-assigned versus 6% of supine-assigned infants.
  • Safety analyses reclassified infants by received exposure, yielding groups of 180 versus 264 and sacrificing the protection of randomisation for causal safety comparisons.
  • Only 3 intubations occurred, so the trial cannot establish an effect on invasive ventilation or other hard patient-centred outcomes.
  • Single-country practice, selected otherwise healthy infants and evolving RSV prevention constrain transportability.

Interpretation & Why It Matters

• Routine Practice

  PROPOSITIS does not support mandating prolonged prone positioning for every young infant with bronchiolitis receiving HFNC. The prespecified intention-to-treat strategy did not significantly reduce escalation, blinded treatment failure, oxygen requirements, support duration or length of stay.
• Residual Uncertainty

  The trial also does not prove that prone positioning is ineffective. The primary point estimate was compatible with a potentially important benefit, and the absolute risk-difference confidence interval extended to a 12.89-percentage-point reduction, but the evidence was imprecise and lacked corroboration from secondary outcomes.
• Feasibility Is Part of Effectiveness

  Poor tolerance is not merely protocol “noise”; it is part of the real-world effect of adopting a routine proning strategy in awake infants. The intention-to-treat estimate appropriately includes agitation, cannula displacement concerns, care interruptions and early return to supine.
• Selected Responders

  The lower escalation rate among infants who accumulated at least 8 prone hours is hypothesis-generating, not proof that proning works “when tolerated”. Tolerance, duration and clinical improvement occur after randomisation and are entangled with the risk of early failure.
• Future Research

  Further trials should predefine the estimand, incorporate objective respiratory-effort or lung-mechanical phenotyping, distinguish a short physiological response test from prolonged treatment, and account for falling HFNC-failure rates. The accompanying editorial similarly emphasised phenotypic heterogeneity, clinician experience and incomplete tolerance as central design issues.²

Controversies & Other Evidence

• The per-protocol “responder” interpretation is causally fragile: the published per-protocol population was redefined post hoc to exclude infants who did not achieve the required exposure rather than reassigning them. Early deterioration both increases escalation risk and shortens the opportunity to remain prone, creating reverse causation and selection bias. Excluding 25% of prone-assigned infants but only 6% of supine-assigned infants can generate an apparently favourable comparison even without a true biological effect.²
• Escalation was a management decision as well as a disease outcome: 112 infants fulfilled blinded adjudicated failure criteria, but only 80 were escalated. With only 3 intubations, the primary endpoint largely captured when clinicians chose to initiate non-invasive positive pressure, a threshold influenced by local practice and growing confidence with HFNC.
• The background event rate changed faster than the trial assumptions: in closely related French studies with similar entry criteria and baseline severity, HFNC failure or escalation fell from approximately 51% in TRAMONTANE to 39% in TRAMONTANE 2 and approximately 25% by PROPOSITIS. This secular drift plausibly reflects broader HFNC use, changing case selection and increased clinician comfort, and it explains why a trial powered around a 39% control rate became less definitive.²³⁴
• Bronchiolitis is unlikely to have a uniform positional response: the preceding crossover physiology study showed a group-level reduction in respiratory effort, but only 8 of 14 infants responded. PROPOSITIS did not measure the respiratory-effort phenotype that might identify such responders, while the accompanying editorial highlighted age, virus, mechanics, body habitus and inflammatory phenotype as plausible effect modifiers.¹²
• More prone time cannot simply be prescribed away: the median exposure barely exceeded the protocol target, 42.4% were permanently returned supine before 24 hours, and infants who switched early had a high escalation rate. This may reflect intolerance, impending failure or clinician concern; the trial cannot distinguish these mechanisms. Routine sedation to force adherence would constitute a different intervention with a different risk-benefit balance.
• Subsequent evidence: PROPOSITIS was newly published in 2026, and no confirmatory clinical-outcome randomised trial has yet established a benefit in a prospectively defined responder phenotype. The evidence therefore remains insufficient to make prolonged prone positioning a standard component of bronchiolitis care.

Summary

• PROPOSITIS randomised 451 young infants with moderate to severe bronchiolitis receiving HFNC across 15 French paediatric intermediate/intensive care units; 446 entered the primary analysis.
• Escalation to non-invasive or invasive ventilation occurred in 15.0% with prone versus 20.8% with supine positioning: adjusted OR 0.66; 95% CI 0.40 to 1.07; P=0.09.
• Blinded treatment failure, oxygenation, respiratory-support duration, unit stay and hospital stay were not significantly different; only 3 infants were intubated.
• Intervention delivery was difficult: 42.4% were permanently switched supine before 24 hours and 20.2% accumulated less than 8 prone hours. The favourable per-protocol result is vulnerable to major post-randomisation selection bias.
• Serious adverse events were rare and unrelated to positioning; skin lesions were more frequent with prone exposure but were predominantly minor.

Overall Takeaway

PROPOSITIS is an important and unusually rigorous test of a simple physiological intervention, but it does not justify routine prolonged prone positioning for all infants with bronchiolitis receiving HFNC. Its principal contribution is to show that possible benefit, poor tolerance, clinician-dependent escalation and a rapidly falling background event rate must all be incorporated into the treatment effect; selected benefit remains plausible, but unproven.

Bibliography

• 1.Baudin F, Emeriaud G, Essouri S, et al. Physiological effect of prone position in children with severe bronchiolitis: a randomized cross-over study (BRONCHIO-DV). J Pediatr. 2019;205:112-119.e4.
• 2.Kohne JG, Conway SR, Shein SL. Prone positioning in viral bronchiolitis—back to the back? JAMA. 2026;336(4):290-292.
• 3.Milési C, Essouri S, Pouyau R, et al; Groupe Francophone de Réanimation et d’Urgences Pédiatriques. High flow nasal cannula (HFNC) vs nasal continuous positive airway pressure (nCPAP) for the initial respiratory management of acute viral bronchiolitis in young infants: a multicenter randomized controlled trial (TRAMONTANE study). Intensive Care Med. 2017;43(2):209-216.
• 4.Milési C, Pierre AF, Deho A, et al; GFRUP Respiratory Study Group. A multicenter randomized controlled trial of a 3-L/kg/min vs 2-L/kg/min high-flow nasal cannula flow rate in young infants with severe viral bronchiolitis (TRAMONTANE 2). Intensive Care Med. 2018;44(11):1870-1878.