PROSpect’s Negative Findings Should Change How We Design Pediatric Critical Care Trials Forever

A pediatric intensivist at one of the 75 participating ICUs had already made the call. The child, moderate-to-severe PARDS, early in the disease course, was being positioned prone. It was the reasonable thing to do. The adult literature supported it. The institutional protocol allowed it. And for years, absent definitive pediatric data, that judgment call had been the standard: extrapolate from adults, document your rationale, and hope the physiology transferred. Then PROSpect published in The Lancet, and the honest answer to “does this work in children?” became: no.

The PROSpect trial, Prone and Oscillation Pediatric Clinical Trial, enrolled 424 children with early, moderate-to-severe pediatric acute respiratory distress syndrome (PARDS) across 75 pediatric intensive care units in 18 countries. It used a two-by-two factorial, response-adaptive, Phase 3 randomized controlled trial design to test two interventions simultaneously: prone positioning and high-frequency oscillatory ventilation (HFOV). The primary outcome was ventilator-free days at 28 days. Neither intervention improved it. The trial stopped early due to futility and slow enrollment, having reached 424 of a planned 600 patients. The conclusion was unambiguous: routine use of prone positioning or HFOV as first-line strategies in this population is not recommended.

That verdict matters for bedside medicine. But for the clinical trial community, the more important story lives inside the methodology, because PROSpect is among the most operationally ambitious pediatric critical care trials in recent memory, and understanding how it was built is the prerequisite for understanding what comes next.

The Design Decision That Made This Possible

Pediatric critical care is among the hardest environments in which to run a randomized controlled trial. The patient population is small relative to adult ICU populations, disease acuity demands rapid protocol decisions, and the ethical threshold for enrolling children is appropriately high. These constraints have historically produced underpowered trials, opportunistic observational studies, and the kind of expert consensus that fills guidelines when evidence is absent. The PALICC-2 guidelines, the Second International Guidelines for the Diagnosis and Management of Pediatric Acute Respiratory Distress Syndrome, published in Pediatric Critical Care Medicine in February 2023, addressed prone positioning and HFOV with the careful language of committees working around evidence gaps, not through them.

PROSpect attacked those constraints directly with two design choices that deserve close attention from anyone planning a high-acuity trial.

The factorial design was the first. Rather than run two sequential trials, prone positioning in one, HFOV in another, the investigators tested both interventions within a single trial structure, assigning participants to one of four arms: supine plus conventional ventilation, prone plus conventional ventilation, supine plus HFOV, or prone plus HFOV. This is not merely an efficiency gain. Factorial designs in this setting generate interaction data: does prone positioning work differently when combined with HFOV than when combined with conventional ventilation? That question cannot be answered by two separate trials. It requires the two-by-two structure. For sponsors considering multi-intervention trials in rare or high-acuity populations, CNS, neonatal, pediatric oncology, factorial methodology offers the chance to answer two questions simultaneously without doubling enrollment burden on a constrained patient pool.

The response-adaptive randomization was the second. As the trial accumulated data, allocation probabilities shifted based on emerging outcome signals, increasing the proportion of patients randomized to better-performing arms. This is the kind of adaptive feature that has been discussed in regulatory circles while the field has been slow to implement in practice. PROSpect shows the implementation is achievable across 18 countries and 75 sites. That is not a trivial demonstration.

What 424 Children Across 18 Countries Actually Proved

Specifically, 126 patients were assigned to the supine arm, with the remaining patients distributed across the prone and HFOV combinations. The trial found no evidence of benefit for prone positioning on ventilator-free days at 28 days, and no evidence of benefit for HFOV. These results held across the factorial arms, meaning the interaction between the two interventions did not rescue either one, prone plus HFOV did not produce an effect that prone alone or HFOV alone missed.

Here is the counterintuitive point that the field needs to sit with: this negative finding is more valuable than a positive finding would have been. Consider what a positive result would have delivered, confirmation of a practice already widely used, incremental validation of the adult extrapolation, perhaps a modest update to the PALICC guidelines. A positive PROSpect would have been welcome and quickly forgotten. The negative result does something harder and more important: it forces de-adoption. It tells the intensivist in that PICU that the call she made, prone positioning as first-line strategy, rested on borrowed evidence that did not transfer. The physiological rationale for prone positioning in adults (improved ventilation-perfusion matching, recruitment of posterior lung segments) apparently does not produce the same ventilator-free day gains in the pediatric ARDS phenotype. That gap between mechanistic plausibility and clinical outcome is exactly what a Phase 3 trial exists to expose.

The trial’s early termination at 424 of 600 planned patients raises a legitimate methodological question: does the futility stopping boundary hold at a sample size that fell 30% short of the original plan? The investigators and the DSMB concluded yes, and the response-adaptive design’s interim monitoring framework supported that conclusion. But trial designers should note the enrollment challenge, slow accrual in pediatric critical care is not a logistical failure specific to PROSpect; it is a structural feature of the space. The 75-site, 18-country footprint was assembled precisely to fight that problem, and it still produced an enrollment shortfall. Any sponsor planning a comparable high-acuity pediatric trial needs to build that friction into their power assumptions from day one, not treat 600 planned as 600 expected.

The Operational Lessons That Don’t Make the Abstract

Running a response-adaptive randomized trial across 75 pediatric ICUs in 18 countries requires a randomization and trial supply management infrastructure that most CROs treat as an afterthought until the first interim analysis triggers an allocation shift. In PROSpect, that shift had to propagate in near-real time to sites operating across multiple time zones, regulatory jurisdictions, and institutional review frameworks. The fact that the trial completed, even at 424 of 600, is a proof of concept for global adaptive trial operations in a pediatric critical care setting. That proof has direct relevance for sponsors designing CNS trials in pediatric epilepsy, neonatal hypoxic-ischemic encephalopathy, or pediatric septic shock, where the patient population is similarly constrained and the clinical urgency is similarly high.

The site activation model also deserves scrutiny. Seventy-five PICUs is a large footprint for any pediatric trial, and maintaining protocol fidelity across that network, ensuring that “prone positioning per protocol” means the same thing in a PICU in São Paulo as in Toronto as in Rotterdam, requires a site training and monitoring architecture that cannot be improvised. Protocol deviations in positioning trials are particularly insidious because the intervention is procedural rather than pharmacological: there is no plasma concentration to audit, no pill count to verify. The monitoring burden falls on process documentation, staff training records, and real-time observation, all of which scale poorly across 18 countries without a deliberately built oversight infrastructure.

For sponsors in the decentralized and hybrid trial space, PROSpect offers an uncomfortable reminder: some trials cannot be decentralized. A study of prone positioning in a ventilated pediatric patient is, by definition, a site-bound, high-acuity intervention. The patient cannot be at home. The intervention cannot be remote. The monitoring cannot be passive. Before the field reflexively applies DCT methodology to every indication, PROSpect is a useful case study in what rigorous site-based trial operations look like when the clinical context demands them, and in the organizational investment required to sustain them across a global network.

The practical directive for sponsors designing adaptive trials in rare or pediatric populations is this: pre-specify your futility boundaries with the same rigor you apply to your efficacy boundaries, build your enrollment projections around observed accrual rates from comparable trials rather than theoretical site capacity, and plan for your response-adaptive randomization system to be tested at interim, not just at baseline. PROSpect’s DSMB triggered a futility stop at 70% enrollment. Your DSMB can too. If your statistical analysis plan is not written to handle that scenario cleanly, the trial’s credibility suffers regardless of the science.

Back in that PICU, the intensivist now has an answer she did not have before PROSpect. The prone positioning call she made, reasonable, evidence-informed by adult data, defensible, turns out to have been wrong for her patient population. That is not a criticism of her clinical judgment. It is precisely what Phase 3 randomized evidence exists to correct. The question for the trial design community is whether the next therapeutic hypothesis in pediatric critical care waits another decade for a definitive answer, or whether the methodological template PROSpect established, factorial, response-adaptive, 75 sites, 18 countries, early futility monitoring, gets replicated before the evidence gap claims more borrowed certainty and more children treated on extrapolation rather than proof.

References

1. The Lancet, “Prone positioning and high-frequency oscillatory ventilation in early paediatric acute respiratory distress syndrome (PROSpect): an international, two-by-two factorial, response-adaptive, phase 3, randomised controlled trial”
2. Clinical Trial Vanguard, “PROSpect Proved Two Treatments Don’t Work. The Real Story Is What That Took.”
3. Clinical Trial Vanguard, “PROSpect’s Negative Findings Are the Most Important Data in Pediatric Critical Care This Decade”
4. Clinical Trial Vanguard, “When the Standard of Care Fails the Trial: What PROSpect’s Negative Findings Reveal About Pediatric Critical Care’s Evidence Crisis”
5. Pediatric Experts, “PALICC-2 Executive Summary: Second International Guidelines for the Diagnosis and Management of Pediatric Acute Respiratory Distress Syndrome,” Pediatric Critical Care Medicine 24(2):143-168, February 2023