Clinical development · planning tool
From hypothesis to powered efficacy study.
AIRCHILL has no demonstrated patient efficacy. This page turns the current indication hypotheses into explicit study-design assumptions that can be challenged: control event rate, expected absolute benefit, statistical power, attrition, recruitment speed, follow-up and cost. The output is planning arithmetic, not a protocol or regulatory decision.
Development logic
Three gates before an efficacy claim.
Can the intended thermal dose be delivered safely?
Technical performance, airway/pulmonary safety, workflow, temperature trajectory, ventilation stability and stopping rules. Typically tens of patients, not powered for clinical benefit.
Which population and thermal strategy deserve a pivotal test?
Randomized multicentre work to select timing, target, duration and phenotype. Biomarkers and clinical endpoints can estimate event rates and plausible effect size for the confirmatory study.
Patient-relevant endpoint, prespecified estimand, adequate power.
The pivotal question must match the intended claim. Final sample size should be locked with a statistician and regulators using validated software and current control-event data.
Indication concepts
What an efficacy study would need to prove.
| Indication | Research population | Comparator | Preferred patient-relevant endpoint | Follow-up | Key design issue |
|---|---|---|---|---|---|
| Cardiac arrest | Adults requiring invasive ventilation after OHCA; phenotype/timing prespecified. | Same ventilation and guideline-based temperature care without active AIRCHILL cooling. | Favourable neurological outcome at ~90 days; blinded outcome assessment. | 3 months | Very early randomisation, emergency-consent pathway, avoid mixing fundamentally different arrest phenotypes. |
| Ischaemic stroke | Intubated LVO patients undergoing EVT; randomise before or at thrombectomy. | Standard anaesthesia/ventilation and temperature management. | 90-day mRS; ordinal shift is generally preferable to a dichotomy. | 3 months | The web calculator uses mRS 0–2 as a binary approximation; a pivotal ordinal design should be simulation-based. |
| Neonatal HIE | Separate paediatric programme only; eligible neonates receiving current standard whole-body cooling. | Current standard servo-controlled neonatal hypothermia. | Death or moderate/severe neurodevelopmental impairment. | 18–24 months | AIRCHILL is not neonatal-validated; paediatric engineering and a separate regulatory programme are prerequisites. |
| Heat stroke | Selected severe/intubated patients where adjunct respiratory cooling can ethically be studied. | Best available rapid active cooling. | Clinical recovery / organ-failure-free survival; mortality as key secondary or larger confirmatory endpoint. | 30 days | Water immersion remains best-supported for exertional heat stroke when feasible; recruitment is seasonal and sparse. |
| Severe TBI | Intubated severe TBI with prespecified phenotype, bleeding and haemodynamic safeguards. | Guideline-based neurocritical care without active AIRCHILL cooling. | GOSE at 6 months; ordinal analysis preferred. | 6 months | ICP reduction is not sufficient; function and safety must drive the claim. |
| Concussion / mild TBI | Separate non-invasive technology pathway, not the current AIRCHILL ventilator. | Usual concussion management. | Persistent symptom burden / time to clinical recovery or return to activity. | 1–3 months | Typical patients are not intubated; this should not be used as an AIRCHILL efficacy programme. |
External design benchmark · PRINCESS2
The most useful lesson is not an effect size. It is the architecture of the trial.
PRINCESS2 does not provide evidence for AIRCHILL. It is, however, unusually relevant as a design benchmark for a prehospital cooling-device programme: a narrow cardiac-arrest phenotype, on-scene randomisation, a time-critical intervention, treatment continuity into hospital care, blinded neurological follow-up, a prespecified pilot phase and independent safety/futility oversight.
Do not mix arrest populations by convenience.
PRINCESS2 restricts the confirmatory question to OHCA with an initial shockable rhythm after earlier studies generated a subgroup signal. For AIRCHILL, intended population and timing phenotype should be frozen before pivotal powering; materially different arrest phenotypes should be separated or prospectively stratified.
Record the clock, not just the temperature.
The protocol randomises at the scene and requires transnasal cooling within 20 minutes of EMS arrival. AIRCHILL feasibility work should prospectively capture emergency call, CPR, airway management, randomisation, device start, ROSC, transport, hospital handover and the first measurable thermal effect. Delay is an exposure variable, not background noise.
Prehospital treatment cannot end at the ambulance door.
PRINCESS2 continues the assigned cooling intervention through transport and hospital arrival until transition to institutional temperature management. AIRCHILL should likewise prespecify device-to-ICU handover, ventilation continuity, temperature trajectory and rewarming so that treatment gaps do not become an uncontrolled co-intervention.
Test conduct and safety before testing efficacy.
The first 100 PRINCESS2 participants formed a prespecified pilot focused on protocol adherence and safety; primary and secondary efficacy endpoints were deliberately not analysed. AIRCHILL should use the same logic: predefined feasibility and safety gates before a larger patient-outcome trial.
Standardise what happens after ROSC.
Post-resuscitation care, neuroprognostication and withdrawal-of-life-support procedures are protocolised, while the 90-day neurological assessment is blinded to treatment allocation. This matters because downstream ICU decisions can otherwise overwhelm the effect of an early device intervention.
Ambulance operations are part of device performance.
The 2026 pilot reported 92% overall protocol adherence and no device-related serious adverse events, but it also documented minor treatment interruptions during real-world prehospital and in-hospital use. AIRCHILL pilot testing should therefore stress fixation, energy/gas reserve, circuit handling, condensation, alarms, transport vibration, reserve runtime and handover—not only benchtop cooling capacity.
Study-family sources: PRINCESS2 protocol v1.1 · PRINCESS2 design paper · PRINCESS2 100-patient pilot (2026) · PRINCESS randomized trial · time-to-cooling subanalysis · PRINCE + PRINCESS pooled analysis. These studies test transnasal evaporative cooling, not AIRCHILL.
Operational evidence requirements
A prehospital device study has to prove more than a temperature change.
For AIRCHILL, usability, timing, measurement quality, ventilation continuity and handover are part of the clinical evidence package. These are not substitutes for patient outcomes; they are the variables that determine whether an efficacy result can be interpreted and reproduced.
Measure a trajectory, not a single temperature.
Prespecify the measurement hierarchy, sampling frequency, clock synchronisation and the thermal variables that matter: time to first measurable effect, rate of change, depth, exposure duration, distribution and rewarming. Continuous device timestamps should be retained so thermal dose can be reconstructed patient by patient.
Use errors and set-up burden are safety data.
Capture critical tasks, set-up time, failed starts, interruptions, alarm handling, circuit changes and recovery from foreseeable use errors in the actual EMS and transport environment. Human-factors engineering should run in parallel with clinical development rather than after the device design is frozen.
The cooling intervention must not degrade the primary life-support function.
Track airway pressure, flow, tidal volume, minute ventilation, FiO2, gas exchange, resistance, condensation and clinically relevant interruptions. Any cooling benefit would be uninterpretable if achieved at the cost of unstable ventilation.
Every important clock should be captured.
Record emergency call, EMS arrival, CPR milestones, airway placement, randomisation, device start, first thermal effect, ROSC, departure, hospital arrival and transition to definitive temperature management. Timing is part of the intervention exposure, not merely a baseline characteristic.
Design the ambulance-to-hospital handover into the protocol.
Prespecify what happens to ventilation and thermal management during transport, emergency-department transfer, imaging or catheter procedures, ICU admission and eventual rewarming. Treatment gaps and crossovers should be measured as protocol variables.
Choose sites that can actually deliver the intended exposure.
Feasibility should include EMS workflow, expected eligible volume, transport times, training burden, hospital handover capability and protocol adherence. A site mix dominated by exposures too short to deliver the intended thermal dose can make a technically valid device look clinically inert.
Conduct and safety before neurological efficacy.
Use a prespecified early-feasibility cohort with go/no-go criteria for device deployment, protocol adherence, ventilation stability, thermal performance and device-related adverse events before expanding into an outcome-powered trial.
Collect resource use alongside efficacy.
For later confirmatory work, prospectively capture ventilation days, ICU and hospital length of stay, major complications and discharge disposition in addition to neurological outcomes. These variables can support later health-economic analyses without changing the primary clinical claim.
Primary references: ISO 14155:2026 · IEC 62366-1:2015 + Amd 1:2020 · FDA human factors guidance · PRINCESS2 pilot · ERC–ESICM 2025 post-resuscitation care.
Interactive power, recruitment & budget lab
Change the assumptions. Watch the trial change.
The calculator uses a two-sided superiority approximation for two independent proportions. It is deliberately transparent and conservative: final protocol power must be reproduced in validated statistical software and adapted to the actual endpoint model.
Sample-size method: normal approximation using Cohen’s h for two independent proportions with the selected allocation ratio, then inflated for attrition. The tool does not adjust for interim analyses, multiplicity, covariates, centre effects, competing risks, ordinal endpoints, cluster randomisation or non-inferiority margins. Those require a protocol-specific statistical model.
Planning framework: EU MDR 2017/745 Annex XV; ISO 14155:2026; ICH E9 / E9(R1); FDA pivotal-device-study guidance and IDE guidance. All default control rates, recruitment speeds and costs on this page are planning anchors and remain editable. They are not regulatory commitments, quotations or AIRCHILL efficacy data. See the site-wide source, assumption and data-gap register.