Skip to main content

Foundational Trials

HYPERION

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Targeted Temperature Management for Cardiac Arrest with Nonshockable Rhythm
  • Acronym: HYPERION
  • Year: 2019
  • Journal published in: The New England Journal of Medicine
  • Citation: Lascarrou JB, Merdji H, Le Gouge A, et al. Targeted Temperature Management for Cardiac Arrest with Nonshockable Rhythm. N Engl J Med. 2019;381(24):2327-2337.

Context & Rationale

  • Background
    • Non-shockable rhythms (asystole/pulseless electrical activity) constitute a large proportion of cardiac arrests and are associated with very poor neurological outcomes despite return of spontaneous circulation.
    • Earlier hypothermia-era evidence was largely derived from shockable out-of-hospital cardiac arrest cohorts; evidence in non-shockable rhythms, non-cardiac aetiologies, and in-hospital cardiac arrest was limited and conflicting.
    • Temperature management is resource-intensive and may introduce harms (infection, haemodynamic instability, arrhythmia), so a dedicated RCT in this high-mortality subgroup was methodologically and clinically important.
  • Research Question/Hypothesis
    • In comatose adults resuscitated from cardiac arrest with a non-shockable rhythm, does targeted hypothermia at 33°C (for 24 hours) improve favourable neurological outcome at 90 days compared with targeted normothermia at 37°C (for 48 hours)?
  • Why This Matters
    • This is a high-burden population with a low baseline probability of meaningful recovery; even small absolute gains in good neurological survival could be practice-changing.
    • Clarifies whether “deeper” cooling (33°C) offers incremental benefit beyond active normothermia/fever prevention strategies in non-shockable arrest.

Design & Methods

  • Research Question: Does moderate therapeutic hypothermia (33°C) versus targeted normothermia (37°C) improve favourable neurological outcome (CPC 1–2) at day 90 in comatose adults after non-shockable cardiac arrest?
  • Study Type: Investigator-initiated, pragmatic, multicentre, randomised, controlled superiority trial; open-label with blinded outcome assessment; conducted in 25 ICUs in France; pre-trial protocol/design published1.
  • Population:
    • Setting: ICU admission after successful resuscitation from in-hospital or out-of-hospital cardiac arrest with a non-shockable rhythm (asystole, pulseless electrical activity, or unknown/not shocked).
    • Key inclusion: age ≥18 years; coma (Glasgow Coma Scale ≤8); no-flow time ≤10 minutes; low-flow time ≤60 minutes; time from cardiac arrest to screening ≤300 minutes.
    • Key exclusions (examples): moribund condition; Child–Pugh class C cirrhosis; pregnancy/breast-feeding; high risk of bleeding; lack of health insurance/guardianship constraints; severe haemodynamic instability (defined as norepinephrine >1 μg/kg/min or epinephrine >1 μg/kg/min); prior inclusion in another cardiac-arrest RCT; logistical constraints.
  • Intervention:
    • Targeted hypothermia: core temperature target 33°C (acceptable range 32.5–33.5°C) for 24 hours after randomisation.
    • Cooling commenced as soon as possible using available temperature management devices (surface and/or intravascular per site practice), with continuous core temperature monitoring.
    • Controlled rewarming after the 24-hour maintenance phase, then temperature maintained in the normothermic range (36.5–37.5°C) through 48 hours post-randomisation.
    • Anti-shivering strategy allowed (sedation/analgesia, and neuromuscular blockade if required); sedation was routinely recommended only during the first 12 hours in the protocol.
  • Comparison:
    • Targeted normothermia: core temperature target 37°C (acceptable range 36.5–37.5°C) actively maintained for 48 hours after randomisation (including active fever prevention).
    • Co-interventions for post–cardiac arrest care were provided according to local practice, including haemodynamic support, ventilation strategies, and neurological prognostication pathways.
  • Blinding: Treating teams were unblinded (temperature management is difficult to mask); neurological outcome at day 90 (CPC) was assessed by a blinded assessor (psychologist) to reduce detection bias.
  • Statistics: A total of 584 patients were required to detect a 9% absolute increase in favourable neurological outcome at day 90 (from 14% to 23%) with 80% power and a two-sided alpha of 5%; interim analyses used a conservative Peto–Haybittle boundary (P<0.001) with final alpha adjusted to 0.049; primary analysis followed a prespecified modified intention-to-treat approach (excluding post-randomisation withdrawals of consent; missing day-90 status classified as dead in the prespecified plan).
  • Follow-Up Period: 90 days (primary endpoint), with in-ICU and in-hospital outcomes also collected.

Key Results

This trial was not stopped early. It completed enrolment (584 randomised; 581 analysed after 3 post-randomisation withdrawals of consent; missing day-90 status was classified as death per prespecified plan).

Outcome Hypothermia (33°C) Normothermia (37°C) Effect p value / 95% CI Notes
Primary: favourable neurological outcome (CPC 1–2) at day 90 29/284 (10.2%) 17/297 (5.7%) Absolute risk difference 4.5 percentage points 95% CI 0.1 to 8.9; P=0.04 Blinded day-90 assessment; low event rates; missing day-90 status counted as dead per prespecified plan.
Death by day 90 231/284 (81.3%) 247/297 (83.2%) Absolute risk difference −1.9 percentage points 95% CI −8.0 to 4.4; P not reported No mortality benefit despite improvement in CPC 1–2.
Death in ICU (time-to-event analysis) 228/284 (80.3%) 239/297 (80.5%) HR 0.93 95% CI 0.78 to 1.10 No clear ICU survival effect.
Survival to hospital discharge (time-to-event analysis) 56/284 (19.7%) 50/297 (16.8%) HR 1.19 95% CI 0.81 to 1.74 Directionally favours hypothermia but imprecise.
Vasopressor requirement (days 0–7) 233/284 (82.0%) 241/297 (81.1%) HR 1.01 95% CI 0.92 to 1.11; P=0.81 No signal for excess haemodynamic support needs.
Severe cardiac arrhythmia (days 0–7) 35/284 (12.3%) 31/297 (10.4%) HR 1.19 95% CI 0.74 to 1.91; P=0.48 No statistically significant difference; CIs compatible with harm or no effect.
Ventilator-associated pneumonia (days 0–28) 44/284 (15.5%) 35/297 (11.8%) HR 1.34 95% CI 0.86 to 2.01; P=0.19 Numerically higher with hypothermia; imprecise estimate.
Mechanical ventilation duration (overall) Median 4.5 days (IQR 2.0–7.0) Median 4.0 days (IQR 2.0–7.0) Not reported Not reported Reported as descriptive summary; no between-group effect estimate provided.
ICU length of stay (overall) Median 4.0 days (IQR 2.0–9.0) Median 4.0 days (IQR 2.0–10.0) Not reported Not reported Similar central tendency; high early mortality limits interpretability.
  • Hypothermia increased favourable neurological outcome at 90 days from 5.7% to 10.2% (absolute difference 4.5 percentage points; 95% CI 0.1 to 8.9; P=0.04), but mortality remained very high (~82%) and numerically similar between groups.
  • Physiological “separation” was achieved: mean core temperature between 12 and 24 hours was 33.5 ± 1.1°C in the hypothermia group vs 37.0 ± 0.7°C in the normothermia group.
  • Safety signals were not definitive; ventilator-associated pneumonia was numerically higher under hypothermia, while severe arrhythmias and vasopressor use were broadly similar.

Internal Validity

  • Randomisation and allocation: Centralised, web-based randomisation with concealment; stratified by centre and location of arrest (in-hospital vs out-of-hospital); balance achieved across key baseline variables (median age 67 years; male 65.1% vs 63.3%; circulatory shock 56.0% vs 60.6%).
  • Dropout/exclusions: 3 post-randomisation withdrawals of consent were excluded from the modified intention-to-treat analysis; 3 total losses to follow-up were classified as dead per prespecified plan (1 in hypothermia; 2 in normothermia).
  • Performance/detection bias: Temperature strategy was necessarily unblinded to clinicians, creating scope for differential co-interventions; however, day-90 neurological outcome was assessed by a blinded assessor to mitigate detection bias.
  • Protocol adherence: All randomised patients received assigned intervention; early discontinuation of temperature management occurred more often in the hypothermia group (36/284; 12.7%) than the normothermia group (9/297; 3.0%).
  • Separation of the variable of interest: Median time from randomisation to target temperature was 317 minutes (IQR 214–477) with hypothermia vs 50 minutes (IQR 0–150) with normothermia; mean core temperature (12–24 hours) 33.5 ± 1.1°C vs 37.0 ± 0.7°C.
  • Timing and dose: Randomisation occurred relatively late post-arrest (median 232.5 minutes vs 219.0 minutes from arrest to randomisation), which may attenuate temperature-related neuroprotection if early reperfusion injury is a dominant mechanism.
  • Baseline illness severity and heterogeneity: High severity and heterogeneity were present (asphyxial aetiology ~55%; significant proportion in-hospital arrests; high prevalence of shock), which improves relevance but complicates mechanistic interpretation and may dilute benefit.
  • Withdrawal of life-sustaining therapy (WLST): Among those who died, WLST preceded death in 93.2% (hypothermia) vs 95.0% (normothermia); median time from randomisation to WLST decision was 5 days (IQR 3–6) vs 4 days (IQR 2–6), respectively, highlighting the potential for self-fulfilling prognostication effects even with structured assessment.
  • Statistical rigour: Prespecified interim monitoring and alpha adjustment were used; the primary endpoint was statistically significant but with small event counts and a confidence interval close to the null, increasing fragility and sensitivity to misclassification or small biases.

Conclusion on Internal Validity: Overall, internal validity is moderate: randomisation and outcome blinding strengthen causal inference, but open-label care, high WLST prevalence, late randomisation, and a fragile primary effect estimate limit confidence in the magnitude and reproducibility of benefit.

External Validity

  • Population representativeness: Includes both out-of-hospital and in-hospital non-shockable arrests, and a high proportion of non-cardiac/asphyxial causes (more representative of non-shockable epidemiology than earlier hypothermia trials), but with stringent no-flow/low-flow limits and exclusion of marked haemodynamic instability.
  • Applicability: Requires reliable ICU-based temperature management capability and structured post-arrest care; results likely generalise best to similar systems with high bystander CPR rates and protocolised neuroprognostication, and may translate less well to resource-limited settings or prolonged low-flow contexts.

Conclusion on External Validity: Generalisability is moderate: the enrolled cohort reflects many real-world non-shockable arrests, but strict inclusion thresholds and system-of-care dependencies may limit transfer to settings with longer downtimes or fewer temperature-management resources.

Strengths & Limitations

  • Strengths: Pragmatic multicentre design; explicit focus on non-shockable rhythm (historically under-tested); robust “temperature separation”; blinded day-90 neurological outcome assessment; prespecified interim monitoring and analysis plan.
  • Limitations: Open-label delivery with potential co-intervention and prognostication bias; borderline primary effect with small number of favourable outcomes; late randomisation relative to arrest; high WLST prevalence; mortality unchanged and absolute benefit small; pneumonia numerically higher under hypothermia.

Interpretation & Why It Matters

  • Clinical signal
    In non-shockable arrest, 33°C hypothermia produced a small absolute increase in good neurological outcome (CPC 1–2), without an accompanying mortality reduction; clinicians must weigh modest potential neurological benefit against complexity and possible infectious risk.
  • Interpretive nuance
    The comparison was not “cooling vs no temperature control” but rather 33°C vs actively maintained 37°C with fever prevention; the study therefore tests incremental depth of cooling rather than the value of temperature management per se.
  • Practice context
    In the current era, where fever prevention is standard and subsequent large trials/meta-analyses question routine hypothermia, HYPERION is best interpreted as hypothesis-strengthening for a narrowly defined subgroup rather than definitive endorsement of universal 33°C cooling for all non-shockable arrests.

Controversies & Subsequent Evidence

  • Borderline statistical signal with low event counts: The absolute risk difference for CPC 1–2 was 4.5 percentage points with a 95% CI of 0.1 to 8.9 and P=0.04; the small number of favourable outcomes increases fragility and susceptibility to minor biases or misclassification, a point emphasised in high-level commentaries23.
  • Open-label care and WLST interplay: Because temperature assignment could not be blinded to treating clinicians and most deaths followed WLST, concerns remain that differences in sedation, neurological prognostication timing, or perceived prognosis could influence observed CPC outcomes even with blinded assessment; contemporaneous editorial interpretation highlighted both the importance and the limitations of this signal4.
  • How to reconcile with later “normothermia-era” evidence: TTM2 (predominantly shockable out-of-hospital cardiac arrest) found no benefit of hypothermia versus normothermia with fever prevention, and reported higher arrhythmia risk with hypothermia, shifting the evidentiary centre of gravity away from routine 33°C cooling across unselected arrests5.
  • Meta-analytic synthesis after HYPERION: Recent systematic reviews and an individual patient data meta-analysis integrating modern temperature-control trials have generally supported fever prevention and questioned routine hypothermia, with heterogeneous effects and no consistent improvement in patient-centred outcomes across arrest subtypes (including non-shockable rhythms)67.
  • Guideline evolution: Post–cardiac arrest care guidance in Europe now places greater emphasis on active fever prevention and careful temperature control rather than routine induction of deep hypothermia for all patients, reflecting the totality of evidence including HYPERION and later trials/meta-analyses89.

Summary

  • HYPERION randomised 584 comatose adults after non-shockable cardiac arrest to 33°C for 24 hours vs 37°C for 48 hours, with day-90 CPC 1–2 as the primary endpoint.
  • Favourable neurological outcome at 90 days was 10.2% with hypothermia vs 5.7% with normothermia (absolute difference 4.5 percentage points; 95% CI 0.1 to 8.9; P=0.04).
  • Mortality was extremely high and similar between groups (81.3% vs 83.2% by day 90), underscoring the severity of illness and limiting downstream recovery opportunities.
  • Temperature separation was achieved (mean 12–24 hour core temperature 33.5 ± 1.1°C vs 37.0 ± 0.7°C), but early discontinuation was more common in the hypothermia group (12.7% vs 3.0%).
  • Subsequent evidence and guideline shifts have generally prioritised fever prevention and careful temperature control over routine induction of 33°C hypothermia across unselected cardiac arrest populations.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • Favourable neurological outcome (CPC 1–2) was rare in both arms, making absolute treatment effects highly sensitive to small shifts in outcomes and post-arrest care practices.
  • The control strategy included active normothermia with fever prevention; interpretation should avoid conflating this with “no temperature management”.

Overall Takeaway

HYPERION is the landmark dedicated RCT in non-shockable cardiac arrest demonstrating a small, statistically borderline improvement in day-90 favourable neurological outcome with 33°C hypothermia compared with actively maintained 37°C normothermia, without a mortality benefit. In the post–TTM2 era, it remains an important, hypothesis-shaping signal for selected non-shockable cohorts, but the totality of evidence and guideline evolution increasingly favour meticulous fever prevention and structured post–cardiac arrest care over routine deep hypothermia.

Overall Summary

  • 33°C for 24 hours increased CPC 1–2 at 90 days (10.2% vs 5.7%) but did not reduce mortality (~82% at 90 days).
  • Temperature separation was clear (33.5 ± 1.1°C vs 37.0 ± 0.7°C at 12–24 hours), yet early discontinuation was more frequent with hypothermia.
  • Subsequent trials/meta-analyses and modern guidelines emphasise fever prevention and caution against routine hypothermia for all-comers.

Bibliography