Skip to main content

Foundational Trials

TTM2

Image: Shutterstock / ame435

Publication

  • Title: Hypothermia versus Normothermia after Out-of-Hospital Cardiac Arrest
  • Acronym: TTM2
  • Year: 2021
  • Journal published in: The New England Journal of Medicine
  • Citation: Dankiewicz J, Cronberg T, Lilja G, et al. Hypothermia versus Normothermia after Out-of-Hospital Cardiac Arrest. N Engl J Med. 2021;384(24):2283-2294.

Context & Rationale

  • Background
    • Two early, small RCTs (2002) reported improved outcomes with induced hypothermia (32–34°C) after witnessed out-of-hospital cardiac arrest (OHCA) with shockable rhythms, and this drove broad international adoption of temperature management.
    • Subsequent practice became heterogeneous (different targets, durations, devices, and fever prevention strategies), complicating inference about “hypothermia” as a discrete intervention.
    • The 2013 TTM (33°C vs 36°C) trial found no clear difference, strengthening the hypothesis that benefit may derive primarily from fever avoidance rather than induced hypothermia per se.
    • Post-cardiac arrest ICU care (coronary angiography/PCI, haemodynamic optimisation, ventilation practices, and protocolised neuroprognostication) evolved substantially, potentially diluting or modifying any hypothermia signal.
  • Research Question/Hypothesis
    • Whether active induced hypothermia to 33°C (with controlled rewarming) improves survival and functional outcome compared with targeted normothermia with early treatment of fever in unconscious adults after OHCA.
  • Why This Matters
    • Cooling to 33°C is resource-intensive (devices, sedation, shivering control) and plausibly harmful (arrhythmias, haemodynamic instability, infection, coagulopathy), so the net clinical value must be demonstrated.
    • Temperature control sits at the centre of post-resuscitation bundles worldwide; clarifying the incremental value of 33°C over active fever prevention has direct protocol and guideline implications.

Design & Methods

  • Research Question: In unconscious adults after OHCA with sustained ROSC, does targeted hypothermia at 33°C improve 6-month outcomes compared with targeted normothermia with early treatment of fever?
  • Study Type: International, multicentre, investigator-initiated, randomised, parallel-group trial; open-label intervention with blinded structured outcome assessment for neurological/function outcomes; conducted in post-cardiac arrest ICU care pathways.1
  • Population:
    • Setting: Post-resuscitation care after OHCA with admission for intensive care, randomisation within 180 minutes of sustained ROSC.1
    • Key inclusion criteria: Adults (≥18 years); OHCA of presumed cardiac or unknown cause; sustained ROSC (≥20 minutes without chest compressions); unconsciousness defined as inability to obey verbal commands (FOUR motor score <4) after sustained ROSC; eligible for ICU care without limitations; inclusion within 180 minutes of ROSC.1
    • Key exclusion criteria: Unwitnessed arrest with initial asystole; admission temperature <30°C; ECMO prior to ROSC; suspected/obvious pregnancy; intracranial bleeding; severe COPD requiring long-term home oxygen therapy.1
    • Screening yield (trial conduct): 4355 screened; 1900 randomised; common reasons for non-randomisation included time >180 minutes from ROSC (n=794), non-cardiac cause (n=441), not unconscious (n=248), and care limitations (n=237).2
  • Intervention:
    • Temperature target: 33°C.
    • Phase structure: Phase 2 (0–40h) with mandatory sedation; Block A (28h) rapid cooling and maintenance at 33°C; Block B (12h) controlled rewarming at ~0.33°C/hour to normothermia.1
    • Cooling methods: Closed-loop intravascular or surface devices recommended; initial adjunct cooling with 4°C crystalloid permitted (max 30 mL/kg); pragmatic allowance for other approved devices and adjuncts (e.g., ice packs), with a requirement that a device be used for temperature control.1
    • Post-rewarming fever prevention: If still comatose, maintain 36.5–37.7°C until 72h post-randomisation; avoid active warming.1
    • Neuroprognostication: Protocolised delayed prognostication at 96 hours or later, using a conservative multimodal approach aligned with ERC recommendations (clinical exam and ancillary testing).1
  • Comparison:
    • Temperature strategy: Targeted normothermia with early treatment of fever.
    • Target and threshold: Device target 37.5°C; conservative measures (including antipyretics) permitted; activation of a temperature management device if a recorded core temperature ≥37.8°C during the intervention period (first 40h).1
    • Equating co-interventions: Same phase structure and mandatory sedation framework (0–40h), followed by tapering/discontinuation of sedation (40–96h) and continuation of normothermic range (36.5–37.7°C) until 72h if still comatose; if a device was used for fever treatment, continue fever control for an additional 32 hours.1
  • Blinding: The temperature intervention was not blinded to treating clinicians (inherently difficult); structured 6-month functional assessment was performed by assessors blinded to allocation; mortality is an objective endpoint.1
  • Statistics: Power calculation: 1862 patients required to detect a 7.5% absolute reduction in 180-day mortality (from 55% to 47.5%) with 90% power (beta 0.10) at a two-sided 5% significance level; planned enrolment 1900. Primary analysis was intention-to-treat (with post-randomisation consent-related exclusions described below), using generalised linear mixed modelling adjusted for stratification variables (site and co-enrolment in the TAME trial), with risk ratios reported; survival analysed using Cox regression.1
  • Follow-Up Period: 6 months for primary/secondary outcomes (with longer-term follow-up planned separately).1

Key Results

This trial was not stopped early. Enrolment reached the planned sample size; outcomes were assessed to 6 months.

Outcome Hypothermia (33°C) Normothermia (≤37.8°C) Effect p value / 95% CI Notes
Death at 6 months (primary outcome) 465/925 (50%) 446/925 (48%) RR 1.04 95% CI 0.94 to 1.14; P=0.37 Primary outcome data missing for 11/1861 patients
Modified Rankin Scale 4–6 at 6 months 488/881 (55%) 479/866 (55%) RR 1.00 95% CI 0.92 to 1.09 Structured mRS assessment missing in 82/1747 assessed patients (unstructured/insufficient assessment)
Poor functional outcome at 6 months (trial-defined) 495/918 (54%) 493/911 (54%) RR 1.00 95% CI 0.91 to 1.08 Functional outcome unavailable in 21 patients (7 vs 14)
Time-to-death Not reported Not reported HR 1.08 95% CI 0.95 to 1.23 Survival analysis reported as hazard ratio
Arrhythmia causing haemodynamic compromise 222/927 (24%) 152/921 (17%) RR 1.45 95% CI 1.21 to 1.75; P<0.001 Key safety signal favouring normothermia
Pneumonia 330/927 (36%) 322/921 (35%) RR 1.02 95% CI 0.89 to 1.17; P=0.75 No clear between-group difference
Sepsis 99/926 (11%) 83/922 (9%) RR 1.18 95% CI 0.89 to 1.57; P=0.25 Directionally higher with hypothermia; imprecise
Bleeding 44/927 (5%) 46/922 (5%) RR 0.95 95% CI 0.64 to 1.41; P=0.79 No signal of excess bleeding
Skin complications related to device 10/927 (1%) 5/922 (1%) RR 1.99 95% CI 0.69 to 5.74; P=0.20 Rare; estimates imprecise
  • Induced hypothermia to 33°C did not reduce mortality at 6 months versus targeted normothermia with active fever treatment (RR 1.04; 95% CI 0.94 to 1.14; P=0.37).
  • Functional outcome at 6 months was similar between groups (mRS 4–6: 55% vs 55%; RR 1.00; 95% CI 0.92 to 1.09).
  • Hypothermia increased clinically important arrhythmias with haemodynamic compromise (24% vs 17%; RR 1.45; 95% CI 1.21 to 1.75; P<0.001).

Internal Validity

  • Randomisation and allocation
    • Randomised in a 1:1 ratio to hypothermia vs normothermia, with stratification by site and co-enrolment in the TAME trial.1
    • Allocation concealment was maintained through central trial processes prior to assignment; selection bias risk is low given narrow enrolment window (≤180 minutes) and objective eligibility thresholds.1
  • Dropout / exclusions (post-randomisation)
    • Of 1900 randomised, 39 were excluded from the intention-to-treat population due to consent not obtained/withdrawn (n=37) or double randomisation (n=2), leaving 1861 analysed in the trial intention-to-treat population.2
    • Primary outcome (mortality) data were missing for 11/1861 patients; missingness was small but not zero, and post-randomisation exclusions introduce potential (albeit likely limited) bias.
  • Performance and detection bias
    • Treating teams were unblinded, so co-interventions could theoretically diverge; however, both groups used the same phase structure with mandatory sedation in the intervention period, which reduces differential co-intervention risk.1
    • Outcome assessment for neurological/functional endpoints used blinded structured interviews; mortality is objective.
  • Protocol adherence and separation of the variable of interest
    • Median time from randomisation to reaching 34°C in the hypothermia group was 3 hours (IQR 2 to 4).2
    • Neuromuscular blockade within 72 hours was used more frequently with hypothermia (586/892; 66%) than normothermia (395/879; 45%), consistent with greater shivering management requirements.2
    • Antipyretic exposure differed: acetaminophen/paracetamol use within 72 hours was 515/892 (58%) with hypothermia vs 621/879 (71%) with normothermia, reflecting the fever-treatment comparator design.2
  • Baseline characteristics and clinical severity
    • Groups were broadly similar at baseline (age 64±13 vs 63±14 years; male 80% vs 79%).
    • Cardiac arrest characteristics were comparable (bystander witnessed 91% vs 92%; bystander CPR 82% vs 78%; median time to sustained ROSC 25 minutes in both groups; median time to randomisation 136 vs 133 minutes).
    • Admission shock was present in 261/930 (28%) vs 275/931 (30%), and lactate was 5.9±4.4 vs 5.8±4.2 mmol/L, indicating substantial but not extreme physiologic derangement.
  • Timing and dose
    • The intervention was delivered early (≤180 minutes from ROSC by design), with a pragmatic cooling approach and a prespecified expectation to reach target rapidly (within ~90 minutes in most participants) in the hypothermia arm.1
    • The hypothermia “dose” was 33°C for 28 hours followed by controlled rewarming over 12 hours (total phase 2 length 40 hours); this is a clinically meaningful dose aligned with contemporary practice patterns.1
  • Adjunctive therapy and downstream care signals
    • Time to extubation or death from randomisation was slightly longer in the hypothermia group (median 3.7 days; IQR 1.4 to 7.7) than the normothermia group (median 3.2 days; IQR 1.2 to 6.6), consistent with deeper sedation/shivering control requirements.2
    • ICU length of stay was similar (median 4.0 vs 3.6 days) and hospital length of stay was similar (median 6.9 vs 6.5 days), suggesting limited downstream distortion of major utilisation endpoints.2
  • Outcome assessment completeness
    • Functional outcome was assessed at 6 months in 1747 patients; structured mRS data were available in 1747 (881 vs 866), with some missing or non-structured assessments contributing to differing denominators.
  • Statistical rigour
    • Sample size targets were achieved; analysis was aligned with an a priori plan using risk ratios for binary endpoints and pre-specified subgroup analyses.
    • Protocol documents indicate planned adjustment for stratification variables (site and TAME co-enrolment) and conservative approach to multiple secondary outcomes (primary conclusion based on the primary outcome).1

Conclusion on Internal Validity: Overall, internal validity appears moderate-to-strong given early randomisation, prespecified co-intervention structure, protocolised prognostication, and near-complete primary outcome follow-up; the major threats are the unavoidable open-label design and post-randomisation exclusions due to consent/double randomisation.2

External Validity

  • Population representativeness
    • Typical ICU-managed OHCA cohort in high-functioning emergency response systems (high bystander-witnessed rate 91–92% and high bystander CPR 78–82%).
    • Substantial proportion of shockable rhythms (72–75% shockable first rhythm), so inference is most robust for this phenotype, though nonshockable rhythms were included (25–28%).
    • Exclusions (unwitnessed asystole, severe COPD on home oxygen, ECMO prior to ROSC, intracranial bleeding, pregnancy, temperature <30°C) limit applicability to those subgroups.
  • Applicability across systems
    • Both strategies required structured temperature monitoring, the capacity to deploy devices, and prolonged sedation pathways, which may not translate directly to low-resource settings.
    • Findings apply specifically to “33°C versus active fever prevention/early treatment” rather than “cooling versus no temperature management”.

Conclusion on External Validity: Generalisability is good for adult OHCA patients admitted unconscious to ICUs capable of active fever prevention; it is more limited for in-hospital arrests, patients with unwitnessed asystole, ECMO populations, and settings without device-based temperature control infrastructure.

Strengths & Limitations

  • Strengths:
    • Large pragmatic international trial with contemporary post-cardiac arrest ICU care.
    • Active comparator (targeted normothermia with fever treatment) isolates the incremental effect of 33°C over and above fever prevention.
    • Prespecified intervention phasing (40h) and delayed neuroprognostication at ≥96h to mitigate self-fulfilling prophecy.
    • Clinically meaningful, patient-centred outcomes at 6 months (mortality and disability).
    • Low missingness for primary outcome with transparent accounting for post-randomisation exclusions.
  • Limitations:
    • Open-label delivery introduces potential performance bias, including differential sedation/shivering management and clinician perceptions.
    • Post-randomisation exclusions (consent not obtained/withdrawn; double randomisation) mean the analysis was not a “pure” full ITT of all randomised patients.
    • Comparator was not “no cooling”: it embedded active fever prevention and device-activated cooling above 37.8°C, narrowing contrast to “33°C vs fever control”.
    • Population skewed towards witnessed arrests and shockable rhythms in high-resource systems, which may not represent lower-survival OHCA systems or different case mixes.
    • Only one hypothermia target (33°C) and one duration/rewarming profile were tested; results do not directly address alternative targets/doses.

Interpretation & Why It Matters

  • Clinical practice
    • Routine induced hypothermia to 33°C cannot be justified on outcome benefit grounds when compared against an active normothermia/fever-treatment strategy in similar OHCA populations.
    • Active fever prevention remains a core component of post-resuscitation care, but the default “target 33°C” paradigm is not supported by these data.
  • Mechanistic inference
    • The neutral result strengthens the view that much of the historical signal attributed to hypothermia may have reflected avoidance of fever and standardisation of post-arrest care rather than a robust temperature-dependent neuroprotective effect.
  • Risk–benefit
    • Harms (notably arrhythmia with haemodynamic compromise: 24% vs 17%) shift the balance away from 33°C as a default strategy when no compensatory clinical benefit is demonstrated.

Controversies & Subsequent Evidence

  • Comparator interpretation
    • The trial tested 33°C versus active normothermia with fever treatment, not “cooling versus no temperature management”; this is central to how the result should be applied clinically.3
  • Temperature separation and contamination (design feature, not protocol deviation)
    • Normothermia management included device-triggered cooling at ≥37.8°C, which reduces between-group temperature contrast compared with older “passive control” trials and makes the result more directly relevant to modern fever-avoidance strategies.3
  • Withdrawal of life-sustaining therapy and prognostication bias
    • Despite delayed, protocolised prognostication, withdrawal decisions remain a plausible pathway for bias in post-cardiac arrest trials; the trial’s conservative prognostication timing is a key mitigation feature, but not a complete solution to the self-fulfilling prophecy problem.3
  • Post-randomisation exclusions
    • Excluding 39 randomised patients for consent/double randomisation creates a methodological tension with strict ITT principles, although the magnitude is small relative to total enrolment and unlikely to reverse the primary conclusion.2
  • Clinical interpretation emphasised in high-level critique
    • Arguments centred on generalisability (excluded phenotypes and system factors) and on framing the result as “no incremental benefit of 33°C over fever prevention” rather than “temperature no longer matters”.45

Summary

  • In unconscious adults after OHCA, induced hypothermia to 33°C did not improve 6-month survival compared with targeted normothermia with early fever treatment (50% vs 48%; RR 1.04; 95% CI 0.94 to 1.14; P=0.37).
  • Disability outcomes were similar between strategies (mRS 4–6: 55% vs 55%; RR 1.00; 95% CI 0.92 to 1.09).
  • Hypothermia increased arrhythmia with haemodynamic compromise (24% vs 17%; RR 1.45; 95% CI 1.21 to 1.75; P<0.001).
  • The comparator was an active fever-prevention strategy with device-activated cooling at ≥37.8°C, making conclusions most applicable to modern ICU pathways.
  • Methodological strengths include early enrolment, prespecified temperature phases, and delayed neuroprognostication; limitations include open-label delivery and post-randomisation exclusions.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Overall Takeaway

TTM2 is “landmark” because it directly tested whether cooling unconscious OHCA survivors to 33°C adds clinical benefit over a contemporary, active normothermia/fever-treatment strategy delivered within a structured ICU pathway. The answer was neutral for survival and disability, with a clear signal of increased haemodynamically significant arrhythmias under hypothermia, reshaping modern post-cardiac arrest temperature management away from routine 33°C and towards robust fever prevention.

Overall Summary

  • 33°C hypothermia did not improve 6-month mortality or disability versus targeted normothermia with fever treatment.
  • Hypothermia increased arrhythmia with haemodynamic compromise.
  • The key inference is “no incremental benefit of 33°C over active fever prevention” in similar OHCA ICU populations.

Bibliography

Results Presentation

Critical Care Reviews Livestream, June 16th, 2021

Last updated January 27th, 2026