Skip to main content

Foundational Trials

TAME

Image: Shutterstock / ame435

Publication

  • Title: Mild Hypercapnia or Normocapnia after Out-of-Hospital Cardiac Arrest
  • Acronym: TAME
  • Year: 2023
  • Journal published in: The New England Journal of Medicine
  • Citation: Eastwood GM, Schneider AG, Suzuki S, et al. Mild hypercapnia or normocapnia after out-of-hospital cardiac arrest. N Engl J Med. 2023;389(1):45-57.

Context & Rationale

  • Background
    • Death and long-term disability after out-of-hospital cardiac arrest (OHCA) are largely mediated by post-cardiac arrest brain injury, compounded by haemodynamic instability and systemic reperfusion injury.
    • Arterial carbon dioxide tension (PaCO2) is a potent, rapidly modifiable determinant of cerebral blood flow and cerebrovascular tone, and therefore a plausible lever for early neuroprotection in the post-ROSC period.
    • Mild hypercapnia can augment cerebral blood flow via vasodilatation, but could plausibly increase intracranial pressure, worsen acidaemia, and provoke haemodynamic/arrhythmic complications; the net clinical effect is uncertain.
    • Pre-trial observational data linked dyscarbia—particularly hypocapnia—to worse outcomes after cardiac arrest, but these studies were susceptible to confounding by illness severity and treatment intensity.1
    • A preceding phase II randomised trial (CCC) demonstrated feasibility of targeting mild hypercapnia in this population, but it was underpowered to determine patient-centred outcome benefit or harm.2
  • Research Question/Hypothesis
    • Does actively targeting mild hypercapnia (PaCO2 50–55 mm Hg) for 24 hours after randomisation improve 6-month neurological recovery versus targeting normocapnia (PaCO2 35–45 mm Hg) in comatose, mechanically ventilated adult OHCA survivors?
    • Hypothesis: targeted mild hypercapnia would increase the proportion of patients with a favourable 6-month neurological outcome by improving cerebral perfusion in the early post-resuscitation phase.
  • Why This Matters
    • Ventilator targets are ubiquitous, low-cost, and scalable; even small absolute improvements in neurological recovery would be practice-changing at population level.
    • Conversely, if deliberate hypercapnia is ineffective (or harmful), the findings directly inform bedside ventilation targets and strengthen the rationale to focus on avoiding extremes rather than “treating” PaCO2 upward.
    • Methodologically, TAME tests a physiological hypothesis at definitive trial scale, addressing a common translational gap in critical care where mechanistic plausibility does not reliably predict patient-centred benefit.

Design & Methods

  • Research Question: In adult comatose survivors of OHCA, does targeted mild hypercapnia (PaCO2 50–55 mm Hg) for 24 hours compared with targeted normocapnia (PaCO2 35–45 mm Hg) improve favourable neurological outcome at 6 months?
  • Study Type: Investigator-initiated, randomised, multicentre, international, parallel-group clinical trial; open-label intervention delivery with blinded assessment of neurological prognosis and 6-month outcomes; ICU-based post-resuscitation care context.
  • Population:
    • Setting: 63 ICUs across 17 countries; recruitment March 2018 to September 2021.
    • Inclusion criteria: adult (≥18 years); OHCA; sustained ROSC for ≥20 minutes; comatose on admission (FOUR motor score 0–3); mechanically ventilated; randomisation within 180 minutes of ROSC.
    • Key exclusions: unwitnessed asystole; admission tympanic temperature <30°C; extracorporeal membrane oxygenation before ROSC; pregnancy; suspected intracranial bleeding; severe COPD with home oxygen therapy.
  • Intervention:
    • Targeted mild hypercapnia: PaCO2 50–55 mm Hg for 24 hours beginning at randomisation.
    • Implementation: recommended deep sedation (target RASS −4); arterial blood gases at least every 4 hours using alpha-stat approach (no temperature correction); end-tidal CO2 used to guide ventilator adjustments between ABGs.
    • Ventilator, sedation, and neuromuscular blockade were clinician-directed, but within protocol framing to achieve and maintain the PaCO2 target.
  • Comparison:
    • Targeted normocapnia: PaCO2 35–45 mm Hg for 24 hours beginning at randomisation.
    • Implementation mirrored the intervention arm (ABG frequency, alpha-stat approach, ETCO2-guided adjustments), differing only by PaCO2 target range.
  • Blinding: Treating clinicians were unblinded; assessors of neurological prognosis (protocol-guided assessment at ≥96 hours after randomisation) and 6-month outcomes were blinded; statisticians and authors were blinded during analysis via masked-group manuscripts.
  • Statistics: A total of 1624 patients were required to detect an 8% absolute increase in favourable neurological outcome (from 50% to 58%) with 90% power at a two-sided alpha level of 0.05; sample size inflated to 1700 to allow for consent withdrawal and loss to follow-up; primary analyses were intention-to-treat excluding participants who withdrew consent; mixed-effects models accounted for centre and co-enrolment in TTM2; multiple imputation prespecified for missingness; a single interim analysis at 850 enrolments used a symmetric O’Brien–Fleming boundary (two-sided P=0.005) for efficacy or harm.3
  • Follow-Up Period: 6 months (≈180 days) for neurological outcome, quality of life, and cognitive assessments; mortality also assessed through 6 months.

Key Results

This trial was not stopped early. A single planned interim analysis was conducted; enrolment continued to the prespecified sample size.

Outcome Mild hypercapnia (PaCO2 50–55) Normocapnia (PaCO2 35–45) Effect p value / 95% CI Notes
Favourable neurological outcome at 6 months (GOS-E 5–8) 332/764 (43.5%) 350/784 (44.6%) Adjusted RR 0.98 95% CI 0.87 to 1.11; P=0.76 Primary outcome; structured blinded assessment.
Dichotomised favourable neurological outcome at 6 months (using all available data when structured assessment not possible) 348/788 (44.2%) 365/806 (45.3%) Adjusted RR 0.98 95% CI 0.87 to 1.11 Secondary approach intended to reduce missingness.
Poor functional outcome at 6 months (mRS 4–6) 407/762 (53.4%) 400/779 (51.3%) Adjusted RR 1.05 95% CI 0.95 to 1.15 Includes death (mRS 6).
Death within 6 months 393/816 (48.2%) 382/832 (45.9%) Adjusted RR 1.05 95% CI 0.94 to 1.16 No evidence of survival benefit or harm at conventional thresholds.
EQ-VAS at 6 months (survivors only) 76.4 (74.1 to 78.6) 74.5 (72.2 to 76.7) Mean difference 1.9 95% CI −0.9 to 4.7 Higher scores indicate better perceived health.
EQ-VAS at 6 months (all patients; non-survivors imputed 0) 35.8 (32.6 to 39.0) 36.8 (33.6 to 40.0) Mean difference −1.0 95% CI −5.0 to 3.0 Preplanned approach to handle competing risk of death.
Pneumonia (ICU days 1–7) 335/821 (40.8%) 358/831 (43.1%) Not reported P=0.35 Prespecified adverse event definition.
Severe bleeding (GUSTO severe) 16/829 (1.9%) 14/839 (1.7%) Not reported P=0.69 Prespecified adverse event.
Suspected/confirmed raised intracranial pressure necessitating normocapnia (ICU days 1–7) 13/829 (1.6%) 19/839 (2.3%) Not reported P=0.31 Prespecified serious adverse event.
  • Targeted mild hypercapnia did not improve the primary patient-centred endpoint: favourable neurological outcome 43.5% vs 44.6%; adjusted RR 0.98; 95% CI 0.87 to 1.11; P=0.76.
  • Mortality at 6 months was similar: 48.2% vs 45.9%; adjusted RR 1.05; 95% CI 0.94 to 1.16.
  • Sensitivity excluding co-enrolment in TTM2 was concordant: favourable neurological outcome 267/599 (44.6%) vs 275/615 (44.7%); adjusted RR 1.00; 95% CI 0.87 to 1.15.

Internal Validity

  • Randomisation and allocation: Central randomisation with site incorporated in modelling; allocation concealment preserved until assignment; co-enrolment in TTM2 treated as a prespecified covariate in analysis.
  • Drop out or exclusions: 1700 randomised; 24 participants were excluded from intention-to-treat analyses due to withdrawal of consent; primary outcome available for 1548 participants with 6.9% missingness; mortality within 6 months available for 1648 participants with 0.9% missingness.3
  • Performance/detection bias: Intervention delivery was unblinded, creating scope for differential co-interventions; however, neurological prognostication (≥96 hours) and 6-month outcome assessment were blinded, reducing detection bias for subjective endpoints.
  • Protocol adherence: Target PaCO2 was achieved by 764/818 (93.4%) in the mild hypercapnia group and 785/818 (95.9%) in the normocapnia group; median hours in target range were 4 (IQR 2–6) vs 7 (IQR 4–10); protocol deviations included early discontinuation of PaCO2 targeting in 68 (8.2%) vs 25 (3.0%), with pH <7.1 contributing to discontinuation in 12 vs 3 participants.3
  • Baseline characteristics: Groups were broadly comparable and clinically severe: age 61.2±14.3 vs 61.6±13.3 years; shockable rhythm 581/829 (70.1%) vs 608/839 (72.5%); median time arrest-to-ROSC 26 (IQR 17–40) vs 25 (IQR 16–39) minutes; first measured PaCO2 52.8±17.3 vs 52.5±20.3 mm Hg.
  • Heterogeneity: Large, geographically diverse ICU trial; prespecified subgroup analyses were performed (age, sex, time to ROSC, initial rhythm, shock on admission) without a reported qualitative signal of effect modification sufficient to alter interpretation.
  • Timing: Randomisation occurred a median of 154 vs 151 minutes after arrest; intervention window therefore began after early post-ROSC care had already occurred, potentially limiting ability to influence the earliest reperfusion phase (but equally in both groups).
  • Dose: The “dose” of hypercapnia was defined as a narrow PaCO2 target (50–55 mm Hg) maintained for 24 hours; overlap in achieved PaCO2 distributions and limited time in target range (median 4 hours) may have attenuated any treatment effect.
  • Separation of the Variable of Interest: Clear separation in achieved PaCO2 was demonstrated: time-weighted mean PaCO2 49.8±4.7 mm Hg vs 40.8±4.2 mm Hg; minimum PaCO2 39.6±6.7 vs 32.3±6.0; maximum PaCO2 59.8±7.8 vs 50.8±6.8; hypocapnia (<35 mm Hg) occurred in 3.1% vs 15.8% of ABGs and severe hypocapnia (<25 mm Hg) in 0.2% vs 0.5%.3
  • Key delivery aspects: Deep sedation was recommended; neuromuscular blockade was used more often in the mild hypercapnia group (558/824 [67.7%] vs 490/829 [59.1%]) with longer duration (median 140 [IQR 60–330] vs 104 [IQR 60–221] minutes), consistent with higher ventilatory control demands.
  • Adjunctive therapy use: Therapeutic temperature management use was similar (651/829 [78.6%] vs 661/839 [78.8%]); coronary angiography and other ICU therapies were broadly comparable, reducing likelihood that co-interventions explain the null effect.
  • Outcome assessment: The primary endpoint (GOS-E at 6 months) was determined by blinded assessors; when structured GOS-E assessment was not possible, a prespecified binary adjudication using medical/interview records was performed by blinded assessors.
  • Statistical rigour: Trial protocol and statistical analysis plan were finalised prior to completion of enrolment; prespecified mixed-effects modelling incorporated centre and co-enrolment; multiple imputation was used for missing data and sensitivity analyses included ordinal GOS-E modelling.

Conclusion on Internal Validity: Overall, internal validity appears moderate-to-strong: randomisation and blinded outcome assessment were robust, and PaCO2 separation was clear; the main threats are open-label co-interventions, protocol non-adherence/overlap in achieved PaCO2, and clinician-directed withdrawal-of-life-sustaining therapy.

External Validity

  • Population representativeness: The cohort reflects a high-functioning OHCA pathway: predominantly witnessed arrests with high bystander CPR and shockable rhythms; exclusions (e.g., unwitnessed asystole, severe COPD, profound hypothermia, suspected intracranial bleeding) narrow applicability to all-comer OHCA.
  • Applicability: The intervention requires frequent ABG sampling, ETCO2 monitoring, and ventilator adjustments; this is feasible in well-resourced ICUs but may be harder to implement in resource-limited systems or where sedation/paralysis practices differ.
  • Clinical setting transferability: Findings apply most directly to comatose, intubated OHCA survivors admitted to ICU within ~3 hours of arrest; extrapolation to in-hospital cardiac arrest, non-cardiac aetiologies, or patients managed without invasive ventilation is uncertain.

Conclusion on External Validity: Generalisability is moderate to contemporary post-arrest ICUs with similar case-mix and monitoring capacity, but more limited for non-shockable/unwitnessed arrests, severe chronic lung disease, and lower-resource environments.

Strengths & Limitations

  • Strengths: Definitive-scale sample size (n=1700) with pragmatic international ICU recruitment; clinically meaningful primary endpoint at 6 months; blinded neurological prognostication and outcome assessment; prespecified modelling accounting for centre effects and TTM2 co-enrolment; demonstrable physiological separation in PaCO2 exposure.
  • Limitations: Open-label intervention with potential for differential co-interventions (sedation/paralysis in particular); intervention commenced a median ~2.5 hours after arrest with baseline PaCO2 already elevated (~53 mm Hg), potentially limiting biological leverage; overlap in achieved PaCO2 and limited time in target range; withdrawal-of-life-sustaining therapy was clinician-directed (risk of centre-level practice variation); primary outcome missingness 6.9% (mitigated by prespecified imputation and alternative-data adjudication).

Interpretation & Why It Matters

  • Clinical implication
    Routine deliberate targeting of mild hypercapnia (PaCO2 50–55 mm Hg) for 24 hours after OHCA is not supported by patient-centred outcomes; maintaining normocapnia and avoiding hypocapnia remains the defensible default.
  • Mechanistic insight
    The absence of benefit despite clear PaCO2 separation suggests that augmenting cerebral blood flow via mild hypercapnia is insufficient to overcome post-arrest brain injury biology in unselected patients, or that any haemodynamic/cerebral advantages are offset by countervailing physiological costs.
  • Trialist takeaway
    TAME exemplifies the need for large, methodologically rigorous RCTs to adjudicate physiologically attractive interventions; feasibility signals from phase II studies did not translate into improved neurological recovery at scale.

Controversies & Subsequent Evidence

  • Timing and biological leverage: Median time to randomisation was 154–151 minutes, with first measured PaCO2 ~53 mm Hg, implying that an “early hypocapnia avoidance” effect may already have occurred before randomised targeting began; this raises uncertainty about whether a still-earlier intervention window would be required to test the core neuroperfusion hypothesis.
  • Intervention fidelity versus clinical separation: Although time-weighted PaCO2 separation was substantial (49.8±4.7 vs 40.8±4.2 mm Hg), the protocolised target range was achieved for a limited duration (median 4 vs 7 hours), and early discontinuation of targeting was more frequent in the mild hypercapnia group (8.2% vs 3.0%), which could dilute any effect size.
  • Open-label care and withdrawal practices: Clinician discretion for sedation/paralysis and withdrawal-of-life-sustaining therapy introduces potential centre-level practice heterogeneity; protocolised prognostication was planned at ≥96 hours but could not be performed universally, which remains a generic threat to internal validity in post-arrest trials.
  • Subsequent syntheses: Post-TAME systematic reviews/meta-analyses continue to show outcome associations with dyscarbia across critically ill cohorts and after cardiac arrest, but remain constrained by confounding and exposure misclassification typical of observational gas-exposure analyses.45
  • Guideline integration: Contemporary treatment recommendations and post-resuscitation care guidelines incorporate the neutral TAME findings and emphasise maintaining normocapnia and avoiding extremes, rather than routine deliberate hypercapnia in unselected OHCA survivors.67

Summary

  • In 1700 comatose, mechanically ventilated OHCA survivors, targeting mild hypercapnia (PaCO2 50–55 mm Hg) for 24 hours did not improve favourable neurological outcome at 6 months versus targeting normocapnia (PaCO2 35–45 mm Hg).
  • Primary outcome was neutral: 43.5% vs 44.6%; adjusted RR 0.98; 95% CI 0.87 to 1.11; P=0.76.
  • Mortality was similar: 48.2% vs 45.9%; adjusted RR 1.05; 95% CI 0.94 to 1.16.
  • Physiological separation in PaCO2 was clear, but time in the protocol target range was limited and early discontinuation of targeting occurred more often with mild hypercapnia.
  • TAME shifts practice away from routine deliberate post-arrest hypercapnia and supports a pragmatic focus on avoiding hypocapnia and extreme hypercapnia within comprehensive post-cardiac arrest care.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • Only DOI-verifiable (within the available source corpus) post-TAME guideline documents were listed above; additional society statements and guideline updates may exist and should be cross-checked against the latest critical care and resuscitation guideline repositories.

Overall Takeaway

TAME is a landmark definitive trial because it tested a widely discussed physiological neuroprotection strategy—deliberate mild hypercapnia—at international scale with blinded neurological outcome assessment. Despite clear PaCO2 separation, there was no improvement in 6-month neurological recovery or survival, shifting practice away from routine targeted hypercapnia and towards maintaining normocapnia while avoiding hypocapnia and extremes as part of comprehensive post-cardiac arrest care.

Overall Summary

  • Targeted mild hypercapnia (PaCO2 50–55 mm Hg for 24 hours) after OHCA did not improve 6-month neurological outcomes or survival versus targeted normocapnia, despite clear physiological separation.

Bibliography

Results Presentation

Critical Care Reviews Meeting 2023

Last updated January 27th, 2026