Skip to main content

Foundational Trials

TOMAHAWK

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Angiography after Out-of-Hospital Cardiac Arrest without ST-Segment Elevation
  • Acronym: TOMAHAWK
  • Year: 2021
  • Journal published in: The New England Journal of Medicine
  • Citation: Desch S, Freund A, Akin I, Behnes M, Preusch MR, Zelniker TA, et al. Angiography after out-of-hospital cardiac arrest without ST-segment elevation. N Engl J Med. 2021;385:2544-2553.

Context & Rationale

  • Background
    • Acute coronary occlusion can precipitate out-of-hospital cardiac arrest (OHCA), yet post-ROSC ECG frequently lacks ST-segment elevation, and the prevalence of a treatable culprit lesion in this phenotype is variable.
    • Observational registries (including PROCAT II) suggested potential associations between early angiography/PCI and improved outcomes, but these data are vulnerable to confounding by indication, survivorship (immortal-time) bias, and selection effects in cath lab activation.1
    • The COACT randomised trial in comatose OHCA with shockable rhythms and without ST-segment elevation showed no survival advantage for immediate over delayed angiography, leaving uncertainty for broader cohorts (including nonshockable rhythms and more heterogeneous causes).2
  • Research Question/Hypothesis
    • A strategy of immediate coronary angiography (with revascularisation as indicated) would reduce 30-day all-cause mortality versus an initial intensive care assessment with delayed or selective angiography, among resuscitated OHCA patients without ST-segment elevation.
    • The trial was designed as a pragmatic “strategy” comparison rather than a test of PCI for angiographically defined culprit lesions.3
  • Why This Matters
    • OHCA is common, resource-intensive, and dominated by competing early risks (recurrent arrest, shock, hypoxic-ischaemic brain injury), so unnecessary emergency cath lab activation may divert focus from core post-resuscitation bundles.
    • Determining whether “routine immediate cath” is beneficial, neutral, or harmful directly informs ED/ICU-to-cath triage pathways and guideline recommendations for a large patient population without STEMI.
    • Clarifies whether prior observational signals represent causal benefit or bias, and whether early invasive procedures introduce iatrogenic delay or harm.

Design & Methods

  • Research Question: In resuscitated OHCA without ST-segment elevation, does an immediate coronary angiography strategy reduce 30-day all-cause mortality compared with an initial intensive care assessment with delayed or selective angiography?
  • Study Type: Investigator-initiated, prospective, randomised, multicentre, open-label, parallel-group strategy trial; 31 enrolling sites (Germany, plus one site in Denmark); post-resuscitation management in ICU settings (cardiac ICU and general medical ICU).
  • Population:
    • Inclusion: Adults aged ≥30 years; documented resuscitated OHCA of possible coronary origin; successful return of spontaneous circulation; no ST-segment elevation on post-resuscitation ECG.
    • Key exclusions: ST-segment elevation or new left bundle branch block; in-hospital cardiac arrest; obvious extracardiac cause (e.g., trauma, severe intracranial haemorrhage, intoxication, suffocation/drowning); persistent haemodynamic or electrical instability requiring immediate angiography (e.g., cardiogenic shock or life-threatening arrhythmias felt to be due to acute ischaemia); pregnancy; enrolment in another interfering interventional trial.
  • Intervention:
    • Immediate coronary angiography with revascularisation (PCI/CABG) as indicated by angiographic findings.
    • Operational intent: minimise delay by directing patients to the cath lab early after admission; protocol target was coronary visualisation within 60 minutes of presentation (real-world timing reported in the trial).
  • Comparison:
    • Initial intensive care assessment with delayed or selective angiography.
    • Angiography could be performed after ≥24 hours, or earlier if pre-specified “urgent” criteria emerged (e.g., cardiogenic shock, refractory electrical instability, evolving ischaemia, or new ST-segment elevation).
  • Blinding: Open-label treatment assignment; outcomes other than death were adjudicated by a clinical events committee (death is objective).
  • Statistics: A total of 558 patients were planned to detect a hazard ratio of 0.674 (30-day mortality 46% to 34%) with 90% power at a two-sided 5% significance level (allowing 5% dropout); group-sequential design with one interim analysis after 109 deaths; primary analysis used a (modified) intention-to-treat approach with Cox regression stratified by trial site; prespecified sensitivity analyses included per-protocol and as-treated analyses.
  • Follow-Up Period: 30 days for the primary endpoint; protocol planned additional follow-up to 12 months.

Key Results

This trial was not stopped early. Recruitment reached 554 participants (of 558 planned), and the primary analysis was performed after completion of 30-day follow-up in the analysed cohort.

Outcome Immediate angiography Delayed/selective angiography Effect p value / 95% CI Notes
All-cause mortality at 30 days (primary) 143/265 (54.0%) 122/265 (46.0%) HR 1.28 95% CI 1.00 to 1.63; P=0.06 Primary analysis population excluded 20 consent-withdrawals requiring deletion and 4 major inclusion/exclusion violations.
Composite: death or severe neurologic deficit at 30 days 164/255 (64.3%) 138/248 (55.6%) RR 1.16 95% CI 1.00 to 1.34; P=Not reported Denominators differ due to missing neurological status data; severe neurologic deficit defined by CPC 3–5.
Severe neurologic deficit at 30 days (among survivors with available CPC) 21/112 (18.8%) 16/126 (12.7%) RR 1.48 95% CI 0.82 to 2.67; P=Not reported Neurological outcome ascertainment incomplete; interpret with the composite outcome.
Myocardial infarction (type 1–3) at 30 days 0/248 (0.0%) 2/250 (0.8%) RR 0 95% CI 0 to 1.93; P=Not reported Very low event rates; one group had zero events.
Moderate or severe bleeding (BARC 2–5) 12/260 (4.6%) 8/232 (3.4%) RR 1.34 95% CI 0.57 to 3.14; P=Not reported Safety population (as treated).
Stroke at 30 days 4/258 (1.6%) 5/242 (2.1%) RR 1.13 95% CI 0.33 to 3.84; P=Not reported Safety population (as treated).
Acute renal failure requiring renal-replacement therapy 49/259 (18.9%) 38/241 (15.8%) RR 1.14 95% CI 0.78 to 1.68; P=Not reported Safety population (as treated).
Length of ICU stay Median 7 d (IQR 3–11) Median 8 d (IQR 4–13) HLE −1 d 95% CI −2 to 0; P=Not reported Hodges–Lehmann estimate for between-group difference in medians.
  • There was no evidence of benefit for routine immediate angiography; the point estimate was compatible with modest harm (HR 1.28; 95% CI 1.00 to 1.63; P=0.06).
  • The composite of death or severe neurologic deficit was more frequent with immediate angiography (RR 1.16; 95% CI 1.00 to 1.34).
  • Exploratory subgroup signal: nonshockable initial rhythm HR 1.64 (95% CI 1.06 to 2.54) vs shockable rhythm HR 1.14 (95% CI 0.84 to 1.53); interaction P=0.15.

Internal Validity

  • Randomisation and Allocation:
    • Central, web-based block randomisation (secuTrial) with randomly changing block sizes; stratified by trial site.
    • Allocation concealment was maintained until randomisation was completed, minimising selection bias.
  • Dropout or Exclusions (post-randomisation):
    • Randomised: 554 patients.
    • Primary analysis population: 530 patients (265 per group).
    • Excluded from final analysis: 24 patients total — 20 withdrew informed consent and demanded complete deletion (14 immediate; 6 delayed) and 4 had major inclusion/exclusion violations (2 in-hospital arrest in immediate; 2 STEMI in delayed).
    • Impact: creates a modified intention-to-treat cohort; the absolute number is small, but post-randomisation exclusion can introduce bias if withdrawal correlates with early prognosis or care burden.
  • Performance/Detection Bias:
    • Open-label design could influence co-interventions (e.g., timing of temperature management) and clinician behaviour, including thresholds for neurological prognostication and treatment limitation.
    • Primary outcome (death) is objective; other outcomes were adjudicated by a clinical events committee.
  • Protocol Adherence and Separation of the Variable of Interest:
    • Coronary angiography performed: 253/265 (95.5%) vs 165/265 (62.2%); RR 1.54; 95% CI 1.41 to 1.67.
    • Time from admission to angiography: median 2.9 h (IQR 2.2–3.9) vs 46.9 h (IQR 26.1–116.6).
    • PCI: 60/265 (22.6%) vs 50/265 (18.9%); RR 1.20; 95% CI 0.86 to 1.67.
    • Thrombectomy: 22/265 (8.3%) vs 10/265 (3.8%); RR 2.20; 95% CI 1.06 to 4.57.
    • Targeted temperature management used: 209/265 (78.9%) vs 218/265 (82.3%); RR 0.96; 95% CI 0.88 to 1.05.
    • Time from admission to initiation of temperature management: median 153 min (IQR 110–221) vs 119 min (IQR 80–165); HLE 33 min; 95% CI 21 to 50.
  • Baseline Characteristics:
    • Median age: 69 (IQR 59–78) vs 71 (IQR 60–79); female sex 30.2% vs 29.1%.
    • Shockable initial rhythm: 53.3% vs 58.7%; median time from collapse to ROSC: 15 min (IQR 10–20) vs 15 min (IQR 8–20).
    • Markers of initial severity were similar (e.g., lactate 4.6 vs 4.4 mmol/L; pH 7.26 vs 7.28).
    • There was a modest prognostic imbalance in initial rhythm distribution (more shockable rhythms in the delayed/selective arm), which could favour the control arm.
  • Heterogeneity:
    • Broad inclusion (shockable and nonshockable rhythms; variable arrest characteristics) increases relevance but likely increases clinical heterogeneity of underlying aetiologies and treatment responsiveness.
  • Timing and Dose:
    • The protocol aspired to coronary visualisation within 60 minutes; actual median time was 2.9 hours in the immediate arm, reflecting pragmatic logistics.
    • The delayed/selective arm still underwent angiography in 62.2%, potentially attenuating strategy separation for patients ultimately selected for cath.
  • Crossover:
    • Safety (as-treated) population: 7 crossed from immediate to delayed/selective; 22 crossed from delayed/selective to immediate.
    • Crossover direction predominantly increased early angiography in the delayed arm, which would tend to bias toward no difference for any benefit of immediate angiography.
  • Outcome Assessment:
    • Death is objective; neurological outcomes (CPC) are clinically meaningful but depend on follow-up completeness and may be influenced by care limitation practices in open-label contexts.
  • Statistical Rigor:
    • Pre-specified group-sequential design with an interim analysis; primary Cox model stratified by site; sensitivity analyses (per-protocol and as-treated) were consistent with the main direction of effect (HR >1) and did not reverse conclusions.

Conclusion on Internal Validity: Overall, internal validity appears moderate-to-strong given robust central randomisation, meaningful separation in timing of angiography, and an objective primary endpoint; limitations include an open-label design and a modified intention-to-treat analysis due to post-randomisation consent withdrawals and eligibility violations.

External Validity

  • Population Representativeness:
    • Includes both shockable and nonshockable rhythms (shockable 53–59%), mirroring real-world OHCA case-mix more closely than shockable-only cohorts.
    • Excludes patients with ST-segment elevation/new LBBB and those with haemodynamic or electrical instability requiring immediate angiography, so results apply to “clinically stable” OHCA without STEMI where the decision is discretionary.
  • Applicability:
    • Conducted in European centres with access to coronary angiography and ICU-based post-arrest care; likely generalisable to similar systems with 24/7 cath capability and structured post-ROSC bundles.
    • In systems with longer cath-lab delays or constrained resources, routine immediate angiography is even less likely to confer benefit and may delay critical care priorities.
    • Findings should not be extrapolated to patients with strong clinical evidence of an ongoing coronary culprit (e.g., evolving ischaemia with shock/refractory arrhythmias), who were deliberately excluded.

Conclusion on External Validity: Generalisability is good for comatose OHCA without ST-elevation in whom immediate cath is optional and resources resemble European tertiary centres, but limited for unstable patients requiring emergent angiography and for settings without rapid integrated cath/ICU pathways.

Strengths & Limitations

  • Strengths:
    • Clinically important, pragmatic strategy question in a high-stakes population.
    • Multicentre design across diverse ICU types; broad OHCA inclusion including nonshockable rhythms.
    • Meaningful separation in angiography timing (median 2.9 h vs 46.9 h) and high procedural adherence in the immediate arm.
    • Pre-specified statistical plan including interim analysis and multiple sensitivity analyses.
  • Limitations:
    • Open-label design with potential performance bias (co-interventions, prognostication and treatment-limitation decisions).
    • Modified ITT due to post-randomisation consent withdrawals requiring deletion and eligibility violations.
    • High angiography rate in the delayed/selective arm (62.2%) potentially reduced the ability to detect a benefit confined to patients ultimately selected for angiography.
    • Power calculation assumed a large mortality reduction; smaller benefits (or harms) are more difficult to confirm with precision.
    • Not applicable to patients with cardiogenic shock/electrical instability mandating immediate angiography.

Interpretation & Why It Matters

  • Clinical implication
    For resuscitated OHCA without ST-segment elevation and without an indication for emergent cath, a delayed/selective angiography strategy is at least as effective as routine immediate angiography, with a signal compatible with harm for the routine immediate strategy.
  • System implication
    The default pathway can shift from “cath first” to “ICU stabilisation and triage”, preserving cath lab resources for patients most likely to benefit while prioritising post-resuscitation bundles.
  • Mechanistic signal
    Immediate angiography achieved earlier coronary visualisation but was associated with later initiation of targeted temperature management (median 153 vs 119 minutes), a plausible pathway for iatrogenic delay in time-critical critical care interventions.
  • Methodological point
    Strategy trials in post-arrest care must anticipate consent-withdrawal–driven data deletion and incorporate robust sensitivity analyses for missingness and competing risks; TOMAHAWK explicitly addressed these challenges in its statistical plan.

Controversies & Subsequent Evidence

  • Interpreting a “near-significant” harm signal:
    • The primary endpoint did not meet conventional statistical significance (P=0.06), yet the hazard ratio and confidence interval are compatible with clinically meaningful harm; interpretation depends on views about dichotomising results by an arbitrary threshold.
    • The composite endpoint (death or severe neurologic deficit) crossed the null at its lower confidence limit (RR 1.16; 95% CI 1.00 to 1.34), adding to concern that immediate angiography may not be benign in this population.
  • Control arm “contamination” versus pragmatic design:
    • Delayed/selective angiography was common (62.2%), meaning the comparison was not “cath versus no cath” but “routine immediate” versus “triage-driven use and timing”, which likely reflects real-world best practice and limits any incremental benefit of universal early cath.
  • Applicability to unstable phenotypes:
    • Patients with haemodynamic/electrical instability requiring immediate angiography were excluded by design; results should not be applied to those with cardiogenic shock, refractory ventricular arrhythmias, or other high-suspicion coronary presentations.
  • Consistency with other randomised evidence:
    • Two pilot RCTs in similar OHCA without ST-segment elevation (DISCO and PEARL) were underpowered but did not demonstrate a convincing clinical benefit for routine immediate angiography, aligning directionally with the neutral-to-harm signal seen in TOMAHAWK.45
  • Meta-analyses:
    • Contemporary syntheses of randomised trials, including systematic reviews/meta-analyses and network meta-analyses, report no improvement in survival or favourable neurological outcome with routine immediate angiography compared with delayed/selective strategies, reinforcing a “selective cath” approach in the absence of STEMI or instability.67
  • Consensus statements and evolving guidance:
    • Expert consensus statements increasingly recommend immediate angiography primarily for patients with ST-segment elevation or clear clinical evidence of ongoing ischaemia/instability, and otherwise support delayed/selective angiography integrated within post-resuscitation critical care pathways.8

Summary

  • In 554 resuscitated OHCA patients without ST-segment elevation, routine immediate angiography did not reduce 30-day mortality compared with delayed/selective angiography (54.0% vs 46.0%; HR 1.28; 95% CI 1.00 to 1.63; P=0.06).
  • The composite of death or severe neurologic deficit occurred more often with immediate angiography (64.3% vs 55.6%; RR 1.16; 95% CI 1.00 to 1.34).
  • Strategy separation was substantial (angiography 95.5% vs 62.2%; median time to angiography 2.9 h vs 46.9 h), supporting that the null result is unlikely to be due to complete overlap of care pathways.
  • Immediate angiography was associated with later initiation of temperature management (median 153 vs 119 minutes), a plausible competing-care delay.
  • Findings apply to OHCA without STEMI and without haemodynamic/electrical instability mandating emergent cath; they support a triage-driven (delayed/selective) invasive approach.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Overall Takeaway

TOMAHAWK is a landmark post-cardiac arrest strategy trial demonstrating that routine immediate coronary angiography does not improve 30-day survival in OHCA patients without ST-segment elevation and may be associated with worse clinical outcomes. It supports a modern triage paradigm: prioritise post-resuscitation critical care and pursue angiography selectively, reserving emergent cath for patients with clear clinical evidence of an acute coronary culprit or instability.

Overall Summary

  • Routine immediate coronary angiography after OHCA without ST-segment elevation did not reduce 30-day mortality and showed a signal compatible with harm.
  • Strategy separation was clear (median 2.9 h vs 46.9 h to angiography), supporting the credibility of the null/harm finding.
  • Findings reinforce a selective angiography approach integrated with post-resuscitation critical care priorities.

Bibliography