Bayesian Reanalyses of the Trials TOMAHAWK and COACT.

BACKGROUND

The timing of coronary angiography in patients with successfully resuscitated out-of-hospital cardiac arrest and missing ST-segment elevations on the electrocardiogram has been investigated in 2 large randomized controlled trials, TOMAHAWK (Angiography After Out-of-Hospital Cardiac Arrest Without ST-Segment Elevation) and COACT (Coronary Angiography After Cardiac Arrest Trial). Both trials found neutral results for immediate vs delayed/selective coronary angiography on short-term all-cause mortality. The TOMAHAWK trial showed a tendency towards harm with immediate coronary angiography, though not statistically significant with traditional frequentist methods. Probabilistic analyses of both trials may enable greater clinical understanding of the trial findings.

OBJECTIVES

The purpose of this study was to perform reanalyses of both trials within a Bayesian framework.

METHODS

Post hoc analyses of both multicenter randomized controlled trials were performed in both cohorts separately and combined. The primary endpoint, 30-day all-cause mortality, was analyzed using Bayesian logistic regression. A spectrum of priors included "flat," "neutral," "optimistic," and "pessimistic" priors based on assumptions made when designing both trials.

RESULTS

In the TOMAHAWK trial, immediate coronary angiography showed a very high posterior probability of increased mortality between 90% and 97% across all priors. The ORs across all priors were directed towards harm. Similarly, COACT showed odds ratios ranging from 0.98 to 1.11 for the 30-day mortality endpoint. When combining both trials, immediate coronary angiography showed a high probability of increased mortality between 83% and 95%, again with ORs across all priors indicating a direction towards harm.

CONCLUSIONS

Bayesian reanalyses showed a very high probability of increased 30-day mortality risk with immediate compared with delayed/selective coronary angiography in the TOMAHAWK trial and combined trial cohort. These findings may shift the current understanding of both trials from a "neutral" towards a likely "harmful" effect of immediate coronary angiography after successfully resuscitated out-of-hospital cardiac arrest without ST-segment elevations. Therefore, adoption of a delayed strategy of coronary angiography might be preferred in clinical practice until the results of the DISCO (Direct or Subacute Coronary Angiography in Out-of-Hospital Cardiac Arrest) trial become available.