23, 24, 74
The circularity problem: the algorithm was validated in cohorts that withdrew treatment on the basis of the tests being validated. Two cohorts that broke that circle (ProNeCA 2019 and Gotze 2026) reproduced the rule but not the certainty. Separately, the PRAGUE-OHCA biomarker sub-analysis concludes that prognostication algorithms should reflect the resuscitation method, which this one does not.
On ECMO the algorithm's trigger time is not a calendar hour: sedation is deeper, doses are higher because of circuit sequestration, and clearance is slower, so the clock starts at demonstrated sedation clearance. Magnetic resonance imaging - the most sensitive item - is unavailable while the patient is cannulated, and the enolase threshold used in the algorithm was not derived in ECMO or ECPR patients. Use the two-or-more rule; do not quote its specificity to a family.
[VERIFICATION REQUIRED - the guideline document itself was NOT retrieved. What follows is the algorithm as described in its validation studies, not as worded in the source.] Prognostication is undertaken in a comatose patient at 72 hours or later, with sedation, neuromuscular blockade, hypothermia and metabolic confounders excluded; prediction of a poor outcome requires two or more concordant unfavourable findings drawn from bilaterally absent pupillary and corneal reflexes, status myoclonus, high neuron-specific enolase, bilaterally absent N20 cortical responses, suppressed or burst-suppression electroencephalographic background, and diffuse anoxic injury on imaging.
Neuroprognostication after cardiac arrest