Prediction of good functional outcome decreases diagnostic uncertainty in unconscious survivors after out-of-hospital cardiac arrest

Purpose: To explore modifications of the 2021 European Resuscitation Council/European Society of Intensive Care Medicine (ERC/ESICM) guideline algorithm for neuroprognostication after cardiac arrest to improve its prognostic accuracy. Methods: Post-hoc analysis of four prospective multicentre studies (TTM, TTM2, KORHN and ProNeCA). We raised the Glasgow Coma Scale motor (GCS-M) inclusion threshold at 72 h after cardiac arrest from the current GCS-M < 4 to GCS-M < 6 (all unconscious patients). Secondly, we included good outcome predictors (GCS-M 4–5, neuron-specific enolase < 17 µg/L, benign electroencephalography patterns ≤ 72 h post-arrest and normal magnetic resonance imaging at 72–168 h post-arrest) in the algorithm. Functional outcome was assessed dichotomously at six months, including modified Rankin Scale 0–3, Cerebral Performance Category 1–2 or Glasgow Outcome Scale 4–5 (no symptoms to moderate disability) as good outcome. Results: We analysed 3,388 patients, of whom 2,079 had GCS-M < 4 at ≥ 72 h. Of the 874 patients identified by the 2021 ERC/ESICM poor outcome criteria, 870 had poor functional outcome (specificity: 99.6% [95%CI 99.0–99.9]). Using the GCS-M < 6 threshold, 366 more patients entered the algorithm (N = 2,445). Seven more patients with poor outcomes were identified, with close to identical specificity. Good outcome predictors thereafter identified 673 patients with potential recovery, of whom 411 (61%) had a good functional outcome at six months. With the updated algorithm, the number of prognosticated patients with an indeterminate prognosis decreased from 1,205/2,079 (58%) to 891/2,445 (36%). Conclusion: Raising the GCS-M inclusion threshold and adding favourable predictors to the 2021 ERC/ESICM prognostication algorithm reduced prognostic uncertainty without increasing falsely pessimistic predictions.