Chapter 20 — ECPR Patient Selection

Every selection criterion weighed one at a time. Dominant evidence base is SAVE-J II (2,157 ECPR patients, 36 Japanese centres) and its secondary analyses, plus a 2025 systematic review of 4,353 patients — all observational, all from patients already selected for ECPR. Two controversies: (1) the ELSO example selection criteria discriminated at AUC 0.63 (0.54–0.72) against a multivariable score's 0.79 in the only head-to-head test — hold thresholds loosely, build in a senior override; (2) prognosis versus treatment effect — a causal-forest analysis found LOWER pH and HIGHER lactate identified the patients ECPR helped most, inverting the metabolic criteria, and giving Chapter 11's 'prognosis is not indication' a quantitative counterpart. Central new findings: signs of life are the strongest pre-cannulation variable (38.0% vs 8.1%; NPV 0.96) and a non-shockable patient WITH them outperformed a shockable patient WITHOUT them (38.7% vs 4.1%); the arrival rhythm beats the initial rhythm (VF that is asystole on arrival, 2.7%, aOR 0.08); comorbidity exclusions moved the yield only 21.5% to 22.2%; acute aortic disease and primary cerebral disease had NO survivors and appear in no criteria list; hypothermic arrest is a different disease. Three numerical problems flagged and none corrected (two odds ratios computed across a zero cell; a non-monotonic centre-volume gradient with an implausible OR of 0.983 at P=0.006). Seventh bibliographic-metadata problem logged (Linz/Lunz spelling). All rows Verified = No — abstracts only. An addendum was appended to Chapter 19 §19.9 correcting the volume threshold.