Taekwon Kim, Sang-Hun Lee, Kyungsub Song, Jonghoon Yoo
Static prognostic scores for extracorporeal cardiopulmonary resuscitation have been developed for pre-cannulation risk estimation and are frequently repurposed for post-cannulation decisions, where early treatment response may be more informative. We evaluated 3 widely used pre extracorporeal membrane oxygenation (ECMO) scores (TiPS65, RESCUE-in-hospital cardiac arrest (IHCA), and Survival After Veno-Arterial ECMO) for in-hospital mortality in a single-center cohort of 158 adults treated with extracorporeal cardiopulmonary resuscitation (2015-2022). Using a prespecified 24-hour landmark cohort, we assessed whether 24-hour lactate clearance (LC) provides incremental prognostic information beyond a static score. All 3 scores showed only modest post-cannulation discrimination (area under the receiver operating characteristic curve 0.574-0.668). In the landmark cohort (n = 105; complete cases n = 103), LC ≥65% was independently associated with survival after adjustment for RESCUE-IHCA (adjusted OR 3.34, 95% CI 1.45-7.68; P = .004). Adding LC to RESCUE-IHCA produced a modest improvement in discrimination (Δarea under the receiver operating characteristic curve 0.069) with mixed reclassification signals (integrated discrimination improvement 0.037, P = .032; category-based net reclassification improvement 0.178, P = .18). On decision curve analysis, the combined model showed a slightly higher net benefit at threshold probabilities of 40%-60%, although bootstrap confidence intervals for the difference included zero across all thresholds examined. These exploratory findings highlight the limitations of static pre-ECMO scores for post-cannulation prognostication and are hypothesis-generating: a pragmatic 2-step concept baseline risk estimation followed by early dynamic reassessment requires confirmation in prospective, multicenter cohorts before any clinical use.