(1531) Identifying Subphenotypes Using Latent Class Analysis in Patients With OHCA Who Undergo ECPR

Introduction: Out-of-hospital cardiac arrest (OHCA) remains low survival rates. Recent evidence suggests that extracorporeal cardiopulmonary resuscitation (ECPR) may improve outcomes for OHCA patients. However, optimal patient selection criteria and implementation strategies for ECPR are still unclear. This study aimed to identify sub-phenotypes among patients with OHCA who underwent ECPR and investigate their association with clinical outcomes.

Methods: This multi-center, retrospective, observational study used the Japanese Association for Acute Medicine OHCA registry. We included adult patients with OHCA who received veno-arterial extracorporeal membrane oxygenation (ECMO) during cardiac arrest. The primary outcome of this study was 30-day survival. The secondary outcome was favourable neurological outcome 30 days after cardiac arrest.　Latent class analysis (LCA) was performed to identify sub-phenotypes based on 14 variables including pre- and in-hospital factors before ECMO initiation. Multivariate logistic regression analysis was used to assess the association between sub-phenotypes and 30-day survival and favourable neurological outcomes.

Results: A total of 1,675 patients were included. The 30-day survival rate for eligible patients was 20.8%. LCA identified four sub-phenotypes with distinct clinical characteristics (A–D). The variables with high discriminative power in the LCA were the time from hospital arrival to ECMO initiation, pre-hospital shock delivery, and the initial cardiac rhythm monitored. Thirty-day survival rates varied significantly among the sub-phenotypes (p = 0.001), with Group B showing the highest survival rate (33.3%). Favourable neurological outcomes at 30 days also differed significantly (p = 0.036), with Group D having the highest rate (14.1%). Logistic regression analysis showed that Groups C and D had significantly higher odds of 30-day survival than Group A, with adjusted odds ratios of 1.89 (95% confidence interval: 1.33—.69) and 1.76 (95% confidence interval: 1.09—.83), respectively.

Conclusions: The sub-phenotypes identified using simple pre-hospital and in-hospital factors before ECMO initiation were associated with clinical outcomes. The time from hospital arrival to ECMO pump-on and initial cardiac rhythm were highly related to identifying sub-phenotypes.