External validation of ECPR prognostic models derived from pre-ECMO indicators in patients undergoing extracorporeal cardiopulmonary resuscitation - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Resusc Plus . 2026 Apr 1;29:101310. doi: 10.1016/j.resplu.2026.101310 Search in PMC Search in PubMed View in NLM Catalog Add to search External validation of ECPR prognostic models derived from pre-ECMO indicators in patients undergoing extracorporeal cardiopulmonary resuscitation Xingxing Li Xingxing Li a Department of Extracorporeal Life Support Center, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, Henan 450000, China Find articles by Xingxing Li a, † , Shujie Yan Shujie Yan b Department of Cardiopulmonary Bypass, Fuwai Hospital, National Centre for Cardiovascular Disease, Chinese Academy of Medical Sciences & Peking Union Medical College, National Clinical Research Centre for Cardiovascular Diseases, Beijing 100037, China Find articles by Shujie Yan b, † , Chuang Liu Chuang Liu a Department of Extracorporeal Life Support Center, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, Henan 450000, China Find articles by Chuang Liu a , Hui Zhao Hui Zhao a Department of Extracorporeal Life Support Center, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, Henan 450000, China Find articles by Hui Zhao a , Yangchao Zhao Yangchao Zhao a Department of Extracorporeal Life Support Center, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, Henan 450000, China Find articles by Yangchao Zhao a , Qianqian Sun Qianqian Sun a Department of Extracorporeal Life Support Center, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, Henan 450000, China Find articles by Qianqian Sun a , Junjie Zhao Junjie Zhao a Department of Extracorporeal Life Support Center, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, Henan 450000, China Find articles by Junjie Zhao a , Qiaoni Zhang Qiaoni Zhang b Department of Cardiopulmonary Bypass, Fuwai Hospital, National Centre for Cardiovascular Disease, Chinese Academy of Medical Sciences & Peking Union Medical College, National Clinical Research Centre for Cardiovascular Diseases, Beijing 100037, China Find articles by Qiaoni Zhang b , Gang Liu Gang Liu b Department of Cardiopulmonary Bypass, Fuwai Hospital, National Centre for Cardiovascular Disease, Chinese Academy of Medical Sciences & Peking Union Medical College, National Clinical Research Centre for Cardiovascular Diseases, Beijing 100037, China Find articles by Gang Liu b , Yuan Teng Yuan Teng b Department of Cardiopulmonary Bypass, Fuwai Hospital, National Centre for Cardiovascular Disease, Chinese Academy of Medical Sciences & Peking Union Medical College, National Clinical Research Centre for Cardiovascular Diseases, Beijing 100037, China Find articles by Yuan Teng b , Jian Wang Jian Wang b Department of Cardiopulmonary Bypass, Fuwai Hospital, National Centre for Cardiovascular Disease, Chinese Academy of Medical Sciences & Peking Union Medical College, National Clinical Research Centre for Cardiovascular Diseases, Beijing 100037, China Find articles by Jian Wang b , Zhenzhen Li Zhenzhen Li b Department of Cardiopulmonary Bypass, Fuwai Hospital, National Centre for Cardiovascular Disease, Chinese Academy of Medical Sciences & Peking Union Medical College, National Clinical Research Centre for Cardiovascular Diseases, Beijing 100037, China Find articles by Zhenzhen Li b , Luyu Bian Luyu Bian b Department of Cardiopulmonary Bypass, Fuwai Hospital, National Centre for Cardiovascular Disease, Chinese Academy of Medical Sciences & Peking Union Medical College, National Clinical Research Centre for Cardiovascular Diseases, Beijing 100037, China Find articles by Luyu Bian b , Guowei Fu Guowei Fu a Department of Extracorporeal Life Support Center, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, Henan 450000, China Find articles by Guowei Fu a, ⁎⁎ , Bingyang Ji Bingyang Ji b Department of Cardiopulmonary Bypass, Fuwai Hospital, National Centre for Cardiovascular Disease, Chinese Academy of Medical Sciences & Peking Union Medical College, National Clinical Research Centre for Cardiovascular Diseases, Beijing 100037, China Find articles by Bingyang Ji b, ⁎ Author information Article notes Copyright and License information a Department of Extracorporeal Life Support Center, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, Henan 450000, China b Department of Cardiopulmonary Bypass, Fuwai Hospital, National Centre for Cardiovascular Disease, Chinese Academy of Medical Sciences & Peking Union Medical College, National Clinical Research Centre for Cardiovascular Diseases, Beijing 100037, China ⁎ Corresponding author at: Fuwai Hospital, National centre for Cardiovascular Disease, Chinese Academy of Medical Sciences & Peking Union Medical College, National Clinical Research centre for Cardiovascular Diseases, No. 167 Beilishi Road, Xicheng District, Beijing 100000, China. [email protected] ⁎⁎ Corresponding author at: Department of Extracorporeal Life Support Center, The First Affiliated Hospital of Zhengzhou University, No. 1 Jianshe Road, Zhengzhou, Henan 450052, China. [email protected] † The first two authors contributed equally to this study. Received 2026 Feb 9; Revised 2026 Mar 21; Accepted 2026 Mar 25; Collection date 2026 May. © 2026 The Author(s) This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). PMC Copyright notice PMCID: PMC13092058 PMID: 42011229 Graphical abstract Open in a new tab Keywords: Extracorporeal Cardiopulmonary Resuscitation (ECPR), Cardiac arrest, Prognostic models, External validation, Outcomes Abstract Objective To externally validate previously published prognostic models developed exclusively from pre-extracorporeal cardiopulmonary resuscitation (ECPR) variables in a contemporary ECPR cohort. Methods We conducted a bicenter retrospective external validation of four published pre-ECPR prognostic models (Lee, RESCUE-IHCA, CHIU-S1, and CHIU-S2) in adult patients treated with ECPR between January 2015 and December 2024. Model performance was evaluated for in-hospital survival and favorable neurological outcome (FNO; Cerebral Performance Category 1–2) in the overall cohort, in-hospital cardiac arrest (IHCA), and cardiac-origin cardiac arrest (Cardio_CA) subgroups. Discrimination (the area under the receiver operating characteristic curve, AUROC), calibration, overall model fit (Brier score), and decision curve analysis (DCA) were assessed. For point-based CHIU scores, validation focused on observed outcome rates across predefined risk strata. Results Among 214 patients, 79.0% (169/214) had IHCA; survival to discharge was 45.8% and FNO occurred in 24.8%. Discrimination for survival was modest across models (overall ECPR AUROC 0.608–0.709; IHCA 0.586–0.672; Cardio_CA 0.591–0.689) but was higher for FNO (overall ECPR 0.709–0.764; IHCA 0.696–0.744; Cardio_CA 0.698–0.718). The Lee model showed poor calibration with slopes far below 1, whereas RESCUE-IHCA model underestimated survival but demonstrated better calibration (slopes close to 1), higher overall accuracy (lower Brier scores) and broader clinical utility (wider net-benefit ranges in DCA). CHIU models provided limited risk separation between adjacent strata. Conclusions In this external validation, pre-ECPR models showed modest performance, with better discrimination for neurological outcome than for survival. RESCUE-IHCA showed the most favorable overall performance. Future studies should develop and validate more robust, transportable tools. Introduction Cardiac arrest (CA) remains a major cause of mortality, with survival rates of 25–40% for in-hospital CA (IHCA) and around 10% for out-of-hospital CA (OHCA). 1 , 2 , 3 Conventional cardiopulmonary resuscitation (CCPR) yields sustained return of spontaneous circulation (ROSC) in 24.6–40% of cases, yet prognosis is poor for refractory CA patients. 75.1–92.9% of survivors experiencing significant neurological impairment. 4 , 5 , 6 Extracorporeal cardiopulmonary resuscitation (ECPR), using veno-arterial extracorporeal membrane oxygenation (VA ECMO) during CPR, has emerged as a rescue therapy. By providing immediate circulatory and respiratory support, ECPR can perserve end-organ perfusion and facilitate treatment of reversible causes of arrest. Current guidelines recognize ECPR as a potential option for selected patients, with meta-analyses showing reduced mortality (OR 0.67, 95% CI 0.51–0.87) and improved neurological outcomes compared to CCPR alone. 7 , 8 , 9 , 10 However, real-world ECPR survival remains at 31%, 11 highlighting the need for optimization. Given ECPR's resource intensity and variable outcomes, appropriate patient selection is crucial. Decisions must be made rapidly with limited early information, creating demand for pre-ECPR risk stratification tools. Although several clinical factors, such as arrest duration, initial rhythm, and age, are known to influence ECPR outcomes, 12 clinicians still lack easily applicable bedside tools to integrate these variables and support real-time decision-making. Several prognostic models 12 , 13 , 14 , 15 based exclusively on pre-ECPR variables have been developed. However, external validation remains limited, 14 and the generalizability (transportability) of these models across different arrest settings (OHCA vs IHCA), regions, and practice patterns remains uncertain. Therefore, we aimed to conduct a simultaneous, head-to-head external validation of four major pre-ECPR prognostic models within the same contemporary adult multicenter cohort. By comprehensively assessing discrimination, calibration, and clinical utility across the overall population and relevant subgroups, this comparative approach minimizes cohort-level confounding. Furthermore, by rigorously evaluating the extent to which model performance attenuates in a external setting, we seek to provide a crucial clinical message regarding the transportability of these scores—ultimately guiding clinicians on whether these models can be safely relied upon for strict ECPR triage. Methods Study design and participants We conducted a retrospective bicenter cohort study to externally validate pre-ECPR prognostic models. The study was approved by the institutional ethics committee (Ethics Numbers: 2022-KY-0272 and 2024-2517), with a waiver of informed consent owing to the retrospective design. Reporting followed the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) statement. 16 Adult patients who consecutively underwent ECPR between January 2015 and December 2024 were identified from electronic medical records. Inclusion criteria: age ≥18 years and receipt of ECPR. Exclusion criteria: missing survival/CPC data or ECMO duration <3 h. Literature search and identification of candidate prognostic models A systematic PubMed search (January 2014–October 2025) identified prognostic models for adult ECPR populations. Eligible studies developed/validated multivariable models predicting discharge survival and/or favorable neurological outcome (FNO). Reports were excluded if they involved non-ECPR populations, did not present an eligible prediction model, evaluated incompatible outcomes, used a case-control design, had unavailable full text, or were reviews/meta-analyses. Of 1343 records, 107 underwent full-text review. Twenty studies addressed ECPR outcome prediction, yielding 18 distinct models. Models were excluded if they targeted non-relevant populations, included post-ECMO predictors (i.e., were not strictly pre-ECPR), were not externally validated as fixed and reproducible models (e.g., non-validated machine-learning approaches), had outcome definitions that did not match the prespecified validation endpoint, or required predictors that were unavailable or inconsistently defined in the validation dataset. Ultimately, only four models based exclusively on pre-ECPR variables were eligible for validation ( Fig. B.1, Table A.1 ). Key characteristics of the four included models are summarized in Table 1 , including model name, publication year, country, study design, cohort size, predicted outcome, model form, derivation method, reported discrimination, and predictor variables. The area under the receiver operating characteristic curve (AUROC) reported in the original derivation studies were 0.88 for Lee, 0.719 for RESCUE-IHCA, 0.79 for CHIU-S1, and 0.83 for CHIU-S2. Table 1. The key characteristics of the original development cohorts and model specification. Feature Dimension LEE RESCUE-IHCA CHIU-S1 CHIU-S2 Basic Model Information Model Name (Year of publication) Lee (2017) RESCUE-IHCA (2022) CHIU-S1 (2025) CHIU-S2 (2025) Nationality South Korea United States Taiwan, China Taiwan, China Study Design Retrospective Single-center Retrospective Multi-center Retrospective Single-center Retrospective Single-center Patient Cohort ( n ) 111 1075 149 149 Population Type OHCA 73.9% IHCA 26.1% Pure IHCA OHCA 24.8% IHCA 55.0% EDCA 20.1% OHCA 24.8% IHCA 55.0% EDCA 20.1% Predicted Outcome (Survival Rate) Survival: 18.9% Mortality: 71.5% Survival: 40.9% Survival: 40.9% Model Features & Validation Model Form Form Scoring System Scoring System Scoring System Derivation Method Logistic Regression Bayesian Regression Logistic Regression Logistic Regression Derivation AUC (95% CI) 0.88(0.80–0.93) 0.719 (0.680–0.757) 0.79 (CI not provided) 0.83 (CI not provided) Internal Validation AUC 0.86 (0.80–0.93) Not reported Not reported Not reported External Validation AUC Not reported 0.676 (0.606–0.746) Not reported Not reported Comparison of Predictors Age ✓ (beta = -0.076) ✓ ✓ (score = 5) Initial cardiac rhythm ✓ (beta = 2.49) ✓ ✓ (score = 7) ✓ (score = 7) CPR duration ✓ (beta = -0.033) – ✓ (score = 5) ✓ (score = 7) CA duration – ✓ – – ROSC before ECPR ✓ (beta = 1.754) – – – PH before ECMO – – – ✓ (score = 6) CA time of day – ✓ – – Renal insufficiency – ✓ – – Illness category – ✓ – – CA Location – – ✓ (score = 5) – Open in a new tab Notes: “✓” indicates that the variable is included in the corresponding prediction model; “–” indicates that the variable is not included in the corresponding prediction model; Beta represent the logistic regression coefficients from the original studies; Score values represent the weighted points assigned to each variable in the scoring systems. OHCA, out-of-hospital cardiac arrest; IHCA, in-hospital cardiac arrest; EDCA, emergency department cardiac arrest; AUC, area under the curve; CI, confidence interval; CPR, cardiopulmonary resuscitation; CA, cardiac arrest; ROSC, return of spontaneous circulation; ECPR, extracorporeal cardiopulmonary resuscitation; ECMO, extracorporeal membrane oxygenation. Institutional ECPR protocol ECPR candidacy and management adhered to institutional protocols and contemporary international recommendations. 8 The ECPR team was activated for patients with refractory cardiac arrest, defined as failure to achieve sustained ROSC despite advanced life support, including three consecutive unsuccessful defibrillation attempts for shockable rhythms or 10 min of resuscitation for non-shockable rhythms. Additional eligibility criteria generally included age below 75 years, absence of significant frailty or major comorbidities, and witnessed cardiac arrest with immediate bystander or professional cardiopulmonary resuscitation (no-flow time ≤5 min). Percutaneous cannulation under real-time ultrasound guidance is the preferred approach (approximately 90% of cases), whereas open surgical cutdown is reserved strictly as a bailout option for difficult anatomy or failed percutaneous access. Data collection Data were extracted from electronic medical records using a standardized case report form, following original predictor definitions ( Table A.2 ). Variables included: 1. Baseline: age, sex, comorbidities, and illness category. 2. Arrest & Resuscitation: location, witness status, initial rhythm, presumed etiology, bystander CPR, and duration of low-flow/no-flow durations. 3. Pre-ECMO physiology: hemodynamic parameters (e.g., MAP), arterial blood gas (e.g., pH, PaO 2 /FiO 2 ratio, lactate), and biochemical results. Importantly, to ensure strict adherence to pre-cannulation prognostic scoring, all physiological and laboratory parameters were strictly defined as the last recorded values immediately prior to the ECMO flow initiation (obtained during ongoing CPR or the periarrest period). A two-stage blinded extraction process minimized bias. First, two authors extracted all baseline and pre-cannulation variables with outcomes masked. Data quality was ensured via range checks and source verification. Second, outcomes were independently assessed by two different authors blinded to predictor information. Outcomes included in-hospital survival and FNO at hospital discharge. FNO was defined as CPC 1 or 2. Statistical analysis Descriptive and univariable analysis A complete-case approach was applied for descriptive and univariable analyses. Continuous variables were assessed for normality (Shapiro–Wilk test) and variance homogeneity (Levene’s test). Normally distributed variables are presented as mean ± SD; others as median (IQR). Categorical variables are reported as frequencies (percentages). Group comparisons used Student’s t -test (normal data) or Mann–Whitney U test (non-normal), and chi-square or Fisher’s exact test for categorical variables. All tests were two-sided, with P < 0.05 considered significant. Implementation of ECPR prognostic models All models were implemented as published, without updating or recalibration. Predictors were mapped using prespecified definitions, with complete-case analysis. Lee model : Individual predicted probabilities were calculated using the published logistic regression equation: l o g ( p / ( 1 - p ) ) = 1.402 - 0.076 × ( a g e , y e a r s ) - 0.033 × ( C P R d u r a t i o n , h o u r s ) + 1.754 × ( a n y R O S C e v e n t ) + 2.490 × ( f i r s t d o c u m e n t e d r h y t h m ) CPR duration converted to hours; rhythm coded as shockable/PEA vs. asystole. RESCUE-IHCA model : The RESCUE-IHCA score was computed by summing the published point values for age, renal insufficiency, time of day, illness category, rhythm, and arrest duration. Mortality probabilities from the published lookup table were converted to survival probabilities: P survival = 1 − P mortality . CHIU-S1 and CHIU-S2 models: Both implemented as point-based scores. For each patient, total scores were calculated from predictor point assignments. Patients were stratified into high-, medium-, and low-risk categories using published risk-group probability estimates. Quantifying the model’s predictive performance and clinical utility Models were validated for in-hospital survival in the overall ECPR population and subgroups: IHCA and presumed cardiac-origin CA (Cardio_CA). Sensitivity analyses used FNO as an alternative endpoint. The predictive performance of the models was quantified in terms of discrimination, calibration, and overall fit. Clinical usefulness was evaluated using decision curve analysis (DCA). The analyses followed current methodological recommendations. 17 Discrimination was quantified by the concordance (c) statistic, which was equivalent to the area under the receiver operating characteristic curve (AUROC), with 95% confidence intervals calculated by DeLong’s method. A value of 1 indicates perfect discrimination, while a value of 0.5 indicates no better than chance. In practice, a c statistic above 0.6 suggests useful discrimination in challenging contexts, whereas values approaching 0.8 indicate strong predictive capabilities in situations with clear predictors. Calibration was assessed through multiple metrics, including: (1) calibration-in-the-large (CITL), evaluating the calibration intercept, which ideally should be close to zero. (2) observed/expected (O/E) ratios, which compared the observed outcomes to the expected probabilities derived from the model, indicating how well the model predicts the actual outcomes. (3) Calibration slope, assessing the relationship between predicted probabilities and observed outcomes, with ideal values close to one. (4) Integrated calibration index (ICI), which quantified the average absolute difference between predicted probabilities and observed outcomes. Overall model fit was quantified using the Brier score. The Brier score ranges from 0 for perfect predictions to 0.25 for a non-informative model that predicts a 50% probability for all subjects. 18 Clinical utility was evaluated through DCA, which reports net benefit across threshold probabilities. For CHIU models (point-based risk stratification tools), formal calibration, overall fit, and DCA were not applicable. Validation focused on comparing observed outcome rates across predefined risk strata. Analyses were conducted in R (version 4.5.2; R Foundation for Statistical Computing) using the following packages: pROC, rms, DescTools, dcurves, and dplyr. Results Comparisons of patient characteristics A total of 214 patients were included ( Fig. 1 ). Missing data for candidate predictors were shown in Table A.3 . Median age was 51 years; 72.4% (155/214) were male. IHCA accounted for 79.0% (169/214) of cases; Cardio_CA for 80.4% (172/214). Overall in-hospital survival was 45.8% (98/214); FNO at discharge was 24.8% (53/214). Comparisons between survivors/non-survivors and favorable/poor neurological outcomes are presented in Table 2 and Table A.4 , respectively. Fig. 1. Open in a new tab Flowchart for ECPR population . ECPR, Extracorporeal cardiopulmonary resuscitation; ECMO, extracorporeal membrane oxygenation; IHCA, In-hospital cardiac arrest; CPC, Cerebral Performance Category. Table 2. Comparison of patient characteristics between survivors and non-survivors. Variables Total ( n = 214) Survivors ( n = 98) Non-survivors ( n = 116) P value Demographics & History Age, years 50.99 ± 14.58 51.39 ± 14.06 50.66 ± 15.07 0.715 Male, n (%) 155 (72.4%) 69 (70.4%) 86 (74.1%) 0.543 BMI, kg/m 2 24.56 (23.03–27.30) 24.50 (22.79–27.09) 24.62 (23.11–27.52) 0.900 Hypertension, n (%) 94 (43.9%) 40 (40.8%) 54 (46.6%) 0.400 Diabetes, n (%) 62 (29%) 29 (29.6%) 33 (28.4%) 0.854 Heart disease, n (%) 61 (28.5%) 29 (29.6%) 32 (27.6%) 0.746 Renal insufficiency, n (%) 10 (4.7%) 2 (2%) 8 (6.9%) 0.176 Cerebrovascular Disease, n (%) 23 (10.7%) 12 (12.2%) 11 (9.5%) 0.516 Cardiac Arrest Details Initial non-asystole rhythm, n (%) 148 (70.8%) 75 (78.9%) 73 (64%) 0.018 Initial shockable rhythm, n (%) 104 (49.8%) 60 (63.2%) 44 (38.6%) <0.001 Cardio_CA, n (%) 172 (80.4%) 84 (85.7%) 88 (75.9%) 0.071 CA Type (IHCA), n (%) 169 (79.0%) 79 (80.6%) 90 (77.6%) 0.588 CA Location (ER), n (%) 77 (36.0%) 37 (37.8%) 40 (34.5%) 0.619 CA Time of Day, n (%) 0.915 7:00–14:59 82 (38.7%) 38 (39.6%) 44 (37.9%) 15:00–22:59 99 (46.7%) 45 (46.9%) 54 (46.6%) 23:00–06:59 31 (14.6%) 13 (13.5%) 18 (15.5%) CA Time (Day/Night), n (%) 116 (54.7%) 53 (55.2%) 63 (54.3%) 0.896 CA Time (Weekday), n (%) 79 (37.3%) 40 (41.7%) 39 (33.6%) 0.288 Illness category, n (%) 0.105 Medical 170 (79.4%) 72 (73.5%) 98 (84.5%) Surgical cardiac 25 (11.7%) 16 (16.3%) 9 (7.8%) Surgical non-cardiac 19 (8.9%) 10 (10.2%) 9 (7.8%) ROSC, n (%) 96 (48.7%) 17 (17.7%) 24 (21.1%) <0.001 CA duration, min 35.00 (22.75–60.00) 28.00 (18.00–40.00) 41.50 (30.00–70.00) <0.001 CPR duration, min 34.50 (22.75–60.00) 26.50 (18.00–40.00) 41.00 (30.00–68.75) <0.001 Physiological & Lab Measures MAP, mmHg 52.00 (41.00–60.00) 56.33 (45.25–65.00) 47.00 (40.00–57.00) <0.001 MAP ≥ 60, n (%) 46 (24.6%) 30 (36.6%) 16 (15.2%) 0.013 PaO 2 /FiO 2 , mmHg 90.00 (57.90–169.39) 105.00 (70.35–182.00) 79.17 (51.10–164.25) 0.044 Hemoglobin, g/L 125.25 ± 32.43 121.68 ± 28.68 127.89 ± 34.87 0.254 Platelet count, ×10 9 /L 201.00 (140.00–260.25) 215.00 (155.50–280.00) 190.00 (126.00–252.00) 0.216 Bilirubin, μmol/L 10.50 (6.55–16.20) 11.09 (7.35–16.03) 10.43 (6.35–16.50) 0.938 Creatinine, μmol/L 94.50 (71.70–144.48) 84.00 (67.00–103.00) 107.42 (72.50–173.04) 0.015 Lactate, mmol/L 11.45 (5.53–15.00) 9.80 (4.88–14.83) 12.05 (6.85–15.00) 0.093 Glucose, mmol/L 14.90 (9.00–20.18) 15.10 (9.25–20.75) 14.60 (8.80–19.80) 0.408 PH 7.21 (7.04–7.34) 7.24 (7.14–7.38) 7.14 (6.98–7.30) 0.002 pH ≥ 7.31, n (%) 66 (31.3%) 37 (38.5%) 29 (25.2%) 0.038 PCO2, mmHg 40.00 (30.70–54.00) 39.50 (30.92–53.10) 42.00 (30.70–54.00) 0.747 PO2, mmHg 83.00 (57.00–147.00) 98.85 (65.50–128.75) 75.00 (49.60–160.00) 0.088 HCO3 – , mmol/L 15.70 (11.50–20.90) 16.80 (12.35–22.45) 14.35 (10.80–18.88) 0.047 Interventions & Outcomes PCI, n (%) 91 (42.5%) 49 (50.0%) 42 (36.2%) 0.042 ECMO duration, hours 61.62 (36.29–112.72) 71.00 (47.33–115.33) 44.00 (26.94–105.33) <0.001 Hospital stay, days 9.97 (2.60–24.11) 22.4 8(13.49–36.60) 3.63 (1.86–9.95) <0.001 ECMO outcome, n (%) 117 (54.7%) 98 (100%) 19 (16.4%) <0.001 CPC 1–2, n (%) 53 (24.8%) 53 (54.1%) 0 (0%) <0.001 Open in a new tab Notes: Continuous variables are presented as mean ± standard deviation for normally distributed data or median (interquartile range) for non-normally distributed data. Categorical variables are presented as frequency (percentage). All physiological and laboratory variables (including MAP, PaO 2 /FiO 2 ratio, and blood gases) represent the last recorded values strictly immediately prior to the initiation of ECMO flow (during ongoing CPR or the periarrest period, accurately reflecting the pre-ECPR status. BMI, body mass index; CA, cardiac arrest; Cardio_CA, cardiac-origin cardiac arrest; IHCA, in-hospital cardiac arrest; ER, emergency room; ROSC, return of spontaneous circulation; CPR, cardiopulmonary resuscitation; MAP, mean arterial pressure; PaO 2 /FiO 2 , ratio of arterial oxygen partial pressure to fractional inspired oxygen; PCI, percutaneous coronary intervention; ECMO, extracorporeal membrane oxygenation; CPC, Cerebral Performance Category. Survivors had significantly shorter CA duration (median 28.0 [IQR 18.0–40.0] vs. 41.5 min [30.0–70.0]; p < 0.001) and CPR duration (26.5 [18.0–40.0] vs. 41.0 min [30.0–68.75]; p < 0.001). They more frequently presented with initial non-asystole rhythm (78.9% vs. 64.0%; p = 0.018) and shockable rhythm (63.2% vs. 38.6%; p < 0.001). Survivors also had higher mean arterial pressure (56.33 [45.25–65.00] vs. 47.00 mmHg [40.00–57.00]; p < 0.001), higher PaO 2 /FiO 2 ratio (105.00 [70.35–182.00] vs. 79.17 [51.10–164.25]; p = 0.044), higher arterial pH (7.24 [7.14–7.38] vs. 7.14 [6.98–7.30]; p = 0.002), higher HCO 3 − (16.80 [12.35–22.45] vs. 14.35 mmol/L [10.80–18.88]; p = 0.047), and lower serum creatinine (84.00 [67.00–103.00] vs. 107.42 μmol/L [72.50–173.04]; p = 0.015). ROSC (48.7% vs. 21.1%; p < 0.001) and PCI therapy (50.0% vs. 36.2%; p = 0.042) were more common among survivors. ECMO duration (71.0 [47.33–115.33] vs. 44.0 h [26.94–105.33]; p < 0.001) and hospital stay (22.48 [13.49–36.6] vs. 3.63 days [1.86–9.95]; p < 0.001) were longer in survivors. Derivation cohorts differed in age and arrest location. The Lee cohort ( n = 111) was younger (survivors 47.0 ± 14.8 vs. non-survivors 57.9 ± 14.6 years; p = 0.003) and predominantly OHCA (73.9%). 12 The RESCUE-IHCA cohort comprised pure IHCA cases with a modest age difference (survivors median 58 vs. non-survivors 61 years; p = 0.019). 13 The CHIU cohort ( n = 149) had mixed arrest location (OHCA 24.8%, IHCA 55.0%, EDCA 20.1%) with no significant age difference. 15 Our validation cohort was younger overall (median 51 years) and predominantly IHCA (79.0%), contrasting with the OHCA-dominant Lee cohort and heterogeneous CHIU cohort. Quantifying the model’s predictive performance Due to the complete-case analysis approach, the exact number of patients validated varied slightly depending on the specific model being evaluated, as detailed in Table A.5 . Discrimination The discrimination performance of the Lee and RESCUE-IHCA models is presented in Tables 3 and A.6 , and the discrimination performance of the CHIU models is presented in Table A.7 . Table 3. Predictive performance of Lee and RESCUE-IHCA models for survival at discharge in the external validation cohorts. Performance measure ECPR IHCA Cardio_CA Lee RESCUE Lee RESCUE Lee RESCUE Discrimination C statistic (AUROC) 0.608 (0.527–0.688) 0.709(0.636–0.782) 0.586 (0.493–0.679) 0.672 (0.587–0.758) 0.591 (0.501–0.682) 0.689 (0.607–0.772) Calibration Calibration-in-the-large −0.349(−0.709–0.010) 0.573(0.278–0.868) −0.439(−0.847--0.031) 0.521 (0.189–0.853) −0.155 (−0.559–0.25) 0.696 (0.366–1.026) Observed/expected 0.893 (0.748–1.044) 1.378 (1.169–1.575) 0.872 (0.713–1.031) 1.333 (1.114–1.556) 0.953 (0.787–1.131) 1.459 (1.24–1.697) Calibration slope 0.215 (0.051–0.379) 1.127(0.647–1.608) 0.176 (−0.007–0.359) 0.975(0.432–1.519) 0.188 (0.01–0.366) 1.025 (0.505–1.544) Integrated calibration index 0.199 (0.142–0.270) 0.117 (0.067–0.184) 0.23 (0.160–0.315) 0.111 (0.062–0.188) 0.214 (0.149–0.292) 0.148 (0.083–0.221) Overall fit Brier score 0.290(0.248–0.332) 0.231 (0.202–0.260) 0.306 (0.261–0.354) 0.237 (0.208–0.270) 0.297(0.251–0.343) 0.245 (0.213–0.278) Open in a new tab Notes: Values are reported as estimate (95% CI). AUROC, the area under the receiver operating characteristic curve; ECPR, extracorporeal cardiopulmonary resuscitation; IHCA, in-hospital cardiac arrest; Cardio_CA, cardiac origin cardiac arrest. Overall ECPR Cohort For survival to discharge the discrimination was modest across models, with AUROCs ranging from 0.608 (Lee) to 0.709 (RESCUE-IHCA). Discrimination improved for FNO, with AUROCs in the fair-to-good range from 0.709 (Lee) to 0.764 (RESCUE-IHCA), and the AUROC of 0.764 was the highest among all models. IHCA cohort For survival to discharge, the discrimination was modest across models, with AUROCs ranging from 0.586 (Lee) to 0.672 (RESCUE-IHCA). Discrimination improved for FNO, with AUROCs in the fair-to-good range from 0.696 (Lee) to 0.744 (RESCUE-IHCA). Cardio_CA cohort For survival to discharge, the discrimination was modest across models, with AUROCs ranging from 0.591 (Lee) to 0.689 (RESCUE-IHCA). Discrimination improved for FNO, with AUROCs in the fair-to-good range from 0.698 (Lee) to 0.718 (RESCUE-IHCA). Calibration Survival to discharge The Lee model demonstrated near-acceptable overall calibration with O/E ratios slightly below 1 (0.893–0.953) and negative CITL, indicating a tendency to overpredict survival; however, its calibration slopes were far below 1 (0.176–0.215), suggesting overly extreme predictions, and its ICI ranged from 0.199 to 0.230. In contrast, the RESCUE-IHCA model systematically underestimated survival, with O/E ratios greater than 1 (1.333–1.459) and positive CITL, but its calibration slopes were close to 1 (0.975–1.127) and its ICI was lower (0.111–0.148) than that of the Lee model ( Table 3 , Fig. 2 ). Fig. 2. Open in a new tab Calibration plots for survival in the external validation cohorts . Calibration plots comparing observed and predicted survival probabilities for the LEE models (panels A, C, E) and RESCUE-IHCA models (panels B, D, F) in external validation cohorts. Panels A and B: calibration for the overall ECPR cohort; Panels C and D: calibration for the IHCA cohort; Panels E and F: calibration for the cardiac-origin CA cohort. ECPR, Extracorporeal cardiopulmonary resuscitation; IHCA, In-hospital cardiac arrest. FNO The Lee model overestimated FNO, with O/E ratios of 0.468–0.515, negative CITL values (−1.665 to −1.915), slopes below 1 (0.446–0.478), and higher ICI values (0.248–0.279). In contrast, the RESCUE-IHCA model showed closer agreement, with O/E ratios of 0.724–0.798, less negative CITL values (−0.493 to −0.348), slopes above 1 (1.329–1.694), and lower ICI values (0.068–0.103) ( Table A.6, Fig. B.2 ). Overall model fit Overall model fit assessed by the Brier score is shown in Tables 3 and A.6 . For survival, RESCUE-IHCA consistently showed lower Brier scores than Lee across ECPR (0.231 vs 0.290), IHCA (0.237 vs 0.306), and Cardio_CA (0.245 vs 0.297). For FNO, RESCUE-IHCA also outperformed Lee with substantially lower Brier scores in ECPR (0.161 vs 0.269), IHCA (0.167 vs 0.287), and Cardio_CA (0.174 vs 0.270), indicating smaller mean squared prediction error. Clinical utility DCA assessed the clinical utility of the Lee and RESCUE-IHCA models across different threshold probabilities for clinical decision-making. Survival to discharge For survival prediction, both models demonstrated clinical utility within specific threshold probability ranges ( Fig. 3 ). In the overall ECPR cohort, the Lee model showed net benefit from threshold probabilities of 0.38–0.52, while the RESCUE model provided utility from 0.29 to 0.72, indicating broader clinical applicability. Similar patterns were observed in the IHCA cohort (Lee: 0.40–0.51; RESCUE: 0.36–0.73) and Cardio_CA cohort (Lee: 0.4–0.52; RESCUE: 0.43–0.73). The RESCUE model consistently demonstrated wider ranges of clinical utility across all patient populations. Fig. 3. Open in a new tab Decision curve analysis evaluating the clinical utility of the LEE and RESCUE-IHCA prognostic models for predicting survival at discharge in three patient cohorts . Panel A: Analysis in the overall ECPR cohort. Panel B: Analysis in the in–hospital cardiac arrest (IHCA) cohort. Panel C: Analysis in the cardiac-origin CA cohort. In each panel, the net benefit of using each model to guide treatment decisions (“Treat per Lee” and “Treat per RESCUE”) is plotted against the threshold probability of survival. The curves are compared with two reference strategies: treating all patients (“Treat all”) and treating no patients (“Treat none”). Higher net benefit indicates greater clinical utility of the model within the corresponding threshold range. FNO For FNO prediction, the models showed more limited but still clinically meaningful utility ( Fig. B.3 ). In both ECPR and IHCA cohorts, the Lee provided net benefit from 0 to 0.30, while the RESCUE-IHCA model showed substantially broader utility (0–0.72). In the Cardio_CA cohort, the Lee provided net benefit from 0 to 0.33, while the RESCUE-IHCA model provided utility from 0 to 0.51. Comparison of risk stratification for the CHIU models Table A.7 presents a comparison of risk stratification for the CHIU models. CHIU-S1 showed limited ability to discriminate between the high-risk and medium-risk groups in the validation cohorts: the observed survival rates were similar (ECPR: 27.8% vs. 33.0%; IHCA: 27.8% vs. 35.8%; Cardio_CA: 36.4% vs. 37.5%). CPC rates were also comparable. CHIU-S2 showed insufficient discrimination between the medium- and low-risk groups for survival. Observed survival rates differed modestly (ECPR: 47.6% vs. 55.6%; IHCA: 47.0% vs. 52.4%; Cardio_CA: 52.2% vs. 55.9%), indicating weaker risk stratification in intermediate-to-low risk range. Discussion This bicenter validation study of 214 ECPR patients conducted a simultaneous, head-to-head comparison of four pre-ECPR prognostic models (LEE, RESCUE-IHCA, CHIU-S1, CHIU-S2) across three cohorts (overall ECPR, IHCA, Cardio_CA) for survival and FNO at discharge. Discrimination for survival was modest and attenuated relative to original cohorts, while discrimination for FNO was generally higher. Calibration varied substantially across models and cohorts. Among the evaluated tools, RESCUE-IHCA demonstrated the most robust overall performance, with lower Brier scores and the widest range of clinical utility on decision curve analysis, particularly for neurological outcome prediction. CHIU-S1 showed slightly better risk discrimination than CHIU-S2 but remained inferior to RESCUE-IHCA. Among the four prognostic models, only the RESCUE-IHCA model has previously been externally validated. RESCUE-IHCA was developed using 1075 patients in the derivation cohort and 197 patients in the external validation cohort, achieving an AUC of 0.719 in derivation cohort and 0.676 in the validation cohort. 13 Subsequent validation in an Asian single-center retrospective cohort of 324 adult IHCA patients (in-hospital mortality 71%) showed comparable discrimination (AUC 0.63), supporting cross-regional transportability, albeit with modest performance. 14 In our IHCA cohort, RESCUE-IHCA yielded an AUC of 0.672 for survival to discharge, which was comparable to its original external validation. Notably, RESCUE-IHCA performed best in our overall ECPR cohort for favorable neurological outcome (AUC 0.764), suggesting that this IHCA-derived score may generalize particularly well for neurological prognostication in broader ECPR populations. All four models relied mainly on established clinical variables (including age, CA/CPR duration, initial rhythm, early metabolic markers) that are known to influence ECPR outcomes, as supported by prior registries and observational studies. 12 , 13 , 15 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 However, the predictive performance of these models varied, and overall discrimination remained modest. The modest performance may reflect signal dilution from factors not captured at baseline. Key contributors include (i) substantial heterogeneity across OHCA and IHCA settings, (ii) unmeasured CPR quality beyond arrest duration (i.e., the effectiveness of low-flow perfusion), and (iii) post-cannulation determinants including post-resuscitation care, evolving organ failure, ECMO-related complications, and center-specific practice patterns, that strongly shape outcomes but are absent from pre-ECPR tools. 27 , 28 , 29 , 30 OHCA and IHCA differ markedly in baseline profiles and resuscitation logistics, so prognostic scores derived in one setting may not transport well to the other. Compared with IHCA, OHCA more often involves delayed or variable initiation of CPR with longer low-flow times, while IHCA uniquely includes etiologies such as post-cardiotomy arrest and is generally supported by faster escalation and more controlled in-hospital resuscitation conditions. 31 , 32 , 33 These structural differences shift baseline risk, rhythm distribution, and the severity of ischemia–reperfusion/metabolic derangement, thereby degrading discrimination and calibration when models are applied across arrest settings. To address this, we prespecified and evaluated model performance within the IHCA population and within a cardiac-etiology cohort. Consistent with these considerations, our validation cohort was IHCA-heavy (79%), and applying OHCA-derived models in this context predictably contributed to calibration drift. In particular, the Lee model, derived predominantly from an OHCA cohort, showed comparatively weaker performance in our largely IHCA population, plausibly due to this case-mix mismatch. Conversely, the stronger performance of RESCUE-IHCA is consistent with better alignment between its derivation setting and our cohort composition. Arrest duration is an imperfect indicator of ischemic exposure because it does not quantify perfusion quality during CPR. Prior ECPR studies suggest that outcome differences are partly attributable to low-flow duration and the effectiveness of CPR, particularly in OHCA where CPR quality may be more variable during transport and early resuscitation. 34 , 35 In our cohort, intra-arrest MAP was higher among survivors, and achieving MAP ≥ 60 mmHg was associated with better survival to discharge, indicating that patients with similar low-flow times can have substantially different end-organ perfusion and prognosis depending on CPR effectiveness. Current pre-ECPR models rarely incorporate objective CPR-quality variables such as intra-arrest hemodynamics, end-tidal CO 2 , or oxygenation dynamics, even though intra-arrest physiological derangement (for example, lactate during CPR) has been linked to outcomes in ECPR-treated IHCA populations. 36 This omission likely attenuates prognostic performance, especially when CPR quality varies across teams, transport conditions, and institutions. 35 Even when baseline risk is well characterized, outcomes after ECPR are strongly determined by post-cannulation care and center capability—particularly the delivery of comprehensive post–cardiac arrest care (hemodynamics, oxygenation/ventilation, temperature control, and neuroprognostication) and the prevention and management of downstream ICU complications (e.g., bleeding, stroke, limb ischemia, infection, and acute kidney injury). 29 , 30 These processes vary across institutions because of differences in systems of care, protocols, staffing, cannulation logistics, ICU expertise, and treatment-limitation practices (including the timing and rationale for withholding/withdrawing life-sustaining therapy). 29 , 35 Because such center-level determinants are not captured in pre-ECPR prediction tools, calibration and transportability may degrade when models are applied to centers with different practice patterns. In contrast to survival prediction, neurological outcome prediction showed relatively better performance, which is biologically plausible given the direct link between neurological injury and early ischemia, as reflected by pre-ECPR variables. 37 , 38 Neurological status is also a more patient-centered and comparatively less system-dependent endpoint, whereas survival to discharge is strongly influenced by downstream, center- and culture-dependent factors (ECMO management, withdrawal practices, discharge policies, and resource constraints). Collectively, this supports prioritizing standardized neurological prognostication and incorporating neurological endpoints when developing and implementing ECPR risk tools. Future models should move beyond static snapshots and incorporate dynamic intra-arrest trajectories and objective CPR-quality metrics (e.g., end-tidal CO 2 and intra-arrest MAP, lactate trends/clearance, oxygenation dynamics). Architecturally, separate OHCA and IHCA pathways are warranted, and routine center-specific recalibration should be planned to mitigate calibration drift. Limitations Several limitations should be acknowledged. First, the relatively small sample size ( n = 214) and limited outcome events (e.g., favorable neurological outcomes) may reduce precision, especially for subgroup comparisons. Nevertheless, considering the relative rarity and complexity of ECPR, this cohort still presents a comparatively large single-center population. Second, the characteristics of ECPR populations may vary across centers. In our cohort, IHCA cases were more prevalent and the proportion of patients with an initial rhythm of asystole was slightly higher, which may have influenced the external validation performance of these prediction models. Third, as this study uses observational data from historical cohorts, a self-fulfilling bias may exist, where prognostic factors influence clinical decisions (e.g., treatment intensity or withdrawal), thereby affecting outcomes and model evaluation. Conclusion In this contemporary bicenter adult ECPR cohort, pre-ECPR prognostic models showed modest discrimination for survival but better performance for neurological outcomes. RESCUE-IHCA demonstrated the most favorable overall performance, possibly reflecting the predominance of IHCA cases in our cohort. Pre-ECPR scores should not be used as stand-alone criteria but rather as adjunctive decision-support tools. Future efforts should focus on developing more robust and transportable models. CRediT authorship contribution statement Xingxing Li: Writing – original draft, Visualization, Software, Methodology, Formal analysis, Data curation, Conceptualization. Shujie Yan: Writing – review & editing, Validation, Supervision, Formal analysis, Conceptualization. Chuang Liu: Writing – original draft, Software, Methodology, Formal analysis, Data curation. Hui Zhao: Supervision, Project administration, Conceptualization. Yangchao Zhao: Methodology, Data curation, Conceptualization. Qianqian Sun: Software, Methodology, Formal analysis, Data curation. Junjie Zhao: Software, Methodology, Formal analysis, Data curation. Qiaoni Zhang: Software, Methodology, Formal analysis, Data curation. Gang Liu: Software, Methodology, Formal analysis, Data curation. Yuan Teng: Software, Methodology, Formal analysis, Data curation. Jian Wang: Software, Methodology, Formal analysis, Data curation. Zhenzhen Li: Software, Methodology, Formal analysis, Data curation. Luyu Bian: Software, Methodology, Formal analysis, Data curation. Guowei Fu: Writing – review & editing, Resources, Project administration, Conceptualization. Bingyang Ji: Writing – review & editing, Resources, Project administration, Funding acquisition. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgments The study was supported by grants from Beijing Natural Science Foundation (grant number L251085) and the China National High Level Hospital Clinical Research Funding (2025-GSP-GG-9). The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. We thank the following for their help and advice. Glossary ECPR extracorporeal cardiopulmonary resuscitation ECMO extracorporeal membrane oxygenation FNO favorable neurological outcome IHCA in-hospital cardiac arrest Cardio_CA cardiac-origin cardiac arrest OHCA out-of-hospital cardiac arrest AUROC area under the receiver operating characteristic curve AUC area under the curve DCA decision curve analysis CA cardiac arrest CCPR conventional cardiopulmonary resuscitation ROSC return of spontaneous circulation VA ECMO veno-arterial extracorporeal membrane oxygenation CPR cardiopulmonary resuscitation CPC Cerebral Performance Category CITL calibration-in-the-large O/E observed/expected ICI Integrated calibration index EDCA emergency department cardiac arrest ER emergency room MAP mean arterial pressure PaO 2 /FiO 2 ratio of arterial oxygen partial pressure to fractional inspired oxygen PCI percutaneous coronary intervention BMI body mass index CI confidence interval PEA pulseless electrical activity VF ventricular fibrillation pVT pulseless ventricular tachycardia Footnotes Appendix A Supplementary material to this article can be found online at https://doi.org/10.1016/j.resplu.2026.101310 . Contributor Information Guowei Fu, Email: [email protected]. Bingyang Ji, Email: [email protected]. Appendix A. Supplementary material The following are the Supplementary material to this article: Supplementary Data 1 mmc1.docx (488.7KB, docx) References 1. Do S.N., Luong C.Q., Pham D.T., Nguyen C.V., Ton T.T., Pham T.T., et al. Survival after out-of-hospital cardiac arrest, Viet Nam: multicentre prospective cohort study. Bull World Health Organ. 2021;99:50–61. doi: 10.2471/BLT.20.269837. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 2. Penketh J., Nolan J.P. In-hospital cardiac arrest: the state of the art. Crit Care. 2022;26:376. doi: 10.1186/s13054-022-04247-y. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. Bertic M., Worme M., Foroutan F., Rao V., Ross H., Billia F., et al. Predictors of survival and favorable neurologic outcome in patients treated with eCPR: a systematic review and meta-analysis. J Cardiovasc Transl Res. 2022;15:279–290. doi: 10.1007/s12265-021-10195-9. [ DOI ] [ PubMed ] [ Google Scholar ] 4. Alao D.O., Mohammed N.A., Hukan Y.O., Al Neyadi M., Jummani Z., Dababneh E.H., et al. The epidemiology and outcomes of adult in-hospital cardiac arrest in a high-income developing country. Resusc Plus. 2022;10 doi: 10.1016/j.resplu.2022.100220. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Koukousaki D., Kosmopoulos M., Sebastian P., Kalra R., Alexy T., Gutierrez A., et al. Effect of time to extracorporeal cardiopulmonary resuscitation on survival outcomes in in-hospital cardiac arrest patients. Resuscitation. 2025;217 doi: 10.1016/j.resuscitation.2025.110857. [ DOI ] [ PubMed ] [ Google Scholar ] 6. Tsao C.W., Aday A.W., Almarzooq Z.I., Anderson C.A.M., Arora P., Avery C.L., et al. Heart disease and stroke statistics-2023 update: a report from the American Heart Association. Circulation. 2023;147:e93–e621. doi: 10.1161/CIR.0000000000001123. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 7. Dezfulian C., Cabañas J.G., Buckley J.R., Cash R.E., Crowe R.P., Drennan I.R., et al. Part 4: systems of care: 2025 American Heart Association guidelines for cardiopulmonary resuscitation and emergency cardiovascular care. Circulation. 2025;152:S353–S384. doi: 10.1161/CIR.0000000000001378. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 8. Soar J., Böttiger B.W., Carli P., Jiménez F.C., Cimpoesu D., Cole G., et al. European Resuscitation Council guidelines 2025 adult advanced life support. Resuscitation. 2025;215(Suppl 1) doi: 10.1016/j.resuscitation.2025.110769. [ DOI ] [ PubMed ] [ Google Scholar ] 9. Cjw L., K R., Rr L., Mj H., Y C., R L., et al. Extracorporeal cardiopulmonary resuscitation versus conventional cardiopulmonary resuscitation in adults with cardiac arrest: a comparative meta-analysis and trial sequential analysis. Lancet Respir Med. 2023;11 doi: 10.1016/S2213-2600(23)00137-6. [ DOI ] [ PubMed ] [ Google Scholar ] 10. Heuts S., Ubben J.F.H., Kawczynski M.J., Gabrio A., Suverein M.M., Delnoij T.S.R., et al. Extracorporeal cardiopulmonary resuscitation versus standard treatment for refractory out-of-hospital cardiac arrest: a Bayesian meta-analysis. Crit Care. 2024;28:217. doi: 10.1186/s13054-024-05008-9. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 11. Registry Dashboard. ECMO. Extracorporeal Membrane Oxygenation; n.d. https://www.elso.org/registry/elsoliveregistrydashboard.aspx (accessed December 30, 2025). 12. Lee S.W., Han K.S., Park J.S., Lee J.S., Kim S.J. Prognostic indicators of survival and survival prediction model following extracorporeal cardiopulmonary resuscitation in patients with sudden refractory cardiac arrest. Ann Intensive Care. 2017;7:87. doi: 10.1186/s13613-017-0309-y. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Tonna J.E., Selzman C.H., Girotra S., Presson A.P., Thiagarajan R.R., Becker L.B., et al. Resuscitation using ECPR during In-Hospital Cardiac Arrest (Rescue-IHCA) mortality prediction score and external validation. JACC Cardiovasc Interv. 2022;15:237–247. doi: 10.1016/j.jcin.2021.09.032. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Ho Y.-J., Su P.-I., Chi C.-Y., Tsai M.-S., Chen Y.-S., Huang C.-H. External validation of the RESCUE-IHCA score as a predictor for in-hospital cardiac arrest patients receiving extracorporeal cardiopulmonary resuscitation. West J Emerg Med. 2024;25:894–902. doi: 10.5811/westjem.18601. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 15. Chiu C.-C., Chang Y.-J., Chiu C.-W., Chen Y.-C., Hsieh Y.-K., Hsiao S.-W., et al. Comparing clinical only and combined clinical laboratory models for ECPR outcomes in refractory cardiac arrest. Sci Rep. 2025;15:2915. doi: 10.1038/s41598-025-87200-7. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Collins G.S., Reitsma J.B., Altman D.G., Moons K.G.M. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. BMJ. 2015;350 doi: 10.1136/bmj.g7594. [ DOI ] [ PubMed ] [ Google Scholar ] 17. Riley R.D., Archer L., Snell K.I.E., Ensor J., Dhiman P., Martin G.P., et al. Evaluation of clinical prediction models (part 2): how to undertake an external validation study. BMJ. 2024;384 doi: 10.1136/bmj-2023-074820. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Haider H., Hoehn B., Davis S., Greiner R. Effective ways to build and evaluate individual survival distributions. arXiv. 2018 doi: 10.48550/arXiv.1811.11347. [ DOI ] [ Google Scholar ] 19. Okada A., Okada Y., Kandori K., Nakajima S., Matsuyama T., Kitamura T., et al. Application of the TiPS65 score for out-of-hospital cardiac arrest patients with initial non-shockable rhythm treated with ECPR. Resusc Plus. 2023;16 doi: 10.1016/j.resplu.2023.100458. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Kawauchi A., Okada Y., Aoki M., Inoue A., Hifumi T., Sakamoto T., et al. Clinical prediction score for patients with initial nonshockable rhythm receiving extracorporeal cardiopulmonary resuscitation for out-of-hospital cardiac arrest: development and internal validation. J Am Heart Assoc. 2025;14 doi: 10.1161/JAHA.125.042734. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Park S.B., Yang J.H., Park T.K., Cho Y.H., Sung K., Chung C.R., et al. Developing a risk prediction model for survival to discharge in cardiac arrest patients who undergo extracorporeal membrane oxygenation. Int J Cardiol. 2014;177:1031–1035. doi: 10.1016/j.ijcard.2014.09.124. [ DOI ] [ PubMed ] [ Google Scholar ] 22. Okada Y., Kiguchi T., Irisawa T., Yamada T., Yoshiya K., Park C., et al. Development and validation of a clinical score to predict neurological outcomes in patients with out-of-hospital cardiac arrest treated with extracorporeal cardiopulmonary resuscitation. JAMA Netw Open. 2020;3 doi: 10.1001/jamanetworkopen.2020.22920. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Li Z., Gao J., Wang J., Xie H., Guan Y., Zhuang X., et al. Mortality risk factors in patients receiving ECPR after cardiac arrest: development and validation of a clinical prognostic prediction model. Am J Emerg Med. 2024;76:111–122. doi: 10.1016/j.ajem.2023.11.048. [ DOI ] [ PubMed ] [ Google Scholar ] 24. Ryu J.-A., Chung C.R., Cho Y.H., Sung K., Jeon K., Suh G.Y., et al. Neurologic outcomes in patients who undergo extracorporeal cardiopulmonary resuscitation. Ann Thorac Surg. 2019;108:749–755. doi: 10.1016/j.athoracsur.2019.03.033. [ DOI ] [ PubMed ] [ Google Scholar ] 25. Redfors B., Byttner A., Bengtsson D., Watson P., Lannemyr L., Lundgren P., et al. The pre-ECPR score: developing and validating a multivariable prediction model for favorable neurological outcomes in patients undergoing extracorporeal cardiopulmonary resuscitation. J Cardiothorac Vasc Anesth. 2024;38:3018–3028. doi: 10.1053/j.jvca.2024.09.009. [ DOI ] [ PubMed ] [ Google Scholar ] 26. Kim T.W., Ahn J., Ryu J.-A. Machine learning-based predictor for neurologic outcomes in patients undergoing extracorporeal cardiopulmonary resuscitation. Front Cardiovasc Med. 2023;10 doi: 10.3389/fcvm.2023.1278374. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Ng P.Y., Li A.C.C., Fang S., Lin J.C.R., Ip A., Chan W.M., et al. Predictors of favorable neurologic outcomes in a territory-first extracorporeal cardiopulmonary resuscitation program. ASAIO J. 2022;68:1158–1164. doi: 10.1097/MAT.0000000000001620. [ DOI ] [ PubMed ] [ Google Scholar ] 28. Taira T., Inoue A., Ijuin S., Nishimura T., Moriyama T., Omoda M., et al. Acute kidney injury after extracorporeal cardiopulmonary resuscitation for out-of-hospital cardiac arrest. Crit Care Med. 2025 doi: 10.1097/CCM.0000000000006946. [ DOI ] [ PubMed ] [ Google Scholar ] 29. Naito H., Sakuraya M., Hongo T., Takada H., Yumoto T., Yorifuji T., et al. Prevalence, reasons, and timing of decisions to withhold/withdraw life-sustaining therapy for out-of-hospital cardiac arrest patients with extracorporeal cardiopulmonary resuscitation. Crit Care. 2023;27:252. doi: 10.1186/s13054-023-04534-2. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 30. Ait Hssain A., Chalkias A., Vahedian-Azimi A., Elmelliti H., Alamami A., Tawel R., et al. Survival rates with favorable neurological outcomes after in-hospital and out-of-hospital cardiac arrest: a prospective cohort study. Intensive Crit Care Nurs. 2025;87 doi: 10.1016/j.iccn.2024.103889. [ DOI ] [ PubMed ] [ Google Scholar ] 31. Wang C.-H., Chou N.-K., Becker L.B., Lin J.-W., Yu H.-Y., Chi N.-H., et al. Improved outcome of extracorporeal cardiopulmonary resuscitation for out-of-hospital cardiac arrest – a comparison with that for extracorporeal rescue for in-hospital cardiac arrest. Resuscitation. 2014;85:1219–1224. doi: 10.1016/j.resuscitation.2014.06.022. [ DOI ] [ PubMed ] [ Google Scholar ] 32. D’Arrigo S., Cacciola S., Dennis M., Jung C., Kagawa E., Antonelli M., et al. Predictors of favourable outcome after in-hospital cardiac arrest treated with extracorporeal cardiopulmonary resuscitation: a systematic review and meta-analysis. Resuscitation. 2017;121:62–70. doi: 10.1016/j.resuscitation.2017.10.005. [ DOI ] [ PubMed ] [ Google Scholar ] 33. Avalli L., Maggioni E., Formica F., Redaelli G., Migliari M., Scanziani M., et al. Favourable survival of in-hospital compared to out-of-hospital refractory cardiac arrest patients treated with extracorporeal membrane oxygenation: an Italian tertiary care centre experience. Resuscitation. 2012;83:579–583. doi: 10.1016/j.resuscitation.2011.10.013. [ DOI ] [ PubMed ] [ Google Scholar ] 34. Avalli L., Maggioni E., Formica F., Redaelli G., Migliari M., Scanziani M., et al. Favourable survival of in-hospital compared to out-of-hospital refractory cardiac arrest patients treated with extracorporeal membrane oxygenation: an Italian tertiary care centre experience. Resuscitation. 2012;83:579–583. doi: 10.1016/j.resuscitation.2011.10.013. [ DOI ] [ PubMed ] [ Google Scholar ] 35. Dezfulian C., Cabañas J.G., Buckley J.R., Cash R.E., Crowe R.P., Drennan I.R., et al. Part 4: Systems of Care: 2025 American Heart Association guidelines for cardiopulmonary resuscitation and emergency cardiovascular care. Circulation. 2025;152:S353–S384. doi: 10.1161/CIR.0000000000001378. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. D’Arrigo S., Cacciola S., Dennis M., Jung C., Kagawa E., Antonelli M., et al. Predictors of favourable outcome after in-hospital cardiac arrest treated with extracorporeal cardiopulmonary resuscitation: a systematic review and meta-analysis. Resuscitation. 2017;121:62–70. doi: 10.1016/j.resuscitation.2017.10.005. [ DOI ] [ PubMed ] [ Google Scholar ] 37. Yannopoulos D., Bartos J., Raveendran G., Walser E., Connett J., Murray T.A., et al. Advanced reperfusion strategies for patients with out-of-hospital cardiac arrest and refractory ventricular fibrillation (ARREST): a phase 2, single centre, open-label, randomised controlled trial. Lancet. 2020;396:1807–1816. doi: 10.1016/S0140-6736(20)32338-2. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 38. Jeong D., Tak Lee G., Eun Park J., Yeon Hwang S., Gun Shin T., Do Shin S., et al. Severity of metabolic derangement predicts survival after out-of-hospital cardiac arrest and the likelihood of benefiting from extracorporeal life support. Emergencias. 2024;36:408–416. doi: 10.55633/s3me/093.2024. [ DOI ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Supplementary Data 1 mmc1.docx (488.7KB, docx) Articles from Resuscitation Plus are provided here courtesy of Elsevier ACTIONS View on publisher site PDF (1.3 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top