Age-Dependent Benefit of ECPR in OHCA: An Interaction Analysis

Abstract

ECPR effectiveness in OHCA remains inconsistent, and age-related effect modification is not well defined. This retrospective study analyzed 470 adult patients (2011-2025), comparing ECPR (n=71) to CCPR (n=399). Using multivariable regression and inverse probability of treatment weighting (IPTW), we found that increasing age independently associated with higher mortality (aOR: 1.064). Also, a significant interaction between age and treatment strategy was identified (p<0.001). ECPR provided a substantial mortality reduction in younger patients, but this benefit progressively attenuated with age. These findings demonstrate that ECPR effectiveness is strongly age-dependent. Consequently, age should be considered as a continuous effect modifier rather than a rigid eligibility threshold for ECPR selection.

INTRODUCTION

Out-of-hospital cardiac arrest (OHCA) remains a major global public health challenge and is associated with persistently high mortality despite advances in resuscitation care.1 Survival after OHCA varies widely and is influenced not only by patient characteristics but also by the organization of systems of care.2 The implementation of coordinated strategies, including specialized cardiac arrest centers and multidisciplinary post–cardiac arrest management, has been associated with substantial improvements in survival, with reported increases from approximately 21% to nearly 40% in selected systems, underscoring the importance of system-level interventions beyond conventional resuscitation techniques.3,4,5 Extracorporeal cardiopulmonary resuscitation (ECPR) has emerged as an advanced rescue strategy for patients with refractory cardiac arrest who fail to achieve sustained return of spontaneous circulation with conventional cardiopulmonary resuscitation (CCPR).6

By providing temporary mechanical circulatory and respiratory support through extracorporeal membrane oxygenation (ECMO), ECPR maintains systemic perfusion during ongoing resuscitation and serves as a physiological bridge that allows time for identification and treatment of potentially reversible causes of cardiac arrest.7 Observational studies and meta-analyses have suggested that ECPR may be associated with improved survival compared with CCPR in selected patient populations.8 However, the effectiveness of ECPR in out-of-hospital cardiac arrest remains inconsistent across clinical settings. While favorable outcomes have been reported from experienced centers and registry-based studies, randomized controlled trials have demonstrated more modest benefits, raising concerns regarding the generalizability to routine practice.9 Patient selection has therefore emerged as a critical determinant of ECPR effectiveness. Among candidate selection factors, age is frequently incorporated into institutional eligibility criteria and may fundamentally modify the balance between survival benefit and physiological or neurological resilience following prolonged resuscitation.10 Despite its clinical importance, the extent to which age modifies the association between ECPR and outcomes in real-world OHCA populations remains insufficiently defined.11 Accordingly, this study aimed to evaluate the real-world effectiveness of ECPR in patients with non-traumatic OHCA treated at a single tertiary medical center. Using a retrospective cohort design with multivariable adjustment and inverse probability of treatment weighting, we compared in-hospital mortality between patients receiving ECPR and those managed with CCPR. We further examined whether patient age modified the association between resuscitation strategy and outcome, with the goal of providing clinically relevant insights into patient selection and resource allocation for ECPR in routine practice.

MATERIALS AND METHODS

1. Study population

We conducted a retrospective analysis of medical records from Kaohsiung Armed Forces General Hospital, a local hospital in Taiwan, including patients aged 20 years or older who received CPR for OHCA in the Emergency Department (ED) from 2011 to 2025. Patients with do-not-resuscitate (DNR) orders or traumatic injuries were excluded, resulting in a final study cohort of 470 individuals (Fig. 1). The exposure of interest was the administration of ECMO during the ED phase, while the primary outcome was in-hospital mortality (death before hospital discharge). All patients received standard CPR protocols based on American Heart Association guidelines. Of the total cohort, 71 patients were treated with ECMO. Age, incidence of witnessed cardiac arrest, location of cardiac arrest, and baseline clinical condition as determined by emergency department doctors and nurses were among the patient data gathered from our hospital’s database.

FIG. 1. Flow diagram of patient selection and study cohort. Because the dataset was extracted from an institutional database in which eligibility criteria had already been applied, the number of patients excluded at each step was not available.

2. ECPR indications

At our hospital, approximately ten minutes after advanced cardiac life support is started, cardiac surgeons and the patient’s visiting emergency doctor together decide whether to start ECPR. Criteria in favor of ECPR use include: patients under 55 years old without major trauma or obvious body rigidity and livor mortis are always considered for ECMO, while patients aged 55-80 are assessed on a case-by-case basis. Acute intracranial hemorrhage or infarct, uncontrollable bleeding, terminal stage malignancy, ventilator dependency for more than a month, major trauma, a do-not-resuscitate order, and patients with preexisting multiorgan dysfunction are not considered suitable candidates for ECPR. On the other hand, patients aged >80 years were not considered candidates for ECPR. All ECMO procedures are performed by qualified cardiac surgeons.

3. Statistical analysis

Data were analyzed using SPSS software (IBM Inc, Armonk, NY, USA), and this retrospective study focused on patients with non-traumatic out-of-hospital cardiac arrest. The association between the use of ECPR and clinical outcomes was evaluated using logistic regression models.

4. Outcome definitions

For regression analyses, in-hospital mortality was coded as death (1) and survival (0). Neurological outcomes, such as cerebral performance category, were not included in the primary analysis due to incomplete documentation and heterogeneity in post-resuscitation neurological assessment across the study period.

5. Covariates and confounder selection

Covariates included in multivariable analyses were selected a priori based on clinical relevance and existing literature. These included age, sex, and underlying etiology of cardiac arrest. These variables were chosen because of their established associations with survival after outof-hospital cardiac arrest and their potential influence on the clinical decision to initiate extracorporeal cardiopulmonary resuscitation.

6. Propensity score estimation and inverse probability of treatment weighting

To account for baseline imbalances between treatment groups, propensity scores representing the probability of receiving extracorporeal cardiopulmonary resuscitation were estimated using a logistic regression model incorporating age, sex, and underlying etiology. Inverse probability of treatment weighting was applied to generate a weighted pseudo population in which measured covariates were balanced between groups. Covariate balance before and after weighting was assessed using standardized mean differences, with values less than 0.1 considered indicative of adequate balance. To ensure the validity of the IPTW model, we assessed the common support of propensity scores between groups via visual inspection. Stabilized weights were applied to the IPTW model to minimize the potential influence of extreme weights caused by clinical selection patterns, such as the inherent preference for ECPR in younger, healthier patients.

7. Interaction analysis

To evaluate potential effect modification by age, an interaction term between treatment group and age was prespecified and included in the inverse probability of treatment weighted logistic regression model.

8. Ethical considerations

This study was approved by the Institutional Review Board of Kaohsiung Armed Forces General Hospital. The requirement for informed consent was waived due to the retrospective nature of the study.

RESULTS

1. Characteristics of patients at baseline

A total of 470 patients with out-of-hospital cardiac arrest were included in the analysis, of whom 71 received ECPR and 399 received CCPR. At baseline, patients in the CCPR group were significantly older than those in the ECPR group (mean age 65.6±15.6 vs. 47.6±13.0 years, p<0.001). The ECPR group also had a higher proportion of male patients (87.3% vs. 66.2%, p<0.001). Underlying etiology differed markedly between groups (p<0.001), with cardiac etiology predominating in the ECPR group, whereas infectious and respiratory etiologies were more frequent in the CCPR group (Table 1). Substantial baseline imbalance between treatment groups was observed, as reflected by large standardized mean differences for several covariates.

TABLE 1. Patient clinical characteristics of the ECPR and CCPR groups.

2. Multivariable logistic regression analysis

In the multivariable logistic regression analysis adjusting for age, sex, and underlying etiology, treatment group was independently associated with in-hospital mortality (Table 2).

TABLE 2. Multivariable adjusted odds ratios for in-hospital mortality.

Using ECPR as the reference group, patients treated with CCPR had a significantly lower risk of in-hospital mortality (adjusted odds ratio [aOR], 0.025; 95% confidence interval [CI], 0.015-0.040; p<0.001). This observation is likely attributable to the fact that the CCPR group includes patients who achieved early return of spontaneous circulation (ROSC) without requiring ECPR, whereas the ECPR group exclusively comprised patients with refractory cardiac arrest. Therefore, our analysis focuses on the age-treatment interaction to better characterize ECPR’s effectiveness across the age spectrum. Increasing age was associated with higher mortality (aOR per year, 1.064; 95% CI, 1.061-1.067; p<0.001). Male sex was associated with lower mortality risk (aOR, 0.654; 95% CI, 0.609-0.701; p<0.001). Compared with cardiac etiology, respiratory, infectious, and other etiologies were all independently associated with increased mortality risk.

3. IPTW-adjusted sensitivity analysis

As a sensitivity analysis, inverse probability of treatment weighting (IPTW) based on propensity scores was applied to account for baseline imbalances between treatment groups. In the IPTW-adjusted logistic regression model, ECPR (compared with CCPR) was significantly associated with in-hospital mortality, with an odds ratio of 1.940 (95% CI, 1.261-2.985; p=0.003), indicating that the association between resuscitation strategy and outcome was robust across analytical approaches.

4. Interaction between age and treatment strategy

To evaluate potential effect modification by age, an interaction term between treatment group and age was included in the IPTW-adjusted logistic regression model.

A significant interaction between age and treatment group was observed (p for interaction <0.001), indicating that the effect of ECPR on outcome varied significantly by age. Specifically, for each additional year of age, the odds ratio for ECPR relative to CCPR increased by a factor of 1.088, suggesting an age-dependent modification of treatment effect.

As illustrated in Fig. 2, the adjusted odds ratio for ECPR versus CCPR increased progressively with advancing age, from 0.03 at age 40, to 0.10 at age 55, and 0.35 at age 70, indicating a diminishing relative benefit of ECPR among older patients.

FIG. 2. Adjusted predicted probability of in-hospital mortality according to age by treatment group. Estimates were derived from an inverse probability of treatment weighted (IPTW) logistic regression model including a treatment-by-age interaction term and adjusted for sex and underlying etiology. The curves illustrate a progressive attenuation of the relative survival benefit associated with extracorporeal cardiopulmonary resuscitation with increasing age.

DISCUSSION

In this cohort of 470 patients with out-of-hospital cardiac arrest, the resuscitation strategy, namely extracorporeal cardiopulmonary resuscitation with 71 patients versus conventional cardiopulmonary resuscitation with 399 patients, emerged as a critical independent predictor of in hospital mortality. With mortality defined as the primary outcome coded as 1 equals death, our multivariable analysis demonstrated that although conventional cardiopulmonary resuscitation appeared to be associated with a lower mortality risk in the overall cohort, with an adjusted odds ratio of 0.025 and a 95 percent confidence interval of 0.015 to 0.040, this relationship was profoundly modified by a highly significant interaction between age and treatment strategy, with a p-value less than 0.001.

In real-world practice, candidate selection for ECPR is inherently non-random, often favoring younger patients with cardiac etiologies. While this results in significant baseline imbalances, it reflects contemporary resuscitation protocols and ethical considerations. Our study utilizes interaction modeling to move beyond simple group comparisons, successfully quantifying how the marginal benefit of ECPR is modified by physiological aging regardless of initial selection bias.

Our data clearly indicate that the survival benefit of extracorporeal cardiopulmonary resuscitation is strongly age dependent. In younger patients, for example at 40 years of age, extracorporeal cardiopulmonary resuscitation was associated with a remarkably low mortality risk, with an adjusted odds ratio of 0.03. However, this protective effect diminished substantially with advancing age, reaching an adjusted odds ratio of 0.35 by 70 years of age. These findings are consistent with the per protocol analysis of the INCEPTION trial published in 2024, which demonstrated that extracorporeal cardiopulmonary resuscitation significantly reduced 30 day mortality, with an adjusted odds ratio of 2.41 for survival, when strict adherence to inclusion criteria was maintained.12,13

The baseline characteristics of our cohort reflect well recognized clinical selection patterns. Patients receiving extracorporeal cardiopulmonary resuscitation were significantly younger, with a mean age of 47.6 years compared with 65.6 years in the conventional cardiopulmonary resuscitation group, and were predominantly male, accounting for 87.3 percent of cases. This male predominance closely mirrors large scale multicenter registries such as SAVE J II, where males comprised approximately 82 percent of the cohort.14,15 Such distributions likely reflect epidemiological realities, as males experience two to three times higher rates of out-of-hospital cardiac arrest due to cardiac etiologies and shockable rhythms, which is consistent with recommendations from the American Heart Association and ELSO 2025 guidelines.16,17

Although ethical discussions have raised concerns regarding a potential, implicit bias favoring perceived physical robustness in males when assigning invasive therapies, our adjusted analyses suggest that treatment selection in our cohort largely reflects the higher ischemic burden and cardiac arrest profiles observed in this demographic.18 Moreover, the predominance of cardiac etiologies in the extracorporeal cardiopulmonary resuscitation group aligns with the ideal candidate characteristics endorsed by the European Resuscitation Council 2025 guidelines, further supporting the external validity of our findings.19

Given the substantial baseline imbalances between treatment groups, as evidenced by large standardized mean differences, we employed inverse probability of treatment weighting to approximate a randomized treatment allocation. After weighting, extracorporeal cardiopulmonary resuscitation remained significantly associated with mortality outcomes, with an odds ratio of 1.940, a 95 percent confidence interval of 1.261 to 2.985, and a p-value of 0.003.

Methodological studies in extracorporeal cardiopulmonary resuscitation research have increasingly advocated for inverse probability of treatment weighting over propensity score matching, as it preserves the full sample size and more effectively addresses confounding by indication, wherein clinicians preferentially assign extracorporeal cardiopulmonary resuscitation to patients perceived to have more favorable prognoses. By generating a pseudo population that more closely resembles randomization, our weighted model offers a more precise estimate of the treatment effect and bridges the gap between real world observational data and trends observed in randomized controlled trials.

The interaction analysis revealed a clear age dependent modification of treatment effect, with the mortality odds ratio of extracorporeal cardiopulmonary resuscitation relative to conventional cardiopulmonary resuscitation increasing by a factor of 1.088 for each additional year of age. Beyond circuit related complications, this progressive attenuation appears to be driven largely by the heightened vulnerability of the aging brain to ischemic injury.20

Aging is associated with lower hypoxic thresholds for blood brain barrier disruption, resulting in four to six fold increases in fibrinogen extravasation and microglial activation following ischemia.21 In elderly patients, downregulation of tight junction proteins such as claudin 5 and ZO 1 further exacerbates post arrest vasogenic edema, a process that restoration of macro perfusion through extracorporeal cardiopulmonary resuscitation cannot adequately counterbalance.21,22 Reduced ischemic preconditioning capacity in the aged brain amplifies reperfusion injury and closely parallels the observed decline in mortality reduction as patients approach 70 years of age.23

Importantly, the interaction suggests that aging functions as a force multiplier for ischemic damage. While younger brains may tolerate a limited duration of low flow during extracorporeal cardiopulmonary resuscitation initiation, the aged brain is more prone to rapid and irreversible microvascular collapse and mitochondrial failure.24 This biological susceptibility provides a plausible explanation for the marked attenuation of extracorporeal cardiopulmonary resuscitation’s protective effect from an adjusted odds ratio of 0.03 at 40 years of age to 0.35 at 70 years of age.

The progressive increase in the mortality adjusted odds ratio from 0.03 at age 40 to 0.35 at age 70 carries important implications for clinical decision making and resource allocation. Although an adjusted odds ratio of 0.35 still suggests a mortality reduction with extracorporeal cardiopulmonary resuscitation at 70 years of age, emerging data from the JAAM OHCA Registry published in 2025 highlight a growing neurological disconnect in elderly patients.25,26

In individuals aged 75 years or older, survival to hospital discharge may approach that of younger cohorts, yet favorable neurological outcomes defined as cerebral performance category 1 or 2 are exceedingly rare. This survival with severe neurological impairment, often categorized as cerebral performance category 3 to 5, is frequently driven by age exacerbated hypoxic ischemic encephalopathy.27 Our findings therefore underscore the ethical challenge of survival at any cost. As the adjusted odds ratio approaches unity with advancing age, extracorporeal cardiopulmonary resuscitation in elderly patients may primarily delay death or result in prolonged states of impaired consciousness, including persistent vegetative states.

These observations are consistent with recent reports indicating that neurological recovery plateaus in elderly patients despite successful the return of spontaneous circulation.25 Consequently, the cost effectiveness of extracorporeal cardiopulmonary resuscitation in older populations appears to be less favorable, as survival gains increasingly fail to translate into meaningful quality-adjusted life years.22 This reinforces the need to integrate frailty and neurological resilience assessments into real time extracorporeal cardiopulmonary resuscitation decision making.

Our multivariable model further identified respiratory and infectious etiologies as independent predictors of higher mortality. From a pathophysiological perspective, non-cardiac, out-of-hospital cardiac arrest is frequently accompanied by a profound systemic inflammatory response syndrome and extensive endothelial glycocalyx shedding. Although VA-ECMO effectively stabilized macro perfusion in cardiac arrest cases, it likely failed to adequately counteract the second hit inflammatory cascade and microcirculatory shunting characteristic of infectious or respiratory arrests.28

Persistent hyperlactatemia observed after extracorporeal cardiopulmonary resuscitation in these subsets is therefore more plausibly attributed to mitochondrial dysfunction rather than residual oxygen debt.29 This mechanistic distinction may explain why the mortality reduction associated with extracorporeal cardiopulmonary resuscitation appears to be largely confined to cardiac etiologies, in which definitive cause reversal and adjunctive neuroprotective strategies such as targeted temperature management are most effective.30

This study has several limitations. First, the generalizability of our findings may be restricted due to the single-center retrospective design. Clinical decision-making regarding ECPR may therefore reflect local expertise and practice patterns, even though standardized cardiopulmonary resuscitation and extracorporeal membrane oxygenation protocols were consistently applied throughout the study period. Second, owing to the retrospective nature of our database, complete information on initial cardiac rhythm was not available for all patients. In addition, detailed time-course variables, such as exact no-flow time, low-flow duration, and precise timing of ECMO initiation, could not be comprehensively captured. These unmeasured temporal factors may have introduced residual confounding. Third, neurological outcomes beyond hospital discharge were not available. Although survival to discharge is a widely accepted and clinically meaningful endpoint in OHCA research, the lack of long-term neurological and functional follow-up limits interpretation regarding the quality of survival, particularly in elderly patients. Fourth, while we utilized IPTW with stabilized weights, unmeasured confounding by indication remains a possibility, as clinicians may select ECPR candidates based on perceived physiological robustness beyond recorded variables. Additionally, the lack of data on initial cardiac rhythm and exact low-flow duration is a constraint. However, the high statistical significance of the age-treatment interaction (p<0.001) suggests that chronological age serves as a dominant proxy for biological resilience, maintaining its predictive value across our cohort. Finally, as with all observational studies, unmeasured confounders related to clinician judgment and patient selection for ECPR may have influenced outcomes despite adjustment.

Despite these limitations, this study provides real-world evidence highlighting a clinically relevant age-dependent modification of ECPR effectiveness in OHCA. Future multicenter, prospective studies incorporating standardized resuscitation timelines and neurological outcome assessments are warranted.

In conclusion, this real-world cohort of non-traumatic, out-of-hospital cardiac arrest, the effectiveness of ECPR was found to be strongly age dependent. While ECPR was not associated with a universal mortality benefit across the entire adjusted population, a profound age–treatment interaction was identified (p for interaction <0.001). The survival advantage of ECPR was most pronounced in younger patients, but this marginal benefit attenuated progressively with advancing age. Younger patients derived a clear survival benefit from ECPR, whereas this benefit progressively attenuated with advancing age. These findings indicate that age should be considered a continuous effect modifier rather than a rigid eligibility threshold when selecting candidates for ECPR. Further prospective multicenter studies incorporating standardized resuscitation metrics and neurological outcomes are warranted.