Introduction

Cardiogenic shock remains the leading cause of mortality in patients with acute myocardial infarction (AMI). Temporary mechanical circulatory support (tMCS) devices are frequently employed to manage AMI complicated by cardiogenic shock. The intra-aortic balloon counterpulsation (IABP) is the most widely utilized tMCS1,2, though Impella (Abiomed, Danvers, MA, USA) and Extracorporeal membrane oxygenation (ECMO) are increasingly used in clinical practice.

AMI is known to increase the risk of both ischemic and hemorrhagic stroke3,4,5. However, limited data exist on the impact of tMCS device on stroke risk in patients with AMI. Our study addresses this gap by analyzing the National Inpatient Sample (NIS) to investigate: 1. Stroke rates: The incidence of overall, ischemic, and hemorrhagic stroke among AMI hospitalizations treated with different devices (IABP, Impella, ECMO) compared to those managed without tMCS. 2. Temporal trends: changes in the incidence of overall, ischemic, and hemorrhagic stroke in AMI hospitalizations over time. 3. Predictors of stroke: Independent predictor of overall, ischemic, and hemorrhagic stroke in AMI hospitalizations. 4. subgroups risks: comparative risks of overall, ischemic, and hemorrhagic stroke in subgroup populations treated with different tMCS devices compared to those managed without tMCS.

The findings of this study will clarify the clinical implications of tMCS device use on stroke risk in AMI patients, addressing a critical knowledge gap in this field.

Methods

Data source

The data for this study were sourced from the NIS6, the largest publicly available all-payer administrative claim-based database in the United States. The study period span time from the implementation of the International Classification of Diseases, 10th Revision Clinical Modification Codes (ICD-10-CM) in the 4th quarter of 2015 through December 2020. Developed by the Agency for Healthcare Research and Quality (AHRQ) as part of Healthcare Cost and Utilization Project (HCUP), the NIS contains de-identified data from a 20% stratified random sample of U.S. inpatient hospitalization. discharge weight provided in the dataset were used to generate national estimates for all hospitalizations.

Study population

We conducted a retrospective analysis of ST-segment elevation myocardial infarction (STEMI) and non-ST-segment elevation myocardial infarction (NSTEMI) hospitalizations using data from the NIS between October 1, 2015, and December 31, 2020. Hospitalizations were identified using ICD-10-CM codes for diagnosis and procedure, which are listed in the Supplementary Table 1. The code used to identify hospitalizations of acute ischemic stroke and hemorrhage stroke has been validated by prior studies7,8. Medical comorbidities were identified using the Elixhauser comorbidity index refined for the ICD-10-CM9. Hospitalizations were stratified into 4 mutually exclusive cohorts based on tMCS devices utilizations: 1. no tMCS use: Managed with medical therapy alone, 2. IABP only: IABP as the sole tMCS device, 3. Impella only: Impella as the sole tMCS device, 4. ECMO with/without other tMCS: ECMO, irrespective of concurrent IABP or Impella use.

Statistical analysis

Categorical variables were compared using chi-square (χ2) tests, while non-normally distributed continuous variables were analyzed with the Wilcoxon rank-sum test6. To evaluate the association between temporary mechanical circulatory support (tMCS) device utilization and stroke risk, univariable Cox regression was performed for ischemic and hemorrhagic stroke as separate outcomes. Variables with a p-value < 0.10 in univariable analysis were included in a stepwise forward multivariable Cox regression model to identify independent predictors of stroke. The entry threshold for the stepwise model was set at p < 0.10, prioritizing statistical significance over hypothesized biological mechanisms to minimize selection bias.

The multivariable Cox regression model adjusted for the following covariates: age, sex, race, hypertension, diabetes mellitus, obesity, smoking status, atrial fibrillation, prior stroke, prior PCI, prior myocardial infarction, prior coronary artery bypass grafting, chronic lung disease, peripheral arterial disease, family history of coronary artery disease, hypothyroidism, hospital size, hospital location and teaching status, primary insurance, IABP only use, Impella only use, ECMO use.

Cochran-Armitage test was used to assess trends in stroke incidence over time. Subgroup analyses were performed to estimate the odd ratios (ORs) of overall, ischemic, and hemorrhagic stroke across predefined subgroups, with results visualized as forest plots (reported as Ors and 95% confidence interval [CIs]). Statistical significance was defined as a two-tailed p-value < 0.05. All analyses were performed using SAS 9.4 (SAS Institute, Cary, NC).

Ethical considerations

As NIS contains de-identified patient data, institutional review board (IRB) approval and informed consent were waived. All analyses adhered to the Healthcare Cost and Utilization Project (HCUP) data use agreement. The authors vouch for the accuracy and completeness of the data and analyses.

Results

Study flow chart and hospitalization characteristics

The study flow chart, depicted in Fig. 1, illustrates the process of selecting hospitalizations for analysis. Initially, a total of 974,216 AMI hospitalizations were extracted between the 4th quarter of 2015 and 2020 NIS database. After excluding 54,517 AMI hospitalizations with elective admission, 6274 with COVID-19 infection, 1296 with both IABP and Impella used but without ECMO, and 463 unspecified tMCS device usage, a final cohort of 911,666 hospitalizations were enrolled. After applying the discharge weight provided by the NIS database, the final cohort of the study included a total of 4,370,069 hospitalizations with medical treatment only, 136,005 with IABP only, 41,560 with Impella only, and 10,695 with ECMO used during the hospitalization. A proportional Venn graph illustrating the proportion of hospitalizations with different modes of tMCS is presented in Supplementary Fig. 1.

Fig. 1
figure 1

Study flow chart. COVID-19 coronavirus disease 2019, ECMO extracorporeal membrane oxygenation, IABP intra-aortic balloon pump, ICD international classification of diseases, NIS national inpatient sample, STEMI ST segment elevation myocardial infarction.

Full size image

Table 1 presents the baseline characteristics of AMI hospitalization stratified by tMCS devices utilization. Overall, hospitalizations treated with tMCS are more likely to occur in large urban teaching hospitals. Significant differences were observed in management strategies, including angiography, percutaneous coronary intervention, and coronary artery bypass grafting. The median age of hospitalization treated with IABP only or Impella only was comparable to those receiving medical therapy alone, However, hospitalizations treated with ECMO involved younger patients, tMCS-treated hospitalizations were also more likely involve male patients and present with STEMI. These hospitalizations had higher rates of cardiogenic shock and cardiac arrest, particularly among ECMO-treated patients. ECMO-treated hospitalizations exhibited fewer comorbidities, such as hypertension, diabetes mellitus, chronic lung disease, hypothyroidism, and dementia. They are also less likely to have a history of smoking, prior MI, prior coronary artery bypass surgery (CABG), or prior stroke.

Table 1 Baseline characteristics of AMI hospitalizations.
Full size table

Stroke rates and trends in study groups

Table 2 presents the stroke rates across different treatment groups. The overall stroke rates were 3.41% for hospitalizations treated without tMCS, 3.46% for those with IABP only, 4.51% for the impella only group, and 13.34% for the ECMO group, respectively. Specifically, the ischemic stroke rates for these groups were 2.93% (no tMCS), 3.17% (IABP only), 3.96% (Impella only), and 9.91% (ECMO), respectively. The hemorrhagic stroke rates were 0.69%, 0.59%, 0.87%, and 4.77% for the same groups. Table 2 also includes additional in-hospital outcomes analyzed in the study.

Table 2 In-hospital outcomes in tMCS treated AMI hospitalizations.
Full size table

Figure 2 reports the quarterly rates of overall stroke, ischemic stroke, and hemorrhagic stroke for each group. The Impella-only group demonstrated an increasing trend in both overall and ischemic stroke rates, whereas rates remained stable in all other groups throughout the study period. Notably, the ECMO group sustained a strikingly elevated stroke risk, which was consistent across the entire study period for all stroke types (overall, ischemic, and hemorrhagic).

Fig. 2
figure 2

Annual rate of overall, ischemic, and hemorrhagic stroke. ECMO extracorporeal membrane oxygenation, IABP intra-aortic balloon pump.

Full size image

Predictor of stroke

Univariate Cox regression analysis identified ECMO as the strongest predictor of overall stroke (OR 4.25, 95% CI [3.75–4.83]). Other significant predictors included: age > 75 years (vs. < 50 years; OR 1.62, 95%CI [1.54–1.69]), Atrial fibrillation (OR 1.59, 95% CI [1.56–1.63]), Female sex (OR 1.41, 95% CI [1.38–1.44]), Impella use (OR 1.35, 95% CI [1.21–1.49]), and prior history of stroke (OR 1.32, 95%CI [1.28–1.38]). In the stepwise forward multivariate Cox regression, after adjustment, ECMO use remained strongly associated with stroke (aOR 3.04, 95% CI [2.66–3.48]), followed by Impella use (aOR 1.79, 95% CI [1.61–2.00]), and atrial fibrillation (aOR 1.34, 95%CI [1.31–1.38]). Notably, IABP use showed no significant association with stroke in either univariate or multivariate regressions analyses (Table 3). Additional details on ischemic and hemorrhagic stroke regression analyses are provided in Supplementary Tables 2 and 3 respectively.

Table 3 Univariate and multivariate regression for the risk of overall stroke.
Full size table

Subgroup analysis

The findings demonstrate a strong association between ECMO use and a significantly elevated risk of overall, ischemic, and hemorrhagic stroke. Subgroup analyses were conducted to evaluate stroke risk across specific populations, including: gender, age, race, hypertension, diabetic mellitus, atrial fibrillation, and STEMI presentations. The analysis revealed that hospitalization with age younger than 50 years old, those without hypertension, and those presented with STEMI are at particularly high risk of stroke, as shown in Fig. 3 (p for interaction < 0.001 for all). Subgroup analysis to assess the risk of ischemic and hemorrhagic stroke is shown in Supplementary Figures S2 and S3. Subgroup analysis for hospitalizations with only Impella use and only IABP use are shown in Supplementary Figures S4–S9.

Fig. 3
figure 3

Forest plot for subgroup analysis of overall stroke risk treated with ECMO to those without tMCS. ECMO extracorporeal membrane oxygenation, OR odds ratio, STEMI ST segment elevation myocardial infarction, tMCS temporal mechanical circulatory support.

Full size image

Discussion

To our best knowledge, this represents the first large-scale study employing robust analytical methods to evaluate stroke risk across different tMCS devices used in AMI hospitalizations. The key findings include: 1. IABP use alone was not associated with an increased risk of overall, ischemic, and hemorrhagic stroke; 2. Impella use alone demonstrated a modest but significant increase in the risk of overall, ischemic, and hemorrhagic stroke; 3. ECMO use associated with a markedly elevated risk of overall, ischemic, and hemorrhagic stroke; 4. Among ECMO-treated hospitalizations, those with age younger than 50 years old, absence of hypertension, or STEMI presentation had the highest stroke risk.

IABP remains the predominant tMCS modality for AMI hospitalizations, despite ongoing debate regarding its survival benefits10,11, However, our findings provide reassuring evidence that IABP use is not associated with an elevated stroke risk- a critical consideration in clinical decision-making.

The use of impella has increased substantially, particularly in the post-IABP-SHOCK II trial era10,11. However, initially, the presumed survival benefit from its superior hemodynamic support lacks robust clinical validation. This is underscored by the Dagmar et al. exploratory trial, which demonstrated that routine Impella use failed to reduce 30-day mortality compared to IABP12. The DanGer-Shock trial represents the first large-scale randomized controlled trial demonstrating improved outcomes with Impella support in STEMI patients complicated by cardiogenic shock. While these findings will likely promote broader adoption of Impella for acute AMI circulatory support, the associated increased stroke risk observed in Impella-treated patients necessitates further investigation. The Impella device’s larger profile and requirement for higher-caliber insertion catheters inherently elevate the risk of vascular complications. Moreover, its implantation necessitates traversal of both the aortic arch and aortic valve, a procedural characteristic that may predispose to increased cerebrovascular events. Consistent with the findings reported by Thakkar et al.13, our analysis confirmed a significantly higher stroke risk in Impella-treated hospitalizations compared to IABP therapy.

Existing data on ECMO use in AMI-related cardiogenic shock remain limited14,15,16,17. The recent ECLS-ECMO trial18 demonstrated that early routine ECMO failed to reduce 30-day mortality compared to standard medical therapy alone. Consistent with prior reports14,19, our finding confirm that ECMO is associated with a substantially elevated stroke risk, predominantly ischemic stroke20. Notable, a study of 153 VA-ECMO patients reported 8.4% acute ischemic stroke and 3.9% hemorrhagic stroke rates21, aligning closely with our observed stroke incidence.

The pathophysiological mechanisms underlying the increased stroke risk associated with ECMO are multifactorial and involve several processes: 1. Altered hemodynamics: Similar to Impella, ECMO generates non-physiological blood flow, which disrupts normal circulation patterns and promotes thrombus formation. 2. Device-induced coagulopathy: The foreign surfaces of ECMO circuit (e.g. gas exchange membranes, hollow fibers) activate the coagulation cascade, increasing the risk of clot formation. 3. Hemolysis: Shear stress from blood flow through the oxygenator causes red blood cell destruction releasing free hemoglobin and other prothrombotic factors, even subclinical hemolysis may elevate stroke risk22. 4. Left ventricular complications: In VA-ECMO, retrograde aortic flow form peripheral cannulation increases LV afterload, leading to LV distention, blood stasis, intracardiac thrombus formation, these thrombi can embolized to the brain, causing stroke16.

Moreover, while anticoagulation is mandatory to prevent circuit thrombosis in Impella and ECMO patients, it simultaneously elevates the risk of hemorrhagic stroke. This creates a critical clinical dilemma—balancing thromboprophylaxis against bleeding complications remains one of the most challenging aspects of managing patients on these devices.

Given the devastating consequences of stroke in tMCS-supported patients, it is crucial that further research be conducted to better understand the risk factors, incidence, and prevention strategies associated with stroke in patients with Impella and ECMO use. In addition, advances in device technology should focus on minimizing the risk of stroke, this could involve the development of more biocompatible materials that reduce the formation of blood clots, as well as improving flow dynamic patterns. Moreover, optimizing anticoagulation protocols and monitoring strategies could help strike a balance between preventing clotting and minimizing bleeding complications.

Limitations

While this study provides important insights into stroke risk across tMCS modalities, several limitations warrant consideration: The administrative database lacked granular clinical data, such as hemodynamic, metabolite, medication, biochemistry, and imaging data, and there were no long-term follow-up data in the NIS dataset. Coding errors and underreporting of secondary diagnoses were also potential sources of bias. Additionally, as with all retrospective observational studies, even with robust confounder adjustment, residue confounding was inherent. Hemodynamic parameter is not available in the NIS dataset, and the indications for ECMO, IABP, and Impella may vary between hospitals; therefore, selection bias was inevitable. Moreover, more granular data about the severity of stroke and Rankin Scales were not available. However, the NIS database has been extensively validated in previous publications23,24. This study included the largest sample of hospitalization comparing the rate of overall, ischemic, and hemorrhagic stroke, robust analyses were performed by multivariable regression and subgroup analysis.

Conclusion

This study investigates the rates, trends, and predictors of ischemic and hemorrhagic stroke in AMI hospitalizations treated with various tMCS devices. By elucidating the association between tMCS device utilization and stroke risk, our findings can guide clinical decision-making by identifying high-risk patients or device-specific risks, enabling tailored therapeutic strategies, and inform future research by highlighting knowledge gaps in device safety profiles and mechanisms of stroke in tMCS-supported AMI patients.