Search Results (585)

Background: Acute brain injury (ABI) is a frequent complication of extracorporeal membrane oxygenation (ECMO), but early detection is limited by sedation, imaging constraints, and low sensitivity of conventional neuroimaging. We hypothesized that plasma metabolomics could identify ECMO-mode-specific metabolic shifts and biomarkers preceding ABI. Methods: Untargeted plasma metabolomics was performed in 70 participants across two centers: 30 healthy controls, 17 critically ill controls, and 23 ECMO patients [14 venovenous (VV) and 9 venoarterial (VA)]. Plasma was collected within 24 h and 7 days after cannulation. Fold-change analyses and partial least squares discriminant analysis were used to define metabolic differences and identify metabolites associated with subsequent ABI. Results: ABI occurred in seven ECMO patients, including five venoarterial and two venovenous ECMO patients. ECMO support was associated with broad alterations in circulating lipid metabolism, including changes in sphingomyelins, lysophospholipids, and monoacylglycerols. PLS-DA demonstrated metabolomic separation between ECMO patients and controls. Among ECMO patients, three structurally related glycerophospholipids—GPI (18:0/18:2), GPC (16:0/18:2), and GPE (16:0/18:2)—were significantly decreased before ABI diagnosis. ABI was also associated with broader reductions in phosphatidylethanolamines, phosphatidylinositols, lysophospholipids, and polyunsaturated fatty acids. Conclusions: Early reductions in membrane-associated phospholipids were associated with subsequent ABI during ECMO support, suggesting that alterations in circulating lipid homeostasis may identify neurological vulnerability before clinical or radiographic recognition of injury. Plasma metabolomics may provide a complementary approach for early neurological risk stratification and support future biomarker development. Full article

Background: Seasonal influenza is typically a self-limiting illness in healthy young adults, but severe and potentially life-threatening complications can occur. Influenza A infection may predispose patients to secondary bacterial infections, including Staphylococcus aureus pneumonia and bacteremia, which can rapidly progress to necrotizing pneumonia, acute respiratory distress syndrome (ARDS), and cardiovascular dysfunction. Case Presentation: We report the case of a young adult with influenza A(H3) infection complicated by Staphylococcus aureus bacteremia, necrotizing pneumonia, ARDS, and biventricular dysfunction. The patient experienced rapid cardiopulmonary deterioration requiring advanced critical care support, including extracorporeal membrane oxygenation (ECMO) and left ventricular unloading. Serial point-of-care ultrasound (POCUS) was central to clinical management, enabling early recognition of worsening respiratory and cardiac function and guiding timely escalation of supportive strategies. Conclusions: This case highlights the potential severity of influenza-associated complications even in young and previously healthy individuals. Repeated POCUS assessment can play a pivotal role in detecting rapid cardiopulmonary deterioration and guiding advanced interventions. Efficient hub-and-spoke organization may further support timely referral and management of critically ill patients requiring ECMO and specialized care. Full article

Background: Acute brain injury (ABI) is a major cause of mortality and morbidity during extracorporeal membrane oxygenation (ECMO), yet early diagnosis remains challenging because neuroimaging is often impractical in critically ill patients. We evaluated whether quantitative electroencephalography (qEEG) combined with machine learning could identify ABI and predict mortality in patients receiving ECMO. Methods: Consecutive adult ECMO patients who underwent a standardized neuromonitoring protocol with continuous EEG during sedation interruption were retrospectively analyzed. Quantitative EEG features and clinical variables were extracted and used to train multiple machine-learning classifiers with leave-one-subject-out cross-validation. Results: Fifty-seven patients were included (mean age 56 years; 54% male), including 41 supported with venoarterial ECMO, 15 with venovenous ECMO, and one with venoarterial-venous ECMO. ABI occurred in 21 patients (37%), of whom 70% had ischemic injury. Models incorporating qEEG achieved higher point estimates than those using clinical variables alone for ABI detection (best area under the curve (AUC) 0.769, 95% confidence interval (CI) 0.638–0.883, vs. 0.681), although the difference did not reach statistical significance. Frontal theta power and interhemispheric asymmetry were the EEG features most strongly associated with ABI. qEEG features also carried prognostic information for 30-day mortality (best AUC 0.864). Conclusions: Machine-learning analysis of continuous qEEG acquired during standardized sedation interruption may provide a noninvasive bedside approach for identifying ECMO patients at increased risk of ABI and short-term mortality and may help prioritize urgent neuroimaging and neurological intervention. Full article

Background: Extracorporeal membrane oxygenation (ECMO) is a life-saving organ support for severe circulatory or respiratory failure, but it is frequently complicated by bleeding. Pulmonary haemorrhage and consecutive airway obstruction are major contributors to morbidity and mortality. Flexible bronchoscopy (FB) is indispensable for diagnosing and managing pulmonary bleeding; however, its safety and efficacy in ECMO patients remain insufficiently studied. This analysis evaluated the feasibility, safety, and short-term physiological effects of FB in critically ill adults supported with ECMO. Methods: In this single-centre retrospective study, all adult ECMO patients who underwent FB for endobronchial bleeding or obstructive clot formation between January 2020 and December 2024 at the University Hospital Zurich were included. The procedural indications, diagnostic findings, haemostatic interventions, anticoagulation management, and oxygenation parameters before and after FB were analysed. The bleeding severity was graded according to the level of intervention required for haemostasis. Results: Sixteen patients (median age: 51.5 years, 69% male) underwent 34 FB procedures while on ECMO (69% veno-arterial). Most procedures were therapeutic (71%) and performed for haemorrhage (35%) or airway obstruction (21%). Clot removal was attempted in 85% of bronchoscopies, achieving partial or complete clearance in most cases. The median PaO₂/FiO₂ ratio improved from 142.8 mmHg to 161.4 mmHg (post/pre ratio: 1.36); requiring a lower FiO2 (p = 0.049). Of the procedures performed for active bleeding, 24% were classified as severe, and bronchial blockers were required in 26%. Bleeding control failed in 12% of cases. The ICU mortality was 62.5%, with pulmonary haemorrhage directly causing death in 25%. Conclusions: FB is feasible in ECMO patients and provides a measurable short-term improvement in oxygenation. However, the high mortality underscores the severity of the underlying disease. Repeat FB likely reflects the disease burden rather than procedural inefficacy. Full article

Background/Objectives: Femoro-femoral veno-arterial extracorporeal membrane oxygenation (VA-ECMO) is associated with limb ischemic complications. Routine near-infrared spectroscopy (NIRS) monitoring and a more standardized distal perfusion strategy were introduced at our institution in 2016. We compared ischemia-related interventions, amputations, and early mortality between the treatment eras. Methods: Consecutive patients undergoing femoro-femoral VA-ECMO from 2012 to 2024 were analyzed retrospectively and stratified by VA-ECMO initiation before 2016 or from 2016 onward. The primary endpoint was a peripheral ischemic vascular complication requiring surgical or interventional therapy; an inclusive sensitivity definition additionally incorporated four clinically plausible but less certain events. Period-specific rates were estimated using Poisson regression adjusted for baseline distal perfusion cannula (DPC) status and extracorporeal cardiopulmonary resuscitation (eCPR). Results: The analytic cohort comprised 270 patients (45 pre-2016 and 225 post-2016). The primary endpoint occurred in 50 patients (18.5%, 95% confidence interval (CI) 14.1–23.7): 3/45 (6.7%, 95% CI 1.4–18.3) before 2016 and 47/225 (20.9%, 95% CI 15.8–26.8) thereafter. Adjusted rates were 4.6% (95% CI 1.0–20.5) and 18.9% (95% CI 13.6–26.3), with an adjusted risk ratio of 4.11 (95% CI 1.06–17.93). The inclusive sensitivity analysis yielded similar estimates (adjusted risk ratio 3.31, 95% CI 1.05–11.47). Five patients underwent lower-limb amputation (1/45 pre-2016 and 4/225 post-2016). Kaplan–Meier 30-day mortality estimates were 56.8% overall (95% CI 50.8–62.9), 58.9% pre-2016 (95% CI 44.9–73.5), and 56.4% post-2016 (95% CI 49.9–63.1). Conclusions: Ischemia-related interventions were more frequently recorded after 2016, while amputations remained rare and 30-day mortality was similar. Because monitoring, DPC practice, and other aspects of care changed concurrently, and NIRS was used without a standardized trigger algorithm, the reasons for the observed era difference cannot be determined. The findings are exploratory and do not establish causal effects of NIRS or DPC use. Full article

Background: Isavuconazole, a broad-spectrum triazole antifungal, exhibits high lipophilicity and extensive plasma protein binding, properties that may predispose it to sequestration within extracorporeal membrane oxygenation (ECMO) circuits. This study aimed to characterize the pharmacokinetics (PK) of isavuconazole and to evaluate drug sequestration within the ECMO circuit in critically ill patients receiving ECMO support. Methods: We conducted a prospective, exploratory, single-center observational study including critically ill patients receiving ECMO (veno-venous (VV) or veno-arterial (VA)) and treated with intravenous isavuconazole. Serial blood samples were collected simultaneously from the patient’s arterial line and from pre- and post-membrane oxygenator sampling sites. Non-compartmental analysis was performed on arterial line samples to estimate PK measures, and concentration differences across sampling sites were analyzed to estimate circuit-related drug loss. PK/pharmacodynamic (PD) target attainment was assessed using established efficacy thresholds (AUC_0–24/MIC ≥ 25 and Cmin > 2 mg/L). Results: A total of 41 plasma samples from 3 critically ill patients (2 VV-ECMO, 1 VA-ECMO) were included in the analysis. Limited, component-specific isavuconazole loss was observed in tubing and connectors (6.62% ± 20.8%, p = 0.294) and across the entire ECMO circuit (7.74% ± 20.2%, p = 0.211). Likewise, no relevant concentration difference was detected across the membrane oxygenator (0.849% ± 6.15%, p = 0.642). Interindividual variability was observed across PK parameters, particularly in measures of elimination and distribution. All patients achieved predefined PK/PD efficacy targets, with mean Cmin and AUC_0–24/MIC of 3.07 ± 0.261 mg/L and 85.3 ± 4.03, respectively, and none exceeded the established toxicity threshold. Conclusions: Preliminary results showed variable concentration differences across ECMO sampling sites, with no consistent pattern of isavuconazole loss across the ECMO circuit under the conditions evaluated. All patients achieved predefined PK/PD efficacy targets using currently recommended dosing regimens; however, interindividual PK variability was observed, supporting the potential value of therapeutic drug monitoring (TDM) to guide individualized dosing decisions in this population. Larger population PK studies are warranted to further characterize determinants of isavuconazole exposure during ECMO support and refine evidence-based dosing strategies. Full article

Background: Sex-based differences in complications after venoarterial extracorporeal membrane oxygenation (VA-ECMO) for cardiogenic shock are not well defined. We compared 30-day coded acute kidney injury and other short-term outcomes between female and male patients receiving ECMO. Methods: We performed a retrospective multicenter cohort study using the TriNetX U.S. Collaborative Network from 2012 through 2025. Adults with cardiogenic shock supported with VA-ECMO were identified. Patients with a diagnosis-coded AKI (ICD-10-CM N17) recorded on or before the index ECMO procedure were excluded, and female and male cohorts were then matched 1:1 by propensity score on 23 characteristics. The primary endpoint was diagnosis-coded AKI between day 1 and day 30 after ECMO initiation. Secondary endpoints were all-cause mortality, newly diagnosis-coded sepsis, and newly diagnosis-coded ischemic stroke. Results: Among 11,229 adults meeting cohort criteria, 3773 were women, and 7456 were men. After exclusion of 8272 patients with previously coded AKI, 1152 women and 1805 men were eligible, and 1100 patients were matched in each group. Diagnosis-coded AKI occurred in 222 women (20.2%) and 276 men (25.1%) (risk ratio, 0.80; 95% confidence interval [CI], 0.69–0.94; hazard ratio [HR], 0.78; 95% CI, 0.65–0.93; p = 0.005). All-cause mortality was identical between groups (29.5% vs. 29.5%; risk ratio, 1.00; 95% CI, 0.88–1.14). Newly coded sepsis (5.1% vs. 6.3%) and newly coded ischemic stroke (4.0% vs. 3.2%) did not differ significantly. In an unadjusted Aalen–Johansen analysis performed in the unmatched eligible cohorts, the 30-day cumulative incidence of coded AKI was 21.6% among women and 27.8% among men. Conclusions: In this propensity-matched federated electronic health record cohort of adults with cardiogenic shock receiving VA-ECMO and without previously coded AKI, recorded female sex was associated with a lower 30-day risk of diagnosis-coded AKI. Mortality, newly coded sepsis, and newly coded ischemic stroke were similar. These findings are hypothesis-generating and support further investigation of sex-associated differences in datasets with granular renal and ECMO-specific variables. Full article

Background: Early extubation after bilateral lung transplantation (LT) may reduce intensive care unit (ICU) complications, but evidence from heterogeneous real-world cohorts and of its impact on mid-term functional recovery remains limited. Methods: We conducted a single-centre observational study of 149 consecutive adult bilateral LT recipients (February 2016–February 2023). Patients extubated within 24 h were assigned to the early extubation (EE) group (n = 63, 42%) (extubated within 24 h of the end of surgery), while those extubated later comprised the late extubation (LE) group (n = 86, 58%) (extubated beyond 24 h). Multivariable logistic regression identified predictors of late extubation. Outcomes included postoperative extracorporeal membrane oxygenation (ECMO), pneumonia, ICU length of stay, and spirometric parameters at 9–12 months after LT. Results: Higher Lung Allocation Score (LAS; adjusted OR 1.19, 95% CI 1.02–1.38) and intraoperative red blood cell (RBC) transfusions (adjusted OR 1.47, 95% CI 1.04–2.06) independently predicted late extubation. Compared with the LE group, EE recipients required less postoperative ECMO (2% vs. 23%; p = 0.008), had shorter inhaled nitric oxide treatment (7 vs. 17 h; p = 0.006), lower pneumonia rates (8% vs. 23%; p = 0.043), and shorter ICU stays (6 vs. 9 days; p = 0.005). In-hospital and 1-year mortality were similar between groups. At 9 ± 1 months, EE recipients showed better volumetric lung recovery, with higher FVC as a percentage of pre-transplant baseline (78.0% vs. 69.5%; p = 0.048) and higher TLC percentage predicted (77% vs. 68%; p = 0.015). Airflow indices and respiratory muscle strength did not differ. Conclusions: In a broadly inclusive LT cohort, higher LAS and intraoperative RBC transfusion independently predicted late extubation. Early extubation was associated with lower postoperative support requirements and was associated with higher FVC relative to pretransplant baseline, an association that should be interpreted in light of the unadjusted comparison and baseline heterogeneity rather than as evidence of a causal benefit. Full article

Vasoplegia remains one of the most challenging and incompletely understood complications across the spectrum of mechanical circulatory support (MCS). Initially described following cardiopulmonary bypass, it is increasingly recognized in patients supported with veno-arterial extracorporeal membrane oxygenation (V-A ECMO) and combined unloading strategies such as ECPELLA (ECMO and Impella support) as well as in patients bridged to heart transplantation with temporary or durable mechanical circulatory support (MCS). Despite occurring in different clinical settings, these syndromes share common pathophysiological features, including systemic inflammation, endothelial dysfunction, glycocalyx degradation, dysregulated nitric oxide signaling, neurohormonal imbalance, microcirculatory impairment and severe vasomotor dysregulation. Although vasoplegia is commonly considered a technology-specific complication, growing evidence suggests that CPB-associated vasoplegia, postcardiotomy vasoplegic syndrome, ECMO-related vasodilatory shock and distributive shock during ECPELLA may represent distinct manifestations of a common pathobiological process driven by blood–artificial surface interactions, ischemia–reperfusion injury, hemolysis and immune activation. This narrative review proposes a unified framework of vasoplegia across the continuum of MCS. Key mechanistic pathways and current therapeutic strategies, including catecholamines, vasopressin, angiotensin II, methylene blue and hydroxocobalamin, are discussed. By integrating evidence from cardiac surgery, critical care and mechanical circulatory support, vasoplegia is presented as a unifying syndrome of extracorporeal circulation. This perspective may support earlier recognition, phenotype-based management and the development of more targeted therapeutic strategies in a clinically significant yet underexplored area of cardiovascular critical care. Full article

Background and Clinical Significance: Peripheral veno-arterial extracorporeal membrane oxygenation (VA-ECMO) substantially alters aortic flow dynamics, generating catastrophic false-positive pathology on standard imaging. We report a case of ECMO-induced artifacts mimicking a Stanford type A aortic dissection (TAAD), which led to an unnecessary exploratory sternotomy. Case Presentation: A 67-year-old man underwent extracorporeal cardiopulmonary resuscitation (ECPR) for a shockable out-of-hospital cardiac arrest. Post-resuscitation chest computed tomography angiography (CTA) and preoperative transesophageal echocardiography (TEE) demonstrated a prominent flap-like structure in the ascending aorta, prompting emergency sternotomy. Intraoperative exploration revealed no intimal tear. Subsequent evaluation confirmed an acute anterior myocardial infarction, managed with coronary intervention. Following a dismal neurological prognosis due to hypoxic encephalopathy, VA-ECMO was palliatively withdrawn on day 9, and the patient expired on day 19. The interaction between retrograde ECMO flow and varying levels of intrinsic cardiac function dictates the topology of flow disturbances. Absent native flow creates contrast layering within the aortic root, whereas preserved native flow creates a volatile downstream watershed zone. Based on these distinct phenotypes, we propose a novel conceptual framework for tailor-made imaging strategies titrated to native flow strength—such as temporary ECMO flow reduction for preserved native output, or circuit contrast injections for profound cardiac depression. Conclusions: ECMO-related artifacts present substantial diagnostic pitfalls. Clinicians should adopt a context-aware approach, integrating multi-modality imaging with hemodynamic status to implement individualized, physiologically guided imaging protocols. Full article

Objective: Intentional pharmaceutical poisoning is one of the most common methods of self-harm among adolescents and frequently results in pediatric intensive care unit (PICU) admission because of the potential for severe pharmaceutical-related toxicity. Although mortality is generally low, a subset of patients may require advanced intensive care interventions. Methods: This retrospective observational cohort study was conducted in the pediatric intensive care unit of a tertiary referral hospital between January 2022 and January 2026 and included adolescents admitted to the PICU due to intentional pharmaceutical poisoning. Demographic characteristics, psychiatric history, clinical findings, laboratory parameters, administered treatments, and clinical outcomes were evaluated. The primary outcome measure was defined as the requirement for a major PICU intervention, including invasive or noninvasive mechanical ventilation, vasoactive support, continuous renal replacement therapy (CRRT), therapeutic plasma exchange, extracorporeal membrane oxygenation (ECMO), intensive care-level seizure management, or clinically significant arrhythmia requiring intensive care support. Patients with and without major PICU intervention requirements were compared. Results: A total of 125 adolescent patients were included in the study, and the majority were female. Multiple-pharmaceutical ingestion and a history of psychiatric diagnoses were common comorbidities. Although the clinical course was stable in most patients, 15 patients (12%) required major pediatric intensive care interventions. Patients requiring major interventions had lower Glasgow Coma Scale scores, higher lactate levels, more frequent symptomatic presentation, and higher rates of Poisoning Severity Score (PSS) ≥ 3. Overall mortality in the cohort was low. Conclusions: Although most adolescents admitted to the PICU due to intentional pharmaceutical poisoning experienced mild-to-moderate toxicity, a subgroup required advanced intensive care support. Neurological impairment at presentation, symptomatic clinical presentation, and markers of increased physiologic stress appeared to be associated with more severe clinical courses. In pediatric poisonings, evaluating the need for advanced intensive care support rather than focusing solely on mortality may better reflect clinical severity. Full article

Extracorporeal membranes oxygenation (ECMO) has been proposed as a useful tool to support ablation of unstable ventricular arrhythmias (VAs). The aim of this study is to assess the clinical outcome of cardiopulmonary support of VAs catheter ablation during a long-term follow-up. In this retrospective observational study, we included 47 patients referred to our center for catheter ablation of repeated episodes of hemodynamically unstable sustained VAs between April 2016 and February 2025. All patients underwent catheter ablation, supported by ECMO, of ventricular arrhythmias symptomatic for syncope or presyncope. The primary endpoint is overall cardiovascular death, including death due to heart failure, cardiogenic shock or ventricular arrhythmias. In particular, arrhythmic death was defined as death occurring during an electrical storm. After a median follow-up of 28 (7–63.5) months, cardiovascular death occurred in 26 patients (55.3%) but arrhythmic death befell only 11 patients (23.4%). All deaths occurred within 6–7 years of follow-up. No difference exists between ischemic and non-ischemic cardiomyopathy with regard to primary endpoints. Moreover, arrhythmic recurrences occurred in 21 patients (44.7%), among whom only 15 (31.9%) had ICD shocks; 25 patients (55.3%) encountered further hospitalizations. ECMO may facilitate procedural mapping and acute ablation success in selected high-risk patients, while long-term prognosis remains mainly driven by advanced heart failure. Full article

Objective: Extracorporeal Membrane Oxygenation (ECMO) has long been used in the treatment of acute respiratory and circulatory failure by providing time for damaged organs to recover. The aim of this study was to evaluate the safety and feasibility of interhospital transport of pediatric patients with acute respiratory failure who had undergone venovenous extracorporeal membrane oxygenation (VV ECMO) initiated at the referring facilities. Subjects and methods: Because of the critical condition of these patients, the high risk associated with transport, and the failure of conventional therapies, ECMO was initiated at the referring center. After cannulation, the patients were transported by ground ambulance to the Pediatric Intensive Care Unit in Poznań for further treatment. Results: Fourteen patients aged 2 months to 11 years with acute respiratory failure were transferred to our ECMO center. The mean time from decision to departure was 7.62 h, and the mean ICU stay before transfer was 4.14 days. The mean transport distance was 157.5 km. No mortality occurred during transport, and no serious adverse events were reported. Two technical complications were noted. Conclusions: Interhospital transport of pediatric patients on VV ECMO initiated at referring centers was feasible and safe, with favorable outcomes in patients who have exhausted conventional intensive care options. Effective collaboration between referring hospitals, ECMO centers, and emergency medical services was essential for optimal results. Full article

Extracorporeal cardiopulmonary resuscitation (ECPR) has emerged as a promising strategy for selected patients with refractory cardiac arrest, improving survival and the likelihood of favorable neurological outcomes. However, neurological prognostication in this setting remains highly challenging and insufficiently standardized. The pathophysiological complexity of ECPR, including global ischemia–reperfusion injury, altered cerebral perfusion, systemic inflammation, anticoagulation and prolonged sedation, limits the reliability of conventional post-cardiac arrest prognostic tools. This narrative review provides a focused and clinically oriented synthesis of current evidence on brain injury and neuroprognostication in patients undergoing veno-arterial extracorporeal membrane oxygenation (V-A ECMO) for cardiac arrest. Key determinants of neurological outcome across pre-ECMO and peri-resuscitation phases are examined, alongside the role and limitations of multimodal monitoring strategies, including neurological examination, electroencephalography, neuroimaging, cerebral oximetry and circulating biomarkers. Particular attention is given to the timing of prognostication and the risk of premature or inaccurate predictions leading to self-fulfilling prophecies. Emerging data suggest that neurological recovery in ECPR patients may be delayed, supporting a more cautious and time-adapted approach. A pragmatic, multimodal framework for neurological assessment in this population is outlined. By addressing current gaps and proposing a structured approach, this review aims to inform clinical decision making and contribute to improved neurologically meaningful survival in ECPR-treated cardiac arrest. Full article

Background: Biotrauma from invasive mechanical ventilation (IMV) and extracorporeal membrane oxygenation (ECMO) drives systemic inflammation, metabolic dysregulation, and organ dysfunction in critically ill patients. Therefore, this study aimed to identify clinical and metabolomic features associated with ICU mortality in patients receiving IMV or ECMO, as these remain incompletely characterized. Methods: The retrospective analysis included 30 ICU patients on IMV and 22 on ECMO. Metabolomic and proteomic profiling were performed using ultra-high-performance liquid chromatography coupled with high-resolution mass spectrometry (UHPLC-HRMS), and serum spectral analysis by Fourier-transform infrared spectroscopy (FTIRS). Significant variables were incorporated into multivariate logistic regression models, ranked by AIC, AUC, and statistical significance. Model performance was evaluated using stratified 5-fold cross-validation. Final models were adjusted for relevant demographic and clinical covariates. Results: The IMV cohort showed discriminatory FTIRS wavenumbers across all preprocessings, and 155 metabolites plus 14 proteins were significantly altered, with unadjusted models achieving mean AUCs above 0.9. The ECMO cohort showed discriminatory FTIRS wavenumbers in one preprocessing, and 15 metabolites plus 3 proteins were highlighted. FTIRS, metabolomic, and proteomic models reached mean AUCs of 0.967, 0.867, and 0.783, respectively, with lower stability during cross-validation. Adjustment for demographic and clinical covariates reduced model robustness. Conclusions: Stronger and more reproducible molecular signatures related to ICU mortality were observed in the IMV cohort, whereas the ECMO cohort showed reduced model stability, likely reflecting increased biological heterogeneity and small sample size. These findings support the utility of integrated omics for characterizing critical illness and outcome stratification, while reinforcing the need for validation in larger and independent cohorts. Full article