Daily Ards Research Analysis
Three ARDS-focused studies stand out today: a meta-analysis shows AI can diagnose and subphenotype ARDS with high accuracy yet poor external validation; a VV-ECMO cohort finds major bleeding is common but not independently associated with 90-day mortality; and a retrospective analysis supports subclavian vein access as a viable alternative cannulation site in VV-ECMO.
Summary
Three ARDS-focused studies stand out today: a meta-analysis shows AI can diagnose and subphenotype ARDS with high accuracy yet poor external validation; a VV-ECMO cohort finds major bleeding is common but not independently associated with 90-day mortality; and a retrospective analysis supports subclavian vein access as a viable alternative cannulation site in VV-ECMO.
Research Themes
- AI-enabled ARDS diagnosis and subphenotyping
- VV-ECMO complications and outcomes in ARDS
- Cannulation strategies for VV-ECMO
Selected Articles
1. Systematic review and meta-analysis of artificial intelligence models for diagnosing and subphenotyping ARDS in adults.
Across 63 studies, AI models achieved pooled sensitivity 0.89, specificity 0.88, and AUROC 0.90 for ARDS detection, with imaging-based and deep learning approaches performing best. However, high heterogeneity, limited external validation, and poor calibration reporting constrain clinical deployment; subphenotyping evidence remains exploratory.
Impact: This synthesis sets realistic performance expectations for AI in ARDS and clarifies methodological gaps (calibration, external validation) that must be addressed for clinical adoption.
Clinical Implications: AI decision support for ARDS diagnosis and subphenotyping could be integrated into workflows once models are prospectively validated with robust calibration and generalization across sites and imaging platforms.
Key Findings
- Pooled sensitivity 0.89, specificity 0.88, and AUROC 0.90 for ARDS detection across 63 studies.
- High heterogeneity (I² > 85%) and frequent lack of external validation (29/63 studies).
- Imaging-based and deep learning models outperformed non-imaging and traditional ML approaches.
- Calibration reporting was missing in 47% of studies.
- Only 7 studies (18%) explored subphenotyping, identifying hyper- and hypoinflammatory profiles.
Methodological Strengths
- Comprehensive multi-database search with quantitative pooling and PROBAST-based bias assessment
- Stratified analyses by model type, data modality, and COVID-19 context
Limitations
- High between-study heterogeneity and scarce external validation limit generalizability
- Calibration metrics were often unreported, hindering clinical thresholding
Future Directions: Prospective, multi-center external validations with standardized reporting (e.g., TRIPOD/PROBAST), calibration assessment, and head-to-head comparisons of imaging and non-imaging models for real-time ARDS triage.
BACKGROUND: Artificial intelligence (AI) has emerged as a promising tool to improve the diagnosis and characterization of ARDS, including the identification of subphenotypes. OBJECTIVES: To evaluate the diagnostic performance and methodological quality of AI models for identifying ARDS and its subphenotypes in adults. METHODS: We conducted a systematic review and meta-analysis of 63 studies (n = 135,762) published between 2013 and 2024 in PubMed, Embase, and the Cochrane Library. Extracted outcomes included sensitivity, specificity, AUROC, and validation methods. Risk of bias was assessed with PROBAST, and AI-specific metrics (overfitting, generalization, interpretability, discrimination, calibration) were reported. RESULTS: Pooled sensitivity was 0.89 (95 % CI 0.84-0.93), specificity 0.88 (95 % CI 0.83-0.92), and AUROC 0.90 (95 % CI 0.86-0.94), with high heterogeneity (I² > 85 %). Twenty-two studies (31 %) were rated high quality, with sensitivity 0.86 (95 % CI 0.82-0.89) and specificity 0.82 (95 % CI 0.78-0.85). Deep learning models (n = 14) achieved sensitivity 0.91, while machine learning models (n = 19) showed 0.87. Imaging-based models (n = 15) outperformed non-imaging approaches. COVID-19 studies (n = 9) reported sensitivity 0.90 with comparable AUROC and specificity. Only seven studies (18 %) investigated subphenotyping, identifying hyperinflammatory and hypoinflammatory profiles with potential therapeutic relevance. Calibration reporting was missing in 47 % and external validation in most (29/63). CONCLUSION: AI models for ARDS demonstrate promising diagnostic accuracy but are limited by poor calibration and scarce external validation. Subphenotyping remains exploratory but suggests opportunities for real-time patient stratification. Prospective validation and standardized reporting are essential for clinical adoption.
2. Outcome of patients with COVID-19 supported by veno-venous extracorporeal membrane oxygenation with major bleeding: a single centre experience.
In 151 COVID-19 patients on VV-ECMO, major bleeding occurred in 48.3%, with longer ECMO duration as the only independent predictor. Notably, major bleeding (including intracranial hemorrhage) did not independently increase 90-day mortality, whereas kidney replacement therapy did.
Impact: Challenges assumptions that major bleeding invariably worsens mortality in VV-ECMO, informing risk–benefit considerations for anticoagulation and transfusion strategies.
Clinical Implications: ECMO teams should monitor and mitigate prolonged ECMO runs and consider the prognostic impact of kidney replacement therapy. Anticoagulation strategies may be individualized without assuming that any major bleeding will necessarily worsen 90-day mortality.
Key Findings
- Major bleeding occurred in 48.3% (73/151) of VV-ECMO patients with COVID-19.
- Longer ECMO duration independently increased bleeding risk (OR 1.32; 95% CI 1.14–1.53).
- Major bleeding, including intracranial hemorrhage, was not independently associated with higher 90-day mortality.
- Kidney replacement therapy independently increased 90-day mortality (OR 4.48; 95% CI 1.83–10.98).
Methodological Strengths
- Defined primary outcomes with multivariable logistic regression to identify independent predictors
- Single-center cohort with consistent COVID-19-era management protocols
Limitations
- Retrospective single-center design with potential residual confounding
- Findings limited to COVID-19 ARDS and may not generalize to non-COVID VV-ECMO
Future Directions: Prospective, multicenter studies to validate bleeding–mortality relationships and to evaluate anticoagulation protocols stratified by ECMO duration and organ support needs.
BACKGROUND: Patients with severe COVID-19 often require veno-venous extracorporeal membrane oxygenation (VV-ECMO) due to acute respiratory distress syndrome (ARDS). Major bleeding complications are common and linked to worse outcomes, though specific risk factors in COVID-19 remain unclear. METHODS: A retrospective analysis of 151 critically ill patients with COVID-19 on VV-ECMO (March 2020-December 2021) was conducted. The primary outcome was major bleeding (fatal bleeding, haemoglobin drop ≥ 20 g/L RESULTS: Major bleeding occurred in 73/151 patients (48.3%). Only a longer ECMO duration [OR 1.32 (95% CI 1.14-1.53; p < 0.001)] was identified as an independent risk factor. Kidney replacement therapy independently influenced 90-day mortality [OR 4.48 (95% CI 1.83-10.98;p = 0.001). However, major bleeding, intracranial haemorrhage, higher burden of co-morbidity and mean aPTT before major bleeding were not associated with an increased 90-day mortality risk. CONCLUSION: Major bleeding events, including intracranial haemorrhage, are common in patients with COVID-19 being supported by VV-ECMO. However, our data does not demonstrate a direct association between major bleeding and increased 90-day mortality.
3. Outcomes of Subclavian Vein Cannulation in Venovenous Extracorporeal Membrane Oxygenation: A Single-Center Retrospective Study.
In 157 VV-ECMO patients, subclavian vein cannulation showed no increased cannulation-specific or ECMO-related complications compared with other access sites. Mixed-effects logistic regression supported SCV as a safe alternative when conventional access is limited.
Impact: Provides pragmatic data supporting subclavian access as an alternative route, potentially expanding cannulation options when femoral or jugular access is not feasible.
Clinical Implications: ECMO teams can consider subclavian cannulation without expecting higher complication rates, enabling individualized cannulation strategies in challenging anatomies or when mobilization is prioritized.
Key Findings
- Single-center retrospective analysis of 157 VV-ECMO patients stratified by subclavian vs non-subclavian cannulation.
- No difference in site-related adverse events or ECMO-related complications between SCV and non-SCV groups.
- Mixed-effects logistic regression used to compare complication risks across cannulation sites.
Methodological Strengths
- Use of mixed-effects logistic regression to account for cannulation events
- Clear definition of cannulation-specific and ECMO-related adverse outcomes
Limitations
- Retrospective single-center design with potential selection bias regarding access site choice
- Incomplete reporting of subgroup sizes and certain baseline characteristics
Future Directions: Prospective multicenter comparisons of cannulation sites assessing mobility, rehabilitation outcomes, and long-term complications, with standardized anticoagulation and device strategies.
OBJECTIVE: Venovenous (VV) extracorporeal membrane oxygenation (ECMO) cannulation may pose venous access challenges. Subclavian vein (SCV) cannulation is an alternative site. This study hypothesizes that SCV access is a safe alternative in VV ECMO. METHODS: This is a single-center, retrospective study of peripheral VV ECMO that was stratified by SCV cannulation. Each site was considered as a separate cannulation event. Descriptive statistics, groupwise comparisons, and mixed-effects logistic regression were used. Primary endpoints included cannulation-specific complications and ECMO-related adverse events. RESULTS: From 2020 to 2023, 157 patients were supported with VV ECMO. The cohort was 57% male patients with a median age of 44 (34 to 56) years and a median body mass index of 34 (28 to 44) kg/m CONCLUSIONS: Between the SCV and non-SCV groups, there was no difference in site-related adverse events or complications. This study supports the use of SCV cannulation as a viable alternative in patients supported with VV ECMO.