Daily Sepsis Research Analysis
Analyzed 49 papers and selected 3 impactful papers.
Summary
Top findings today span precision immunotherapy, interpretable AI prognostication, and metabolomics-driven mechanisms in sepsis. A double-blind RCT shows presepsin-guided anakinra reduces organ dysfunction and mortality in pneumonia; a multicenter interpretable ML model predicts mortality in sepsis-associated AKI; and metabolomics identifies succinate as a key mediator of sepsis-associated encephalopathy with in vivo validation.
Research Themes
- Precision immunotherapy guided by biomarkers in sepsis
- Interpretable machine learning for critical care prognostication
- Metabolomics linking systemic metabolism to neuroinflammation in sepsis
Selected Articles
1. Efficacy of anakinra in reducing progression to organ dysfunction in patients with pneumonia (INSPIRE): a randomised, double-blind, placebo-controlled, phase IIa trial.
In a double-blind phase IIa RCT of 60 hospitalized pneumonia patients enriched by presepsin (>350 pg/mL) and qSOFA=1, anakinra 100 mg daily for 10 days reduced progression to organ dysfunction (20% vs 50%; p=0.011) and 90-day mortality (20.0% vs 43.3%; p=0.029) versus placebo. Serious TEAEs were fewer with anakinra and none were treatment-related.
Impact: Demonstrates a biomarker-guided immunotherapy strategy that significantly improves hard outcomes in pneumonia at risk for sepsis-related organ dysfunction.
Clinical Implications: Supports presepsin-guided IL-1 blockade as a precision therapy to prevent deterioration in pneumonia. If replicated in larger RCTs, this could inform sepsis bundle updates and personalized immunomodulation.
Key Findings
- Primary endpoint (SOFA +2 by day 7 and/or 90-day death) reduced from 50.0% (placebo) to 20.0% (anakinra); p=0.011
- 90-day mortality reduced from 43.3% to 20.0%; p=0.029
- Serious TEAEs occurred less frequently with anakinra (33.3% vs 50%) and none were treatment-related
- Cytokine production (TNFα, IFNγ) by blood mononuclear cells was attenuated with anakinra
Methodological Strengths
- Randomized, double-blind, placebo-controlled design with ITT analysis
- Biomarker-enriched population (presepsin) enables precision targeting and mechanistic plausibility
Limitations
- Small, single phase IIa sample size (n=60) limits generalizability
- Conducted in pneumonia with qSOFA=1; applicability to broader sepsis phenotypes requires study
Future Directions: Larger, multicenter phase III trials to confirm efficacy, refine presepsin thresholds, and assess timing/duration; head-to-head comparisons with other immunomodulators.
BACKGROUND: Early recognition of risk of death in pneumonia and start of precision immunotherapy to improve outcomes is an unmet need. We hypothesized that a precision strategy approach combining early recognition of interleukin (IL)-1 activation coupled with Anakinra treatment may improve pneumonia outcome. METHODS: INSPIRE is a prospective, double-blind randomized placebo-controlled trial which recruited hospitalized adults with community-acquired or hospital-acquired pneumonia, with qSOFA (quick sequential organ failure assessment) equal to 1 and plasma presepsin (soluble CD14) more than 350 pg/ml. Patients were 1:1 randomised to standard-of-care medication plus either subcutaneous placebo or subcutaneous Anakinra 100 mg once daily for 10 days. The primary endpoint was progression into organ dysfunction defined as meeting at least one of the following two conditions; (i) at least 2-point increase of the baseline SOFA score by day 7 and/or (ii) death by day 90. This was analyzed in the intention-to-treat (ITT) population. This trial is registered with the EU Clinical Trials Register (2022-002390-28) and ClinicalTrials.gov (NCT05785442). FINDINGS: Patients were enrolled between March 2023 and June 2024; 30 patients in the placebo arm and 30 patients in the Anakinra arm were the ITT population. The primary endpoint was met in 50·0% (15 patients) of placebo-treated and in 20·0% (6 patients) of Anakinra-treated patients (difference 30·0% [5·9 to 49%]; p: 0·011). 90-day mortality was 43·3% (13 patients) and 20·0% (6 patients) (difference 23·3% [0 to 43·7%]; p: 0·029). The overall incidence of serious treatment-emergent adverse events (TEAEs) in the placebo group was 50% and in the Anakinra group 33·3%. None of the serious TEAEs was judged to be related to treatment assignment. Production of TNFα and ΙFNγ by blood mononuclear cells was decreased. INTERPRETATION: Presepsin-guided Anakinra treatment prevents progression of pneumonia to organ dysfunction and death. The mechanism of benefit is associated with attenuation of cytokine production. FUNDING: Hellenic Institute for the Study of Sepsis; PHC Europe BV; Swedish Orphan Biovitrum AB.
2. Widely targeted metabolomics and machine learning identify succinate as a key metabolite in sepsis-associated encephalopathy.
Plasma metabolomics with machine learning identified 12 discriminatory metabolites, with succinate progressively increasing from health to sepsis to SAE and correlating with severity. In a CLP mouse model, exogenous succinate worsened cognition, neuronal injury, and microglial activation, supporting a mechanistic role of succinate in SAE.
Impact: Bridges human metabolomic signatures with functional validation, highlighting succinate as a potential biomarker and therapeutic target for sepsis-associated encephalopathy.
Clinical Implications: Succinate and related pathways may enable risk stratification for SAE and inspire metabolic interventions; translation requires prospective validation and interventional studies.
Key Findings
- Identified 12 discriminatory plasma metabolites across healthy, sepsis, and SAE groups; succinate showed stepwise increases linked to severity
- Exogenous succinate aggravated cognitive deficits, neuronal injury, and microglial activation in CLP mice
- Findings connect systemic metabolic remodeling to neuroinflammation and dysfunction in sepsis
Methodological Strengths
- Integrated human metabolomics with in vivo functional validation in a standard CLP model
- Machine learning improved metabolite selection and interpretability
Limitations
- Modest human sample size and cross-sectional design limit causal inference
- Exogenous succinate may not fully recapitulate endogenous metabolic dynamics
Future Directions: Prospective validation of succinate as a biomarker, exploration of inhibitors/modulators of succinate signaling, and interventional trials targeting metabolic pathways in SAE.
Sepsis-associated encephalopathy (SAE) is a common and serious complication of sepsis that leads to acute brain dysfunction and long-term cognitive impairment. We used widely targeted LC-MS/MS plasma metabolomics in 29 healthy controls, 32 sepsis patients, and 27 SAE patients, combined with machine learning, to define metabolic patterns across these groups. This approach identified 12 discriminatory metabolites, with succinate showing a stepwise increase from health to sepsis to SAE and associations with clinical severity scores. To test its functional relevance, we used a cecal ligation and puncture (CLP) mouse model and found that exogenous succinate supplementation aggravated cognitive deficits, neuronal injury, and microglial activation. Together, these findings link systemic metabolic remodeling to brain inflammation and dysfunction in sepsis and suggest that succinate and related pathways may help stratify SAE risk and provide mechanistic entry points for future therapeutic exploration.
3. Prediction of hospital mortality in sepsis-associated acute kidney injury using a machine-learning approach: a multicenter study using SHAP interpretability analysis.
Using five international ICU databases, an interpretable gradient boosting model predicted hospital mortality in 27,485 S-AKI patients with AUCs ~0.73–0.78 across external cohorts and strong calibration. SHAP pinpointed SAPS II and SOFA thresholds as major risk inflection points, supporting personalized risk communication and triage.
Impact: Delivers a generalizable, interpretable prognostic tool for S-AKI across diverse settings, addressing a key gap in precision risk stratification.
Clinical Implications: May inform early discussions on goals of care, resource allocation, and enrollment in interventional trials; integration into EHRs could enable bedside decision support pending prospective impact evaluation.
Key Findings
- Developed a GBM model with AUC 0.770 (training), 0.731 (internal validation), and 0.732–0.778 across external cohorts
- Excellent calibration and minimal overfitting (3.9% AUC difference)
- SHAP identified SAPS II (>60) and SOFA (>15) as dominant risk drivers; DCA showed net benefit across 4–82% thresholds
Methodological Strengths
- External validation across four independent international cohorts with consistent performance
- Use of SHAP for transparent feature attribution and clinical interpretability
Limitations
- Retrospective design with potential confounding and coding heterogeneity across databases
- No prospective clinical impact assessment or randomized evaluation of model-guided care
Future Directions: Prospective impact studies, calibration drift monitoring, and integration with treatment effect modeling to support dynamic decision-making.
BACKGROUND: Sepsis-associated acute kidney injury (S-AKI) represents a critical complication with high mortality rates in intensive care units. Current risk stratification tools lack precision and interpretability for clinical decision-making. This study aimed to develop and validate interpretable machine learning models for predicting hospital mortality in S-AKI patients. METHODS: This retrospective cohort study utilized five international critical care databases: Medical Information Mart for Intensive Care (MIMIC)-IV ( RESULTS: Among 27 485 S-AKI patients, hospital mortality was 27.5%. Boruta identified 21 consensus features including severity scores [Simplified Acute Physiology Score II (SAPS II), Sequential Organ Failure Assessment (SOFA), OASIS], vital signs and laboratory parameters. Gradient Boosting Machine emerged as optimal with area under the curve (AUC) values of 0.770 (training), 0.731 (internal validation) and 0.732-0.778 across four external validation cohorts. The model demonstrated excellent calibration and minimal overfitting (3.9% AUC difference). Decision curve analysis revealed superior clinical utility across probability thresholds of 4%-82%. SHAP analysis identified SAPS II as the most important predictor, with scores >60 and SOFA >15 associated with substantially increased mortality risk. Complete case analysis confirmed model robustness (AUC 0.766-0.847). CONCLUSIONS: The interpretable machine learning model demonstrated excellent performance and robust generalizability for S-AKI mortality prediction across five international databases. SHAP analysis provided clinically meaningful insights supporting personalized risk stratification and evidence-based clinical decision-making.