Daily Sepsis Research Analysis
Analyzed 44 papers and selected 3 impactful papers.
Summary
An international randomized trial suggests benzylpenicillin may be preferable to anti-staphylococcal penicillins for penicillin-susceptible Staphylococcus aureus bacteraemia, with similar mortality and less acute kidney injury. A genomic study links neonatal invasive Group B Streptococcus disease to a CC17 lineage profile, while severity is driven mainly by clinical factors. An interpretable, externally validated machine learning model stratifies mortality risk in sepsis-associated AKI patients on CRRT.
Research Themes
- Antimicrobial therapy optimization in bloodstream infections
- Pathogen genomics and lineage-specific virulence in neonatal sepsis
- AI-driven prognostic stratification in sepsis-associated organ failure
Selected Articles
1. Benzylpenicillin versus flucloxacillin or cloxacillin for the treatment of penicillin-susceptible Staphylococcus aureus bacteraemia (SNAP): an international, multicentre, open-label, non-inferiority randomised controlled trial.
In adults with PSSA bacteraemia, benzylpenicillin had a high posterior probability of non-inferiority for 90-day mortality versus flucloxacillin/cloxacillin (adjusted OR 0.67; 95% CrI 0.35–1.28), and significantly reduced acute kidney injury (AKI) risk (adjusted OR 0.50). Recruitment was stopped early due to excess AKI in the anti-staphylococcal penicillin arm.
Impact: This pragmatic, multicentre RCT provides actionable evidence favoring benzylpenicillin over anti-staphylococcal penicillins for PSSA bacteraemia, balancing efficacy with reduced nephrotoxicity.
Clinical Implications: When PSSA is confirmed, benzylpenicillin should be preferred over flucloxacillin/cloxacillin, given comparable mortality and lower AKI risk. Robust susceptibility testing workflows are essential to enable timely agent selection.
Key Findings
- 90-day mortality: 14% with benzylpenicillin vs 22% with flucloxacillin/cloxacillin (adjusted OR 0.67; 95% CrI 0.35–1.28).
- AKI occurred in 11% (benzylpenicillin) vs 22% (anti-staphylococcal penicillins); adjusted OR 0.50; high probability of superiority.
- Recruitment ceased early due to increased AKI in the flucloxacillin/cloxacillin group; formal non-inferiority criterion was not met as the CrI exceeded the NI margin.
Methodological Strengths
- International, multicentre randomized platform with Bayesian analysis.
- Clinically relevant primary endpoint (90-day all-cause mortality) and predefined dosing regimens.
Limitations
- Open-label design with potential performance bias.
- Early termination and credible interval crossing the NI margin limit definitive non-inferiority conclusions.
Future Directions: Prospective implementation studies assessing outcomes and AKI with benzylpenicillin-first strategies, and optimization of rapid diagnostics to confidently identify PSSA at the bedside.
BACKGROUND: Previously considered rare, penicillin-susceptible Staphylococcus aureus (PSSA) bacteraemia has re-emerged worldwide. Although benzylpenicillin might offer advantages in terms of pharmacokinetic and adverse effect profiles, anti-staphylococcal penicillins are recommended for serious infections because of concern over undetected penicillin resistance. We aimed to compare benzylpenicillin with anti-staphylococcal penicillins (cloxacillin or flucloxacillin) for the treatment of PSSA bacteraemia in adults. METHODS: This investigator-initiated, international, multicentre, open-label, non-inferiority, randomised controlled trial was conducted within the PSSA silo of the backbone domain of the ongoing S aureus Network Adaptive Platform (SNAP) trial. We enrolled patients of all ages who were admitted with S aureus bacteraemia at 67 hospitals across Australia, New Zealand, Canada, Israel, the Netherlands, the UK, Singapore, and South Africa. Herein, we report results for adult patients (aged ≥18 years). Participants were randomly allocated 1:1 to receive either benzylpenicillin, or flucloxacillin or cloxacillin. Trial staff, staff caring for participants, and participants were aware of the treatment allocated and received. Cloxacillin was used only where flucloxacillin was not available (ie, Canada, Israel, Singapore, and South Africa). Recommended standard dosing for benzylpenicillin was 1·8 g intravenously once every 4 h or 2·4 g once every 6 h; for flucloxacillin 2·0 g intravenously once every 6 h; and for cloxacillin 2·0 g intravenously once every 4 h. The primary outcome was all-cause mortality 90 days after platform entry, assessed in the intention-to-treat population, including all patients with available data. The primary outcome was analysed by use of a hierarchical Bayesian logistic regression model, with non-inferiority of benzylpenicillin defined as an adjusted odds ratio (OR) of less than 1·20. Analyses occurred after every 500 participants reached 90-day follow-up. The SNAP trial is registered with ClinicalTrials.gov (NCT05137119) and is ongoing.
2. Lineage-encoded genomic determinants of neonatal invasive Group B Streptococcal disease.
Neonatal invasive GBS disease was strongly associated with CC17 (65% of invasive vs 28% of colonizing isolates). Presence of hylB, hvgA, and ermB correlated with invasiveness, whereas pilus island 1 was negatively associated. Among invasive cases, LTOD related to prematurity, early-onset disease, and lower birthweight; no bacterial genomic features remained associated with severity after multiple-testing correction.
Impact: This study refines our understanding of lineage-specific genomic architectures driving neonatal GBS invasiveness while highlighting that severe outcomes are largely host- and context-driven.
Clinical Implications: Enhanced surveillance and targeted preventive strategies (e.g., vaccine candidates and diagnostics) should prioritize CC17-linked determinants. Risk stratification for severe disease should emphasize prematurity and early-onset features rather than bacterial genotype alone.
Key Findings
- CC17 accounted for 65% of invasive isolates versus 28% of colonizing isolates, indicating strong lineage association with invasiveness.
- hylB, hvgA, and ermB were positively associated with invasive disease; pilus island 1 showed a negative association.
- Among invasive cases, LTOD was associated with prematurity, early-onset disease, and lower birthweight; no genomic markers predicted severity after FDR correction.
Methodological Strengths
- PacBio long-read whole-genome sequencing enabling allele-level resolution.
- Serotype-matched colonizing controls and FDR-adjusted statistical analysis.
Limitations
- Observational design without host-genomic or immunologic data limits causal inference.
- Univariable logistic models may not fully adjust for confounding; external validation across diverse settings is needed.
Future Directions: Integrate host-genomic and immunophenotypic data with bacterial genomics to model invasion risk; evaluate CC17-focused vaccine antigens and diagnostics in prospective studies.
OBJECTIVES: Group B Streptococcus (GBS) is a leading cause of invasive infection in newborns and infants. Yet the genomic basis of invasiveness and disease severity remains incompletely defined. We aimed to identify bacterial genomic features associated with invasive GBS disease and to explore associations between bacterial genomics and severe clinical outcomes. METHODS: We performed PacBio long-read whole-genome sequencing of 381 GBS isolates, including 127 invasive isolates obtained from blood cultures of infants with GBS disease and 254 serotype-matched colonizing isolates obtained from rectovaginal swabs of pregnant women. Gene presence/absence and allele-level variation were analyzed for virulence, adhesion, hemolysis, regulatory, and antimicrobial resistance genes. Associations with invasive disease, life-threatening organ dysfunction (LTOD), and meningitis were assessed using univariable logistic regression with false discovery rate correction. Correlation matrices were constructed to identify lineage-associated genomic profiles.
3. Risk stratification for in-hospital mortality in sepsis-associated acute kidney injury patients receiving continuous renal replacement therapy: an interpretable, externally validated machine learning study.
A GBM-based model using routinely available variables achieved AUCs of 0.756 (internal) and 0.752 (external) for in-hospital mortality prediction in SA-AKI patients on CRRT, outperforming SOFA and SAPS II. SHAP identified urine output, serum creatinine, and age as key predictors; calibration varied across cohorts, highlighting the need for local recalibration.
Impact: Provides a pragmatic, interpretable, and externally validated risk tool for a high-mortality sepsis subgroup, with potential to guide triage, communication, and resource allocation.
Clinical Implications: Clinicians can use model-informed risk stratification to prioritize monitoring, consider tailored CRRT strategies, and inform goals-of-care discussions, while ensuring local validation and calibration before deployment.
Key Findings
- Multicenter development cohort (n=1,217) and external validation cohort (n=332); GBM achieved AUC 0.756 (internal) and 0.752 (external).
- Model outperformed conventional scores (SOFA, SAPS II) in external validation; SHAP highlighted urine output, creatinine, and age as top contributors.
- Calibration was affected by population heterogeneity, emphasizing the need for local recalibration for clinical deployment.
Methodological Strengths
- External validation across geographically distinct cohorts and SHAP-based interpretability.
- Robust feature selection (LASSO, Boruta) and comparison against multiple ML models and clinical scores.
Limitations
- Retrospective design with potential residual confounding and missing data bias.
- Single-country external cohort; no prospective impact evaluation or clinical utility trial.
Future Directions: Prospective impact studies integrating the model into CRRT decision workflows, with site-specific recalibration and assessment of downstream clinical outcomes.
Patients with sepsis-associated acute kidney injury (SA-AKI) requiring continuous renal replacement therapy (CRRT) have a high risk of in-hospital mortality, and early risk stratification may support timely clinical decision-making and efficient resource allocation. In this retrospective study, we developed and validated a prognostic model for SA-AKI patients receiving CRRT using data from the Medical Information Mart for Intensive Care IV (MIMIC-IV version 3.1 United States, 2008-2022) and the eICU Collaborative Research Database (eICU-CRD United States, 2014-2015), with external validation in an independent cohort from the intensive care unit of the Second Affiliated Hospital of Anhui Medical University (AYEFY-ICU China, 2021-2024). Candidate variables were selected using the least absolute shrinkage and selection operator (LASSO) and Boruta algorithms, and eight machine learning models were constructed and compared. Model interpretability was assessed using SHapley Additive exPlanations (SHAP). A total of 1,217 patients from the MIMIC-IV and eICU-CRD databases and 332 patients from the AYEFY-ICU cohort were included, and ten predictors were ultimately identified. Among the evaluated models, the gradient boosting machine (GBM) showed strong performance, with AUCs of 0.890, 0.756, and 0.752 in the training, internal validation, and external validation cohorts, respectively.