Daily Sepsis Research Analysis

Colistin (COL) is a cationic cyclic peptide that disrupts the membranes of Gram-negative bacteria and is often used as a last resort antibiotic against multidrug-resistant strains. The emergence of plasmid-borne mcr genes, which confer transferable COL resistance, has raised serious concerns, particularly in strains also carrying extended-spectrum β-lactamase and carbapenemase genes. Standard antimicrobial susceptibility testing (AST), performed in enriched bacteriological media, indicates no activity of COL against mcr+ strains, leading to its exclusion from treatment regimens. However, these media poorly reflect in vivo physiology and lack host immune components. Here we show that COL retained bactericidal activity against mcr-1+ Escherichia coli, Klebsiella pneumoniae, and Salmonella enterica when tested in tissue culture medium containing physiological bicarbonate. COL enhanced serum complement deposition on bacterial surfaces and synergized with human serum to kill pathogens. At clinically achievable concentrations, COL killed mcr-1+ strains in freshly isolated human blood and was effective as monotherapy in a murine E. coli bacteremia model. These findings suggest that COL, currently dismissed based on conventional AST, may offer clinical benefit against mcr-1+ infections when evaluated under more physiological conditions - warranting reconsideration in clinical microbiology practices and future trials for high-risk patients.

PURPOSE: Despite optimal antimicrobial therapy, the treatment failure rate of hospital-acquired pneumonia (HAP) routinely reaches 40% in critically ill patients. Subphenotypes have been identified within sepsis and acute respiratory distress syndrome with important predictive and possibly therapeutic implications. We derived prognosis subphenotypes for HAP and explored whether they were associated with biological markers and response to treatment. METHODS: We separately analysed data from four cohorts of critically ill patients in France (PNEUMOCARE, n = 511, ATLANREA, n = 401), Netherlands (MARS, n = 1351) and Europe-South America (ENIRRI, n = 900) to investigate HAP heterogeneity using unsupervised clustering based on clinical and routine biological variables available at HAP diagnosis. Then, we developed a machine learning-based workflow to create a simplified classification model using discovery data sets. This model was validated by applying it to an independent replication data set from an international randomized clinical trial comparing linezolid and tedizolid for the treatment of HAP (VITAL, n = 726 patients). The primary outcome was the association of subphenotypes with 28-day all-cause mortality. Secondary analyses included subphenotype associations with treatment failure at test-of-cure, respiratory microbiome and cytokine profiles in the ATLANREA subgroup, and treatment response in the VITAL trial. RESULTS: We tested twelve metrics and determined that a two-cluster model best fits all cohorts. HAP subphenotype 2 had greater disease severity, lower body temperature, and worse PaO2/FiO2 ratio than subphenotype 1 patients. Although the prevalence of subphenotype 2 ranged from 26.9 to 66.9% across the four derivation cohorts, the rates of 28-day mortality and treatment failure at test-of-cure were consistently higher to subphenotype 1 (p < 0.01 for all comparisons). Subphenotype 2 was associated with greater respiratory microbiome dysbiosis and higher levels of proinflammatory cytokines in the ATLANREA cohort, as well as with statistically significant tedizolid effect modification in the VITAL trial (Relative Risk of treatment failure with tedizolid = 1.52; 95% CI 1.12-2.06 in subphenotype 1 vs. = 0.98; 95% CI 0.7-1.38 in subphenotype 2). CONCLUSIONS: We identified two robust clinical subphenotypes by extensively analyzing HAP data sets. Their associations with respiratory microbiome composition, systemic inflammation, and treatment efficacy in independent data sets highlight their potential for prognostic value and predictive enrichment in future clinical trials aimed at personalized therapies.

OBJECTIVE: To develop an electronic medical record (EMR) data processing tool that confers clinical context to machine learning (ML) algorithms for error handling, bias mitigation, and interpretability. MATERIALS AND METHODS: We present Trust-MAPS, an algorithm that translates clinical domain knowledge into high-dimensional, mixed-integer programming models that capture physiological and biological constraints on clinical measurements. EMR data are projected onto this constrained space, effectively bringing outliers to fall within a physiologically feasible range. We then compute the distance of each data point from the constrained space modeling healthy physiology to quantify deviation from the norm. These distances, termed "trust-scores," are integrated into the feature space for downstream ML applications. We demonstrate the utility of Trust-MAPS by training a binary classifier for early sepsis prediction on data from the 2019 PhysioNet Computing in Cardiology Challenge, using the XGBoost algorithm and applying SMOTE for overcoming class-imbalance. RESULTS: The Trust-MAPS framework shows desirable behavior in handling potential errors and boosting predictive performance. We achieve an area under the receiver operating characteristic curve of 0.91 (95% CI, 0.89-0.92) for predicting sepsis 6 hours before onset-a marked 15% improvement over a baseline model trained without Trust-MAPS. DISCUSSIONS: Downstream classification performance improves after Trust-MAPS preprocessing, highlighting the bias reducing capabilities of the error-handling projections. Trust-scores emerge as clinically meaningful features that not only boost predictive performance for clinical decision support tasks but also lend interpretability to ML models. CONCLUSION: This work is the first to translate clinical domain knowledge into mathematical constraints, model cross-vital dependencies, and identify aberrations in high-dimensional medical data. Our method allows for error handling in EMR and confers interpretability and superior predictive power to models trained for clinical decision support.