Sepsis, October 26 edition

Using untargeted metabolomics, genetic manipulation, biochemical analysis, and a murine peritonitis-sepsis model, this study identified L-histidine as a metabolic adjuvant that restores carbapenem activity against NDM-producing bacteria. L-histidine inhibited NDM directly through zinc chelation and indirectly by increasing nickel accumulation through nikABCDE, and its combination with meropenem improved survival and reduced bacterial burden in mice.

Impact: The study proposes a readily translatable, dual-mechanism strategy to extend the usefulness of an existing last-resort antibiotic against NDM-mediated carbapenem resistance. It advances the field beyond conventional enzyme inhibition by linking antibiotic resensitization to bacterial metal homeostasis.

Clinical Implications: L-histidine could become a candidate adjunct to meropenem or related carbapenems for NDM-producing infections, potentially enabling reuse of existing antibiotics. Clinical translation requires pharmacokinetic and pharmacodynamic studies, confirmation of activity across diverse NDM variants and bacterial species, and controlled human trials.

Future Directions: Future work should define pharmacokinetics, optimal dosing, resistance emergence, activity against diverse NDM variants, and interactions with the human microbiome. Multicenter clinical studies should evaluate safety and efficacy of L-histidine–carbapenem combinations in patients with confirmed NDM-producing infections.

This retrospective population-based study created RADAR-Septic Shock from Catalonia’s administrative hospital database and captured 37,855 septic shock episodes from 2018 to 2024. The system estimated an incidence of 80.28 cases per 100,000 population in 2024, found 36.7% overall in-hospital mortality and 49.5% mortality for hospital-acquired disease, and identified substantial risk-adjusted variation among hospitals and territorial clusters.

Impact: This study demonstrates a scalable infrastructure for population-level sepsis surveillance and outcome benchmarking rather than relying on isolated hospital reports. Its ability to identify temporal trends and interhospital variation can directly support quality improvement, resource planning, and sepsis policy.

Clinical Implications: Health systems can use standardized administrative surveillance and risk-adjusted observed-to-expected mortality ratios to identify hospitals or regions requiring targeted sepsis quality-improvement interventions. The framework may also inform staffing, critical-care capacity, antimicrobial stewardship, and public-health planning, although coding validity and local calibration must be confirmed.

Future Directions: Future studies should validate septic shock coding against clinical records, incorporate physiologic and treatment-timing data, and assess whether benchmarking leads to measurable improvements in mortality and process-of-care indicators. International harmonization could enable cross-system comparisons using common definitions and calibration methods.

This systematic review evaluated 45 studies of AI and machine-learning models for postoperative sepsis and related complications after major abdominal surgery. Although some models showed promising performance, only 15 studies performed strict temporal, geographic, external-center, or independent-investigator validation; calibration and decision-curve analyses were uncommon, and no study prospectively demonstrated that model-guided care improves clinical outcomes.

Impact: The review provides an important negative conclusion: predictive discrimination alone is insufficient to justify clinical deployment of AI for postoperative sepsis. By separating clinical sepsis from related endpoints and emphasizing external validation, calibration, workflow integration, and clinical benefit, it establishes a more rigorous implementation standard.

Clinical Implications: Hospitals should not adopt postoperative sepsis AI prediction tools solely on the basis of development-set AUC or complex algorithms. Before implementation, models should undergo multicenter external validation, calibration assessment, decision-curve analysis, prospective workflow testing, evaluation of alert burden, and demonstration that their use improves patient-centered outcomes.

Future Directions: Future studies should use standardized Sepsis-3 or otherwise clinically justified outcome definitions, preregister model-development and validation protocols, perform multicenter temporal and geographic external validation, report calibration and net benefit, and conduct prospective impact studies. Implementation research should evaluate whether predictions trigger effective interventions without causing alert fatigue or inequitable care.