Beyond the sepsis label: heterogeneity, subphenotypes, and the path to precision therapy

Sepsis is increasingly recognized as a biologically heterogeneous syndrome rather than a single uniform inflammatory condition. Patients meeting the same clinical definition may differ substantially in immune competence, endothelial injury, metabolic adaptation, organ vulnerability, temporal trajectories, and likelihood of response to host-directed therapies. This narrative review synthesized relevant literature identified through PubMed, Scopus, Google Scholar. Priority was given to landmark clinical and translational studies, consensus papers, methodological articles, randomized trials, secondary trial analyses, and high-quality reviews directly related to sepsis heterogeneity, subphenotypes, endotypes, immune dysfunction, multi-omics classification, machine learning-assisted stratification, and treatment-effect heterogeneity. The reviewed evidence shows that sepsis heterogeneity is reproducible across bedside clinical phenotypes, longitudinal trajectories, immune-cell-defined endotypes, transcriptomic, proteomic, metabolomic, metagenomic, and machine learning-derived classifications. Clinical models, including early pragmatic phenotypes and dynamic trajectories, are scalable and close to bedside use, whereas omics-based approaches provide deeper biological resolution but remain limited by cost, turnaround time, platform dependence, and incomplete concordance across layers. Immune low-response states, endothelial/coagulation injury, metabolic stress, organ-specific vulnerability, and PIICS-related non-resolution emerged as recurrent biological motifs. Evidence for phenotype-guided therapy, including corticosteroids, and precision immunotherapy, remains promising but largely exploratory, with most treatment-response signals requiring prospective validation. Sepsis subphenotyping should be interpreted as a staged framework for precision decision-making rather than as a competition between rival taxonomies. Future progress requires reproducible, time-sensitive, clinically measurable treatable traits that can be prospectively tested to guide therapy, organ support, trial enrichment, and long-term recovery research.