Abstract
Background
Sepsis is characterized by profound immune dysregulation, yet the biological significance and clinical application of peripheral immune phenotyping in sepsis remain incompletely defined. This study aimed to define internally identified immune phenotypes in sepsis and evaluate their prognostic and translational relevance.
Methods
In this single-centre retrospective cohort study, 374 adult patients with sepsis or severe infection were included. Six peripheral immune markers (total T cells, CD4
+
T cells, CD8
+
T cells, total B cells, NK cells, and monocyte HLA-DR) together with clinically relevant laboratory indicators were analysed to characterize immune and clinical-laboratory heterogeneity in sepsis. Internal validity metrics were used to determine the optimal cluster number. The prognostic value of immune phenotypes was benchmarked against routine laboratory indicators through exhaustive combinatorial clustering and repeated cross-validated logistic regression. Associations with illness severity (SOFA, APACHE II, and NUTRIC) were further assessed.
Results
Two internally stable immune phenotypes were identified: an immune-preserved phenotype (n = 306) and an immune-suppressed phenotype (n = 61). The immune-suppressed phenotype exhibited globally reduced lymphocyte subsets and lower monocyte HLA-DR, accompanied by lower platelet and white blood cell counts, higher inflammatory markers, and slightly higher SOFA scores. However, in-hospital mortality did not differ significantly between phenotypes (39.0% vs 30.6%, *P* = 0.224). Immune phenotyping achieved limited prognostic separation (Δ = 8.9 percentage points), substantially lower than laboratory-based combinations (Δ = 63.3 points). In predictive modelling, the six-marker immune model showed the weakest discrimination (AUC = 0.646 ± 0.066), whereas SOFA alone outperformed all six-marker combinations (AUC = 0.747 ± 0.055). Lactate, as a severity-associated laboratory indicator, demonstrated a stronger prognostic association than immune parameter suggesting that immune suppression and physiological stress represent complementary dimensions of sepsis pathobiology.
Conclusion
Peripheral immune phenotyping identifies biologically distinct immune-preserved and immune-suppressed states in sepsis. Although these immune phenotypes provide limited additional prognostic discrimination beyond established severity scores and routine laboratory indicators, they provide valuable insights into immune dysfunction and represent a potential framework for immune stratification. Further longitudinal studies are needed to determine whether these immune phenotypes can inform future personalized immunomodulatory strategies.