BackgroundSepsis induces profound derangements in the immune system, including lymphopenia, which correlates with immunosuppression and poor prognosis. However, most evaluations of immunosuppression in sepsis patients rely on static, sporadic lymphocyte counts, lacking dynamic modeling over the disease course. This study aimed to apply latent class mixed modeling on longitudinal lymphocyte counts to uncover heterogeneous trajectory phenotypes in sepsis patients and assess their predictive value for clinical outcomes.ResultsFour lymphocyte trajectory phenotypes were identified in the retrospective cohort (n=2,149) and externally validated (n=2,388): high–declining (α, 3.8%), stable–medium (β, 69.3%), high–increasing (γ, 3.2%), and stable–low (δ, 23.8%). The α phenotype exhibited the highest disease severity and mortality (25.9%) compared with other phenotypes in both cohorts. In the prospective cohort (n=1,056), all lymphocyte subset counts differed among phenotypes on admission (P <.001) and were lower in non-survivors (P<.05). Multivariable regression demonstrated that age, Acute Physiology and Chronic Health Evaluation-II score, heart rate, natural killer cell count, infection source, and lymphocyte trajectory phenotype were independent predictors of 28-day mortality. A nomogram combining these variables provided individualized risk estimations.ConclusionsThe lymphocyte trajectories delineated novel dynamic phenotypes associated with divergent sepsis outcomes. Incorporating longitudinal trajectory modeling and lymphocyte subsets may improve prognostic risk assessment and guide the selection of immunotherapies tailored to specific immune phenotypes in sepsis patients.Clinical trial registrationhttps://www.chictr.org.cn/showproj.aspx?proj=18277, identifier ChiCTR-40 ROC-17010750.
Read full abstract