Trajectories of Coagulation-Fibrinolysis Balance Reveal Hemostatic Heterogeneity in Sepsis: Derivation and Multicenter Validation

Background: The imbalance between coagulation activation and fibrinolysis drives sepsis-induced coagulopathy and organ dysfunction. However, this dynamic process is highly heterogeneous. We aimed to identify longitudinal coagulation-fibrinolysis trajectory phenotypes and to develop a novel prothrombin time (PT)–D-dimer Z-balance score to quantify this evolving hemostatic imbalance. Methods: This study included a single-center training cohort (n=1672) and a prospective multicenter external validation cohort (n=1393) of adult septic ICU patients. Joint group-based trajectory modeling (GBTM) was applied to daily PT and D-dimer measurements over the first 7 days. Hemostatic, inflammatory and prognostic profiles were compared across phenotypes. Additionally, a daily Z-balance score [Z(PT) − Z(D-dimer)] was constructed to quantify the evolving hemostatic imbalance. Results: Joint modeling reveals three phenotypes: Compensate (Class 1, 51.9%; preserved balance), Coupled (Class 2, 35.5%; intensive but balanced coagulation-fibrinolysis activation), and Consumptive (Class 3, 12.6%; profound coagulation factor consumption with blunted fibrinolysis). The Consumptive phenotype exhibited the most severe systemic inflammation, organ failure, and the highest 28-day mortality (52.1%). Multivariable analysis confirmed Class 3 as an independent predictor of 28-day mortality in both the training (adjusted HR 2.69) and validation cohorts (adjusted HR 3.78). The novel Z-balance score revealed distinct dynamic patterns: Class 3 was characterized by persistently positive yet declining Z-balance (a delayed, partial fibrinolytic response relative to coagulation). Conclusions: Joint longitudinal modeling of PT and D-dimer reveals three distinct coagulation‑fibrinolysis phenotypes in sepsis. The Consumptive phenotype, characterized by profound coagulation and inflammation activation alongside a positive PT-D-dimer Z-balance, identifies a high-risk subpopulation with markedly increased mortality.