Hemodynamic trajectory phenotypes may modify associations between corticosteroid timing and outcomes in septic shock

Corticosteroid trials in septic shock show conflicting results, with patient heterogeneity and treatment timing as likely contributors. We evaluated whether associations between corticosteroid timing and clinical outcomes differed by hemodynamic trajectory phenotype defined by dynamic time warping clustering of Blood Pressure-Vasopressor Index (BPRI) trajectories across three ICU cohorts. We conducted a retrospective cohort study using MIMIC-IV as the derivation cohort and eICU-CRD and PINES as validation cohorts. Adult patients with septic shock receiving vasopressors were included. Hemodynamic phenotypes were identified from 168-h BPRI trajectories. Corticosteroid timing was classified as early (≤ 6 h) or late (> 6 h) after sepsis onset. In MIMIC-IV, inverse probability of treatment weighting with cause-specific Cox regression was used to estimate associations between timing and 28-day mortality, hospital length of stay, and ventilation duration within each phenotype. Validation cohorts were analyzed using multivariable Cox regression, and cohort-specific estimates were summarized using random-effects meta-analysis with I2. Two phenotypes were identified in MIMIC-IV: Deteriorated (N=2,082, 38%), with declining hemodynamics and higher mortality, and Improved (N=3,361, 62%), with stable or rising trajectories. Similar trajectory patterns were observed in eICU-CRD and PINES. In the Deteriorated phenotype, late corticosteroids were associated with lower 28-day mortality in MIMIC-IV (HR 0.88 [0.78-0.98]) and shorter ventilation duration (HR 0.65 [0.58-0.73]) compared with early administration. The mortality association was not consistently replicated in validation cohorts, whereas the ventilation association was observed in PINES Deteriorated patients. In the Improved phenotype, late steroids showed no consistent mortality or hospital length-of-stay association. Exploratory hemodynamic response analysis showed a more favorable BPRI slope for late versus early steroids in the Deteriorated phenotype. Hemodynamic trajectory phenotypes identified clinically meaningful variation in the relationship between corticosteroid timing and outcomes in septic shock. In the Deteriorated phenotype, delayed corticosteroid initiation was associated with lower 28-day mortality and shorter recovery, whereas the Improved phenotype showed no mortality advantage from delayed treatment. These findings support incorporating dynamic hemodynamic trajectories into future studies of corticosteroid timing in septic shock.