Early versus delayed enteral nutrition in septic shock: a target trial emulation study

The optimal timing of enteral nutrition (EN) initiation in septic shock remains uncertain. Although early EN is recommended for critically ill patients in current guidelines, evidence specific to septic shock is limited and inconsistent. Observational studies evaluating treatment timing are also vulnerable to immortal time bias and time-dependent confounding. Target trial emulation offers a structured approach to estimating trial-like effects from observational data while reducing these biases. We therefore emulated a target trial to assess the association between EN timing and clinical outcomes in septic shock and to explore whether this association varied with vasopressor exposure.

We conducted a retrospective target trial emulation using the MIMIC-IV critical care database. Adult patients admitted to the ICU who met Sepsis-3 criteria and developed septic shock within the first 96 h after ICU admission were eligible. Two treatment strategies were defined according to the timing of EN initiation after ICU admission: early EN (0-48 h) and delayed EN (48-96 h). A clone-censor-weight (CCW) framework was used to emulate randomization. Patients were cloned at baseline and censored when their observed treatment course deviated from the assigned strategy. Stabilized inverse probability weights were estimated using pooled logistic regression models incorporating baseline characteristics and lagged time-varying clinical variables measured at 6-h intervals. The primary outcome was 28-day mortality. Weighted Cox proportional hazards models were used to estimate hazard ratios (HRs) comparing delayed versus early EN. Secondary outcomes included ventilator-free days within 28 days (VFD28). Interaction analyses evaluated whether norepinephrine exposure modified the association between EN timing and mortality.

A total of 4,003 patients with septic shock were identified. Within the first 96 h after ICU admission, EN was initiated in 1,301 patients, including 575 within 48 h and 726 between 48 and 96 h. In the weighted CCW analysis, delayed EN was associated with an increased risk of 28-day mortality compared with early EN (HR 1.21, 95% CI 1.15-1.27; p < 0.001). The estimated 28-day mortality risk was 30.0% in the early EN group and 35.0% in the delayed group. The weighted Kaplan-Meier curves separated early after ICU admission and remained apart over 28 days. Delayed EN was also associated with fewer ventilator-free days (mean difference -1.75 days, 95% CI -2.36 to -1.04). A significant interaction was observed between EN timing and norepinephrine exposure (p for interaction < 0.001). The excess mortality associated with delayed EN was greatest among patients receiving lower vasopressor doses and became less pronounced at higher levels of vasopressor support. Sensitivity analyses were broadly consistent with the primary findings.

In this target trial emulation study of septic shock, initiation of enteral nutrition within 48 h was associated with lower 28-day mortality and more ventilator-free days than initiation between 48 and 96 h. The association between EN timing and survival varied according to vasopressor exposure, with the greatest apparent benefit of early EN observed among patients receiving lower levels of vasopressor support. These findings suggest that the timing of enteral nutrition in septic shock may need to be tailored to the degree of hemodynamic instability.