Enteral Feeding Timing in Sepsis

Evidence

Two recent retrospective cohorts examined when enteral nutrition began during sepsis, but they asked the question in different populations and used different clocks. Their adjusted results point in different directions, making the comparison useful for understanding both the signal and its uncertainty.

The newer study followed 2,205 adults with sepsis in a tertiary-hospital intensive care unit in Hebei, China, from 2015 through 2024. All had started enteral nutrition within seven days of admission. The investigators classified 1,500 patients as receiving early nutrition within two days and 705 as receiving delayed nutrition from day two through day seven. They used propensity-score matching, multivariable Cox regression, and inverse-probability weighting to reduce measured differences between groups.

In that cohort, early enteral nutrition was associated with higher mortality at 28 days (hazard ratio 1.44, 95% confidence interval 1.08 to 1.92) and 60 days (hazard ratio 1.45, 95% confidence interval 1.11 to 1.89). Gastric retention was also more frequent in the early group (odds ratio 1.77, 95% confidence interval 1.14 to 2.79). The association appeared more pronounced in several subgroups, including patients with severe sepsis, abdominal infection, or vasopressor support at a norepinephrine-equivalent dose of at least 0.1 micrograms per kilogram per minute. These are associations from an abstract-level evidence source, not proof that earlier feeding caused harm.

The second study used the Chinese Database in Intensive Care at Zhongda Hospital and focused more narrowly on septic shock. It included 1,855 adults who received vasopressors within 24 hours of ICU admission between 2014 and March 2025. Early enteral nutrition meant starting within the first 24 hours; 578 patients met that definition, while 1,277 started later. The study excluded patients who died within 48 hours and those with gastrointestinal surgery, active bleeding, obstruction, bowel ischemia, or another enteral-feeding contraindication.

Before adjustment, 28-day mortality was 16.6% in the early group and 21.1% in the later group. The groups differed at baseline, however. Early-fed patients were older, had slightly higher APACHE II scores, more pneumonia and invasive ventilation, and lower vasopressor intensity. After adjustment for age, sex, body mass index, vasopressor intensity, APACHE II score, SOFA score, and lactate, early feeding was not significantly associated with lower 28-day mortality (hazard ratio 0.832, 95% confidence interval 0.654 to 1.059; p=0.135).

Vasopressor-stratified results also remained uncertain. Below 15 micrograms per minute in norepinephrine equivalents, early feeding showed a nonsignificant trend toward lower mortality (hazard ratio 0.738, 95% confidence interval 0.523 to 1.042; p=0.084). At or above that threshold, there was no discernible association (hazard ratio 0.926, 95% confidence interval 0.660 to 1.298; p=0.655). The early group had longer ICU and hospital stays, a result the authors noted could partly reflect baseline differences and the competing effect of earlier deaths shortening stays in the later group.

Analysis — Timing May Reflect Clinical Selection

The cross-study pattern is not a simple contest between “early” and “late” feeding. It is an analysis signal that the label “early” may capture different clinical decisions in different cohorts. One study compared feeding within two days against feeding during days two through seven across adults with sepsis; the other used a 24-hour boundary in patients already receiving vasopressors for septic shock. The second cohort also excluded deaths within 48 hours and several gastrointestinal conditions, while the first abstract describes a broader group that began nutrition within seven days. Those design choices can change who is eligible to appear in either timing group.

The studies nevertheless converge on one narrower point: vasopressor exposure and hemodynamic severity are important modifiers or confounders when feeding timing is compared with mortality. The Hebei cohort found a stronger adverse association among patients needing at least 0.1 micrograms per kilogram per minute of norepinephrine equivalent. The Zhongda cohort found no significant adjusted mortality association overall and no benefit signal at higher vasopressor intensity. A plausible, still unproven interpretation is that feeding time partly records clinicians’ assessment of stability and gastrointestinal tolerance rather than acting as an isolated exposure. Randomized evidence designed around clearly defined stability, nutrition dose, and tolerance endpoints would be needed to separate that selection process from any effect of timing itself.

Limitations

Both studies were retrospective, observational, and based at single centers in China, so neither can establish causality or guarantee that the findings transfer to other ICUs. Propensity methods and regression address measured variables but cannot remove unmeasured confounding or confounding by indication. The timing definitions were not equivalent, and the source available for the newer Hebei cohort was limited to its abstract.

Important nutrition details were also incomplete. The Zhongda study did not analyze total calorie or protein delivery, and its vasopressor-intensity measure may not represent exact pharmacologic equivalence for every patient. Its exclusion of deaths within 48 hours can introduce survivor selection, while subgroup estimates were imprecise and did not reach statistical significance. The Hebei subgroup findings likewise need prospective confirmation. Taken together, these studies identify a research problem and possible effect modifiers; they do not establish a universal feeding-timing rule or a patient-level treatment recommendation.