This multicenter, randomized controlled trial aims to evaluate whether early initiation of supplemental parenteral nutrition (SPN) (on day 4 of ICU admission) compared to late initiation (on day 8) can reduce 28-day all-cause mortality in critically ill patients. Secondary objectives include assessing the effects of early SPN on long-term functional outcomes (2-year EQ-5D-5L scores), mortality at various timepoints (ICU, 28-day, 90-day, 180-day, and 2-year), ICU length of stay, serum prealbumin levels (days 8 and 28), skeletal muscle measurements (diaphragm thickness, biceps, quadriceps, and rectus femoris cross-sectional area), incidence of ICU-acquired infections, adverse events (e.g., hyperglycemia, hypoglycemia, dyslipidemia, liver dysfunction), and ventilator-free days within 28 days.

Eligible patients are adults (≥18 years) with at least one organ failure (SOFA score ≥2) within 24 hours of ICU admission, expected to remain in the ICU for >72 hours, and unable to reach 50% of caloric target via enteral nutrition (EN) within the first 72 hours. A total of 946 patients will be enrolled and randomized to receive either early SPN (day 4) or late SPN (day 8) in addition to standard EN. Follow-up will include in-hospital assessments and telephone follow-ups at 28, 90, 180 days, and 2 years. The study hypothesizes that early SPN may improve short-term survival and nutritional/muscle status, though its impact on long-term quality of life remains to be determined.

This is a multicenter, randomized, controlled clinical trial designed to evaluate the impact of early supplemental parenteral nutrition (SPN) versus late SPN initiation on clinical outcomes in critically ill patients. The primary objective is to determine whether initiating SPN on day 4 of ICU admission, compared to day 8, can reduce 28-day all-cause mortality. Secondary objectives include assessing the effects of early SPN on functional status (2-year EQ-5D-5L), mortality at various time points (ICU, 28-day, 90-day, 180-day, and 2-year), ICU length of stay, nutritional markers (prealbumin levels on days 8 and 28), muscle mass indicators (diaphragm thickness, biceps, quadriceps, and rectus femoris cross-sectional area), incidence of ICU-acquired infections, ventilator-free days, and metabolic adverse events (such as hyperglycemia, hypoglycemia, hyperlipidemia, and liver dysfunction).

Baseline and clinical process data will be collected for all enrolled patients within the first 24 hours of ICU admission. This includes demographic data (age, sex, height, weight), severity scores (APACHE II, SOFA), and nutritional status.

Additional ICU-related data will be recorded, including:

Use of sedation and analgesia (yes/no), and Richmond Agitation-Sedation Scale (RASS) scores Use of vasopressors (yes/no), type of agent(s), dosage, and duration Vital signs, fluid balance, and mechanical ventilation status Contact information for patient surrogates or family members will be obtained to facilitate follow-up at Day 28, Day 90, Day 180, and 2 years after ICU admission.

Randomization Method A free recruitment model was adopted nationwide. The randomization plan was developed by the study statistician. Dynamic block randomization stratified by participating centers was used to allocate participants to either the early or late SPN group in a 1:1 ratio. The size of the permutation blocks was variable and blinded to the treating clinicians and nursing staff.

Statistical Analysis Continuous variables will be presented as mean ± standard deviation or median (interquartile range [IQR]) depending on distribution characteristics and compared using the independent samples t-test or Wilcoxon rank-sum test. Categorical variables will be reported as counts (percentages) and compared using Pearson's chi-square test or Fisher's exact test as appropriate. The Kaplan-Meier method and log-rank test will be used to compare 2-year survival between groups.

To assess whether baseline nutritional risk modifies the effect of the intervention, interaction terms will be incorporated into multivariate logistic regression models (for mortality) and linear regression models (for physical function). The significance of interaction terms will be tested using the likelihood ratio test.

Sensitivity analyses will treat NRS-2002 and mNUTRIC scores as ordinal variables and age as a continuous variable, repeating the regression models. Additionally, the Random Forest (RF) algorithm will be applied to explore potential nonlinear interactions between nutritional risk and intervention effects.

All statistical analyses will be conducted using JMP software (SAS Institute Inc., Cary, NC, USA). A two-sided P-value < 0.05 will be considered statistically significant, while the threshold for interaction terms will be set at P < 0.1 to enhance detection sensitivity.

Sample Size Calculation Sample size estimation was performed using PASS software. Based on previous literature, the 28-day mortality rate was assumed to be 8% in the early SPN group and 14% in the late SPN group. Assuming a 1:1 allocation ratio, a two-sided α of 0.05, and a statistical power (1-β) of 0.90, 430 patients were required per group. Accounting for a 10% dropout rate, the final total sample size was set at 946 patients.

Inclusion Criteria：

1. Age ≥ 18 years
2. Failure of one or more organ systems within 24 hours of ICU admission, defined as Sequential Organ Failure Assessment (SOFA) score ≥ 2 in any single organ system
3. Expected ICU stay > 72 hours
4. Able to initiate early enteral nutrition (EN) within 48 hours of ICU admission
5. Enteral nutrition providing <50% of target caloric requirements within 72 hour Exclusion Criteria

1.Full oral intake or tolerance of full-dose EN; no need for SPN 2.Receiving palliative care or expected to die within 72 hours 3.Pregnant or lactating women 4.Long-term use of systemic corticosteroids or other immunosuppressive agents 5.Known malignancy currently receiving radiotherapy or chemotherapy 6.Contraindications to parenteral nutrition (PN) 7.Participation in another interventional clinical trial

Participants are blinded to group assignment to reduce expectation bias. The clinical staff administering the nutrition intervention are not blinded due to the nature of the intervention.