From March 10, 2020, to September 15, 2022, we conducted a single-center, prospective, double-blind, randomized control trial at Zhongshan Hospital, Fudan University, China. The trial protocol and statistical analysis plan are available online (Additional File 1). Ethical approval (B2019-279) was obtained from the Ethics Committee of Zhongshan Hospital. Written informed consent was obtained from all the participants. The trial was registered at ClinicalTrials.gov (ChiCTR2000029635) and followed the Consolidated Standards of Reporting Trials guidelines. Trial oversight was monitored by the Shenkang Development Center, which was independent of the investigators.

Patients aged 65–80 years with GC, a body mass index of 18.5 to 28.0 kg/m2, and an American Society of Anesthesiologists physical status class I–II who were scheduled for curative resection were eligible for the trial. Patients were excluded if they had received preoperative chemoradiotherapy; had metabolic disease (i.e., diabetes or thyroid disease), previous major gastrointestinal or other major surgeries, combined metastatic cancer, or cachexia (weight loss > 20% in the last month or an albumin level < 30 g/L) affecting recovery; or had undergone surgical procedure and experienced pathological diagnosis changes. The detailed exclusion criteria are described in the trial protocol.

Eligible patients were randomly assigned using computer-generated random numbers (SPSS 22.0; IBM SPSS Inc., Armonk, NY, USA). An independent researcher concealed the allocations in consecutively numbered, sealed, opaque envelopes until informed consent was obtained. The physician in the pharmacy unit at the intravenous drug allocation center prepared the PN infusions and subsequently delivered them in unified PN bags of uniform volume and labeling to the unknown investigators. No visual or physical differences were detectable that could reveal the treatment allocation. Only the pharmacist was aware of the group allocation and acted as the unblinded investigator. The anesthesiologist, surgeon, patients, trained follow-up investigators, and data collectors were blinded to the treatment allocation. The statistician was independent and blinded to the interventions.

This trial was based on ERAS protocols (Additional Table 1 in Additional File 2) which were implemented by a multidisciplinary team.

An epidural catheter was inserted between thoracic vertebral levels T8 and T11 before the induction of general anesthesia. Following successful internal jugular vein and radial artery catheter access, general anesthesia was induced with sufentanil 0.2 μg/kg, plasma target-controlled infusion (TCI) of propofol 3 μg/mL, plasma TCI of remifentanil 3 ng/mL, and rocuronium 0.6 mg/kg, and maintained by combined intravenous–inhalation anesthesia. Epidural blockade was established with 0.375% ropivacaine 6–10 mL and maintained with 0.375% ropivacaine at 3–5 mL/h during surgery.

Using standard intraoperative monitoring, the mean arterial pressure and heart rate were controlled within ± 20% of baseline values, and end-tidal CO2 was maintained at 30–40 mmHg. Balanced acetated Ringer’s solution was administered at 5–7 mL/kg/h. Colloids were initiated in patients with significant blood loss but without renal impairment. A warm air heater was used to maintain perioperative normothermia.

Postoperative pain management was achieved by patient-controlled epidural analgesia, including 0.12% ropivacaine and 0.4 μg/mL sufentanil, targeting a resting pain score of < 4 (score = 0 [no pain], score = 10 [worst pain imaginable]). Postoperative care followed the local clinical practice.

The formulated solutions have already been marketed, and we formulated the intervention solution according to the drug instructions. The pharmacy at the intravenous drug allocation center prepared the PN solutions according to standard operating procedures.

The BAF formulation consisted of the following components: 780 mL BCAA solution (3AA), which included 3.38 g L-isoleucine, 4.12 g L-leucine, and 3.15 g L-valine per 250 mL, with the national medicine permission number (NMPN) H19993799; 100 mL arginine hydrochloride solution (5 g per 20 mL), with NMPN H31021692; 120 mL alanyl-glutamine solution (10 g per 50 mL), with NMPN H20053877; and 2 mL vitamin B6 solution (0.1 g per 2 mL), with NMPN H41021737.

The solutions were subsequently administered via a central venous catheter. Intraoperatively, patients in the high-proportion BCAA (HBCAA) and control groups received 500 mL BAF and 500 mL Ringer’s solution, respectively, within 2 h after induction. Postoperatively, patients in the HBCAA group were administered 1000 mL BAF and 100 mL dextrose (50%), whereas the control group patients were administered 1000 mL dextrose (5%) and 100 mL sodium chloride (0.9%). Therefore, the dose and intervention volume of dextrose were the same across the two groups. From postoperative day (POD) 1, patients received 1100 mL of the intervention solutions at a rate of 200 mL/h, which was reduced to 550 mL during fluid diet initiation until semi-liquid intake or hospital discharge (Additional Figure 1 in Additional File 2). Additional nutrients, including potassium chloride injection and compounded vitamin injection, were supplemented by assessing each patient’s fluid status, electrolyte levels, and nutritional needs. Based on these assessments, the attending physician adjusted it to ensure that both groups received the recommended PN.

The primary outcome was the standardized LOS, defined as the time from surgery to discharge, based on the following criteria: stable vital signs, good wound healing without infection or considerable pain, free ambulation, sufficient oral intake of a semi-liquid diet, and no requirement for intravenous therapy.

The secondary outcomes were indicators closely related to primary outcomes, clinically significant to reflect short-term postoperative prognosis, be measurable and have easily accessible data, including the time to first flatus and defecation (postoperative gastrointestinal function recovery time), time to liquid and semi-liquid diet initiation, postoperative ambulation, weight loss (from baseline to POD 5), frequency and intensity of postoperative nausea and vomiting (PONV), time to gastrointestinal tube and urinary catheter removal, major postoperative complications (anastomotic leak, arrhythmia, congestive heart failure, respiratory failure, pneumonia, wound infection, bleeding, renal failure, intestinal obstruction, and deep vein thrombosis), hospitalization cost, readmission, and mortality rate within 28 days postoperatively.

The levels of the following indicators were measured preoperatively and on POD 1, 3, and 5 (at 6:00 am): albumin, prealbumin, blood glucose, free fatty acids (FFAs), CD4 + /CD8 + ratio, interleukin (IL)-6, D-lactate, alanine aminotransferase, aspartate aminotransferase, total bilirubin, blood urea nitrogen, serum creatinine, urinary 3-methylhistidine (3-MH), and urinary creatinine. The ratio of urinary 3-MH to creatinine was used as a nutritional indicator of protein metabolism. Blood D-lactate, urinary 3-MH, and urinary creatinine levels were analyzed using a D-lactate colorimetric assay kit (MAK058; Sigma-Aldrich Co. LLC, St. Louis, MO, USA), a urinary 3-MH ELISA kit (abx257295; Abbexa LTD, CB4 0GJ, UK), and a urinary creatinine colorimetric assay kit (500701; Cayman Chemical, Ann Arbor, MI, USA), respectively. The remaining indicators were measured at the laboratory at the study site.

Follow-up personnel assessed the clinical indicators related to the patient’s outcomes twice daily (8:00 am and 4:00 pm). After discharge, complications and readmission status were monitored via a follow-up phone call 28 days post-surgery.

Unexpected experiences observed in patients during the trial, whether or not considered related to the solutions, were reported as adverse events. Events and rescue interventions were recorded and reported according to the adverse event procedures. Serious adverse events were reported individually to the sponsor, the Ethics Committee, and the health administrative department within 24 h.

Previous studies revealed that the overall standardized LOS after surgery for gastrointestinal tumors was 7.7 days, with a standard deviation of approximately 1.7 days. Moreover, the between-group difference was clinically valuable if it reached 1 day. Hence, we used a targeted standardized LOS decrease of 1 day, a two-tailed type I error rate of 5%, and a statistical power of 80% to obtain a sample size of 65 patients in each group. The sample size was increased to 150 to accommodate participant withdrawal and loss to follow-up.

The full analysis set adhered to the intention-to-treat principle. Variables were reported as number (percentage), mean (standard deviation), or median (interquartile range). The normality of continuous data was assessed using the Shapiro–Wilk test. Between-group comparisons were performed using the chi-square or Fisher exact test and the Mann–Whitney U test for categorical and continuous variables, respectively. Time-based measurements within each group were compared using a repeated-measures analysis of variance.

For each hypothesis, a two-tailed P < 0.05 was considered statistically significant. Statistical analyses were conducted using SPSS (version 27.0). The sample size was estimated using the Power Analysis and Sample Size Statistical Software (Version 11, Kaysville, UT, USA).