The medical data of consecutive patients who underwent PD between January 2020 and July 2024 in the Department of Pancreatic Surgery, Nanjing Drum Tower Hospital, were collected in this retrospective cohort study. Patients were classified into EEN combined with the SPN group and the PN group. Demographics, preoperative and postoperative laboratory tests, and postoperative complications were collected. This study was approved by the revised Declaration of Helsinki and approved by our hospital’s Ethics Committee (2024–786).

The inclusion criteria were: (1) age ≥18 years, (2) complete clinical data, (3) underwent conventional PD, (4) no evidence of locally unresectable or other active cancers at diagnosis, and (5) Nutritional Risk Screening 2002 (NRS-2002) score ≥3.

The exclusion criteria were: (1) inflammatory bowel disease, (2) combined other organ resections, (3) severe preoperative infections, and (4) severe renal dysfunction.

All PDs were operated by two experienced surgical teams (The same surgeons participated in both surgical teams, and surgeons were cross-trained). All patients underwent either a conventional Whipple’s procedure with digestive reconstruction by Child’s approach, which included pancreatic-intestinal (Blumgart anastomosis), biliointestinal (hepatic ducts and jejunum with consecutive anterior and posterior wall stitches), and gastroenterostomy (Billroth II anastomosis). Standard perioperative management was implemented for all patients. In this study, all patients received a prophylactic intravenous injection of ceftriaxone, 30 min prior to surgery, continuing for 48 h postoperatively. Drain amylase, bacterial smear, and microbiological culture with antibiotic sensitivity were conducted on postoperative days (PODs) 1, 3, 5, and 7. A full abdominal enhanced computed tomography (CT) scan was performed on POD 7. The clinical team made individualized decisions about drainage tube removal based on symptoms, CT results, and biochemical indicators. Patients identified as with nutritional risk preoperatively received oral intact protein-based enteral nutrition supplements, providing approximately 300–500 kcal/day for 3–5 days.

In patients in the EEN + SPN group, a gastric tube was routinely placed intraoperatively. During surgery, nasojejunal nutrition tubes were placed in the jejunum to facilitate postoperative EN delivery. EN was initiated within 24 h following abdominal surgery according to standard protocols based on ESPEN guidelines. Within this time frame, a 5% glucose and 0.9% sodium chloride solution was administered at a rate of 1.25 to 1.67 mL/kg/h via the nasojejunal feeding tube. On POD 2, EN was provided at a low rate, 500 mL of enteral formula (providing 1 kcal/mL of energy, comprising 16% protein, 35% fat, and 49% carbohydrates), and 250 mL of 5% glucose and 0.9% sodium chloride solution were administered. On POD 3, 500–1,000 mL of EN was administered, and the PN formula was adjusted based on the amount of EN given to meet the remaining energy requirements.

In the PN group, the gastric tube was not routinely placed, and central venous catheters (CVCs) were placed after anesthesia. The PN included intravenous lipid emulsions, micronutrients, amino acids, glucose, and vitamins that were administered continuously, 24 h a day for up to 7 days from the first day after PD. The total caloric intake was fixed at 25–30 kcal/kg/d, the proportion of protein calories with a nitrogen ratio of 120–150:1. Clinical pharmacists adjusted the caloric and protein supplements, as well as fluid and electrolyte levels, based on the patient’s weight, laboratory indicators, and dietary intake. The PN was administered via CVCs. As the patient gradually resumed adequate oral intake, the PN was tapered off.

The clinical data from medical records included demographics (age, sex, comorbidity, and body mass index); preoperative jaundice; preoperative biliary drainage; the Global Leadership Initiative on Malnutrition (GLIM) criteria; and preoperative laboratory data (total protein [TP], serum albumin [Alb], prealbumin [PA], hemoglobin [Hb], and bilirubin). Additionally, intra-operative variables were recorded, including operating time, volume of blood loss, vessel resection, pancreatic texture, and surgical method. Postoperative nutritional indices, such as TP, Alb, and Hb, were also assessed, along with the duration of postoperative hospital stay, hospitalization costs, and pathological diagnosis.

The primary outcome was the incidence of postoperative complications. If patients experienced multiple complications, they were recorded only once according to the highest-grade complication based on the Clavien-Dindo classification system, and severe complications were defined as grade≥III (16). The secondary outcomes encompassed a comparison of postoperative complications between the two groups, including DGE, BL, CL, AP, PPH, and infectious complications. Clinically relevant postoperative pancreatic fistula (CR-POPF, grade B/C), biliary leak (BL), postoperative acute pancreatitis (PPAP), delayed gastric emptying (DGE), chylous fistula (CL), and post-pancreatectomy hemorrhage (PPH) were diagnosed based on the International Study Group for Pancreatic Surgery (ISGPS) (17–21). Exploratory outcomes included postoperative nutritional indices TP, Alb, and Hb. The analysis also encompassed wound infection, bacteremia, intra-abdominal infection, urinary tract infection, and pneumonia (22).

All statistical analyses were carried out using SPSS 27.0 software and R version 4.0. Categorical variables were reported as frequencies and percentages, and with the X² or Fisher’s exact test applied for the comparison. Quantitative variables were characterized by mean±standard deviation or median (interquartile range, IQR) contingent upon the assessment of normal distribution, and t-test or Mann–Whitney U-test was used for the analyses. To assess the effect of nutrition therapy on clinical outcomes, logistic regression was utilized. Odds ratio (OR) with 95% confidence interval (CI) and p-values were documented. A p-value of less than 0.05 was considered statistically significant.

To mitigate potential confounding effects arising from baseline differences between the two cohorts, we implemented a propensity score-matched (PSM) analysis approach to adjust for baseline characteristics. We matched the two groups in a 1:1 ratio with a caliper width of 0.2. The standardized mean difference (SMD) of less than 0.2 was used to examine the degree of PSM. With an SMD threshold of <0.20, indicating a successful balance between groups and absence of significant inter-group differences (Figure 1). To assess the robustness of our findings, we conducted a sensitivity analysis using inverse probability of treatment weighting (model 2) to mitigate the impact of confounding variables and reduce the likelihood of selection bias. Among patients in the EEN + SPN group, we examined the association between POD 3 EN kcal/kg/d and postoperative complications using

logistic regression with restricted cubic splines (RCS). Splines were used for four knots at the 5th, 35th, 65th, and 95th percentiles of POD3EN to allow for potential non-linearity. A two-proportion Fisher’s exact test was performed using PASS 2025 software to evaluate statistical power (α = 0.05). The analysis included 116 patients in the treatment group (event proportion: 0.034) and 132 in the control group (event proportion: 0.160), yielding a proportion difference of −0.126. The resulting statistical power was 0.8818, indicating an 88.2% probability of detecting a true difference between groups and confirming the high sensitivity of the study design.

A total of 248 patients who underwent PD were initially enrolled in the study. Patients were divided into two groups based on perioperative nutritional therapy. The characteristics of all patients are shown in Table 1. There were 116 (46.8%) patients in EEN + SPN group, 132 (53.2%) in PN group. The study cohort consisted of 143 males (57.7%) and 105 females (42.3%), with an average age of 66.5 ± 9.8 years. Before surgery, the levels of total direct bilirubin (DB) and bilirubin (TB) were 10.7 (2.3, 86.1) umol/L and 24.0 (10.5, 116.5) umol/L, respectively. A total of 111 (44.8%) were diagnosed with preoperative jaundice, and 69 (27.8%) underwent preoperative biliary drainage (PBD). Postoperative complications occurred in 173 patients (69.8%). Postoperative severe complications (Clavien-Dindo grade≥III) occurred in 25 patients (10.1%), intra-abdominal infections were observed in 103 patients (41.5%), and CR-POPF developed in 50 patients (20.2%).

As shown in Table 2, the PN group had a longer operative time, a higher preoperative incidence of GLIM-defined malnutrition, and a substantial number of PPPD compared to the EEN + SPN group. Additionally, there were statistically significant differences in rates of vascular resection and pancreatic texture between the groups. To adjust for baseline confounders, a 1:1 PSM analysis was performed. After PSM, the standardized mean differences for all preoperative covariates were less than 0.20. The final cohort comprised the EEN + SPN group (59 patients) and the PN group (59 patients). All baseline characteristics were comparable after PSM. Details on sensitivity analyses are provided in the Supplementary Appendix, Supplemental Digital Content 1, https://10.6084/m9.figshare.28559741.

No statistically significant differences were observed in the incidence of complications between the two groups, either before or after PSM (all p > 0.05, Table 3). Before PSM, the incidence of severe postoperative complications was 3.4% in the EEN + SPN group, compared to 15.9% in the PN group (p = 0.002, Table 3). After PSM, the incidence of severe complications was 5.1% in the EEN + SPN group and 20.3% in the PN group (p = 0.027, Table 3). Both before and after PSM, the incidence of severe complications was significantly higher in the PN group. Before PSM, the incidence of intra-abdominal infection was significantly higher in the PN group (34.5% vs. 47.7%, p = 0.035). However, after PSM, the difference was no longer statistically significant (33.9% vs. 42.4%, p = 0.448). Both before and after PSM, there was no significant difference in CD-I complications between the EEN + SPN and PN groups (before PSM: 7.8% vs. 9.8%, p = 0.724; after PSM: 10.2% vs. 3.4%, p = 0.600). Similar results were observed for CD-II complications (before PSM: 54.3 vs. 47.7%, p = 0.364; after PSM: 55.9% vs.39.0%, p = 0.097) and CR-POPF (before PSM: 17.2% vs. 22.7%, p = 0.283; after PSM: 22.0% vs. 20.3%, p = 1.000). No significant differences were found in the remaining complications between the two groups (Table 3).

The body weight of the patients in this study ranged from 40 to 86.5 kg. On POD 3, patients in the EEN + SPN group received a mean (SD) energy intake of 30 (4.5) kcal/kg/d, in contrast to the PN group, which had a mean (SD) energy intake of 23.4 (6.0) kcal/kg/d. On POD 7, patients in the EEN + SPN group received a mean (SD) energy intake of 22.3 (5.5) kcal/kg/d, while those in the PN group received a mean (SD) energy intake of 21.5 (5.5) kcal/kg/d. Throughout this period, the mean (SD) protein intake was 1.3 (0.3), 1.0 (0.2) g/kg/day in the EEN + SPN group and 1.0 (0.2), 0.8 (0.2) g/kg/day in the PN group. Before and after PSM, the average daily energy and protein intake for both groups 1 week postoperatively is provided in Figure 2. The period on POD 3 represented a dynamic titration phase, with caloric intake for all patients stabilizing within the target range (25–30 kcal/kg/day) at POD 5. In a longitudinal analysis of patients receiving nutritional support, RCS regression found no evidence of a linear or non-linear dose–response relationship between enteral nutrition and the risk of complications. The results showed that neither the linear term (coefficient = −0.485, SE = 0.608, p = 0.427) nor the two non-linear spline terms (3.095, SE = 3.197, p = 0.335; −9.851, SE = 10.761, p = 0.362) were statistically significant, indicating no reliable evidence for either a linear or non-linear association. Further analysis using piecewise logistic regression (breakpoint = 8.224) to compare single-slope and dual-slope models showed no significant effects on either side of the breakpoint (<8.224: OR = 0.94, 95% CI: 0.60–1.47, p = 0.784; ≥8.224: OR = 1.18, 95% CI: 0.79–1.76, p = 0.415). Additionally, dose-based ROC analysis demonstrated limited discriminative ability (AUC = 0.550, 95% CI: 0.439–0.661). (Figure 3, Supplementary Table S3).

The trends in postoperative nutritional indicators for both groups are shown in Figure 4. Mean (SD) POD 1 TP (55.2 [5.7] vs. 53.4 [4.7], p = 0.010), Hb (114.3 [15.6] vs. 108.9[14.6], p = 0.005); mean (SD) POD 5 TP (57.7 [6.7] vs. 54.9 [5.4], P < 0.001); mean (SD) POD 7 TP (60.6 [8.0] vs. 56.6[6.6], p = 0.008), Alb (35.0 [3.0] vs. 33.9 [3.4], P < 0.001), were significantly higher in the EEN + SPN group than in the PN group. However, after PSM, none of the nutritional indicators showed statistically significant differences between the two groups. (All nutritional indicator analyses are shown in Supplementary Table S4.)