Early Postoperative Intake Linked To Complications After GI Cancer Surgery
Key Takeaways
- Early postoperative energy intake showed a nonlinear association with overall complications from surgery for GI cancer.
- Infectious complications followed broadly similar directional patterns, and exploratory subgroup analyses of the energy intake-overall complication association found no significant interaction by GLIM-defined malnutrition, surgery type, or preoperative neoadjuvant therapy.
The prospective cohort study enrolled 642 adults at Peking University Cancer Hospital from January 2022 through June 2023. Eligible patients underwent elective gastrectomy or partial colectomy for histologically confirmed gastrointestinal malignancy. Nutritional intake during postoperative days 1 and 2 included oral, enteral, and parenteral sources. Primary exposures were energy intake (kcal/kg/day) and protein intake (g/kg/day) using adjusted body weight.
The primary endpoint was overall postoperative complications after postoperative day 2 and within 30 days, while infectious complications were secondary; 84 patients (13.1%) had overall complications, and 48 (7.5%) had infectious complications.
Energy intake showed statistically significant nonlinearity for overall complications, with p for nonlinearity of 0.018. The modeled nadir was 14.4 kcal/kg/day, and the estimated range associated with the lowest predicted risk was 12.8 to 20.1 kcal/kg/day. Protein intake did not show statistically significant nonlinearity, with p equal to 0.166, a nadir at 0.81 g/kg/day, and a modeled range of 0.55 to 1.51 g/kg/day. Generalized additive and quadratic models were supportive sensitivity analyses, and discrimination was modest, with RCS AUC values of 0.644 for energy and 0.617 for protein. Models were adjusted for age, BMI, sex, cancer stage, surgery type, GLIM-defined malnutrition, and neoadjuvant therapy. Overall, the energy signal was clearer than the protein pattern.
For infectious complications, both exposures showed broadly similar directional patterns, but confidence intervals were wider and model discrimination was lower. The infectious outcome analysis did not show statistically significant nonlinearity, with energy p equal to 0.19 and protein p equal to 0.20. Exploratory subgroup analyses of the energy intake-overall complication association did not detect statistically significant interaction by GLIM-defined malnutrition, surgery type, or preoperative neoadjuvant therapy.
The authors described the findings as hypothesis-generating and noted the limits of a single-center observational design, intake-based exposure measurement, possible residual confounding, and constrained generalizability in a cohort with a high laparoscopic proportion. Those considerations limited causal interpretation and broader extrapolation.