Correlation analysis between preoperative inflammation-nutrition-related indices and the risk of pancreatic fistula formation following distal pancreatectomy

Objective To investigate the associations of preoperative C-reactive protein-to-albumin ratio (CAR), inflammatory burden index (IBI), and C-reactive protein-albumin-lymphocyte (CALLY) index with clinically relevant postoperative pancreatic fistula (CR-POPF) after distal pancreatectomy (DP), and to compare their predictive value, providing a reference for preoperative risk assessment. Methods We retrospectively included 300 patients who underwent DP between June 2022 and June 2024. CR-POPF was defined according to the 2016 International Study Group of Pancreatic Surgery (ISGPS) criteria as Grade B or Grade C POPF. Multivariable logistic regression was used to evaluate the associations of CAR, IBI, and CALLY with CR-POPF, and restricted cubic splines were used to assess dose-response relationships. Predictive performance of the preoperative baseline model and models incorporating each index was compared using receiver operating characteristic (ROC) analysis with DeLong tests, calibration assessment, Brier scores, decision curve analysis, and 1,000-resample bootstrap internal validation. Results CR-POPF occurred in 84 of 300 patients (28.0%); among these events, 81 (96.4%) were Grade B and 3 (3.6%) were Grade C. In preoperative-adjusted analyses, higher CAR (odds ratio [OR] 230.244, 95% confidence interval [CI] 45.053–1176.652; P < 0.001) and IBI (OR 1.015, 95% CI 1.006–1.024; P < 0.001) were associated with higher odds of CR-POPF, whereas higher CALLY was associated with lower odds (OR 0.979, 95% CI 0.967–0.991; P = 0.001). The baseline model had an area under the ROC curve (AUC) of 0.786 (95% CI 0.726–0.845). Adding CAR increased the AUC to 0.894 (95% CI 0.857–0.931; P < 0.001 vs baseline), compared with 0.821 for IBI (P = 0.004) and 0.827 for CALLY (P = 0.016). Bootstrap-corrected AUCs were 0.763 for the baseline model, 0.879 for baseline+CAR, 0.796 for baseline+IBI, and 0.805 for baseline+CALLY. The CAR-extended model also had the lowest Brier score (0.120) and showed the greatest net benefit across much of the examined threshold range. Conclusions Preoperative CAR, IBI, and CALLY were associated with CR-POPF after DP. Among the three indices, CAR showed the strongest predictive performance when added to the preoperative clinical model. These findings may provide a reference for preoperative risk assessment, although external validation is needed.