Dual-energy CT-derived fibrosis biomarkers predict pancreatic fistula after pancreaticoduodenectomy: a collagen proportional area mediation analysis.

PURPOSE: To investigate whether dual-energy CT-derived iodine concentration ratio and extracellular volume fraction (ECV) predict clinically relevant postoperative pancreatic fistula (CR-POPF) after pancreaticoduodenectomy, and whether their associations with CR-POPF are statistically consistent with histologically quantified pancreatic fibrosis.

MATERIALS AND

METHODS

This retrospective single-center study included 133 patients who underwent pancreaticoduodenectomy with preoperative DECT between November 2022 and September 2025. DECT parameters and collagen proportionate area (CPA) at the resection margin were assessed. Logistic regression and ROC analyses identified predictors of CR-POPF. Bootstrap internal validation assessed optimism-corrected discrimination and calibration. Spearman correlation and exploratory mediation analyses evaluated relationships among DECT parameters, CPA, and CR-POPF.

RESULTS

Among 133 patients (mean age, 59.9 ± 12.3 years; 55.6% male), 35 developed CR-POPF. Multivariable analysis identified ICA/V and ECV as independent predictors (both p < 0.01). The combined model showed strong performance (apparent AUC = 0.97; optimism-corrected AUC = 0.96) and outperformed either parameter alone (both p < 0.01). Internal validation showed good calibration (optimism-corrected Brier score, 0.04; slope, 0.92; intercept, -0.04). CPA correlated positively with ECV (ρ = 0.63, p < 0.01) and negatively with ICA/V (ρ = -0.59, p < 0.01). Exploratory mediation analysis suggested that these associations were consistent with a fibrosis-related pathway.

CONCLUSION

DECT-derived ICA/V and ECV are promising biomarkers for predicting CR-POPF. Their correlations with histological fibrosis and exploratory mediation findings support a biologically plausible association with CR-POPF risk and suggest potential value for noninvasive preoperative assessment of pancreatic texture and fibrosis before surgery and perioperative risk stratification.