Preoperative endoscopic ultrasound (EUS) strain elastography for the prediction of intraoperative pancreatic texture after pancreaticoduodenectomy: a prospective diagnostic accuracy study with exploratory analysis of clinically relevant postoperative pancreatic fistula

Background Postoperative pancreatic fistula (POPF) is the most serious complication of pancreaticoduodenectomy (PD); established risk tools require intraoperative variables. We assessed the diagnostic accuracy of preoperative endoscopic ultrasound (EUS) strain elastography for intraoperative pancreatic texture and, exploratorily, predictors of clinically relevant POPF (CR-POPF). Methods Prospective single-centre study of 100 consecutive patients undergoing elective PD. The index test was preoperative EUS strain ratio (SR), threshold ≤ 2.5 prespecified before enrolment; the reference standard was intraoperative texture recorded independently by two blinded surgeons, with neck-margin fibrosis grade as secondary reference. CR-POPF (ISGPS 2016 Grade B/C) was adjudicated blind to EUS. Accuracy is reported with Wilson 95% confidence intervals (CIs). An exploratory index used Firth-penalised regression restricted to preoperative variables, with bootstrap validation. Reporting follows STARD, TRIPOD, and STROBE. Results Soft texture occurred in 58 (58%). SR ≤ 2.5 gave 51 true positives, 7 false positives, 7 false negatives, 35 true negatives: sensitivity 87.9% (95% CI 77.1–94.0), specificity 83.3% (69.4–91.7), predictive values 87.9 and 83.3%, accuracy 86.0% (77.9–91.5), and area under the curve (AUC) 0.91 (0.84–0.96). SR rose monotonically with fibrosis grade (2.0 ± 0.6 to 5.8 ± 1.9; Spearman ρ = +0.82, p < 0.001); mean histogram elasticity fell ( ρ = − 0.79, p < 0.001). CR-POPF occurred in 16 (16%). Non-PDAC pathology was unassociated with CR-POPF (9/48 vs 7/52; odds ratio 1.48, 0.51–4.35; p = 0.59). Six preoperative variables were entered simultaneously (EUS-SR, duct diameter, albumin, haemoglobin, body mass index, serum amylase; amylase carried zero points in the rounded score). Apparent AUC was 0.87 (0.79–0.94), optimism-corrected 0.78, calibration slope 0.72, and Brier score 0.096. Conclusions Preoperative EUS strain elastography at a prespecified threshold identified intraoperative pancreatic texture accurately and tracked histological fibrosis, supporting it as an objective surrogate for parenchymal consistency. The index rests on 16 events, is hypothesis-generating, and requires external multicentre validation.