BACKGROUND: Preoperative differentiation between malignant and benign pancreatic tumors can be difficult. Consequently, a proportion of patients undergoing pancreatoduodenectomy for suspected malignancy will ultimately have benign disease. The aim of this study was to compare preoperative clinical and diagnostic characteristics of patients with unexpected benign disease after pancreatoduodenectomy with those of patients with confirmed (pre)malignant disease. METHODS: We performed a multicenter retrospective cohort study in 1,629 consecutive patients undergoing pancreatoduodenectomy for suspected malignancy between 2003 and 2010 in 11 Dutch centers. Preoperative characteristics were compared in a benign:malignant ratio of 1:3. Malignant cases were selected from the entire cohort by using a random number list. A multivariable logistic regression prediction model was constructed to predict benign disease. RESULTS: Of 107 patients (6.6 %) with unexpected benign disease after pancreatoduodenectomy, 86 fulfilled the inclusion criteria and were compared with 258 patients with (pre)malignant disease. Patients with benign disease presented more often with pain (56 vs. 38 %; P = 0.004), but less frequently with jaundice (60 vs. 80 %; P < 0.01), a pancreatic mass (13 vs. 54 %, P < 0.001), or a double duct sign on computed tomography (21 vs. 47 %; P < 0.001). In a prediction model using these parameters, only 19 % of patients with benign disease were correctly predicted, and 1.4 % of patients with malignant disease were missed. CONCLUSIONS: Nearly 7 % of patients undergoing pancreatoduodenectomy for suspected malignancy were ultimately diagnosed with benign disease. Although some preoperative clinical and imaging characteristics might indicate absence of malignancy, their discriminatory value is insufficient for clinical use.