Methods
Patients
Eight fellowship-trained pancreatobiliary surgeons performed 672 consecutive PDs between October 1993 and November 2008 in our department; the period of observation was 1993 to 2011. We excluded patients who underwent palliative bypass or pancreatic resections for pancreatic cancer in the body and tail of the pancreas, distal cholangiocarcinoma, duodenal carcinoma, neuroendocrine tumors, cyst-adenocarcinoma, solid and papillary tumors, and metastatic tumors. The final pathological diagnosis confirmed ductal pancreatic adenocarcinoma (PDAC) in 195 (29%) of the remaining patients. The demographic characteristics are summarized in Table 1.
Operations
The head resection surgeries analyzed in the study included 69.7% pylorus-preserving pancreatoduodenectomies (PPPD) and 30.3% standard Kausch-Whipple pancreatoduodenectomies (Whipple). The decision for one of the approaches (either Whipple or PPPD) was made during the operation. The primary goal of every operation was en bloc R0 tumor resection. In all the patients, a lymphadenectomy was performed along the hepatoduodenal ligament, common hepatic artery, vena cava, interaortocaval and right side of the superior mesenteric artery. In cases with portal vein involvement, a venous resection was performed to achieve R0-resection. Patients with arterial infiltration by the tumor were stated to be locally irresectable. Thrombosis of the portal vein was always a contraindication for pancreatic head resection. The two-layer invagination technique was used for pancreatic anastomosis in all the cases as previously described. We routinely placed drains intraoperatively. All the patients were staged preoperatively with CT and/or MRI and transabdominal ultrasound, and the PD patients were routinely observed at the Intensive Care Unit (ICU). The drains were removed after exclusion of a postoperative pancreatic fistula (POPF). Postoperative complications were treated symptomatically.
Data Collection
The medical records from a prospective database of patients who underwent PDs for PDAC were analyzed retrospectively for each case. In accordance with the guidelines for human subject research, approval was obtained from the Ethics Committee at the Carl Gustav Carus University Hospital. All the operated patients singed inform consent agreements before surgery. The survey data were complemented with the clinical notes of the patients' physicians and surgeons. Details regarding the deceased patients were obtained from family members or from the general practitioner. The postoperative follow-up time was three years or until the death of the patient.
Patient characteristics and parameters used for statistical analysis are listed in the supplementary information (Additional file 1: Table S1). The postoperative events and clinical outcomes were recorded prospectively and analyzed retrospectively. The tumor-stage designation was categorized according to the TNM system of the Union Internationale Contre le Cancer (UICC 2007).
Definitions
Perioperative mortality was defined as in-hospital mortality. Postoperative pancreatic hemorrhage (PPH) was categorized according to the ISGPS consensus definition Delayed gastric emptying (DGE) was classified according to the definition suggested by the ISGPS. Postoperative pancreatic fistula (POPF) was defined according to the ISGPF criteria.
Statistical Analysis
The statistical analyses were performed using SPSS for Windows, version 15.0 (SPSS, Inc., Chicago, IL). All clinical and pathological characteristics were stratified to build categorical or nominal variables. CEA and CA19-9 were grouped according to the cutoff values used in our center (cut-off levels CEA and CA 19–9: ≤3 ng/ml and ≤75 U/ml). Other variables such as age and BMI were grouped according to previous publications. The thresholds used for categorization were based on previously described thresholds in the literature and/or recursive partitioning as previously described. Continuous data are presented as 95% confidence intervals (95% CI) and standard deviation (SD). The univariate examination of the relationship between the assessed criteria and survival was performed with a χ-test. To assess the impact of the different parameters on survival, we utilized a 3-year survival rate. The estimates of patient survival were generated using the Kaplan-Meier method. The comparisons of survival were performed using the log-rank test. Student's t-tests (ratio scale) and Fisher's exact tests (ordinal scale) were utilized for comparisons between groups. Ordinal-scaled variables were compared using the chi-square test. Significant factors (at P < 0.10) at the univariate level were entered into the multivariate model. A Cox regression analysis with stepwise backwards elimination based on the likelihood ratios was employed to test for independent predictors of survival. A p-value <0.05 was considered significant.