Preoperative Pancreatic Stiffness by Real-time Tissue Elastography to Predict Pancreatic Fistula After Pancreaticoduodenectomy

Patients and Methods

Patients and surgery. The study population consisted of 17 consecutively enrolled patients who underwent subtotal stomach-preserving pancreaticoduodenectomy (SSPPD) at the Hiroshima Red Cross Hospital and Atomic Bomb Survivors Hospital between July 2014 and February 2016. Preoperative RTE imaging could not be performed in one patient because of difficulty obtaining clear images of the pancreatic body. In the remaining 16 patients, one laboratory technician (K.M.) and one surgeon (N.H.) measured the pancreatic liver fibrosis (LF) index by RTE of the pancreatic body one day prior to SSPPD.

SSPPD was performed as previously described by Tatsumi C et al. (8). The pancreatic parenchyma was surgically divided, and reconstruction was performed using the modified Child's method. The main pancreatic duct was anastomosed to the jejunum with 6-0 PDS intermittent sutures. A 4-8 French (Fr) polyvinyl catheter with multiple side holes (MD-41515 pancreatic duct tube; Sumitomo Bakelite, Tokyo, Japan) was inserted as an external drain or an internal stent in all patients. Next, 3-0 TE Prolene sutures were used to anastomosis the pancreas and the seromuscular layer of the jejunum. The external drain was typically removed on postoperative day 21. Then, end-to-side hepaticojejunostomy was performed by single-layer anastomosis using continuous or interrupted sutures with 5-0 or 6-0 PDS-II 10-15 cm distal to the pancreaticojejunostomy site. Gastrojejunostomy was performed via the antecolic route by two-layer anastomosis with 4-0 PDS-II and 3-0 silk 35-45 cm distal to the hepaticojejunostomy site. Next, polyethylene glycolic acid felt (Neoveil® 50×50 mm; Gunze, Kyoto, Japan) and fibrin glue (Bolheal®, The Chemo-Sero-Therapeutic Research Institute, Kumamoto, Japan) were applied to the pancreaticojejunostomy site. Nineteen Fr closed drains (J-VAC Suction Reservoir; Johnson and Johnson, Somerville, NJ, USA), from which postoperative drain fluid was collected, were inserted at the anterior pancreaticojejunostomy site. We graded the PF using the International Study Group of Pancreatic Fistula (ISGPF) classification (9).

Diagnosis of tactile sense-based pancreatic hardness was performed by two hepatobiliary-pancreatic surgeons (T.M. and H.K.) who were blinded to the preoperative pancreatic RTE stiffness value. The study protocol conformed with the updated ethical guidelines of the 2013 Declaration of Helsinki and was approved by our Institutional Review Board.

Preoperative pancreatic stiffness by real-time elastography. Preoperative RTE was performed using HI VISION Ascendus® (Hitachi-Aloka Medical, Tokyo, Japan, Figure 1A) in the pancreas mode LF index. Patients were fasting and were placed in the supine position. Conventional B-mode ultrasonography was used to visualize the pancreatic body parenchyma at a depth of 7 cm from the transducer (10). The region of interest of the strain image was approximately 2.5 cm² and was located at the pancreatic body above the confluence of the abdominal aorta, distant from the pancreatic tumors (Figure 1B). The LF index and Mean relative strain value (MEAN) were measured in each patient (11). The pancreatic LF index was measured three times in each of the 16 patients; the mean LF index of the three values was calculated and used in the analyses.

Pathological assessment of the pancreas. Pancreatic tissue samples were obtained 0.1-0.5 cm from the pancreatic stump, fixed in formalin, and stained with Masson's trichrome (MT). The slides were evaluated by an experienced pathologist (W.H.) and blinded to each patient's clinical data, LF index, and postoperative course. The degree of pancreatic fibrosis was indicated as a percentage of fibrosis of the slides (Figure 1B), calculated as the ratio of fibrosis (blue-stained area) to acinous cells (red-stained area) in the ×40 field of vision using video diagnosis software (ImageJ 1.50i; Wayne Rasband, National Institutes of Health, Bethesda, MD, USA,). We choose 10 sites at random and examined the correlation between the percentage of fibrosis and the LF index.

Statistical analysis. Continuous data were expressed as mean±standard deviation (SD). Quantitative and categorized variables were compared using the Wilcoxon rank-sum test and Fisher's exact test, respectively. The correlations of the LF index, percentage of fibrosis, and relative strain were evaluated using Spearman's rank correlation test. Multivariate analysis was performed to evaluate the ability of the preoperative LF index to predict the risk of postoperative PF. Because of their clinical significance, seven potentially important pre- and intraoperative factors were included in the analysis: age (≤82 years vs. >82 years), sex, presence of diabetes mellitus (yes vs. no), body mass index (>21.4 kg/m² vs. ≤21.4 kg/m²), amount of blood loss (>245 ml vs. ≤245 ml), main pancreatic duct (MPD) diameter (≤3 mm vs. >3 mm), and LF index (≤1.91 vs. >1.91). The cut-off values were calculated by receiver operating characteristic (ROC) curve analysis regarding the presence of PF. The results were expressed as adjusted odds ratios (OR), and p values were calculated using the likelihood ratio test. Values of p<0.05 were considered statistically significant. All statistical analyses were performed using JMP software version Pro12 (SAS Institute, Cary, NC, USA).