ORIGINAL ARTICLE. A Preoperative Biliary Stent Is Associated With Increased Complications After Pancreatoduodenectomy

Transcription

1 ORIGINAL ARTICLE A Preoperative Biliary Stent Is Associated With Increased Complications After Pancreatoduodenectomy Martin J. Heslin, MD; Ari D. Brooks, MD; Steven N. Hochwald, MD; Lawrence E. Harrison, MD; Leslie H. Blumgart, MD; Murray F. Brennan, MD Background: A preoperative biliary stent is commonly used after the initial evaluation of the patient with a periampullary mass. Objective: To evaluate the effect of a preoperative biliary stent on operative difficulty, postoperative complications, and length of hospital stay after a pancreatoduodenectomy. Design: A retrospective review of a prospectively collected consecutive series. Setting: The Memorial Sloan-Kettering Cancer Center s Surgical Service, New York, NY. and Methods: Seventy-four patients underwent pancreatoduodenectomy between March 1, 1994, and February 15, Thirty-five did not receive a biliary stent, and 39 received a biliary stent prior to medical evaluation. We analyzed patient, nutritional, laboratory, and operating room factors. Univariate analysis was by Student t test, 2 test, and Fisher exact test; multivariate analysis was by logistic regression. Significance was defined at P.05. Main Outcome Measures: Operative time, amount of blood loss, complications, and length of hospital stay. Wound complications were defined as cellulitis, superficial infections, and deep infections. Intra-abdominal complications were defined as intra-abdominal abscesses and pancreatic or biliary fistula. Results: Groups were equivalent for tumor size, risk of comorbidity, time spent in the operating room, and amount of blood loss. There was 1 perioperative death. with a stent had significantly lower bilirubin (P.03) and aspartate aminotransferase (P.04) levels and a significantly increased risk of nodal positivity (P.05). The patients with a biliary stent had an increased risk of wound or abdominal complications on univariate (P.003) and multivariate (P.02) analysis and tended toward a prolonged hospital stay (P.04, Wilcoxon signed rank test). Conclusions: A preoperative biliary stent was associated with an increased risk of wound or intraabdominal complications; a stent may prolong the length of hospital stay. However, length of time under anesthesia, amount of blood loss, and transfusion requirements were not altered. A biliary stent should be used with a high degree of selectivity in the management of patients with resectable periampullary masses. Arch Surg. 1998;133: From the Department of Surgery, Memorial Sloan-Kettering Cancer Center, New York, NY. Dr Heslin is now with the Section of Surgical Oncology, University of Alabama at Birmingham. IN RETROSPECTIVE studies, the depth of jaundice has been associated with an increased risk of complications following operations. 1-4 Prospective randomized trials to evaluate the effect of a percutaneous transhepatic biliary stent on normalizing elevated serum bilirubin levels have demonstrated no decrease in complications, 1,5 and some have reported increases in morbidity associated with the procedure. 6 Recently, a randomized trial was performed to evaluate a preoperative endoscopic biliary stent; the trial demonstrated no benefit in the group of patients who received the stent. 7 The study included patients with high and low levels of obstruction due to malignant neoplams; the patients represented a heterogeneous group for preoperative comorbidity, location of the lesion, and resectability. This study determined whether a preoperative biliary stent increased the operative difficulty, rate of complications, and length of hospital stay of patients after they had undergone a pancreatoduodenectomy. The study population was a homogeneous group of patients without evidence of antecedent infection; preoperative jaundice was not a criterion for enrollment, although most patients had documented hyperbilirubinemia prior to the operation. 149

2 PATIENTS AND METHODS Between March 1, 1994, and February 15, 1996, 74 patients underwent pancreatoduodenectomy at Memorial Sloan-Kettering Cancer Center, New York, NY, within the confines of a nutritional trial. The criteria for enrollment into this analysis included no evidence of infection (white blood cell count /L, no fever, and no evidence of bacteremia), no immunosuppressive medication, and no history of abdominal or pelvic irradiation; finally, those enrolled had to have undergone a pancreatoduodenectomy for a presumed or histologically proved periampullary malignant neoplasm. All patients had normal coagulation profiles prior to operation. After admission to the hospital, a medical history was obtained from the patients and a physical examination, laboratory analyses, and a nutritional assessment were performed. Informed consent for operation was obtained from all patients. All patients had documentation of their serum bilirubin levels prior to undergoing pancreatoduodenectomy; then, they underwent operation for presumed periampullary malignant neoplasms. Prior to the evaluation, data documenting the serum bilirubin level were available for 60 patients. These data were unavailable for 14 patients, 9 who did not receive a stent (all those who did not receive a stent were also referred to as the no stent group ) and 5 who did receive a stent (all those who did receive a stent were also referred to as the stent group ). Twenty-six patients in the no stent group had data available for their serum bilirubin levels; 23 (89%) of these patients had levels above the normal range ( 18 µmol/l [ 1.0 mg/dl]). Thirty-four patients in the stent group had data available for their serum bilirubin levels; all except 1 (97%) had levels above the normal range ( 18 µmol/l [ 1.0 mg/dl]). A preoperative biliary stent was defined as either an endoscopic or percutaneous transhepatic stent that crossed the ampulla of Vater. Of the 39 patients who received a stent, 34 had an endoscopically placed prosthesis and 5 underwent percutaneous transhepatic placement. A careful analysis of factors associated with perioperative morbidity was undertaken. The patient factors analyzed were age, sex, and comorbidity. Comorbidity was defined as the presence of diabetes mellitus (insulin dependent or non insulin dependent), coronary artery disease, or hypertension. The nutritional factors analyzed included percentage of weight lost, preoperative serum albumin level, and nutrition risk index (NRI). The NRI is defined as follows: (15.9 the albumin level [in grams per liter])+ (0.417 the percentage of the usual body weight). 8 with an NRI of 97.5 or more but less than are classified as borderline nutritional risk and show no stigmas of malnutrition. 9 The specific laboratory data analyzed included alkaline phosphatase level; aspartate aminotransferase level; serum urea nitrogen level; serum creatinine level; white blood cell count; and total lymphocyte count, defined as the white blood cell count multiplied by the percentage of lymphocytes in the differential blood cell count. When used, early enteral feeding was begun on postoperative day 1, advanced as tolerated to a target of 25 ml/kg every 24 hours, and continued until the patient was eating 1000 kcal ( J) or consuming at least 1000 ml of a liquid diet. Eight surgeons performed these operations; 72% of the operationswereperformedby2surgeons(l.h.b. andm.f.b.). Most operations were standard pancreatoduodenectomies involvingdistalgastrectomywithpancreatojejunostomy, choledochojejunostomy, and gastrojejunostomy. There were subtle variations, including Roux-en-Y reconstruction of the gastrojejunostomy and pylorus-preserving procedures RESULTS DEMOGRAPHICS Seventy-four patients underwent pancreatoduodenectomy during a 2-year period within the confines of a prospective, randomized trial to evaluate the benefit of early enteral feeding. The demographic data for the patients are listed in Table 2. There was a significantly increased percentage of women compared with men in the stent group. NUTRITIONAL AND BIOCHEMICAL DATA Table 3 lists the nutritional and biochemical parameters measured. There was a statistically significant difference in the NRI between the groups. There were significantly decreased preoperative bilirubin and aspartate aminotransferase levels in the stent group, as would be expected (P=.03 and P=.04, respectively). In the no stent group, the mean bilirubin level of patients at presentation was no different than the mean preoperative bilirubin level. This was compared with the mean bilirubin level in the stent group; in these patients, the bilirubin level at presentation was significantly (P=.03) increased compared with the preoperative bilirubin level. This suggests that on average the stent served its purpose by lowering the bilirubin level and correcting the liver function test results prior to the operation. There were no other significant (P.05) differences in biochemical abnormalities between the groups. OPERATIVE AND PATHOLOGIC DATA In an effort to evaluate the potential additional operative difficulty with the inflammation that commonly accompanies stent placement, operating room factors were analyzed (Table 4). There were no differences between the groups in the length of time under anesthesia, amount of blood lost in the operating room, or the number of patients who needed an intra-operative blood transfusion. The length of time under anesthesia was prolonged on average and does not represent the true time of pancreatoduodenectomy as most patients underwent preincision laparoscopy. Pathologic factors are given in Table 4. There were no significant differences in the site of the lesions between the groups. There were no significant differences between the groups in the proportion of patients with malignant tumors, in the average tumor size, or in the number of positive margins. The 150

3 procedures with pancreatogastrostomy reconstruction, depending on surgeon preference. Routine bile cultures were not obtained and, therefore, could not be correlated to the infectious complications. The operating room factors analyzed were the length of time under anesthesia (which included the time necessary for preincision laparoscopy), the amount of blood lost in the operating room, and the need for a blood transfusion. The pathologic factors analyzed were the presence of malignant neoplasms, tumor size (measured as the single greatest diameter), the presence of nodal metastasis, and the presence of a positive margin. Complications were evaluated and recorded by a physician not affiliated with the surgical team (A.D.B. and S.N.H.). This involved daily patient rounds by a surgeon not associated with the team to evaluate the condition of the patient and laboratory and radiological data when appropriate. Cellulitis was defined as erythema that did not require opening and packing but that did require the use of antibiotic therapy. Superficial wound complications were defined as erythema necessitating opening and packing without débridement. Deep wound infections were defined as an open and packed wound that required débridement. Wound dehiscence was defined as an open wound that had fascial separation. A biliary or pancreatic fistula was defined as persistent bile or pancreatic fluid drainage that continued at 30 ml/d or more and that did not resolve by postoperative day 7. Intra-abdominal abscesses were defined as an intra-abdominal collection associated with fever that required either percutaneous or operative drainage yielding positive culture results. Gastrointestinal tract anastomotic leaks were defined as leaks from either the gastrojejunostomy or jejunojejunostomy. The analysis of wound and abdominal complications is the most conservative in that some of the patients who received a stent were counted as having had a complication when data from them were summarized in the wound or abdominal category; in fact, these patients had multiple wound or abdominal complications during their hospitalization. Other complications listed but not included in the wound or abdominal complications category in Table 1 were pneumonia (defined as radiological and culture evidence of infection), reexploration for any cause (specifically bleeding, hepatic artery thrombosis, bowel necrosis, and wound débridement), and death (hepatic artery thrombosis). Complications not listed, but possibly included, in the any complication category in addition to those listed were prolonged ileus, gastric atony, atelectasis, pleural effusion, urinary tract infection, abdominal wall infection at the jejunostomy tube site, leak at the jejunostomy tube site, evisceration, intestinal obstruction, pulmonary embolism, myocardial infarction, cerebral vascular incident, deep venous thrombosis, chylous leak, renal failure, congestive heart failure, gastrointestinal tract hemorrhage, bowel necrosis, pneumothorax, necrotizing fasciitis, colitis due to Clostridium difficile, line sepsis, acute cardiac arrhythmias, tracheobronchitis, and abdominal wall hematomas. The length of hospital stay was measured from the date of the operation until discharge from the hospital in days, regardless of social service constraints. Statistical analyses were performed using computer software (SPSS for Windows, version 7.0, SPSS Inc, Chicago, Ill). Continuous variables were compared by the Student t test, categorical variables were compared by the 2 or Fisher exact test, and multivariate analysis used logistic regression. Length of hospital stay was compared using the log-rank test and the Wilcoxon signed rank test and plotted by the Kaplan- Meier method. The time until discharge from the hospital (length of hospital stay) does not represent a normal distribution; therefore, the Wilcoxon signed rank test may be a more appropriate analysis with a Kaplan-Meier representation of the curves. The mean length of hospital stay was compared by the Student t test. Significance was defined at P.05. patients in the stent group had a significantly increased rate of nodal positivity compared with those in the no stent group. COMPLICATIONS Wound and abdominal complications that might be related to the presence of an indwelling stent are listed in Table 1. There was no significant increased rate of complications in any individual end points followed. However, when wound or abdominal complications were combined, there was a highly significant increase in the risk of complications in the patients who had received a stent preoperatively (P=.003) (Figure 1). When significant univariate factors were subjected to a multivariate analysis, the presence of a stent was the only factor that significantly correlated with an increased risk of wound or abdominal complications (P.02). A comparison of patients in the stent group who had any 1 complication with those in the no stent group who had any 1 complication revealed that there was a significantly increased rate of complications in the patients in the stent group. These complications included those listed in Table 1 and those listed in the Subjects and Methods section but not separately analyzed. LENGTH OF HOSPITAL STAY An analysis of the time until discharge from the hospital by the Wilcoxon signed rank test and a plot of these differences by the Kaplan-Meier method revealed a statistically significant difference between the 2 groups (P.04) (Figure 2). There was no statistical difference by the log-rank test. There were no significant differences in the length of hospital stay when analyzed by comparing the mean (±SEM) of each of the groups (Table 1). COMMENT We have shown in a multivariate analysis of patient, nutritional, biochemical, operative, and pathologic factors that the presence of a biliary stent prior to pancreatoduodenectomy is significantly (P.05) associated with an increased risk of wound or abdominal complications. Selected patients who are technically candidates for pancreatoduodenectomy, without fever, signs of bacteremia, or notable coagulation abnormalities, should not require the placement of a routine biliary stent, as the presence of jaundice and abnormalities of liver function test results alone are not associated with increased complications or prolonged length of hospital stay. 151

4 Table 1. Data Regarding Complications, Deaths, and Length of Hospital Stay Characteristics Who Did Not Receive a Stent Who Did Receive a Stent (n=39) P * Wound or abdominal complications Cellulitis Superficial infections Deep infections Wound dehiscence Fistulae (pancreatic or biliary) Intra-abdominal abscesses Gastrointestinal tract anastomotic leak Other complications Pneumonia Reexploration Death Summary No. of wound or abdominal complications No. of any complications Length of hospital stay, d Median (Wilcoxon signed rank test) Mean (±SEM) (Student t test) 12±1 13±1.41 Total *Ellipses indicate data not applicable. can be listed in more than 1 category if more than 1 complication has occurred. are only listed once regardless of the number of complications that an individual patient sustains for wound or abdominal complications and any complications. Any complications is defined as a complication listed in this table as well as those not listed but included in the and Methods section. The role of preoperative biliary drainage prior to operation has been examined previously. The difference in most of these trials compared with the present study is mainly that the previous trials enrolled patients with infectious and coagulation abnormalities in addition to obstructive jaundice with lesions at varying locations in the biliary tree. McPherson et al 6 randomized 65 patients to preoperative percutaneous transhepatic biliary drainage (PTD) followed by laparotomy vs laparotomy alone. This study included high and low levels of malignant biliary obstructions and various curative and palliative surgical procedures. Drainage reduced the serum bilirubin level as expected, but there was no difference in the overall postoperative morbidity or mortality. This fact, in addition to the patients who had complications related to PTD prior to the operation, suggested that PTD did not improve survival or decrease morbidity. Hatfield et al 5 randomized 57 patients to external PTD vs surgery alone. This study included patients regardless of clinical status, most of whom had carcinoma of the head of the pancreas; however, most patients underwent palliative bypass surgery and not resection. This study demonstrated that PTD decreased the serum bilirubin level; however, postoperative complications were Table 2. Patient Demographic Characteristics and Comorbidity* Characteristics and Comorbidity Who Did Not Who Did (n=39) Age, y 62±2 67±2.08 Female sex 11 (31) 22 (56).03 Diabetes mellitus 5 (14) 11 (28).15 Hypertension 10 (29) 15 (38).37 Coronary disease 4 (11) 6 (15).62 *Data are given as the number (percentage) of patients in each group unless otherwise specified. Data are given as the mean (±SEM). similar and complications of the drainage procedure were significant. Pitt et al 1 randomized 79 patients, of whom 75 eventually underwent operation. The location of the lesions produced low and high levels of biliary obstructions, and approximately 75% of the lesions were found to be malignant. There was no difference between the groups for postoperative complications, although those patients randomized to receive a stent had an increased cost of hospitalization. In this series, at least 33 patients had a periampullary carcinoma and only 7 underwent pancreatoduodenectomy, which would be insufficient to subset analyze further. Lygidakis et al 10 consecutively studied 38 patients and randomly assigned them to stent vs no stent using endoprostheses only. Approximately 50% of the whole cohort had fever and positive bile culture results, 25% with positive blood culture results. This study found that preoperative drainage decreased the incidence of postoperative complications. The presence of clear signs of infection without drainage prior to the operation is the likely reason for the increased rate of complications in the undrained group. If positive blood or bile culture results or both, which in and of themselves we would not consider a complication, were taken out of the complication classification, it is unlikely that there would be a significant difference between the groups. Either way, this study included patients with cholangitis who we would not offer operation without consideration of preoperative drainage and antibiotic therapy. Lai et al 7 randomized 87 patients to undergo either early elective surgery or endoscopic biliary drainage followed by exploration. This study included high and low levels of malignant biliary obstructions. The group of patients who underwent endoscopic drainage had significant reductions of hyperbilirubinemia and serum albumin concentrations. There was no difference in morbidity or mortality between groups or for the subset of patients with obstructions distal to the common hepatic duct. Twenty-three pancreatoduodenectomies were performed in the study, which again would be too small a subset to further analyze. The present study compares 2 homogeneous groups of patients. The major difference is the presence of a biliary stent with associated normalization of serum P 152

5 Table 3. Nutritional and Biochemical Parameters* Parameters Who Did Not Who Did (n=39) Nutritional Weight loss, % 6±1 8±1.28 NRI 104±1 100±1.03 Early enteral feeding 18 (51) 26 (66).18 Biochemical Albumin, g/l 41±1 39±1.06 Alkaline phosphatase, U/L 313±40 249±20.16 AST, U/L 87±14 53±8.04 Ever jaundiced 27 (77) 38 (97).008 Presentation bilirubin, µmol/l (mg/dl) 118±18 (6.9±1.0) 160±14 (9.3±0.8).09 Days jaundiced 21±3 18±2.36 Preoperative bilirubin, µmol/l (mg/dl) 114±20 (6.7±1.2) 58±12 (3.4±0.7).03 BUN, mmol/l (mg/dl) 5.5±0.5 (15±1) 5.5±0.5 (15±1).99 Creatinine, µmol/l (mg/dl) 85.7±3.5 (0.97±0.04) 85.7±3.5 (0.97±0.04).99 WBC, 10 9 /L 7.3± ± TLC (WBC the percentage of lymphocytes) 1373± ± *Data are given as the mean (±SEM) unless otherwise specified. NRI indicates nutrition risk index; AST, aspartate aminotransferase; BUN, serum urea nitrogen; WBC, white blood cell count; and TLC, total lymphocyte count. Data are given as the number (percentage) of patients in each group. Ever jaundiced indicates the presence of clinical jaundice prior to the operation (n=27, patients who did not receive a stent; n=37, patients who did receive a stent). Presentation bilirubin indicates the serum bilirubin level at presentation to the referring physician (n=27, patients who did not receive a stent; n=34, patients who did receive a stent). Days jaundiced indicates the number of days from presentation to the operation during which the patients had jaundice (n=27, patients who did not receive a stent; n=36, patients who did receive a stent). Preoperative bilirubin indicates the serum bilirubin level at preoperative testing. P Table 4. Operative and Pathologic Characteristics* Characteristics Who Did Not Receive a Stent Who Did Receive a Stent (n=39) Operating room (OR) features Anesthesia time, min 325±20 373±23.11 OR blood loss, ml 1080± ± OR blood transfusion 12 (34) 13 (33).93 Location Pancreas 27 (77) 27 (69).99 Ampulla 3 (9) 4 (10).93 Bile duct 2 (5) 8 (21).06 Duodenum 3 (9) 0 (0).34 Pathologic features Malignant neoplasm 32 (91) 37 (95).67 Tumor size, cm 3.1± ± Node positive 20 (57) 31 (79).05 Margin positive 9 (26) 11 (28).88 *Data are given as the number (percentage) of patients unless otherwise specified. Data are given as the mean (±SEM). Data were incomplete for these variables. The number of patients with data for these variables were as follows: n=32, patients who did not receive a stent; and n=37, patients who did receive a stent. There were nodal metastases. There was a tumor at the margin of the resection on the final pathologic examination. bilirubin level and liver function test results in the stent group. There are more women in the stent group, the clinical significance of which is unknown and likely irrelevant. The NRI in the stent group is statistically P No. of With Wound or Abdominal Complications Who Did Not Who Did Figure 1. Data for the number of patients with wound or abdominal complications by group are provided (patients who did not receive a stent, n=35; patients who did receive a stent, n=39). The difference between the 2 groups was significant ( P=.003). lower than that in the no stent group. The NRI was originally used to stratify patients based on nutritional status in the Veterans Affairs Total Parenteral Nutrition Cooperative Study Group trial evaluating the effects of perioperative total parenteral nutrition. 9 An analysis of that study revealed that only patients who had an NRI less than 83.5 had an increased risk of perioperative complications. Both groups of patients in the present study have mean NRIs of 100 or more; despite statistical differences, it is likely that there are no clinically significant differences. Either way, the NRI did not provide additional prognostic information when studied by multivariate analysis. There were significantly (P.05) more patients in the stent group who had ever had jaundice. Five pa- 153

6 Proportion of Discharged From the Hospital Who Did Not Who Did Length of Hospital Stay, d Figure 2. A Kaplan-Meier graph of the time until discharge from the hospital (patients who did not receive a stent, n=35; patients who did receive a stent, n=39). The difference between the 2 groups was significant ( P.04, Wilcoxon signed rank test; and P=.24, log-rank test). tients in the no stent group had abnormal bilirubin levels (range, µmol/l [ mg/dl]) in the months preceding the operation, which resolved spontaneously. Recognizing that the data at presentation are incomplete, serum bilirubin level, either at presentation or preoperatively, did not correlate with outcome in contradistinction to previous reports. 1,11 Therefore, we believed that all pancreatoduodenectomies should be included for analysis, regardless of whether an abnormal serum bilirubin level was documented. The patients with a preoperative stent had a significantly (P.05) higher rate of microscopic nodal positivity. Biologically, it would be conjecture to suggest that the stent was associated with the presence of nodal metastasis. The identification of a greater number of positive nodes is likely to be a statistical aberration due to relatively small numbers. A relatively small but consistent increase in the length of hospital stay in the stent group would be hard to prove when only analyzing the difference in the means, not because they are not different but because the range over which most patients are discharged from the hospital is relatively small and is not a normal distribution. The Wilcoxon signed rank test examines early differences in timeuntil-event data and weighs the differences in the curves by the number of patients at risk at any one time. Examining these data in this way uses time until discharge from the hospital as the event variable that demonstrated a significant difference (Figure 2). A separate way to emphasize the difference between the groups would be to summarize the total length of hospital stay between the 2 groups, which is done in Table 1. Recognizing that there are 4 more patients in the stent group, there is an approximately 20% increase in the total days spent in the hospital by the group of patients who received a stent preoperatively. Even if the median hospital stay for 4 patients is added to the total length of the hospital stay for the no stent group, there is a 13% increase (52 days) in the number of days spent in the hospital in the stent group. In summary, the strengths of this study include a well-characterized, almost homogeneous group of patients, all of whom underwent pancreatoduodenectomy in an institution in which this operation is performed frequently, thereby minimizing technical errors due to inexperience. were selected by the absence of infectious comorbidity and coagulopathy. The postoperative complications were objectively and consistently reviewed by a physician outside the surgical team, the absence of which is often the problem with retrospective reviews of complication data. There was a significantly (P.05) increased rate of wound and abdominal complications in the group of patients who received a stent preoperatively. The length of hospital stay was mildly prolonged in the stent group, and any prolongation of hospitalization in today s economic climate would be viewed as suboptimal. The preoperative use of a biliary stent in the patient with a technically resectable periampullary malignant neoplasm should be applied with great selectivity given the association with increased complications and prolonged length of hospital stay. Dr Heslin was a Kristen Ann Carr Fellow from 1995 to 1996 at the Memorial Sloan-Kettering Cancer Center, New York. We thank Lianne Latkany, RD, for providing outstanding nutritional assessment and data management skills; and the faculty, fellows, and staff at Memorial Sloan-Kettering Cancer Center who helped make this study possible. Reprints: Murray F. Brennan, MD, Department of Surgery, Memorial Sloan-Kettering Cancer Center, 1275 York Ave, New York, NY REFERENCES 1. Pitt HA, Gomes AS, Lois JF, Mann LL, Deutsch LS, Longmire WP Jr. Does preoperative percutaneous biliary drainage reduce operative risk or increase hospital cost? Ann Surg. 1985;201: Armstrong CP, Dixon JM, Taylor TV, Davies GC. Surgical experience of deeply jaundiced patients with bile duct obstruction. Br J Surg. 1984;71: Dixon JM, Armstrong CP, Duffy SW, Davies GC. Factors affecting morbidity and mortality after surgery for obstructive jaundice: a review of 373 patients. Gut. 1983;24: Su CH, P eng FK, Lui WY. Factors affecting morbidity and mortality in biliary tract surgery. World J Surg. 1992;16: Hatfield AR, Tobias R, Terblanche J, et al. Preoperative external biliary drainage in obstructive jaundice: a prospective controlled clinical trial. Lancet. 1982;2: McPherson GA, Benjamin IS, Hodgson HJ, Bowley NB, Allison DJ, Blumgart LH. Pre-operative percutaneous transhepatic biliary drainage: the results of a controlled trial. Br J Surg. 1984;71: Lai EC, Mok FP, Fan ST, et al. Preoperative endoscopic drainage for malignant obstructive jaundice. Br J Surg. 1994;81: Buzby GP, Williford WO, Peterson OL, et al. A randomized clinical trial of total parenteral nutrition in malnourished surgical patients: the rationale and impact of previous clinical trials and pilot study on protocol design. Am J Clin Nutr. 1988; 47(suppl): The Veterans Affairs Total Parenteral Nutrition Cooperative Study Group. Perioperative total parenteral nutrition in surgical patients. N Engl J Med. 1991;325: Lygidakis NJ, van der Heyde MN, Lubbers MJ. Evaluation of preoperative biliary drainage in the surgical management of pancreatic head carcinoma. Acta Chir Scand. 1987;153: Trede M, Schwall G. The complications of pancreatectomy. Ann Surg. 1988;207: