Pancreaticojejunostomy versus controlled pancreaticocutaneous fistula in pancreaticoduodenectomy for perianipullary carcinoma

Pancreaticojejunostomy Versus Controlled Pancreaticocutaneous Fistula in Pancreaticoduodenectomy for Periampullary Carcinoma Petachia Reissman, MD, Yaron Perry, MD, Abigo Cuenca, MD, Allan Bloom, MD, Ahmed Eid, MD, Eitan Shiloni MD, Abraham Rivkind, MD, Arieh Durst, MD, FACS, Jerusalem, Israel

BACKGROUND: Anastomotic leak of the pancreaticojejunostomy is a major cause of morbidity and mortality following pancreaticoduodenectomy. Reports have described a large variety of techniques for performing this anastomosis and managing the pancreatic stump. In an attempt to obviate the pancreaticojejunostomy, we prospectively studied the technique of ligating the pancreatic duct and using external drains to create a temporary controlIed pancreaticocutaneous listula. PATENTS AND METHODS: Thirty-five consecutive patients who were to undergo pancreaticoduodenectomy for periampullary carcinoma were prospectively randomized to one of two groups: pancreaticojejunostomy (PJ) (n = 18) or controlled pancreaticocutaneous listula (CPF) (n = 17). The groups were well matched for age, sex, coexisting medical illnesses, type of tumor, and preoperative condition. Except for the management of the pancreatic remnant, all patients in both groups underwent an identical procedure. Major morbidity, length of hospitalization, duration of the controlled pancreatic fistula, and mortality were analyzed over a mean follow-up interval of 26 months (range 5 months to 7.5 years). RESULTS: The CPF group experienced lower overall operative morbidity rates than the PJ group (24% versus 56%, P ~0.01). Two patients (11%) in the PJ group and none in the CPF group died (P = NS). Half the morbidity in the PJ group and both mortalities were related to anastomotie leak. The CPF and PJ groups left the hospital after mean stays of 26.4 and 42.2 days respectively (P ~0.01). CONCLUSIONS: Compared to pancreaticojejunal anastomosis, creation of a temporary controlled pancreaticocutaneous listula in patients who undergo pancreaticoduodenectomy for periampullary malignancy has no appreciable risk. It is associated with reduced morbidity and shorter length of hospitalization. From the Department of General Surgery, Hadassah University Hospital, Jerusalem, Israel. Requests for reprints should be addressed to Petachia Reissman, MD, Department of Colorectal Surgery, Cleveland Clinic Florida, 3000 West Cypress Creek Road, Fort Lauderdale, Florida 33309. Manuscript submitted December 22,1993 and accepted in revised form April 28, 1994. THE AMERICAN

he pancreaticojejunalanastomosishas always been a T major concernin pancreaticoduodenectomy.Surgeons have attempted a variety of techniquesfor managingthe pancreatic stump, all with the goal of lowering the incidence of anastomoticleak and related complications.‘-* They include invaginating end-to-endpancreaticojejunostomy (dunking-type or telescoping),end-to-side, duct-tomucosapancreaticojejunostomy,isolation of the pancreaticojejunal anastomosison a separateRoux jejunal limb, pancreaticogastrostomy,total pancreatectomy, and injection of the duct with occlusive substances.Nevertheless, failure or disruption of this anastomosis,leading to leakageandabdominalsepsisor hemorrhage,remainsthe most common causeof morbidity and mortality following pancreaticoduodenectomy.1-6 Rather than try to improve the safety of pancreaticojejunostomy, we prospectively assesseda technique that avoids it altogether: ligation of the pancreatic duct and oversewingthe pancreatic stumpwith placementof drains next to the stump in order to create a controlled pancreaticocutaneousfistula (CPF). PATIENTS

AND

METHODS

Between January 1985 and February 1993, 35 consecutive patients who were scheduled for pancreaticoduodenectomy for periampullary carcinoma were randomly assignedto one of two groupsafter informed consent.The PJ group (n = 18) underwent end-to-end pancreaticojejunostomy (PJ) usinga double-layer 3-O silk-suturedanastomosis with invagination of the pancreatic cut end into thejejunum, or with a dunking-type anastomosis. The CPF group (n = 17) underwent primary closure of the pancreatic duct with prolene 5-Orunning suture, and oversewing of the cut edge of the pancreatic stump with 3-O silk interrupted horizontal sutures.Two lo-mm closed-suction drains (Jackson-Pratt Biometrix Ltd., Jerusalem,Israel) were placed superiorly and inferiorly to the pancreatic stump and were brought out separately through the abdominal wall to create a CPF. One of the drains was removed when output volume fell to ~20 mL per 24 hours, which occurred 5 to 12 days postoperatively (mean 8.7). The seconddrain was left in place until output dropped to ~10 mL per 24 hours, and removed at an outpatient clinic visit. Drains were removed according to output regardless of position. Patient data is summarized in Table I. Both groups were closely matched for age, sex distribution, type of tumor, coexisting medical illnesses, and preoperative laboratory data. JOURNAL

OF SURGERY”

VOLUME

169 JUNE

1995

585

PANCREATICODUODENECTOMY

IN

PERIAMPULLARY

CARCINOitWREISSMAN

TABLE I

ET AL

TABLE II Major Complications and Outcome for the Pancreaticojejunostomy (PJ) and Controlled Pancreaticocutaneous Fistula (CPF) Groups CPF ,;p)g) (n q 17) N (%) Oo Anastomotic leak 5 (28) Abdominal abscess 6 (33) 3 (18)

Preoperative Patient Data for the Pancreaticojejunostomy (PJ) and Controlled Pancreaticocutaneous Fistula (CPF) Groups CPF (n IJJ*] (n = 17) Mean age fy) 60.4 58.7 Range 43-73 36-77 1 l/7 Sex (M/F) 1w Diagnosis Pancreatic carcinoma 8 9 Distal CBD carcinoma 2 1 Ampullary carcinoma 8 7 kaboratqry data’ Brkrubrn fmg/dL) 10.1 + 1.0 9.4 + 0.8 Albumin fg/dL) 3.4 f 0.2 3.6 + 0.4 Prothrombin time (s) 11.9 * 0.9 12.1 + 0.7 Hemoglobin (g/dL) 10.9 ” 0.4 11.7 + 0.5 Medical disorders Diabetes mellitus 2 3 lschemic heart disease 1 1 COPD 0 1

an&x

sepsis

Gastrointestinal bleeding Total morbidity’ Mortality Postop hospital stay fd) Range Duration of CPF tdl Range fd)

3 (17)

2 (12)

10 (56) 2 (11) 42.2 f 1.6 21-86 NA

4 (24) 0 26.4 + 1.1 14-90 73.4 * 8.6 21-150

‘Total number of patients who suffered one or more complications. NA = not applicable.

‘Laboratory data reported as mean * standard error. CBD = common bile duct; COPD = chronic obstructive pulmonary disease.

lower in the CPF group when comparedto the PJ group (18% versus 33%, P <0.05). The incidence of gastrointestinalhemorrhagewasnot significantly different between the two groups(CPF 12%, PJ 17%). All thesepatientsrespondedwell to medicaltherapy. The overall morbidity rate in the CPF group was 24%, markedly lower than the 56% observed in the PJ group (P ~0.01). The CPF group also had a shortermeanpostoperativehospitalization, 26.4 days (range14to 90) versus42.2 days(range21 to 86) (P ~0.01). CPF occurredin 16 (94%) of the 17patientsin this group, and persistedfor 21 days to 5 months (mean 73.4 f 8.6 days). One patient in this group did not develop a CPFaS, and no drainagewas noted through the drains. No patient in this group developedan intra-abdominalabscessor collection after the dram was removed. No patient who was nondiabetic prior to surgery developed postoperativediabetes.There were no clinical manifestationsof pancreaticexocrine insufficiency in patients with CPF. Two patients (11%) in the PJ group died, both as a result of anastomotic leak of the pancreaticojejunostomy, sepsis,and multi-organ failure. There were no fatalities in the CPF group (P = NS). The meanfollow-up period was 26 months (range 5-7.5 years.) Long-term survival during the follow-up period was similar in both groups. The median survival was 14.5 months in the PJ group (range 5 months to 6 years), and 12 monthsin the CPF group (range5 monthsto 7.5 years).

All patients received prophylactic perioperative broadspectrum antibiotics. Biliary drainage by endoscopicretrograde cholangiographicstent placementwas performed in patients who had an expected interruption of >14 days for surgery (10 patients in the PJ group and 8 in the CPF group). The samesurgical team performed all the operations. Except for the managementof the pancreatic remnant, the techniqueof pancreaticoduodenectomywasidentical in all patients. Postoperativetotal parenteralnutrition wasadministeredfor a period of 10 to 61 days (mean 17.6). Major postoperative complications such as anastomotic leak, abdominal abscess,sepsis,gastrointestinalhemorrhage, and fatality were recorded and analyzed, as were length of postoperativehospitalstay and, in the CPF group, duration of controlled pancreatic fistula. PJ anastomotic leak was defined as(1) recovery of 240 mL/d of amylaserich fluid (>lO times normal plasmalevel) from the peripancreatic drains for more than 7 days, or (2) radiologically demonstrableleakage. All patients in the CPF group received oral pancreatic enzyme supplementafter the resumption of normal oral intake to avoid manifestationsof pancreatic exocrine insufficiency. Close monitoring of blood and urine glucose levels was performed in both groups from the time of surgery until discharge,and continued with lessintensity during the follow-up period. The follow-up period for all patients ranged from 5 COMMENTS monthsto 7.5 years (mean26 months). Fisher’s exact test More than 50 years have passedsinceAllan 0. Whipple and the t-test were usedfor statistical analysis. first reported his techniquefor pancreaticoduodenectomy, but the ideal managementof the pancreaticstumphasstill RESULTS not beendetermined.2~5,7*8 Most of the acceptedtechniques Table II and Figure 1 summarizeour results.Five pa- include anastomosisbetweenthe pancreatic stumpandjetients (28%) from the PJ group developedleaks,of whom junum. Despite continuous improvement of these tech2 (11%) required reoperation. Both of thesepatients also niques-as well as in the perioperative supportive mansuffered from an abdominalabscessand sepsis.The rate of agementof the patients-the rate of anastomoticleakage abdominal abscessformation and sepsiswas significantly still rangesbetween 4.2% and 26.4%.5*6,9-11 586

THE AMERICAN

JOURNAL

OF SURGERY”

VOLUME

169

JUNE

1995

PANCFiEATICODUODENECTOMY

IN PERL4MPULLAFtY

q 60

CARCINOMAiFWSSMAN

ET AL

CPF HPJ

Days

P
50

h

Figure. Major complications and hospital stay. NS = no statistical significance; PJ = pancreaticojejunal anastomosis; CPF = controlled pancreaticocutaneous fistula, ^. til = gastrolntestlnal. I

Anastomotic leak is a major causeof fatal postoperative complications. Patients with leaks may develop abdominal abscess, sepsis,or hemorrhage,andhave mortality rates as high as 75%.*,‘*-17 Isolation of the pancreatic stump and oversewing of its cut end with external drainageobviates the pancreaticojejunostomy. Nevertheless,the few studiesemploying this technique have been nonrandomized, and they have reported varying results. Goldsmith et all8 performed pancreaticoduodenectomy with ligation of the pancreaticduct and oversewing of the pancreaticstumpin 45 patients.They comparedthis group with 34 patients who had the sameprocedure with pancreaticojejunal anastomosis.The operative mortality and morbidity were similar in both groups. Later comparison of pancreashistology also failed to show any significant differences between the groups. Hines and Burn~‘~ reported 19 patients who underwent pancreaticduct ligation and external drainageduring pancreaticoduodenectomy.A transientcontrolled pancreaticocutaneousfistula occurredin only 3. No additional mortality or morbidity was noted when this group was compared with a similar group of patientswho had undergonea pancreaticojejunostomy.In a more recent series,Funovics et al* reported 19 patientsin whom external drainagealonewithout ligation of the duct or oversewing tbe pancreatic stump-was performed due to local morphologic findings that could endangera pancreaticojejunalanastomosis. The authorsusedPenrosedrainsto createthe CPF. They compared this group with three other groupswho had undergone pancreaticoduodenectomywith different methodsof PJ: end-to-end,end-to-side,and Roux-en-Y. Although the patientswith external pancreaticdrainagehad higher morbidity, their mortality rate was significantly lower than that associatedwith any of the three methodsof pancreaticojejunostomy. The authorsconcluded that the low mortality THE AMERICAN

rate justified the increasedmorbidity, and that controlled pancreaticocutaneous drainageis preferablein casesof endangeredpancreaticanastomosis. In our study, we prospectively assessed the creation of a CPF versus PJ end-to-end anastomosisin two groups of randomized patients.The overall morbidity and mortality ratesfor the combinedgroupswere 40% and 6%. Among the 5 patientswho suffered an anastomoticleak, however, 2 died. This 40% rate is comparablewith other reports of mortality between 20% and 75% in patients who developed pancreatic anastomoticleak.2,12-‘5 As might be expected, anastomoticleak was the responsiblefactor in the significantly higher morbidity and longer hospitalization experiencedby the PJ group. The CPF was transient and lastedfor a meanperiod of 73.4 days. It was well tolerated by all patients with only minor inconvenience. They regularly emptied the “grenade” container of the drain and kept a record of the amountof drainagewithout difficulties. Since the majority have been using the same container from the time of surgery, the prolongedperiod of the CPF was not associated with additional cost. Skin irritation at the drain site was minimal and easily controlled with local wound care. No patient developed drain-relatedabdominal- or fistulatract infection. The null incidence of such infection may have beendue to our useof closedsuctiondrainsthat evacuate the pancreaticfluids into a closedodor-free container, avoiding all direct contact of the highly irritative fluids with the tissuesalong the tract or the skin. Such infections were experiencedin previousseriesthat usedPenrosedrains.We alsooversewedthe pancreaticremnantin addition to ligating the duct, which may have reducedthe initial amountof drainage.As in the few previous reportsemploying a similar technique, no caseof pancreatitiswas noted. The lossof pancreaticexocrine function in the CPF group waseasilymanagedby oral pancreaticenzymes,which were JOURNAL

OF SURGERY”

VOLUME

169

JUNE

1995

587

PANCREATICODUODENECTOMY

IN PERIAMPULLARY

CARCINOMAlREISSMAN

started as soon as oral feeding was resumed. No patient in this group developed clinical manifestations of pancreatic exocrine insufficiency. Interestingly, 3 patients of the PJ group who developed persistent loose bowel movements after resumption of oral feeding responded to oral administration of pancreatic enzymes similar to the regimen used in the CPF group. These findings confirm the suggestion, made by several authors, that pancreatic exocrine dysfunction can occur after pancreaticojejunostomy.19~20 Our patients’ endocrine function was not affected, since none of the nondiabetic patients in either group developed diabetes mellitus following surgery. Despite the cost of local wound care at the drain site and supplementary pancreatic enzymes in the CPF group, this procedure was still economical. The source of savings was the significant reduction in morbidity and length of hospitalization compared with the PJ group. Several authors have reported that the morphological appearance of pancreatic tissue may influence the leakage rate after pancreaticojejunostomy, and that fibrotic firm pancreatic stumps leak less. V’ This factor played no role in our series, since normal pancreatic tissue was noted in 14 of 18 patients in the PJ group and 15 of 17 patients in the CPF group. The method of managing the pancreatic remnant did not affect long-term survival, with median survival of 14.5 months and 12 months in the PJ and CPF groups. Comparable long-term survival might have been expected, since the patients in both groups had resectable tumors and underwent similar procedures in terms of margins and extent of resection. In conclusion, our results indicate that the creation of a transient controlled pancreaticocutaneous fistula in patients who undergo pancreaticoduodenectomy does not increase operative risk. This procedure is associated with reduced morbidity and mortality and shorter hospitalization. It may be a viable alternative to traditional pancreaticojejunal anastomosis. REFERENCES 1. Monge JJ, Judd ES, Gage RP. Radical pancreaticoduodenectomy. A 22 year experience with the complications, mortality and survival rates. Ann Surg. 1964;160:71 l-722.

588

ET AL

2. Funovics JM, Zoch G, Wenzl E, Schulz F. Progress in reconstruction after resection of the head of the pancreas. Surg Gynecol Obstet. 1987;164:545-548. 3. Trede M, Schwall G. The complications of pancreatectomy. Ann Surg. 1988;207:3947. 4. Warshaw AL, Swanson RS. Pancreatic cancer in 1988. Possibilities and probabilities. Ann Surg. 1988;208:541-545. 5. Braasch JW. Pancreaticoduodenal resection. Curr Probl Surg. 1988;25:321-363. 6. Matsumoto Y, Fujii H, Miura K, et al. Successful pancreaticojejunal anastomosis for pancreaticoduodenectomy. Surg Gynecol Obstet. 1992;175:555-562. 7. Peters JH, Carey LC. Historical review of pancreaticoduodenectomy. Am J Surg. 1991;161:219-225. 8. Whipple AO, Parsons WW, Mullins CR. Treatment of carcinoma of the ampulla of vater. Ann Surg. 1935:102:763-768. 9. Crist DW, Sitzman JV, Cameron JL. Improved hospital morbidity, mortality and survival after the Whipple procedure. Ann Surg. 1987;206:358-365. 10. Hines LH, Bums RP. Ten years experience treating pancreatic and peri-ampullary cancer. Am Surg. 1976;42:4414l7. 11. Miedema BW, Sarr MG, Van Heerden JA, et al. Complications following pancreaticoduodenectomy: current management. Arch Surg. 1992;127:945-950. 12. Gilsdorf RB, Spanos P. Factors influencing morbidity and mortality in pancreatoduodenectomy. Ann Surg. 1973;177:332-337. 13. Braasch JW, Gray BN. Considerations that lower pancreaticoduodenectomy mortality. Am J Surg. 1977;133:480-484. 14. Aston SJ, Longrnire WP Jr. Pancreaticoduodenal resection. Arch Surg. 1973;106:813-817. 15. Warren KW, Choe DS, Plaza J, Relihan M. Results of radical resection for periampullary cancer. Am Surg. 1975;181:534-550. 16. Lerut JP, Gianello PR, Otte JB, et al. Pancreaticoduodenal resection. Surgical experience and evaluation of risk factors in 103 patients. Ann Surg. 1984;199:432437. 17. Hertner FP, Cooperman AM, Ahlbom TN, et al. Surgical experience with pancreatic and periampullary cancer. Ann Surg. 1982;195:274-281. 18. Goldsmith HS, Ghosh BC, Huvos AG. Ligation versus implantation of the pancreatic duct after pancreaticoduodenectomy. Surg Gynecol Obstet. 1971;132:87-92. 19. Lai ECS, King W IB, Longmire WP Jr, Tompkins RK. Pancreaticojejunostomy after Whipple’s operation-in vivo evaluation of long term patency: a preliminary report. Surgery. 1987;101:544-548. 20. Fish JC, Smith LB, Williams RB. Digestive function after radical pancreaticoduodenectomy. Am J Surg. 1969;117:40-45.

THE AMERICAN JOURNAL OF SURGERYa VOLUME 169 JUNE 1995