c¿-

An Overview

Rainer W. G. Gruessner, MD, PhD, David E. R. Sutherland, MD, PhD, FACS,

Christoph Troppmann, MD, Enrico Benedetti, MD, Nadey Hakim, MD,

David L. Dunn, MD, PhD, FACS, and Angelika C. Gruessner, MS, PhD

enteric-drained (39%) versus bladder-drained (18%)transplants. Graft and patient survival rates were signif-icantIy 10wer for recipients with (versus without) intra-abdominal infection. Outcome was better after bacte-rial (versus fungal) infections. For SPK recipients,those not on dialysis before fue transplant had signifi-cantIy higher graft survival than those on dialysis. Vas-cular graft thrombosis occurred in 12% of all recipi-ents. The rate was significantIy higher for PAK (21%)than for PTA (10%) and SPK (9%) recipients. It wassignificantIy 10wer for recipients of grafts with donoriliac Y-graft reconstruction (versus all other types ofarterial reconstruction) and with right-sided (versusleft-sided) graft placement.' Of note, patient survivalwas not different for recipients with versus withoutvascular graft thrombosis. The incidence of anasto-motic or duodenal stump leaks was 10%; of fuese re-cipients, 70% required relaparotomy. Patient and graftsurvival rates were no different for recipients with ver-sus without leaks.

Background: Pancreas transplants are still associatedwith the highest surgical complication rate of all rOD-tinely performed salid organ transplants. To date, theimpact of serious surgical complicatioDS in the cyclo-sporine era on perioperative patient morbidity, graftand patient survival, and hospital costs has not beenanalyzed in detail.

Study Design: We retrospectively studied surgical com-plications after 445 consecutive pancreas transplants(45% simultaneous pancreas-kidney [SPK], 24% pan-creas after kidney [PAK], and 31 % pancreas transplantalone [PTA]). Of these, 80% were primary transplants,20% were retransplants. Cadaver donors were used in92%, living related donors in 8%. To develop guide-lines for their prevention and management, we studiedthe impact of significant surgical complications (intra-abdominal infections, vascular graft thrombosis, andanastomotic leak) requiring relaparotomy on graft andpatient survival.

Results: Relaparotomy was required after 32% of allpancreas transplants (SPK: 36%, PAK: 25~, PTA: 16%[p = 0.04]). Perioperative mortality was 9%. Graft andpatient survival rates were significantIy lower for recip-ients with (versus without) relaparotomy. The mostcornmon procedures were drainage of intra-abdominalabscess with graft necrosectomy (50% of all relaparoto-mies) and transplant pancreatectomy (34%). The mostcornmon causes of relaparotomy were intra-abdominalinfection, vascular graft thrombosis, and anastomoticleak. lntra-abdominal infection occurred in 20% (SPK:18%, PAK: 24%, PTA: 20% [p = NS]). The rate wassignificantIy higher for living related donor (42%) ver-sus cadaver donor (18%) recipients and for those with

Conclusions: Serious surgical complications occurred in35% of pancreas recipients and had a significant impacton patient and graft survival. Based on multivariate riskfactor analyses, we recornmend the following: donorsover 45 years and those dying of cerebrocardiovasculardisease should not be usedj recipients over 45 years andthose with a history of cardiac disease should be con-sidered for a kidney transplant alone (KTA)j surgicaltechnique for graft procurement, preparation, and im-plantation should be meticulous; right-sided implanta-tion and arterial Y-graft reconstruction should be per-formed when possible, since they had the highestsuccess ratesj when complications require relapa-rotomy, the focus must switch from graft salvage to lifepreservationj and the threshold for pancreatectomy

.,should be low. Diagnosis should be timely, and treat-ment and relaparotomy expeditious. These comer-stones of success should help de crease the risk ofsurgical complications and mortality after pancreastransplants. (j Am Coll Surg 1997jI85:128-144.@ 1997 by the American College of Surgeons)

Received October 28, 1996; Revised March 3, 1997; Accepted March10,1997.From the Department ofSurgery, University ofMinnesota, Minneap-olis, MN.Presented in part at the Papers Sessions of the 82nd Annual ClinicalCongress of the American College of Surgeons, San Francisco, CA,October 6-11,1996.Correspondence lO: Dr. Rainer W. G. Gruessner, University of Min-nesota, Department of Surgery, Box 90 UMHC, 420 Delaware Sto SE,Minneapofis, MN 55455.

ISSN 1072-7515/97/$17.00Pll SI072-7515(97)00015-X

@ 1997 by fue American College of SurgeonsPublished by Else\ier Science Inc. 128

< J-~ "( 4. ¿ 1 ~

129Gruessner et al RISK OF PANCREAS TRANSPlANTATION

ond, to study the spectrum of the complicationsthat pan creas transplant surgeons have to manage.We also analyzed in detail the three most commoncauses for surgical reexploration: intra-abdominalinfection, vascular graft thrombosis, and anasto-motic leak. Because of the multitude of donar andrecipient risk factors that potentially affect pan-creas transplant outcome, we applied multivariateanalysis to all three recipient categories: simulta-neous pancreas-kidney transplant (SPK), pan-creas after kidney transplant (PAK), and pancreastransplant alone (PTA). Special consideration wasgiven to fue technical aspects of the transplantprocedure itself, their implications for infectious andvascular complications, and fue treatment of post-

transplant surgical complications.

Methods

Study population. We reviewed the most serioussurgical complications (ie, all those requiring re-operation) after 445 consecutive pancreas trans-

plants done at our institution between ]anuary 1,1986 and]uly 31,1994. Cadaver donors were usedin 409 (92%) transplants, and living related do-pors in 36 (8%) transplants. There were 357(80%) primary transplants and 88 (20%) retrans-plants. Of the latter, 68 were first, 17 second, and

3 third retransplants.Of these 445 pancreas transplants, 199 (45%)

were SPK, 138 (31%) were PTA, and 108 (24%)were PAK Ofthe 380 recipients, 375 were white, 3were Native American, and 2 were black. The av-erage recipient age was 33.:!: 6 years (range, 19-59years). For management of pancreas exocrine se-cretions, bladder drainage was used in 399 (90%)transplants, enteric drainage in 43 (10%), andduct injection in 3 (1 %). For cadaver donar trans-plants, bladder drainage was most common (389of 409; 94%); for living related donar transplantsenteric drainage was most common (21 of 36;58%). Donor and recipient selection criteria have

previously been detailed (9).

Surgical techniqueDonor operation (10). Combined pancreas-

liver- kidney specimens were procured as follows.At laparotoJDY, the presence of an aberrant oraccessory right or left hepatic artery is ruled out by

inspection and palpation of the hepatoduodenaland gastrohepatic ligaments. Mter a Kocher ma-neuver, the hepatoduodenalligament is dissectedalong the superior border of the pancreas to iden-

Pancreas transplantation is currently the onlytreatmentof type 1 diabetes mellitus that routinelyand consistently restores continuous normoglyce-mia and normalizes longterm hemoglobin Al Clevels. Pancreas transplantation has gained signif-icant popularity ayer the past decade. In 1995, forthe first time, ayer 1,000 pancreas transplantswerereported-from the United States alone-to theInternational Pancreas Transplant Registry,which has data on a total of 8,500 pancreastransplants worldwide between June 1976 andJuly 1996 (1). Overall 1-year patient survivalrates now exceed 90%; pancreas graft survivalrates are 70% (2). The increasingly widespreadapplication of pancreas transplantation is a re-sult of improved outcomes secondary to techni-cal refinements of the transplant procedure it-self (eg, bladder drainage technique); morepotent immunosuppressive regimens (eg, qua-druple therapy for induction, new immunosup-pressants such as tacrolimus and mycophenolatemofetil); more accurate diagnosis of rejection(eg, percutaneous, transcystoscopic, and laparo-scopic biopsies); more efficient yet less toxicantiviral (eg, ganciclovir), antifungal (eg, flu-conazole), and antibacterial (eg, triazoles)agents for prophylaxis and treatment of post-transplant infection, and better selection of do-nors and recipients (eg, limiting the number ofdonors and recipients > 45 years of age) (3-9).

Pancreas transplantation was originally con-ceived as a definitive surgical therapy for a meta-bolic disease, but it remains a challenge to sur-geons because of the diversity and complexity ofsurgical complications. According to a recent In-ternational Pan creas Transplant Registry report,between 11% and 21 % of all pancreas grafts arelost because of surgical complications such asintra-abdominal infection, vascular graft thrombo-gis, and anastomotic leak (1, 2). Minimizing surgi-cal complications would not only de crease mor-bidity and mortality rates, but algo provideadditional impetus for even more widespread ap-plication, especially because pancreas transplanta-tion is not considered life saving.

.In our series of 445 consecutive pancreas trans-plants betweenJanuary 1,1986 andJuly 31,1994(cyclosporine era), we reviewed the most serioussurgical complications, (ie, all those requiring re-operation). The purpose of this retrospective re-view was twofold: first, to assess the perioperativemorbidity and implications for longterm graft andpatient survival of surgical co_mplications, and sec-

130 J MI COLL SURG AUGUST 1997 VOLUME 185:128-144

tify the common hile duct, common hepatic ar-tery, and portal vein. The common bile duct andthe gastroduodenal artery are identified, ligated,and divided. The common hepatic artery is mobi-lized in a retrograde fashion toward the celiac axis;the splenic artery takeoff is identified and dis-sected circumferentially. The takeoff of the celiacaxis from the aorta is identified, and the suprace-liac aorta is looped.

The splenocolic ligament and the retroperito-neal attachments of the spleen are divided. Withthe spleen as a handle to minimize mechanicaltrauma, the pancreas is then dissected off its retro-peritoneal attachments. During mobilization ofthe lower pancreatic margin, the inferior mesen-teric vein is ligated and divided. The pancreaticdissection and mobilization are continued to theleft lateral aspect of the abdominal aorta.

A nasogastric tube is advanced across the pylo-rus into the duodenum, which is flushed with 250mL of normal saline solution containing cefazolinsodium (4,000 mg/L), amikacin (2,000 mg/L),and amphotericin B (200 mg/L). The nasogastrictube is pulled back into the stomach. The duo de-num is stapled and transected proximally and dis-tally at the ligament of Treitz using a GIA (gastro-intestinal anastomosis) stapler.

The donor is systemically heparinized, and thesupraceliac abdominal aorta is cross-clamped. Theliver, pancreas, and kidneys are flushed in situ withUniversity of Wisconsin preservation solution viathe lower abdominal aorta. The intra-abdominalorgans are cooled with topical ice slush. Systemicvenous venting is done through the right atriumor the infrarenal vena cava. At the same time, aportal flushing cannula is inserted into the portalvein, through either the inferior or the superiormesenteric vein.

Next, the mesenteries of the small bowel andcolon are divided, using a GIA stapler. Sufficientdistance is left between the pan creas and the sta-pIe lines to permit gafe re-ligation of the mesen-teric root ex vivo on the back table. The liver,pancreas, and kidneys are then excised sequen-tially: usually the celiac axis is retained with theliver graft after the splenic artery is divided close toits origin; the superior mesenteric artery (SMA) isprocured on an aortic CarTel patch with the pan-creasgraft. The pan creas is then placed in Univer-sity of Wisconsin preservation solution at 4°C.

If a replaced or accessory right hepatic arterycomes off the SMA, it is carefully dissected free

from the posterior surface of the pancreas. At thetime of pancreas removal, the SMA is dissected justdistalto fue right hepatic artery takeoff. The aorticCarrel patch is left bearing fue celiac axis and theproximal SMA with the liver graft, and the distalSMA with fue pan creas graft.

Pan creas- kidney specimens were procured asfollows. The operation starts as described above,but fue liver is not mobilized, and fue distalproper hepatic artery is ligated and divided, leav-ing the gastroduodenal artery intacto The com-mon hepatic artery is mobilized in a retrogradefashion, leaving the splenic artery intact and divid-ing fue left gastric artery. The aortic takeoff of fueceliac axis and fue SMA are identified. Mter theaortic flush is begun, the portal vein is transectedclase to fue porta hepatis to decompress pancre-atic outflow. The pancreas is then removed, leav-ing the celiac axis and fue SMA takeoffs on acommon aortic Carrel patch. AlI other steps arethen done as previously described.

Pancreas graft preparation (10, 11). The backtable preparation is always done with the pan creasimmersed in University of Wisconsin preservationsolution at 4°C. Mter re~oval of the spleen, fuevascular bundles of the transverse colonic andsmall bowel mesentery are re-ligated and short-ened. The proximal duodenal staple line is over-sewn and inverted using Prolene sutures.

Three techniques are used for arterial re-construction after combined pancreas-liverprocuremént:l. Y-graft reconstruction (most frequen t), using

the donar iliac artery bifurcation (after short-ening fue SMA, an end-to-end anastomosis tothe donar external iliac artery is created; fuedonar iliac artery is anastomosed end-to-end tothe splenic artery, leaving fue proximal com-mon donar iliac artery for anastomosis to fuerecipient inflow vessel)

2. End-to-side anastomosis between fue splenic ar-tery and fue SMA

3. Interposition graft, using a segment of donarinternal or external iliac artery between thesplenic artery and the SMA

Mter pan creas- kidney procurement, if the ce-liac axis and SMA have been procured with the

~ pan creas (aortic Carrel patch), no further dissec-tion or vascular reconstruction is done.

When venous reconstruction is necessary, fueportal vein is only rarely lengthened, because webelieve that a short portal vein decreases the riskof venous thrombosis by kinking or impingement.~

131Gruessner et al RISK OF PANCREAS TRANSPlANTATION

abdominal cavity is irrigated with a total of 4 L ofsolution containing cephalothin sodium (1,000mg/L) and amphotericin B (10 mg/L). Usually,no drains are inserted.

General postoperative caTe (14). Postoperatively,blood glucose levels are determined at fue bedsideat least every 2 hours. Insulin is administered in-travenously, as needed, to keep blood sugar levels< 150 mg/dL during the first 14 days after fuetransplant. For intravenous fluid requirements< 300 mL/hr, D5 W 0.5 normal saline plus 10 mEqNaHCO3/L is given; for intravenous fluid require-ments> 300 mL/hr, DI W 0.5 normal saline plus10 mEq NaHCO3/L. For all recipients, a nasogas-tric tube is placed at the time of transplant and isdiscontinued when the postoperative ileus re-salves, usually between dar 5 and 7. At that time, aclear liquid diet is started, which is advanced to aregular diet by fue time of hospital discharge,usually by dar 15.

Generally, recipients take acetylsalicylic acid(162 mg/day orally) for the first year after thetransplant.

Antimicrobial prophylaxis and treatment (8). Perio-peratively, all recipients in our series were givenantimicrobial prophylaxis. They received imipenem-cilastatin (Primaxin, Merck & Co. Inc., West Point,PA) alone (500 mg intravenously) or imipenem-cilastatin plus vancomycin (Vancocin, Eli Lilly &Co., Indianapolis, IN) (1,000 mg) at the time ofanesthetic induction and for 7 days after fue trans-planto Antimicrobial prophylaxis was adjusted ac-cording to fue results of duodenal culture. Onedouble-strength tablet per dar of trimethopri~-sulfamethoxazole (Bactrim, Hoffmann-LaRoche, Inc.,Nutley, NJ) was routinely'administered in the ab-sence of documented allergy to sulfonamides.

Antifungal prophylaxis, implemented in August1991, consisted of fluconazole (Diflucan, RoerigDivision, Pfizer Inc., New York, NY) (400 mg/ darintravenously), at fue time of anesthetic inductionand for 7 days after fue transplant. Postoperativenystatin swish and swallow (Nilstat, Lederle, Stan-dard Products, Wayne, NJ) or clotrimazole (My-celex, Troches, Bayer Pharmaceuticals, West Ha-ven, CT) was administered indefinitely to reducefungal colonization of the gastrointestinal tract.

Antiviral prophylaxis, as part of a controlledtrial, was administered to prevent infection soc-ondary to cytomegaiovirus and other herpes viruses.It consisted of acyclovir (Zovirax, Glaxo WellcomeInc., Research Triangle Park, NC) or gancicloVir(Cytovene, Roche Laboratories, Nutley, NJ).

If necessary, the portal vein extension graft is cre-ated using a segment of donar iliac vein, which isanastomosed end-to-end to the donar portal vein.

Recipient operation (12). The pancreas graft isusually placed intra-abdominally in the right sirleof fue pelvis for two reasons: dissection of thecommon iliac vessels is easier compared with theleft sirle and fue natural position of the right iliacvessels, which is vein lateral to the artery, do es notrequire vascular realignment, though on the leftsirle it might.

The common, external, and internal iliac arter-ies are completely mobilized; for right-sided dis-sections fue cecum is mobilized. For completemobilization of fue vein, we par particular atten-tion to division of all internal iliac veins. Mter thebladder is mobilized, nonuremic recipients aresystemically heparinized.

For the inflow vessel we generally use the com-mon or external iliac artery. Alternatively, the in-ternal iliac artery, the aorta, or the inferior mes-enteric artery can be used. If the graft is placed onthe right sirle, the iliac vein remains lateral to theiliac artery. If the graft is placed on the left sirle,the iliac vein is positioned medial to the iliacartery. For complete mobilization of the vein, theinternal iliac artery is ligated and divided. Vascularclamps are placed proximally and distally anta fueiliac vasculature. A venotomy and (more proxi-mally) an arteriotomy are then performed. Fourstay sutures are placed in the arterial wall forsubsequent anastomosis: any arterial plaques orintimal flaps are tacked at this point. The head offue pan creas is oriented in a caudad direction andthe tail is oriented in a cephalad direction. Vascu-lar anastomoses are created using running 6-0Prolene sutures for fue arterial and venous anas-tomases. Before unclamping, the pan creas graft isreperfused with 25 g of mannitol given intrave-nously to the recipient.

Next, the duodenocystostomy is created. It iseither hand-sewn (two-Iayer technique using poly-dioxanone for the inner and Prolene sutures forfue outer layer) or done with fue end-to-end anas-tomosis stapler (13). The duodenocystostomy isdone between the antimesenteric lateral aspect offue duodenum and the posterior-superior aspectof the bladder dome. The oren distal duodenalend is shortened (leaving about 10 cm of graftduodenum), and fue stump is stapled, oversewn,and inverted.

Mter the recipient operation is completed, the

VOLUME 185:128-144AUGUST 1997J AM COLL SURG

Immunosuppression (4). Prednisone (2 mg/kg/dar) was initiated intraoperatively for all recipi-ents and was tapered to 0.25 mg/kg/day by 6

months after the transplant. Azathioprine (Imu-Tan, Glaxo Wellcome Inc.) (2.5 mg/kg/day) wasgiven intravenously during the operation andcontinued orally (2.5 mg/kg/ dar) indefinitely if

the white blood cell count was 3,000 cells/mm3or more. Beginning on the dar after the opera-tion, a 7- to 14-day induction course of Minne-sota antilymphocyte globulin, or, after 1992, an-

tithymocyte globulin (both 20 mg/kg/dayintravenously) or OKT3 (Orthoclone OKT3, OrthoBiotech Inc., Raritan, NJ) (5 mg/kg/day intrave-

nously) was administered.PAK and PTA recipients were given cyclospor-

ine A (CsA) (Sandimmune, Sandoz Pharmaceuti-cals, East Hanover, NJ) (3 mg/kg/day intrave-nously) during fue operation and continuing for 5

to 7 days until oral administration was possible.

SPK recipients were given CsA (8 mg/kg/dayorally) beginning on dar 5. If the kidney allograftdysfunctioned, the introduction of CsA was de-layed until function was adequate or until the

antibody induction course was completed.For all recipients, maintenance therapy con-

sisted of prednisone, azathioprine, and CsA. CsAlevels were adjusted during fue induction andmaintenance phases to achieve whole blood levelsof200 to 250 ng/mL for fue first 6 months, 150 to200 ng/mL for the second 6 months, and 100 to

150 ng/mL thereafter, using high-pressure liquidchromatography. ..'

Rejection: diagnosis and treatment (4). Pancreas

rejection after solitary bladder-drained transplantswas defined by a decrease of urinary amylase levels

of 25% or more from baseline on two consecutivemeasurements or by diagnosis from a biopsy. Re-jection after solitary enteric-drained or duct-injected transplants was defined by an increase in

serum glucose or serum amylase levels or by diag-nosis from a biopsy. Irrespective of the manage-

ment of exocrine pancreas secretions, rejectionafter SPK transplants was frequently diagnosed byan increase in serum creatinine levels. Kidney bi-opsies were obtained whenever rejection was clin-ically suspected. In general, other causes of graftdysfunction (eg, infection, leak) were ruled out by

imaging studies and clinical findings.Surgical complications. Relaparotomy was defined

as any reoperative procedure involving fue intra-peritoneal or the retroperitoneal space, done dur-

132ing the first 3 months after transplant or duringthe initial pancreas transplani hospital stay if itexceeded 3 months. Causes of relaparotomy (de-termined by reviewing preoperative. operative.and postoperative findings) were classitied as fol-lows: infection (culture-proved intra-abdominalonly), vascular graft thrombosis. anastomotic andduodenal stump leak. pancreatitis. bleeding. andother causes (eg. acute cholecystitis). Relaparoto-mies were categorized as elective versus nonelec-tive, and nonrelated versus related to the preced-

ing pancreas transplant.The three most common indications for early

relaparotomy-intra-abdominal infection with orwithout graft pancreatitis. vascular graft thrombo-sis, and anastomotic and duodenal stump leaks-were studied in detail. Only symptomatic patientswith documented culture-positive intra-abdominalinfections were included in this study. Cultures~were obtained from aspirations guided by com-puted tomography (CT) or ultrasonography, or atthe time of abdominal exploration. Clinical symp-toms included fever, purulent wound drainage.ileus, diarrhea, acute abdomen, and sepsis. Throm-bosis was diagnosed by clinical symptoms, imagingstudy results, intraoperative findings, or a combina-tion thereof. Clinical symptoms included graft ten-demess, rapidly declining or absent urinary amylaselevels. unexplained hyperglycemia. rapidly and un-expectedly increaSing insulin requirements, dark he-matufia aríd unexplained hemoperitoneum, leuko-cytosis, and thrombocytopenia. Imaging studiesincluded color duplex Doppler ultrasonography,pancreas perfusion scintigraphy, cr scan (withintravenous contrast enhancement of the pan-creas graft), and arteriography. The thromboseswere differentiated (arterial versus venous) basedon pre- and intraoperative findings, imaging studyresults, and the histopathologic reporto Anasto-motic and duodenal stump leaks after bladder-drained. whole-organ, cadaver donar pancreastransplants were diagnosed by clinical symptoms,imaging study results, laboratory findings, orcombination thereof. Clinical symptoms includeabdominal pain, abdominal distention, fever, decreased urine output, peritonitis. and improv

..:0 ment after placement of Foley catheter. Imaginstudies included cystogram and CT scan with bladder contrasto Laboratory findings included an increase in serum amylase, decrease in urine amtase, or increase in serum creatinine levels, orcombination thereof. The leaks were differen .

Gruessner et al 133RISK OF PANCREAS TRANSPLANTATION

ated (anastomotic versus duodenal stump) basedon pre- and intraoperative findings, as well as im-aging study results.

Statistical analysis. Categorical variables were an-alyzed using the chi-square test, and when appli-cable, Fisher's exact test. Continuous variableswere analyzed parametrically using the t test andnonparametrically using the Mann-Whitney Utest.

Pancreas graft and patient survival rates werecalculated according to the Kaplan-Meier method.For pancreas grafts, fue time of graft loss wasdetermined by return to exogenous insulin useafter insulin independence; for kidney grafts, byreturn to permanent dialysis. For PAK, kidneygraft survival rates refer to the date of fue pancreastransplant. Calculation ofpatient survival includeddeaths occurring after pancreas and kidney graftloss. Survival rates were compared among groupsusing the generalized Wilcoxon test. For all uni-variate statistical tests, p values < 0.05 were con-sidered significant.

Risk factors for graft and patient survival werestudied (respectively) in four different regressionanalyses: all recipients, SPK only, PAK only, andPTA only. Variables included donar age (per 10-year increments), recipient age (per 10-year incre-ments), preservation time, transplant number(primary versus retransplant), recipient category(SPK, PTA, PAK), duration of diabetes (per 10-year increments), transplant year (per 1-year in-crements), human leukocyte antigen mismatch(number ofmismatches), cause of death (cerebro-vascular versus traumatic), and risk factors specificto certain surgical complications (eg, thrombosisafter left-sided versus right-sided pancreas graftplacement). For all multivariate analyses, p value< 0.15 was considered significant.

Results

Relaparotomy. Relaparotomy was required after32% of all pan creas transplants. The relaparotomyrate was highest for SPK (36%), followed by PAK(25%) and PTA (16%) (SPKand PAK versus PTA:p < 0.03). Multiple relaparotomies were necessaryfor 9% of all recipients: 12% for SPK, 9% for PAK,and 4% for PTA (p = NS). The most commonindications for relaparotomy were intra-abdominalinfection with or without graft pancreatitis, vascu-lar graft thrombosis, anastomotic and duodenalstump leaks, and bleeding. In all three recipientcategories, the highest relaparotomy rate was for

infection with or without graft pancreatitis. AlIrelaparotomies were nonelective, and all were re-lated to the preceding pancreas transplant. AlIwere done as oren operations, with no laparo-scopic procedures.

For 84 SPK recipients, 153 relaparotomies (1.8per recipient) were done; for 35 PAK recipients,57 were done (1.6); and for 23 PTA recipients, 27were done (1.2).

The most common reoperative procedures weredrainage of intra-abdominal abs~ess and pancre-atic fluid with graft pancreas necrosectomy (a totalof 119 relaparotomies: SPK, 84; PAK, 26; PTA, 9);transplant pancreatectomy (a total of 81 relapa-rotomies: SPK, 42; PAK, 25; PTA, 14); intra-abdominal hemostasis with evacuation of hema-toma (a total of 24 relaparotomies: SPK, 13; PAK,7; PTA, 4); and repair of leak at the duodenocys-tostomy or duodenal stump (a total of 20 relapa-rotomies: SPK, 13; PAK, 6; PTA, 1).

Patient survival rates at 1 and 5 years with versuswithout relaparotomy were as follows: for SPK,77% and 65% versus 87% and 78% (p = 0.04); forPAK, 79% and 76% versus 98% and 81 %(p = 0.02); for PTA, 80% and 74% versus 95% and70% (p = 0.03) (Figs. 1-3).

Graft survival rates at 1 and 5 years with versuswithout relaparotomy were as follows: for SPK,32% and 20% versus 82% and 70% (p = 0.0001);for PAK, 11 % and 11 % versus 71 % and 38% (p =0.0001); for PTA, 26% and 8% versus 60% and35% (p = 0.0001) (Figs. 1-3).

Overall Cox regression analysis showed that therisk of death was significantIy higher for SPK ver-sus PAK (RR = 1.6, P =.0.1) and for SPK versusPTA (RR = 1.9, P = 0.07) recipients. Cox regres-sion analyses showed the following significant riskfactors for death: for SPK, older recipient age(RR = 1.4, P = 0.05), retransplant (RR = 2.3,P = 0.05), relaparotomy for infection with or with-out graft pancreatitis (RR = 2.2, P = 0.03), andrelaparotomy for anastomotic leak (RR = 2.3,P = 0.09); for PAK, older recipient age (RR = 2.6,P = 0.01), retransplant (RR = 6.2, P = 0.001), andrelaparotomy for thrombosis (RR = 7.6, P =0.007); for PTA, long preservation time (RR = 3.6,P = 0.01) and relaparotomy for bleeding (RR =21.7, P = 0.002).

Overall Cox regression analysis showed that therisk of graft loss was significantIy higher for PAKversus SPK (RR = 1.8, P = 0.0004) and for PTAversus SPK (RR = 1.8, P = 0.0004) recipients.

~

134 J AM COLL SURG AUGUST 1997

100 '8-g~ ~E5 558 888 8 ~~ 8 8 e381

,7575

5050 I

~c

iü>

.~~:n

(

~o

~>

.~~/)

25

o

o I , , , , , ,o 6 12 18 24 30 36

p=O,O27 Patlen! Months Posttransplantp=O,OOO1 Gral!o

p=O.O41 Patientp=O.OOO1 Graft

6 12 18 24 30

Months Posttransplant

36

FIG 3. Patient and pancreas graft survival rates for pancreastransplant alone recipients with versus without relaparotorny.

FIG 1. Patient and pan creas graft survival rates for sirnul-taneous pancreas and kidney recipients with versus withoutrelaparotorny.

Cox regression analyses showed fue following sig-nificant risk factors for graft loss: for SPK, olderdonor age (RR = 1.1, P = 0.13), relaparotomy forthrombosis (RR = 22, P = 0.001), and relapa-rotomy for intra-abdominal infection with or with-out graft pancreatitis (RR = 3.6, P = 0.0001); forPAK, older donar age (RR = 1.2, P = 0.13),retransplant (RR = 1.5, P = 0.13), relaparotomyfor thrombosis (RR = 15, P = 0.0001), relapa-rotomy for intra-abdominal infection and graftpancreatitis (RR = 9.3, P = 0.004), relaparotomyfor anastomotic and duodenal leaks (RR = 5.6,P = 0.007), and relaparotomy for bleeding(RR = 3.6, P = 0.06); for PTA recipients. relapa-rotomy for thrombosis (RR = 9.4, P = 0.02).

¡neTa-abdominal inJection. lntra-abdominal infec-tion occurred after 20% of all pancreas trans-plants. Most infections were diagnosed within fuefirst 3 months after the transplant: 61 % within the

first month, 79% within the first 2 months, and96% within the first 3 months after fue transplant.In 40% of recipients with infections, fue diagnosiswas made during the initial hospitalization, and60% were readmitted at various intervals post-transplant. Intra-abdominal infections were gener-alized (ie, diffuse peritonitis) in 51% and localizedin 49%. In 30%, an anastomotic or duodenalstump leak was prese};1t; in 30%, transplantpancreatitis. Of recipients with infections, 76%had a functioning graft at fue time of diagnosis.Causes of graft failure before intra-abdominal in-fection occurred were vascular graft thrombosis(14%), rejection (7%), and graft pancreatitis(3%). The incidence ofintra-abdominal infectionwas highest after graft thrombosis: 27% of recipi-ents with vascular graft thrombosis had an intra-abdominal infection.

By recipient category, fue rate of intra-abdominal infection was 18% for SPK, 24% forPAK, and 20% for PTA recipients (p = NS). Bydonar type, fue rate was significantly higher forliving related donar (42%) versus cadaver donarrecipients (18%) (p = 0.001). By duct manage-ment technique, fue rate was significantly higherfor enteric-drained (39%) versus bladder-drainedtransplants (18%) (p = 0.001). In fue SPK cate-gory, neither dialysis dependence nor dialysismode affected fue rate; of recipients with infections,21 % were on peritoneal dialysis, 46% on hemo-dialysis, and 33% not on dialysis at all (p = 0.3).

Of 89 intra-abdominal infections, 48 (54%)were caused by bacterial species only, 13 (15%) byfungal species only, and 28 (31 %) by bacterial andfungal species. Of sol e bacterial infections, fuemost common microorganisms were StaPhylococcusspecies; of sale fungal infections, Candida albicans

1001

75

~~;~~;;;~~~~=

-ePatient without Relaparotomy.¡,.Patient with Relaparotomy-Pancreas without Relaparotomy

"'Pancreas with Relaparotomy

50(/)

~o

iV>

.~~

25 ~ 1-

...v , , , , ,o 6 12 18 24 30 36

p=O.O20 Patlentp=O.OOO1 Gral! Months Posttransplant

FIc 2. Patient and pancreas graft survival rates for pancreasafter kidney recipients with versus without relaparotorny.

Gruessner et al RISK OF PANCREAS TRANSPLANTATION 135

la>

';:...:J(/)

~

FIG 5. Patient and pancreas graft sulVival rates for pan-creas after kidney recipients with versus without intra-abdominal infection.

infections were most common. Staphylococcus spe-cies and C. albicans were the most common causesof mixed infections. Of note, positive donar duo-denal cultures were not associated with increasedintra-abdominal infection rates; for recipients withversus those without infection posttransplant, cul-tures taken from the donar duodenum were pos-itive in 17% versus 18% of cases (p = NS).

Of 89 intra-abdominal infections, 80 (90%) re-quired exploratory laparotomies, 56% required 1relaparotomy, 29% required 2, and 15% requiredmore than 2. A transplant pancreatectomy wasperformed in 70% of recipients undergoing re-laparotomy, Only 9 (10%) intra-abdominal infec-tions were successfully treated with percutaneousdrainage of the peripancreatic abscess.

Pancreas graft loss from intra-abdominal infec-tion was a majar risk factor for a subsequent pan-creas retransplant. Of 13 pan creas retransplantsdone after a previous graft loss from infection, 12(92%) retransplants were again complicated byinfection. Of note, 10 of those infections at re-transplant were caused by the same microbial spe-cies that had caused failure of the previous graft,after an interval of 1-5 years.

Intra-abdominal infection had a significant im-pact on both patient and graft survival. Overallpatient survival at 1 and 3 years was 76% and 74%for recipients with versus 92% and 84% withoutinfection (p = 0.005) (Figs. 4-6). By recipientcategory, patient survival was significantly loweronly for SPK, but not PAK and PTA, recipients.Overall graft survival at 1 and 3 years was 17% and11 % for recipients withversus 63% and 50% with-out infection (p = 0.0001) (Figs. 4-6). Graft sur-

vival was significantly lower for recipients with in-fection regardless of fue recipient category. ForSPK recipients, the rate of graft loss from intra-abdominal infection was not significantly differ-ent for those on peritoneal dialysis versus thoseon hemodialysis (p = 0.1); but it was higher forthose on hemo- or peritoneal dialysis versusthose not on dialysis (p = 0.04).

Outcomes after sale bacterial infections wasmore favorable than after sale fungal or combinedfungal and bacterial infections. Graft functioncontinued in 44% of recipients with sol e bacterialversus only 22% with sale fungal or combinedfungal and bacterial infections (p = 0.03). Graftpancreatectomy was required in 56% of recipientswith sale bacteria! versus 78% with sale fungalor combined fungal and bacterial infections(p = 0.03). Mortality was, 6% for recipients with

80-~-KJ--fb.-6--aJ75-11\

'fIm>

">...::1(/)

~o

~--60-

-o- Palienl without Infection,. Palient with Infection--o-- Pancreas without Infection

--t.- Pancreas with Infection40

--D-- Patient without Infection-Patient with Infection

~~'- :- ~~~creas without Infection-u~~~s with Infection

--~

ti!>os;...:J(/)

~20-

p=0.1620 Palien!p=0.0001 Gral!

O I , , , , , ,O 6 12 18 24 30 36

Months Posttransplant

o

p=O.OO6 Pa!ien!p=O.OOO1 Gral!

6 12 18 24

Months Posttransplant

30 36

FIG 4. Patient and pancreas graft survival rates for simul-taneous pancreas and kidney recipients with versus withoutintra-abdominal infection.

FIG 6. Patient and pancreas graft survival rates for pan-creas transplant alone recipients with versus without intra-abdominal infection.

136 J AM COLL SURG c --

Table l. lncidence ofVascular Thrombosis by RecipientCategory and Donor Type

AUGUST 1997 VOLUME 185:128-144

Thromboses

%

18 (9/9)18 (9/9)O

21 (9/12)18 (6/12)O3 (3/0)

14 (4/10)9 (2/7)1 (l/O)4 (1/3)

20.619.8

30.010.1

8.410.019.0

SPK (n = 198)CAD, whole-organ (n = 196)CAD, segmental (n = 2)

PAK (n = 102)CAD, whole-organ (n = 91)CAD, segmental (n = 1)LRD, segmental (n = 10)

PTA (n = 138)CAD, whole-organ (n = 107)CAD, segmental (n = 10)LRD, segmental (n = 21)

SPK, simultaneous pancreas-kidney transplant; CAD. cadaver donor; PAK,pancreas after pre\ious kidney transplant; LRD, li\ing related donor; l'TA,pancreas transplant alone.

sale bacterial versus 20% with sol e fungal or com-bined fungal and bacterial infections (p = 0.05).

For SPK recipients, the risk of intra-abdominalinfection was increased by older donar age(RR = 1.03, P = 0.001), older recipient age(RR = 1.05, P = 0.02), enteric (versus bladder)drainage (RR = 0.24, P = 0.02),andvasculargraftthrombosis (RR = 25.9, P = 0.001). For PAKrecipients, the risk was increased by older donarage (RR = 1.05, P = 0.01), retransplant (versusprimary transplant) (RR = 2.19, P = 0.03), andvascular graft thrombosis (RR = 5.1, P = 0.01).For PTA recipients, the risk was increased byolderrecipient age (RR = 1.15, P = 0.0007), enteric(versus bladder) drainage (RR = 0.09, P = 0.06),long preservation time (RR = 1.12, P = 0.07), andvascular graft thrombosis (RR = 1.2,1> = 0.0001).

Vascular graft thrombosis. The overall incidence ofvascular graft thrombosis was 12%; venous vascu-lar graft thrombosis (7%; n = 31) was slightlymore frequent than arterial graft thrombosis (5%;n = 22). Thromboses occurred early after trans-

plant: 47% within 3 days, 70% ,vithin 7 days, and90% within 21 days. The overall thrombosis rate wasnot significantIy different for primary (12%; n = 41)versus retransplant (14%; n = 12) recipients.

By recipient category, \'ascular gratt thrombosiswas significantIy higher for PAK (21 %) versus PTA(10%) and SPK (9%) recipients (p < 0.05) (Table1). For cadaver donar whole-organ grafts, fue rateof vascular graft thrombosis \\'aS lowest for PTArecipients (8%). The thrombosis rate was higherfor living related donar (23%) versus cadaver do-nor (11 %) recipients, albeit not significantIy(p = 0.08). Likewise, fue thrombosis rate washigher for living related donar and cadaver donarrecipients of segmental (18%) versus whole-organgrafts(II%) (p=0.2).

By type of vascular pancreas graft reconstruc-tion, the lowest thrombosis rate was after Y-graftreconstruction (10%), and when no arterial re-construction was required and an aortic Carrelpatch was used (10%). This rate was significantlylower than after construction of an end-to-sideanastomosis between the splenic artery and SMA(21 %) or after placement of an interposition graftbetween the splenic artery and SMA (16%) (Table2). Portal vein extension grafts did not in creasethe incidence of vascular graft thrombosis; 11 % ofthe grafts without versus 15% with portal veinextension grafts thrombosed (p = NS).

By pancreas graft placement, the rate ofvascularthrombosis was 9% for medial placement (ie, in-flow through the aorta of fue inferior mesentericartery), 10% for right-sided placement, 21% forleft medial placement (ie, medial to fue sigmoidcolon mesentery), and 37% for left lateral place-ment (ie, lateral to the sigmoid colon mesentery).The difference between left-sided placement ver-sus all other implantation sites was statisticalIy sig-

Table 2. Incidence of Vascular Thrombosis for Whole-organ Cadaver Donor Grafts by Type of Arteria) Reconstruction

OtherNo reconstruction

Thromboses

n n %

Y-graft Interposition gtaft

Thromboses

h n %

End-to-sideanastomosis

Thromboses

n n %% %

SPK (n = 196)PAK (n = 91)PTA (n = 107)

n

1384

*Interposition graft betWeen splenic artery and gastroduodenal artery; superior mesenteric artery anastomosed to recipient inflo". vessel.tlnterposition graft betWeen splenic artery and right renal artery.sIncludes 2 pan creases with individual anastomosis of splenic artery or superior mesenteric artery to recipient inflow vessels (no thromboses): 1 pancreas with

splenic anerycoming offthe aorta (anatomic anomaly, no superior mesenteric artery) , the other with superior mesenteric arteryonly (anatomic anomaly, no splenic

artery).¡End-to-side anastomosis of splenic artery to celiac axis (functional aortic Carrel patch).SPK, simultaneous pancreas-kidney transplant; PAK, pancreas after previous kidney transplant; PTA, pancreas transplam alone.

Gruessner et al 137RISK OF PANCREAS TRANSPLANTATION

Table 3. Incidence of Vascular Thrombosis by Graft Placement

MedianRight sirle,, Thromboses

n %

Left lateral*

Thromboses

n %

Left medialt

Thromboses

n % %nn n

o100 6 o17O

9O

o 1 11892

21

oo

oo

281

61

21100

1817

53

2 4033

566

101

11

3325

oo

oo

34

3O

33O

4O

SPK (n = 190)Whole-organ (n = 196)Segmental (n = 2)

PAK (n = 102)Whole-organ (n = 91)Segmental (n = 11)

PTA (n = 138)Whole-organ (n = 107)Segmental (n = 31)

9727

74

715 o

*Lateral lo Ihe sigmoid colon mesentery.tMedial lo Ihe sigmoid colon mesentery.SPK, simultaneous pancreas-kidney Iransplanl; PAK, pancreas afler previous kidney transplant; PTA, pancreas lransplant alone.

nificant (p = 0.01). For living related donar grafts,the thrombosis rate for right-sided placement(16%) was lower versus left-sided or median place-ment (50%) (Table 3).

Duct management technique (bladder versusenteric drainage), preservation solution (silica-gel-fractionated plasma versus University ofWisconsinsolution), and type of recipient arterial inflow ves-gel (common versus external versus internal iliacartery versus aorta versus other vessels) did notaffect the vascular graft thrombosis rate by univar-iate analysis. Nor did the dialysis status at trans-plant for SPK recipients; the thrombosis rate was8% for those on dialysis versus 10% for those noton dialysis.

Patient survival rates were not significantly lowerfor recipients with versus without vascular graftthrombosis. For SPK recipients, patient survivalrates at 1 and 3 years were 91 % and 91 % withversus 83% and 77% without thrombosis (p =0.4); for PAK recipients, 86% and 76% with versus93% and 88% without throinbosis (p = 0.4); forPTArecipients, 83% and 83% with versus 94% and89% without thrombosis (p = 0.3).

Cox regression analysis showed that PAK recip-ients had a significantly higher risk for graftthrombosis than SPK recipients (RR = 0.3,P = 0.002). For SPK recipients, significant inde-pendent risk factors for graft thrombosis were car-diocerebrovascular cause of donar death (RR = 5.2,P = 0.02), end-to-side anastomosis of the splenicartery to the superior mesenteric artery (RR = 3.6,P = 0.14), and the use of an aortic Carrel patch(RR = 3.5, P = 0.05). For P AK recipients, the graftthrombosis risk was increased by older donar age(RR = 1.05, P = 0.05), cardiocerebrovascular

cause of donar death (RR = 5.2, P = 0.03), leftmedial pancreas graft placement (RR = 5.0,P = 0.03), and graft pancreatitis (RR = 12.7,P = 0.001). For PTA recipients, the graft throm-bosis risk was increased by older donar age (RR =1.2, P = 0.05), fue use of an interposition graftbetween the splenic artery and fue superior mes-enteric artery (RR > 25, P = 0.03), end-to-sideanasteomosis of fue splenic artery and fue supe-rior mesenteric artery (RR > 25, P = 0.01), andgraft pancreatitis (RR = 7.7, p = 0.17).

Anastomotic and duodenal stump leaks. Anasto-motic and duodenal stump leaks occurred after10% of all bladder-drained, whole-organ, cadaverdonar pancreas transplants. Of fue 42 duodenalleaks, 29% occurred early (mean, 11 days; range,1-28 days), and 71 % late (mean, 7 months; range,1-36 months). The site of the leak was the duode-nocystostomy site (true bladder anastomotic leaks)in 36%, the stapled proximal duodenal stump in19%, and fue stapled distal duodenal stump in9%. In the remaining 36%, it was impossible toidentify the exact site of the leak because of con-servative treatment or indeterminate diagnosticstudies (cystogram and CT scan).

The most common symptoms were abdominalraiD (88%), abdominal distention (69%), rever(57%), decreased urine output (31 %), peritonitis(26%), improvement after placement of a Foleycatheter (24%), and vomiting (12%). Serum amy-lase levels rase in 48%, urine arnylase levels de-creased in 38%, and serum creatinine levels rasein 33%. The incidence of leaks did not differ by

recipient category.We operated on 30 recipients with leaks (71 %).

In 23 of them, the leak was oversewn; in 1 fue leak

138 J AM COLL SURG AUGUST 1997 VOLUME 185:128-144

was already sealed by the time of fue operation. Infue other 6 recipients, graft pancreatectomy wasrequired because of duodenal blowout in conjunc-tion with graft thrombosis (n = 1), a large perlo-ration of fue duodenum and bleeding (n = 1),and peritonitis (n = 4). We treated 12leaks (29%)conservatively with an indwelling catheter for 1-2months until resolution.

In 10% leaks recurred 1-5 months after the firstleak had be en diagnosed. In 2 recipients a bladderanastomosis was converted to enteric drainage; in1 fue duodenocystostomy was taken clown withprimary bladder closure, resection of fue duo-denal graft, and ligation of the pancreatic duct;and in 1 the leak was oversewn.

Discussion

The rate of technica! failures for pan creas trans-plants is higher than for any other routinely per-forrned solid organ transplant. Nonimmunologicgraft failure, in particular, has hampered morewidespread application of solitary pan creas trans-plants (PTA, PAK). Severa! factors contribute tosurgical complications after pan creas transpIants:

l. The underlying disease itself which, for exam-pIe, makes patients more prone to infectionsand, in the presence of secondary complica-tions of diabetes mellitus, increases cardiacand cerebrovascular morbidity and mortaIityperitransplant

2. The transplant procedure, whi& invoIves twohollow viscera (duodenum; bladder or intestine)

3. The transpIanted organ, with its broad spec-trum of potential surgicaI compIications suchas pancreatitis, infection, necrosis, pseudocyst,and fistula

4. Extended (compared with kidney or Iiver trans-plants) anti-T-cell induction therapy

5. Specific risk factors pertaining to the recipientcategory, such as uremia in SPK recipients orchronic pretransplant immunosuppression inPAK rec~pients

For all these reasons, it is not surprising that ouroverall relaparotomy rate was 32%. This figure iscomparable to the reIaparotomy rate reported byother pancreas transplant centers, such as fueIowa group (28%) and fue Nebraska group (36%)(15,16). The Wisconsin group reported an overallsurgical complication rate of 31 %, fue majorityrequiring reIaparotomy (24%) (17). By recipient

category, our relaparotomy rate was significantlyhigher for SPK and PAK (versus PTA) recipients.This finding reflects the technically more demand-ing dual-organ transplant procedure for uremic orpreuremic SPK recipients, and the preex.istingchronically immunosuppressed state of P AK recip-ients. Likewise, the rate of multiple relaparoto-mies was higher for SPK and PAK (versus PTA)recipients.

Relaparotomy for surgical complications had asignificant impact on patient and graft suIVivalrates. Both were significantly lower at 1 and 5 yearsafter transplant for recipients with versus withoutrelaparotomy, irrespective of the recipient cate-gorro Multivariate analysis showed that the risk ofdeath was significantly higher for SPK and PAK(versus PTA) recipients. Cox regression analysesshowed that for SPK recipients, relaparotomy forinfection or anastomotic leak increased the risk ofdeath; for PAK recipients, relaparotomy for throm-bosis; and for PTA recipients, relaparotomy forbleeding.

The most significant risk factors for graft sur-vival by recipient category were relaparotomy forthrombosis or infection (SPK) and relaparotomyfor thrombosis (PTA). For PAK.recipients, relapa-rotomy for any cause was associated with ahigher risk of graft loss, which again can beexplained by .their long standing chronic pre-transplant immunosuppression.

The impact of relaparotomy on perioperativemorbidity and mortality includes an increase inmedian hospital charges: for SPK recipients, by$65,700; for PAK recipients, by $45,100; and forPTA recipients, by $35,800. Thus, the cost of apancreas transplant with (versus without) relapa-rotomy was increased by 65% for SPK, 59% forPAK, and 46% for PTA recipients. In fact, in aprevious analysis, we showed that relaparotomywas the most important risk factor for an increasein hospital charges (18).

The most frequent indications for relaparotomyin our series were intra-abdominal infection, vas-cular graft thrombosis, and anastomotic and duo-denalleaks (19). This finding is in line with pre-vious reports from other pancreas transplantcenters (15-17), but we were interested in thespecific impact on patient and graft outcome ofeach of these indications.

Intra-abdominal infection occurred after 20%of all pancreas transplants. We have recently re-ported on contributing factors, such as postreper-~

Gruessner et al 139RlSK OF PANCREAS TRANSPlANTATION

fusion pancreatitis and peripancreatitis (with localrelease of proinflammatory substances, providingan optimal environment for growth of microor-ganisms) , contamination of the donor duodenum,microbial translocation through fue duodenalwaII, prolonged duration of surgery, perioperativeimmunosuppression, underlying disease, and sur-gical complications (9). Our infection rate is com-parable to reports from fue Wisconsin group(17%) and fue Iowa group (18%) (15,20).

In our series, at the time of diagnosis, only 76%of recipients with infections still ha~ a functioninggraft. We noted a particularly high incidence ofinfection after ~t pancreatectomy for thrombo-sis (27%). Interestingly, transplant pancreatitiswas diagnosed in 30% of recipients with infec-tions, although it is difficult to retrospectively ana-lyze which carne first, graft pancreatitis with sub-sequent infection or perigraft infection withsubsequent pancreatitis. Intra-abdominal infec-tion was fue most common indication for single ormultiple relaparotomies and the second mostcommon indication for graft pancreatectomy. Ithad a more significant impact on costs than anyother surgical complication.

By recipient category, we found a slightly higherrate of infection for PAR recipients. This findingmatch es fue previous report by the Nebraskagroup (17). It can be explained by the longtermchronic immunosuppressed state of PAK recipi-ents. As with our first experience with intra-abdominal infections that we reported 10 yearsago, the infection rate in this series was againhigher for enteric-drained (39%) versus bladder-drained (18%) transplants (21). The rate was alsohigher for living related donar recipients (42%),most of whom had enteric-drained transplants,versus cadaver donar recipients (18%), most ofwhom had bladder-drained transplants. For SPKrecipients, the infection rate was not different forrecipients on hemo- or peritoneal dialysis versusthose not on dialysis (22). Thus, recipients onperitoneal dialysis should not be excluded fromintraperitoneal pancreas graft placement.

One of our concerns regarding fue microbiol-ogy of intra-abdominal infection was not con-firmed in our retrospective review: positive donorcultures were not associated with an increasedinfection rate (23). This finding mar reflect thefacts that recipients were prophylactically on anti-biotics for 7 days and that antibody coverage waschanged according to sensitivity test results. Duo-

denal culture results appear to be useful for de-termining fue length, specificity, and microbialtarget of perioperative antimicrobial prophylaxis.

In our series, 90% of recipients with infectionsrequired exploratory laparotomy; of fuese, 45%underwent multiple relaparotomies and 70% aneventual graft pancreatectomy. As previouslyshown, pancreas graft loss resulting from infectionwas the majar risk factor for a pan creas retrans-plant; most pancreas retransplants done after fueprevious graft was lost owing to infection wereagain complicated by infection (24). The infec-tions were caused by fue same microbial species,even though fue retransplan ts were done 1-5 yearsafter fue first transplant. Do fuese microorganismssurvive dormantly in the abdominal cavity for suchlong asymptomatic periods and then become ac-tive again at fue time of retransplant when immu-nosuppression is restarted? We did not find anyintra-abdominal abscesses, but sensitivity test re-sults were identical for microbial species causingintra-abdominal infections after fue first trans-plant and after the retransplant. This finding sug-gests that fuese microorganisms persist even afterthe first infection resolves clinically.

The overall incidence of just fungal and com-bined fungal and bacterial infections in our serieswas 46%. It is known from nontransplant surgerythat the pancreas is particularly susceptible to fun-gal infections; fue reported infection rate is 23%after pancreatic surgery for benign diseases (25).Yet we noticed a marked difference in outcomesafter fungal versus bacterial infections (8). Graftpancreatectomy was required in 78% of recipientswith sale fungal, or combined fungal and bacte-rial, infections versus 56% of recipients with salebacterial infections. Furthermore, mortality was20% for recipients with fungal, or combined fun-gal and bacterial, infections versus 6% with salebacterial infections.

Infections significantly decreased overall patientand graft survival at 1-3 years posttransplant. Pa-tient survival rates by recipient category were sig-nificantly lower only for SPK recipients, but notfor PAK or PTA recipients. Yet graft survival wassignificantly lower regardless of the recipientcategory.

Of note, for SPK recipien ts, fue rate of graft lossowing to infection was higher for those on (hemo-or peritoneal) dialysis than for those not on dial-ysis. This is again a strong argument for preemp-tive SPK transplants in an attempt to decreaseperitransplant morbidity and overall costs.

140 J AM COLL SURG AUGUST 1997 VOLUME 185:128-144

Ifwe compare the resultsof our current analysiswith our first experience with infections published10 years ago, we must admit that, despite improve-ments in surgical technique and perioperativecare, fue incidence of infections has not changed.But the percentage of recipients who recover frominfection with graft function has increased from19% to 34%, and fue mortality rate has decreasedfrom 27% to 12%. Reasons may include greaterexperience in routinely managing this (frequent)complication and increased vigilance, leading toearlier diagnosis and treatrnent.

Next to intra-abdominal infection, vascular graftthrombosis was fue most common indication forrelaparotomy. For graft pancreatectomy, vasculargraft thrombosis was the most common indication.This finding is in line with IPTR data showing thatthrombosis accounts for almost 60% of all nonim-munologic graft failures (2). In our series, fueoverall incidence of vascular graft thrombosis was12%. This incidence is comparable to reportsfrom other centers, where graft thrombosis rangesbetween 10% and 39% (26-29). Only the Wiscon-sin group has consistently reported a < 1 % inci-dence, but for SPK recipients only (30).

In our series, PAR recipients had a thrombosisrate twice as high (21%) as PTA (10%) or SPK(9%) recipients. A higher thrombosis rate for PARrecipients (39%) was algo reported by fue Ne-braska group (29). One might speculate that, inthis category, either the longterm chronic immu-nosuppression required for the kidney transplantunfavorably alters the recipient's -coagulatory sys-tem (owing to the procoagulan t effects of CsA andprednisone) or the technical adaptations andmodifications (in light of a right-sided kidneytransplant) required for the subsequent pancreastransplant increases fue risk of thrombosis. Wealgo observed a higher thrombosis rate for seg-mental graft recipients. Although flow throughfue splenic artery and vein does not differ forsegmental versus whole-organ grafts, the shortersplenic artery and vein in segmental grafts allowfor less flexibility and are more prone to bendingor twisting. In addition, fue transverse pancreaticartery, which supplies fue body and the tail of thepan creas, and which receives collateral flow from "fue SMA (31), is not revascularized in segmentalgrafts. Thus, fue blood supply of the segrnentalgraft is critically dependent on the splenic artery;any partial or complete thrombosis would be moredetrimental than for whole-organ grafts. The ad-

vantage of the dual blood supply of the whole-organ graft has algo been shown for recipients withstable graft function: even with clinically occultthrombosis of fue SMA or fue splenic artery, doc-umented on angiogram, their graft function wasmaintained by collateral flow (32). It is thereforenot surprising that fue thrombosis rate for seg-mental grafts was twice that of cadaver donargrafts.

We algo observed that fue type of arterial recon-struction affected fue incidence of thrombosis.The lowest thrombosis rate was after Y-graft recon-struction or when an aortic Carrel patch was used.When the end-to-side anastomosis or interpositiongraft techniques were used, fue rates were signifi-cantly higher, most likely owing to less spatial flex-ibility than with a Y-graft or owing to preexistingdonar arteriosclerotic disease (ie, when fue donol;Y-graft could not be used for that reason). Incontrast to arterial reconstruction, portal vein re-construction (ie, extension grafts) had no impacton the incidence of thrombosis.

Another important risk factor for thrombosiswas fue implantation ~ite. Usually, fue side ofchoice is fue right iliac system, with complete ma-bilization of the right iliac vein, which allows cre-ation of a tension-free venous anastomosis lateralto the arterial anastomosis (12); however, whenthe previous kidney graft has be en placed in theright side or with a retransplant, fue left iliacsystem or'the aorta has to be used. On fue left side,the sigmoid colon is in fue way, and the graft hasto be positioned either medially (ie, auto the com-mon iliac artery) or late rally (ie, anta fue externaliliac artery). The left common iliac vein is tetheredby the internal iliac artery and is much more pos-terior than its right counterpart. This explains thethrombosis rate of 10% for right-sided place-ment versus 21% for left medial placement and37% for left lateral placement for cadaver donargrafts. Likewise, for living relative-donated grafts,fue thrombosis rate was 16% for right-sided versus50% for left-sided placement (11). Interestingly,median placement (ie, the aorta) had a low risk ofgraft thrombosis; it should be considered for pa-tients in whom left-sided placement appears diffi-.culto

In our experience, cross-clamping of fue dis-tal abdominal aorta or the proximal iliac arteryhas never had any detrimental consequences forkidney grafts located distally to fue temporary oc-clusion (such as for PAK recipients) (11). Butstrong consideration must be given to left-sided

Gruessner et al RISK OF PANCREAS TRANSPLANTATION 141

with versus without anastomotic or duodenalleaks.

We have confined this retrospective analysis tothe most serious surgical complications after pan-creas transplants. But other surgical complicationsdeserve a brief mention.

Graft pancreatitis and graft peripancreatitis aremost common for recipients with intra-abdominalinfection or vascular graft thrombosis. Complica-tions of graft pancreatitis include pancreatic ab-scess, pancreatic necrosis, perigraft infection, ster-ile pancreatic and peripancreatic fluid collections,and pseudocyst. Graft pancreatitis can be relatedto donar risk factors (hemodynamic instability,vasopressure administration) , procurement injury,perfusion injury (excessive amounts of flush vol-time or perfusion pressure), preservation injury,and reperfusion injury. Vasoconstriction, intravas-cular coagulation, and increased endothelial per-meability mar result in pancreatic edema, which,in tufO, can impair venous drainage. With pancre-atic microcirculation progressively impaired, graftthrombosis can resulto

Superficial wound infections usually require re-exploration only in combination with deep (ie,intra-abdominal) infections (35, 36). We previ-ously reported an 18% incidence of superficialwound infections; 45% were in combination withdeep wound infections. We identified prolongedoperating time, enteric drainage, and older donarage as risk factors for superficial infection. Themicrobiologic spectrum was almost identical tothat of intra-abdominal infections.

Complications of subsequent pan creas trans-plant-specific procedures have been detailed else-where. To diagnose rejection, we routinely performCT -guided percutaneous biopsy; if technically notpossible (eg, overlying bowel), we alternatively per-form transcystoscopic or laparoscopic biopsy (4, 7).Of 75 recipients who underwent percutaneous bi-opsy, only 1 needed surgical exploration for bleed-ing (37); of the 83 recipients who underwent tran-scystoscopic or laparoscopic biopsy, none neededexploration. For graft biopsies, surgical complica-tions after bladder-to-enteric convergían have beenrare (38).

Conclusions

graft placement for uremic type I diabetic patientsif a later pancreas transplant is contemplated.Such placement has become routine at our insti-tution and allows for safer right-sided placementof the subsequent pancreas graft.

Surprisingly, vascular graft thrombosis did nothave an impact on fue patient survival rate at 1 and3 years, irrespective of fue recipient category. Thisfinding can be explained by the facts that almost90% of graft thromboses occurred during the ini-tial hospitalization, they were rapidly diagnosed,and graft pancreatectomy was expeditiously done.

The third most common indication for relapa-rotomy was an anastomotic or duodenal stumpleak. Such a leak occurred in 10% of all bladder-draiped, whole-organ, cadaver donar pancreastransplants (33). This rate is in line with a re-ported incidence of 7-14% at other pancreastransplant centers. Several factors might accountfor early leaks: 1. the donar duodenum itself, ow-ing to reperfusion edema, ischemic damage, ormicrobial colonization; 2. impaired wound heal-ing, owing to immunosuppressive therapy or fueunderlying disease; or 3. technical complicationsat the sites owing to surgical manipulation. Lateduo den al leaks, in contrast, are less likely to becaused by technical complications, but rather ow-ing to infection (eg, cytomegalovirus), rejection,or ischemia.

For early and late duodenal leaks, surgical re-pair was frequently required. In our series, only29% of recipients with a leak (all small) weretreated conservatively. Most of the leaks were over-sewn, and we rarely converted from bladder toenteric drainage at the time ofleak repair. But fueWisconsin group reported an enteric conversionrate of up to 75% for recipients operated on foranastomotic or duodenal leaks (34). The differ-ence can be explained by our large population ofsolitary pancreas transplant recipients; entericconvergían would make it impossible to monitortheir urinary amylase levels for rejection. We per-form enteric conversion for duodenalleaks only ifthey are toa large to be oversewn or if there areadditional complications related to fue bladderdrainage (eg, intractable metabolic acidosis, dys-uria, or urethritis).

Only 9% of recipients with duodenalleaks losttheir grafts-a much lower rate than for recipientswith intra-abdominal infection or vascular graftthrombosis. Patient survival rates at 1 and 3 yearsafter transplant were no different for recipients

Donors must be appropriately selected with regardto age and cause of death. Older donor age was arisk factor for intra-abdominal infection, vascular

142 ] AM COLL SURG AUGUST 1997 VOLUME 185:128-144

graft thrombosis, anastomotic or duodenal leaks,and relaparotomy. Donors over 45 years of ageshould not be routinely used, particularly if theircause of death is cardiocerebrovascular disease(39). Accepting older donors to improve HIA-matching cannot be recommended; the tradeoff isa higher graft loss rate and, even more important,a higher mortality rateo

Second, recipients over 45 years of age andthose with a history of coronary disease (myocar-dial infarction, coronary bypass, or angioplasty)should be considered for a kidney transplantalone, rather than an SPK transplant. Previously(9, 39), and again in this analysis, we showed thatolder recipient age not only has a negative impacton technical failure, but algo on patient and graftsurvival rates. Recipients with preexisting cardiacdisease have a significantly higher risk of dying fromcardiocerebrovascular complications (with a func-tioning graft) than recipients without cardiac riskfactors. Patients with advanced diabetic nephropathyshould undergo their SPK transplant before dial-ysis is initiated (ie, preemptively). This policy min-imizes dialysis-related and dialysis access-relatedcomplications, as well as intra-abdominal compli-cations and their disastrous consequences. Recip-ients who were not on dialysis before the trans-plant were not at higher risk for graft thrombosis.

Third, meticulous surgical technique duringgraft procurement, preparation, andimplantationis paramount. During the donor operation, un-necessary mechanical trauma and excessive perfu-sion pressures and flush volumes must be avoided.For reconstruction of the arterial paqcreas inflow,the Y-graft technique should be used. Intraperito-neal placement of the pancreas on the right sirleand the kidney on the left (SPK) is now standard.Diabetic kidney recipients who are candidates fora subsequent pancreas transplant (PAK) shouldundergo a left-sided kidney transplant. During thepan creas recipient operation, minimal mechani-cal manipulation of the graft will de crease thelikelihood of infection, pancreatitis, and bleeding.The duodenocystostomy and the closure of theduodenal stumps must be done meticulously, with-out any technical shortcuts or compromises, tominimize leaks and their adverse consequences.

Fourth, for pancreas graft-related complicationsrequiring reláparotomy, the focus must switch,.from graft salvage to life preservation. Graft pan-createctomy should be done when thrombosis oc-curs or when clinically symptomatic intra-abdominalinfection with or without graft pancreatitis develops.If a leak or bleeding requires relaparotomy, surgical

coITection should only be considered if dIe site caneasily be identified at fue time of relaparotomy, if itcan be safely repaired, and if there are no otherconCUITent complications such as infection. If surgi-cal coITection of a leak or bleeding is attempted, fuerisk of a relaparotomy must be weighed against fueincreased risk of graft loss and death, particularly forSPK and PAK recipients. In contrast to graft-relatedcomplications, other intra-abdominal complicationssuch as acute cholecystitis can be surgically managedwithout increased risk of graft loss and death.

Fifth, more efIective strategies to avert intra-abdominal infections need to be developed. Wepreviously showed that prophylactic antifungaltherapy with intravenous fluconazole can reducethe incidence of intra-abdominal fungal infections(8) associated with a significantly higher morbidityand mortality than sale bacterial infections. We,are cuITently investigating fue role of continuousirrigation and suction systems for recipientS withintra-abdominal infection and graft pancreatitis. Ifinfection can be eradicated by such treatment,graft pancreatectomy could be avoided and mor-tality reduced. Our efIü;rts continue to focus onstrategies that ultimately improve prophylaxis andtreatment of intra-abdominal infection, whichayer fue years has remained the most commoncomplication after pan creas transplants.

Sixth, surgical complications requiring relapa-rotomy cover a broad spectrum of procedures (in-volving the vascular, hepatobiliary, gastric, intesti-nal, and urinary tract systems and fue graftsthemselves). Thus, pancreas transplant surgeonsmust be prepared to handle adequately surgicalcomplications in diverse areas. Timely diagnosisand knowledge of potential complications, alongwith expeditious treatment and relaparotomy, arefue cornerstones of successful management of sur-

gical complications.By applying these principIes, pancreas graft sur-

vival rates should exceed 75% at 1 year posttrans-planto Future research mar reduce fue incidenceand consequences of surgical complications, sothat pancreas transplantation can be more widelyapplied to nonuremic diabetic patients and canbecome an option for uremic high-risk type I

...insulin-dependent patients who are currentlynot considered for this procedure. Further-more, more efficient yet less toxic immunosup-pressants might decrease not only the rate ofimmunologic graft failure but algo the incidenceof surgical complications. It remains to be seen

Gruessner et al RISK OF PANCREAS TRANSPlANTAllON 143

if the two new immunosuppressants currentlyused for pancreas transplantation, tacrolimusand mycophenolate mofetil (40, 41), will ulti-mately help decrease the risk of surgicalcomplications.

Acknowledgments

The authors would lile to thank Barbara Mc-Donald for the preparation of fue manuscript andMary Knatterud for editorial assistance.

References

pancreas-kidney transplantation. Incidence, risk factors andimpact on long-term outcome. Transplantation 1996;61:1323-30.

15. Douzdjian V, Abecassis MM, Cooper JL, et al. Incidence,management and significance of surgical complications afterpancreatic transplantation. Surg Gynecol Obstet 1993;177:451-6.

16. Ozaki CF, Stratta RJ, Taylor RJ, et al. Surgital complicationsin solitary pan creas and combined pancreas-kidney tran~plantation. AmJ Surg 1992;164:546-51.

17. EckhoffDE, and Sollinger HW. Surgical complications aftersimultaneous pancreas-kidney transplant with bladderdrainage. In: Terasaki PI, and CeckaJM, eds. Clinical Tran~plants 1993. Los Angeles: UCLA Tissue Typing Laboratory;1993:185-191.

18. Gruessner AC, Troppmann C, Sutherland DER, andGruessner RWG. Donor and recipient risk factors signifi-cantly affect cost of pancreas transplants. Transplant Proc1997;656-57.

19. Troppmann C, Gruessner AC, Sutherland DER, et al. Surgi-cal complications requiring early relaparotomy after pan-creas transplantation: a multivariate risk factor and economicimpact analysis. Ann Surg. (In Press).

20. Sollinger HW, Ploeg RJ, Eckhoff DE, et al. Two hundredconsecutive simultaneous pancreas-kidney transplants withbladder drainage. Surgery 1993;114:736-44.

21. Hesse Uj, Sutherland DER, Najarian JS, and Simmons RL.Intra-abdominal infections in pancreas transplant recipients.Ann Surg 1986;203:153-62.

22. PapaloisBE, Troppmann C, Gruessner AC, Benedetti E, et al.Chronic pretransplant peritoneal dialysis and intra-abdominal infection after simultaneous pancreas-kidneytransplants. Arch Surg 1996;131:761-6.

23. Troppmann C, Gruessner AC, Benedetti E, et al. Positiveduodenal segment cultures are not associated with increasedsurgical complications after whole organ, bladder-drainedpancreas transplantation in three recipient categories. Tran~plant Proc 1995;27:3101-3.

24. Benedetti E, Troppmann C, Gruessner AC, et al. Pancreasgraft loss from intra-abdominal infection: a risk factor for asubsequent pancreas transplant. Arch Surg 1996;131:1054-60.

25. Blenkharn JI, and Blumgart LH. Infectious compJicationsfollowing pancreatic surgery. Dig Surg 1985;2:65-9.

26. Wright FH, SmithJL, and Corry RJ. Postoperative complica-tions of pancreas transplan~tion. Diabetes 1989;39:236-7.

27. DiCarlo V, Staudacher C, Cristallo M, et al. Pancreas andkidney transplantation: fue San Raffaele Hospital (Milan,ltaly) experience. Diabetologia 1991;34:S11-513.

28. Martin X, Lefrancois N, MarechalJM, et al. Pancreas tran~plantation in Lyon: overall results. Diabetologia 1991;34:S8-S10.

29. BynonJS, Stratta RJ, Taylor RJ, et al. Vascular reconstructionin 105 consecutive pan creas transplants. Transplant Proc1993;25:3288-9.

30. Sollinger H. Pancreatic transplantation and vascular grartthrombosis. (Editorial).J Am Coll Surg 1996;182:362-3.

31. Lippert H, Pabst R, and Pabst H, eds. Arterial Variations inMan. Munich, Germany:J.F. Bergmann Verlag; 1985.

32. Tyden G, CalissendorffB, Billing H, et al. The vascular supplyin human pancreaticoduodenal grafts: an angiographicstudy. Transplant Proc 1990;22:593.

33. Hakim N, Gruessner AC, Papalois BE, et al. Duodenal com-plications in bladder-drained pancreas transplants. Surgery1997;618-24.

34. Sollinger HW, Messing EM, Eckhoff DE, et al. Urologicalcomplications in 210 consecutive simultaneous pancreas-kidney transplants with bladder drainage. Ann Surg 1993;218:561-70.

l. Gruessner AC. Results of the International Pancreas Trans-plant Registry (IPTR). Personal communication (1996).

2. Gruessner AC, and Sutherland DER. Pancreas transplantresults in the United Network for Organ Sharing (UNOS)United States of America (USA) registry compared with non-USA data in the International Registry. In: Terasaki PI andCeckajM, eds. Clinical Transplants 1994. Los Angeles: UCLATissue Typing Laboratory; 1994:47-68.

3. Sollinger HW, Cook H, Kamps D, et al. Clinical and experi-mental experience with pancreaticocystostomy for exocrinepancreatic drainage in pancreas transplantation. TransplProc 1989;21:2837-8.

4. Gruessner RWG, and Sutherland DER. Clinical diagnosis inpancreatic allograft rejection. In: Solez K, Racusen LC, Bill-ingham ME, eds. Solid organ transplant rejection: mecha-nisms, pathology, and diagnosis. New York: Marcel Dekker,Inc.; 1996:455-89.

5. Gruessner RWG, Burke GW, Stratta R, et al. A multicenteranalysis of the first experience with FK506 for induction andrescue therapy after pancreas transplantation. Transplanta-tion 1996;61:261-73.

6. Benedetti E, Najarian jS, Gruessner A, et al. Correlationbetween cystoscopic biopsy results and hypoamylasuria inbladder-drained pancreas transplants. Surgery 1995; 118:864-72.

7. West M, and Gruessner RWG. Laparoscopic biopsy after pan-creaticoduodenal transplantation: a first reporto Transplan-tation 1996;62:1684-7.

8. Benedetti E, Gruessner AC, Troppmann C, et al. Intra-abdominal fungal infections after pancreas transplantation:incidente, treatment and outcome.j Am Coll Surg 1996;183:307-16.

9. Gruessner RWG, Dunn DL, Gruessner AC, et al. Recipientrisk factors have an impact on technical failure and patientand graft survival rates in bladder-drained pancreas trans-plants. Transplantation 1994;57:1598-1606.

10. Gruessner RWG, and Sutherland DER. Pancreas transplanta-tion: Part 1: The donor operation. Surg Rounds 1994;17:311-24.

11. Troppmann C, Gruessner AC, Benedetti E, et al. Vasculargraft thrombosis after pancreatic transplantation: univariateand multivariate operative and nonoperative risk factor anal-ysis.j Am Coll Surg 1996;182:285-316.

12. Gruessner RWG, and Sutherland DER. Pancreas transplanta-tion: Part 11: The ~ecipient operation. Surg Rounds 1994;17:383-91.

13. Pescovitz MD, Dunn DL, and Sutherland DER. Use of thecircular stapler in construction of the duodenocystosto~y fordrainage into the bladder in transplants involving the wholepancreas. Surg Gynecol Obstet 1989;169:169-71.

14. Troppmann C, Gruessner AC, Papalois BE, et al. Delayedendocrine pancreas graft function after simultaneous

144 J AM COLL SURG AUGUST 1997 VOLUME 185:128-144

indications, risk factors and outcome. Transplant Proc 1993;25:1179-81.

39. Gruessner RWG, Troppmann C, Barrou B, et al. Assessmentof donor and recipient risk factors on pancreas transplantoutcome. Transplant Proc 1994;26:437-8.

40. Gruessner RWG, 5utherland DER, Drangstveit MB, et al. Useof FK 506 in pancreas transplantation. Transplant Int 1996;9:5251-5257.

41. Gruessner RWG, Burke GW, 5tratta R, et al. A multicenteranalysis of the first experience with FK 506 for induction andrescue therapy after pancreas transplantation. Transplanta-tion 1996;61:261-73.

35. Douzdjian V, Abecassis MM, Cooper JL, et al. Incidence,management and significance of surgical complications afterpancreatic transplantation. Surg Gynecol Obstet 1993;177:451-6.

36. Everett JE, Wahoff DC, Statz C, et al. Characterization andimpact on wound infection after pancreas transplantation.Arch Surg 1994;129:1310-7.

37. Aideyan OA, Schmidt A, Trenkner SW, et al. CT-guidedpercutaneous pancreas transplant biopsy. Radiology 1996;201:825-8.

38. Gruessner RWG, Stephanian E, Dunn DL, et al. Cystoentericconversion afterwhole pancreaticoduodenal transplantation: