Randomized Trial of Emergency Mesocaval and Portacaval Shunts for Bleeding Esophageal Varices Ronald A. Malt, MD, Boston, Massachusetts William M. Abbott, MD, Boston, Massachusetts Andrew L. Warshaw, MD, Boston, Massachusetts Thomas J. Vander Salm, MD, Worcester, Massachusetts William L. Smead, MD, Boston, Massachusetts
Shunts for the emergency control of bleeding esophageal varices must decompress the collateral portal venous system with speed and certainty. In practice the options capable of meeting this goal are overwhelmingly the end-to-side portacaval shunt and the mesocaval shunt with an interposition H graft; shunts using peripheral vessels in the splanchnic system are unreliable. Because the mesocaval shunt has been advocated as being easier to do [I-101 and preserving hepatic function in cirrhotic patients better than the portacaval shunt [II], we compared the utility of the two shunts. Material and Methods
From July 1973 to May 1977,twenty-four patients were entered into a randomized trial. Criteria for entry were the presence of exsanguinating hemorrhage from esophageal varices either (1) controlled by balloon tamponade or parenteral vasopressin or (2) uncontrollable by these means, followed by the clinical decision that the patient’s interest would best be served by prompt decompression of the portal system. In all but one instance the varices resulted from Laennec’s cirrhosis; the noncirrhotic patient had occlusion of the portal vein by a small carcinoma of the pancreas. Endoscopic examination in each patient showed esophageal varices. Clot was present on the varices in half the patients, and varices were present without gastritis or gastroduodenal ulceration in the other half. The absence of arterial sources of hemorrhage was confirmed by angiography in fifteen patients. No patient with evidence of bleeding gastritis, ulcer, or Mallory-Weiss syndrome was From the Surgical Services, Massachusetts General Hospital, and the Da Harvard Medical School, Boston, Massachusetts. Reprint requests should be addressed to Ronald A. Malt, MD, Massachusetts General Hospital, Boston, Massachusetts 02114. Presented at the Fifty-EighthAnnual Meeting of the New England Surgical Society,Portsmouth, New Hampshire, September 30-October 2, 1977.
admitted to the study. All were service patients. After a decision was made that urgent portal decompression was required and a patient met criteria for admission to the study, a sealed notice dictating the choice of operation was opened as the patient was readied for the operating room. Twelve patients were assigned to each operation, but because a low and immobile duodenum prevented a mesocaval shunt in one patient, there were actually eleven mesocaval shunts and thirteen portacaval shunts. None of the patients with mesocaval shunts and three of those with portacaval shunts were included in the results ofshunting at this hospital from 1966 through 1973
WI. The technics of shunting were standard [11,13]. The mesocaval %Igrafts were mostly 19 mm diameter standard weight knitted [email protected]
prostheses. One of us or a designate was present at all operations. Patency of the shunts was proved in all five autopsy examinations after portacaval shunting and in four autopsy examinations after mesocaval shunting. One mesocaval shunt thrombosed in the patient with pancreatic carcinoma. Laboratory data recorded were the last obtained before operation. Anesthesia times were recorded from the moment the anesthetist began to care for the patient to the time the patient arrived in the recovery room. Operation times were about half the anesthesia times but were not separately recorded. Deaths during the same hospitalization or within four weeks were scored as postoperative deaths. Significance was assessed by cross-tabulating sets of variables. The usual x2 test was replaced by Fisher’s exact test foi small-sample 2 -X 2 tables  and by computing Goodman-Kruskal’s y, a measure of correlation, when sensitivity, to directional trends was desired . All p values were two-tailed except as indicated. The small size of the sample, a result of the scarcity of patients meeting criteria for admission to this study, biases against p values reaching 0.05 and suggests relaxing the criterion of statistical significance. Exact p values are reported to permit judgments of practical significance.
The American Journal of Surgery
Emergency Shunts for Bleeding Esophageal Varices
The distribution by sex and age in the two groups is shown in Table I. Six of thirteen patients (46 per cent) died after portacaval shunts and eight of eleven patients (73 per cent) died after mesocaval shunts. This sample was representativeof all patients with emergency shunts treated here since 1966.Predictors of postoperative death were examined both in this group of randomized patients from 1973 to 1977 (n = 24) and in a larger group derived by adding results of thirty-two nonduplicated emergency shunts from the 1966-1973 study (n = 56) . Fatality rateswere similar: 53 per cent in 1966-1973 and 58 per cent in 1973-1977. The two best predictors of death were the serum albumin concentrations (combined group, p < 0.001; present study, p = 0.03) and the duration of anesthesia (combined group, p = 0.03; present study, p = 0.03). (Tables II and III.) Unlike the results of the previous study , which included patients with elective shunts, the presence of ascites did not correlate with the fatality rate (combined group, p > 0.50; present study, p > 0.05), and serum bilirubin concentration was a less strong predictor (combined group, P = 0.09; present study, p = 0.37). Better correlates were globulin levels (combined groups, p = 0.09; present study, p = 0.09) and Child’s score  (combined group, p = 0.10; present study, p = 0.12). Besides being representative of all patients undergoing emergency shunts, characteristics of the patients randomized for emergency portacaval or mesocaval shunts were in themselvesequivalent.The mortality data from the two groups may, therefore, be compared. The sex and age distributions were the same. (Table I.) The preoperative condition was equal, whether assessedby Child’s criteria  or the Massachusetts General Hospital scale . (Table IV.) Moreover, the groups were equivalent with respect to serum albumin concentration (p > 0.50), the best predictor of death.
Age and Sex Dlstrlbutlon Portacaval
Years 30-39 40-49 50-59 80-89 70-79
8% 27% 31% 27% 38% 18% 15% 9% 8% 18% X * = 2.9, p > 0.5.
Sex Male Female
31% 36% 69 % 64% Fisher’s exact test: p > 0.5.
Correlatlons of Albumln Level with Survival Albumin (gm/dl) 3.1-3.5
23.6 Lived Died Total
11(79%) 3(21%) 14(100%) n=54
8 (47%) 6 (26%) 17(74%) 9(53%) 17(100%) 23(100%) y=O.63 p 0.5
2 3 4 5 8
23% 31% 15% 31% 0 y = 0.32
9% 18% 36% 27% 9% p = 0.24
Malt et al
the mesocaval shunt was a better operation, importance must be attached to this p value. Fatality rates correlated with the length of anesthesia. The rate was 25 per cent for anesthetic times less than 5 hours and 75 per cent for longer operations (x2 = 5.5, p = 0.02). The slightly longer median anesthetic times for mesocaval shunts (5.9 versus 5.1 hours, p = 0.11) is probably attributable not to the operations themselves but to performing necessary ancillary procedures such as cholecystectomy during mesocaval shunts and ligating persistent bleeding gastric varices.
The results favor portacaval shunts over mesocaval shunts for the emergency control of bleeding esophageal varices in the spectrum of patients operated on in an urban hospital. The p value of 0.18 for a lower operative mortality rate after portacaval shunts (46 versus 73 per cent) in this small randomized study suggests the possibility of lower p values with larger samples. Indeed, if the true mortality rate for all emergency portacaval shunts were 51 per cent as from 1966 to 1977, the postoperative mortality rate after mesocaval shunting in the present study was significantly worse (p = 0.05, one tail). In any event, the mesocaval shunt was in no way superior. Given equal options, we now prefer the emergency end-to-side portacaval anastomosis as being faster, easier, and less demanding to those trained in its use. This shunt avoids use of a prosthesis or autograft and may have functional advantages over side-to-side shunts. Although emergency shunts add to immediate risk of death  and therefore should be avoided in favor of elective shunts, if possible, the survival rate five years or longer after emergency shunting is the same as after elective shunts [12,16]. If mesocaval shunts have special virtue, it is the ease with which they may sometimes be constructed rather than because they maintain hepatic function. Evidence that a mesocaval shunt preserves hepatic function is arguable . If the mesocaval shunt is functionally like a side-to-side portacaval shunt, the survival rate should actually be less good, as it was in a randomized comparison of end-to-side and sideto-side shunts . Even if it were not like direct side-to-side shunts, the hypothetical lines of flow [II] would supply the liver with blood from the spleen rather than from the splanchnic bed. Worse than selective exclusion of splanchnic blood from the liver, these same hemodynamics may be responsible for mesocaval shunts decompressing portal hypertension
in the splanchnic system, but not in the esophageal varices  or, in our experience, gastric varices. In addition, the mesocaval shunt is scarcely immune from technical problems. Angulation and occlusion of the superior mesenteric vein and impingement upon the duodenum are recognized hazards. Aberrant anatomy probably makes the operation impossible in 6 per cent of patients and more than usually difficult in another 28 per cent . Unrecognized pyelophlebitis may have obliterated the vein. For these reasons, among others, occlusion is reported in 4 to 12 per cent of mesocaval shunts by experienced and enthusiastic surgeons [I-11,131. The world-wide rate of thrombosis is said to be 32 per cent 1161,and higher failure rates have been recorded 1211. In contrast, occlusion has occurred in only 1 of 100 &d-to-side portacaval shunts at our hospital. Nonetheless, we consider an emergency or elective mesocaval shunt superior to a portacaval shunt in specific instances. These include the problem of a patient with a portal vein that cannot be joined to the vena cava without interfering with smooth lines of hemodynamic flow, with an occluded portal vein but an open mesenteric system, with obliterative infrahepatic fibrosis, with a large caudate lobe, or with massive obesity. The mesocaval shunt may also be advantageous in relieving the Budd-Chiari syndrome and in minimizing postoperative respiratory handicap in the patient with severe pulmonary disease. Surgeons who have not received training in traditional shunting operations will find it helpful if they are forced to decompress the portal system. We recognize, too, that many experienced surgeons prefer the mesocaval shunt to the portacaval shunt and achieve excellent results [1-11,131. Their success does not make the mesocaval shunt a panacea. Although we rarely see patients referred with acute failures of portacaval shunts, the situation is not the same after mesocaval shunts. Summary
A randomized trial of emergency portacaval or mesocaval shunting was conducted in twenty-four cirrhotic patients bleeding from esophageal varices. The group of eleven patients having mesocaval shunts was comparable to the group having portacaval shunts in age distribution, sex, and preoperative physical condition. Postoperative fatality rates were 46 per cent after portacaval shunting and 73 per cent after mesocaval shunting. Results favor portacaval shunts for the emergency control of bleeding varices in the spectrum of patients seen in an urban hospital.
The American Journal ol Surgery
Emergency Shunts for Bleeding Esophageal Varices
References 1. Lord JW Jr: Mesocaval shunts: techniques and results. Vascular Surgery (Rutherford FIB, ed). Philadelphia, WB Saunders, 1977. 2. Gleidman ML: The mesocaval shunt for portal hypertension. Am J Gastroenterol58: 323, 197 1. 3. Stlpa A, Thau A, Cavallaro A, et al: A technique for mesentericocaval shunt. Surg Gynecol Obstet 137: 285, 1973. 4. Giles GR, Brennan TG, Losowsky MS: Interposition teflon mesenteric caval shunt for bleeding oesophageal varlces. Br J Surg 80: 849, 1973. 5. Rosenberg N, Konigsberg SF, Noronha J: Mesocaval and portacaval shunts using panelled bovine arterial H grafts. Arch Surg 109: 754, 1974. 8. Smith M, Tuft TJ, Davidson AR, et al: Mesentericocaval “jump” graft in managemeM of portal hypertension: experience with 24 cases. Br Med J 3: 705.1974. 7. Thompson BW, Read RC, Casali RE: Interposition grafting for portal hypertension. Am J Surg 130: 733, 1975. 8. Huguet C, Benhamida F, Levy VG: Les anastomoses n&sentericocaves par prosthese de dacron interposee. Nouv Presse Med4: 1481, 1975. 9. Nay HR, Fitzpatrick H: Mesocaval “H” graft using autogenous vein graft. Ann Surg 183: 114, 1978. 10. Filtzer HS, Rossi R, Wolfort FG: Experience with interposition mesocaval shunt for management of variceal bleeding. An+ Surg 112: 593, 1977. 11. Drapanas T, LoCicero J Ill, Dowling JB: Hemodynamics of the interposition mesocaval shunt. Ann Surg 181: 523, 1975. 12. Malt RA, Sczczerban J, Malt RB: Risks in therapeutic portacaval and splenorenal shunts. Ann Surg 184: 279, 1978. 13. Drapanas T: Interposition mesocaval shunt for treatment of portal hypertension. Ann Surg 178: 435, 1972. 14. Blakrck l-M Jr: Social Statistics, 2nd ed. New York, f&Graw-Hill, 1972. 15. Child CG Ill, Turcotte JG: Surgery and portal hypertension. The Liver and Portal Hypertension (Child CG, ed). Philadelphia. WB Saunders, 1984. 18. Drloff MJ, Duguay LR, Kosta LD: Criteria for selection of patients for emergency portacaval shunts. Am J Surg 134: 148, 1977. 17. Galambos JT, Warren WD, Rudman D, et al: Selective and total shunts in the treatment of bleeding varices. N Eng/ J Med 295: 1089, 1978. 18. Resnick RH, lber FL, lshihara AM, et al: A controlled study of the therapeutic portacaval shunt. Gastroenterology 87: 843, 1974. 19. Witte CL, Dvitt TW. wile MR. et al: Left sided segmental portal hypertension following mesocaval interposition shunt. Surg Gynecol Obstet 145: 189, 1977. 20. Holyoke EA, Davis WC, Harry RD: Surgical anatomy of the mesocaval shunt. Surgery 78: 528, 1975. 2 1. Craven JI, Webster DJT, Smith PM: Prosthetic mesocavalshunts in portal hypertension (letter). Lancet 2: 370, 1978.
Discussion James M. Foster (Hartford, CT): Although beautifully presented by Dr. Malt, this paper left me with considerable concern about the validity of the study and the conclusions drawn. Questions can be raised because one patient had portal vein occlusion by pancreatic cancer, because two thirds of the patients were female (in contrast to most other reviews of portal hypertension), because the individual indications for emergency operation varied from patient to patient, and most important, because the numbers in each group were so small. Did the fourteen patients die
veklme 135, April 1978
because of technical failure to control hemorrhage, because of inadequate hepatocellular reserve, or because the hemodynamic results of shunting compromised an already borderline situation? This whole issue is a foggy one, and those that champion one operation over another often fail to consider the ever-changing and continuing effects of scarring, regeneration, and arteriovenous fistulization in progressive liver disease. However, I share Dr. Malt’s prejudice that, if there is any place for emergency portosystemic shunting, then the easiest one technically is the one to choose (usually the end-to-side portacaval). We must await more evidence to either discard or accept the emergency mesocaval technic, although the present data should help to cool the ardor of the enthusiast. Perhaps a more important question than which shunt to do is whether to perform shunt in patients in Child’s C classification-as an emergency or electively. On a costeffective basis, on a statistical basis, and on a philosophical basis, I am closer to answering no to that question than I have been in the past, and the authors’ data reinforce that position. Gerald 0. Strauch (Stamford, CT): I was especially happy to have the opportunity to read Dr. Malt’s manuscript because it certainly agrees with my biases. Enthusiastic reports of experience with interposition mesocaval shunts several years ago led a number of us in Connecticut to become interested in this shunt, and we recently reviewed our experience with twenty-eight patients with regard to immediate results. We found that eleven of those twenty-eight patients died, for an overall mortality rate of 39 per cent. Among Child’s class B and C patients, death occurred in ten of twenty-three patients (43 per cent). Among nonelective shunts, eleven of eighteen patients (61 per cent) died, which agrees pretty closely with the Massachusetts General series presented today. Among a number of other adverse factors that we looked at, only massive transfusion seemed to relate directly to mortality; and, of course, massive transfusion relates importantly to emergency shunts. One technical problem with this shunt that deserves mention is the high anastomosis of the tributaries of the superior mesenteric vein, an anatomic variant occurring in about 30 per cent of patients. In these instances one has an extremely short segment of the main trunk of the vein to deal with at the operating table, requiring use of one of the tributaries for part of the anastomosis to the graft if the recommended large caliber of graft is to be inserted. We found that because of this there was a tendency for surgeons to fudge on the caliber of the graft that they used. This, of course, is against the rules and may have been responsible for one of two thrombosed grafts in the series. In general, we concur that in nonelective situations the interposition mesocaval shunt does not offer much advantage over other shunts that are available today. W. McDermott (Boston, MA): In a brief review of our own experience fifteen years ago, we randomized a group
Malt et al
of approximately 100 unselected patients. Regardless of their condition, stage of liver disease, or the actual feasibility of surgery, they were assigned to either medical management or to one of two surgical groups-ligation of varices or emergency shunt, Whether or not surgery was actually carried out did not affect the ultimate analysis. The overall mortality in the three groups was recorded as follows: ligation of varices 69 per cent; emergency shunt 70 per cent; and medical management 67 per cent. Obviously this left us without any valid conclusions except to emphasize the inherently lethal nature of the problem. In ruminating on this problem, I reviewed the contributions by Peter, Potter, and Murphy in an attempt to find a “law” that might fit the situation. Finding none, I was forced to construct a new principle: “When the number of critical variables exceeds the number of patients in a series, the conclusion then becomes a hypothesis.” Willard C. Johnson (Boston, MA): In our own experience twenty-five patients were randomized into a prospective study to evaluate intravenous versus SMA-infused vasopressin. The results were recently reported (Ann Surg 186: 369,1977). What I want to talk about today is what happened to this group of patients. Before therapy, the men had bled an average of 8 units of blood, and 90 per cent were Child’s class C. The results of this study were that after thirty days 64 per cent of them were still alive. Only 16 per cent had bled to death, and all of these patients had a severe coagulopathy. Elective selective splenorenal shunting was recommended after control of the acute bleeding episode. None of those patients currently alive is encephalopathic, whereas our overall incidence of encephalopathy after splenorenal shunting is one in seven-
teen. The one year survival rate for those twenty-five patients is 48 per cent. We have a new approach to patients who have failed to respond to vasopressin therapy. We currently recommend transhepatic obliteration of the coronary vein by embolization with gel foam strips soaked in Sotradecol, or heattreated blood. This therapy has effectively halted bleeding in seven of eight patients. Ronald A. Malt (closing): I agree with everyone. The qualifications implicit in our assessment of the controversies were obvious, I hope. We shall never find enough patients who meet the rigid criteria for inclusion in this kind of study fbr n to be large enough to permit attaining a conventional standard of statistical significance. Therefore, one has to make inferences from trends rather than hoping to reach a predetermined p value. These points are explained by Peto et al in Br J Cancer 34: 585,1976. Since the points raised by the generous and perceptive discussers are valid and brook little argument, I,would answer specifically only the question not addressed in the manuscript: the issue of survival, raised by Dr. Johnson (before condensation of his remarks). At our hospital, if patients survive an emergency shunt for eighteen months, their incidence of encephalopathy, or of any other complication, is the same as that of patients who have survived an elective operation. The difference is in the survival rate between the two groups within the first eighteen postoperative months. Many more of the patients with emergency shunts die during that period. After that time, survivors are similar to those surviving elective shunts.
The American Journal ol Surgery