Introduction

Emergency treatment of bleeding esophageal varices (BEV) in patients with cirrhosis of the liver is of singular importance because of the high mortality rate surrounding the episode of acute bleeding.1–9 The most widely used emergency treatment of BEV is endoscopic sclerotherapy (EST) or endoscopic variceal ligation (EVL), with or without the addition of pharmacologic measures.10–12 When it is believed that portal decompression is needed, transjugular intrahepatic portosystemic shunt (TIPS) has become the most widely used procedure of choice despite the facts that, as we have pointed out previously, TIPS has a high rate of stenosis and occlusion, a resultant high incidence of portal-systemic encephalopathy (PSE), and limited durability. TIPS occlusion rate has been reduced by the recent introduction of the polytetrafluorethylene-coated stent, but the rates of occlusion and PSE are still much higher than the incidences of these serious complications following portacaval shunt in all of our studies.

Emergency portacaval shunt (EPCS) is rarely used today because of the belief, unsubstantiated by randomized controlled trials involving unselected patients, that EPCS causes frequent portal-systemic encephalopathy and liver failure.4,13–21 Consequently, portacaval shunt (PCS) has been relegated solely to the salvage of failed endoscopic and pharmacologic treatment. An important question is: is the regimen of EST or ligation with rescue PCS for failure to control BEV superior to EPCS? A unique opportunity to compare the regimen of EST with rescue PCS with EPCS was provided by our randomized controlled trial (RCT) of EST versus EPCS known as the San Diego Bleeding Esophageal Varices Study.

From April 8, 1988 to December 31, 2005, we conducted a RCT in 211 unselected, consecutive patients with cirrhosis and acute BEV in whom emergency and long-term EST was compared with direct EPCS, otherwise known as total shunt. The trial was a community-wide endeavor that involved patients referred from four adjacent counties to the University of California, San Diego (UCSD) Medical Center. In two recent publications, we described the study in detail and reported the outcomes first with regard to control of bleeding and survival22 and second with regard to the development of PSE.23 This report focuses on a comparison of outcomes following the regimen of EST with rescue PCS to outcomes following EPCS.

Patients and Methods

The reader is directed to our two recent publications22,23 that provide detailed descriptions of the following methods and protocols used in this RCT:

  1. 1.

    Design of study24,25

  2. 2.

    Patient eligibility

  3. 3.

    Definitions of:

    1. (a)

      Bleeding esophageal varices

    2. (b)

      Unselected patients (all comers)

    3. (c)

      Emergency EST

    4. (d)

      Long-term EST

    5. (e)

      Emergency portacaval shunt

    6. (f)

      Failure of emergency primary therapy

    7. (g)

      Failure of long-term therapy

    8. (h)

      Rescue therapy

    9. (i)

      Informed consent

  4. 4.

    Randomization

  5. 5.

    Diagnostic workup26

  6. 6.

    Quantitative Child’s classification27,28

  7. 7.

    Initial emergency therapy during workup

  8. 8.

    Endoscopic sclerotherapy

  9. 9.

    Emergency portacaval shunt29

  10. 10.

    Lifelong follow-up

  11. 11.

    Quantitation of PSE

In addition, the RCT involved the following protocols that have not been described previously.

Rescue Portacaval Shunt

Rescue PCS was performed in 50 patients as soon as possible after failure of EST was declared. Direct side-to-side PCS was done in 46 patients (92%), and direct end-to-side PCS was done in four patients (8%). Operative technique and intraoperative pressure measurements were identical to those used in EPCS.

QOL Score

Quality of life (QOL) was measured by assessing the following factors: (1) liver function as determined by quantitative Child’s risk class; (2) development of recurrent PSE; (3) number of PSE episodes; (4) units of packed red blood cell (PRBC) transfusion for upper gastrointestinal bleeding; (5) number of hospital readmissions; (6) days of hospitalization during readmission; (7) return to work, including housekeeping; (8) abstinence from alcoholism; and (9) portacaval shunt patency. These nine factors were weighted numerically so as to produce a QOL score in which the lower the score, the better the QOL.

Direct Cost of Care

All hospital and outpatient facility charges and all professional fee bills from UCSD and from referring hospitals and physicians were obtained continuously for every patient entered into the study for 10 years.

Figure 1 is a Consort flow diagram that shows the overall design and conduct of the RCT.22,23

Fig. 1
figure 1

Consort flow diagram showing the overall design and conduct of the prospective randomized controlled trial.22,24,25

Full size image

Statistical Analysis

The comparison between Emergency and Rescue PCS groups used Fisher’s exact test for binary outcomes (e.g., control of bleeding, incidence of recurrent PSE) and Wilcoxon rank-sum test (WRT) for continuous outcomes (e.g., units of PRBC transfusion, number of recurrent PSE episodes, number of hospital readmissions). The length of survival was compared using Gehan–Wilcoxon rank test. The change in Child’s class was compared for each time interval using the exact WRT, adjusted for ties. The average change in Child’s class during the first 5 years was computed by averaging the duration of time in years spent by the patients at risk (alive) in each category (improved, unchanged, or worse). The comparison of the cause of recurrent PSE episodes used Pearson’s chi-squared test. The overall quality of life score was computed for each group and each year by adding up the scores of the nine components. This score was compared between the two groups assuming a Poisson (log-linear) model, with different means for the different categories, and a constant group effect. At the beginning of the study, it was decided in advance not to perform an interim analysis of the data.

Results

EPCS Versus EST—Outcome Data

Our recent publications should be consulted for detailed data on the clinical characteristics of the 211 patients, findings on upper endoscopy and liver biopsy, results of laboratory blood tests, data on rapidity of therapy, data on control of bleeding, operative and endoscopic data, data on PSE, and data on survival.22,23 There were no significant differences in the clinical characteristics of the two groups on entry in the RCT. Cirrhosis was demonstrated by liver biopsy in all patients. Definitive treatment was initiated in <24 h after onset of bleeding in all patients. EPCS controlled bleeding permanently in all patients, while EST achieved permanent control of bleeding in only 20%. Survival rates were significantly higher after EPCS than after EST at all time intervals and in all Child’s classes (p < 0.001). Patients with the most severe liver disease in Child’s risk class C realized substantial long-term survival after EPCS.

The incidence of recurrent PSE following EST was 35%, which was more than twice the 15% incidence following EPCS (p < 0.001). EST patients had a total of 179 episodes of PSE and 146 PSE-related hospital admissions compared with EPCS patients who had 94 episodes of PSE and 87 hospital admissions (p = 0.003). Recurrent UGI bleeding, which was rare in the EPCS group, was a major causative factor of PSE in the EST patients.

EST with Rescue PCS Versus EPCS

Patient Characteristics

Table 1 summarizes the clinical characteristics at the time of entry in the San Diego BEV study of the 105 patients who were randomized to EPCS and the 50 patients who failed EST and underwent rescue PCS. There were no significant differences between the two groups in any important characteristics of cirrhosis and BEV. Thirty-one patients failed EST but did not undergo rescue PCS for various reasons, most prominent of which were death from recurrent BEV at home or at a distant hospital and death from massive recurrence of BEV before a rescue PCS could be done. As others have found, failure of patients to take advantage of rescue treatment reflects the realities of treating BEV in the cirrhotic population. Although these 31 patients were excluded from the analysis, their deaths have a negative impact on the concept of rescue PCS for failed EST.

Table 1 Clinical Characteristics at Study Entry in Patients with Cirrhosis and Bleeding Esophageal Varices Undergoing Primary EPCS or EST with Rescue PCS
Full size table

Table 2 summarizes data on rapidity of therapy and indicates clearly that all patients underwent rapid diagnosis and treatment upon entry in the RCT. Median time from onset of bleeding to the start of therapy was <24 h in both groups of patients. The time from initial contact at UCSD Medical Center to start of therapy was <8 h in every patients in the EST group and in 102 of the 105 patients in the EPCS group. Active bleeding was observed within 4 h of entry in the study in 83% of the 155 patients.

Table 2 Rapidity of Therapy of Patients with Cirrhosis and Bleeding Esophageal Varices Randomized to EPCS or EST followed by Rescue PCS
Full size table

Control of Bleeding

Table 3 provides data on control of BEV by EPCS and by EST with rescue PCS. EPCS promptly and permanently controlled bleeding in every patient. In contrast, EST failed to control bleeding in any of the 50 patients, and that is why they underwent rescue PCS. Failure of EST in 106 patients in the EST group was based on one or more of the criteria established in advance by the study protocol, which included: (1) in 15 patients, variceal bleeding continued or recurred during the first 7 days after initial EST and required ≥6 U blood transfusion; (2) in 47 patients, recurrent variceal bleeding required ≥8 U of blood transfusion during any 12-month period after the index hospitalization; (3) in 27 patients, variceal bleeding recurred after an experienced co-investigator faculty gastroenterologist had previously declared that the esophageal varices were obliterated or gone. In eight of these same patients, recurrent bleeding required ≥8 U of blood transfusion, so they met two criteria of failure.

Table 3 Control of Bleeding in Patients with Cirrhosis and Bleeding Esophageal Varices Randomized to EPCS or EST followed by Rescue PCS
Full size table

Table 3 also summarizes the requirement for PRBC transfusions. Overall, patients treated by EST with rescue PCS required almost twice the number of PRBC transfusions as patients treated by primary EPCS (p < 0.001).

Survival

Table 4 shows data on survival in the two groups of patients, and Fig. 2 shows 15-year Kaplan–Meier estimated survival plots. All patients in the EST-rescue PCS group and 98 of the 105 patients in the EPCS group were eligible for ten or more years of follow-up. The remaining seven EPCS patients had follow-up for 9.4–9.9 years. No patients were lost to follow-up. After the first year, there were highly significant differences in the survival rates of the two study groups at all long-term time intervals. The 5-, 10-, and 15-year survival rates in the EST-rescue PCS group were 22%, 16%, and 0%, respectively, and in the EPCS group were 72%, 46%, and 46%, respectively (p < 0.001). Median survival was 6.15 years in patients randomized to EPCS compared to 3.1 years in EST-rescue PCS patients (p < 0.001). Hepatic failure was the primary cause of death in 44% of patients who underwent EST with rescue PCS compared to 22% of patients who received primary EPCS. In contrast to the entire group of 106 EST patients in which 26% died from variceal bleeding, none of the 105 EPCS patients died of bleeding.

Table 4 Survival of Patients with Cirrhosis and Bleeding Esophageal Varices Randomized to EPCS or EST Followed by Rescue PCS
Full size table
Fig. 2
figure 2

Kaplan–Meier estimates of overall survival after emergency portacaval shunt (EPCS, n = 105) and after failed endoscopic sclerotherapy (EST) with rescue portacaval shunt (PCS, n = 50).

Full size image

As expected, the survival rate was related to the severity of liver disease at the time of entry in the study, as expressed by quantitative Child’s risk classes. In the EST group with rescue PCS, 5-year survival rates in Child’s classes A, B, and C were 36%, 15%, and 20%, respectively, and 10-year survival rates in Child’s classes A, B, and C were 29%, 12%, and 10%, respectively. In contrast, in the EPCS group, the corresponding survival rates in Child’s classes A, B, and C were 89%, 76%, and 53% at 5 years and 62%, 47%, and 30% at 10 years. The differences in favor of EPCS were highly significant (p = 0.005 to p < 0.001).

Median survival of patients who failed EST and underwent a rescue PCS was 3.01 years compared to median survival of 2.36 years in the 38% of patients who failed EST but did not undergo a rescue PCS. Importantly, median postoperative survival following rescue PCS was only 1.99 years compared to 6.18 years following primary EPCS (p < 0.001).

Portal Systemic Encephalopathy

Table 5 shows data on PSE in the two groups of patients. Calculations of the incidence of PSE are based on patients who were discharged from the index hospitalization and survived more than 30 days after study entry since deaths on or before 30 days were considered indeterminate and unrelated to PSE. As we have reported previously, the incidence of PSE was 35% in the primary EST group and 15% in the primary EPCS group (p = 0.001).23 The difference in incidence of PSE was even greater when the primary EPCS group with its 15% PSE incidence was compared to the EST-rescue PCS group in which the PSE incidence was 43% (p < 0.001). Furthermore, as shown in Table 5, the number of episodes of PSE per patient and per year and the number of hospital readmissions per patient and per year were all significantly more frequent in the EST-rescue PCS group than in the EPCS group (p < 0.001). Additionally, the EST-rescue PCS patients with PSE had a median survival from the time of study entry of 3.44 years, which was longer than the 2.45 years of survival of the patients free of PSE, but the difference was not significant. In contrast, the patients in the primary EPCS group had a significantly longer survival than those in the EST-rescue PCS group (p < 0.001), and their median survival was 5.18 years for those with PSE and 10.43 years in those free of PSE (p < 0.001).

Table 5 Recurrent Portal-Systemic Encephalopathy in Patients with Cirrhosis and Bleeding Esophageal Varices Randomized to EPCS or EST Followed by Rescue PCS
Full size table

Dietary indiscretion with regard to protein restriction was the most frequent cause of recurrent PSE in both groups of patients. Portal hypertension-related UGI bleeding, usually from BEV, was the main cause of PSE in 23% of the episodes of PSE in the EST-rescue PCS group and was a contributing cause in an additional 16%. PSE episodes occurred more frequently prior to performance of rescue PCS than after rescue PCS. UGI bleeding was infrequently responsible for PSE in patients randomized to EPCS, occurring in only 8% of the patients even though they survived more than twice as long as the EST-rescue PCS patients (p < 0.001).

Change in Liver Function

Improvement or worsening of liver function was determined by serial quantitative measurements of Child’s risk class monthly during the first year after study entry and every 3 months thereafter. An increase or decrease in two or more Child’s class points reflected, respectively, improvement or worsening of liver function. Table 6 presents a summary of yearly changes in Child’s risk class using Child’s class on study entry as a baseline and combining Child’s classes A, B, and C. Results in patients randomized to EPCS are compared to results in the EST-rescue PCS patients. In every year, there was a statistically significant difference between the EPCS group and the EST-rescue PCS group, with the patients randomized to EPCS having more improvement and less worsening of liver function than the patients in the EST-rescue PCS group (p = 0.008 to <0.001). Overall, the 1- to 5-year average change in Child’s classes comparing EPCS versus EST-rescue PCS, respectively, showed improvement in 59% versus 32% and worsening in 10% versus 33% (p < 0.001). The differences in liver function between the EPCS and EST-rescue PCS groups were particularly striking in Child’s class C where improvement in liver function was most important. Five years after entry in the RCT, liver function had improved in 94% of the EPCS group compared to 65% in the EST-rescue PCS group, and liver function had worsened in 4% of the EPCS group compared to 30% of the EST-rescue PCS group. The difference in favor of EPCS was significant (p < 0.001).

Table 6 Changes in Child’s Class Compared to Child’s Class on Study Entry in Patients with Cirrhosis and Bleeding Esophageal Varices Randomized to EPCS or EST Followed by Rescue PCS
Full size table

Quality of Life Score

Table 7 summarizes data on QOL for 5 years in the 105 patients randomized to EPCS and the 50 patients who failed EST and underwent rescue PCS. QOL score was based on nine criteria shown at the bottom of Table 8. In the comparison, a lower score indicates a better QOL. Overall, during each year and for the entire 5-year period of study, QOL was significantly better, i.e., the QOL score was lower in the EPCS group than in the EST-rescue PCS group (p < 0.001).

Table 7 Quality of Life Score Based on Nine Criteria in Survivors Who Were Discharged from the Index Hospitalization After Undergoing EPCS or EST Followed by Rescue PCS (Lower Score is Better QOL)
Full size table
Table 8 Total Facility and Professional Fee Charges for Patients with Cirrhosis and Bleeding Esophageal Varices Randomized to EPCS or EST Followed by Rescue PCS
Full size table

Direct Costs of Care

Table 8 summarizes the total charges over a 10-year period for hospitalization and outpatient care in thousands of US dollars in patients randomized to EPCS and those randomized to EST-rescue PCS. The mean grand total charges over the entire length of the study were $150,400 in the EPCS patients and $263,600 in the EST-rescue PCS patients, a highly significant difference (p < 0.001). More importantly, the mean grand total charges per year amounted to $39,200 in the EPCS patients and $216,700 in the EST-rescue PCS patients, 5.5 times greater (p < 0.001).

Discussion

Comment is warranted regarding the use of EST rather than EVL in this RCT. In 1988 when the San Diego BEV Study was initiated, EST was a mainstay of therapy of BEV and the sole form of endoscopic therapy in use. When EVL was introduced generally, as well as at our institution, we were well into our RCT and our investigators and senior advisors made the unanimous decision not to change from EST to EVL. That decision has received strong support from studies published in 2003, 2005, and 2006 that have questioned replacement of EST by EVL. In a survey reported in 2003 of 93 gastroenterologists who treated 725 patients with BEV, EST was used more frequently than EVL for control of BEV and as frequently as EVL for initial control of acute bleeding.11 Trials published in 2005 and 1999 reported a significantly higher failure rate with band ligation of actively bleeding varices and an overall higher recurrence rate of varices treated by EVL.12,30 Moreover, EST has been reported to be more cost-effective if active variceal hemorrhage is present at the index endoscopy procedure, as was the case in our RCT.30 It is noteworthy that none of nine randomized clinical trials summarized in 2005 observed a statistically significant difference in survival rate between EVL and EST.12 In a meta-analysis of emergency EST in 40 trials involving 4031 patients reported by Triantos et al.10 in 2006, there was no statistically significant difference in survival rate between EVL and EST. The authors concluded that “the conclusive evidence for substituting banding ligation or the combination of vasoconstrictors with sclerotherapy as better therapeutic approaches has not been provided in randomized trials. Sclerotherapy can remain a gold standard in variceal bleeding….”

It is widely agreed that portal-systemic shunts are very effective in controlling BEV. The results of our RCT confirm such effectiveness since both EPCS and rescue PCS promptly and permanently controlled BEV in every patient. Nevertheless, according to numerous statements in the literature, surgical shunts control bleeding at the expense of an unacceptably high rate of PSE as well as progressive liver failure, and that is the main reason why portal-systemic shunts have been relegated to a secondary salvage role for use solely as a last resort when endoscopic and pharmacologic measures have failed.4,13–21 The results of our RCT, which involved unselected, consecutive cirrhotic patients with all degrees of liver dysfunction, including patients in Child’s class C, contradict the widely held beliefs about the appropriate role of portal-systemic shunts. According to our findings which have been reported in detail recently,23 the incidence of PSE following EPCS was significantly lower (15%) than the incidence following primary EST (35%) or after EST with rescue PCS (43%). The protocol of our RCT describes the requirements for achieving a low incidence of PSE.23 These are: (1) diagnosis and EPCS within 24 h of onset of BEV; (2) operation by surgeons experienced in portal hypertension surgery; (3) postoperative care in an ICU by trained and experienced nurses and physicians; and (4) regular, long-term follow-up that includes concerted efforts to promote abstinence from alcohol and repeated emphasis on reasonable restriction of dietary protein intake. It is our conviction that these requirements can be fulfilled by most trained surgeons and by most general hospitals in the USA.

Regarding the matter of post-shunt liver failure, the concept that direct portacaval shunts cause liver failure because of diversion of essential portal blood flow began over a century ago with the animal experiments of Eck and Hahn and associates in Pavlov’s laboratory31,32 and has been suggested repeatedly but not substantiated since then.33–35 The concept has led to the invention of a number of operations that are purported to maintain portal blood flow to the liver while overcoming portal hypertension. These include distal splenorenal shunt, small-diameter prosthetic, H-graft portacaval shunt, and small-diameter direct side-to-side portacaval shunt. However, the concept is contradicted by two important hemodynamic facts. The first is that whether or not a PCS is constructed, BEV arise as a consequence of progressive diversion of a substantial portion of venous blood flow away from the liver and into portal-systemic collaterals so that, with regard to creation of a PCS, the cirrhotic liver with BEV is markedly different from the normal liver. The second important hemodynamic fact is that a fundamental physiologic response to diversion of portal venous flow is a compensatory increase in hepatic arterial blood flow to the liver.36–38 It is not possible by any currently available practical method to predict the adequacy of hepatic arterial compensation prior to performance of a PCS. Substantial data from preoperative and intraoperative measurements of both pressure and blood flow in the portal vein in large numbers of patients have failed to show a correlation between any hemodynamic measurements performed prior to PCS and survival, hepatic function, or development of PSE after PCS.36–40 Our studies of portal vein hemodynamics before PCS showed no statistically significant correlation between pre-shunt maximum perfusion pressure and post-shunt survival, liver function, hepatic failure, or development of PSE.38 It is noteworthy that Burchell and colleagues in their extensive intraoperative hemodynamic studies observed the largest post-shunt increments in compensatory hepatic arterial flow following side-to-side PCS, the procedure performed in 99 of the 105 EPCS patients in our RCT. In the final analysis, the long-term improvement in liver function following EPCS observed in the current trial provides the most meaningful and objective information regarding the effect of portal venous flow diversion on the cirrhotic liver. Each year for 5 years after EPCS, liver function improved in 59–65% of patients, and liver function declined in only 8–11%.

The San Diego BEV Study provided a unique opportunity in a RCT to compare EPCS, a treatment that is infrequently used today, with a conventional treatment regimen consisting of rescue PCS following failure of EST to control BEV. Not only did EPCS prove to be superior to EST, but also, by every measure of effectiveness, EPCS proved to be significantly better than the combination of EST with rescue PCS. How can this striking difference be explained? A likely explanation is that patients who require rescue PCS are much more severely ill than patients who undergo a diagnostic workup and a definitive operation within 24 h of the onset of bleeding. They are poorer candidates for operation or, for that matter, for any other form of rescue therapy. There is little doubt that persistent variceal bleeding, repeated readmissions to the hospital, and repeated bouts of PSE in the EST patients take their toll. In point of fact, by the time rescue PCS was required, many of the patients had experienced a decline in liver function reflected by a negative change in Child’s risk class. Furthermore, one third of the patients who failed EST died before having the opportunity to undergo rescue PCS, a common occurrence in programs that treat cirrhotic patients with BEV.

Kahn et al.,5 in their extensive review of emergency treatment of BEV, identified serious shortcomings in many of the reported studies. The San Diego BEV Study was designed to overcome these shortcomings and was unique in the following respects: (1) the 211 patients with acute BEV were unselected and consecutive; (2) physicians from four California counties with a population of 8.5 million agreed to participate in the study; (3) the diagnostic workup was completed rapidly in a mean 3.1–4.4 h, entirely at the bedside in the ICU; (4) unlike any reported study to date, definitive treatment with EST or EPCS was started within 8 h of study entry in 208 of 211 patients; (5) the surgeons and gastroenterologists were experienced senior faculty physicians; (6) follow-up was 100%, was regular, and extended for 9.4 to more than 10 years or until death; (7) concerted, organized, and often successful efforts were made to control dietary protein intake and alcoholism; (8) PSE was determined and prevented according to a clearly defined protocol by a “blinded” senior gastroenterologist; and (9) consistent with our past experience following EPCS, only 2 of 105 patients developed shunt occlusion, which prevented recurrent BEV and PSE.

Conclusion

In this RCT of emergency treatment of acute BEV in 211 unselected, consecutive patients with cirrhosis of all grades of severity, EPCS was strikingly superior to EST as well as to the combination of EST and rescue PCS in regard to all outcome measures, specifically control of bleeding, survival, incidence of PSE, improvement in liver function, quality of life, and cost of care. These results contradict the widespread belief that PCS, otherwise known as total shunt, is associated with a high incidence of PSE and causes liver failure. Moreover, these results call into question the widespread practice of relegating PCS solely to salvage failure of endoscopic therapy of BEV and strongly support the use of EPCS as the first line of emergency treatment of BEV in cirrhosis.