In the absence of treatment, variceal hemorrhage recurs in approximately 70% of patients after the index bleeding episode [5, 12, 13]. This risk is greatest in the first 48 hours after initial hemostasis and gradually subsides thereafter [13-15]. The long-term course is punctuated by recurrent bleeding episodes, hepatic decompensation, and, eventually, death [13, 14]. Thus, rebleeding episodes may occur early (within a few days) or late (weeks to months) after an episode of variceal hemorrhage.

The primary objective of our randomized trial was to compare the efficacy and safety of TIPS with those of repeated endoscopic sclerotherapy for the prevention of late variceal rebleeding in clinically stable persons who survived an episode of esophageal variceal bleeding.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

Active variceal hemorrhage was defined as emesis of coffee-ground material or bright red blood with or without melena or hematochezia, along with a decrease in hemoglobin level of at least 2 g/dL caused by bleeding varices. Bleeding was considered variceal in origin if actively bleeding varices or varices with stigmata of bleeding [16, 17] were seen during endoscopy and if no other lesions were noted that could explain the bleeding. Survivors of an episode of active esophageal variceal hemorrhage were considered for inclusion if they were clinically stable and were not actively hemorrhaging (absence of hemorrhage was indicated by a stable hemoglobin level and no need for transfusions) for at least 72 hours. Before randomization, all patients were assessed clinically and by Doppler ultrasonography. We excluded patients who had portal venous thrombosis, ultrasonographically evident hepatoma, and end-stage cancer or systemic diseases that would limit a patient's life span to less than 1 year. Other exclusion criteria were failure to obtain informed consent, pregnancy, and a history of noncompliance with treatment.

Randomization

Therapy with ß-blockers was discontinued in all patients who were taking these agents before study entry. Patients who met entry criteria were randomly assigned to either TIPS or sclerotherapy. Patients were not stratified in any manner before being randomly assigned in a 1:1 ratio, which was done by using a computer-generated randomization code. Randomization was performed by selecting a sealed opaque envelope that contained the code for the study group. The study groups were balanced for every 20 randomly assigned patients. For all patients, the primary intervention was initiated within 72 hours of randomization.

Primary Interventions

Transjugular Intrahepatic Portosystemic Shunts

Transjugular intrahepatic portosystemic shunts were created with Wallstents (Schneider, Inc., Plymouth, Minnesota) using standard techniques described elsewhere [10, 18]. Special care was taken to ensure that only the central portions of the right or left branches of the portal vein were used for creation of the intrahepatic tract in order to optimize hemodynamics and minimize turbulence in the stent. After the TIPS were created, the stents were dilated with an 8-mm balloon catheter. By using methods described in previous reports [18, 19], pressures were measured in the inferior vena cava; hepatic vein; and proximal, middle, and distal stent and portal vein before and after TIPS were created. If the portosystemic gradient did not decrease to less than 12 mm Hg or the completion portogram did not demonstrate excellent flow through the stent, the stent was dilated to 10 mm and pressure was measured again. If the gradient was less than 12 mm Hg, the stent was dilated to 12 mm. Parallel stents were not used in any patient. The left gastric vein was not embolized.

Sclerotherapy

Patients assigned to sclerotherapy were treated with 2-mL intravariceal freehand injections of 5% Na morrhuate, for a total of 12 to 20 mL per session. Patients received sclerotherapy every 2 to 3 weeks until all varices were obliterated. After obliteration, surveillance endoscopy was performed every 3 months for 1 year and every 6 months thereafter. If varices recurred, sclerotherapy was again performed at 2- to 3-week intervals until varices were obliterated.

Follow-up and Data Collection

All patients were followed in the clinical research center outpatient clinic at the Medical College of Virginia by a team of hepatologists and trained nurses. Data were collected from patients during clinic visits, visits to the endoscopy-angiography suite for sclerotherapy, or hospitalizations for any reason. After TIPS were created, patients were seen every month for the first 6 months and every 3 months thereafter. Patency and function of the stents were assessed by Doppler ultrasonography on day 1, week 1, months 1 and 3, and every 3 months thereafter. In the TIPS group, all surviving patients had angiography every 6 months. Angiography was also done if variceal hemorrhage recurred or if results of one of the noninvasive tests suggested that the stent was not patent. Patients undergoing sclerotherapy were seen in the endoscopy suite until varices were obliterated. They were then seen in clinic every month until 6 months from randomization and every 3 months thereafter.

Endoscopy was performed before TIPS were created and on day 7, months 1 and 3, and every 3 months thereafter. Varices were photographed in the retroflexed position, at the gastroesophageal junction, and 2 and 5 cm above the gastroesophageal junction. These photographs were later evaluated in a blinded manner by four investigators. Varices were graded as follows: 0, absent; 1, trace (flattens on insufflation); 2, small (minor impingement on lumen after insufflation); 3, greater than 25% of luminal impingement after insufflation; and 4, lumen-occluding varices. Little variability was seen between repeated observations by the same endoscopists (Spearman correlation coefficient = 0.89; P < 0.001) or different endoscopists (rank correlation coefficients > 0.85; P < 0.001).

Secondary Interventions

Occlusion of TIPS was identified if angiography showed occlusion of the stent. Stenosis of TIPS was defined by angiographic definition of the contour of the shunt (>25% impingement of lumen). Recurrent portal hypertension was arbitrarily defined as a portosystemic gradient greater than 12 mm Hg in patients whose portosystemic gradient immediately after TIPS creation was less than 12 mm Hg. In patients whose portosystemic gradient after TIPS creation was greater than 12 mm Hg, recurrent portal hypertension was defined as a 25% increase in portal pressure. Patients with TIPS occlusion due to shunt thrombosis were treated with angiographic clot removal [20]. Recurrent portal hypertension caused by stent stenosis was managed by dilatation of the stent and placement of additional stents in a series to bridge the stenosed segment.

Management of Recurrent Variceal Hemorrhage

Recurrent hemorrhage was defined as hematemesis or melena occurring after randomization. All cases of recurrent hemorrhage were examined by endoscopy within 24 hours of hospitalization. Regardless of study group, all patients who had bleeding esophageal varices, esophageal varices with stigmata of bleeding, and gastric varices that were contiguous with esophageal varices were initially injected with 12 to 20 mL of 5% Na morrhuate.

Patients in the sclerotherapy group who rebled from varices were offered TIPS if 1) they continued to bleed actively [that is, had bright red blood in nasogastric aspirate and required transfusion of >3 units of packed red blood cells every 8 hours to maintain a hemoglobin level of approximately 9 g/dL] or 2) after they stopped bleeding. Because, in theory, one would not expect varices to be present or to rebleed in the presence of widely patent TIPS, patients in the TIPS group who rebled from varices were studied by angiography either immediately (if they continued to bleed actively despite sclerotherapy) or within 48 hours of control of active hemorrhage. Any TIPS stenoses were corrected by balloon dilatation and further stenting within the original stent. Active hemorrhage that continues despite adequate portal decompression is an indication for embolization of the left gastric vein; however, embolization was not necessary in any patient.

End Points and Data Analysis

Data were stored in a spreadsheet that was created using Lotus software (Boston, Massachusetts) and were analyzed by using software from Epistat, Inc. (Richardson, Texas). The principal end points studied were rebleeding and survival, and secondary end points were complications and rates of rehospitalization. Intention-to-treat analyses were done. Between-group differences in survival, rebleeding, and encephalopathy were assessed by the Wilcoxon and log-rank tests. Relative risks were also determined, and CIs [21, 22] were used to confirm equivalence when no significant differences between treatment groups were found. Continuous data (such as portosystemic pressure gradients and age) were compared by using a t-test, and categorical data (such as presence or absence of ascites) were compared by using a chi-square test. Means ±SDs were calculated for normally distributed continuous variables, and median values were calculated for variables that were not normally distributed. Survival curves for death, rebleeding, and encephalopathy were estimated by using Kaplan-Meier analysis [23, 24]. A fixed covariate proportional hazards model was used to assess the role of patient characteristics at study entry in predicting rebleeding or death. In addition to the interventions used, the covariates were selected on the basis of previously published data [16, 25-27]. A time-dependent analysis was also done to study the potential influence of therapy for recurrent bleeding on survival. A two-tailed P value of 0.05 was considered statistically significant.

Sample Size

Sample size estimates were calculated using software from Epistat, Inc. Repeated sclerotherapy and surgical portacaval shunts are associated with rebleeding rates of approximately 50% [28-32] and 9% to 22% [1-4], respectively. When our study began, no data existed to provide any reasonable estimates of the rebleeding rates after TIPS creation. By using data from surgical shunts, we estimated that 40 patients in each group would be required to demonstrate a decrease in rebleeding rate from 50% to 20%, with and ß values of 80%.

Results

Top Methods Results Discussion Author & Article Info References

Between June 1991 and July 1994, 132 consecutive patients with bleeding esophageal varices were seen by the hepatology service at the Medical College of Virginia. Of these, 128 were treated by urgent sclerotherapy and 4 underwent variceal band ligation. Thirty-two of the 132 patients could not be stabilized by endoscopic treatment and required definitive therapy within 72 hours; thus, these patients were excluded from our study. One hundred patients remained hemodynamically stable for more than 72 hours and met entry criteria. Twenty of these 100 patients were excluded because of portal venous thrombosis (n = 6), hepatoma (n = 3), florid alcoholic hepatitis or history of noncompliance with treatment (n = 6), and refusal to give informed consent (n = 5).

Forty-one patients were randomly assigned to the TIPS group, and 39 were assigned to the sclerotherapy group. At study entry, both groups were similar with respect to age, sex, severity and cause of liver disease, time from active hemorrhage, number of patients having an index bleeding episode, and the presence and degree of renal dysfunction (Table 1). Four patients in the TIPS group and 2 patients in the sclerotherapy group were receiving ß-blockers before study entry; therapy with these drugs was stopped at study entry.

Technical Outcome

Transjugular Intrahepatic Portosystemic Shunts

In 39 of 41 patients, TIPS procedures were successful. Technical failure to create TIPS resulted from failure to enter the portal vein in one patient and previously undetected portal venous thrombosis with cavernous transformation in the other. These patients were treated with sclerotherapy using the protocol for the sclerotherapy group. In most patients, 1 to 2 stents were used; the stents were dilated to achieve an internal diameter of 10 mm. After TIPS procedures were done, the portosystemic pressure gradient immediately decreased from 23 ± 3 mm Hg to 11 ± 2.5 mm Hg (P < 0.001) (Figure 1).

View larger version (16K): [in this window] [in a new window]   Figure 1. Long-term effects of transjugular intrahepatic portosystemic shunts (TIPS) on portosystemic pressure gradients. The shunts produced an initial highly significant decrease in the gradient (P < 0.001). However, portal hypertension recurred in most patients; this condition was amenable to balloon dilatation. Data are the mean ±SD.

Before randomization, 30 of 41 patients in the TIPS group had grade 3 or 4 esophageal varices and 9 of 41 had additional gastric varices. By day 7 after the TIPS procedure, esophageal varices had disappeared in 30 of 39 patients and had decreased to grade 1 to 2 in 6 patients. Three cases of varices did not resolve: one because of portal venous thrombosis and two because of kinking of the stent lumen; these varices were amenable to angiographic thrombectomy and stent dilatation, respectively. A spontaneous splenorenal shunt was seen in 6 of 9 patients with gastric varices. Although gastric varices without spontaneous splenorenal shunts disappeared after TIPS were created, three of six cases of varices in patients with such shunts did not resolve, despite a portosystemic gradient less than 12 mm Hg and a patent shunt (mean flow, 139 mL/min) because of preferential splenic venous flow through such collaterals.

Thirty patients underwent routine angiography at 6 months. An additional 4 patients had angiography because of rebleeding during the first 6 months. Portal pressures were significantly elevated in 20 of 34 patients because of stent stenosis (Figure 1); these patients, however, responded to balloon dilatation and deployment of additional stents to shore up the dilated shunt. Over time, portal hypertension tended to recur because of neointimal hyperplasia but remained amenable to correction by dilatation of the stent.

Sclerotherapy

In the sclerotherapy group, 28 of 39 patients had grade 3 or 4 esophageal varices. Varices were obliterated in 36 of 41 cases (2 of these patients were in the TIPS group) after a mean of 4.2 ± 1.2 sessions using an average volume of 16 mL of 5% Na morrhuate per session. At study entry, nonsclerosed gastric varices were present in the fundus of 6 patients. During follow-up, varices recurred in 10 of 36 patients. New gastric varices developed in 5 patients, and portal gastropathy either remained the same or worsened in all patients in the sclerotherapy group.

Effects on Rebleeding

Median follow-up was 956 days for the TIPS group and 990 days for the sclerotherapy group. During follow-up, 5 patients in the TIPS group and 3 patients in the sclerotherapy group underwent orthotopic liver transplantation; these patients were censored at the time of transplantation. Two patients from the TIPS group and 1 patient from the sclerotherapy group were lost to follow-up. Ten patients in the TIPS group and 10 patients in the sclerotherapy group had at least one episode of gastrointestinal bleeding (hematemesis or melena) (Figure 2) (P > 0.2; generalized Wilcoxon test). According to univariate analysis, the relative risk for rebleeding was 0.95 (95% CI, 0.44 to 2.03). The lack of significant differences was also confirmed by multivariate analysis (relative risk, 0.96 [CI, 0.74 to 2.2]); in this analysis, timing of randomization, previous use of ß-blockers, Child-Pugh class, and variceal grade were included as covariates. The groups did not differ with respect to the severity of rebleeding episodes or transfusion requirements during such episodes.

View larger version (11K): [in this window] [in a new window]   Figure 2. Kaplan-Meier analysis of the risk for rebleeding, plotted as the probability of remaining free of bleeding over time. No significant differences were seen between the two groups. TIPS = transjugular intrahepatic portosystemic shunts.

Transjugular Intrahepatic Portosystemic Shunts

In the TIPS group, one patient developed recurrent hemorrhage due to congestive gastropathy, one developed alcoholic cardiomyopathy and reflection of high systemic venous pressures into the portal bed, one had bleeding gastric varices despite patent TIPS and a patent portal vein, and one had TIPS thrombosis. The latter patient was managed by angiographic thrombectomy and urgent liver transplantation. In the other six patients, variceal rebleeding was associated with recurrent portal hypertension due to neointimal hyperplasia and stent stenosis, which was corrected by balloon dilatation and placement of additional stents in a series extending across the stenosed segment.

To further analyze the potential risk factors for recurrent hemorrhage in all patients undergoing TIPS creation regardless of the original treatment group, we used a proportional hazards model to evaluate the portosystemic gradient, percentage change in portal pressures, presence or absence of flow into the left gastric vein, flow velocity through the TIPS, Child-Pugh class, and history of alcohol abuse. Of these factors, only a history of alcohol abuse was associated with an increased risk (2.8-fold compared with other liver diseases) (P = 0.007) for developing recurrent variceal hemorrhage.

Sclerotherapy

In the sclerotherapy group, bleeding recurred because of variceal hemorrhage in eight patients, portal gastropathy in one patient, and a gastric lipoma in one patient. Three of eight cases of recurrent variceal hemorrhage were caused by gastric varices. In the other five patients, rebleeding occurred before varices were obliterated (n = 3) or after varices were first obliterated (n = 2). No bleeding occurred from sclerotherapy-induced ulcers.

Two of the eight patients with recurrent variceal hemorrhage died before definitive therapy could be instituted: One patient died of exsanguination, and the other died of myocardial infarction. The other six patients crossed over to the TIPS group; TIPS were successfully created in all patients, and the portosystemic gradient decreased to 11 ± 3 mm Hg. During follow-up, four of these six patients experienced at least one other episode of variceal hemorrhage caused by development of stent stenosis. One patient died of recurrent variceal hemorrhage after crossover to TIPS.

Effects on Survival

In contrast to rebleeding, substantial differences in survival were seen between the two groups. Twelve deaths occurred in the TIPS group, and 7 occurred in the sclerotherapy group (Table 2); the median duration of survival was 260 days (CI, 118 to 630 days) for patients assigned to the TIPS group and 1004 days (CI, 740 to 1173 days) for those assigned to the sclerotherapy group. For patients in the TIPS group, the relative risk for death was 1.68 (CI, 0.7 to 3.7). Five patients in the TIPS group and three patients in the sclerotherapy group died of variceal hemorrhage; three and two patients, respectively, died of sepsis. In the TIPS group, all three cases of sepsis (caused by pneumonia in two patients and probable cholangitis in one patient) occurred several months after the TIPS procedure and were not related to the procedure. Two patients in each group died of progressive liver failure. In the TIPS group, one death occurred from iatrogenic traversal of the liver capsule followed by hemoperitoneum that required surgical repair and from a hepatoma that became clinically evident 6 months after randomization. Overall, TIPS were associated with a significantly higher risk for death (Figure 3) (P = 0.02 by generalized Wilcoxon test; P = 0.03 by log-rank analysis).

View larger version (11K): [in this window] [in a new window]   Figure 3. Kaplan-Meier analysis of the risk for death, plotted as the probability of remaining alive over time. Patients in the sclerotherapy group had a definite survival advantage over those in the transjugular intrahepatic portosystemic shunts (TIPS) group (P = 0.03).

A proportional hazards model that included age, Child-Pugh class, encephalopathy, alcohol abuse, and the primary intervention was used to examine potential predictors of survival. None of these variables independently predicted death. Similar results were obtained when the interventions used for recurrent bleeding were incorporated into the model and analyzed as a time-dependent covariate proportional hazards model.

Deaths Related to Recurrent Variceal Hemorrhage

Transjugular Intrahepatic Portosystemic Shunts

Of the five deaths from variceal hemorrhage in the TIPS group, one was caused by multiorgan dysfunction and bleeding gastric varices that occurred before the patient could be transferred to the Medical College of Virginia. A second patient with recurrent varices caused by heart failure died of a ruptured mediastinal varix. Two patients who presented with severe alcoholic hepatitis and recurrent variceal hemorrhage died even though hemostasis was achieved. The fifth patient had previously unsuspected portal venous thrombosis; thus, TIPS could not be created. This patient died after hemorrhage recurred within 7 days of randomization.

Sclerotherapy

In the sclerotherapy group, three patients died of variceal hemorrhage. One death occurred because of exsanguinating hemorrhage, the second because of severe variceal hemorrhage and myocardial infarction within 24 hours of randomization, and the third because of recurrent hemorrhage and aspiration pneumonia developing 3 months after the patient crossed over to TIPS as a result of nonfatal recurrent variceal bleeding. This patient was not considered stable enough for transfer to the Medical College of Virginia and died at another institution.

Complications and Rehospitalization Rates

The two groups had a similar rate of rehospitalization (Table 3). New-onset or acutely worsening encephalopathy occurred in 12 patients in the TIPS group and 5 patients in the sclerotherapy group; this complication accounted for 28 and 9 rehospitalizations in these groups, respectively (P = 0.03 by log-rank analysis). Gastrointestinal bleeding and sepsis were the other major indications for rehospitalization in the TIPS group; hemolytic anemia and alcoholic hepatitis were less frequent causes of rehospitalization. In the sclerotherapy group, post-sclerotherapy oozing that necessitated overnight observation was the most common indication for rehospitalization. Although several patients undergoing sclerotherapy required rehospitalization for decompensated ascites, none of the patients in the TIPS group developed ascites.

Transjugular Intrahepatic Portosystemic Shunts

Portosystemic encephalopathy was the principal complication in patients undergoing creation of TIPS. Eighteen episodes of this complication occurred within 30 days after TIPS creation; 14 of these were associated with a reversible precipitating factor but not with bleeding. Encephalopathy that occurred more than 30 days after the creation of TIPS was related to gastrointestinal bleeding in 8 of 10 cases. Hemolysis associated with TIPS was noted in 5 patients but was severe in only 1.

Sclerotherapy

The most common complication in patients undergoing sclerotherapy was chest discomfort after sclerotherapy. Although 5 patients reported dysphagia, a stricture could be documented by barium studies in only 3 of these patients. These 3 patients responded to an average of three through-the-scope balloon dilatation sessions and were able to maintain their weight. Shallow esophageal ulcers were noted in 22 of 39 patients, but none of these ulcers bled or perforated.

Discussion

Top Methods Results Discussion Author & Article Info References

The ideal treatment for variceal hemorrhage remains elusive [25, 33]. In our study, TIPS were associated with a higher mortality rate than was sclerotherapy, despite similar rebleeding rates in the two groups. Rebleeding was primarily related to recurrence of portal hypertension after TIPS because of stent stenosis caused by neointimal hyperplasia. We recently showed this process to be due to proliferation of cells with smooth-muscle phenotype that synthesize and deposit collagen I and III in the stent lumen [34]. It is important to note that rebleeding occurred despite frequent assessment and correction of TIPS patency. Further research is needed to prevent stent stenosis.

Failure of gastric varices to resolve and recurrence of variceal hemorrhage (which occurred in one patient) are other potentially important predictors of outcome. We have recently embolized such persistent gastric varices with varying success (Unpublished data). Finally, reflection of high systemic venous pressures into the portal tree after TIPS creation upon development of heart failure may also increase the propensity for rebleeding. Clearly, patients with overt heart failure should be managed by alternate methods.

To date, one study comparing TIPS with sclerotherapy has been published in full-length form [35]; four others comparing TIPS with 1) sclerotherapy [36, 37], 2) sclerotherapy and propranolol [38], or 3) endoscopic band ligation [39] have been reported in abstract form. Our randomized study differs from these studies in that it provided longer follow-up (several years) on the outcome of patients after creation of TIPS.

These other studies reported a lower rate of rebleeding in patients who underwent creation of TIPS compared with patients receiving other treatment. Although we observed no significant differences in the relative risk for rebleeding, the wide CIs of the relative risk (0.44 to 2.03) diminish the power of our study and suggest that equivalence of the two treatments in this respect cannot be established beyond doubt. It must also be noted that all of these other studies were limited by a small sample size, as was our study.

It is unlikely that operator inexperience contributed to rebleeding after creation of TIPS in our study; rebleeding rates before and after 1993 were similar. In addition, 7 of 10 episodes of rebleeding were due to neointimal hyperplasia and occurred months after initially adequate portal decompression. Finally, it is worth noting that the rates of rebleeding after TIPS creation were only slightly higher than those reported in the other studies and may reflect our much longer duration of follow-up. Indeed, the principal difference between our study and the other studies is the lower rates of rebleeding in our sclerotherapy group.

Many factors may explain why rates of rebleeding in patients undergoing sclerotherapy were much lower in our study than in the other studies. First, randomization occurred later after cessation of active hemorrhage in our study; this difference reflects the referral pattern of the authors' institution. In addition, we believe that individualizing the sclerotherapy regimen by volume and frequency also helped to decrease the incidence of complicated sclerotherapy ulcers, which could contribute to further rebleeding.

A major difference between our study and the others is the higher mortality rate we observed in patients undergoing TIPS creation. Although significant differences were noted by both Wilcoxon and log-rank analyses, multivariate analyses did not show that the interventions used were independent determinants of death. These differences may have been due to the small sample size, which, as mentioned above, weakened the power of our study; they may also have been due to baseline differences that, although not statistically significant, may have confounded treatment differences. It is also important to note that failure to achieve statistical significance by all of the analytical methods used does not necessarily imply lack of clinical significance. Thus, despite the contrary results achieved with univariate and multivariate analysis, the substantial differences in median survival between the two groups suggest that the findings may indeed be clinically important. In addition, a meta-analysis presented at a recent consensus conference on portal hypertension (Grace N. Personal communication) reported a statistically significant survival advantage for endoscopic treatment compared with TIPS.

An explanation of the differing survival outcomes in our study may be related to the larger number of patients and the longer follow-up. The failure to identify a specific predictor of death in our study may be due to the multiplicity of factors that might affect outcome. Clearly, it will be important to interpret these data in the context of other studies once enough patients have been followed for several years in various centers.

Our study and the other reported studies have several potential limitations. First, as has been noted, none of the studies had large sample sizes, which reduces the power to detect clinical differences. In our study, episodes of active bleeding were treated with sclerotherapy because this procedure can be performed easily and effectively at the bedside during active hemorrhage. Thus, 25% of patients in the TIPS group also received sclerotherapy after study entry. Similarly, crossover occurred in the other direction (sclerotherapy to TIPS); this may also introduce bias. However, bias was avoided because we did an intention-to-treat analysis and used the first rebleeding episode as the primary end point. A time-dependent proportional hazards model indicated that treatment for such rebleeding did not have any demonstrable effects on survival or further rebleeding. The potential value of TIPS as salvage therapy in such patients may have been underestimated by the small numbers of patients who crossed over. It seems unlikely that operator inexperience or a "learning curve" phenomenon affected the outcome of our study.

Both sclerotherapy and creation of TIPS are associated with unique profiles of morbidity [40, 41]. We previously reported that encephalopathy and hemolysis developed after creation of TIPS [42-44]. Most complications of sclerotherapy were relatively minor and were well within the limits described in the literature [41, 45].

In conclusion, sclerotherapy and creation of TIPS were similar for the prevention of recurrent (late) episodes of variceal rebleeding. The latter was associated with a higher long-term mortality rate. Both procedures have a unique spectrum of complications and similar rates of rehospitalization. Our data indicate that in patients who have remained stable for several days after initial resolution of esophageal variceal hemorrhage, sclerotherapy may have some advantage over TIPS as first-line treatment to prevent recurrent variceal hemorrhage. We believe that until controlled trials comparing TIPS with surgical salvage treatment are published, TIPS should be created only for patients in whom sclerotherapy fails.

Presented in part at the annual meeting of the American Association for Study of Liver Diseases in New Orleans, Louisiana, May 1994.

Dr. Freedman: Winter Park Memorial Hospital, Imaging Services Department, 124 East Wellboune Avenue, Winter Park, FL 32789.

Drs. Luketic and Shiffman and Ms. Simmons: Medical College of Virginia Hospital, MCV Station 980341, 11th and Marshall Streets, Richmond, VA 23298-0341.

Dr. Purdum: 1111 Granville Road, Charlotte, NC 28207.

Dr. Cole: Department of Radiology, HSC-L4, State University of New York at Stony Brook, Stony Brook, NY 11790.

Dr. Tisnado: H.H. McGuire Medical Center, Department of Radiology, 1201 Broad Rock Boulevard, Richmond, VA 23249.