The transjugular intrahepatic portosystemic shunt (TIPS) procedure is a nonsurgical procedure in which an expandable metal prosthesis is used to connect an intrahepatic portal vein with an adjacent hepatic vein [3, 4]. In our initial report on 100 patients having this procedure [3], TIPS stent placement was technically successful in 96 patients and variceal hemorrhage was controlled in 88 of 94 patients [3]. Furthermore, the 30-day mortality rate was only 13% in patients treated with TIPS. These data, as well as those of other researchers who used radiographically placed portosystemic stents, suggest that TIPS might be more cost-effective than sclerotherapy; however, data comparing the two procedures are limited [5-10]. Therefore, we did a randomized, controlled trial in patients with massive acute variceal hemorrhage in an effort to compare the two therapies for the prevention of recurrent variceal hemorrhage.

Methods

Top Methods Results Discussion References

From November 1991 through December 1995, we enrolled 49 adults who had cirrhosis and endoscopically documented bleeding from esophageal varices. We excluded an estimated additional 250 patients with bleeding varices whom we had seen during the study period (vide infra). Our study protocol was approved by the Committee on Human Research of the University of California, San Francisco. Patients who were admitted to San Francisco General Hospital, University of California Medical Center, and Veterans Affairs Medical Center (all located in San Francisco, California) with massive or submassive acute gastrointestinal tract hemorrhage from large esophageal varices were approached for consent and randomization within 24 hours of admission. Massive hemorrhage was defined as bleeding associated with shock (systolic blood pressure < 80 mm Hg). Submassive hemorrhage was defined as hemorrhage associated with postural vital sign changes (upright pulse rate increased by 20 beats per minute compared with supine pulse rate; upright systolic blood pressure decreased by 20 mm Hg compared with supine blood pressure).

The 250 excluded patients were excluded for the following reasons: They were prisoners; they were younger than 18 or older than 75 years of age; they had had a cerebrovascular accident within 3 months before the onset of bleeding; they refused to accept blood products; or they had gastric variceal hemorrhage, electrocardiographic changes compatible with acute myocardial infarction, a PO₂ less than 70 mm Hg or an arterial pH of 7.20 or less on room air at the time of evaluation for eligibility, a serum creatinine level of 221 µmol/L or more, a prothrombin time at least 5 seconds longer than control (despite the use of fresh frozen plasma), a platelet count less than 50 x 10⁹/L, stage IV hepatic encephalopathy, cancer other than skin cancer, the acquired immunodeficiency syndrome (AIDS) or advanced AIDS-related complex, sepsis, pneumonia, peritonitis, clinical evidence of alcoholic hepatitis, a serum bilirubin concentration of 7 mg/dL or more, thrombosis of the portal vein, thrombosis of the hepatic veins, or thrombosis of the inferior vena cava as determined by Doppler ultrasonography.

All patients received endoscopic sclerotherapy at the time of the initial endoscopic procedure that established the source of hemorrhage as esophageal varices. Patients were deemed eligible for participation if they presented with hemodynamically submassive or massive hemorrhage and were found to have large (>1 cm across) distal esophageal varices with cherry red spots, hematocystic spots, or red wale signs. Before randomization, all patients had patency of the portal venous system (main, right, and left portal veins and the splenic vein) and hepatic veins determined by real-time color and pulse-wave Doppler ultrasonography.

After we obtained informed consent, we used serially numbered, sealed, opaque envelopes to randomly assign patients either to repeated sclerotherapy or to TIPS. If neither the patient nor the patient's next of kin was able to give informed consent, a patient advocate was designated to consider the invitation to participate. Patients randomly assigned to sclerotherapy received treatment every 2 to 7 days during the initial hospitalization; treatment consisted of 0.5- to 2.0-mL injections of ethanolamine oleate solution per varix. Repeated endoscopy and sclerotherapy treatments were done weekly after discharge from the initial hospitalization. As much as 30 mL of ethanolamine oleate solution was used per treatment session. All visible varices were injected within the distal 7 to 10 cm of the esophagus. In patients who developed sclerotherapy-associated ulcers, repeated endoscopy was scheduled to be done 2 to 7 days after the notation of ulcers to assess interval healing.

Patients assigned to the TIPS group had the procedure within 48 hours of randomization; the procedure was performed by one of six radiologists skilled in this procedure, as described elsewhere [3]. Catheterization of the hepatic vein was done through the right internal jugular vein. A tract between a suitable hepatic vein and a suitable portal vein was established by needle set (Ring TIPS set, Cook, Inc., Bloomington, Indiana), dilated with a balloon over a guidewire, and then maintained by one or more expandable metal mesh stents (Wallstent, Schneider, Inc., Minneapolis, Minnesota). The adequacy of the portosystemic shunt was documented by contrast injection and manometric measurement at the time of the initial procedure. A target portal vein-to-hepatic vein pressure gradient of 12 mm Hg or less was achieved in all cases. Persistent varices opacified at portal venography after adequate stenting were occluded by using embolization coils.

Preprocedural data recorded prospectively included sex, age, vital signs at admission, physical examination findings at admission (including presence of encephalopathy and ascites), nutritional status, results of laboratory tests, and Child-Pugh score [11]. Prospectively identified outcome variables after randomization included death, rebleeding, liver transplantation, total transfusion requirements, onset and presence of encephalopathy, cost of managing variceal hemorrhage after randomization, total duration of hospitalization for variceal hemorrhage and any related encephalopathy, and complications of therapy. Nutritional status was defined as "malnourished" if the patient had gross muscle wasting, had cachexia, or had lost at least 10% of body weight during the previous 6 months. "Bleeding" after randomization was defined as bloody or "coffee grounds" emesis (hemetemesis, melenemesis) or liquid black stools (melena) with a decrease in hematocrit sufficient to warrant transfusion. Hepatic encephalopathy was defined clinically by the presence of asterixis, gross disorientation or agitation, or frank somnolence or coma in the absence of another identifiable cause. Presence of ascites was determined by both ultrasonography and clinical assessment (shifting dullness, fluid wave, gross distention) for all patients initially and for patients in the TIPS group having follow-up Doppler ultrasonography. For patients assigned to sclerotherapy, presence of ascites was subsequently determined by clinical criteria alone.

Follow-up information was obtained through face-to-face interviews, telephone interviews, or chart reviews and was obtained from the patient, family, physician, or all three sources. The total cost of health care per patient was calculated as the sum of all real costs for inpatient and outpatient hospital care, including hospital expenditures and costs for professional services from the day of randomization until death or the last follow-up visit. In addition, all outpatient costs for endoscopic sclerotherapy, Doppler ultrasonography, and stent revision during the follow-up period were included. We used the actual cost to the hospital or medical staff, or both, of providing a service or procedure (rather than billing charges or collections). For example, the cost of an endoscopic sclerotherapy session was determined by summating the following: 1 hour of a gastroenterologist's time plus benefits (derived from personnel pay records); 2 hours of a registered nurse's time plus benefits; the invoice cost of disposable sclerotherapy catheters, bite blocks, and intravenous tubing; the pharmacy costs of all drugs, including the sclerosant agent; the estimated depreciation of an Olympus GIT-IT100 videoendoscope (Lake Success, New York); endoscopic processing costs; 1 hour of recovery room personnel time; and costs of recovery supplies.

Outcome variables were compared, using the intention-to-treat paradigm, by either the Fisher exact test or a two-tailed Mann-Whitney nonparametric test for independent samples. Survival was defined as the time from randomization to death from any cause. The survival curve for each treatment group was estimated by using the Kaplan-Meier method, and the survival curves were compared by using the log-rank test. The Cox proportional-hazards model was used to test statistical hypotheses and to study the effect of other possible covariates on outcome. The primary end point was death. In addition to treatment with TIPS or sclerotherapy, other potential covariates were examined: age, hematocrit, prothrombin time, platelet count, bilirubin concentration, albumin level, Child-Pugh score, number of hours from admission to therapeutic procedure (TIPS or initial sclerotherapy), and number of units of blood received before randomization. Univariate analysis was initially used to determine potentially significant covariates. Forward stepwise selection was then done to determine the optimum Cox proportional-hazards model. Adjusted survival curves were calculated using the method of Breslow [12] to adjust for differences in baseline characteristics between the TIPS and sclerotherapy groups and thus compare the groups on a more equal footing. The proportionality assumption used by the Cox models was tested for all covariates in all models [13].

Sample Size

On the basis of our previous work [1, 2], we anticipated a 75% rate of rebleeding after endoscopic sclerotherapy. On the basis of our previous published report [3], we anticipated rebleeding in the TIPS group to be 25%. To achieve an of 0.05 (two-tailed) and a ß of 0.10, we estimated that we would need a sample size of 20 patients per group.

Results

Top Methods Results Discussion References

Clinical and Laboratory Data

Twenty-five patients were randomly assigned to sclerotherapy, and 24 were randomly assigned to TIPS. Blood transfusion requirements before randomization were higher in the TIPS group than in the sclerotherapy group (5.3 ± 0.5 units compared with 4.0 ± 0.5 units; P = 0.016) (Table 1). The TIPS group also had a worse Child-Pugh score (P = 0.049). No other pretreatment clinical or laboratory variables were significantly different in the two groups, including time from admission to randomization (35.4 ± 5.6 hours compared with 37.4 ± 4.7 hours). Time in the hospital before randomization was spent resuscitating, assessing patient eligibility by endoscopy, performing Doppler ultrasonography, and collecting laboratory data. According to protocol design, patients assigned to sclerotherapy received their first treatment at the index endoscopy, 17.3 ± 4.7 hours after admission. Those assigned to TIPS had stent placement 59.5 ± 6.7 hours after admission.

Outcome of Index Hospitalization

Twenty-one of 25 patients in the sclerotherapy group and 19 of 24 patients in the TIPS group were discharged alive from the index hospitalization (P > 0.2). Moreover, 4 patients in the sclerotherapy group and 5 patients in the TIPS group died within the first month after hospitalization for the index variceal hemorrhage (P > 0.2). The patients who died within 30 days after randomization did not differ from those who survived in terms of Child-Pugh score, laboratory variables, or varix size. Portosystemic pressure gradient measurements taken before and after TIPS were available for 22 of the 24 patients in whom TIPS stent placement was successful. Mean portal vein pressures decreased from 23.9 ± 1.9 mm Hg to 7.73 ± 0.52 mm Hg (P < 0.001) after the initial TIPS procedure.

Long-Term Follow-Up

The clinical course of the study patients is shown in Table 2. During follow-up (approximate mean, 570 days), varices were found to be obliterated in 11 of the 25 patients assigned to sclerotherapy. Seventeen of 18 patients assigned to TIPS had varices that were found to be obliterated on a follow-up endoscopy procedure that was done not according to protocol design but because of recurrent upper gastrointestinal bleeding, anemia, or symptoms of reflux. Varices were therefore more likely to be obliterated in patients assigned to TIPS than in those assigned to sclerotherapy (P = 0.001). The number of episodes of rebleeding from varices was also significantly lower in patients assigned to TIPS than in those assigned to sclerotherapy (0.17 ± 0.10 compared with 0.88 ± 0.25 per patient; P = 0.016). However, total days of hospitalization for variceal hemorrhage, number of hemorrhagic events in the upper gastrointestinal tract after randomization, days of hospitalization for any upper gastrointestinal tract bleeding, number of blood units received after randomization, and total transfusion requirements after randomization did not differ significantly between the groups.

New or worsened portosystemic encephalopathy was noted during follow-up in 11 of 25 patients receiving sclerotherapy and 12 of 24 patients receiving TIPS (P > 0.2). Furthermore, no significant difference was seen between the groups in the prevalence of encephalopathy after randomization (Table 2) or the number of days of hospitalization for encephalopathy.

Six patients initially assigned to receive sclerotherapy crossed over to receive TIPS because of recurrent variceal hemorrhage that did not respond to repeated sclerotherapy. One patient initially assigned to receive TIPS crossed over to receive sclerotherapy because stent placement was technically impossible. Twenty-two of 24 patients assigned to receive TIPS had follow-up Doppler ultrasonographic examinations of the portal vein. These 22 patients had 3.05 ± 0.57 Doppler ultrasonographic evaluations (median, 2.0; range, 1 to 10). Four patients were found to have stent stenosis or occlusion, or both, as evidenced by decreased portal venous flow velocity. These 4 patients had five separate balloon angioplasty sessions on days 90, 190, 222, 295, and 497 after initial stent placement. Three patients each had a single successful angioplasty revision. One patient had angioplasty revision of the TIPS on two separate occasions and required placement of a new TIPS prosthesis on day 300.

No significant difference was seen in crude long-term survival between the sclerotherapy and the TIPS groups, as measured by the log-rank test (P > 0.2). In Cox models, the unadjusted relative hazard ratio of the TIPS group was 1.02 (95% CI, 0.38 to 2.7). As noted above, several clinical variables were worse in the TIPS group at baseline despite randomization. The final Cox proportional-hazards model (Table 3) includes the covariates prothrombin time and blood transfusion requirements before randomization. Other clinical variables, including the Child-Pugh score, did not prove to be independent predictors of survival. The estimated hazard ratio for TIPS in the final Cox model was 0.53 (CI, 0.18 to 1.5). The adjusted survival curves are shown in the (Figure 1). No strong evidence argued against the assumption of proportionality (P > 0.2), although survival rates appeared to be similar for the first 2 years. The TIPS group showed a survival advantage only after that time.

View larger version (17K): [in this window] [in a new window]   Figure 1. Adjusted survival curves for patients randomly assigned to transjugular intrahepatic portosystemic shunt (TIPS) or sclerotherapy.

Health care costs did not differ significantly between the groups (Table 2). Health care costs for patients assigned to sclerotherapy were $27 540 ± $5088; those for patients assigned to TIPS were $29 790 ± $3422 (P > 0.2). When calculated as health care costs per survival day, no statistically significant difference was seen between the two groups ($354 ± $108 for the sclerotherapy group compared with $475 ± $154 for the TIPS group; P > 0.2).

Discussion

Top Methods Results Discussion References

Patients with massive hemorrhage from esophageal varices present a considerable challenge to clinicians around the world. Our patient sample was largely male, middle-aged, white, and actively alcoholic, a sample that is representative of most patients in the United States with bleeding varices. Our patients required transfusion of substantial amounts of blood for resuscitation, and nearly half of them required endotracheal intubation for airway protection so that initial diagnostic and therapeutic endoscopy could be done safely. Despite randomization, the TIPS group was sicker, had required more transfusions, and had a worse Child-Pugh score at the time of randomization. The two groups were otherwise well matched in terms of demographic, clinical, and laboratory characteristics.

More reliable obliteration of varices and significantly fewer episodes of repeated variceal hemorrhage were seen in patients assigned to receive TIPS. Other variables, however, including days of hospitalization for bleeding, transfusion requirements after randomization, prevalence of portosystemic encephalopathy, and health care costs, did not differ significantly between the groups.

A trend toward improved survival (which was not statistically significantly different in the two groups) was noted in the TIPS group (hazard ratio, 0.53 [CI, 0.18 to 1.5]). Our sample size may have limited our power to show statistically significant improvements in survival. As suggested by the adjusted survival curves, any survival advantage may be substantial only after 2 years. It is also possible that no true difference exists between the two forms of therapy, despite the trend that we noted.

Six other studies have assigned patients to receive sclerotherapy or TIPS [5-10] (Table 4). Four of these studies reported that TIPS-treated patients had significantly less rebleeding from varices than did patients receiving sclerotherapy. In addition, Cabrera and coworkers [6] noted a decrease in the number of days of hospitalization and transfusion requirements for patients assigned to receive TIPS. Sanyal and colleagues [5] reported a higher mortality rate for patients having TIPS, whereas the other studies noted no difference in survival between TIPS- and sclerotherapy-treated patients. Three of the six studies, unlike our own, noted an increase in the prevalence of encephalopathy after TIPS compared with sclerotherapy.

Endoscopic band ligation, which was introduced after our study was started, may result in a lower rate of rebleeding and decreases in the number of hospital days and transfusion requirements compared with endoscopic sclerotherapy [14-18]. Several recent studies comparing sclerotherapy and esophageal band ligation for the treatment of variceal hemorrhage showed a reduction of approximately 33% in the number of treatment sessions required to obliterate varices by banding compared with sclerotherapy. In theory, if the treatment period required to obliterate varices is shorter with band ligation, patients having this procedure would be less likely to have rebleeding and, accordingly, would probably require fewer days of initial hospitalization. If fewer treatment sessions are required for band ligation than for sclerotherapy, it is also possible that overall health costs after index bleeding (unlike in our study) would be reduced in patients receiving band ligation compared with patients receiving sclerotherapy. Further studies are required to compare the efficacy of band ligation with that of TIPS.

It is possible that the addition of nonselective ß-blockers or nitrates, or both, would reduce the rates of rebleeding in patients receiving sclerotherapy and patients receiving TIPS. However, the only randomized trial that has compared TIPS with sclerotherapy using supplemental ß-blockade showed higher rates of rebleeding than were seen in the other six trials (including our own), which did not use ß-blockers [10].

Although liver transplantation is a highly effective treatment for patients with cirrhosis and associated portal hypertension, most patients with esophageal variceal hemorrhage are not candidates for liver transplantation because of active alcoholism. Only 1 of our 49 study patients ultimately had liver transplantation. Thus, TIPS and sclerotherapy are appropriate treatment options for most cirrhotic patients with variceal hemorrhage (particularly those whose underlying disease renders them poor candidates for transplantation) and for patients awaiting liver transplantation.

Portosystemic encephalopathy is a potential concern in patients undergoing a TIPS procedure [3-10, 18]. However, we found no difference in the incidence of new or worsened clinically determined portosystemic encephalopathy or in the overall prevalence of encephalopathy after randomization between patients randomly assigned to receive TIPS and those randomly assigned to receive sclerotherapy. We speculate that the high degree of severe underlying liver disease in our entire patient sample at large may have obscured potential differences in encephalopathy between the TIPS group and the sclerotherapy group. Perhaps in a patient sample with less severe initial hepatic dysfunction, portosystemic encephalopathy would be evident more frequently in patients treated with TIPS.

The management of acutely ill patients with active variceal hemorrhage remains problematic. Moreover, outpatient follow-up of these patients is critical. For example, patients treated with sclerotherapy must return for multiple sclerotherapy sessions. Although patients treated with TIPS may be expected to do well initially, these patients also require careful follow-up (including ultrasonographic surveillance of their shunts) because TIPS stenosis or occlusion, or both, may occur at any time. It is possible that if patients treated with TIPS are followed for longer than they were in our study, a greater prevalence of occlusion and, consequently, an overall higher cost of health maintenance would be revealed. On the other hand, if the incidence of TIPS stenosis or occlusion tapers off with time, this cost would be decreased. As noted above, the adjusted survival curves suggest improved survival in the TIPS group 2 years after the initial intervention. At present, data on survival over a longer period are not available.

How should the physician approach patients with massive or submassive variceal hemorrhage? Our study and others indicate that continued vigorous sclerotherapy or immediate TIPS stent placement are both acceptable. However, it must be emphasized that both treatment programs require good patient compliance during long-term follow-up to prevent additional episodes of variceal hemorrhage. The following types of patients may be best served by TIPS: patients unable to tolerate endoscopy; patients with little hemodynamic reserve; noncompliant patients or patients who reside far from medical care facilities; and patients with esophageal strictures, tumors, or other contraindications to repeated endoscopy.

In contrast, endoscopic sclerotherapy may be best for patients with extensive portal or hepatic vein thrombosis; anatomy unsuitable for TIPS; superior vena cava thrombosis; biliary obstruction; cholangitis; allergy to radiographic contrast medium; the Budd-Chiari syndrome; untreatable inferior vena cava webs; or other contraindications to TIPS, such as pulmonary hypertension.

Patients who do not respond to one of these methods may cross over to receive treatment with the other. Sclerotherapy and TIPS, together with newer techniques, such as banding or banding combined with sclerotherapy, are attractive options. Choosing the best therapy for the management of bleeding esophageal varices in a given patient requires a careful evaluation of key clinical, anatomic, and procedural features.

Previously presented at Digestive Disease Week, San Diego, California, 14-19 May 1995. Previously published in abstract form in Gastroenterology (1995; 108:A1045).

From the University of California at San Francisco, San Francisco General Hospital, University of California Medical Center, and Veterans Affairs Medical Center, San Francisco, California.

Drs. Ring, Gordon, LaBerge, and Wall: Radiology Department, Box 0628, M-361, University of California, San Francisco, CA 94143-0628.

Dr. Olcott: Stanford University School of Medicine, Veterans Affairs Palo Alto Health Care System, 3801 Miranda Avenue, DRC/114, Palo Alto, CA 94304.

Dr. Sandhu: Department of Radiology, University of North Carolina, 206 Main Hospital, Box 7510, Chapel Hill, NC 27599-7510.

Dr. Morgan: Wake Medicine Service, University of North Carolina, Wake Medical Center, 3024 New Bern Avenue, Suite 301, Raleigh, NC 27601.

Dr. Ostroff: Gastroenterology Consult Service, Box 1007, University of California, San Francisco, San Francisco, CA 94143-1007.

Dr. Rockey: Liver Center Laboratory, Sand Building, Room 334, Research Drive, Duke University Medical Center, Durham, NC 27710.

Dr. Bacchetti: Box 0840, University of California, San Francisco, San Francisco, CA 94143-0840.

Dr. Lake: Box 0780, M-896, University of California, San Francisco, San Francisco, CA 94143-0780.

Dr. Somberg: Box 0630, C-349, University of California, San Francisco, San Francisco, CA 94143-0780.

Ms. Doherty: Box 0538, S-357, University of California, San Francisco, San Francisco, CA 94143-0538.

Dr. Davila: Division of Gastroenterology, Laboratory Surge Building, Room P-304, Stanford University Medical Center, Stanford, CA 94305.

Dr. McQuaid: Veterans Affairs Medical Center 1A-1, Box 0120, 4150 Clement Street, San Francisco, CA 94121.