The Transjugular Intrahepatic Portosystemic Stent-Shunt Procedure for Variceal Bleeding

Patients and Preoperative Treatment

Table 1. Table 1. Clinical Characteristics of 100 Patients Assigned to Treatment with the Transjugular Intrahepatic Portosystemic Stent-Shunt Procedure.

We attempted to create a transjugular intrahepatic portosystemic shunt in 100 consecutive patients with cirrhosis and recurrent variceal bleeding. Twelve patients were excluded because of contraindications such as portal-vein thrombosis (three patients), severe hepatic encephalopathy (two patients), hepatocellular carcinoma (two patients), and stenosis of the celiac trunk (one patient); two patients refused the procedure, and two died within 24 hours of admission. Portal-vein thrombosis and stenosis of the celiac trunk were diagnosed by duplex sonography and confirmed angiographically. Hepatocellular carcinoma was diagnosed by ultrasonography and confirmed histologically. Ninety patients were treated electively, and 10 on an emergency basis -- i.e., for a recurrence of severe variceal bleeding within the preceding 48 hours and hemorrhagic shock. The patients' characteristics are shown in Table 1. Most of the patients had undergone endoscopic sclerotherapy and were referred to our hospital because of treatment failure. Acute bleeding episodes were treated by endoscopic sclerotherapy with polidocanol or n-butyl-2-cyanoacrylate (Histoacryl, Braun, Melsungen, Germany) or by esophageal balloon tamponade. The 10 patients given emergency measures were treated within 48 hours of admission; all the other patients were in stable condition at the time of treatment. The patients were followed for 3 to 36 months (mean [±SD], 12 ±6); none were lost to follow-up.

The study protocol was approved by the local ethics committee. All patients were informed about the procedure in detail and gave written consent.

Before treatment, duplex sonography (UM9, ATL, Solingen, Germany) was performed with a phased-array transducer (3.5 MHz) and a wall filter of 50 Hz. The sampling size was equal to the vessel diameter, and the angle between the Doppler beam and the longitudinal axis of the vessel was less than 60 degrees. Measurements were performed in triplicate in fasting patients by the same expert operator (reproducibility, ±10 percent). Blood flow in the portal vein, the first-order right and left portal branches, and the shunt was estimated by multiplying the mean blood-flow velocity by the cross-sectional area of the vessel measured by planimetry during real-time imaging.

Hepatic encephalopathy was assessed one to three days before the procedure. Encephalopathy that was mild (stage I, minimal mental disability) or moderate (stage II, moderate mental disability and somnolence) was assessed by the Number Connection Test¹³ and the Mini-Mental State test¹⁴. The results of the Number Connection Test were weighted for age¹⁵. The Mini-Mental State Test evaluates short-time memory, simple calculations, handwriting, and orientation semiquantitatively. Severe encephalopathy (stage III, stupor; stage IV, coma) was assessed by clinical examination.

In patients with tense ascites, 2 to 4 liters of ascitic fluid was removed by paracentesis to improve respiration and to facilitate catheterization of the hepatic vein. Ascites was classified sonographically as minimal (fluid restricted to the abdominal gutter and around the liver, <1 liter), moderate (fluid in the flanks, 1 to 3 liters), or severe (fluid in the entire abdominal cavity, >3 liters).

All patients received lactulose in a dosage that induced two loose stools per day. Blood products were given if the hemoglobin level fell below 9 g per liter, the prothrombin time rose above 22 seconds, or the platelet count fell below 30,000 per cubic millimeter.

Operative Technique and Treatment after Shunting

After sonographic localization of the portal bifurcation¹² and premedication (with midazolam or meperidine [pethidine]), the right hepatic vein was catheterized transjugularly and a needle introduced through the catheter (modified Ross needle RM 15-55.0; 9-French catheter [0-65-45-M-NS-TJC, Cook, Monchengladbach, Germany]). The needle was then advanced under fluoroscopic control, and after puncture of a main branch of the portal vein, a guide wire (Amplatz super stiff 0.035, Meditech, Watertown, Mass.) was introduced through the needle, the catheter was advanced into the portal vein, and the needle was removed. After measurement of the portal venous pressure gradient (i.e., portal venous pressure minus the pressure in the inferior vena cava measured above the hepatic veins), the needle track was dilated with a balloon (Match-35, 7-mm diameter [Schneider, Bulach, Switzerland]) and angiography was performed. Thereafter, a 45-cm 11-French introducer sheath (Cook) was introduced, through which an expandable stent (Palmaz-stent, P308M [Johnson & Johnson, Norderstedt, Germany]) mounted on a balloon catheter (Olbert French 7.5 [Meadox Surgimed A/S, Stenlose, Denmark]) was advanced and expanded to a diameter of 8 to 12 mm. If necessary, this procedure was repeated to place additional stents. The final diameter was adjusted to achieve a portal venous pressure gradient of less than 12 mm Hg and a marked reduction or loss of variceal opacification. Substantial variceal perfusion in spite of reduction of the portal venous pressure gradient to less than 12 mm Hg was observed in eight patients and was treated by embolization of the varices: an end-hole catheter (4 French) was placed 1 to 2 cm into each varix, and a mixture of n-butyl-2-cyanoacrylate and lipiodol (iodized oil) (0.8:1.0, vol/vol) was injected under fluoroscopic control. At the time of stent placement, 1000 U of heparin was given intravenously to all patients except those with severe coagulopathy (see below) or accidental extrahepatic punctures perforating the liver capsule. After final portography and pressure measurements, the sheath for jugular access was removed.

During the first postoperative week, 5000 to 15,000 U of heparin was infused for 12 hours to prolong the partial-thromboplastin time slightly (to 40 to 50 seconds; normal, <40). Thereafter, low-molecular-weight heparin (0.3 ml of fraxiparine given once a day subcutaneously) was administered for one month to avoid early thrombosis of the stent. Anticoagulative measures were not performed in patients with severe coagulopathy (prothrombin time, >22 seconds; platelet count, <30,000 per cubic millimeter); they were also carried out in patients with recent bleeding who had an esophageal balloon tube in place, except in two patients in whom active bleeding was seen angiographically during the procedure. Dietary restrictions (limits on protein, fluid, and sodium intake) were instituted only for overt hepatic encephalopathy or fluid retention. Lactulose treatment was continued in all patients. With few exceptions, patients were discharged one week after the treatment, examined one month later, and examined thereafter at three-month intervals according to an outpatient protocol that included duplex sonography, blood-chemistry measurements, and assessment of hepatic encephalopathy.

Statistical Analysis

Results are expressed as means ±SD. Student's t-test and the chi-square test were used to determine statistical significance. The Kaplan-Meier method was used to calculate cumulative rates of stenosis or occlusion, rebleeding, and survival.

Technique

A transjugular intrahepatic portosystemic shunt was achieved by placement of a stent in 93 of the 100 patients. This treatment failed in seven patients because it was not possible to puncture a main branch of the portal vein: two of these patients underwent surgical treatment (portacaval shunt), and five had endoscopic sclerotherapy. The overall duration of the procedure, including obtaining jugular access, was a mean of 1.2 ±0.3 hours (range, 30 minutes to 3 hours).

Hemodynamic Function

Table 2. Table 2. Portal and Shunt Hemodynamic Function in 92 Patients Receiving Stents.

Table 2 shows the results of direct pressure recording and duplex sonography. The shunt immediately reduced the portal venous pressure gradient by 57 ±14 percent, with a mean shunt diameter of 9.1 ±1.1 mm as determined sonographically; this reduction was mainly due to a reduction in the portal pressure. The pressure in the inferior vena cava increased significantly (P<0.05), probably because of expansion of the central blood volume, and thus contributed to the reduction in the portal venous pressure gradient. The portal-vein diameter was not changed by the shunt. Duplex sonography revealed a 2.5-fold increase in portal flow velocity and portal blood flow. On average, the blood flow in the stent exceeded the portal venous flow by 300 ml per minute because of the additional blood flow from intrahepatic arterioportal communications in patients with retrograde intrahepatic portal blood flow. Retrograde flow was found in 20 percent of the patients before shunting and in 81 percent shortly after the treatment. However, during the following six months, antegrade portal blood flow returned in half of these patients.

Complications

In two patients the stent became dislocated; it was then placed in a peripheral hepatic vein or the iliac vein. In two other patients, a catheter became trapped in the stent mesh and broke off. These technical complications occurred early in the study and did not seriously affect any patient's general condition.

Early clinical complications were encountered in 15 patients and consisted of bleeding (6 patients with intraperitoneal hemorrhage, 4 with biliary hemorrhage, and 3 with hematoma of the liver capsule) and migration of a stent into the pulmonary artery (2 patients). One of these pulmonary embolizations was caused by migration of a stent placed in the iliac vein two days earlier (see above). Neither patient had clinical symptoms, and radiologic evaluation showed that their pulmonary arteries were patent. Among the patients with bleeding complications, bleeding ceased with conservative management in all but two patients. One patient had disseminated intravascular coagulopathy before treatment and died of intraperitoneal hemorrhage within eight hours. The other patient had recurrent episodes of hemobilia, which ceased after the coagulopathy was treated with vitamin K.

Clinical Follow-up

Hepatic Encephalopathy

Overall, the rate of hepatic encephalopathy increased from 10 percent before treatment to 25 percent (23 patients) after treatment. New hepatic encephalopathy of stage I, II, or III developed in 16 patients. Sixteen of the 23 patients with this condition responded to medical treatment (protein restriction and administration of lactulose, paromomycin, and branched-chain amino acids), but 7 did not. The encephalopathy progressed in three of these seven patients before they died of hepatic failure and sepsis; the remaining four patients had mild encephalopathy (grade I or II). In all 23 patients, hepatic encephalopathy appeared during the first three months after placement of the stent.

Patients with hepatic encephalopathy were significantly older than patients without this condition (64 ±8 vs. 54 ±12 years, P<0.05). Fifteen of the 23 patients with encephalopathy (65 percent) were more than 60 years old. These older patients had a 40 percent probability of hepatic encephalopathy, which did not correlate with the Child-Pugh class; patients 60 years old or younger had a probability of 13 percent, and all eight patients who had hepatic encephalopathy were in class C. The mean diameter of the shunts of the patients with hepatic encephalopathy was slightly but not significantly larger than the mean in patients without hepatic encephalopathy (9.3 ±0.8 vs. 9.0 ±1.2 mm, P>0.2).

Liver Function and Child-Pugh Class

Table 3. Table 3. Results of Liver-Function Tests, Child-Pugh Classes and Mean Scores, and Degree of Ascites in 90 Patients before Stent Placement and during Follow-up of Three to Six Months.

Within one week of the shunting procedure, 18 patients had an increase of more than 3 mg per deciliter (52 μmol per liter) in their bilirubin concentrations and 10 patients had levels of aminotransferase activity more than 10 times normal. However, test results improved in most of the patients and had not changed significantly three to six months later (Table 3). Their Child-Pugh scores¹⁶ improved significantly, because of the reduction in ascites (see below).

Ascites

Overall, the treatment was associated with a reduction in ascites in 47 of 53 patients (89 percent) (Table 3). After three months, severe ascites was not detected in any patient. Nine patients (17 percent) required long-term treatment with diuretic agents.

Shunt Patency and Rebleeding

Figure 2. Figure 2. Cumulative Rates of Shunt Stenosis and Shunt Occlusion, and Cumulative Percentages of Patients Free of Variceal Rebleeding among 93 Patients Receiving Stents (Calculated by the Kaplan-Meier Method).

Follow-up with duplex sonography allowed noninvasive detection of stenosis or occlusion of the shunt. In our experience, reduction of the flow in the shunt to less than 1000 ml per minute and the portal flow velocity to less than 10 cm per second was related to shunt insufficiency. According to these criteria, 31 patients (33 percent) had either stenosis (21 patients) or occlusion (10 patients) of the shunt; these complications were confirmed angiographically and treated successfully in all cases by repeat balloon dilation, local thrombolysis, or implantation of an additional stent. Seventy-six percent of the stenoses and occlusions developed at the draining hepatic vein, and 24 percent inside the stent. Thrombosis of the stem of the portal vein was not observed. Of the 31 patients with shunt insufficiency, 10 (11 percent of all patients) had rebleeding from varices and 21 did not. However, all patients with shunt insufficiency had evidence of varices on endoscopy or angiography. Nonvariceal bleeding, found in 12 percent, was due to gastric erosions or sclerosing ulcers. The cumulative rates of stenosis and occlusion and the percentage of patients free of variceal rebleeding (Figure 2) were, respectively, 13, 7, and 92 percent at six months (74 patients) and 33, 15, and 82 percent at one year (31 patients). All stenoses or occlusions first appeared during the first year of follow-up.

Patients with stenosis or occlusion of the shunt had slightly smaller shunt diameters (8.8 ±1.1 vs. 9.3 ±1.1 mm, P = 0.16), significantly higher platelet counts (135,000 ±60,000 vs. 108,000 ±58,000 per cubic millimeter, P<0.01), and significantly lower prothrombin times (14.7 ±1.0 vs. 15.6 ±0.5 seconds, P<0.001) than those without these complications.

Mortality

Figure 3. Figure 3. Cumulative Survival of 93 Patients Receiving Stents, According to Child-Pugh Class (Calculated by the Kaplan-Meier Method), and Number of Patients at Risk.

Ten patients died. Three of the 10 patients treated on an emergency basis died within 30 days of treatment, for an early mortality of 30 percent in this group. Only one patient died of procedure-related causes. Late deaths were due to hepatic failure in six patients who continued to abuse alcohol and to heart failure in one 84-year-old patient. None of the patients died of variceal bleeding. At autopsy, five of the six patients who died of hepatic failure had patent shunts that were covered by a smooth neointima; one patient had a fresh thrombotic occlusion. Life-table analysis of the 100 patients that was based on intention to treat revealed an overall one-year survival rate of 85 percent. The 93 patients receiving stents had an overall one-year survival rate of 87 percent, with rates of 100, 86, and 73 percent among patients in Child-Pugh classes A, B, and C, respectively (Figure 3). No significant difference was found between patients with alcoholic cirrhosis and those with nonalcoholic cirrhosis (85 percent vs. 91 percent). Three of the seven patients (43 percent) who were not treated successfully died of variceal rebleeding 51, 80, and 81 days later.

The current treatments for variceal bleeding -- endoscopic sclerotherapy and surgical shunting -- are associated with high rates of rebleeding and hepatic encephalopathy, respectively. In addition, survival may not be significantly influenced by either treatment¹⁷. Since our results are based on an uncontrolled study, one should use caution in interpreting and comparing them with those of these other treatments.

The incidence of hepatic encephalopathy among our patients was comparable to that among patients with surgical shunts. Among patients in whom a distal splenorenal shunt was created -- the shunting procedure with the lowest incidence of encephalopathy -- the cumulative one-year rates of hepatic encephalopathy in four studies ranged from 4 to 25 percent^18-21. However, these studies included younger patients (mean age, 43 to 54 years) and fewer patients with Child-Pugh class C cirrhosis than did our study.

As shown previously, the incidence of hepatic encephalopathy depends on the shunt diameter²². The best results (9 percent) were obtained with 8-mm portacaval H grafts, which maintained some degree of portal hepatic perfusion. In our study, patients with encephalopathy had larger shunt diameters than patients without the disorder, but the difference was not significant. Fortunately, during six months of follow-up, endothelialization of the stent caused some degree of narrowing, which may explain why the incidence of hepatic encephalopathy was highest during the first three months and declined thereafter.

Our intervention was designed to reduce the portal venous pressure gradient to less than 12 mm Hg and to eliminate variceal perfusion. We did not take full advantage of the potential of the procedure by sequentially expanding the stent. In future studies, placement of small-caliber stents (7 to 8 mm) combined with embolization of the coronary veins should be considered in patients over 60 years of age and patients with Child-Pugh class C cirrhosis, to reduce the risk of hepatic encephalopathy. In our study, the rate of hepatic encephalopathy among such older patients did not correlate with their Child-Pugh class, suggesting that factors other than hepatic function may be important.

Comparing the rate of variceal rebleeding after the placement of transjugular intrahepatic portosystemic stents with the rate after the creation of surgical shunts revealed similar results. During follow-up periods of 2 to 11 years, variceal rebleeding was found in 0 to 21 percent of the patients given surgical shunts^17-32; cumulative 1-year rates were reported in only two studies^18,19 (4 and 12 percent for portacaval shunts and 18 and 30 percent for distal splenorenal shunts). When the stent-shunt procedure was compared with endoscopic sclerotherapy, the latter was associated with a much higher incidence of variceal rebleeding, with rates of 30 to 65 percent at one year^33-39. Better understanding of the pathogenesis of stenosis or occlusion of the shunt may help to decrease the rate of rebleeding and the need for redilation.

Like surgical shunts,^18,23,26 the transjugular intrahepatic portosystemic stent is often associated with an initial rise in bilirubin and ammonia concentrations. Since the prothrombin time and albumin concentration remain unchanged, this increase may be due at least in part to a decrease in hepatic perfusion rather than to a deterioration of hepatocellular function. In patients whose liver function indicates a need for transplantation, the stent may even facilitate the operation by reducing the pressure in the portal vein and portosystemic collaterals beforehand⁴⁰.

The early mortality after implantation of the transjugular intrahepatic portosystemic stent (3 percent) was much lower than that after surgical shunting (10 percent)⁴¹ and was limited to patients who had emergency procedures. Our cumulative one-year survival (87 percent) was comparable to rates in studies of surgical shunts, which ranged from 70 to 90 percent^{18-21,23,24,26,31,42,43}. However, our study population included older patients (15 were more than 70 years old), more patients treated on an emergency basis, and a higher percentage of patients in Child-Pugh class C than most of the surgical trials. In studies of endoscopic sclerotherapy, one-year survival rates ranged from 55 to 90 percent,^33-39 but the two studies with the best results excluded patients in Child-Pugh class C^35,38. Since most of the patients whom we treated had undergone endoscopic sclerotherapy before receiving a transjugular intrahepatic portosystemic stent, it may be more appropriate to compare our trial with trials of endoscopic sclerotherapy with salvage shunting. In the study of Warren et al.,²⁶ the one-year survival rate among patients who underwent sclerotherapy with salvage distal splenorenal shunting was 88 percent, a rate thus more comparable to ours.

Recently, 150 other patients treated in our hospital underwent implantation of a transjugular intrahepatic portosystemic stent, with complete success. The findings from the short follow-up of these patients and others recently described⁴⁴ agree with those we have reported here.