Potential conflict of interest: Nothing to report.
The clinical usefulness of assessing hemodynamic response to drug therapy in the prophylaxis of variceal rebleeding is unknown. An open-labeled, uncontrolled pilot trial was performed to evaluate the feasibility and efficacy of using the hemodynamic response to pharmacological treatment to guide therapy in this setting. Fifty patients with acute variceal bleeding underwent a hepatic venous pressure gradient (HVPG) measurement 5 days after the episode. Nadolol and nitrates were initiated, and a second HVPG was measured 15 days later. Responder patients (≥20% decrease in HVPG from baseline) were maintained on drugs, partial responders (≥10% and <20%) had banding ligation added to the drugs, and nonresponders (<10%) received a transjugular intrahepatic portal-systemic shunt (TIPS). Mean follow-up was 22 months. Eight patients (16%) did not receive the second HVPG, 6 of them because of early variceal rebleeding. Of the other 42 patients, 24 were classified as responders (57%); 10 as partial responders (24%), who had banding added; and 8 as nonresponders (19%), who received a TIPS. Patients with cirrhosis of viral etiology compared to alcoholic cirrhosis tended to present more early rebleedings, less response to drugs and needed more TIPS. Variceal rebleeding occurred in 22% of all patients but only in 12% of patients whose hemodynamic response was assessed. The 3 therapeutic groups were not different. In conclusion, using hemodynamic response to pharmacological treatment to guide therapy in secondary prophylaxis to prevent variceal bleeding is feasible and effectively protects patients from rebleeding. In this context, viral cirrhosis seems to present a worse outcome than alcoholic cirrhosis. (HEPATOLOGY 2006;44:806–812.)
The recommended prophylaxis for variceal rebleeding includes pharmacological therapy—nonselective β-blockers alone or with nitrates—endoscopic band ligation, or a combination of both.1, 2 For patients treated with drug therapy, evaluating the hemodynamic response by repeated measurement of the hepatic venous pressure gradient (HVPG) has been recommended.2, 3 However, the clinical utility of this approach is unknown, the best therapies for those who do not respond to drugs have not determined, and very few studies have addressed this strategy. Only 2 studies have prospectively used the hemodynamic response to nonselective β-blockers to guide prophylaxis for variceal rebleeding.4, 5 One of the studies included only 14 patients treated after an episode of variceal bleeding.4 In the other study,5 the repeat HVPG was performed 3 months after starting propranolol, by which time 50% of the patients had already rebled, died, or were unavailable for pressure measurements.
We conducted an open-labeled, uncontrolled pilot trial in which the hemodynamic response to pharmacological treatment was used to guide therapy in secondary prophylaxis to prevent variceal bleeding. The aim of this pilot trial was to determine if prophylactic treatment assigned according to response to pharmacotherapy was effective in preventing rebleeding. To optimize the clinical results of this study, nonselective β-blockers were administered at the maximum tolerated doses, the time frame to assess hemodynamic response was shortened to less than 3 weeks, and the different therapeutic options were selected according to the level of hemodynamic response achieved.
Between January 2001 and December 2005, 149 patients with cirrhosis were admitted to the Bleeding Unit of our hospital because of acute esophageal variceal bleeding. Bleeding was considered of esophageal variceal origin if the emergency endoscopy performed within 12 hours of admission showed any of the accepted criteria defining variceal bleeding.6 Diagnosis of cirrhosis was based on the usual clinical, biochemical, and sonographic criteria. Acute bleeding was treated with somatostatin perfusion for 5 days with or without endoscopic therapy. All patients received oral norfloxacin prophylaxis for 7 days. A total of 99 patients (Fig. 1) were excluded from the trial because at least one of the following features was present: age older than 75 years (n = 16), Child-Pugh score > 12 (n = 12), failure to control index bleeding (n = 8), prior therapy with β-blockers and nitrates or endoscopic variceal obliteration (n = 25), contraindications to β-blockers or nitrates (n = 6), advanced hepatocellular carcinoma (n = 22), severe associated conditions (n = 12), portal thrombosis (n = 14), and an HVPG < 10 mm Hg (n = 2). The remaining 50 (33%) patients were included in the study after giving written informed consent. Time of inclusion was the moment of the first hemodynamic study. The trial was approved by the ethics committee of our institution.
Five days after the index bleeding, a first hepatic hemodynamic study was performed after stopping somatostatin. Pharmacological therapy with β-blockers and nitrates was subsequently started. Once the maximum doses of both drugs were achieved, a second hemodynamic study was scheduled 5-7 days later. According to the hemodynamic response observed, patients were classified into 1 of 3 groups: responders, when the HVPG was ≤12 mm Hg or had decreased at least 20% from baseline values2; partial responders, when the HVPG had decreased at least 10% but less than 20%, and nonresponders when the decrease in the HVPG was less than 10%. Responder patients were maintained on the same pharmacological therapy, partial responders were treated with banding ligation in addition to drugs, and nonresponders had a transjugular intrahepatic portal-systemic shunt (TIPS) inserted.
Patients were followed at 3 months intervals at the outpatient liver clinic. History and examination assessed alcohol abstinence, compliance, and side effects. The study was continued for 3 months following the enrollment of the last patient. The primary end point was clinically significant rebleeding from portal hypertensive sources as defined by Baveno IV,1 which is equivalent to failure of secondary prophylaxis. Under these circumstances, alternative therapy was instituted (e.g., banding in those on drug therapy, TIPS, or transplantation).
Continuous pharmacological therapy was started immediately in all patients after the first hemodynamic study. Nadolol was given orally at an initial dose of 40 mg once daily. The dose was subsequently increased for 5 days until intolerance appeared, the heart rate descended to 55 beats per minute, or a 160-mg dose was reached. Oral isosorbide mononitrate was then started at 20 mg once daily at bedtime, followed by 20 mg of BD for 2 days, which was finally increased to 40 mg of BD if tolerated.
Patients assigned to banding underwent ligation within 24 hours after the second hemodynamic study, in which they were classified as partial responders. Ligation was performed with commercial multiband devices at 2-week intervals until variceal obliteration had been achieved. Once varices were eradicated, patients underwent a follow-up endoscopy at 3 months and then at 6-month intervals thereafter. Additional sessions of ligation were conducted if varices reappeared.
TIPS placement was carried out within 72 hours of the second hemodynamic study in patients classified as nonresponders. The portal-systemic pressure gradient was reduced to < 12 mm Hg. After the procedure, patients underwent follow-up Doppler ultrasound controls on day 7 and then every 3 months, and a control upper endoscopy was performed at 3 months. Angiography with measurement of the portal-systemic pressure gradient was also conducted at 6 and 12 months after TIPS insertion and whenever shunt dysfunction was suspected. The first 2 patients received uncoated prostheses (Wallstent; Boston Scientific, Barcelona, Spain), and the remaining patients PTFE-coated prostheses (Viatorr Endoprothesis; WL Gore, Flagstaff, AZ).
Hemodynamic studies were performed after an overnight fast. Under local anesthesia, a venous introducer was placed in the right femoral vein or internal jugular vein by the Seldinger technique. Under fluoroscopy, a 7F balloon-tipped catheter (Boston Scientific Medi-Tech, Natick, MA) was guided into the main right vein for measurements of wedged (occluded) and free hepatic venous pressures. The HVPG was obtained by subtracting the free from the wedged hepatic pressure gradient. All measurement were performed in triplicate, the average taken, and printed tracings of the hemodynamic studies were reviewed by investigators unaware of clinical data.
Overall rebleeding and mortality analysis was conducted on an intention-to-treat basis. Follow-up was censored on the date the patient was last seen or the date of death or of transplantation. Data are expressed as mean ± SD. Frequency data were compared using the χ2 test or the Fischer's exact test, when necessary. Quantitative variable were analyzed with the Student t test. A P value of < .05 was required for statistical significance. The SigmaStat 3.00 statistical package was used.
Patient Features and Hemodynamic Studies.
Basal clinical and hemodynamic features of the 50 patients in the study are shown in Table 1. The majority (92%) of the 27 patients with cirrhosis of alcoholic etiology (with or without viral etiology) were consuming alcohol (>80 g/day) at the time of bleeding. Fifteen patients were already taking β-blockers before the index bleeding, 10 as primary prophylaxis and 5 to prevent variceal rebleeding. The basal hemodynamic study was performed 5.3 ± 1.3 days after the index bleeding. A second HVPG measurement was not carried out in 8 patients (16%). Two of these patients died of progressive liver failure 8 and 15 days after the index bleeding. The other 6 patients showed variceal rebleeding 16.6 ± 4.9 days after the index bleeding and before the second hemodynamic study could be conducted, one of whom died. The remaining 42 patients underwent a second hemodynamic study 15.6 ± 2.9 days after the first study, that is, approximately 21 days after the index bleeding. Table 2 compares the characteristics of the 6 patients who bled before the second HVPG measurement with the 42 patients who underwent the second hemodynamic study. As Table 2 shows, patients who had early rebleeding more frequently had cirrhosis of only viral etiology. Mean nadolol dose was similar for the 6 patients who rebleed before the second HVPG measurement, but mean nitrates dose was significantly lower. At the time of bleeding, the 6 patients who were early rebleeders were already taking the maximum tolerated dose of nadolol. However, 3 of these patients were not taking the maximum dose of nitrates because of intolerance (1 case) or because rebleeding occurred at the beginning of nitrates therapy (2 cases). No differences between patient groups were observed for the rest of the parameters examined.
Table 1. Clinical and Hemodynamic Features of Patients at Time of Inclusion
N = 50
Grade 1, varices flattened by insufflation; grade 2, varices not flattened by insufflation and separated by areas of normal mucosa; grade 3, confluent varices not flattened by insufflation.
Table 2. Comparison of Characteristics at Time of Inclusion Between the 6 Patients Who Rebled and the 42 Patients Who Did Not Rebleed Before the Second Hemodynamic Study
Bleeders Before Second Study (n = 6)
Nonbleeders Before Second Study (n = 42)
58 ± 10
52 ± 10
Alcohol + virus
6.6 ± 1.5
7.6 ± 2.1
2.8 ± 0.3
2.7 ± 0.4
1.4 ± 0.6
2.8 ± 2.8
Prothrombin rate (%)
58 ± 10
55 ± 14
Prior β-blocker prophylaxis
Variceal size (1/2/3)
Treatment of index episode
Endoscopic therapy plus somatostatin
Units of packed red cells
2.4 ± 1.6
3.1 ± 2.1
Basal heart rate (beats/min)
68 ± 10
77 ± 11
Hepatic venous pressure (mm Hg)
9.2 ± 4.5
10.5 ± 5.3
26.6 ± 5.1
29 ± 6.9
17.4 ± 4.2
18.4 ± 4.3
Nadolol dose (mg/day)
100 ± 53
123 ± 46
Nitrates dose (mg/day)
50 ± 35
80 ± 0
As shown in Figure 1, of the 42 patients who underwent a second HVPG measurement, 24 (57%) were classified as responders and maintained on drug therapy, 10 (24%) were partial responders and had banding ligation added to their drug therapy, and 8 (19%) were nonresponders and received a TIPS. Among the 42 patients with hemodynamic response assessment, differences between the 3 hemodynamic response groups were not observed in age, sex, and Child-Pugh distribution. Table 3 presents data on the doses of pharmacological treatment administered, the time to reach the maximum dose of drugs, heart rate, and hepatic venous pressures of the 3 study subgroups. Differences were only observed in the second hemodynamic study results. It is worth mentioning that 7 of the 8 patients (87%) who received a TIPS had cirrhosis of pure viral etiology.
Table 3. Dose of Drug Therapy, Time to Reach Therapeutic Target, Heart Rate, and Hepatic Venous Pressure of the 3 Study Subgroups According to Hemodynamic Response and Treatment Assigned
Responders—Drug Therapy (n = 24)
Partial Responders Banding+Drugs (n = 10)
Nonresponders TIPS (n = 8)
NOTE. No differences among groups were observed, except for the second hemodynamic study.
Table 4 compares the characteristics of responder patients with those of patients who had a low response (partial responders + nonresponders). As can be seen in Table 4, differences were not detected, except for a trend toward more patients with cirrhosis of only viral etiology in the low-response group. A direct comparison of those with cirrhosis of pure alcoholic etiology with those with cirrhosis of pure viral etiology suggested a higher response rate in patients with alcoholism (Table 4). In fact, 13 of the 17 (76%) patients with cirrhosis of pure alcoholic etiology responded to drug therapy, whereas only 6 of the 16 (37%) patients with cirrhosis of pure viral etiology were responders (P = .03).
Table 4. Baseline Comparison Between Responder Patients to Drug Therapy and Low Responders (Partial Responders + Nonresponders)
Responders (n = 24)
Low Responders (n = 18)
P = .03 (Fischer's exact test) for a direct comparison between pure alcoholic and pure viral etiologies.
The 50 patients included in the study were followed up for a mean of 22 months (8 days-60 months). Table 5 summarizes the episodes of rebleeding and mortality in the different groups during follow-up. The overall incidence of rebleeding in all 50 patients was 22%, whereas it was only 12% in patients whose hemodynamic response was assessed. All rebleeding episodes were secondary to variceal bleeding. In the responder group treated with drugs, only 3 patients rebled; 2 of these were patients with alcoholism who had resumed alcohol intake, both of whom died during the bleeding episode. The other patient rebled when drug therapy had to be stopped because of a minor surgical procedure. In addition, 2 patients in the partial responder group treated with banding plus drugs showed variceal rebleeding before the process of variceal eradication had been completed. No rebleeding was observed in the TIPS group. Rebleeding and mortality were not different between responder patients on drug therapy and those in the low-response group (partial responders and nonresponders).
Table 5. Rebleeding and Mortality During Follow-Up in All Patients in the Study, in Patients Who Did and Did Not Undergo the Second Hemodynamic Measurement, and in Patients Distributed According to Different Therapeutic Groups
Side effects related to drug therapy were observed in 10 patients (20%). Five patients reported headaches of short duration, which were attributed to nitrates; 1 patient needed a dose reduction to 40 mg/day. Three patients had dizziness and hypotension, 2 of whom had their nadolol dosing was reduced. One patient reported fatigue and another reported impotence that also resolved after the nadolol dosage was decreased. Only one (10%) of the 10 patients treated with banding ligation presented a transient episode of dysphagia as a side effect of the therapy. Among the group of 8 patients who underwent a TIPS insertion, 3 (37%) had at least one episode of hepatic encephalopathy during follow-up, which resolved with medical therapy in 2 cases and with placement of a coaxial stent in the other. Finally, another 3 patients needed angiographic interventions during follow-up. A complete occlusion was detected at 12 months in 1 patient with an uncovered TIPS, and a second covered TIPS was placed. In 2 patients balloon dilatation of the stent was carried out because stenosis of the prosthesis was observed during routine angiographic controls.
The results of the present trial indicate that our strategy of using the hemodynamic response to pharmacological therapy to guide therapy in secondary prophylaxis of variceal bleeding is feasible and effective in protecting patients from variceal rebleeding. In addition, our data seem to suggest that patients with cirrhosis of viral etiology have more early rebleeding and a worse response to drug therapy than do patients with cirrhosis of alcoholic etiology.
In contrast to previous reports,4, 5 this is the first reported trial of secondary prophylaxis with hemodynamic response–guided therapy that included a substantial number of patients and that optimized therapeutic strategy. In our study, β-blockers were administered at the maximum-tolerated doses without using the criterion of reducing the resting heart rate by 25%. In fact, recent evidence clearly suggests the dose of β-blockers is an independent predictor of hemodynamic response.7 This probably explains the high rate of hemodynamic responders in our study (57%). Another important characteristic of the present report is the short interval between portal hemodynamic evaluations (15 days). Most studies that have evaluated hemodynamic response to pharmacological therapy have performed the second hemodynamic study 2-6 months after the basal study.5–8 As nonresponders to drugs tend to rebleed early, many patients rebleed before their hemodynamic response is evaluated. In the study by Patch et al.5 more than half the patients in their trial could not undergo the second hemodynamic evaluation for this reason. Because of this, early hemodynamic measurements have been strongly recommended.3 Although nobody knows when is early enough and when drugs have reached a stable and significant effect on portal pressure, our experience of waiting only 5-7 days after reaching the maximum dose of β-blockers and nitrates to evaluate the hemodynamic response seems adequate. Finally, instead of rescuing patients who did achieve the target reductions in HVPG (≤20 of basal) with only a single procedure, we decided to further stratify the risk of rebleeding. Obviously, this could be questioned. However, our decision was based on two facts. We were afraid banding would not provide enough protection to nonresponders, as suggested by the very limited experience of the trial of Bureau et al.4 In addition, other studies seemed to indicate that the rebleeding risk varied according to the degree of decrease in the HVPG after drug therapy.9
This was not a controlled trial, and direct comparison with other currently recommended strategies for secondary prophylaxis cannot be done. However, our 22% rebleeding rate with a long follow-up (22 months) is better than that of any other published study of a single therapy (drugs or banding) in this setting.3, 5, 10–12 Only a TIPS insertion as the initial prophylactic treatment provides superior protection.13 By contrast, more recent studies of combination of nadolol plus banding ligation14, 15 have shown similar or lower recurrence rates than our trial. The report of de la Peña et al.15 with a 14% rebleeding rate had however, a much shorter follow-up (16 months). On the other hand, patients who were able to have their hemodynamic response assessed in our study presented an impressive 12% rebleeding rate, without relevant differences across the 3 therapeutic arms, indicating the high efficacy of the strategy utilized. However, the results of the partial response group (banding plus drugs) are difficult to interpret, given the small number of patients included. Larger series are needed to evaluate the efficacy of ligation as rescue therapy for low responders.
Another finding of our study is the difference in outcome depending on the etiology of the cirrhosis. More patients with cirrhosis of viral etiology only in contrast to patients with cirrhosis of pure alcoholic etiology had early rebleeding, occurring before the second hemodynamic evaluation, showed a worse response to drug therapy, and needed TIPS as rescue therapy. Abstinence from alcohol has been associated with hemodynamic improvement or response during follow-up.8, 16 All the patients with alcoholism in the study were undoubtedly abstinent during the 3-week period from the index variceal bleeding to the second hemodynamic study, whereas most of them were consuming alcohol at the time of the index bleeding. The quick response to drug therapy by most of the patients with alcoholism in our study suggests there is an acute and rapidly reversible effect of alcohol intake on portal pressure. This is supported by the experimental studies of animal models that demonstrated an effect of acute alcohol intake on portal pressure,17–19 as well as the studies that found that acute moderate alcohol consumption worsened portal hypertension in patients with cirrhosis.20 However, this possible better outcome in our patients with alcoholism should be interpreted with caution; overall, the numbers of patients in our subgroups were small, and the statistical relevance of the findings was probably low.
To improve and refine the therapeutic strategy we present, two important questions remain. First, how can rebleeding before assessment of the hemodynamic response be avoided? Shortening the time between HVPG measurements seems very difficult. Detecting nonresponders during the first hemodynamic study by evaluating the acute drug effect would be a reasonable approach, but this issue has not been validated. An option might be to perform a banding ligation session in the middle of this period in the patients at higher risk, that is, patients without alcoholic cirrhosis. The second question refers to the appropriateness of the strategy adopted in our study to rescue nonresponders to pharmacological therapy. It is our opinion that banding ligation as the only rescue therapy for nonresponders should be evaluated in view of the acceptable results in our subgroup of low responders. TIPS, as expected, worked very well to avoid rebleeding, but side effects and continuous surveillance were problematic. Another approach for treating nonresponders could be adding other drugs or changing to other drugs known to lower portal pressure. However, this would eventually require a third hemodynamic evaluation.
Current strategies for secondary prophylaxis tend to combine drugs plus banding ligation.1 With this approach hemodynamic studies are avoided, but half of patients receive endoscopic therapy they may not need. By contrast, with the hemodynamic response–guided therapy approach, treatment is individually tailored, but at least 2 HVPG measurements are needed. The final rebleeding rates might not be very different, but controlled trials comparing these two approaches are needed. A third therapeutic alternative derived from our results could be to treat patients with alcoholism only with drugs and patients with viral cirrhosis with a combination of drugs and banding. In conclusion, the results of our trial indicate our strategy of using the hemodynamic response to pharmacological treatment to guide therapy in secondary prophylaxis for variceal bleeding is feasible and effective in protecting patients from variceal rebleeding.