GASTROENTEROLOGY
1990;99:1401-1407
Hemodynamic Events in a Prospective Randomized Trial of Propranolol Versus Placebo in the Prevention of a First Variceal Hemorrhage ROBERTO J. GROSZMANN, JAIME BOSCH, NORMAN D. GRACE, HAROLD 0. CONN, GUADALUPE GARCIA-TSAO, MIGUEL NAVASA, JEANNE ALBERTS, JUAN RODES, ROSEMARIE FISCHER, MAX BERMANN, STEPHEN ROFE, MICHAEL PATRICK, and EMANUEL LERNER Veterans Administration Medical Center, West Haven, Connecticut: Yale University School of Medicine, New Haven, Connecticut; Hospital Clinic i Provincial, Barcelona, Spain; and Faulkner and Lemuel Shattuck Hospital, Boston, Massachusetts
In a double-blind randomized trial, the hemodynamic events following the administration of propran0101(n = 51) or a placebo (n = 51) were prospectively studied in cirrhotic patients with esophageal varices. The hepatic venous pressure gradient, heart rate, and variceal size were determined at the baseline and 3, 12, and 24 months after the beginning of therapy. Baseline values were similar in both groups. At 3 months, the hepatic venous pressure gradient decreased significantly in propranolol-treated patients (from 18.1 * 4.2 to 15.7 + 3.4 mm Hg; P -C 0.05) but not in patients receiving the placebo (19.8 t 6.8 to 17.5 + 5.3 mm Hg; NS). At subsequent time intervals this gradient decreased significantly from the baseline value in both groups. Heart rate decreased significantly in the propranolol-treated group at all times (P < 0.001). Variceal hemorrhage occurred in 13 patients (11 placebo-, 2 propranolol-treated; P < O.Ol),all of whom had a hepatic venous pressure gradient >l2 mm Hg. In 21 patients (14 propranolol-, 7 placebo-treated) the hepatic venous pressure gradient decreased to 512 mm Hg; none of them bled from esophageal varices, and their mortality rate also decreased. Because most of the bleeding events occurred during the first year (10 placebo-, 1 propranolol-treated; P < O.Ol),propranolol seems to have its protective effect during the period associated with the largest reduction in the hepatic venous pressure gradient. Because a reduction in the hepatic venous pressure gradient to t12 mm Hg protects from variceal bleeding and increases the rate of survival,
this should be the aim of the pharmacological apy of portal hypertension.
ther-
ropranolol and other nonselective fl-adrenergic blockers are being used increasingly in the prevention of variceal hemorrhage. Although their efficacy in the prevention of rebleeding from gastroesophageal varices is still controversial, recent studies suggest that these agents are effective in preventing a first variceal hemorrhage (l-4). The mechanism by which /3-adrenergic blockers decrease the risk for variceal bleeding remains unclear. To investigate this question, the sequential portal and systemic hemodynamic events following the administration of propranolol or a placebo in the course of a trial for the prevention of a first variceal hemorrhage were studied in a prospective, randomized, double-blind fashion. The clinical results of this study have been reported elsewhere (5). In addition to information obtained for the propranolol-treated patients, analysis of the data for the placebo-treated patients provides a unique opportunity to assess the natural history of the hemodynamic events in this group of patients.
P
Abbreviations used in this paper: AZBF, azygos blood flow; CO, cardiac output; FHVP, free hepatic venous pressure; HR, heart rate; HVPG, hepatic venous pressure gradient; MAP, mean arterial pressure; WHVP, wedged hepatic venous pressure. 0 1990by the American Gastroenterological Association 0016-5065/90/83.00
1402
GROSZMANN
ET AL.
Methods Subjects One hundred two patients with cirrhosis and portal hypertension were included in this study. Thirty were studied in the West Haven Veterans Administration and the Yale-New Haven Medical Centers, 33 in the Faulkner and Lemuel Shattuck Hospitals in Boston, and 39 in the Hospital Clinic i Provincial in Barcelona, Spain. The diagnosis of cirrhosis was based on liver biopsy and clinical data including compatible histories, confirmatory physical examinations, and clinical tests including liver-spleen scans. There were 73 men and 29 women, with a mean age of 54.2 ? 10.2 years (mean k SD] and a range of 25-72 years. All 102 patients had esophageal varices as diagnosed by endoscopy. Varix size was graded endoscopically according to protrusion of the largest varix into the esophageal lumen at the gastroesophageal junction: grade 1, small varices (l-3 mm) only visible during a Valsalva maneuver; grade 2, small varices (1-3 mm) easily visible without a Valsalva maneuver; grade 3, moderate-sized varices (3-6 mm]; and grade 4, large varices (r6 mm) (6). None of the patients had a history of variceal hemorrhage. The severity of liver disease was graded according to the Child-Turcotte classification by assigning numerical values to the alphabetical grading, in which class A patients were those having a score of 5-8. class B, 9-11, and class C, 12-15 (7). Fifty-nine patients were class A, 35 were class B, and 8 were class C. Of the 102 patients entered in the study, 51 (39 with alcoholic cirrhosis) received propranolol and 51 (41 with alcoholic cirrhosis) received an identical placebo. Patients were followed up for a mean period of 16.3 * 12 months. Patients underwent hepatic and systemic hemodynamic studies before randomization and 3,12, and 24 months after inclusion in the trial. The study was approved by the human investigation subcommittees of the involved hospitals, and patients gave written informed consent to participate in it.
GASTROENTEROLOGY
In 33 patients studied in Barcelona, cardiac output (CO) and azygos blood flow (AZBF) were also measured at each study period. The CO was measured by thermodilution using a balloon-tipped catheter, and the AZBF was measured with a continuous thermodilution technique, as previously described, using a coronary sinus catheter (9).
Titration
Studies
In all patients the dose of propranolol (or placebo) to be used during the clinical study was determined before randomization according to the HVPG and/or HR response to the oral administration of propranolol. Propranolol was titrated to achieve either (a) a 25% reduction in HVPG, (b) a decrease in HVPG to 12 mm Hg or less, or (c) a decrease in HR to 55 beats/min or less. The titration procedure is described elsewhere [lo). Briefly, after baseline measurements were obtained, patients received 40-80 mg of propranolo1 PO; if one of the endpoints had not been met within 4 hours, propranolol was administered at a dose of 80 mg b.i.d. for 1 week, at the end of which a second hemodynamic study was performed. If an endpoint had still not been reached, patients would receive an additional 40-80 mg of propranolol. If at any time during the procedure one of the endpoints was met, the titration would be stopped and the dose would have been determined. If none of the endpoints was attained by the end of the procedure, the patient would receive the maximum dose of 160 mg b.i.d. during the study. Once the dose was determined, patients were randomly assigned to be given propranolol or a placebo. The mean dose of propranolol received by the 51 patients randomized to receive the drug was 132 * 78 mg/day [range, 40-320 mg/day). The median dose was 80 mg/day. For later analyses, patients showing a reduction in HVPG >lO% at any time during the titration period were considered “titration responders,” and patients showing a decrease in HVPG ~10% or an increase in HVPG were considered “titration nonresponders” [lo).
Statistical Hemodynamic
Vol. 99, No. 5
Analyses
Studies
Portal pressure was determined by the hepatic venous pressure gradient (HVPG), which is obtained by subtracting the free hepatic venous pressure (FHVP) from the wedged hepatic venous pressure (WHVP). An HVPG of at least 12 mm Hg was required to enter the study. The technique used at all three centers was the same. After an overnight fast, right hepatic vein catheterization was performed percutaneously through the femoral vein, and pressure was recorded in both the wedged (occluded) and the free position using a balloon catheter [S). The wedged position was checked by the absence of reflux after the injection of 2 mL of contrast medium. Pressures were measured using a previously calibrated strain gauge transducer and recorded on paper. Each pressure reading was recorded at least in duplicate, and for each an electronic mean was obtained. Arterial pressure [with an external sphygmomanometer) and heart rate (HR) were also determined at each study period. Mean arterial pressure [MAP) was calculated using the formula MAP = [systolic + (diastolic x 2)]/3.
Statistical comparisons were performed using Student’s t test for paired and unpaired data, with Bonferroni corrections for multiple comparisons when needed Ill). Intergroup comparisons for each time interval were performed for each hemodynamic study. Within each group, the baseline hemodynamic study was compared with the 3-, 12-, and 24-month studies. Correlations were performed by linear regression. Percentage comparisons were made using x2 and Fisher’s exact test. Cumulative proportions were analyzed with the Mantel-Cox and Wilcoxon tests (12). Results are expressed as mean + SD except when otherwise stated. Results As has been reported in the clinical portion of the study (51, a beneficial effect of propranolol in the prevention of the first variceal hemorrhage was observed. Only 2 of 51 (4%) patients taking propranolol
November
PREVENTION
1990
had a variceal hemorrhage during the study period, whereas 11 of 51 (22%) patients taking placebo bled from varices (P < 0.001). The majority of bleeding episodes (11/13) occurred in the first year of the study. Baseline data for the 102 patients studied are shown in Table 1. Hemodynamic parameters were similar in both the placebo and propranolol groups. There were no differences in the mean baseline HVPG values of the patients entered in New Haven (18.3 f 7.2 mm Hg), Boston (19.2 + 4.4 mm Hg), and Barcelona (18.2 + 5.4 mm Hg). Hepatic Venous Pressure Gradient to Propranolol or Placebo
Response
Of the 102 patients entered into the study, 84 (45 propranolol, 39 placebo) underwent repeat HVPG measurements at 3 months, 61 (35 propranolol, 26 placebo) underwent repeat measurements at 3 and 12 months, and 21(12 propranolol, 9 placebo) underwent repeat HVPG measurements at 3, 12, and 24 months after randomization. Patients did not undergo repeat measurements because of terminating events (e.g., variceal hemorrhage, death, concomitant disease, side effects, noncompliance) or because of completion of the study. As shown in Table 2, compared with the baseline values, 3-month HVPG values decreased significantly in the propranolol group but not in the placebo group. However, at 12 and 24 months, patients in both groups showed significantly decreased HVPG values compared with the baseline values, and there were no differences in HVPG between the propranolol-treated and the placebo-treated groups. When the HVPG values of patients that dropped out of the study are Table 1. Baseline
Data of the 102 Patients
No. of Patients
NOTE. Numbers in parentheses different from 51.
Propranolol
51 54 + 11 35 : 16
51 54 k 9 38 : 13
41: 10 8.3 + 1.9 2.45 + 0.78 4 25 17 5 19.6 k 6.8 26.3 + 7.3 6.7 + 3.5 81 f 12 92 + 11 8.1 f 2.4 (16) 650 f 296 (17)
represent
1403
Table 2. Hepatic Venous Pressure Gradient Values (in mm Hg] at Different Time Periods for Placebo- and Propranolol-Treated Patients Baseline
3mo
12 mo
24 mo
Placebo
19.6 + 6.8
17.5 * 5.3”
15.0 + 4.5b
10.6 t 4.0h
Propranolol
(51) 18.1 t 4.2
(391 15.7 f 3.4c
15.1 + 3.8”
12.9 + 4.3”
(51)
(45)
(351
(121
1261
(91
NOTE. Differences between placebo- and propranolol-treated patients were nonsignificant at all times. Numbers in parentheses represent no. of patients. “P = NS vs. baseline. ‘P < 0.01 vs. baseline. “P -c 0.05 vs. baseline.
analyzed, as can be seen in Table 3, the mean HVPG values of the 13 patients in the placebo group who did not undergo repeat measurements after 3 months were significantly higher than the mean HVPG values of the 10 patients taking propranolol who left the study between 3 and 12 months (20.9 f 5.8 vs. 16.0 k 4.9; P -C0.05). This difference may explain the significant decrease in HVPG shown by the placebo group after 12 months. After 24 months, both study groups had significantly lower HVPG values compared with the baseline values. Unfortunately, the number of patients with repeat measurements 24 months after inclusion in the study is too small for meaningful analysis. It is important to emphasize that reductions in HVPG during the study were due to decreases in the WHVP, and that FHVP did not change significantly. Table 4 shows the sequential HVPG measurements in both groups when only patients undergoing repeat measurements at the different time intervals were considered (paired comparisons). The HVPG values
Studied
Placebo
Age (yrl Sex (M:F] Etiology of cirrhosis (alcoholic: nonalcoholic] Children’s score Variceal size No. grade 1 No. grade 2 No. grade 3 No. grade 4 HVPG (mm Hg) WHVP (mm Hg) FHVP (mm Hg] HR (beats/min] MAP (mm Hg] CO (L/min] AZBF (L/min]
OF FIRST VARICEAL HEMORRHAGE
39: 12 8.0 + 2.2 2.52 k 0.74 4 22 21 4 18.1 2 4.2 24.4 + 5.9 6.3 ? 3.5 80 e 14 94 + 11 8.2 k 1.4 (16) 620 + 210 (16)
P NS NS NS NS NS NS
Table 3. Hepatic Venous Pressure Gradient (in mm HgJ of Patients Remaining in the Study and Leaving It at Different Time Periods Propranolol
Placebo Left the study Baseline
18.1 + 4.2
19.6 L 6.8
(511
(511 21.3 + 10.8 3mo
l17.5 * 5.3 (391
NS NS NS NS NS NS NS
no. of patients when
12 mo
NS -
17.6 f 4.2 (61
(121
I 15.7 k 3.4 (451
20.9 + 5.8 -
r 15.0 + 4.5
10.05 -
16.0 * 4.9 (101
(131
+-I 15.1 + 3.8 (351
(26) 16.1 + 4.5 24 mo
t10.6 t 4.0
(17)
NS -
15.9 + 3.3 (231
(321
(91
NOTE. Numbers in parentheses
I 12.9 t 4.3
represent no. of patients.
1404
GROSZMANN
Table 4.
GASTROENTEROLOGY
ET AL.
Vol. 99, No. 5
Sequential Changes in Hepatic Venous Pressure Gradient [in mm Hg] During the Course of the Study Baseline
3mo
Placebo
19.1 f 5.1(39)
Propranolol
18.2 + 4.2 (45)
17.5 + 5.2 (39) 15.7 + 3.4 (45)
Placebo Propranolol
16.1 ? 4.0 126) 15.4 + 3.1(35)
Placebo Prooranolol
12mo
24 mo
P NS 10.05
15.0 + 4.5 (26) 15.1 + 3.8 (35) 12.8 k 3.8 (9) 13.5 f 4.2 (121
NS NS 10.6 k 4.0 [9] 12.9 + 4.3 112)
NS NS
NOTE. Numbers in parentheses represent no. of patients. “Paired t tests with Bonferroni correction.
decreased significantly at 3 months (compared with baseline) only in the propranolol-treated group. In the placebo-treated group (n = 39 patients with repeat measurements], the changes in HVPG values seemed to have a relationship with the improvement or worsening of liver disease. Thus, of 19 patients taking a placebo in whom HVPG decreased ~10% in the course of the study, 16 (84%) showed an improvement in Child score (decrease of 1 point or greater), whereas of 20 patients in whom HVPG values remained the same (tlO% decrease) or increased, only 5 (25%) showed an improvement in Child score (P < 0.01). This relationship was not observed in the propranolol-treated group (n = 45 patients with repeat measurements], in which Child score improved in 12 of 29 (41%) patients in whom HVPG decreased >lO% and in 9 of 16 (56%) patients in whom HVPG values did not change or increased, a difference that is not significant statistically. Hepatic Venous Pressure Gradient and Variceal Hemorrhage During the study, 13 patients (11 placebo, 2 propranolol] bled from varices, and three patients (placebo-treated group] bled from congestive gastropathy. All 16 patients had an HVPG >12 mm Hg. None of 21 patients showing HVPG values that decreased to 512 mm Hg during the study (14 propranolol, 7 placebo) bled from varices or portal hypertensive gastropathy. None of the hemodynamic parameters measured at baseline or during the course of the study differed between patients who bled from varices and those who did not. Hepatic Venous Pressure Gradient and Variceal Size The HVPG had a tendency to be higher in Fatients with large varices (19.9 f 6.6 mm Hg) than in patients with small varices (17.9 f 4.6 mm Hg) (P < 0.05, 1 tail).
No relationship was found between changes in HVPG and changes in variceal size during the course of the study. However, a significantly larger reduction in variceal size was observed in the group of patients in whom HVPG decreased to 512 mm Hg than in patients that did not show such a decrease (Table 5). Hepatic Venous Pressure Gradient Response and its Relationship to Titration Studies In the propranolol-treated group, a reduction in HVPG >lO% during titration (titration responders) made it possible to predict the HVPG response to propranolol during the study. Whereas 74% of titration responders showed a decrease in HVPG >lO% at 3 months, only 21% of titration nonresponders showed an equivalent response (P < 0.001). Twenty patients had decreases in HVPG to levels t12 mm Hg during titration. Of these, 11 were randomized to receive placebo and 9 to receive propranolol. Of the 9 patients taking propranolol, 5 showed a decrease in HVPG to levels tl2 mm Hg during the study, and of the remaining 4 patients, 2 showed unchanged HVPG, 1 showed an HVPG that decreased to 12.5 mm Hg, and the remaining patient had an increase in HVPG. Titration studies did not predict variceal hemorrhage because of the two patients taking propranolol who bled from varices, one was a titration responder and the other a nonresponder. Other Hemodynamic
Measurements
Heart rate. In the propranolol-treated group, HR decreased significantly throughout the study compared with baseline values and to HR values in the placebo-treated patients (Figure 11. Arterial Pressure. No significant changes in MAP were observed in either group. Cardiac output. Cardiac output decreased significantly at 3 and 12 months in patients taking propranolol. The number of patients with measurements at 24 months is too small for analysis (Figure 2).
November
PREVENTION
1990
OF FIRST VARICEAL HEMORRHAGE
1405
Table 5. Comparison of Patients Who During the Course of a Prospective Trial Showed a Decrease in Hepatic Venous Pressure Gradient to 12 mm Hg or Less With Those Who Did Not Show Such a Decrease HVPG t12 mm Hg (n = 21)
HVPG >12 mm Hg
(n = 63)
P
52 f 9 2.14 + 0.8 7.9 2 2.4 20/21(95%) 90 + 8 17.1 + 2.7
56 k 10 2.44 + 0.74 8.1 k 2.0 44/63 (70%) 94 * 12 19.0 f 5.1
NS NS NS
1990;99:1401-1407
Hemodynamic Events in a Prospective Randomized Trial of Propranolol Versus Placebo in the Prevention of a First Variceal Hemorrhage ROBERTO J. GROSZMANN, JAIME BOSCH, NORMAN D. GRACE, HAROLD 0. CONN, GUADALUPE GARCIA-TSAO, MIGUEL NAVASA, JEANNE ALBERTS, JUAN RODES, ROSEMARIE FISCHER, MAX BERMANN, STEPHEN ROFE, MICHAEL PATRICK, and EMANUEL LERNER Veterans Administration Medical Center, West Haven, Connecticut: Yale University School of Medicine, New Haven, Connecticut; Hospital Clinic i Provincial, Barcelona, Spain; and Faulkner and Lemuel Shattuck Hospital, Boston, Massachusetts
In a double-blind randomized trial, the hemodynamic events following the administration of propran0101(n = 51) or a placebo (n = 51) were prospectively studied in cirrhotic patients with esophageal varices. The hepatic venous pressure gradient, heart rate, and variceal size were determined at the baseline and 3, 12, and 24 months after the beginning of therapy. Baseline values were similar in both groups. At 3 months, the hepatic venous pressure gradient decreased significantly in propranolol-treated patients (from 18.1 * 4.2 to 15.7 + 3.4 mm Hg; P -C 0.05) but not in patients receiving the placebo (19.8 t 6.8 to 17.5 + 5.3 mm Hg; NS). At subsequent time intervals this gradient decreased significantly from the baseline value in both groups. Heart rate decreased significantly in the propranolol-treated group at all times (P < 0.001). Variceal hemorrhage occurred in 13 patients (11 placebo-, 2 propranolol-treated; P < O.Ol),all of whom had a hepatic venous pressure gradient >l2 mm Hg. In 21 patients (14 propranolol-, 7 placebo-treated) the hepatic venous pressure gradient decreased to 512 mm Hg; none of them bled from esophageal varices, and their mortality rate also decreased. Because most of the bleeding events occurred during the first year (10 placebo-, 1 propranolol-treated; P < O.Ol),propranolol seems to have its protective effect during the period associated with the largest reduction in the hepatic venous pressure gradient. Because a reduction in the hepatic venous pressure gradient to t12 mm Hg protects from variceal bleeding and increases the rate of survival,
this should be the aim of the pharmacological apy of portal hypertension.
ther-
ropranolol and other nonselective fl-adrenergic blockers are being used increasingly in the prevention of variceal hemorrhage. Although their efficacy in the prevention of rebleeding from gastroesophageal varices is still controversial, recent studies suggest that these agents are effective in preventing a first variceal hemorrhage (l-4). The mechanism by which /3-adrenergic blockers decrease the risk for variceal bleeding remains unclear. To investigate this question, the sequential portal and systemic hemodynamic events following the administration of propranolol or a placebo in the course of a trial for the prevention of a first variceal hemorrhage were studied in a prospective, randomized, double-blind fashion. The clinical results of this study have been reported elsewhere (5). In addition to information obtained for the propranolol-treated patients, analysis of the data for the placebo-treated patients provides a unique opportunity to assess the natural history of the hemodynamic events in this group of patients.
P
Abbreviations used in this paper: AZBF, azygos blood flow; CO, cardiac output; FHVP, free hepatic venous pressure; HR, heart rate; HVPG, hepatic venous pressure gradient; MAP, mean arterial pressure; WHVP, wedged hepatic venous pressure. 0 1990by the American Gastroenterological Association 0016-5065/90/83.00
1402
GROSZMANN
ET AL.
Methods Subjects One hundred two patients with cirrhosis and portal hypertension were included in this study. Thirty were studied in the West Haven Veterans Administration and the Yale-New Haven Medical Centers, 33 in the Faulkner and Lemuel Shattuck Hospitals in Boston, and 39 in the Hospital Clinic i Provincial in Barcelona, Spain. The diagnosis of cirrhosis was based on liver biopsy and clinical data including compatible histories, confirmatory physical examinations, and clinical tests including liver-spleen scans. There were 73 men and 29 women, with a mean age of 54.2 ? 10.2 years (mean k SD] and a range of 25-72 years. All 102 patients had esophageal varices as diagnosed by endoscopy. Varix size was graded endoscopically according to protrusion of the largest varix into the esophageal lumen at the gastroesophageal junction: grade 1, small varices (l-3 mm) only visible during a Valsalva maneuver; grade 2, small varices (1-3 mm) easily visible without a Valsalva maneuver; grade 3, moderate-sized varices (3-6 mm]; and grade 4, large varices (r6 mm) (6). None of the patients had a history of variceal hemorrhage. The severity of liver disease was graded according to the Child-Turcotte classification by assigning numerical values to the alphabetical grading, in which class A patients were those having a score of 5-8. class B, 9-11, and class C, 12-15 (7). Fifty-nine patients were class A, 35 were class B, and 8 were class C. Of the 102 patients entered in the study, 51 (39 with alcoholic cirrhosis) received propranolol and 51 (41 with alcoholic cirrhosis) received an identical placebo. Patients were followed up for a mean period of 16.3 * 12 months. Patients underwent hepatic and systemic hemodynamic studies before randomization and 3,12, and 24 months after inclusion in the trial. The study was approved by the human investigation subcommittees of the involved hospitals, and patients gave written informed consent to participate in it.
GASTROENTEROLOGY
In 33 patients studied in Barcelona, cardiac output (CO) and azygos blood flow (AZBF) were also measured at each study period. The CO was measured by thermodilution using a balloon-tipped catheter, and the AZBF was measured with a continuous thermodilution technique, as previously described, using a coronary sinus catheter (9).
Titration
Studies
In all patients the dose of propranolol (or placebo) to be used during the clinical study was determined before randomization according to the HVPG and/or HR response to the oral administration of propranolol. Propranolol was titrated to achieve either (a) a 25% reduction in HVPG, (b) a decrease in HVPG to 12 mm Hg or less, or (c) a decrease in HR to 55 beats/min or less. The titration procedure is described elsewhere [lo). Briefly, after baseline measurements were obtained, patients received 40-80 mg of propranolo1 PO; if one of the endpoints had not been met within 4 hours, propranolol was administered at a dose of 80 mg b.i.d. for 1 week, at the end of which a second hemodynamic study was performed. If an endpoint had still not been reached, patients would receive an additional 40-80 mg of propranolol. If at any time during the procedure one of the endpoints was met, the titration would be stopped and the dose would have been determined. If none of the endpoints was attained by the end of the procedure, the patient would receive the maximum dose of 160 mg b.i.d. during the study. Once the dose was determined, patients were randomly assigned to be given propranolol or a placebo. The mean dose of propranolol received by the 51 patients randomized to receive the drug was 132 * 78 mg/day [range, 40-320 mg/day). The median dose was 80 mg/day. For later analyses, patients showing a reduction in HVPG >lO% at any time during the titration period were considered “titration responders,” and patients showing a decrease in HVPG ~10% or an increase in HVPG were considered “titration nonresponders” [lo).
Statistical Hemodynamic
Vol. 99, No. 5
Analyses
Studies
Portal pressure was determined by the hepatic venous pressure gradient (HVPG), which is obtained by subtracting the free hepatic venous pressure (FHVP) from the wedged hepatic venous pressure (WHVP). An HVPG of at least 12 mm Hg was required to enter the study. The technique used at all three centers was the same. After an overnight fast, right hepatic vein catheterization was performed percutaneously through the femoral vein, and pressure was recorded in both the wedged (occluded) and the free position using a balloon catheter [S). The wedged position was checked by the absence of reflux after the injection of 2 mL of contrast medium. Pressures were measured using a previously calibrated strain gauge transducer and recorded on paper. Each pressure reading was recorded at least in duplicate, and for each an electronic mean was obtained. Arterial pressure [with an external sphygmomanometer) and heart rate (HR) were also determined at each study period. Mean arterial pressure [MAP) was calculated using the formula MAP = [systolic + (diastolic x 2)]/3.
Statistical comparisons were performed using Student’s t test for paired and unpaired data, with Bonferroni corrections for multiple comparisons when needed Ill). Intergroup comparisons for each time interval were performed for each hemodynamic study. Within each group, the baseline hemodynamic study was compared with the 3-, 12-, and 24-month studies. Correlations were performed by linear regression. Percentage comparisons were made using x2 and Fisher’s exact test. Cumulative proportions were analyzed with the Mantel-Cox and Wilcoxon tests (12). Results are expressed as mean + SD except when otherwise stated. Results As has been reported in the clinical portion of the study (51, a beneficial effect of propranolol in the prevention of the first variceal hemorrhage was observed. Only 2 of 51 (4%) patients taking propranolol
November
PREVENTION
1990
had a variceal hemorrhage during the study period, whereas 11 of 51 (22%) patients taking placebo bled from varices (P < 0.001). The majority of bleeding episodes (11/13) occurred in the first year of the study. Baseline data for the 102 patients studied are shown in Table 1. Hemodynamic parameters were similar in both the placebo and propranolol groups. There were no differences in the mean baseline HVPG values of the patients entered in New Haven (18.3 f 7.2 mm Hg), Boston (19.2 + 4.4 mm Hg), and Barcelona (18.2 + 5.4 mm Hg). Hepatic Venous Pressure Gradient to Propranolol or Placebo
Response
Of the 102 patients entered into the study, 84 (45 propranolol, 39 placebo) underwent repeat HVPG measurements at 3 months, 61 (35 propranolol, 26 placebo) underwent repeat measurements at 3 and 12 months, and 21(12 propranolol, 9 placebo) underwent repeat HVPG measurements at 3, 12, and 24 months after randomization. Patients did not undergo repeat measurements because of terminating events (e.g., variceal hemorrhage, death, concomitant disease, side effects, noncompliance) or because of completion of the study. As shown in Table 2, compared with the baseline values, 3-month HVPG values decreased significantly in the propranolol group but not in the placebo group. However, at 12 and 24 months, patients in both groups showed significantly decreased HVPG values compared with the baseline values, and there were no differences in HVPG between the propranolol-treated and the placebo-treated groups. When the HVPG values of patients that dropped out of the study are Table 1. Baseline
Data of the 102 Patients
No. of Patients
NOTE. Numbers in parentheses different from 51.
Propranolol
51 54 + 11 35 : 16
51 54 k 9 38 : 13
41: 10 8.3 + 1.9 2.45 + 0.78 4 25 17 5 19.6 k 6.8 26.3 + 7.3 6.7 + 3.5 81 f 12 92 + 11 8.1 f 2.4 (16) 650 f 296 (17)
represent
1403
Table 2. Hepatic Venous Pressure Gradient Values (in mm Hg] at Different Time Periods for Placebo- and Propranolol-Treated Patients Baseline
3mo
12 mo
24 mo
Placebo
19.6 + 6.8
17.5 * 5.3”
15.0 + 4.5b
10.6 t 4.0h
Propranolol
(51) 18.1 t 4.2
(391 15.7 f 3.4c
15.1 + 3.8”
12.9 + 4.3”
(51)
(45)
(351
(121
1261
(91
NOTE. Differences between placebo- and propranolol-treated patients were nonsignificant at all times. Numbers in parentheses represent no. of patients. “P = NS vs. baseline. ‘P < 0.01 vs. baseline. “P -c 0.05 vs. baseline.
analyzed, as can be seen in Table 3, the mean HVPG values of the 13 patients in the placebo group who did not undergo repeat measurements after 3 months were significantly higher than the mean HVPG values of the 10 patients taking propranolol who left the study between 3 and 12 months (20.9 f 5.8 vs. 16.0 k 4.9; P -C0.05). This difference may explain the significant decrease in HVPG shown by the placebo group after 12 months. After 24 months, both study groups had significantly lower HVPG values compared with the baseline values. Unfortunately, the number of patients with repeat measurements 24 months after inclusion in the study is too small for meaningful analysis. It is important to emphasize that reductions in HVPG during the study were due to decreases in the WHVP, and that FHVP did not change significantly. Table 4 shows the sequential HVPG measurements in both groups when only patients undergoing repeat measurements at the different time intervals were considered (paired comparisons). The HVPG values
Studied
Placebo
Age (yrl Sex (M:F] Etiology of cirrhosis (alcoholic: nonalcoholic] Children’s score Variceal size No. grade 1 No. grade 2 No. grade 3 No. grade 4 HVPG (mm Hg) WHVP (mm Hg) FHVP (mm Hg] HR (beats/min] MAP (mm Hg] CO (L/min] AZBF (L/min]
OF FIRST VARICEAL HEMORRHAGE
39: 12 8.0 + 2.2 2.52 k 0.74 4 22 21 4 18.1 2 4.2 24.4 + 5.9 6.3 ? 3.5 80 e 14 94 + 11 8.2 k 1.4 (16) 620 + 210 (16)
P NS NS NS NS NS NS
Table 3. Hepatic Venous Pressure Gradient (in mm HgJ of Patients Remaining in the Study and Leaving It at Different Time Periods Propranolol
Placebo Left the study Baseline
18.1 + 4.2
19.6 L 6.8
(511
(511 21.3 + 10.8 3mo
l17.5 * 5.3 (391
NS NS NS NS NS NS NS
no. of patients when
12 mo
NS -
17.6 f 4.2 (61
(121
I 15.7 k 3.4 (451
20.9 + 5.8 -
r 15.0 + 4.5
10.05 -
16.0 * 4.9 (101
(131
+-I 15.1 + 3.8 (351
(26) 16.1 + 4.5 24 mo
t10.6 t 4.0
(17)
NS -
15.9 + 3.3 (231
(321
(91
NOTE. Numbers in parentheses
I 12.9 t 4.3
represent no. of patients.
1404
GROSZMANN
Table 4.
GASTROENTEROLOGY
ET AL.
Vol. 99, No. 5
Sequential Changes in Hepatic Venous Pressure Gradient [in mm Hg] During the Course of the Study Baseline
3mo
Placebo
19.1 f 5.1(39)
Propranolol
18.2 + 4.2 (45)
17.5 + 5.2 (39) 15.7 + 3.4 (45)
Placebo Propranolol
16.1 ? 4.0 126) 15.4 + 3.1(35)
Placebo Prooranolol
12mo
24 mo
P NS 10.05
15.0 + 4.5 (26) 15.1 + 3.8 (35) 12.8 k 3.8 (9) 13.5 f 4.2 (121
NS NS 10.6 k 4.0 [9] 12.9 + 4.3 112)
NS NS
NOTE. Numbers in parentheses represent no. of patients. “Paired t tests with Bonferroni correction.
decreased significantly at 3 months (compared with baseline) only in the propranolol-treated group. In the placebo-treated group (n = 39 patients with repeat measurements], the changes in HVPG values seemed to have a relationship with the improvement or worsening of liver disease. Thus, of 19 patients taking a placebo in whom HVPG decreased ~10% in the course of the study, 16 (84%) showed an improvement in Child score (decrease of 1 point or greater), whereas of 20 patients in whom HVPG values remained the same (tlO% decrease) or increased, only 5 (25%) showed an improvement in Child score (P < 0.01). This relationship was not observed in the propranolol-treated group (n = 45 patients with repeat measurements], in which Child score improved in 12 of 29 (41%) patients in whom HVPG decreased >lO% and in 9 of 16 (56%) patients in whom HVPG values did not change or increased, a difference that is not significant statistically. Hepatic Venous Pressure Gradient and Variceal Hemorrhage During the study, 13 patients (11 placebo, 2 propranolol] bled from varices, and three patients (placebo-treated group] bled from congestive gastropathy. All 16 patients had an HVPG >12 mm Hg. None of 21 patients showing HVPG values that decreased to 512 mm Hg during the study (14 propranolol, 7 placebo) bled from varices or portal hypertensive gastropathy. None of the hemodynamic parameters measured at baseline or during the course of the study differed between patients who bled from varices and those who did not. Hepatic Venous Pressure Gradient and Variceal Size The HVPG had a tendency to be higher in Fatients with large varices (19.9 f 6.6 mm Hg) than in patients with small varices (17.9 f 4.6 mm Hg) (P < 0.05, 1 tail).
No relationship was found between changes in HVPG and changes in variceal size during the course of the study. However, a significantly larger reduction in variceal size was observed in the group of patients in whom HVPG decreased to 512 mm Hg than in patients that did not show such a decrease (Table 5). Hepatic Venous Pressure Gradient Response and its Relationship to Titration Studies In the propranolol-treated group, a reduction in HVPG >lO% during titration (titration responders) made it possible to predict the HVPG response to propranolol during the study. Whereas 74% of titration responders showed a decrease in HVPG >lO% at 3 months, only 21% of titration nonresponders showed an equivalent response (P < 0.001). Twenty patients had decreases in HVPG to levels t12 mm Hg during titration. Of these, 11 were randomized to receive placebo and 9 to receive propranolol. Of the 9 patients taking propranolol, 5 showed a decrease in HVPG to levels tl2 mm Hg during the study, and of the remaining 4 patients, 2 showed unchanged HVPG, 1 showed an HVPG that decreased to 12.5 mm Hg, and the remaining patient had an increase in HVPG. Titration studies did not predict variceal hemorrhage because of the two patients taking propranolol who bled from varices, one was a titration responder and the other a nonresponder. Other Hemodynamic
Measurements
Heart rate. In the propranolol-treated group, HR decreased significantly throughout the study compared with baseline values and to HR values in the placebo-treated patients (Figure 11. Arterial Pressure. No significant changes in MAP were observed in either group. Cardiac output. Cardiac output decreased significantly at 3 and 12 months in patients taking propranolol. The number of patients with measurements at 24 months is too small for analysis (Figure 2).
November
PREVENTION
1990
OF FIRST VARICEAL HEMORRHAGE
1405
Table 5. Comparison of Patients Who During the Course of a Prospective Trial Showed a Decrease in Hepatic Venous Pressure Gradient to 12 mm Hg or Less With Those Who Did Not Show Such a Decrease HVPG t12 mm Hg (n = 21)
HVPG >12 mm Hg
(n = 63)
P
52 f 9 2.14 + 0.8 7.9 2 2.4 20/21(95%) 90 + 8 17.1 + 2.7
56 k 10 2.44 + 0.74 8.1 k 2.0 44/63 (70%) 94 * 12 19.0 f 5.1
NS NS NS
