Prognostic usefulness of hepatic vein catheterization in patients with cirrhosis and esophageal varices

GASTROENTEROLOGY

1992:102:973-979

Prognostic Usefulness of Hepatic Vein Catheterization in Patients With Cirrhosis and Esophageal Varices CARLO MERKEL, MASSIMO BOLOGNESI, STEFANO BELLON, RENZO ZUIN, FRANC0 NOVENTA, GIANFRANCO FINUCCI, DAVID SACERDOTI, PAOLO ANGELI, and ANGELO GATTA Department of Clinical Medicine, University of Padua, Padua, Italy

Clinical and anamnestic data, Pugh score, and size of esophageal varices were obtained in 129 cirrhotics. Hepatic vein catheterization was performed to measure hepatic venous pressure gradient (HVPG), indocyanine green (ICG) intrinsic hepatic clearance, and hepatic plasma flow. During a follow-up period of up to 60 months, 44 patients experienced gastrointestinal bleeding and 54 died. Applying Cox regression analysis, ICG intrinsic hepatic clearance, Pugh score, previous variceal bleeding, and HVPG were the only significant prognostic determinants of survival. In addition, Cox’s regression analysis showed that HVPG, Pugh score, size of varices, and previous variceal bleeding all contained significant prognostic information regarding risk of gastrointestinal bleeding. The models were validated using a split-sample technique, and prognostic indexes for death and gastrointestinal bleeding were calculated. The prognostic index predicting death had significantly improved prognostic accuracy over a prognostic index calculated excluding the data obtained from hepatic vein catheterization (P < 0.05). In conclusion, prognostic accuracy in cirrhosis with portal hypertension is significantly improved by information obtained from hepatic vein catheterization.

of the measurement of hepatic venous pressure gradient (HVPG) is also uncertainll because a significant prognostic value was observed only in some series.‘2s*3Recently, changes in HVPG after administration of P-blocking agents were found to be predictive of success of treatment in patients withI and without15 previous variceal bleeding. Nevertheless, additional parameters related to liver function and hemodynamics, such as hepatic plasma flow (HPF) and the intrinsic hepatic clearance (IHC) of indocyanine green (ICG),” which might be of prognostic significance, can be determined easily during hepatic vein catheterization, a procedure necessary for the measurement of HVPG. To our knowledge the overall usefulness of the data obtainable by means of hepatic vein catheterization has not previously been assessed. We undertook this study to evaluate prospectively the clinical value of HVPG, HPF, and KG IHC in predicting death and/or gastrointestinal bleeding in patients with cirrhosis and portal hypertension and to assess whether these additional measurements can provide prognostic information not available from the CTP classification and from the size of esophageal varices. Patients and Methods

I

n clinical practice, the prognosis of cirrhosis is usually assessed using the Child-Turcotte-Pugh (CTP) classification,‘*’ which mainly takes into account the severity of liver function derangement. However, a relevant proportion of patients with cirrhosis who develop portal hypertension die of gastrointestinal bleeding. 3-6 As a consequence, additional prognostic indicators of gastrointestinal bleeding and of death from bleeding have been searched for.7 The size and aspect of esophageal varices have recently been emphasized as indicators of risk of bleeding.‘*’ However, the clinical value of the size of esophageal varices in predicting death is uncertain.s*‘O The role

Study Population From October 1983 to June 1986,135 patients with liver disease admitted to the Second Medical Clinic, Department of Clinical Medicine, University of Padua, fulfilled the following entry criteria: (a) histological diagnosis of liver cirrhosis or clinical diagnosis of cirrhosis in the presence of severe clotting abnormalities precluding liver biopsy; (b) presence of esophageal varices at endoscopic examination; (c) age between 18 and 75 years; (d) absence of other diseases having a short prognosis per se; (e) ab0 1992 by the American

Gastroenterological 0016-5065/92/$3.00

Association

974 MERKEL ET AL.

sence of previous medical, endoscopic, or surgical treatment for prophylaxis of first variceal bleeding or prevention of recurrent bleeding; and (f) informed consent to the study and attitude to collaborate in it. Most patients were previously reported in a study on the value of quantitative liver function assessement,17 but that study did not address the question of the value of portal hemodynamic measurements in the prognosis of cirrhosis. The study was carried out in accordance with the Declaration of Helsinki, and approval was obtained from the competent ethics authorities. The reasons for admission of the 135 patients, were ascites and/or peripheral edema in 49, presence of esophageal varices without previous hemorrhage in 9, previous gastrointestinal hemorrhage for which patients were referred for possible treatment in 31, acute gastrointestinal bleeding in 27, and unspecific symptoms in 19. Because stable hemodynamic conditions are essential to obtain reliable values of hemodynamic measurements,‘B~‘g and conceivably also of variceal size,” in patients admitted for gastrointestinal bleeding, these determinations were performed at least 14 days after the bleeding. Therefore, 6 patients who died within 14 days of hemorrhage could not be investigated. The final set consisted of 129 patients. No patient had a known history of varices longer than 1 year. In the 129 patients, median age was 50 years [interquartile range (IQR), 43-57 years]; etiology of cirrhosis was alcoholic in 92, posthepatitic in 22, and cryptogenic in 15. Seventy-two patients had clinical evidence of ascites, and 78 had never experienced variceal bleeding. Median serum albumin level was 35 g/L (3.5 g/dL) [IQR, 32-38 g/L (3.23.8 g/dL)]; median serum bilirubin was 24 pmol/L (1.4 mg/dL) [IQR, 18-30 pmol/L (1.0-1.8 mg/dL)]; median prothrombin index was 59% (IQR, 48%-66%); and median Pugh score was 7 (IQR, 6-9). Medication at the time of the study included diuretics (40 patients), HZ-receptor antagonists (27 patients), digoxin (5 patients), and antibiotics (10 patients). Only 29 of 92 patients with alcoholic cirrhosis claimed abstainence from alcohol in the month preceding admission.

Study Protocol In all patients, the clinical and biochemical data included in the CTP classification were recorded and Pugh score was calculated.’ The size of esophageal varices at endoscopy was classified, according to the Japanese classification” as Fl in 20 patients, F2 in 55 patients, and F3 in 54 patients. Hepatic vein catheterization was performed using a balloon catheter, under fluoroscopic control, as reported elsewhere.” In one patient it was not possible to obtain a stable recording of pressure in the wedged position, and HVPG was not computed. During hepatic vein catheterization, constant infusion of ICG was performed to measure HPF and ICG IHC according to a procedure previously described.22 In three patients in whom the ICG extraction ratio [(A - V)/A] was ~0.10, HPF was not computed according to Rowell.23 The reference range obtained in our laboratory in hospitalized subjects without liver disease was 3-5 mm Hg for HVPG, 370-1060 mL/min for

GASTROENTEROLOGY Vol. 102, No. 3

HPF, and 1080-3850 mL/min for ICG IHC. All clinical, biochemical, endoscopic, and hemodynamic variables were collected within 7 days of each other. Patients were treated according to standard medical practice and were seen as outpatients approximately every 3 months or as inpatients when necessary. If variceal bleeding occurred, they were treated with blood transfusions, esophageal tamponade, and/or sclerotherapy, and were subsequently treated with chronic sclerotherapy (12 patients), P-blocking agents (14 patients), or portal-systemic shunting (9 patients) for prevention of rebleeding; 3 patients died before treatment could be institued, and 6 refused further treatment. The study was terminated on November 1, 1988, when the first included patient reached 60 months of follow-up. Median follow-up was 45 months. Five patients were lost to follow-up after 2, 3, 4, 12, and 26 months. The primary end-point of the study was death. Death was classified as caused by gastrointestinal bleeding if it occurred within 40 days of bleeding, regardless of the severity,24 and as caused by liver failure if progressive impairment of liver function with neurological disturbances occurred. Patients who died of conditions not related to liver disease were censored at the time of death. The secondary end-point of the study was the occurrence of gastrointestinal bleeding, defined as hematemesis or melena observed by a physician, regardless of the severity and source of bleeding. Data Analysis Descriptive statistics are provided as medians and IQR. Comparisons were performed using nonparametric tests. Univariate survival analysis was performed using the Kaplan-Meier methodz5 and the log-rank test.26 Variables contained in the Pugh score were clustered according to the same scoring system. HVPG, ICG IHC, and HPF were clustered at 16 mm Hg, 400 mL/min, and 750 mL/ min, respectively. To establish whether data obtained from hepatic vein catheterization contributed to a better definition of the prognosis in our subjects, after the conventional data had been taken into account, a multiple regression analysis according to the Cox proportional hazard modelz7 was performed. Because the available number of end-points allowed only a limited number of covariates to be analyzed, Pugh score, etiology of disease, treatment received during follow-up in those who bled, size of varices, history of previous variceal bleeding, age, sex, HVPG, HPF, and ICG IHC were studied by Cox’s model. To assess the validity of the model proposed, a split-sample procedure was used.” The clinical usefulness of the data obtained from hepatic vein catheterization was assessed by deriving prognostic indexes that included or excluded such data. Receiver operating characteristic (ROC) curves were drawn, and areas under the curve (AUC) and standard errors of AUC were calculated and compared according to the method of Hanley and McNeil.” In addition the number of cases correctly and incorrectly classified were compared using the McNemar tesL3’

HEPATIC VEIN CATHETERIZATION

March 1992

Results Clinical and Laboratory to Outcome

Data in Relation

At the time of inclusion, median HVPG was 20.2 mm Hg (IQR, 18.2-22.8 mm Hg), median HPF was 735 mL/min (IQR, 563-1081 mL/min), and median ICG IHC was 381 mL/min (IQR, 225-530 mL/ min). During follow-up, 44 patients experienced at least one episode of upper gastrointestinal bleeding. Bleeding was caused by variceal rupture in 32 patients, acute gastric erosions in 7, and peptic ulcer in 1. In 4 patients, no definite diagnosis was obtained. During the study, 54 patients died; the cause of death was gastrointestinal bleeding in 19, liver failure in 20, hepatorenal syndrome in 1, spontaneous bacterial peritonitis in 2, and other infections in 5 (pneumonia, 3 patients; sepsis; meningitis). In 7 patients, death was not related to liver disease (stroke; myocardial infarction, 2 patients; heart failure; biliary cancer; breast cancer; automobile accident). Twelve patients developed hepatocellular carcinoma during follow-up. HVPG was significantly higher in patients who bled during follow-up (21.7 mm Hg; IQR, 18.7-23.9 mm Hg) than in those who did not (19.8 mm Hg; IQR, 17.0-21.9 mm Hg) (P = 0.035). This difference was more evident when only patients without previous bleeding were considered (P = 0.005). HVPG was not significantly different in patients who rebled during follow-up and in those who experienced their first bleeding episode during follow-up. The HVPG was similar in patients who died during follow-up (20.5 mm Hg; IQR, 19.8-23.6 mm Hg) and in patients who survived (20.2 mm Hg; IQR, 16.7-22.0 mm Hg). ICG IHC was not significantly different in patients who bled during follow-up (367 mL/min; IQR, 214512 mL/min) and in those who did not (382 mL/min; IQR, 225-544 mL/min), but it was significantly lower in patients who died (328 mL/min; IQR, 198427 mL/min) than in survivors (424 mL/min; IQR, 257-668 mL/min) (P = 0.01). No difference in HPF was demonstrated either between patients who bled (733 mL/min; IQR, 506-1138 mL/min) and those who did not (737 mL/min; IQR, 569-1000 mL/min) during follow-up or between patients who died (706 mL/min; IQR, 506-1000 mL/min) and those who survived (802 mL/min; IQR, 569-1138 mL/min). A significantly higher proportion of patients with F2 or F3 varices bled during follow-up (41/109) than of patients with Fl varices (3/20) (P = 0.039). Varices were significantly larger in patients with previous variceal bleeding than in patients without previous bleeding (P = 0.01). A significantly higher proportion of patients with previous variceal bleeding bled dur-

AND PROGNOSIS

975

ing follow-up (23/51) than of those with no previous episode of bleeding (21/78) (P = 0.026), and a higher proportion of patients with previous bleeding died (23/51 vs. 24/78). The difference was close to the significance level (P = 0.072). Factors Predicting

Death

According to univariate analysis, the cumulative probability of survival was lower in patients with HVPGs of >16 mm Hg (P < O.Ol), ICG IHCs of ~400 mL/min (P < 0.05) (Figure l), esophageal varices other than Fl (P < 0.05), previous variceal bleeding (P < 0.05), ascites (P < O-01), and a high Pugh score (P < 0.001). It was not related to HPF, sex, or age. Cox multiple regression analysis selected Pugh score, HVPG, ICG IHC, and previous variceal bleeding as the best independent and significant prognostic indicators of survival. The statistical parameters of this model are given in Table 1. The contribution of the parameters obtained from hepatic vein catheterization to overall prognostic accuracy was statistically significant as assessed by the Wald test3’ (HVPG, x2 = 5.81, P < 0.02; ICG IHC, x2 = 5.20, P <

7

A ‘.

‘$

0.8*-

HVPG

<

16 mmHg

6 .2 0.6.. .z0 % & 0.4-y

HVPG

.-f

p<

?j 0.2 E z

>

16 mmHg

0.01

1

oc.:.:.:.:.:.:.:.:.:.1 0

6

12

16

24

30

36

42

48

54

60

months I$+:

:T

123

::,

IGC-IHC

6 :;I++;

E

12

18

24

30

56 41

Figure 1. Kaplan-Meier plots of survival (A) HVPG and (B) ICG IHC.

36

42

< 400

48 33 16

at different

ml/min

54

60

months

levels

of

976

MERKEL ET AL.

Table 1. Statistical

GASTROENTEROLOGY Vol. 102,No. 3

Parameters

of the Cox Model Predicting Death in the 129 Patients With Cirrhosis Improvement

Variable

Coefficient/SE ratio

Coefficient

Pugh score HVPG ICG IHC Previous bleeding

0.2601 0.1030 -0.0008 0.5980

2.41 2.28 -1.82 1.93

0.02). A prognostic index for death (PI-d) was calculated according to the formula PI-d = -0.0008

X ICG IHC + 0.26 X Pugh score + 0.103 X HVPG + 0.598

(if previous

variceal

bleeding

expressing ICG IHC in milliliters HVPG as millimeters of mercury. Factors Predicting

occurred),

per minute

and

Gastrointestinal

patients

at risk

12

18

24

30

36

9.90 4.86 4.04 3.70

P

x2

0.002 0.028 0.045 0.054

10.9 13.9 15.8 17.7

0.001 0.001 0.001 0.001

statistical parameters of this model are given in Table 2. The contribution of HVPG to the overall prognosis was statistically significant as assessed by the Wald test (x2 = 7.34, P < 0.01). A prognostic index predicting gastrointestinal bleeding (PI-b) was computed according to the formula PI-b = 0.103 X HVPG + 0.288 X Pugh score

Bleeding

u

6

P

x2

+ 0.419 x varices degree + 0.927

According to univariate analysis, the cumulative risk of gastrointestinal bleeding was greater in patients with HVPGs of >16 mm Hg (P < 0.05) (Figure 2), larger esophageal varices (P < 0.05), previous variceal bleeding (P < O.Ol), and a high Pugh score (P < 0.05). It was not significantly related to ICG IHC, HPF, ascites, sex, or age. When patients without previous bleeding and those with previous bleeding were considered separately, HVPG was still a predictor of bleeding. Using the multivariate analysis according to Cox regression model, the four variables significant at univariate analysis resulted to contribute significantly to the risk of gastrointestinal bleeding. The

0

Global

42

48

54 60 months

21 108

Figure 2. Kaplan-Meier plots of cumulative probability of not bleeding from the upper gastrointestinal tract at different levels of HVPG.

(if previous expressing

variceal

HVPG as millimeters

bleeding

occurred),

of mercury.

Validity and Clinical Usefulness of Prognostic Indexes Validity of the models was assessed using a split-sample technique. PI-d and PI-b were recalculated in a training set of approximately 60% of randomly selected patients. Individual prognostic indexes were calculated for the test set comprising the remainders. The test set was divided into two groups with PI higher or lower than the median value, and mean estimated survival functions for the patients in the two classes were computed. No significant difference between expected and observed mortality or between expected and observed risk of bleeding was found in patients with different levels of prognostic indexes. To evaluate the clinical usefulness of the addition of the parameters obtained from hepatic vein catheterization to the common data in the prediction of death, PI-d was compared with a prognostic index computed only from Pugh score and previous history of bleeding (PI-c). Sensitivity and specificity in predicting death during follow-up were calculated in eight intervals of prognostic indexes, and ROC curves were drawn. From this analysis, cut-off points of 4.0 for PI-d and 2.8 for PI-c were selected (Figure 3). AUC calculated from PI-d was significantly larger than that calculated from PI-c (AUC = 0.730, SE = 0.050vs.AUC=0.657,SE0.061;z=1.70;P<0.05).In addition, the best cut-off point of PI-d was associated with a significantly smaller misclassification rate

HEPATIC VEIN CATHETERIZATION

March 1992

Table 2. Statistical

Parameters

of the Cox Model Predicting

Gastrointestinal

Bleeding

AND PROGNOSIS

in 129 Patients With Cirrhosis

Improvement Coefficient/SE ratio

Coefficient

Variable HVPG Previous bleeding Pugh score Size of varices

0.1036 0.9270 0.2884 0.4192

2.71 2.87 2.72 1.75

6.27 7.10 6.79 3.26

than PI-c (x2 = 9.00, P < 0.05). A similar analysis was performed to assess the clinical usefulness of PI-b. PI-b was compared with a prognostic index computed excluding HVPG, and ROC curves were drawn. PI-b resulted in a better sensitivity and specificity, and AUC was larger; however, this difference was not significant (AUC = 0.626, SE = 0.054 vs. AUC = 0.586, SE = 0.057; z = 1.11; P = 0.14). The misclassification rate was not significantly affected by removing HVPG. Discussion The results of this study show that HVPG and ICG IHC, obtained through hepatic vein catheterization, are important prognostic indicators in patients with cirrhosis and portal hypertension and add significant information to that available from conventional investigation. In contrast, HPF did not show

~ 0.9

.e

5 0.8 z 0.7 0.6 0.5 0.4 0.3 X ----

0.2 0.1

AUC AUC z = p

<

= 0.730 = 0.657 1.70 0.05

0 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

x2

1

1 -specificity Figure 3. ROC curves of the prediction of death of hepatic cause within 36 months of inclusion according the the prognostic index computed from Pugh score, HVPG, KG IHC, and history of previous variceal bleeding (solid line) and from Pugh score and history of previous variceal bleeding (dashed line).

977

Global P

0.012 0.008 0.009 0.071

x2 5.6 11.9 17.5 19.3

P 0.017 0.003 0.001 0.001

any relevant prognostic implication in the same subjects. Shorter survival in patients with a higher HVPG has been reported in a few studies.‘2,‘3,32 Only the study by Gluud et a1.,13however, addressed the question of whether the same prognostic information might have been obtained in a simpler way without performing hepatic vein catheterization. In that study, HVPG and the size of esophageal varices added significant prognostic information to that obtained from the Child-Turcotte classification. The present study confirms and extends the prognostic value of ICG IHC in predicting survival, which we have previously reported in a shorter series of patients.17 The fact that in our previous study ICG IHC, although exhibiting a remarkable prognostic value, did not improve prognostic ability of the CTP classification17 is probably attributable to the smaller sample size and the shorter follow-up time, resulting in a smaller number of end-points. Statistical significance in these studies is closely related to the number of available end-points31 To our knowledge, the overall usefulness of HVPG and ICG IHC combined has not been studied, although these measurements are usually performed in combination. Pugh score, presence of ascites, and history of previous variceal bleeding are commonly used as predictors of survival in cirrhosis.2,g,‘o,28~33-35 The size of esophageal varices was a predictor of survival in some series,‘3,33 and we have obtained the same result. However, Pugh score, history of previous variteal bleeding, HVPG, and ICG IHC were found to be the best covariates in predicting death from liver disease in such a way that prognostic accuracy could not be improved by adding further data. The relative importance of the data obtained from hepatic vein catheterization in the overall prognostic index may be derived from the coefficient/standard error ratio given in Table 1 .31 HVPG and ICG IHC combined had a weight comparable to that of Pugh score and history of previous variceal bleeding taken together. From a practical point of view, however, the clinical usefulness of the addition of data obtained from hepatic vein catheterization may be better estimated comparing sensitivity and specificity of the whole prognostic index with that of a prognostic index ob-

978

MERKEL ET AL.

tained excluding such data. The AUC drawn according to the complete set of covariates was significantly larger than that drawn excluding HVPG and ICG IHC. Therefore, the addition of data obtained from hepatic vein catheterization improved prognostic efficiency in these patients. Different studies on the relationship between HVPG and risk of gastrointestinal bleeding led to some conflicting results.‘3*36-38 The present investigation show that HVPG is an important predictor of bleeding and that prognostic information arising from both HVPG and size of varices is not redundant because both remained in the final Cox model predicting gastrointestinal bleeding in association with Pugh score and history of previous variceal bleeding. The fact that both HVPG and size of esophageal varices were found to be predictors of bleeding at the same time is consistent with the observation that no statistical relationship exists between these two parameters in patients above the threshold value of 12 mm Hg3’v3’and supports the hypothesis that variceal wall tension, which depends on the product of varix size times pressure, is the major factor determining variceal rupture, as recently suggested by Groszmannl’ on the basis of in vitro studies.3g Unfortunately, a reliable measurement of variceal wall tension is not yet available in clinical practice. The contribution of HVPG to the prognostic index predicting gastrointestinal bleeding was of the same magnitude as that of Pugh score or history of previous variceal bleeding. Because only one of the four prognostic indicators was obtained from hepatic vein catheterization, it is conceivable that a small improvement in prognostic accuracy resulted from the addition of HVPG in the clinical situation. It is likely that in a larger population, improvement in prognostic accuracy might reach statistical significance. In conclusion, in patients with cirrhosis and portal hypertension, hepatic vein catheterization supplied useful prognostic indicators for death and gastrointestinal bleeding, which improved prognostic accuracy obtained from the conventional investigation. If these data are confirmed in further series, in addition to being a valuable tool for clinical research and a means of predicting the efficacy of P-blocking agents,‘4*‘5 hepatic vein catheterization will be considered a clinically useful instrument for the evaluation of such patients.

GASTROENTEROLOGY Vol. 102, No. 3

3.

4.

5.

6.

7.

8.

9.

10.

11. 12.

13.

14.

15.

99:1401-1407. 16. 17.

18. 19.

References 1.

2.

Pugh RNH, Murray-Lyon IM, Dawson JL, Pietroni MC, Williams R. Transection of the oesophagus for bleeding esophageal varices. Br J Surg 1973;60:646-649. Christensen E, Schlichting P, Fauerholdt L, Gluud C, Andersen PK, Juhl E, Poulsen H, Tygstrup N, Copenhagen Study Group for Liver Disease. Prognostic value of Child-Turcotte

criteria in medically treated cirrhosis. Hepatology 1984;4: 430-435. Baker LA, Smith C, Lieberman G. The natural history of esophageal varices. A study of 155 patients in whom varices were diagnosed prior to bleeding. Am J Med 1959;26:229-237. Garceau AJ, Chalmers TC, Boston Inter-Hospital Liver Group. The natural history of cirrhosis. I. Survival with esophageal varices. N Engl J Med 1963;268:469-473. Christensen E, Fauerholdt L, Schlichting P, Juhl P, Poulsen H, Tygstrup N, and CSL. Aspects of the natural history of gastrointestinal bleeding in cirrhosis and the effect of prednisone. Gastroenterology 1981;81:944-952. Merkel C, Bolognesi M, Angeli P, Noventa P, Caregaro L, Sacerdoti D, Gatta A. Prognostic indicators of survival in patients with cirrhosis and esophageal varices, without previous bleeding. Am J Gastroenterol 1989;84:717-722, Beppu K, Inokuchi K, Koyanagy N, Nakayama S, Sakata H, Kitano S, Kobayashi M. Prediction of variceal hemorrhage by esophageal endoscopy. Gastrointest Endosc 1981;27:213-218, North Italian Endoscopic Club for the Study and Treatment of Esophageal Varices. Prediction of first variceal hemorrhage in patients with cirrhosis of the liver and esophageal varices. A prospective multicenter study. N Engl J Med 1988;319:983989. P Cal&. Facteurs predictifs de la premiere hemorrhagic digestive et de la mortalite chez les malades cirrhotiques avec varices oesophagiennes. Gastroenterol Clin Biol 1989;13:54-59. Poupon RE, Thibult N, Barbare JC, Darnis F, Poupon R. La taille des varices oesophagiennes a-t-elle une signification pronostique au tours des cirrhoses alcoholiques? Gastroenterol Clin Biol 1987;11:119-122. Groszmann RJ. Reassessing portal venous pressure measurements. Gastroenterology 1984;86:1611-1617. Vine1 JP, Cassigneul J, Levade M, Pascal JP. Clinical and prognostic significance of porto-hepatic gradient in patients with cirrhosis. Surg Gynecol Obstet 1982;155:347-352. Gluud C, Henriksen JH, Nielsen G, Copenhagen Study Group for Liver Diseases. Prognostic indicators in alcoholic cirrhotic man. Hepatology 1988;8:222-227. Sacerdoti D, Merkel C, Gatta A. Importance of the l-montheffect of nadolol on portal pressure in predicting failure of prevention of rebleeding in cirrhosis. J Hepatol 1991;12:124125. Groszmann RJ, Bosch J, Grace ND, Corm HO, Garcia-Tsao G, Navasa M, Alberts J, Rod&s J, Fisher R, Bermann M, Rofe S, Patrick M, Lerner E. Hemodynamic events in a prospective randomized trial of propranolol versus placebo in the prevention of a first variceal hemorrhage. Gastroenterology 1990:

20.

21.

Keiding S. Hepatic clearance and liver blood flow. J Hepatol 1987;4:393-398. Merkel C, Bolognesi M, Finucci GF, Angeli P, Caregaro L, Rondana M, Gatta A. Indocyanine green intrinsic hepatic clearance as a prognostic index of survival in patients with cirrhosis. J Hepatol 1989;9:16-22. Zimmon DS, Kessler RE. The portal pressure-blood volume relationship in cirrhosis. Gut 1974;15:99-101. Kravetz D, Sikuler E, Groszmann RJ. Splanchnic and systemic hemodynamics in portal hypertensive rats during hemorrhage and blood volume restitution. Gastroenterology 1986;90:1232-1240. Teres J. Diagnosis of variceal bleeding and definition of clinically significant bleeding. In: Burroughs AK, ed. Methodology and reviews of clinical trials in portal hypertension. Amsterdam: Excerpta Medica, 1987:11-15. Japanese Research Society for Portal Hypertension. The gen-

March 1992

HEPATIC VEIN CATHETERIZATION

AND PROGNOSIS

979

era1 rules for recording endoscopic findings on esophageal varices. Jpn J Surg 1980;10:84-87. 22. Merkel C, Sacerdoti D, Finucci GF, Zuin R, Bazzerla G, Bolognesi M, Gatta A. Effect of nadolol on liver haemodynamics and function in patients with cirrhosis. Br J Clin Pharmacol 1986;21:713-719. 23. Rowe11 LB. Measurement of hepatic-splanchnic blood flow in man by dye dilution curves. In: Bloomfield DA, ed. Dye curves. The theory and practice of indicator dilution. Baltimore: University Park Press, 1975:209-225. 24. Serensen TIA. Definition of death in relation to variceal bleeding. In: Burroughs AK, ed. Methodology and reviews of clinical trials in portal hypertension. Amsterdam: Excerpta Medica, 1987:31-35. estimation from incom25. Kaplan EL, Meier P. Nonparametric plete observations. J Am Stat Assoc 1958;53:457-481. 26. Peto R, Pike MC, Armitage P, Breslow NE, Cox DR, Howard SV, Mantel N, McPherson K, Peto J, Smith PG. Design and analysis of randomized clinical trails requiring prolonged observation of each patient. II Analysis and examples. Br J Cancer 1977;35:1-38. 27. Cox DR. Regression models and life tables (with discussion). J R Stat Sot B 1972;34:187-220. 28. Schlichting P, Christensen E, Andersen PE, Fauerholdt L, Juhl E, Poulsen H, Tygstrup N, Copenhagen Study Group for Liver Diseases. Prognostic factors in cirrhosis identified by Cox’s regression model. Hepatology 1983;3:889-895. 29. Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983;148:839-843. 30. MC Nemar Q. Note on the sampling error of the difference between correlated proportions or percentages. Psycometrika 1947;12:153-157, 31. Christensen E. Multivariate survival analysis using Cox’s regression models. Hepatology 1987;7:1346-1358. 32. Vine1 JP, Cassigneul J, Levade M, Voigt JJ, Pascal JP. Assessment of short term prognosis after variceal bleeding in patients with alcoholic cirrhosis by early measurement of portohepatic gradient. Hepatology 1986;6:116-117.

33. Milani A, Marra L, Sicilian0 M, Rossi L. Prognostic significance of clinical and laboratory parameters in liver cirrhosis. Hepatogastroenterology 1985;32:270-272. 34. D’Amico G, Morabito A, Pagliaro L, Marubini E, Liver Study Group of V Cervello Hospital. Survival and prognostic indicators in compensated and decompensated cirrhosis. Dig Dis Sci 1986;31:468-475. 35. Villeneuve JP, Infante-Rivard C, Ampelas M, Pomier-Layrargues G, Huet PM, Marleau D. Prognostic value of aminopyrine breath test in cirrhotic patients. Hepatology 1986;6:928-931. 36. Viallet A, Marleau D, Huet M, Martin F, Farley A, Villeneuve JP, Lavoi P. Hemodynamic evaluation of patients with intrahepatic portal hypertension. Gastroenterology 1975;69:12971300. 37. Garcia-Tsao G, Groszmann RJ, Fisher RL, Conn HO, Atterbury CE, Glickmann M. Portal pressure, presence of gastroesophageal varices and variceal bleeding. Hepatology 1985;5:419-424. 38. Lebrec D, De Fleury P, Rueff B, Nahum H, Benhamou JP. Portal hypertension, size of esophageal varices, and risk of gastrointestinal bleeding in alcoholic cirrhosis. Gastroenterology 1980;79:1139-1144. 39. Polio J, Groszmann RJ. Hemodynamic factors involved in the development and rupture of esophageal varices: a pathophysiologic approach to treatment. Semin Liver Dis 1986;6:318331. Received September 7, 1990. Accepted July 29, 1991. Address requests for reprints to: Carlo Merkel, M.D., Istituto di Medicina Clinica, Clinica Medica 2, Policlinico Universitario, via Giustiniani, 2, I-35100 Padova, Italy. Supported in part by a grant from the Italian Ministery of Education (National Project “Liver Cirrhosis”). The authors thank Professor Albert0 Morabito, Institute ofMedical Statistics and Biometrics, University of Milan, Italy, for helpful suggestions in the statistical analysis. The authors also thank the staff of the wards of the Second Medical Clinic, Department of Clinical Medicine, and the Hepatic Hemodynamic Laboratory, University of Padua, for their expert and dedicated care of the patients.