Cirrhosis and bleeding: the need for very early management q

Transcription

1 Journal of Hepatology 39 (2003) Cirrhosis and bleeding: the need for very early management q Delphine Nidegger 1, Stéphanie Ragot 1, Philippe Berthelémy 2, Claude Masliah 3, Christophe Pilette 4, Thierry Martin 5, Alain Bianchi 6, Thierry Paupard 7, Christine Silvain 8, Michel Beauchant 8, *, A Multicenter Group 1 Institut de Santé Publique Faculté de Médecine de Poitiers, Poitiers, France 2 Service de Gastroentérologie Centre Hospitalier de Pau, Pau, France 3 Service de Gastroentérologie centre Hospitalier Universitaire de Nantes, Nantes, France 4 Service d Hépatologie et de Gastroentérologie Centre Hospitalier et Universitaire d Angers, Angers, France 5 Service de Gastroentérologie Centre Hospitalier de St Nazaire, St Nazaire, France 6 Service de Gastroentérologie Centre Hospitalier du Mans, Le Mans, France 7 Service de Gastroentérologie Centre Hospitalier de Dunkerque, Dunkerque, France 8 Service d Hépatologie Centre Hospitalier Universitaire de Poitiers, BP 577, Poitiers, France Background/Aims: Retrospective studies suggest that the prognosis of patients with cirrhosis and variceal hemorrhage has improved in more recent decades. In a prospective cohort study in which the choice of prophylactic therapy was left to each practitioner, we followed cirrhotic patients with medium/large varices to determine factors predictive of bleeding and death. Methods: Three hundred fourteen patients with grades 2 or 3 esophageal varices (Child A and B/C: 218 and 96) were enrolled. One hundred seventy-three patients had no previous history of variceal bleeding. Only 245 patients (100% of patients with prior variceal hemorrhage, 61% of patients without prior hemorrhage) were receiving some form of prophylactic therapy. The median follow-up was 18 months. Results: There were 76 bleeding events and 14 related deaths (18%); nine of these deaths occurred within 24 h of bleeding onset (two at home, two during hospital transfer, and five in hospital, a mean of 2.5 h after onset; six involved Child C patients). Twenty-five deaths were not due to bleeding but were closely related to cirrhosis. In a Cox model, the presence of tense ascites (relative risk 3.4, 95% confidence interval, CI ) and a prior history of hemorrhage (relative risk 4.4, 95% CI ) were independent predictors of variceal hemorrhage. In patients without a prior history of bleeding, bleeding risk was higher with more prolonged prothrombin time and lower when patients were receiving propranolol. Conclusions: Despite the advent of effective drugs and endoscopic therapy for variceal bleeding, about a quarter of deaths occur very early after bleeding onset, confirming the need for rapid specific management. q 2003 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved. Keywords: Cirrhosis; Variceal bleeding; Octreotide Received 17 December 2002; received in revised form 11 June 2003; accepted 18 June 2003 q The following hospitals and investigators participated in the multicenter group: Alès (Dendale A), Amiens (Vandewalle F), Angers (Pilette C), Angoulême (De Lustrac M), Metz (Perarnau JM), Besançon (Becker MC), Bordeaux (Bernard P and De Lédinghen V), Brest (Nousbaum JB), Caen (Dao T), Clichy (Hillaire S), Dijon (Hillon P), Dreux (Goldfain D), Dunkerque (Paupard T), Evry (Ficher J), Grenoble (Rolachon R), La Rochelle (Faucher P), Le Mans (Bianchi A), Lyon (Mion F), Nancy (Chone L), Nantes (Masliah C), Nice (Tran A), Nîmes (Ribard D), Niort (Staub JL), Orléans (Legoux JL), Pau (Pariente A), Poitiers (Beauchant M, Silvain C), Provins (Chapat O), Rennes (Heresbach D), Rochefort (Charneau J), St Nazaire (Martin T), Tarbes (Druart F), and Vichy (Cassan P). * Corresponding author. Tel.: þ ; fax: þ address: m.beauchant@chu-poitiers.fr (M. Beauchant). 1. Introduction Several recent reports suggest that the prognosis of patients with variceal bleeding has improved over the past four decades [1 3]. McCormick and O Keefe [4] analyzed deaths due to a first bleeding event in the control groups of randomized trials, considering that these patients were likely to have received state-of-the-art treatment. They found that mortality fell significantly, from 65 to 40%, during the period. However, details on the circumstances of bleeding onset were lacking in the studies examined by these authors, and a noteworthy number of patients were lost to follow-up. In the literature, /03/$30.00 q 2003 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved. doi: /s (03)

2 510 D. Nidegger et al. / Journal of Hepatology 39 (2003) bleeding-related mortality is generally estimated after admission to a specialized unit; patients who die at home or during hospital transfer are rarely taken into account [5]. To our knowledge, the risk of death before admission to a specialized unit has not been studied since the widespread adoption of treatments aimed at preventing rupture of esophageal varices. The goal of this prospective study was to identify risk factors for gastrointestinal hemorrhage and death, and treatment modalities, in a cohort of cirrhotic patients with large esophageal varices. 2. Patients and methods March 1996 September 1998, cirrhotic patients were enrolled in a randomized double-blind placebo-controlled trial designed to determine the efficacy of octreotide in bleeding control given at home, before admission to an emergency unit. Three hundred fourteen cirrhotic patients in stable condition and at high risk of bleeding, admitted to 33 centers throughout France, were selected and agreed to participate in the study. After discharge from the hospital, they came back home with the drug or a placebo. If bleeding occurred (hematemesis or melena), they were told to call their general practitioner or a medicalized ambulance in order to receive the allocated drug subcutaneously at home and then to be sent to the nearest emergency unit. The trial was terminated prematurely because of inadequate patient accrual and a low incidence of events. Patients were selected for this study if they were years of age and if they had cirrhosis (all etiologies) and endoscopically documented esophageal varices that were either resistant to insufflation (grade 2), or large and confluent (grade 3). Patients were eligible regardless of whether they had bled previously, but patients with a positive history were only selected 1 week after the last episode and if they were in stable conditions. The choice of primary or secondary prophylaxis was left to the individual investigator. Patients were not selected if they had had portocaval anastomosis or transjugular stenting, or if they had known hepatocellular or extrahepatic carcinoma, or portal vein thrombosis. Patients were seen every 3 months. Follow-up lasted at least 1 year and no more than 2 years in the absence of events (hemorrhage or death). A detailed report of management modalities was available in each case. The diagnosis of cirrhosis was based on liver biopsy or on a combination of physical examination, laboratory tests and imaging. The gravity of liver damage was estimated using the Child-Pugh score [6]. Initial laboratory tests comprised, in addition to parameters required for the Child-Pugh score, liver enzyme activities (ALT, AST and gamma-glutamyl transferase) and a platelet count. The following variables were recorded at each 3-monthly visit: alcohol abstinence (complete, partial, none, or clinically unassessable), complications of cirrhosis during the previous 3 months, and prophylaxis of variceal bleeding. To avoid missing information due to inadequate follow-up, the contact details of each patient s general practitioner were recorded, together with the patient s social security number and place of birth (to obtain the death certificate, if relevant, and the contact details of the physician who signed it). If death occurred before hospital admission, one of the patient s family members, and the attending physician, were interviewed by telephone. The following data were recorded for each gastrointestinal bleeding episode: the times of bleeding onset, hospital admission, and initial endoscopy; detailed endoscopic findings (active bleeding or signs of recent bleeding due to esophageal or gastric varices or other lesions); the treatment used; the number of units of blood and fresh frozen plasma administered; and outcome on day 30. Bleeding control was defined as follows: during the first 24 h following admission: no hematemesis, no decline in the maximal blood pressure exceeding 20 mmhg and/or no rise in the heart rate exceeding 20 bpm on two separate occasions 1 h apart; transfusion requirements,2 blood units, and a hematocrit.27% and/or hemoglobin.9 g/dl. Re-bleeding was defined as hematemesis or melena starting between 24 h and 30 days after initial onset, whatever the need for blood transfusion. Death was considered as related to hemorrhage when it occurred within 30 days after bleeding occurrence. The Child-Pugh score was recorded at admission, together with the laboratory data obtained during the entry phase. All the patients signed an informed consent form, and the protocol was approved by the ethics committee of Poitiers University Hospital (Comité de Protection des Personnes dans la recherche Biomédicale Poitou- Charentes) according to French law Statistical analysis Data were collected using Cob Info software version 6 and were analyzed with Statview software (SAS Institute, Berkeley CA). Continuous variables were expressed as means ^ standard deviation and nominal variables were recorded as crude numbers and frequencies. Student s unpaired t test or Mann Whitney test were used to compare continuous variables. The Chi 2 test was used to identify differences between categorical variables. Fisher s test was used in case of expected values lower than 5. Kaplan Meier curves were used to estimate the frequencies of bleeding and survival. Curves were compared between patient subgroups by using the log-rank test. Cox regression analysis was performed using a backward manual procedure to determine the most important independent variables influencing the risk of bleeding and death. These variables were entered in the maximal model if the P value of the relationship with the dependent variable was less than P values below 0.05 were considered to indicate statistical significance. 3. Results Two hundred forty men and 74 women with an average age of 55 years (median 57 years, range years) were selected. All patients were followed for at least 1 year in the absence of events (hemorrhage or death), and no patients were lost to follow-up. The cirrhosis was due to alcohol in 266 cases (85%), hepatitis in 41 cases (13%, two related to B virus, 38 to C virus and one to B plus C virus), and miscellaneous causes in seven cases (2%); 16 patients were heavy drinkers (more than 80 g of ethanol per day). The median follow-up was 373 days (range days). Liver damage, based on the Child-Pugh score, was moderate (class A (98) or B (120)) in 218 cases (69%) and severe (class C) in 96 cases (31%). At the time of entry, 173 patients had no history of gastrointestinal hemorrhage; 105 (61%) of these patients were receiving prophylaxis, consisting of betablocker monotherapy (n ¼ 90), nitrate monotherapy (n ¼ 7), combined nitrate-betablocker therapy (n ¼ 3), ligation alone or combined with a betablocker (n ¼ 4), and sclerotherapy alone (n ¼ 1). Of the 141 patients with a prior history of bleeding, 140 were receiving prophylaxis, consisting of a betablocker alone in 14 cases, a nitrate alone in one case, sclerotherapy and/or mainly banding in 43 cases, and a combination of betablockers (alone or combined with nitrates) and endoscopic treatment in 82 cases. The trial was terminated prematurely because of inadequate patient accrual and a low incidence of events. During follow-up, 76 (24%) of the 314 selected patients had a bleeding episode. Forty-eight (63%) of these patients received one to three injections of the allocated treatment. Twenty-eight patients received no study drug injections. The most frequently stated reasons for non-respect of the protocol were loss of the product, forgetfulness of

3 D. Nidegger et al. / Journal of Hepatology 39 (2003) the protocol, breaking of vials, patient or physician refusal to inject the treatment during the bleeding episode, and excessively abundant bleeding. Sixteen of the 40 patients in the placebo group and 15 of the 36 patients in the octreotide group received a vasoactive drug in addition to the allocated treatment, even though this was forbidden by the protocol. No significant difference was found between the placebo group and the octreotide group in terms of the 24-h bleeding control rate (27/39 (69%) versus 25/37 (68%), respectively) or the mortality rate on day 30 (6/39 (15%) and 8/37 (22%), respectively) Occurrence of bleeding episodes During follow-up, 76 (24%) of the 314 selected patients had a bleeding episode. Only two patients bled less than 30 days after entry into the study, and one died within the first 24 h, 12 days after entry. Bleeding episodes were more frequent in patients with a prior bleeding history (53/141, 38%) than in patients with no bleeding history (23/173, 13%; P, 0:001). The bleeding risk was 34 and 10% at 1 year and 38 and 13% at 2 years, respectively, in patients with and without a bleeding history (Fig. 1). The timing of treatment measures and endoscopy is shown in Table 1, together with the severity of liver disease and the bleeding treatment modalities. The median interval between bleeding onset and hospital admission was 3 h (0 168 h), and exceeded 8 h in 18/76 (24%) of cases. Bleeding was controlled in 51 patients (67%). Eleven patients re-bled between 24 h and 30 days (all between days 2 and 4), and were treated with sclerotherapy (n ¼ 4), ligature (n ¼ 1), octreotide (n ¼ 4), embolization (n ¼ 1) or transjugular shunting (n ¼ 1) Prognostic factors associated with the bleeding risk Univariate analysis of the entire dataset showed that the following factors had significant prognostic value: the prothrombin time (P, 0:01), serum albumin level (P ¼ 0:01), AST (xn increments) (P ¼ 0:03), a prior history of bleeding from esophageal varices (P, 0:0001), Table 1 Main characteristics of the 76 patients who bled during follow-up Median interval between bleeding onset and hospital admission (h; range): n ¼ 62 (h) Median interval between admission and endoscopy (h; range; n ¼ 62; 82%) (h) 3 (0 168) 6 (0 39) Endoscopic findings (n ¼ 62) a Active bleeding: (%) 27 (44) Gastroesophageal varices 19 Portal hypertensive gastropathy 4 Sclerotherapy-related ulcer 2 Other cause unrelated to portal hypertension 2 Inactive bleeding: (%) 35 (56) Clot on varix (esophagus/cardia/fundus) 15/10/2 Undetermined cause 8 Child-Pugh class at admission: A/B/C (n) b 21/27/28 Treatment of the bleeding episode during the first 24 h (n) Blood transfusion (.4 units, first 24 h) 22 Sclerotherapy 35 Banding 9 Vasoactive drug 61 Balloon tamponade 5 Failure of bleeding control or death at 24 h (n) (%) 24/76 (32) Complications (n) (%) 33 (53) Encephalopathy 15 Spontaneous peritonitis/hepatorenal syndrome 0/0 30-day mortality (n)/mean time after bleeding onset (days, range) 14/2.7 (0 16) a Nine patients died early before endoscopic examination could be done, and five missing data. b Calculated at admission or extrapolated from the last examination in case of early death. and ascites (P ¼ 0:016). In multivariate analysis (Cox model), only a prior history of bleeding from esophageal varices, ascites and AST (xn increments) were independent predictors of the bleeding risk (Table 2). In patients with a history of bleeding, the following variables had significant prognostic value for the bleeding Table 2 Factors predictive of the risk of bleeding in the overall population (n 5 314) Independent variable Coefficients P Relative risk 95% confidence interval Fig. 1. Bleeding-free survival among patients with and without a history of bleeding from esophageal varices. Previous 21.48, [2.6;7.5] bleeding AST a [1.0;1.4] Ascites NS 1.4 [0.7;2.7] , [2.5;5.9] a Per xn increment.

4 512 D. Nidegger et al. / Journal of Hepatology 39 (2003) Fig. 2. Bleeding-free survival according to the presence of ascites in patients with a history of bleeding from esophageal varices. risk in univariate analysis: age (P ¼ 0:02), ascites (P, 0:001), AST (xn) (P ¼ 0:04) and creatinemia (P ¼ 0:01). After adjustment, only abundant ascites significantly influenced the bleeding risk, which was 4.2- fold higher than in patients without ascites (95% confidence interval, CI , P, 0:0001) (Fig. 2). In patients with no history of bleeding, the bleeding risk was higher when the prothrombin time was prolonged (P, 0:005) and lower when the patient was receiving propranolol prophylaxis (P, 0:05). In the Cox model, only the prothrombin time was associated with the risk of bleeding: the risk fell by a factor of 1.9 per increment of 10 units (unit defined as percent of control) Outcome Thirty-nine patients (12%) died during follow-up, 14 (36%) of bleeding and 25 (64%) of another cause. Death occurred within 24 h of bleeding onset in nine cases (two at home, before the ambulance arrived; two during transfer in a medicalized ambulance; and 5 an average of 2.5 h after hospital admission). Compared to the other 67 patients, these patients had significantly a longer prothrombin time and a lower serum albumin level at entry into the study (Table 3). Six of them were Child-Pugh class C (P ¼ 0:22). Eight had alcoholic cirrhosis and 7 were probably abstinent. Seven patients were receiving prophylaxis (betablockers alone (2) or with endoscopic ligation (1) or sclerosis (2), endoscopic treatment alone (2)). None of them had hepatocellular carcinoma diagnosed during follow-up before the bleeding episode. The other five patients died between 1 and 16 days after initial bleeding onset. The other causes of death were hepatic encephalopathy (n ¼ 6), liver failure (n ¼ 4), a hepatorenal syndrome (n ¼ 4), infection (n ¼ 6), stroke (n ¼ 2), hepatocellular carcinoma (n ¼ 1), cancer of the pharynx (n ¼ 1), and hemoperitoneum (n ¼ 1). The patients living conditions did not influence their management; in particular, deaths were not more frequent among patients living alone. In univariate analysis, four factors were significantly associated with the risk of death, namely the prothrombin time (P ¼ 0:009), bilirubinemia (P ¼ 0:01), the Child-Pugh class (P ¼ 0:002) and the severity of ascites (P ¼ 0:03). In Cox multivariate analysis, only the prothrombin time (P, 0:0001) and AST activity (P ¼ 0:04) were significantly associated with the risk of death. 4. Discussion Details on the management of bleeding episodes were available for all the patients in this study. The results confirm the improved prognosis in this setting [1 4]: as regards deaths occurring before hospital admission, bleeding-related mortality was about 20%, a rate lower than that reported in the 1980s [3]. El-Serag et al. [2] found that the 30-day mortality rate improved significantly from 1981 to 1991, despite the fact that patients were older and more likely to have ascites, encephalopathy or spontaneous peritonitis towards the end of the study period. Similar results were obtained in randomized trials during this period [3,4]. This improvement can be attributed to better patient management [7,8], including better prevention of infection [9], careful compensation of blood losses [8], and better Table 3 Main characteristics recorded at the time of selection in nine patients who died during the first 24 h after the bleeding episode: comparison with the 67 other patients who also presented a bleeding episode Death within first 24 h (n ¼ 9) Other bleeding episode (n ¼ 67) P value Age (median) Alcohol abstinence (n; %) a 7 (78) 32 (48) 0.29 Previous bleeding (n; %) 5 (56) 48 (72) 0.44 Initial variceal size (grade 2/grade 3) 7/2 54/ Child-Pugh grade: A/B/C 1/2/6 20/25/ Serum bilirubin (mmoles/l) Prothrombin time (percent control) Serum albumin (g/l) Ascites: absent/moderate/large (n) 3/0/6 33/13/ Encephalopathy (%) 1 (11) 1 (2) 0.22 Bleeding prophylaxis (%) 7 (78) 56 (84) 0.65 a Number of patients who had probable complete or partial alcohol abstinence on clinical judgment at the last visit preceding the bleeding episode.

5 D. Nidegger et al. / Journal of Hepatology 39 (2003) bleeding control with early use of endoscopic treatment [2] and/or vasoactive drugs [10 12]. Our results emphasize the vital importance of the time interval between admission and the start of management [13]. Two-thirds of deaths linked to the bleeding episodes occurred within 24 h after onset, and nearly one-third occurred before admission to a specialized unit. Surprisingly, the interval between bleeding onset and hospital admission was more than 8 h in 29% of cases, even though the patients had been instructed to contact the duty physician or ambulance service so that the study drug could be administered rapidly. However, major difficulties were encountered in our study with early administration of a vasoactive drug at the patient s home, underlining the difficulties encountered in this type of clinical trial. The prognosis did not seem to be influenced by the patients living conditions (especially living alone). Our results show that these patients should be treated very rapidly, at the slightest sign of bleeding, and should be admitted to a specialized unit. Patients must be better informed, and must be taught to recognize the early symptoms of bleeding (melena or malaise). To our knowledge only one study has demonstrated the benefit of vasoactive drug administration during hospital transfer [12]. Other studies must be conducted, especially to test drugs that can readily be administered by a general practitioner and that do not require special hemodynamic monitoring (e.g. somatostatin derivatives). The choice of bleeding prophylaxis was left to the individual investigators in this study. Despite several consensus meetings, an astonishing 39% of patients who had not previously bled were receiving no preventive treatment in our study, even though they were known to have large esophageal varices. The bleeding risk was significantly lower in patients receiving propranolol prophylaxis, confirming the results of a meta-analysis [14]. In contrast, all the patients with a prior history of bleeding received preventive treatment, mainly consisting of ligature or endoscopic sclerosis, alone or in combination with a betablocker. The bleeding risk did not differ according to the type of preventive treatment, as previously reported [15]. The risk of bleeding from esophageal varices has fallen strongly with the advent of prophylactic treatments. The estimated 2-year bleeding risk before the prophylaxis era was 20 30% for a first event and 50 80% for relapses [15]. In our study these rates were only 13 and 38%, respectively, confirming the efficacy of prophylaxis [15]. In contrast, the factors predictive of bleeding and death have not changed in recent years. Patients who have had a first bleeding episode remain at a high risk of relapse [16]. The prothrombin time and ascites are well-known prognostic factors in this setting [17]. Alcohol withdrawal had no prognostic value in our study, but collection of this variable is notoriously unreliable. The influence of alcohol withdrawal is controversial [18,19]. Although oral ethanol intake increases portal pressure and collateral portal blood flow [20], abstinence has only been found beneficial in one study [21]. Using a scoring system, McCormick et al. [19] failed to demonstrate an improvement in the prognosis of alcoholic cirrhotic patients who became abstinent following a variceal bleed. In our study, AST activity measured at entry was significantly associated with the bleeding risk, pointing to a possible role of associated alcoholic hepatitis. Seven of the eight alcoholic patients who died within the first 24 h were probably abstinent, but had more severe liver disease, which suggests that the underlying disease may have advanced to such an extent that the benefit of abstinence is negligible [19]. In conclusion, this study confirms that the management of cirrhosis-related bleeding has improved in recent years. However, despite the advent of effective drug-based and endoscopic treatments, many deaths still occur before hospital admission, suggesting a need for very early specific intervention, if possible at the patient s home. Acknowledgements This study was supported by a grant from Novartis; we thank David Young for editorial assistance. References [1] Carbonell N, Pauwels A, Fourdan O, Serfaty L, Levy VG, Poupon R, et al. Diminution de la fréquence des récidives précoces après rupture de varices oesophagiennes au cours des deux dernières décennies: rôle des traitements vasoactifs et de la ligature endoscopique. Gastroenterol Clin Biol 2001;25:613. (abstract). [2] El-Serag HB, Everhart JE. Improved survival after variceal hemorrhage over an 11-year period in the Department of Veterans Affairs. Am J Gastroenterol 2000;95: [3] Pagliaro L, D Amico G, Pasta L, Tiné F, Aragona E, Politi F, et al. Efficacy and efficiency of treatments in portal hypertension. In: De Franchis R, et al., editors. Portal hypertension II. Oxford: Blackwell Science Ltd; p [4] McCormick PA, O Keefe C. Improving prognosis following a first variceal haemorrhage over four decades. Gut 2001;49: [5] Pascal JP, Calès P and a Multicenter Study Group. Propranolol in the prevention of first upper gastrointestinal tract hemorrhage in patients with cirrhosis of the liver and esophageal varices. N Engl J Med 1987; 317: [6] Pugh RNH, Murray-Lyon IM, Dawson JL, Pietroni MC, Williams R. Transection of the oesophagus for bleeding oesophageal varices. Br J Surg 1973;60: [7] Jalan R, Hayes PC. UK guidelines on the management of variceal haemorrhage in cirrhotic patients. Gut 2000;46(Suppl. III): iii1 iii15. [8] Burroughs AK. General management of the cirrhotic patient with acute variceal bleeding. In: De Franchis R, editor. Portal hypertension III. Oxford: Blackwell Science Ltd; p [9] Bernard B, Grangé JD, NguyenKhac E, Amiot X, Opolon P, Poynard T. Antibiotic prophylaxis for the prevention of bacterial infections in cirrhotic patients with gastrointestinal bleeding: a meta-analysis. Hepatology 1999;29:

6 514 D. Nidegger et al. / Journal of Hepatology 39 (2003) [10] Besson I, Ingrand P, Person B, Boutroux D, Heresbach D, Bernard P, et al. Sclerotherapy with or without octreotide for acute variceal bleeding. N Engl J Med 1995;333: [11] Calès P, Masliah C, Bernard B, Garnier PP, Silvain C, Szostac- Talbodec N, et al. Early administration of vapreotide for variceal bleeding in patients with cirrhosis. N Engl J Med 2001;344: [12] Levacher S, Letoumelin P, Pateron D, Blaise M, Lapandry C, Pourriat JL. Early administration of terlipressin plus glyceryl trinitrate to control active upper gastrointestinal bleeding in cirrhotic patients. Lancet 1995;346: [13] Burroughs AK, Mezzanotte G, Philips A, McCormick PA, McIntyre N. Cirrhotics with variceal hemorrhage: the importance of the time interval between admission and the start of analysis for survival and rebleeding rate. Hepatology 1998;9: [14] D Amico G, Pagliaro L, Bosch J. Pharmacological treatment of portal hypertension: an evidence-based approach. Sem Liver Dis 1999;19: [15] D Amico G, Pagliaro L, Bosch J. The treatment of portal hypertension: a meta-analytic review. Hepatology 1995;22: [16] Thomsen BL, Moller S, Sorensen TIA and The Copenhagen Esophageal Varices Sclerotherapy Project. Optimized analysis of recurrent bleeding and death in patients with cirrhosis and esophageal varices. J Hepatol 1994;21: [17] Christensen E. Prognostic models in chronic liver disease: validity, usefulness and future role. J Hepatol 1997;26: [18] Calès P, Pascal JP. Histoire naturelle des varices oesophagiennes au cours de la cirrhose (de la naissance à la rupture). Gastroenterol Clin Biol 1988;12: [19] McCormick PA, Morgan MY, Phillips A, Yin TP, McIntyre N, Burroughs AK. The effects of alcohol use on rebleeding and mortality in patients with alcoholic cirrhosis following variceal haemorrhage. J Hepatol 1992;14: [20] Luca A, Garcia-Pagan JC, Bosch J, Feu F, Cabaleria J, Groszmann J, et al. Effects of ethanol consumption on hepatic hemodynamics in patients with alcoholic cirrhosis. Gastroenterology 1997;112: [21] Dagradi AE. The natural history of oesophageal varices in patients with alcoholic liver cirrhosis: an endoscopic and clinical study. Am J Gastroenterol 1972;57: