Risk Factors for Hemorrhage From Gastric Fundal Varices

Transcription

1 Risk Factors for Hemorrhage From Gastric Fundal Varices TETSUHIRO KIM, 1 HIROSHI SHIJO, 1 HIROSHI KOKAWA, 1 HIDEO TOKUMITSU, 1 KATSUHIKO KUBARA, 1 KAZUHIRO OTA, 1 NOBUO AKIYOSHI, 1 TAKESHI IIDA, 1 MASANORI YOKOYAMA, 1 AND MAKOTO OKUMURA 1 The incidence and the risk factors of hemorrhage from prospective trials. 8,9 However, the incidence of hemorrhage gastric fundal varices (FV) have not been fully evaluauthors from FV has not been fully investigated. 3,7 Although several ated. We therefore conducted a retrospective and prospective have mentioned the incidence of hemorrhage from study to define the incidence and risk factors FV, 3,7 to our knowledge, there has been no prospective study for such episodes. We investigated 132 patients with cirlarge documenting the cumulative risk for FV hemorrhage in a rhosis and gastric FV. Of these 132 patients, 15 patients number of patients. The aim of this study was to define had hemorrhagic FV at the time of enrollment. The clinical the incidence and the risk factors for hemorrhage from FV. characteristics were compared between these pa- tients and those without a first hemorrhage from FV. In PATIENTS AND METHODS the patients who had never previously bled, the inciber Baseline Clinical Assessment. From January 1985 through Decem- 1995, a total of 1,392 cirrhotic patients consecutively underwent dence and risk factors were prospectively investigated. upper gastrointestinal endoscopy at our institute. Of these patients, The size of FV was greater and red-spot on the FV were 540 patients had esophageal varices, 96 patients had gastric varices, more prevalent in patients with hemorrhagic FV. Child s and 253 patients had both esophageal and gastric varices. In the status was also more severe in these patients. In the 117 patients with gastric varices, 143 patients had FV, while the repatients who had never bled, 34 hemorrhages from FV maining 206 patients had varices in the cardia. In the 143 patients occurred during the follow-up period. The cumulative with FV, 11 patients were subsequently excluded from further evaluation risk for such hemorrhage at 1, 3, and 5 years was 16%, because of their refusal to participate in the study. We thus 36%, and 44%, respectively. A multiple regression analy- investigated a total of 132 patients with FV in the current study. sis (Cox s model) revealed the size of varices, red-spot The subjects consisted of 89 men and 43 women ranging in age from on the FV, and Child s status to be statistically signifihistory was taken and routine liver chemistry values and prothrom- 34 to 82 years (mean age, 59.8 { 10.8 years). At entry, a full medical cant, as well as independent predictors for hemorrhage bin activity were determined for each patient. The patients also unfrom FV. The endoscopic criteria (size of the largest derwent a complete physical examination and were evaluated for the varix and presence of red-spot), as well as the hepatic presence of hepatitis B and hepatitis C virus markers. All patients functional reserve, provide the most essential informa- had cirrhosis diagnosed based on history, physical examination, and tion for predicting a hemorrhage from FV. An estimation liver chemistry values (54 patients) or liver histological features (78 of the probability for hemorrhage from FV based on patients) (or both). The cause of cirrhosis was alcohol abuse in 15 Cox s model may therefore be beneficial in the clinical patients, a hepatitis B virus infection in 14, a hepatitis C virus infecmanagement of patients with high-risk FV. (HEPATOLOGY tion in 97, primary biliary cirrhosis in 3, and cryptogenic cirrhosis 1997;25: ) in 3. The hepatic functional reserve was graded according to the Child-Pugh s classification: patients were class A, 38 patients were class B, and 25 patients were class C. Sixty-one patients had Varices of the gastric fundus (so-called fundal varices [FV]) cirrhosis and concomitant hepatocellular carcinoma (HCC) at entry. are initiated by the dilation of short or posterior gastric veins HCC was diagnosed based on the histological findings from biopsy and are frequently associated with a large gastrorenal specimens and/or angiography. Eleven patients had a history of shunt. 1 The blood flow of such collaterals is abundant, 2 and treatment for esophageal varices by endoscopic injection sclerotherhemorrhage from FV is more serious than that of esophageal apy. All patients were informed of the scientific nature of the investivarices. The treatment of hemorrhagic FV by endoscopic gation and gave their informed consent. The study protocol was ap- proved by the Hospital Ethics Committee. procedures sometimes fails to cease the bleeding, and surgi- Assessment of Endoscopic Findings. The diagnosis of FV was made cal modalities are thus often required for hemostasis. 3-6 As a based on endoscopy with the agreement of two experienced endoscoresult, the mortality rate is high in patients with hemor- pists when one or more distinct venous channels were found in the rhagic FV. 3-6 In addition, because hemorrhagic esophageal gastric fundus. In patients with hemorrhagic FV, the endoscopic findings varices have recently been treated by endoscopic variceal were analyzed after the cessation of bleeding. The size of FV sclerotherapy, the development of FV following the obliteravation was estimated semiquantitatively based on the simultaneous obser- tion of esophageal varices is now being noted with increasing of metal forceps and varices, and was classified as follows: frequency. 3,7 It is therefore important to clarify the incidence small Åõ5 mm; medium Å 5 to 10 mm; and large Åú10 mm. The and risk factors for hemorrhage from FV. presence of red-spot on FV was defined according to the criteria established by Hashizume et al. 7 including reddish mucosal areas or The incidence of hemorrhage from esophageal varices and spots localized on FV (Fig. 1). Our definition of the red-spot did not the risk factors for such episodes have been studied by several therefore include portal hypertensive gastropathy in the fundus that extended diffusely over the fundal mucosa. The presence or absence of esophageal varices was also evaluated. The FV were classified as either FV continuous with esophageal varices (type 1 FV) or solitary Abbreviations: FV, fundal varices; HCC, hepatocellular carcinoma. FV (type 2 FV), as described previously. 3,4 Depending on the time of From the 1 First Department of Internal Medicine, School of Medicine, Fukuoka Univer- the first examination, the FV were then further classified as either sity, Fukuoka, Japan. primary FV (FV present at initial examination) or secondary FV (FV Received April 18, 1996; accepted October 16, occurring after an obliteration of the esophageal varices). Address reprint requests to: Hiroshi Shijo, M.D., First Department of Internal Medicine, Follow-up. Patients with no variceal bleeding were prospectively School of Medicine, Fukuoka University, 45-1, 7-chome Nanakuma, Jonan-ku, Fukuoka, , Japan. investigated. Follow-up endoscopy was performed regularly, initially Copyright 1997 by the American Association for the Study of Liver Diseases. at 3- to 6-month intervals. The entry point of the study was the day /97/ $3.00/0 of the endoscopic diagnosis of FV, and the endpoint was the day a 307 AID Hepa 0059 / 5p1b$$ :30:35 hpta WBS: Hepatology

2 308 KIM ET AL. HEPATOLOGY February 1997 FIG. 1. Red-spot found on the fundal varix. Red-spots were defined as localized reddish mucosal area or spots on the mucosal surface of a varix. In patients without first variceal bleeding, the incidence of hemorrhage from FV was prospectively analyzed. The disease-free curves were plotted by the Kaplan-Meier method and compared by a univariate Cox s proportional hazard model. Variables that achieved statistical significance (P õ.05, two-tailed) in the univariate analyses were subsequently included in the multiple regression analysis (Cox s model). 11 With the significant prognostic variables obtained from the multiple regression analysis, a prognostic index for hemorrhage from FV was calculated for each patient using the equation: l i (t)/l 0 (t) Å exp(b 1 z 1 / rrrr / b p z p ) prognostic index Å b 1 z 1 / rrrr / b p z p where l i (t) represents the hazard rate for the hemorrhage from FV at time t for a particular patient with z variables (z 1 to z p ), l 0 (t) is the hazard at time t for a patient whose z variables are all zero, and b 1 to b p are the regression coefficients of the variables. Higher values for the prognostic index indicate a higher probability of hemorrhage from FV. The details regarding this method have been described in a previous report. 11,12 The statistical analysis system was used for all analyses. 13 RESULTS Retrospective Study. Of the 132 patients, 15 patients had an active hemorrhage from FV, while the remaining 117 patients had no history of hemorrhage from FV at the time of hemorrhage from FV was first observed. Any patients who died of enrollment (Table 1). The variceal size was greater and the unrelated causes, such as hepatic failure or HCC, were withdrawn from the study on the day of death. None of the patients received presence of red-spot was more prevalent in patients with any vasoactive drugs, such as b-blockers or nitrates, during the study hemorrhagic FV. Child s status and prothrombin time were period. The follow-up ended on December 31, 1995, 1 year after the also more severe in patients with hemorrhagic FV. However, enrollment of the last patient in the study. All patients were thus such parameters as the type of FV (type 1 or type 2), classifi- potentially observed for at least 12 months. cation as primary or secondary FV, presence of esophageal Criteria for a Hemorrhage From FV. Whenever, during the follow- varices, age, sex, cause of cirrhosis, presence of ascites, biliruup, a patient was admitted to our institute because of an upper bin level, albumin level, presence of encephalopathy or presgastrointestinal hemorrhage, emergency endoscopy was performed ence of HCC did not differ between the two groups. to establish the source of bleeding. The diagnosis of hemorrhage from Prospective Study. Forty-one patients died of bleeding, he- FV was defined as the presence of hematemesis or melena (or both), together with a decrease in the hemoglobin level of 2 g/dl and patic failure, HCC, or of other causes, and were followed until endoscopic evidence of active bleeding from the FV or a varix with death. Of the 117 patients, 49 (41.9%) were followed for less an adherent clot. In patients who were no longer bleeding at the than 12 months, 40 (34.1%) for 12 to 36 months, and 28 time of endoscopy, FV were considered to be the source of bleeding (23.9%) for more than 36 months. For the entire series, the if no other source could be identified in the presence of distinct, large median and range of follow-up period were 15.2 and FV and no esophageal varices. months, respectively (mean { SD Å 22.9 { 22.6). The median Statistical Methods. The endoscopic, clinical, and laboratory pa- and range of numbers of endoscopic procedures performed rameters including sex, the cause of cirrhosis, the presence of HCC, per patient were 3 and 1-16 sessions, respectively (mean { the presence of esophageal varices, the presence of red-spot on FV, SD Å 4.1 { 3.8). size of FV, type 1 or type 2 FV, primary or secondary FV, presence of ascites, presence of encephalopathy, albumin level, bilirubin level, Forty-three of the 117 patients (36.7%) bled from the upper prothrombin time, and Child s status were compared between paemergency endoscopy were bleeding from FV in 34 patients, gastrointestinal tract during the follow-up. The diagnoses of tients with hemorrhagic FV and those without a first hemorrhage from FV. Both the x 2 and Mann-Whitney tests were used for a comparison of the two groups. ulcer in 2, and bleeding from portal hypertensive bleeding from esophageal varices in 6, bleeding from a gastric gastropathy TABLE 1. Clinical Characteristics of Patients With Hemorrhagic FV and With no First Hemorrhage From FV Bleeder Nonbleeder P Sex (male:female) 12 (80):3 (20) 77 (66):40 (34) NS Etiology of cirrhosis (AL:non-AL) 4 (27):11 (73) 11 (9):106 (91) NS Child s status (A:B:C) 3 (20):9 (60):3 (20) 66 (56):29 (25):22 (19) õ.05 Ascites (absent:mild:severe) 11 (73):3 (20):1 (7) 89 (76):19 (16):9 (8) NS Bilirubin (õ2.0: :ú3.0 mg/dl) 10 (66):2 (14):3 (20) 91 (78):19 (16):7 (6) NS Albumin (ú3.5: :õ2.8 g/dl) 4 (27):7 (46):4 (27) 47 (40):50 (43):20 (17) NS Prothrombin time (ú70:40-70:õ40%) 3 (20):11 (73):1 (7) 59 (50):52 (45):6 (5) õ.05 Encephalopathy (absent:mild:severe) 14 (93):0 (0):1 (7) 99 (85):14 (12):4 (3) NS Presence of HCC (0:/) 9 (60):6 (40) 62 (53):55 (47) NS Esophageal varices (0:/) 11 (73):4 (27) 74 (63):43 (37) NS Primary or secondary FV 14 (93):1 (7) 106 (91):11 (9) NS Type 1 or type 2 FV 3 (20):12 (80) 36 (31):81 (69) NS Size of FV (small:medium:large) 0 (0):9 (60):6 (40) 29 (25):56 (48):32 (27) õ.05 Red-spot on GV (0:/) 5 (33):10 (66) 79 (68):38 (32) õ.05 NOTE. Numbers in parentheses denote percentages. Abbreviations: AL, Alcoholic cirrhosis; NS, not significant; type 1 FV, FV continuous with esophageal varices; type 2 FV, solely FV; GV, gastric varices. AID Hepa 0059 / 5p1b$$ :30:35 hpta WBS: Hepatology

3 HEPATOLOGY Vol. 25, No. 2, 1997 KIM ET AL. 309 FIG. 3. The cumulative survival curve after first bleeding from the FV in FIG. 2. The cumulative percentage of the patients showing no first variceal the 34 patients. bleeding from FV. presence of red-spot on FV. Using this method, the variables in 1. In patients with hemorrhagic FV, 22 patients had active were included in the model in the following order: the size of bleeding from FV, 10 patients had an adherent clot on FV, FV, Child s status, and the presence of red-spot. Because the and 2 patients were no longer bleeding at the time of endosnegligible increase in the significance of the model, 11 the final inclusion of more clinical variables would have led to only a copy. Eighteen of the 34 hemorrhages from FV occurred within the first year of inclusion. The cumulative percentages model included only the Child s status (P Å.014), the size of of hemorrhage from FV at 1, 3, and 5 years were 16%, 36%, FV (P Å.009), and the presence of red-spot (P Å.047) as and 44%, respectively (Fig. 2). The cumulative survival curve significant and independent risk factors (Table 4). after the first bleeding from FV is shown in Fig. 3. Most The prognostic index was calculated for each patient ac- deaths occurred within 1 year after bleeding from FV, and the cording to the following equation: cumulative 1-year survival rate was 48% in these patients. prognostic index Å Child s status / 0.78 Tables 2 and 3 show the results of a univariate analysis of risk factors for hemorrhage from FV. The patients with larger 1 variceal size / the presence of red-spot. varices bled more often than those with smaller variceal col- Using this formula, the Child s status (A Å 0, B Å 1, C Å umns (Fig. 4). Hemorrhage was more frequent in patients in 2) and size of FV (small Å 0, medium Å 1, large Å 2) were whom red-spot had been demonstrated at the time of recruit- introduced as semiquantitative variables, and the presence ment (Fig. 4). The risk of hemorrhage was significantly af- or absence of red-spot as a dichotomous variable (absent Å fected by the severity of liver disease as assessed by the 0, present Å 1). Figure 5 shows the estimated probability of Child s status (Fig. 4), presence of ascites, albumin level, and not bleeding at different levels of the prognostic index. The prothrombin time. However, such parameters as the type of 1-year probability of bleeding as a function of all possible FV, classification as primary or secondary FV, the presence combinations of the two endoscopic variables, size and redof esophageal varices, age, sex, cause of cirrhosis, bilirubin spot, was also calculated separately for the patients in Child level, the presence of encephalopathy, and the presence of class A, B, and C. The results of such calculations are re- HCC did not emerge as significant risk factors. ported in Table 5, which can be used for a quick estimation To further evaluate their prognostic significance with re- of a patient s 1-year chance of bleeding. gard to hemorrhage from FV, the following parameters were included in a multiple regression analysis using a forwardselection DISCUSSION method (Cox s model): Child s status, presence of The most common type of gastric varix is the lesser curve ascites, albumin level, prothrombin time, size of FV, and the varix continuous with esophageal varices, 3 which is formed TABLE 2. Univariate Analysis of the Endoscopic Variables Related to Hemorrhage From FV No. of Patients % Who Bled x 2 Risk Ratio P Esophageal varices Present Absent ( ) Primary or secondary FV Primary Secondary ( ) Type of FV Type 1 FV Type 2 FV ( ) Size of FV Small Medium ( ) Large ( ) Red-spot on FV Present Absent ( ) AID Hepa 0059 / 5p1b$$ :30:35 hpta WBS: Hepatology

4 310 KIM ET AL. HEPATOLOGY February 1997 TABLE 3. Univariate Analysis of the Clinical and Biochemical by the deep submucosal veins arising from the left gastric Variables Related to Hemorrhage From FV vein. 1 Endoscopic sclerotherapy has been shown to be effective No. of % Who in controlling acute bleeding and in obliterating such Patients Bled x 2 Risk Ratio P varices. 3-6 Thus, most of the lesser curve gastric varices have been thought to be harmless. 3-6 However, FV develop because Sex of a dilation of the short gastric and posterior gastric veins Male or direct anastomotic veins between the gastric and retro- Female ( ) Cause of cirrhosis peritoneal veins, and are frequently associated with large Alcoholic gastrorenal shunts. 1 Hemorrhage from FV is serious and en- Nonalcoholic ( ) doscopic sclerotherapy sometimes fails to cease such hemor- Child s status rhage and obliterate FV, resulting in high mortality. 3-6 We A thus concentrated our attention on hemorrhage from FV. B ( ) In the current study, 61 of 132 patients had both FV and C ( ) HCC at the time of enrollment. The complications of HCC Ascites occur more often among patients with hepatitis B and C Absent related cirrhosis than among similar patients with alcoholic Mild ( ) Severe ( ) cirrhosis. 14 The cumulative frequency of HCC was reported Bilirubin (mg/dl) to be 60% 10 years after the detection of viral hepatitis õ related cirrhosis, 14,15 and we often encounter HCC in patients ( ) with cirrhosis and FV. Concomitant HCC has been suggested ú ( ) to be a risk factor for hemorrhage from esophageal vari- Albumin (g/dl) ces. 12,16,17 We thus assumed that the presence of HCC might ú also increase the frequency of hemorrhage from FV, and ( ) therefore included patients with concomitant HCC in the õ ( ) present study. Prothrombin time (%) ú It is generally believed that the overall incidence of hemor ( ) rhage from gastric varices is lower than that of esophageal õ ( ) varices. 3,7 The low incidence seems to be attributable to the Encephalopathy low frequency of hemorrhage from the gastric varices at the Absent lesser curve. However, several authors have reported an ex- Mild ( ) tremely high incidence of hemorrhage from FV. 3,18 It thus Moderate ( ) appears that bleeding is common in patients with FV as well HCC as in patients with large esophageal varices. However, previ- Present ous studies only analyzed a small number of patients with Absent ( ) FV, 3,7 and a type 2 statistical error might well have influenced the results. The current study, which included a large series of patients with FV, disclosed a cumulative bleeding rate at 1, 3, and 5 years of 16%, 36%, and 44%, respectively. The incidence of hemorrhage from risky esophageal varices was shown to be 19% to 40% with similar follow-up periods. 8,9,19 Thus, the incidence of hemorrhage from FV appears to be similar to that from moderate to large esophageal varices. However, the present results indicate that the bleeding risk of FV is not as high as has been previously suggested. 3,18 We analyzed the risk factors for hemorrhage from FV based on a retrospective and prospective investigation. In the retrospective study, we found the size of FV to be greater, the presence of red-spot on FV to be more prevalent, and the Child s status to be more severe in patients with hemorrhagic FV than in those patients enrolled in our study who had not yet demonstrated a hemorrhage from FV. In the prospective study, a multiple regression analysis (Cox s model) disclosed that Child s status, the size of FV, and the presence of red- TABLE 4. A Multivariate Analysis of the Clinical and Biochemical Variables Related to Hemorrhage From FV Regression Coefficient SE x 2 Risk Ratio P Size of FV ( )* ( ) Child s status ( ) ( ) Red-spot on FV ( ).047 FIG. 4. The cumulative percentage of the patients showing no first variceal hemorrhage from FV in relation to Child s status, the size of FV, and the presence or absence of red-spot on the FV. * Small vs. medium. Small vs. large. Class A vs. class B. Class A vs. class C. Absent vs. present. AID Hepa 0059 / 5p1b$$ :30:35 hpta WBS: Hepatology

5 HEPATOLOGY Vol. 25, No. 2, 1997 KIM ET AL. 311 TABLE 5. Estimated 1-Year Percentage Probability of Bleeding as a Function of All Possible Combinations of the Two Endoscopic Variables for Patients in Child Class A, B, or C Class A Class B Class C Gastric Varices Gastric Varices Gastric Varices Small Medium Large Small Medium Large Small Medium Large Red-spot Absent Present spot were all statistically significant and independent pre- This study disclosed the severity of liver disease, as defined dictors of hemorrhage from FV. Because the same risk factors by Child s status, to also be an important predictor for hemor- emerged from both the prospective and retrospective studies, rhage from FV. Patients with a prolonged prothrombin time, these variables thus appear to be reliable risk factors. ascites, jaundice, and impaired protein synthesis may be The size of FV is the most important risk factor for a hemor- prone to develop a hemorrhage from FV. In addition, patients rhage from FV. The present findings correlate with the results with advanced cirrhosis have a greater magnitude of portal from a retrospective study by Hashizume et al. 7 The hypertension, and Child s status is associated with portal size of the esophageal varices is also an important risk factor venous pressure. 25 It is well known that portal venous pres- for hemorrhage from esophageal varices. 8,19 The portal and sure is an important risk factor for a rupture of esophageal variceal pressure is higher and the blood flow of variceal varices. 20 It is thus more likely that Child s status may indi- vessels is greater in patients with larger esophageal varices. 20 rectly denote the magnitude of portal hypertension influencing High variceal pressure is thought to be the main causative hemorrhage from FV. factor of hemorrhage from esophageal varices. 21 In addition Because hemorrhage from FV is more serious and the mortality to variceal pressure, the size of the varices is an important rate is higher than in those with esophageal varices and determinant of the tension exerted on the varix wall. 21 A high lesser curvature gastric varices, 3-6 the estimation of bleeding degree of variceal wall tension has also been suggested to probability may therefore be beneficial in the clinical man- be an important risk factor for hemorrhage from esophageal agement of patients with high-risk FV. 8 The prognostic index, varices. 21 Although we could not measure the pressure or which identifies the groups of patients with a 2-year probabil- blood flow of FV, it is plausible that larger FV may have a ity of bleeding ranging from 5% to 65%, may serve to roughly greater blood flow and/or wall tension and thus result in a identify patients at very high risk of hemorrhage from FV high incidence of hemorrhage. who need a more careful observation. Because some patients We found the presence of a red-spot on FV to be an independent considered to be at low risk do bleed and the number of veryrisk bleeding risk factor. The importance of this sign as a high-risk patients who can be identified with the current factor for hemorrhage from gastric varices has also been model is relatively small, other unknown factors may com- reported by Hashizume et al. 7 The red-color sign, which corre- bine to cause hemorrhage from FV. Most likely, the integration sponds to the dilated, blood-filled channels lying within and of such information as variceal pressure, variceal blood beneath the mucosal epithelium, 22 is also a well-known risk flow, and variceal wall tension may in future help to improve factor for hemorrhage from esophageal varices. 8,19 However, the prognostic index. 21 the vascular structures of the stomach are not the same as Based on our findings, we conclude that the incidence of those of the esophagus, 23,24 and gastric varices usually lie in hemorrhage from FV appears to be similar to that from mod- the gastric submucosa. 22 Thus, a red-color sign is seldom erate-to-large esophageal varices, and that the bleeding risk found in patients with gastric varices. In the current study, of FV is not as high as has been suggested previously. The red-spot was defined as a localized reddish mucosal area or size of varices and red-spots on varices are important risk spots on FV. 7 Although the histological features of red-spot factors. A less-preserved hepatic functional reserve is also a on FV remain unclear, red-spot may denote an injury of the significant and independent predictor. High-risk patients, as gastric mucosa and may influence hemorrhage from FV. defined by a high probability index, therefore need more care- ful observation than those without such risk factors. Acknowledgment: We thank Dr. Katsuro Tachibana for his valuable discussions. REFERENCES 1. Watanabe K, Kimura K, Matsutani S, Ohta M, Okuda K. Portal hemodynamics in patients with gastric varices: a study in 230 patients with esophageal varices and/or gastric varices using portal vein catheterization. Gastroenterology 1988;95: Ohnishi K, Saito M, Sato S, Nakayama T, Takashi M, Iida T, Nomura F, et al. Direction of splenic venous flow assessed by pulsed Doppler flowmetry in patients with a large splenorenal shunt: relation to spontaneous hepatic encephalopathy. Gastroenterology 1985;89: Sarin SK, Lahoti D, Saxena SP, Murthy NS, Makwana U. Prevalence, classification and natural history of gastric varices: a long-term follow-up study in 568 portal hypertension patients. HEPATOLOGY 1992;16: Hosking SW, Johnson AG. Gastric varices: a proposed classification of gastric varices leading to management. Br J Surg 1988;75: Sarin SK, Sachdev G, Nanda R, Misra SP, Broor SL. Endoscopic sclerotherapy in the treatment of gastric varices. Br J Surg 1988;75: Trudeau W, Prindiville T. Endoscopic injection sclerosis in bleeding gastric FIG. 5. Estimated probability of remaining free from bleeding caused by varices. Gastrointest Endosc 1986;32: FV (survival function, Cox s model) for patients with different values for the 7. Hashizume M, Kitano S, Yamaga H, Koyanagi N, Sugimachi K. Endoscopic prognostic index. PI, prognostic index. classification of gastric varices. Gastrointest Endosc 1990;36: AID Hepa 0059 / 5p1b$$ :30:35 hpta WBS: Hepatology

6 312 KIM ET AL. HEPATOLOGY February The North Italian Endoscopic Club for the study and treatment of esopha- 17. Hashizume M, Kitano S, Koyanagi N, Tanoue K, Ohta M, Wada H, Yamaga geal varices. Prediction of the first variceal hemorrhage in patients with H, et al. Endoscopic injection sclerotherapy for 1000 patients with esophacirrhosis of the liver and esophageal varices: a prospective multicentric geal varices: a nine-year prospective study. HEPATOLOGY 1992;15: study. N Engl J Med 1988;319: McCormick PA, Jenkins SA, McIntyre N, Burrough AK. Why portal hyper- 9. Pagliaro L, D Amico G, Sörensen TIA, Lebrec D, Burroughs AK, Morabito tensive varices bleed and bleed: a hypothesis. Gut 1955;36: A, Tiné F, et al. Prevention of first bleeding in cirrhosis: a meta-analysis 19. Beppu K, Inokuchi K, Koyanagi N, Nakayama S, Sakata H, Kitano S, of randomized trials of nonsurgical treatment. Ann Intern Med 1992;117: Kobayashi M. Prediction of variceal hemorrhage by esophageal endoscopy Gastrointest Endosc 1981;27: Pugh RNH, Murray-Lyon IM, Dawson JL, Pietroni MC, Williams R. Tran- 20. Garcia-Tsao G, Groszmann RJ, Fisher RL, Conn HO, Atterbury CE, Glicksection of oesophagus for bleeding oesophageal varices. Br J Surg 1973; man M. Portal pressure, presence of gastroesophageal varices and variceal 60: bleeding. HEPATOLOGY 1985;5: Christensen E. Multivariate survival analysis using Cox s regression 21. Rigau J, Bosch J, Bordas JM, Navasa M, Mastai R, Kravetz D, Bruix J, model. HEPATOLOGY 1987;7: et al. Endoscopic measurement of variceal pressure in cirrhosis: correlation 12. Kokawa H, Shijo H, Kubara K, Nakaoka K, Toriya H, Shirai Z, Okazaki with portal pressure and variceal hemorrhage. Gastroenterology 1989;96: M. Long-term risk factors for bleeding after first course of endoscopic injection sclerotherapy: a univariate and multivariate analysis. Am J Gas- 22. Spence RAJ. The venous anatomy of the lower oesophagus in normal subtroenterol 1993;88: jects and in patients with varices: an image analysis study. Br J Surg 13. Harrell FE. The PHGLM procedure. In: SUGI supplemental library user s 1984;71: guide. Ed 5. Cary: SAS Institute Inc., 1986: Zaman SN, Melia W, Johnson RD, Portmann BC, Johnson PJ, Williams 23. Hashizume M, Kitano S, Sugimachi K, Sueishi K. Three-dimensional view R. Risk factors in development of hepatocellular carcinoma in cirrhosis: of the vascular structure of the lower esophagus in clinical portal hypertenprospective study of 613 patients. Lancet 1985;i: sion. HEPATOLOGY 1988;8: Kew MC, Popper H. Relationship between hepatocellular carcinoma and 24. Hashizume M, Tanaka K, Inokuchi K. Morphology of gastric microcirculacirrhosis. Semin Liver Dis 1984;4: tion in cirrhosis. HEPATOLOGY 1983;3: Poynard T, Lebrec D, Hillon P, Sayegh R, Bernuau J, Naveau S, Chaput 25. Graillon R, Cales P, Valla D, Gaudy D, Geoffroy P, Lebrec D. Influence of JC, et al. Propranolol for prevention of recurrent gastrointestinal bleeding the degree of liver failure on systemic and splanchnic haemodynamics and in patients with cirrhosis: a prospective study of factors associated with rebleeding. HEPATOLOGY 1977;7: on response to propranolol in patients with cirrhosis. Gut 1986;27: AID Hepa 0059 / 5p1b$$ :30:35 hpta WBS: Hepatology