HEPATOLOGY else when^
Vol. 12, No. 3, 1990
months of the procedure (a mean of 4.8 months), four were loet to follow-up by four months, and three were alive at 12 months. Of the eight patients who did not have ES, five were alive at one to six months and three died within five months (mean 1.7 months) after the procedure. This paper expands our knowledge of biliary tract abnormalitiesin Ew-infectedindividuals and mggests that ES may improve biliary tract 8880ciated abdominal pain. COMKENTS
619
cholestatic liver tests, the diagnosis of acalculous cholecystitis ( 6 ) should be considered because it has a significant mortality (7). Biliary tract disorders that are not HIV specific (i.e., gallstone disease) can also occur in the presence of HIV disease and should not be forgotten. Bile duct imaging studies should precede endoscopic retrograde cholangiopancreatography (ERCP) in patients with cholestatic liver tests, and, if normal, suggest that a liver biopsy may be more helpful than endoscopy. Biliary tree dilation is an indication for ERCP, at which time ampullary biopsies (for histology and culture) and bile aspiration (for cytology, cytomegalovirus culture and examination for parasites) should be done. ERCP should be considered in patients with normal imaging studies if the liver biopsy specimen suggests (and hopefully not) extrahepatic obstruction. As with many clinical studies exploring new territories, more questions are raised than answered. What is the prevalence of AIDS cholangiopathy? What is the role of papillotomy in the various subgroups of AIDS cholangiopathy? Do patients with papillary stenosis alone do best? Should asymptomatic patients with dilated ducts have a papillotomy? The role of endoscopic sphincterotomy as therapy for AIDS cholangiopathy and as a means of prolonging survival is yet to be determined.
Sclerosing lesions of the biliary tract are multifactorial in origin (1). Primary sclerosing cholangitis (PSC) appears to be an immunological disorder, and biliary tract sclerosis has been reported in a wide variety of disorders, including histiocytosis X, sarcoidosis, ulcerative colitis and Crohn's disease. It also occurs after injection of scolicidal solution into the bile duct, after hepatic artery embolization or 5-fluorodeoxyuridine infusion and with congenital and acquired immunodeficiency states, such as dysgammaglobulinemiaand graft vs. host disease (1). AIDS is the latest immunological disorder to be associated with sclerosing lesions of the biliary tract. AIDS cholangiopathy represents a spectrum of disorders of multiple etiologies. It probably is not due to HIV per se but rather to the resultant immunodeficiency state after infection with the virus. AIDS cholangiopathy has been reported in intravenous DOUGLAS SIMON, M.D., F.A.C.G. drug users, transfusion recipients, male homosexuals Bronx Municipal Hospital Center and subjects with HIV-2 infection (2), although it most Albert Einstein College of Medicine often appears to be a result of opportunistic pathogens, LAWRENCE J. BRANDT, M.D., F.A.C.G. particularly cryptosporidium and cytomegalovirus. Montefiore Medical Center Cryptosporidium has also been reported to cause scleAlbert Einstein College of Medicine rosing cholangitis in congenital immunodeficiency (3). Bronx, New York 10467 Cytomegalovirus is commonly identified in ampullary biopsy specimens (5 of 12 in this series) and may induce REFERENCES a vasculitis with subsequent ischemia, ulceration and 1. Sherlock S. The syndrome of disappearing bile duds. Lancet warring (4) of the bile duct epithelium. This sequence 198;2:439-496. has been reported in the biliary scarring after liver 2. Roulot D, Valla D, Brun-Vezinet F, Rey M, Clavel F, Degott C, transplantation (5) and also has been identified in Guillian J, et al. Cholangitis in the acquired immunodeficiency syndrome: report of two cases and review of the literature. Cut autopsies of patients with AIDS cholangiopathy (2, 4). 1987;8:1653-1660. The cholangiopathy of AIDS is clearly distinct from 3. Davis J, Heyman M, Farrell L, Kerner J, Kerlan R, Thaler M. that of PSC. In AIDS sclerosing cholangitis intrahepatic Sclerosing cholangitis associated with chronic cryp~poridiosisin ducts show narrowing and beading as in PSC, but unlike a child with congenital immunodeficiency disorder. Am J GastroPSC common bile duct dilatation, papillary stenosis and enteml 1987;82:1196-1202. bacterial infection also occur. Histological changes of the 4. Margolis S, Lonig C, Soave R, Govioni A, Mouradian J, Jacobson I. Biliary tract obstruction in the acquired immunodeficiency synbile duct in AIDS cholangiopathy reveal chronic inflamdrome. Ann Intern Med 1986;105:207-210. mation with focal mucosal ulcerations, whereas the liver 5. Martineau G , Porter K, Corman J, Launois B, Schroter G, Palmer exhibits distorted interlobular bile ducts and portal W, Putnam C, et al. Delayed biliary tract obstruction after orthotopic liver transplant. Surgery 1972;72:604-610. fibrosis. Many of the features of the papillary stenosis seen in this syndrome are similar to sphincter of Oddi 6. Blumberg R, Kelsey P, Perrone T, Dickerson R, LaQuagha M, Ferruci J. Cytomegalovirus and cryptosporidium associated acaldysfunction, including a dilated biliary tree, delayed culous gangrenous cholecystitis. Am J Med 1984;76:1118-1123. contrast drainage, biliary-type symptoms and abnormal 7. LaReJa R, Rotherberg R, Odom J, Mueller S. The incidence of liver tests. Manometric studies have not been reported intra-abdominal surgery in acquired immunodeficiency syndrome: a statistical review of 904 patients. Surgery 1989;105:175-179. in this situation, but the relief of pain, improved contrast drainage and decreased size of the biliary tree after CORTICOSTEROJDTHERAPY OF ALCOHOLIC endoscopic sphincterotomy suggest that such functional HEPATITIS: HOW MANY STUDIES WILL IT TAgE? abnormalities may occur. The evaluation of AIDS patients for cholangiopathy is Carithers RL Jr., Herlong HF, Diehl AM, Shaw EW, similar to that in non-HIV-infected individuals. In any Combes B, Fallon HJ, Maddrey WC. MethylpredAIDS patient with fever, right upper quadrant pain and nisolone therapy in patients with severe alcoholic hep-
620
HEPATOLOGY Elsewhere
atitis: a randomized multicenter trial. Ann Intern Med 1989;110~685-690. ABSTRACT Sfudy Objective: To determine the efficacy of a corticosteroid in reducing the short-termmortality of patients with 8evere alcoholic hepatitis. Design: Randomized, double-blind, placebocontrolled multicenter trial. setting: Four university teaching hospitals. Patients: We enrolled 66 patients with alcoholic hepatitis and either spontaneous hepatic encephalopathy or a discriminant function value greater than 32, calculated using the formula: 4.6(prothrombin time - control time) + serum bilirubin [in pmoUL]/17.1. Fifty-nine patients (89%)completed the study. Two patients withdrew from the trial. The other 64 patients were hospitalized for the duration of the trial: however, treatment was discontinued in 5 patients because of potential drug toxicity. Interventions: Patients were randomly assigned to receive either methylprednisolone (32 mg) or placebo within 7 days of admission. Treatment was given for 28 days. The doses were then tapered over 2 weeks and discontinued. Measurements and Main Results: The endpoint of the study was death. Of the 31 recipients of placebo, 11 (35%)died within 28 days of randomization compared with 2 (6%)of the 36 patients given methylprednisolone ( p = 0.006). The 95% CI for the difference in mortality was 12% to 70%. In the patients with spontaneous hepatic encephalopathy at entry, 9 of 19 recipients of placebo died (47%)compared with 1 (7%)of the 14 patients given methylprednisolone ( p = 0.02). The 95% CI for the difference in mortality was 14% to 66%.The Cox proportionalhazards regression model showed the advantage of methylprednisolone over placebo after adjustment for other potentidy important prognostic variables ( p = 0.004). cOncluuwns: Methylprednisolone therapy decreases short-term mortality in patients with severe alcoholic hepatitis manifested either by spontaneous hepatic encephalopathy or a markedly elevated discriminant function value.
COMMENTS
This is the twelfth published randomized controlled trial (RCT) of short-term corticosteroid treatment for acute alcoholic hepatitis. Why so many? The reason is the inconsistent results of these trials -mostly negative but occasionally tantalizingly positive. This is a relatively common disorder that begs for some sort of effective treatment. Corticosteroid offers the theoretical benefits of suppressing hepatic inflammation, reducing collagen formation, increasing albumin production, enhancing appetite and, if the hepatic damage is in any respect immunologically mediated, perhaps suppressing this. The initial RCT from North Carolina, published in 1971 (11,enrolled 37 patients all with liver biopsy showing acute alcoholic hepatitis before randomization. Treatment with prednisolone, 40 mg/day, or placebo lasted 6 weeks. During a 3-month follow-up, mortality was 6 of 17 among placebo-treated patients
HEPATOLOGY
compared to 1 of 20 in those given prednisolone. In the more severely ill patients with hepatic encephalopathy at randomization, the results were even more favorable for prednisolone treatment with mortality 1 of 9 vs. 6 of 6 in controls (p < 0.01).Additional RCTs soon followed from Seattle, Los Angeles, West Haven and Boston, using various types of corticosteroid, given for varying periods (26 to 42 days). All showed negative results. Taken together the four trials showed a mortality of 29 of 64 (45%) among controls and 25 of 55 (45%) in patients treated with corticosteroid. Not disheartened, the North Carolina investigators published a second RCT in 1978 designed to compare prednisolone treatment with caloric supplementation in patients with severe alcoholic hepatitis and spontaneous encephalopathy (2). Again there seemed to be a beneficial effect from corticosteroid with mortality 2 of 7 in treated patients compared with 7 of 7 in controls (p < 0.01). From 1978 to 1987 there were five more RCTs published, all with negative results. One enrolled only patients with spontaneous hepatic encephalopathy, yet mortality was virtually identical in treated patients (53%) and controls (54%) (3).In another RCT from Baltimore, the overall results showed no significant difference between prednisolone and placebo-treated patients, but a subgroup analysis of patients judged to have a poor prognosis because of high bilirubin and low prothrombin values showed a lower mortality with prednisolone than in controls (1of 7 vs. 6 of 8, p = 0.03) (4).The largest RCT of all was conducted in a number of United States Veterans Administration hospitals with 90 patients randomized to 30 days of prednisolone treatment while 88 patients were given placebo (5). This trial had a good statistical probability (89%)of being able to detect a 20% improvement in short-term mortality. However, no difference was found between treated patients and controls (21% vs. 22%). Long-term mortality was similarly unaffected by prednisolone treatment. This RCT is difficult to fault on methodological grounds and seemed, for many, to settle the issue of corticosteroid benefit. Meanwhile the authors of previous positive or near-positive trials had organized the currently discussed RCT in four centers -Johns Hopkins, Medical College of Virginia, Thomas Jefferson and Southwestern. They enrolled only patients with a relatively poor prognosis based on either the presence of spontaneous hepatic encephalopathy or a discriminant function based on prothrombin and bilirubin values (4). Patients with potentially confounding variables such as gastrointestinal bleeding and active infection were excluded. Their results are convincingly in favor of benefit from corticosteroid in the subgroup of patients they treated. The severity of disease seemed comparable in the treated patients and controls except for a slightly higher serum bilirubin (1 mg/dl) and greater frequency of ascites (20of 31 patients vs. 14 of 35 patients) in the control group. Seven patients were withdrawn from their assigned treatment for various reasons, but only one was lost to follow-up and survival analysis was by the intention-to-treat strategy. There was more rapid
Vol. 12, No. 3, 1990
62 1
HEPATOLOGY Elsewhere
improvement in mean laboratory test results (prothrombin, AST, bilirubin, albumin) in steroid-treated patients, but this was statistically significant only for the first two. It would have been of interest to compare the progression of mean test results with the exclusion of patients who died. The inconsistency in the results of the corticosteroid trials are disconcerting and difficult to explain. Positive treatment trials should have more meaning than negative ones because of the large number of subjects needed in negative trials to exclude confidently the possibility of missing a true benefit (“type 11” error). Only the Veterans Administration multicenter trial contained enough patients to make it statistically unlikely that a true beneficial treatment effect had been overlooked. Other potentially confounding variables that might explain inconsistent results from corticosteroid treatment are discussed by Carithers and colleagues and include differences in age and sex of the patients and accidents of randomization in small trials. Meta-analysis is a technique for combining results of multiple RCTs to increase the statistical power in an effort to detect true treatment effects. It requires a clear formulation of the objectives of the analysis, welldefined inclusion criteria and a certain degree of homogeneity of results. Two groups have recently published the results of meta-analysis of the RCTs of corticosteroid treatment in acute alcoholic hepatitis, including the report of Carithers et al. From analysis of 11RCTs that met inclusion criteria, a “working team” reported to the 1988 International Gastroenterology Congress in Rome an improvement in short-term mortality with corticosteroid averaging 16%(confidence interval - 17 to 46%) (6). In the subgroup of patients with spontaneous encephalopathy, the average gain in survival with corticosteroid was 28% (confidence interval -6.5 to 62%).From meta-analysis of the same 11 RCTs, Imperide and McCullough (7) concluded that overall benefit from corticosteroid treatment, which they termed protective efficacy, was 37% (confidence interval 32% to 41%)with 55% protective efficacy (confidence interval 50% to 59%) in the subgroup with hepatic encephalopathy. The difference in the results of the two metaanalyses is more apparent than real, I believe, and is
largely due to the manner of expressing the improvement in survival. In the first meta-analysis it is expressed as the difference in percent mortality between controls and treated patients, whereas in the second the improvement in survival is expressed as a percentage of the mortality in the control group. Whether the accumulated data from the 12 RCTs and their meta-analyses are sufficient to establish corticosteroid treatment as the standard of practice for acute alcoholic hepatitis is problematic. The skeptic may well ask, if corticosteroid treatment reduces the mortality in this disorder, why isn’t there a demonstrable benefit in terms of shorter hospitalization and more rapid improvement in liver test results in less ill patients? None of the trials has shown this. My own view is that another trial showing results similar to the Carithers report will be needed before corticosteroid treatment will be widely adopted.
TELFERB. REYNOLDS,M.D. University of Southern California School of Medicine Las Angeles, California 90033 REFERENCES 1. Helman
2.
3.
4. 5.
6. 7.
RA, Temko MH, Nye SW, Fallon HJ. Alcoholic hepatitis:
natural history and evaluation of prednisolone therapy. Ann Intern Med 1971;74:311-321. Lesesne HR, Bozymsh EM, Fallon HJ. Treatment of alcoholic hepatitis with encephalopathy: comparisons of prednisolone with caloric supplements. Gastroenterology 1978;74:169-173. Depew W, Boyer T, Omata M, Redeker A, Reynolds T. Double-blind controlled trial of prednisolone therapy in patients with severe alcoholic hepatitis and spontaneous encephalopathy. GastroenterOIOW 1980;78:524-529. Maddrey WC, Boitnott JK, W i n e MS, Weber FLJ, Mezey E, White RI Jr. Corticosteroid therapy of alcoholic hepatitis. Gastroenterology 1978;75:193-199. Mendenhall CL, Anderson S, Garcia-Pont P, Goldberg S, Kiernan T, Seeff LB, Sorrell M, et al. Short-term and long-term survival in patients with alcoholic hepatitis treated with oxandrolone and prednisolone. N Engl J Med 1984;311:1464-1470. Reynolds TB, Benhamou J-P, Blake J, Naccarato R, Orrego H. Treatment of acute alcoholic hepatitis. Gastroenterology Int 1989; 2:208-216. Imperiale TF, McCullough AJ. Corticosteroids reduce mortality from alcoholic hepatitis: a meta-analysis of the randomized trials IAbstract]. HEPAMLCGY 1989;10:580.
Vol. 12, No. 3, 1990
months of the procedure (a mean of 4.8 months), four were loet to follow-up by four months, and three were alive at 12 months. Of the eight patients who did not have ES, five were alive at one to six months and three died within five months (mean 1.7 months) after the procedure. This paper expands our knowledge of biliary tract abnormalitiesin Ew-infectedindividuals and mggests that ES may improve biliary tract 8880ciated abdominal pain. COMKENTS
619
cholestatic liver tests, the diagnosis of acalculous cholecystitis ( 6 ) should be considered because it has a significant mortality (7). Biliary tract disorders that are not HIV specific (i.e., gallstone disease) can also occur in the presence of HIV disease and should not be forgotten. Bile duct imaging studies should precede endoscopic retrograde cholangiopancreatography (ERCP) in patients with cholestatic liver tests, and, if normal, suggest that a liver biopsy may be more helpful than endoscopy. Biliary tree dilation is an indication for ERCP, at which time ampullary biopsies (for histology and culture) and bile aspiration (for cytology, cytomegalovirus culture and examination for parasites) should be done. ERCP should be considered in patients with normal imaging studies if the liver biopsy specimen suggests (and hopefully not) extrahepatic obstruction. As with many clinical studies exploring new territories, more questions are raised than answered. What is the prevalence of AIDS cholangiopathy? What is the role of papillotomy in the various subgroups of AIDS cholangiopathy? Do patients with papillary stenosis alone do best? Should asymptomatic patients with dilated ducts have a papillotomy? The role of endoscopic sphincterotomy as therapy for AIDS cholangiopathy and as a means of prolonging survival is yet to be determined.
Sclerosing lesions of the biliary tract are multifactorial in origin (1). Primary sclerosing cholangitis (PSC) appears to be an immunological disorder, and biliary tract sclerosis has been reported in a wide variety of disorders, including histiocytosis X, sarcoidosis, ulcerative colitis and Crohn's disease. It also occurs after injection of scolicidal solution into the bile duct, after hepatic artery embolization or 5-fluorodeoxyuridine infusion and with congenital and acquired immunodeficiency states, such as dysgammaglobulinemiaand graft vs. host disease (1). AIDS is the latest immunological disorder to be associated with sclerosing lesions of the biliary tract. AIDS cholangiopathy represents a spectrum of disorders of multiple etiologies. It probably is not due to HIV per se but rather to the resultant immunodeficiency state after infection with the virus. AIDS cholangiopathy has been reported in intravenous DOUGLAS SIMON, M.D., F.A.C.G. drug users, transfusion recipients, male homosexuals Bronx Municipal Hospital Center and subjects with HIV-2 infection (2), although it most Albert Einstein College of Medicine often appears to be a result of opportunistic pathogens, LAWRENCE J. BRANDT, M.D., F.A.C.G. particularly cryptosporidium and cytomegalovirus. Montefiore Medical Center Cryptosporidium has also been reported to cause scleAlbert Einstein College of Medicine rosing cholangitis in congenital immunodeficiency (3). Bronx, New York 10467 Cytomegalovirus is commonly identified in ampullary biopsy specimens (5 of 12 in this series) and may induce REFERENCES a vasculitis with subsequent ischemia, ulceration and 1. Sherlock S. The syndrome of disappearing bile duds. Lancet warring (4) of the bile duct epithelium. This sequence 198;2:439-496. has been reported in the biliary scarring after liver 2. Roulot D, Valla D, Brun-Vezinet F, Rey M, Clavel F, Degott C, transplantation (5) and also has been identified in Guillian J, et al. Cholangitis in the acquired immunodeficiency syndrome: report of two cases and review of the literature. Cut autopsies of patients with AIDS cholangiopathy (2, 4). 1987;8:1653-1660. The cholangiopathy of AIDS is clearly distinct from 3. Davis J, Heyman M, Farrell L, Kerner J, Kerlan R, Thaler M. that of PSC. In AIDS sclerosing cholangitis intrahepatic Sclerosing cholangitis associated with chronic cryp~poridiosisin ducts show narrowing and beading as in PSC, but unlike a child with congenital immunodeficiency disorder. Am J GastroPSC common bile duct dilatation, papillary stenosis and enteml 1987;82:1196-1202. bacterial infection also occur. Histological changes of the 4. Margolis S, Lonig C, Soave R, Govioni A, Mouradian J, Jacobson I. Biliary tract obstruction in the acquired immunodeficiency synbile duct in AIDS cholangiopathy reveal chronic inflamdrome. Ann Intern Med 1986;105:207-210. mation with focal mucosal ulcerations, whereas the liver 5. Martineau G , Porter K, Corman J, Launois B, Schroter G, Palmer exhibits distorted interlobular bile ducts and portal W, Putnam C, et al. Delayed biliary tract obstruction after orthotopic liver transplant. Surgery 1972;72:604-610. fibrosis. Many of the features of the papillary stenosis seen in this syndrome are similar to sphincter of Oddi 6. Blumberg R, Kelsey P, Perrone T, Dickerson R, LaQuagha M, Ferruci J. Cytomegalovirus and cryptosporidium associated acaldysfunction, including a dilated biliary tree, delayed culous gangrenous cholecystitis. Am J Med 1984;76:1118-1123. contrast drainage, biliary-type symptoms and abnormal 7. LaReJa R, Rotherberg R, Odom J, Mueller S. The incidence of liver tests. Manometric studies have not been reported intra-abdominal surgery in acquired immunodeficiency syndrome: a statistical review of 904 patients. Surgery 1989;105:175-179. in this situation, but the relief of pain, improved contrast drainage and decreased size of the biliary tree after CORTICOSTEROJDTHERAPY OF ALCOHOLIC endoscopic sphincterotomy suggest that such functional HEPATITIS: HOW MANY STUDIES WILL IT TAgE? abnormalities may occur. The evaluation of AIDS patients for cholangiopathy is Carithers RL Jr., Herlong HF, Diehl AM, Shaw EW, similar to that in non-HIV-infected individuals. In any Combes B, Fallon HJ, Maddrey WC. MethylpredAIDS patient with fever, right upper quadrant pain and nisolone therapy in patients with severe alcoholic hep-
620
HEPATOLOGY Elsewhere
atitis: a randomized multicenter trial. Ann Intern Med 1989;110~685-690. ABSTRACT Sfudy Objective: To determine the efficacy of a corticosteroid in reducing the short-termmortality of patients with 8evere alcoholic hepatitis. Design: Randomized, double-blind, placebocontrolled multicenter trial. setting: Four university teaching hospitals. Patients: We enrolled 66 patients with alcoholic hepatitis and either spontaneous hepatic encephalopathy or a discriminant function value greater than 32, calculated using the formula: 4.6(prothrombin time - control time) + serum bilirubin [in pmoUL]/17.1. Fifty-nine patients (89%)completed the study. Two patients withdrew from the trial. The other 64 patients were hospitalized for the duration of the trial: however, treatment was discontinued in 5 patients because of potential drug toxicity. Interventions: Patients were randomly assigned to receive either methylprednisolone (32 mg) or placebo within 7 days of admission. Treatment was given for 28 days. The doses were then tapered over 2 weeks and discontinued. Measurements and Main Results: The endpoint of the study was death. Of the 31 recipients of placebo, 11 (35%)died within 28 days of randomization compared with 2 (6%)of the 36 patients given methylprednisolone ( p = 0.006). The 95% CI for the difference in mortality was 12% to 70%. In the patients with spontaneous hepatic encephalopathy at entry, 9 of 19 recipients of placebo died (47%)compared with 1 (7%)of the 14 patients given methylprednisolone ( p = 0.02). The 95% CI for the difference in mortality was 14% to 66%.The Cox proportionalhazards regression model showed the advantage of methylprednisolone over placebo after adjustment for other potentidy important prognostic variables ( p = 0.004). cOncluuwns: Methylprednisolone therapy decreases short-term mortality in patients with severe alcoholic hepatitis manifested either by spontaneous hepatic encephalopathy or a markedly elevated discriminant function value.
COMMENTS
This is the twelfth published randomized controlled trial (RCT) of short-term corticosteroid treatment for acute alcoholic hepatitis. Why so many? The reason is the inconsistent results of these trials -mostly negative but occasionally tantalizingly positive. This is a relatively common disorder that begs for some sort of effective treatment. Corticosteroid offers the theoretical benefits of suppressing hepatic inflammation, reducing collagen formation, increasing albumin production, enhancing appetite and, if the hepatic damage is in any respect immunologically mediated, perhaps suppressing this. The initial RCT from North Carolina, published in 1971 (11,enrolled 37 patients all with liver biopsy showing acute alcoholic hepatitis before randomization. Treatment with prednisolone, 40 mg/day, or placebo lasted 6 weeks. During a 3-month follow-up, mortality was 6 of 17 among placebo-treated patients
HEPATOLOGY
compared to 1 of 20 in those given prednisolone. In the more severely ill patients with hepatic encephalopathy at randomization, the results were even more favorable for prednisolone treatment with mortality 1 of 9 vs. 6 of 6 in controls (p < 0.01).Additional RCTs soon followed from Seattle, Los Angeles, West Haven and Boston, using various types of corticosteroid, given for varying periods (26 to 42 days). All showed negative results. Taken together the four trials showed a mortality of 29 of 64 (45%) among controls and 25 of 55 (45%) in patients treated with corticosteroid. Not disheartened, the North Carolina investigators published a second RCT in 1978 designed to compare prednisolone treatment with caloric supplementation in patients with severe alcoholic hepatitis and spontaneous encephalopathy (2). Again there seemed to be a beneficial effect from corticosteroid with mortality 2 of 7 in treated patients compared with 7 of 7 in controls (p < 0.01). From 1978 to 1987 there were five more RCTs published, all with negative results. One enrolled only patients with spontaneous hepatic encephalopathy, yet mortality was virtually identical in treated patients (53%) and controls (54%) (3).In another RCT from Baltimore, the overall results showed no significant difference between prednisolone and placebo-treated patients, but a subgroup analysis of patients judged to have a poor prognosis because of high bilirubin and low prothrombin values showed a lower mortality with prednisolone than in controls (1of 7 vs. 6 of 8, p = 0.03) (4).The largest RCT of all was conducted in a number of United States Veterans Administration hospitals with 90 patients randomized to 30 days of prednisolone treatment while 88 patients were given placebo (5). This trial had a good statistical probability (89%)of being able to detect a 20% improvement in short-term mortality. However, no difference was found between treated patients and controls (21% vs. 22%). Long-term mortality was similarly unaffected by prednisolone treatment. This RCT is difficult to fault on methodological grounds and seemed, for many, to settle the issue of corticosteroid benefit. Meanwhile the authors of previous positive or near-positive trials had organized the currently discussed RCT in four centers -Johns Hopkins, Medical College of Virginia, Thomas Jefferson and Southwestern. They enrolled only patients with a relatively poor prognosis based on either the presence of spontaneous hepatic encephalopathy or a discriminant function based on prothrombin and bilirubin values (4). Patients with potentially confounding variables such as gastrointestinal bleeding and active infection were excluded. Their results are convincingly in favor of benefit from corticosteroid in the subgroup of patients they treated. The severity of disease seemed comparable in the treated patients and controls except for a slightly higher serum bilirubin (1 mg/dl) and greater frequency of ascites (20of 31 patients vs. 14 of 35 patients) in the control group. Seven patients were withdrawn from their assigned treatment for various reasons, but only one was lost to follow-up and survival analysis was by the intention-to-treat strategy. There was more rapid
Vol. 12, No. 3, 1990
62 1
HEPATOLOGY Elsewhere
improvement in mean laboratory test results (prothrombin, AST, bilirubin, albumin) in steroid-treated patients, but this was statistically significant only for the first two. It would have been of interest to compare the progression of mean test results with the exclusion of patients who died. The inconsistency in the results of the corticosteroid trials are disconcerting and difficult to explain. Positive treatment trials should have more meaning than negative ones because of the large number of subjects needed in negative trials to exclude confidently the possibility of missing a true benefit (“type 11” error). Only the Veterans Administration multicenter trial contained enough patients to make it statistically unlikely that a true beneficial treatment effect had been overlooked. Other potentially confounding variables that might explain inconsistent results from corticosteroid treatment are discussed by Carithers and colleagues and include differences in age and sex of the patients and accidents of randomization in small trials. Meta-analysis is a technique for combining results of multiple RCTs to increase the statistical power in an effort to detect true treatment effects. It requires a clear formulation of the objectives of the analysis, welldefined inclusion criteria and a certain degree of homogeneity of results. Two groups have recently published the results of meta-analysis of the RCTs of corticosteroid treatment in acute alcoholic hepatitis, including the report of Carithers et al. From analysis of 11RCTs that met inclusion criteria, a “working team” reported to the 1988 International Gastroenterology Congress in Rome an improvement in short-term mortality with corticosteroid averaging 16%(confidence interval - 17 to 46%) (6). In the subgroup of patients with spontaneous encephalopathy, the average gain in survival with corticosteroid was 28% (confidence interval -6.5 to 62%).From meta-analysis of the same 11 RCTs, Imperide and McCullough (7) concluded that overall benefit from corticosteroid treatment, which they termed protective efficacy, was 37% (confidence interval 32% to 41%)with 55% protective efficacy (confidence interval 50% to 59%) in the subgroup with hepatic encephalopathy. The difference in the results of the two metaanalyses is more apparent than real, I believe, and is
largely due to the manner of expressing the improvement in survival. In the first meta-analysis it is expressed as the difference in percent mortality between controls and treated patients, whereas in the second the improvement in survival is expressed as a percentage of the mortality in the control group. Whether the accumulated data from the 12 RCTs and their meta-analyses are sufficient to establish corticosteroid treatment as the standard of practice for acute alcoholic hepatitis is problematic. The skeptic may well ask, if corticosteroid treatment reduces the mortality in this disorder, why isn’t there a demonstrable benefit in terms of shorter hospitalization and more rapid improvement in liver test results in less ill patients? None of the trials has shown this. My own view is that another trial showing results similar to the Carithers report will be needed before corticosteroid treatment will be widely adopted.
TELFERB. REYNOLDS,M.D. University of Southern California School of Medicine Las Angeles, California 90033 REFERENCES 1. Helman
2.
3.
4. 5.
6. 7.
RA, Temko MH, Nye SW, Fallon HJ. Alcoholic hepatitis:
natural history and evaluation of prednisolone therapy. Ann Intern Med 1971;74:311-321. Lesesne HR, Bozymsh EM, Fallon HJ. Treatment of alcoholic hepatitis with encephalopathy: comparisons of prednisolone with caloric supplements. Gastroenterology 1978;74:169-173. Depew W, Boyer T, Omata M, Redeker A, Reynolds T. Double-blind controlled trial of prednisolone therapy in patients with severe alcoholic hepatitis and spontaneous encephalopathy. GastroenterOIOW 1980;78:524-529. Maddrey WC, Boitnott JK, W i n e MS, Weber FLJ, Mezey E, White RI Jr. Corticosteroid therapy of alcoholic hepatitis. Gastroenterology 1978;75:193-199. Mendenhall CL, Anderson S, Garcia-Pont P, Goldberg S, Kiernan T, Seeff LB, Sorrell M, et al. Short-term and long-term survival in patients with alcoholic hepatitis treated with oxandrolone and prednisolone. N Engl J Med 1984;311:1464-1470. Reynolds TB, Benhamou J-P, Blake J, Naccarato R, Orrego H. Treatment of acute alcoholic hepatitis. Gastroenterology Int 1989; 2:208-216. Imperiale TF, McCullough AJ. Corticosteroids reduce mortality from alcoholic hepatitis: a meta-analysis of the randomized trials IAbstract]. HEPAMLCGY 1989;10:580.
