1096
HEPATOLOGY Elsewhere
radiation to damage DNA stop in G1 before they replicate their DNA in S phase. However, when they express mutant p53 they proceed through G1 and replicate their DNA despite the damage (4). This arrest could allow the cell time to repair its DNA or, if irreversibly damaged, result in permanent clonal deletion. This hypothesis makes a great deal of intuitive sense when all of the data are examined. The lack of such a screening function should not interfere with normal cell growth or differentiation. When exposed to carcinogens, however, cells lacking p53 would attempt to replicate their damaged DNA, immortalizing the alteration, whereas cells with wild-type p53 would be locked in a growth-arrested state and fail to pass on the acquired defects. How is this consistent with the growth arrest in G1 found when tumor cells are transfected with wild-type p53? Tumor cells may have a high level of ongoing genetic abnormalities that normal p53 responds to by inducing growth arrest, or growth arrest may be an artifact of the artificial overexpression of p53. Inequalities in p63 Abnormalities. A p53 knockout may not be equivalent to p53 point mutation. Missense mutations constitute the largest group of naturally acquired p53 abnormalities in tumors and are found to occur throughout the open reading frame; however, they are mostly in the 200 codon region spanning exons 5 through 8. Nonsense (stop) and splicing mutations, frame shifts and deletions of part or all of the open reading frame are also found, but deletions of p53 are actually a very uncommon event. By far the most common alteration is the missense point mutation, implying that mutant p53 commonly plays an active role in human cancer. The high frequency of missense mutations is not just due to a general increased probability of such events, because the types of mutations found in the APC tumor suppressor gene in familial polyposis tend to be deletions or nonsense (stop) mutations. Functionally, mutant p53 can cooperate with ras to transform primary rat embryo fibroblasts and thus appears to have a dominant transforming capability in addition to the tumor-suppressing phenotype. All missense mutations are not equivalent either because both transforming activity and transcriptional activities vary among various different missense proteins. A different spectrum of affected codons in different tumors seems to exist, with a preference for codon 249 in HCC, codon 175 in colon carcinoma and codons 150 to 160 (hot spot IIa) (5) in lung cancer. The latter two are nearly mutually exclusive, with codon 175 mutations being rare in lung cancer and IIa mutations rare in colon cancers. It is possible that the various mutant p53 proteins associated with different tumors affect the ability of the resultant protein to detect genomic alterations specifically associated with the different tumor types, for example those caused by particular carcinogens. The p53 “knockout” may thus be an imperfect model for early events leading to cancer. Interesting questions to ask with these p53 knockout mice include whether these mice have an altered sensitivity to carcinogens, whether the spectrum of
HEPATOLOGY
tumor types resulting from carcinogen treatment is different from normal mice and whether their cells exhibit increased sensitivity to oncogenic viruses that produce proteins that normally inactivate p53. Are there any detectable abnormalities of cell-cycle control, signal transduction, response to stress, cell division, DNA repair or cytogenetic abnormalities? Will tumors from these mice respond to the introduction of wild-type p53? Are they defective in antitumor immunity? The answers to these questions are forthcoming and will provide valuable insight into the role of p53 in normal and tumor cells.
DAVID CARBONE, M.D., PH.D. National Cancer Institute NCI-Navy Medical Oncology Branch Bethesda, Maryland 20889-5105 REFERENCES 1. Lavigueur A, Maltby V, Mock D, Rossant J , Pawson T, Bernstein A. High incidence of lung, bone, and lymphoid tumors in transgenic mice overexpressingmutant alleles of the p53 oncogene. Mol Cell Biol 1989;9:3982-3991. 2. Malkin D, Li FP, Strong LC, Fraumeni JF, Nelson CE, Kim DH, Kassel J, et al. Germ line p53 mutations in a familial syndrome of breast cancer, sarcomas, and other neoplasms. Science 1990;250: 1233-1238. 3. Takahashi T, Carbone D, Takahashi T, Nau MM, Hida T, Linnoila I, Ueda R, et al. Wild-type but not mutant p53 suppresses the growth of human lung cancer cells bearing multiple genetic lesions. Cancer Res 1992;52:2340-2343. 4. Kastan MB, Plunkett BS, Kuerbitz SJ. p53 protein is a cell cycle checkpoint following DNA damage. AACR Proc 1992;33:169. 5. Caron de Fromental C, Soussi T. TP53 tumor suppressor gene: a model for investigating human mutagenesis. Genes, Chromosomes and Cancer 1992;41-15.
HEPATOLOGY ELSEWHERE: TWENTY YEARSAFTER!
Ramond MJ, Poynard T, Rueff B, Mathurin P, Theodore C, Chaput JC, Benhamou JP. A randomized trial of prednisolone in patients with severe alcoholic hepatitis. N Engl J Med 1992;326:507-512. ABSTRACT Buckground. Controlled trials have yielded inconsistent results with regard to the efficacy of corticosteroids in the treatment of alcoholic hepatitis. Three meta-analyses suggest that they may be effective in patients with encephalopathy who have severe liver disease. Methods. We conducted a randomized, double-blind trial comparing 28 days of prednisolone treatment (40 mg per day) with placebo in 61 patients with biopsyproved alcoholic hepatitis and either spontaneous hepatic encephalopathy (n = 19) or a discriminantfunction value higher than 32. The discriminant function used was as follows: 4.6 (prothrombintime control time [in seconds]) + serum bilirubin (inmicromoles per liter)/l7. Fifty-seven of the patients had evidence of cirrhosis on biopsy. The primary end point was death within two months.
HEPATOLOGY Elsewhere
Vol. 16, No. 4, 1992
ResuZfs. One patient was lost to follow-up after 56 days. Treatment was discontinued in two patients because of drugtoxicity. By the 66th day after randomization, 16 of 29 placebo recipients had died (mean [ f SEI survival, 46 f 8 percent), as compared with 4 of 32 preduieolone recipients (survival, 88 f 5 percent) (log-ranktest, 10.9; P = 0.001). The survival advantage for prednisolone persisted after stratification according to center and the presence of encephalopathy, and after adjustment for prognostic factors in a proportional-hazardsmodel. c o n c l u a ~Treatment ~. with prednisolone improves the short-termsurvival of patients with severe biopsyproved alcoholic hepatitis. COMMENTS
The presence of hepatitis on liver biopsy increases the mortality rate in patients with alcoholic liver disease with or without cirrhosis from 7% to 20% (1).If the hepatitis results in severe hepatic decompensation the mortality rate increases to 50% during that hospitalization (2). In the study by Ramond et al., this French group during a 3-yr period were able to study 124 patients with severe liver disease. This was defined as a decriminatory function (DF) score of greater than 32 or spontaneous PSE or both. Using these criteria the authors had previously found a mortality rate of about 50% in such patients. The DF had originally been described by Maddrey et al. (3) and consisted of a mathematical formula using prothrombin time and serum bilirubin. To this reviewer’s knowledge the DF has not been subjected to as careful an evaluation as, or been compared to, other prognostic indices such as the Pugh-Child, the Orrego CCLI or the Copenhagen Cox Regression Index. Nevertheless, it is simple to use and it does appear to predict a 50% mortality rate in this patient population. Of greater importance is the fact that the investigators insisted on a liver biopsy as an entry criteria. Presumably, the reason for this is that previous series have shown that with chronic alcoholic patients with liver disease the liver disease in approximately 10%to 20% is due to another cause (4). This was confirmed in this series. In addition, another 32 patients, or one third, were then excluded for other reasons. The remaining 61 patients (or less than 50%of the original patients) were then randomized with different codes for the sexes. Although unexplained this was presumably related to a worse prognosis in women than in men. Despite this randomization, disparities were seen in some of the important parameters: in the prednisolone group the serum bilirubin levels were lower (213vs. 284 Fmol);the albumin levels were higher (414 vs. 388 Fmol); creatinine levels were lower (83 vs. 103 mmol); and DF were lower (51and 60 at the time of randomization). Thus although none of the individual parameters were significantly different, all four important ones were tilted in favor of the steroid group, which if complexed into an index might have significantly favored the steroid group. Despite this, the steroid-treated group had a significantly reduced mortality. All four deaths occurred
1097
within the first 11days of therapy, but this was 25 days after admission. The authors point out that, as with variceal bleeding, the results might have been different if randomization with start of therapy had occurred closer to the time of admission. Fourteen of the 16 deaths in the placebo group occurred after 11days, or 28 days after admission. Thus to explore the possible mode of steroid action, one first has to look at the causes of death. These included infection, gastrointestinal bleeding and kidney failure (creatinine greater than 200 mmol/L). The former two would exacerbate hepatic hypoxia, thought to be a major factor in the pathogenesis of alcoholic hepatocyte necrosis (51, and renal hypoperfusion, a major factor in the pathogenesis of hepatorenal syndrome, would be exacerbated by the presence of severe hyperbilirubinemia. The frustrating deterioration of liver function during hospitalization in patients with alcoholic hepatitis, once again demonstrated, has constantly perplexed physicians. One explanation, as pointed out by the investigators, could be that hepatocyte damage in this condition is immune mediated because of antibody reaction to antigenic stimulation of acetaldehyde protein adducts (6). Steroid suppression of this reaction might result in decreased hepatocyte necrosis. Patients with alcohol liver disease were shown to have elevated circulating titers to these adducts, and the highest titers were in patients with alcoholic hepatitis (6). However, only 24 patients remained hospitalized, thus raising the important questions of compliance with medication and resumption of alcohol intake. Surprisingly, neither of these factors were referred to by the authors, and no information was provided as to whether either were monitored. A resumption of alcohol intake in the placebo group significantly greater than the steroid group could have affected the mortality. Another important therapeutic issue previously raised is the question of the effect of nutrition in the management of alcoholic hepatitis. Despite several controlled trials showing that nutritional parameters definitely improve the question of an improved survival remains controversial (7). Although the authors addressed this issue with high-calorie and protein intake during hospitalization despite the presence of portosystemic encephalopathy (PSE), this apparently was not controlled after the patients were discharged from the hospital. Numerous controlled trials have previously failed to show a beneficial effect of steroids in alcoholic hepatitis. However, many of these trials suffered from insufficient power. To reduce the mortality rate of 50% by 50%, the calculation of an alpha error of 5% and a beta error of 10% would mean 60 patients in each group. In fact, in the positive trials, steroids reduced the mortality rate by 75%, enabling the investigators to achieve significance with about 30 patients in each group. Nine of the 10 negative trials had randomized less than 60 patients. This recurring error took 20 yr to correct. A recent metaanalysis of the best 11studies showed a signiscant benefit of steroids for patients with severe alcoholic
1098
HEPATOLOGY Elsewhere
hepatitis as defined by spontaneous PSE (8). Thus this study agrees in large measure with this conclusion, extending the benefit to nonencephalopathic patients with severe hepatic decompensation. However, a number of questions remain. What is the mode of therapeutic action of steroid therapy? Is prednisolone, the initial hepatic metabolite of prednisone, more effective? Interestingly, this drug at this dose and for this period of time exactly reproduces the first positive trial 20 years ago! The North Carolina group produced two positive trials with the randomization of only 37 and 14 patients respectively. What irony! What is the optimal dose regime for this therapy? Finally, having salvaged the patients from early death, should one then introduce long-term therapy with agents such as propylthiouracil (9)or colchicine (10)to prevent a death later? That is, should the patient be enrolled in a therapeutic program rather than be allowed to die from the same cause sometime in the future? The answer to these and other questions will have to await further careful studies. In the meantime, we appear to have made significant progress in the acute treatment of severe alcoholic hepatitis. LAURENCE M. BLENDIS, M.D.
The Toronto Hospital University of Toronto Toronto, Ontario M5G 2C4, Canada REFERENCES 1. Orrego H, Blake JE, Blendis LM, Medline A. Prognosis of alcoholic cirrhosis in the presence and absence of alcoholic hepatitis. Gastroenterology 1987;92:208-214. 2. Hardison WG, Lee FI. Prognosis in the acute liver disease of the alcoholic patient. N Engl J Med 1966;275:61-66. 3. Maddrey WC, Boitnott JK, Bedine MS, Weber FL, Meley E, White RJ. Steroid therapy of alcoholic hepatitis. Gastroenterology 1978;73:193-199. 4. McAfee JH, Keefe EB, Lee RG, Rosch J. Transjugular liver biopsy. HEPATOLOGY 1992;15:726-732. 5. Orrego H, Israel Y, Blendis LM. Alcoholic liver disease: information in search of knowledge? HEPATOLOGY 1981;1:267-283. 6. Niemela 0, Klajner F, Orrego H, Vidins E, Blendis LM,Israel Y. Antibodies against acetaldehyde-modified protein epitopes in human alcoholics. HEPATOLOGY 1987;7:1210-1214. 7. Cabre E, Gonzalez-Huix F, Abad-Lacruz A, Esteve M, Acero D, Fernandez-BanaresF, Xi01 X, et al. Effect of total enteral nutrition on the short-term outcome of severely malnourished cirrhotics. Gastroenterology 1990;98:715-720. 8. Imperiale TF, McCulloughAJ. Do corticosteroidsreduce mortality from alcoholic hepatitis? Ann Intern Med 1990;113:299-307. 9. Orrego H, Blake JE,Blendis LM, Compton KV, Israel Y. Longterm treatment of alcoholic liver disease with propylthiouracil. N Engl J Med 1987;317:1421-1427. 10. Kershenobich D, Vargas F, Garcia Tsao G, Tamayo RP, Gent M, Rojkind M. Colchicine in the treatment of cirrhosis of the liver. N Engl J Med 1988;318:1709-1713.
STATISTICS AND CLINICAL TRIALS:THE CASE OF PREDNISOLONE IN ALCOHOLIC HEPATITIS
The paper by b o n d et al. is the latest in a series of studies that address the effect of prednisolone or similar
HEPATOLOGY
compounds on survival in patients with acute alcoholic hepatitis. All are well designed and carefully controlled. This paper is no exception. The recommendations made on the basis of all these publications, however, are inconsistent. Why, in such a deceptively simple disease? First, an element of unpredictability is seen in patients with alcoholic liver disease. It has been known for some time that treatment with steroids lowers life expectancy in patients with alcoholic cirrhosis (11, yet there is a rationale to evaluate their use in patients with alcoholic hepatitis even though many patients’ studies have both cirrhosis and alcoholic hepatitis. Which lesion is dominant? Secondly, the patient sample evaluated may be heterogeneous. Patients could gain entry to this study by exceeding a value of 32 of Maddrey’s discriminant function (D.F.), (2),having encephalopathy or both. Given that the D.F. is composed of only two elements (the serum bilirubin concentration and prothrombin time) and assuming that no patient with a prothrombin ratio greater than 1.5 will have a biopsy, it can be calculated that patients who meet the entry criteria will fall within a range of minimum serum bilirubin concentration of 4.9 mg/dl and a prothrombin time of about 18 sec, or a normal prothrombin time and a bilirubin concentration of at least 32 mg/dl. The fact that 15of the patients meeting these criteria also had encephalopathy does not necessarily modify the homogeneity of this group. However, patients with less abnormal prothrombin times and bilirubin concentrations could qualify for entry primarily on the basis of encephalopathy, suggesting the possibility that the study population was indeed heterogeneous. In this instance the impact of this reservation is small because it accounts for only four patients. A third and serious problem illustrated by this article is the requirement that placebo and treatment groups be truly similar. Ramond et al. have been singularly unfortunate because, despite their painstaking efforts to randomize at the time treatment started, the prednisolone group was much less ill than the control group, thus favoring a treatment effect. True, no “statistically significant” differences were found, but this may not be the issue. The objective is not to prove “non-difference” but reasonable similarity. Both groups of patients started off with similar D.F.s, and the D.F. increased during the waiting period. In the prednisolone group this increase was significant (p < 0.05); in the placebo group this was much more striking (p 0.025).Hence, at randomization the differences between groups could be calculated to have a p value of about 0.1, indicating a substantial probability that the two groups were indeed different. Doubts about the similarity of the two groups are increased when the creatinine values were examined. Expressed as multiples of upper limit of normal at randomization, the prednisolone group had a value of 0.79 f 0.1 mg/dl, and the placebo group had a value of 0.98 ? 0.11 mg/dl. This suggests that almost one half of the placebo-treated patients had an abnormal serum creatinine value at randomization, compared with substantially fewer in the controls.
-=
HEPATOLOGY Elsewhere
radiation to damage DNA stop in G1 before they replicate their DNA in S phase. However, when they express mutant p53 they proceed through G1 and replicate their DNA despite the damage (4). This arrest could allow the cell time to repair its DNA or, if irreversibly damaged, result in permanent clonal deletion. This hypothesis makes a great deal of intuitive sense when all of the data are examined. The lack of such a screening function should not interfere with normal cell growth or differentiation. When exposed to carcinogens, however, cells lacking p53 would attempt to replicate their damaged DNA, immortalizing the alteration, whereas cells with wild-type p53 would be locked in a growth-arrested state and fail to pass on the acquired defects. How is this consistent with the growth arrest in G1 found when tumor cells are transfected with wild-type p53? Tumor cells may have a high level of ongoing genetic abnormalities that normal p53 responds to by inducing growth arrest, or growth arrest may be an artifact of the artificial overexpression of p53. Inequalities in p63 Abnormalities. A p53 knockout may not be equivalent to p53 point mutation. Missense mutations constitute the largest group of naturally acquired p53 abnormalities in tumors and are found to occur throughout the open reading frame; however, they are mostly in the 200 codon region spanning exons 5 through 8. Nonsense (stop) and splicing mutations, frame shifts and deletions of part or all of the open reading frame are also found, but deletions of p53 are actually a very uncommon event. By far the most common alteration is the missense point mutation, implying that mutant p53 commonly plays an active role in human cancer. The high frequency of missense mutations is not just due to a general increased probability of such events, because the types of mutations found in the APC tumor suppressor gene in familial polyposis tend to be deletions or nonsense (stop) mutations. Functionally, mutant p53 can cooperate with ras to transform primary rat embryo fibroblasts and thus appears to have a dominant transforming capability in addition to the tumor-suppressing phenotype. All missense mutations are not equivalent either because both transforming activity and transcriptional activities vary among various different missense proteins. A different spectrum of affected codons in different tumors seems to exist, with a preference for codon 249 in HCC, codon 175 in colon carcinoma and codons 150 to 160 (hot spot IIa) (5) in lung cancer. The latter two are nearly mutually exclusive, with codon 175 mutations being rare in lung cancer and IIa mutations rare in colon cancers. It is possible that the various mutant p53 proteins associated with different tumors affect the ability of the resultant protein to detect genomic alterations specifically associated with the different tumor types, for example those caused by particular carcinogens. The p53 “knockout” may thus be an imperfect model for early events leading to cancer. Interesting questions to ask with these p53 knockout mice include whether these mice have an altered sensitivity to carcinogens, whether the spectrum of
HEPATOLOGY
tumor types resulting from carcinogen treatment is different from normal mice and whether their cells exhibit increased sensitivity to oncogenic viruses that produce proteins that normally inactivate p53. Are there any detectable abnormalities of cell-cycle control, signal transduction, response to stress, cell division, DNA repair or cytogenetic abnormalities? Will tumors from these mice respond to the introduction of wild-type p53? Are they defective in antitumor immunity? The answers to these questions are forthcoming and will provide valuable insight into the role of p53 in normal and tumor cells.
DAVID CARBONE, M.D., PH.D. National Cancer Institute NCI-Navy Medical Oncology Branch Bethesda, Maryland 20889-5105 REFERENCES 1. Lavigueur A, Maltby V, Mock D, Rossant J , Pawson T, Bernstein A. High incidence of lung, bone, and lymphoid tumors in transgenic mice overexpressingmutant alleles of the p53 oncogene. Mol Cell Biol 1989;9:3982-3991. 2. Malkin D, Li FP, Strong LC, Fraumeni JF, Nelson CE, Kim DH, Kassel J, et al. Germ line p53 mutations in a familial syndrome of breast cancer, sarcomas, and other neoplasms. Science 1990;250: 1233-1238. 3. Takahashi T, Carbone D, Takahashi T, Nau MM, Hida T, Linnoila I, Ueda R, et al. Wild-type but not mutant p53 suppresses the growth of human lung cancer cells bearing multiple genetic lesions. Cancer Res 1992;52:2340-2343. 4. Kastan MB, Plunkett BS, Kuerbitz SJ. p53 protein is a cell cycle checkpoint following DNA damage. AACR Proc 1992;33:169. 5. Caron de Fromental C, Soussi T. TP53 tumor suppressor gene: a model for investigating human mutagenesis. Genes, Chromosomes and Cancer 1992;41-15.
HEPATOLOGY ELSEWHERE: TWENTY YEARSAFTER!
Ramond MJ, Poynard T, Rueff B, Mathurin P, Theodore C, Chaput JC, Benhamou JP. A randomized trial of prednisolone in patients with severe alcoholic hepatitis. N Engl J Med 1992;326:507-512. ABSTRACT Buckground. Controlled trials have yielded inconsistent results with regard to the efficacy of corticosteroids in the treatment of alcoholic hepatitis. Three meta-analyses suggest that they may be effective in patients with encephalopathy who have severe liver disease. Methods. We conducted a randomized, double-blind trial comparing 28 days of prednisolone treatment (40 mg per day) with placebo in 61 patients with biopsyproved alcoholic hepatitis and either spontaneous hepatic encephalopathy (n = 19) or a discriminantfunction value higher than 32. The discriminant function used was as follows: 4.6 (prothrombintime control time [in seconds]) + serum bilirubin (inmicromoles per liter)/l7. Fifty-seven of the patients had evidence of cirrhosis on biopsy. The primary end point was death within two months.
HEPATOLOGY Elsewhere
Vol. 16, No. 4, 1992
ResuZfs. One patient was lost to follow-up after 56 days. Treatment was discontinued in two patients because of drugtoxicity. By the 66th day after randomization, 16 of 29 placebo recipients had died (mean [ f SEI survival, 46 f 8 percent), as compared with 4 of 32 preduieolone recipients (survival, 88 f 5 percent) (log-ranktest, 10.9; P = 0.001). The survival advantage for prednisolone persisted after stratification according to center and the presence of encephalopathy, and after adjustment for prognostic factors in a proportional-hazardsmodel. c o n c l u a ~Treatment ~. with prednisolone improves the short-termsurvival of patients with severe biopsyproved alcoholic hepatitis. COMMENTS
The presence of hepatitis on liver biopsy increases the mortality rate in patients with alcoholic liver disease with or without cirrhosis from 7% to 20% (1).If the hepatitis results in severe hepatic decompensation the mortality rate increases to 50% during that hospitalization (2). In the study by Ramond et al., this French group during a 3-yr period were able to study 124 patients with severe liver disease. This was defined as a decriminatory function (DF) score of greater than 32 or spontaneous PSE or both. Using these criteria the authors had previously found a mortality rate of about 50% in such patients. The DF had originally been described by Maddrey et al. (3) and consisted of a mathematical formula using prothrombin time and serum bilirubin. To this reviewer’s knowledge the DF has not been subjected to as careful an evaluation as, or been compared to, other prognostic indices such as the Pugh-Child, the Orrego CCLI or the Copenhagen Cox Regression Index. Nevertheless, it is simple to use and it does appear to predict a 50% mortality rate in this patient population. Of greater importance is the fact that the investigators insisted on a liver biopsy as an entry criteria. Presumably, the reason for this is that previous series have shown that with chronic alcoholic patients with liver disease the liver disease in approximately 10%to 20% is due to another cause (4). This was confirmed in this series. In addition, another 32 patients, or one third, were then excluded for other reasons. The remaining 61 patients (or less than 50%of the original patients) were then randomized with different codes for the sexes. Although unexplained this was presumably related to a worse prognosis in women than in men. Despite this randomization, disparities were seen in some of the important parameters: in the prednisolone group the serum bilirubin levels were lower (213vs. 284 Fmol);the albumin levels were higher (414 vs. 388 Fmol); creatinine levels were lower (83 vs. 103 mmol); and DF were lower (51and 60 at the time of randomization). Thus although none of the individual parameters were significantly different, all four important ones were tilted in favor of the steroid group, which if complexed into an index might have significantly favored the steroid group. Despite this, the steroid-treated group had a significantly reduced mortality. All four deaths occurred
1097
within the first 11days of therapy, but this was 25 days after admission. The authors point out that, as with variceal bleeding, the results might have been different if randomization with start of therapy had occurred closer to the time of admission. Fourteen of the 16 deaths in the placebo group occurred after 11days, or 28 days after admission. Thus to explore the possible mode of steroid action, one first has to look at the causes of death. These included infection, gastrointestinal bleeding and kidney failure (creatinine greater than 200 mmol/L). The former two would exacerbate hepatic hypoxia, thought to be a major factor in the pathogenesis of alcoholic hepatocyte necrosis (51, and renal hypoperfusion, a major factor in the pathogenesis of hepatorenal syndrome, would be exacerbated by the presence of severe hyperbilirubinemia. The frustrating deterioration of liver function during hospitalization in patients with alcoholic hepatitis, once again demonstrated, has constantly perplexed physicians. One explanation, as pointed out by the investigators, could be that hepatocyte damage in this condition is immune mediated because of antibody reaction to antigenic stimulation of acetaldehyde protein adducts (6). Steroid suppression of this reaction might result in decreased hepatocyte necrosis. Patients with alcohol liver disease were shown to have elevated circulating titers to these adducts, and the highest titers were in patients with alcoholic hepatitis (6). However, only 24 patients remained hospitalized, thus raising the important questions of compliance with medication and resumption of alcohol intake. Surprisingly, neither of these factors were referred to by the authors, and no information was provided as to whether either were monitored. A resumption of alcohol intake in the placebo group significantly greater than the steroid group could have affected the mortality. Another important therapeutic issue previously raised is the question of the effect of nutrition in the management of alcoholic hepatitis. Despite several controlled trials showing that nutritional parameters definitely improve the question of an improved survival remains controversial (7). Although the authors addressed this issue with high-calorie and protein intake during hospitalization despite the presence of portosystemic encephalopathy (PSE), this apparently was not controlled after the patients were discharged from the hospital. Numerous controlled trials have previously failed to show a beneficial effect of steroids in alcoholic hepatitis. However, many of these trials suffered from insufficient power. To reduce the mortality rate of 50% by 50%, the calculation of an alpha error of 5% and a beta error of 10% would mean 60 patients in each group. In fact, in the positive trials, steroids reduced the mortality rate by 75%, enabling the investigators to achieve significance with about 30 patients in each group. Nine of the 10 negative trials had randomized less than 60 patients. This recurring error took 20 yr to correct. A recent metaanalysis of the best 11studies showed a signiscant benefit of steroids for patients with severe alcoholic
1098
HEPATOLOGY Elsewhere
hepatitis as defined by spontaneous PSE (8). Thus this study agrees in large measure with this conclusion, extending the benefit to nonencephalopathic patients with severe hepatic decompensation. However, a number of questions remain. What is the mode of therapeutic action of steroid therapy? Is prednisolone, the initial hepatic metabolite of prednisone, more effective? Interestingly, this drug at this dose and for this period of time exactly reproduces the first positive trial 20 years ago! The North Carolina group produced two positive trials with the randomization of only 37 and 14 patients respectively. What irony! What is the optimal dose regime for this therapy? Finally, having salvaged the patients from early death, should one then introduce long-term therapy with agents such as propylthiouracil (9)or colchicine (10)to prevent a death later? That is, should the patient be enrolled in a therapeutic program rather than be allowed to die from the same cause sometime in the future? The answer to these and other questions will have to await further careful studies. In the meantime, we appear to have made significant progress in the acute treatment of severe alcoholic hepatitis. LAURENCE M. BLENDIS, M.D.
The Toronto Hospital University of Toronto Toronto, Ontario M5G 2C4, Canada REFERENCES 1. Orrego H, Blake JE, Blendis LM, Medline A. Prognosis of alcoholic cirrhosis in the presence and absence of alcoholic hepatitis. Gastroenterology 1987;92:208-214. 2. Hardison WG, Lee FI. Prognosis in the acute liver disease of the alcoholic patient. N Engl J Med 1966;275:61-66. 3. Maddrey WC, Boitnott JK, Bedine MS, Weber FL, Meley E, White RJ. Steroid therapy of alcoholic hepatitis. Gastroenterology 1978;73:193-199. 4. McAfee JH, Keefe EB, Lee RG, Rosch J. Transjugular liver biopsy. HEPATOLOGY 1992;15:726-732. 5. Orrego H, Israel Y, Blendis LM. Alcoholic liver disease: information in search of knowledge? HEPATOLOGY 1981;1:267-283. 6. Niemela 0, Klajner F, Orrego H, Vidins E, Blendis LM,Israel Y. Antibodies against acetaldehyde-modified protein epitopes in human alcoholics. HEPATOLOGY 1987;7:1210-1214. 7. Cabre E, Gonzalez-Huix F, Abad-Lacruz A, Esteve M, Acero D, Fernandez-BanaresF, Xi01 X, et al. Effect of total enteral nutrition on the short-term outcome of severely malnourished cirrhotics. Gastroenterology 1990;98:715-720. 8. Imperiale TF, McCulloughAJ. Do corticosteroidsreduce mortality from alcoholic hepatitis? Ann Intern Med 1990;113:299-307. 9. Orrego H, Blake JE,Blendis LM, Compton KV, Israel Y. Longterm treatment of alcoholic liver disease with propylthiouracil. N Engl J Med 1987;317:1421-1427. 10. Kershenobich D, Vargas F, Garcia Tsao G, Tamayo RP, Gent M, Rojkind M. Colchicine in the treatment of cirrhosis of the liver. N Engl J Med 1988;318:1709-1713.
STATISTICS AND CLINICAL TRIALS:THE CASE OF PREDNISOLONE IN ALCOHOLIC HEPATITIS
The paper by b o n d et al. is the latest in a series of studies that address the effect of prednisolone or similar
HEPATOLOGY
compounds on survival in patients with acute alcoholic hepatitis. All are well designed and carefully controlled. This paper is no exception. The recommendations made on the basis of all these publications, however, are inconsistent. Why, in such a deceptively simple disease? First, an element of unpredictability is seen in patients with alcoholic liver disease. It has been known for some time that treatment with steroids lowers life expectancy in patients with alcoholic cirrhosis (11, yet there is a rationale to evaluate their use in patients with alcoholic hepatitis even though many patients’ studies have both cirrhosis and alcoholic hepatitis. Which lesion is dominant? Secondly, the patient sample evaluated may be heterogeneous. Patients could gain entry to this study by exceeding a value of 32 of Maddrey’s discriminant function (D.F.), (2),having encephalopathy or both. Given that the D.F. is composed of only two elements (the serum bilirubin concentration and prothrombin time) and assuming that no patient with a prothrombin ratio greater than 1.5 will have a biopsy, it can be calculated that patients who meet the entry criteria will fall within a range of minimum serum bilirubin concentration of 4.9 mg/dl and a prothrombin time of about 18 sec, or a normal prothrombin time and a bilirubin concentration of at least 32 mg/dl. The fact that 15of the patients meeting these criteria also had encephalopathy does not necessarily modify the homogeneity of this group. However, patients with less abnormal prothrombin times and bilirubin concentrations could qualify for entry primarily on the basis of encephalopathy, suggesting the possibility that the study population was indeed heterogeneous. In this instance the impact of this reservation is small because it accounts for only four patients. A third and serious problem illustrated by this article is the requirement that placebo and treatment groups be truly similar. Ramond et al. have been singularly unfortunate because, despite their painstaking efforts to randomize at the time treatment started, the prednisolone group was much less ill than the control group, thus favoring a treatment effect. True, no “statistically significant” differences were found, but this may not be the issue. The objective is not to prove “non-difference” but reasonable similarity. Both groups of patients started off with similar D.F.s, and the D.F. increased during the waiting period. In the prednisolone group this increase was significant (p < 0.05); in the placebo group this was much more striking (p 0.025).Hence, at randomization the differences between groups could be calculated to have a p value of about 0.1, indicating a substantial probability that the two groups were indeed different. Doubts about the similarity of the two groups are increased when the creatinine values were examined. Expressed as multiples of upper limit of normal at randomization, the prednisolone group had a value of 0.79 f 0.1 mg/dl, and the placebo group had a value of 0.98 ? 0.11 mg/dl. This suggests that almost one half of the placebo-treated patients had an abnormal serum creatinine value at randomization, compared with substantially fewer in the controls.
-=
