Long-term Outcomes of Patients With Autoimmune Hepatitis Managed at a Nontransplant Center

Transcription

1 GASTROENTEROLOGY 2011;140: Long-term Outcomes of Patients With Autoimmune Hepatitis Managed at a Nontransplant Center BARBARA HOEROLDT, ELAINE MCFARLANE,* ASHA DUBE, PANDURANGAN BASUMANI, MOHAMMED KARAJEH,* MICHAEL J. CAMPBELL, and DERMOT GLEESON* *Liver Unit, Sheffield Teaching Hospitals, Sheffield, England; Department of Gastroenterology, Rotherham General Hospital, Rotherham, England; Department of Pathology, Sheffield Teaching Hospitals, Sheffield, England; and Medical Statistics Group, School for Health and Related Research, University of Sheffield, Sheffield, England See editorial on page BACKGROUND & AIMS: The long-term outcomes of patients treated for autoimmune hepatitis (AIH) are considered to be good. However, follow-up data beyond 10 years are limited and confined to tertiary referral centers. We assessed long-term outcomes and determinants of outcome in patients with AIH from a nontransplant center. METHODS: We studied 245 patients (204 women; median age, 56 years; range, years) with AIH (167 definite by International AIH Group criteria) managed at a single nontransplant center from 1971 to RESULTS: 229 patients (93%) achieved normal serum levels of alanine aminotransferase within 12 months after treatment. After a median follow-up period of 9.4 years (range, years), 11 patients received liver transplants (2 subsequently died). Seventy other patients died (30 from liver disease), 15 were censored (moved away, defaulted, or developed primary biliary cirrhosis), and 149 were still being followed up on December 31, Survival rates from all-cause death or transplantation were 82% 3% and 48% 5% after 10 and 20 years, respectively, and from liver-related death or transplantation were 91% 2% and 70% 5%, respectively. The standardized mortality ratio was 1.63 for all-cause death (95% confidence interval [CI], ), 1.86 also considering liver transplant as death (95% CI, ), and 0.91 for non liver-related death (95% CI, ). By Cox regression analysis, liver decompensation, cirrhosis at any time, failure to normalize levels of alanine aminotransferase within 12 months, and 4 relapses per decade were significantly associated with liver-related death or transplant. CONCLUSIONS: Despite a good initial response to immunosuppression, long-term mortality of patients with AIH is greater than that of the general population. Keywords: PBC; Autoimmune Disease; Immunosuppressive Therapy; Prednisolone. Autoimmune hepatitis (AIH) is a chronic immune-mediated liver disease, diagnosis of which is based on a widely accepted scoring system. 1 In untreated patients, the outcome is poor. However, in trials published in the early 1970s, treatment with corticosteroids, with or without azathioprine, dramatically improved survival. 2 Over the past 15 years, several centers have reported experience with AIH Ten-year survival has ranged from 85% to 95%. 3 6 In 2 reports, 3,4 survival of patients with AIH was similar to that of the general population. These data support a widespread view that the outcome of treated AIH is good. However, there are few data on long-term outcome in AIH. Only 3 centers 3,6,7 10 have reported on more than 140 patients. In only 4 reports, 6,9,12,14 one involving only 42 patients, 12 did the mean follow-up period exceed 8 years. The most extensive and long-term studies have come from 2 tertiary referral centers, 6,8 raising the possibility of referral bias. Reported numbers and inclusion criteria have sometimes varied in successive publications from the same center. 3,8,10 Therefore, the overall outcome for AIH beyond the first decade remains ill defined. Suggested adverse prognostic factors for AIH include cirrhosis at diagnosis, 5,15 development of cirrhosis during treatment in some 11 although not other 3 studies, nonwhite ethnicity, 16 female sex, 6 presence of symptoms, 5,17 severe liver dysfunction, 18 less than 10-fold elevation of transaminase levels at presentation, 14 failure to normalize transaminase levels with treatment, 11,13,19 and recurrent relapses. 10,20 Few of these have been independently confirmed. We report here the outcome of a large unselected cohort of patients with AIH, most of them referred from primary care and followed up over several decades in a non liver transplant center. The results suggest that the long-term outcome of AIH is less benign than commonly supposed. We have also analyzed the potential determinants of long-term outcome. Patients and Methods Patients Ours is a nontransplant center. We searched clinical and liver biopsy databases as well as a database Abbreviations used in this paper: AIH, autoimmune hepatitis; IAIHG, International Autoimmune Hepatitis Group; PBC, primary biliary cirrhosis; SMR, standardized mortality ratio by the AGA Institute /$36.00 doi: /j.gastro

2 June 2011 OUTCOME OF AUTOIMMUNE HEPATITIS 1981 relating to a study of AIH epidemiology in Sheffield ( ). We found 259 patients with a clinical diagnosis of AIH presenting since Of these, 245 patients were included in the present study. All had probable (n 78) or definite (n 167) AIH on the pretreatment score, as defined by International Autoimmune Hepatitis Group (IAIHG) revised criteria. 1 Of the 14 patients excluded, 5 had insufficient points for a diagnosis of AIH by IAIHG criteria. In 2 patients just meeting the score for probable AIH, liver biopsy showed primary biliary cirrhosis (PBC). The remaining 7 patients met the IAIHG criteria but had other coexisting liver diseases (n 6; alcoholic liver disease, 1; nonalcoholic fatty liver disease, 2; hepatitis C virus [HCV], 3) or inadequate follow-up data (n 1). The 245 patients include all those (n 192) with AIH attending the department since January 1, 1987, but we may have missed some patients who died before All 245 patients were negative for hepatitis B surface antigen. All 220 patients who were alive and in follow-up since 1995 were HCV antibody negative. None of the remaining 25 patients had any risk factors for HCV. All but 16 patients underwent liver biopsy at diagnosis. Of these 16 patients, 3 patients underwent a liver biopsy 2 to 15 (median, 6) years after diagnosis and 4 others had histology obtained postmortem 2 to 9 years after diagnosis. Histology was consistent with AIH in all cases and did not suggest another diagnosis. Of the 9 patients with no histology at any stage, 5 had baseline AIH scores of 17 to 20, thus meeting the criteria for definite AIH. The remaining 4 patients had probable AIH, with scores of 10 to 14. All 4 were hepatitis B surface antigen and HCV antibody negative, and in all liver test results normalized rapidly on corticosteroid treatment. Two experienced a relapse. All 4 patients had either raised serum immunoglobulin G/globulin and/or positive autoantibodies. Overlap syndromes were present in 15 patients (14 AIH with PBC and 1 AIH with primary sclerosing cholangitis). In all 15 patients, the AIH component of the overlap syndrome necessitated immunosuppressive treatment. A total of 89 patients (36%) had cirrhosis at diagnosis and 47 patients (19%) had clinical decompensation at presentation (both defined in Table 1). Treatment and Definition of Response After diagnosis, 212 of the 245 patients received corticosteroids (210 prednisolone, 1 hydrocortisone, 1 methylprednisolone) specifically for AIH. In 3 of the 212 patients, low-dose corticosteroid treatment for preexisting conditions (systemic lupus erythematosus, Addison disease) was increased on diagnosis of AIH. Continuation of lowdose prednisolone (5 mg/day taken for systemic lupus erythematosus) was considered adequate immunosuppression in a further 2 patients. In 7 patients, initial treatment was elsewhere and details were unobtainable. The remaining 24 patients were not treated at diagnosis; 6 had minimal inflammation on liver biopsy (Ishak necroinflammatory score of 3). 21 Spontaneous normalization of liver test results occurred in 17 patients, and one refused corticosteroid treatment but had subsequent azathioprine treatment after normalization of liver function test results. Of these 24 patients, 12 required corticosteroid treatment at a later stage. Thus, 224 patients overall received corticosteroid treatment specifically for AIH, at a median (range) initial dosage of 30 (10 60) mg/day. As serum transaminase levels decreased, the dosage of prednisolone was tapered to 10 mg/day; median time receiving more than 10 mg/day was 3 (range, 0 30) months. A second immunosuppressive agent was introduced between 4 and 24 weeks later in most patients. Azathioprine was started in 208 patients at an initial dose at 1 mg/kg, which was continued for more than 4 weeks in 188 patients. Biochemical remission was defined as a decrease in serum alanine aminotransferase (ALT) levels to within the normal range ( 55 U/L) within 1 year from diagnosis. Serial ALT values over the first year of treatment were available in 201 patients. Actual values were unavailable in 44 patients, but correspondence in 40 of these indicated that values had normalized (ALT level 55 U/L) within 12 months of starting treatment. Patients in biochemical remission were encouraged to undergo repeat liver biopsy after 2 to 3 years of treatment. For patients in histologic remission (Ishak necroinflammatory score 21 of 3) and those patients who did not undergo follow-up biopsy but whose serum transaminase levels were persistently normal, prednisolone was withdrawn gradually and the dosage of azathioprine was increased to 2 mg kg 1 day 1 or the maximum tolerated dosage below this. In most patients, maintenance azathioprine was continued indefinitely. In 42 patients, azathioprine was changed to another immunosuppressive drug because of side effects and lack of efficacy (Table 2). To assess the association between outcome and azathioprine treatment, the dosage of azathioprine was recorded at each clinic visit. Percentage of follow-up time on azathioprine, mean dosage of azathioprine while on the drug and, where body weights were available (n 220), dosage per kilogram body weight was calculated. Relapse was defined, using IAIHG criteria, 1 as either (1) doubling of the transaminase levels from the nadir (and above normal) together with symptoms or (2) an increase in the transaminase levels to 3 times the upper limit of normal with or without symptoms. Relapse rate was expressed as number of relapses per decade of follow-up. The end of data collection was December 31, 2007, for patients alive without a liver transplant and under active follow-up (n 149). A total of 141 of these patients still attended the Liver Unit, and 8 patients (all older than 80 years) were managed by their family practitioner with liaison as required. Patients moving out of area or defaulting from follow-up (n 12) were censored at their last clinic visit. Patients were censored if follow-up biopsy

3 1982 HOEROLDT ET AL GASTROENTEROLOGY Vol. 140, No. 7 Table 1. Baseline Characteristics of Patients All patients Definite AIH Probable AIH P value a No Age at diagnosis (y) (2.5 87) Age groups (y),n(%) Younger than (4) (19) (35) Older than (41) Female/male 204/41 149/18 55/ Ethnicity (white/other) 229/16 160/7 69/9 Median AIH score 16 (10 23) 19 (16 23) 14 (10 15) Presentation Acute Chronic Incidental Autoantibody profile ANA ( 1:40) SMA ( 1:40) LKM ANA/SMA ( 1:40) None None or titer 1:40 (%) 57 (24) 28 (16.8) 29 (44.2).0006 ALT (14 55 IU/L) b 413 ( ) 498 ( ) 214 ( ) AST (15 41 IU/L) b 385 ( ) 489 (8 2555) 235 ( ) Bilirubin (4 22 mol/l) b 38 (5 560) 56 (5 620) 29 (5 598) Albumin (35 48 g/l) b 33 (17 56) 32 (19 48) 35 (17 56) Globulins (18 36 g/l) b 43 (20 126) 46 (20 126) 39 (23 88) Prothrombin time (9 13 s) b 13 (9 34) 13.1 (99 34) 12.7 ( ) Associated autoimmune disease (excluding 89 (36), 125 conditions 72, 103 conditions 17, 22 conditions.0016 NIDDM) (%) Cirrhosis at presentation c (%) 89 (36) Decompensation at presentation d (%) 47 (19) Overlap syndrome (%) 14 (6) NOTE. Values are expressed as median (range) unless otherwise noted. ANA, antinuclear antibody; SMA, smooth muscle antibody; LKM, liver-kidney microsomal antibody; NIDDM, non insulin-dependent diabetes mellitus. a Definite vs probable AIH. b The values in parentheses indicate the normal range; where upper limits of normal varied during the follow-up period, the highest level is used as the reference value throughout follow-up. c Cirrhosis was confirmed on histology, explant, or postmortem examination in 79 patients. Two patients had a fibrosis score 5 but had bleeding varices and were classed as cirrhotic; the other 8 patients did not undergo biopsy at diagnosis but had varices without evidence of portal vein thrombosis. d Clinical decompensation was defined as variceal bleeding or 2 of the following: varices at endoscopy, visible ascites, marked coagulopathy ( 4 seconds prolonged), and hepatic encephalopathy. showed PBC and no flare of transaminase levels occurred after withdrawal of all immunosuppression (n 3). Patients were also censored at death (n 70) or liver transplantation (n 11). Statistical Analyses Statistical analysis was performed using SPSS for Windows (SPSS Inc, Chicago, IL). Categorical data were summarized as frequencies and percentages and continuous data as medians and ranges. Survival was calculated by life table analysis (Kaplan Meier) and compared between subgroups using the log-rank test. The association of several variables with liver-related death or transplantation was assessed by Cox regression analysis, looking at each variable separately. Variables significantly (P.10) associated were assessed further by backward stepwise Cox multiple regression analysis to establish which were independently associated with this outcome. Similar analyses were performed for the outcome of allcause death or liver transplantation. Standardized mortality ratio (SMR) was calculated using general population age- and sex-specific overall mortality data for South Yorkshire and Humberside from the 1991 National Census. 22 We could not obtain data on population mortality due to liver disease. However, we calculated that only 4% of deaths in the United Kingdom were due to liver disease. Therefore, we calculated SMR for non liver-related deaths by dividing death rates in the AIH cohort by expected total death rates for the general population, accepting that this estimate of SMR will be an underestimate by approximately 4%.

4 June 2011 OUTCOME OF AUTOIMMUNE HEPATITIS 1983 Table 2. Drug Treatment Results Baseline Characteristics Table 1 summarizes baseline characteristics of the 245 patients. As expected, compared with patients with probable AIH, those with definite AIH were more likely to No. No. of patients 245 No details on early treatment available 7 Details available 238 Initial treatment details (n 238) Started corticosteroid treatment 209 Increased preexisting corticosteroid 3 treatment Continued corticosteroid treatment at 2 the same dose No treatment ab 24 No. of patients treated with 212 corticosteroids at diagnosis No. of patients treated with 224 corticosteroids at any stage Median time dose of prednisolone 3 (0 30) 10 mg (mo) Median time to normalization of ALT 3 (0 24) levels (mo) Normalization of ALT levels with first corticosteroid course regardless when started (n 224) Normal 5 3 mo mo mo mo/not normalized 12 Median dosage of corticosteroid 30 (10 60) (mg/day) No. of patients receiving azathioprine 188 (of 208 started) 4 wk Median dosage (mg) 89.2 ( ) Median dosage (mg/kg body wt) 1.2 ( ) Median length (y) 6.7 ( ) Azathioprine stopped 74 Reason stopped Side effects 29 Ineffective 19 Remission 11 Medical 8 Liver transplant 7 No. of patients receiving alternative 42 immunosuppression Alternative immunosuppression c Mycophenolate mofetil 36 Cyclosporine 11 Tacrolimus 1 Cyclophosphamide 1 a Causes of nontreatment: spontaneous normalization of liver function test results, 17; refusal, 1; Ishak necroinflammatory score of 3, 6. b Of these, 12 patients were later treated with corticosteroids when they experienced a flare. c Thirty-six patients received one alternative immunosuppressive agent, 5 patients received 2 alternative immunosuppressive agents, and one patient received 3 alternative immunosuppressive agents. Five patients received further treatment with azathioprine later. be female and to have other autoimmune diseases but were less likely to have absent or low-titer autoantibodies. At diagnosis, 47 patients (19%) had clinical decompensation and 89 (36%) had cirrhosis; these are defined in Table 1. Cirrhosis at diagnosis was absent in 146 patients, and its presence or absence could not be established in 10 patients (9 who had not undergone biopsy at presentation, 1 report unavailable). In 5 of these 10 patients, cirrhosis was confirmed 2 to 22 years later but its time of onset remained unknown. In the remaining 5 patients, presence or absence of cirrhosis was never established. Initial Response to Treatment and Relapse Rate Within 1 year, 229 of the 245 patients achieved clinical and biochemical remission. Median time to normalization of transaminase levels was 3 months. A total of 96 patients subsequently had at least one relapse. In these, median (range) number of relapses per decade of follow-up was 2 (1 6). Relapse rate was 31%, 41%, and 51% after 5, 10, and 20 years, respectively, and did not differ significantly between patients with probable and those with definite AIH. Most patients reattained clinical and biochemical remission following reintroduction of prednisolone with or without increase in the dosage of azathioprine or change to alternative immunosuppression. Development of Cirrhosis Of the 146 patients without cirrhosis at presentation, 24 (16%) developed cirrhosis after a median of 5.7 (range, 1 27) years of follow-up. In 15 patients, development of cirrhosis was diagnosed histologically (12 biopsies, 2 explant liver, 1 postmortem); in 9 patients, cirrhosis was diagnosed based on development of varices (n 6) or hepatocellular carcinoma (n 3). In only 2 of the 24 was there evidence of development of liver disease other than AIH (one had developed PBC and one had developed features of nonalcoholic fatty liver disease on follow-up biopsy). The rate of de novo cirrhosis development was 12% (SE 3%) after 10 years and 34% (SE 8%) after 20 years. The rate was similar in patients with probable and definite AIH (data not shown) and was positively associated on Cox regression analysis with increasing time to initial normalization of serum ALT level (P.02) and with number of relapses per decade (P.005) but was unrelated to body weight or presence of diabetes (not shown). Overall Outcome Overall outcome is summarized in Figure 1. Of the 30 patients who died of liver disease without receiving a transplant, 8 had hepatocellular carcinoma, 21 had liver decompensation, and one had azathio-

5 1984 HOEROLDT ET AL GASTROENTEROLOGY Vol. 140, No. 7 Figure 1. Individual outcomes in autoimmune hepatitis. Of the 40 patients dying of extrahepatic disease, 14 had malignancy, 11 had cardiorespiratory diseases, 2 had sepsis, one had amyloidosis, and one had renal failure and intestinal obstruction. Ten patients died of either old age (n 6 aged years) or cause unknown (n 4). In all 10 of these patients, liver test results were normal before death. Of the 30 patients dying of liver disease, 2 were referred for transplantation but turned down and 28 were not referred for the following reasons: major comorbidity, 10; comorbidity and age older than 70 years, 7; advanced hepatocellular carcinoma, 8; acute decompensation and death, 1; drug toxicity, 1; and reason for nonreferral not clear, 1. The censored patients included 3 patients who developed pure PBC and 12 who were lost to follow-up (10 moved, 2 defaulted). prine-related bone marrow failure and sepsis. In only one patient who died of liver disease (in 1989) was the reason for nonreferral for transplantation unclear from the records. Forty patients died for reasons unrelated to liver disease (Figure 1). As shown in Figure 2A, rates of survival (from all-cause death or liver transplantation) were 82% (SE, 3%) and 48% (SE, 5%) after 10 and 20 years, respectively. Corresponding survival rates from liver-related death or transplantation were 91% (SE, 2%) and 70% (SE 5%). As suggested by Figure 2A, both all-cause (P.001) and liver-related mortality (P.013) were lower during the first decade (n 245) than during the second decade (n 115), and this difference persisted after adjusting for age (patients in the second decade were 10 years older). For all-cause death or transplantation, the hazard rate increased after 7 years of follow-up and was significantly lower during the first decade of follow-up compared with the second decade (Figure 2B). Mortality in the cohort was significantly in excess of that of the general population in Yorkshire and Humberside. SMR data are shown in Table 3. SMR in the cohort as a whole was 1.63 considering actual deaths only and 1.86 considering liver transplantation as death. When only deaths unrelated to liver disease were considered (patients censored in the event of liver-related death or transplantation), the SMR was 0.91 ( ), a value close to unity. This suggests that the excess of deaths in AIH is a result of liver disease. Similar SMR values were obtained for the 192 patients seen since 1987 (consecutive series, Table 3). Also, SMR was higher in patients presenting at a younger age. This was not due to an excess of deaths, but to a lower expected death rate in younger patients. Finally, consistent with Figure 2B, SMR during the second decade of follow-up was significantly higher than during the first decade of follow-up. Factors Associated With Death or Transplantation Factors strongly associated with liver-related death or transplantation included presence of liver decompensation at diagnosis (Figure 3A) and presence of cirrhosis at any stage (either at diagnosis, subsequent development, or time of onset unknown) (Figure 3B). Another adverse prognostic factor was failure to achieve normal serum ALT levels within 12 months of commencing treatment (Figure 3C). Indeed, even considering only patients with available ALT values, survival at 10 years (although not subsequently) was higher in those with a minimum serum ALT level of 0 to 27 U/L within 12 months (n 61) compared with those with values between 28 and 55 U/L (n 128) (100% vs 90%; P.016). Fourthly, a relapse rate exceeding 4 episodes per decade (although not relapse rates lower than this) was associated with a poor outcome (Figure 3D). All of these were confirmed to be independently associated factors on Cox multiple regression analysis (Table 4). Additionally, liver-related death or transplantation was associated with (1) nontreatment with azathioprine (Figure 3E and F), confirmed on Cox regression analysis (Table 4), and (2) presence of cirrhosis at diagnosis, confirmed on univariate analysis but not on Cox regression analysis. Factors not significantly associated with liver-related death or transplantation were age (hazard ratio [95% confidence interval], 1.02 [ ]), time to achieve normal serum ALT level (0.79 [ ]), year of diagnosis (1.0 [ ]), AIH score (1.00 [ ]), probable or definite AIH (0.87 [ ]), presence of overlap syndrome (1.59 [ ]), male sex (0.80 [ ]), nonwhite ethnicity (0.05 [ ]), or treatment with immunosuppressive drugs other than prednisolone and azathioprine (1.32 [ ]).

6 June 2011 OUTCOME OF AUTOIMMUNE HEPATITIS 1985 Figure 2. (A) Overall survival from first presentation. The lower plot shows all-cause death or liver transplant, and the upper plot shows liver-related deaths or transplants, with death due to non liver-related causes censored. Numbers at risk, given below the figure, are identical for both end points, because in analysis of liver-related deaths, patients were censored in the event of a non liver-related death. (B) Smoothed hazard estimate for all-cause death or liver transplantation. Hazard rate increased after 7 years of follow-up and was significantly lower during the first decade of follow-up compared with the second decade (log-rank 2 7.8, df 2, P.02). A similar trend for liver-related death or transplantation was not significant (log-rank ). Both all-cause death/transplantation rate (18% vs 42%; P.001) and liver-related death/transplantation rate (9% vs 27%; P.013) were lower during decade 1 compared with decade 2 of follow-up. Predictive factors for all-cause death or liver transplantation were similar (Table 4). On univariate analysis, they included age, decompensation at presentation, cirrhosis at any time, nontreatment with azathioprine, failure to achieve normal serum ALT levels within 12 months of starting treatment, and relapse rate. Of these, only presence of cirrhosis at any time was not a significant covariate on Cox multiple regression analysis. Sex (0.71 [ ]), nonwhite ethnicity (0.74 [ ]), year of diagnosis (1.09 [ ]), AIH score (1.01 [ ]), time to achieve normal serum ALT levels (0.82 [ ]), presence of overlap (1.61 [ ]), or treatment with immunosuppressive drugs other than prednisolone and azathioprine (0.74 [ ]) were not significantly associated with all-cause mortality. Although patients not treated with azathioprine had higher mortality than those who were (Figure 3E and F), this was due to an excess of early deaths in patients not given azathioprine. With analysis confined to patients followed up for 6 months, survival was unrelated to azathioprine treatment (data not shown). Percentage of follow-up time on azathioprine (Figure 3E), mean dosage (not shown), or mean dosage per kilogram of azathioprine (Figure 3F) over the period taking the drug were not predictive of death (either allcause or liver related) or transplantation, and none of these variables was predictive on Cox regression analysis (not shown). Results were similar with analysis confined to patients (n 203) not treated with other immunosuppressive agents apart from prednisolone and azathioprine (data not shown). Discussion Patients with AIH are generally considered to have a good prognosis, 80%-90% achieving clinical and biochemical remission on immunosuppressive treatment Reported 10-year survival rates are 85% to 95%. 3 6 However, there are few published data on patients with AIH followed up into their second decade. One study of 42 patients followed up for a median of 13.5 years 12 reported a liver-related mortality of 5%. However, in a report 6 from Kings College Hospital, of 238 patients followed up for a mean of 12 years, 10-year survival was 90% but 20-year survival was only 75%. Our cohort is probably representative of AIH in the United Kingdom, because more than 80% of patients were referred from primary care. Our patient population has been stable, with only 5% of patients moving out of the area during follow-up. As in other studies, 8,11,12,20 more than 90% of our patients attained clinical and biochemical remission defined by IAIHG criteria. 1 Despite this, long-term mortality was substantial: 18% and 52% after 10 and 20 years, respectively, which is higher than in other studies. Indeed, as suggested in Figure 2, mortality over the second decade of follow-up exceeded that during the first decade. One possible explanation is that median age at presentation in our cohort was 56 years, which is 12 to 14 years older than that in other studies of unselected patients with AIH. 3-8,11-13 Unsurprisingly, age was associated with all-cause (although not liver-related) mortality (Table 4). In our patients aged younger than 65 years at presentation, 10-year survival without transplant was 88%, which

7 1986 HOEROLDT ET AL GASTROENTEROLOGY Vol. 140, No. 7 Table 3. SMR All deaths All deaths or transplants No. of subjects No. of observed events SMR (95% CI) No. of observed events SMR (95% CI) All ( ) ( ) Age at diagnosis (y) Younger than ( ) ( ) ( ) ( ) Older than ( ) ( ) Presented since ( ) 1.76 ( ) First decade ( ) ( ) Second decade ( ) a ( ) a Third decade ( ) ( ) NOTE. Values in cohort as a whole and in subgroups. Event defined as either death only (from any cause) or as death (any cause) or liver transplant. a P.005 compared with SMR during first decade. is similar to that in the Kings cohort aged younger than 65 years. 6,23 However, corresponding survival after 20 years was only 55%, which is lower than in the Kings cohort. 6,23 In some reports 23,24 with a median follow-up time of 3.5 to 9 years, the outcome of AIH in the elderly was no worse than in those diagnosed younger, although in one UK study, mean age of presentation was 62 years and 5-year survival only 64%. 25 To assess whether mortality in AIH was in excess of that expected for age, we calculated the SMR. SMR has previously been reported as 4.0 in a survey 26 of all patients with a hospital diagnosis of lupoid hepatitis in Denmark between 1979 and This study included only 96 patients, probably an underestimate of the number of cases of AIH in a country of 5.3 million, given the reported prevalence of AIH in Scandinavia In other reports, 3,4 SMR values, although not given, are implied to be close to unity, because mortality in the AIH cohort was similar to that in the general population. We found that SMR in our overall AIH cohort was significantly greater than unity at 1.63 (1.86 if liver transplantation is regarded as death ). SMR was close to unity if only non liver-related deaths were considered. The small (approximately 4%) underestimation of this value 22 because of inclusion of liver-related deaths in the denominator (general population mortality rate) is unlikely to alter this conclusion. We have also found 30 that the rate of extrahepatic cancer in this AIH cohort was not significantly increased compared with that in the general population, as per Regional Cancer Registry data. Others have reported an excess of deaths in AIH from extrahepatic malignancy and cardiovascular disease. 27,31 However, nearly all the excess of deaths in our AIH cohort appears to be due to liver disease. The increased mortality during the second decade of follow-up (Figure 2) is underlined by the significantly higher SMR in the second compared with the first decade of follow-up (Table 3). The previously implied SMR values of close to unity in AIH 3,4 may be because follow-up was not long enough. We found higher SMR values in people presenting at a younger age, not because of increased absolute mortality but rather because of higher excess mortality over expected in younger patients. Unfortunately, younger patients presenting with AIH do not, despite immunosuppressive therapy, have a normal life expectancy. The substantial late liver-related mortality is probably a result of slow progression of liver injury in some patients, resulting eventually in cirrhosis and liver decompensation. Indeed, the rate of de novo cirrhosis development in our cohort was 34% after 20 years. In other studies, cirrhosis developed in 30% to 50% of 3 Figure 3. (A) Association of liver-related death or transplantation with presence of decompensation at presentation. (B) Association of liver-related death or transplantation with presence or development of cirrhosis at any stage (at diagnosis, developing subsequently, or time of onset unknown). Patients with no cirrhosis differed significantly from patients with cirrhosis onset of time unknown ( 2 12; P.01) and from the other groups ( ; P.001). (C) Association of liver-related death or transplantation with normalization of serum ALT levels in the first 12 months of immunosuppressive treatment. (D) Association of liver-related death or transplantation with relapse rate per decade of follow-up. Survival in those without relapse was not significantly different from those with one or more relapses when the latter group was pooled. However, unexpectedly, survival in those without relapses was lower (P.004) in patients without relapse than in those with 0.1 to 1.99 relapses per decade. If patients followed up for less than 1 year (n 15) are excluded, this difference is smaller but remains significant (P.028). (E) Association of liver-related death or transplantation with percent of follow-up time on azathioprine. Data missing in one patient. (F) Association of liver-related death or transplantation with mean dosage of azathioprine per kilogram body weight over the period while on the drug. Total number of patients in whom body weight available was 220.

8 June 2011 OUTCOME OF AUTOIMMUNE HEPATITIS 1987

9 1988 HOEROLDT ET AL GASTROENTEROLOGY Vol. 140, No. 7 Table 4. Variables Significantly Associated in Cox Regression Analysis With Liver-Related Death or Transplantation and With All-Cause Death or Transplantation Liver-related death or transplantation All-cause death or transplantation Hazard ratio (95% confidence interval) P value Hazard ratio (95% confidence interval) P value Decompensation at presentation 5.63 ( ) ( ).001 Failure to normalize serum ALT levels within ( ) ( ).001 months of starting treatment No. of relapses per decade 1.19 ( ) ( ).001 Nontreatment with azathioprine 3.96 ( ) ( ).001 Cirrhosis (at presentation or developing subsequently) 9.96 ( ) ( ).295 Age at presentation (y) ( ) ( ).001 NOTE. Analysis in the 192 patients presenting since 1987 (not shown) gave similar results. patients despite treatment. 3,6,9 12,17 Only one study 13 reported lower rates of cirrhosis development (13%). We assessed the parameters associated with an adverse outcome. Unsurprisingly, they included presence of liver decompensation at presentation. Cirrhosis at presentation has been associated with a poor outcome in some studies 5,15 but not in others. 3,11 In our patients, cirrhosis developing at any time (but not specifically cirrhosis at presentation) was an independent adverse prognostic factor. We found (Figure 3C and Table 4), as have others, 11,13 that failure of serum ALT levels to normalize over 12 months of treatment (presumably a marker of failure to suppress inflammation completely) was also predictive of both all-cause and liver-related death/ transplantation. Indeed, patients with a minimum serum ALT level in the lower normal range (0 27 U/L) had a better outcome over the first decade than those with a high-normal ALT level (28 55 U/L). In patients with chronic hepatitis B, 32 serum ALT values in the higher normal range are associated with more progressive disease. The fourth adverse factor was a high relapse rate. This is not surprising, given the liver damage resulting from repeated relapse episodes. Indeed, relapse rate was also independently associated with development of cirrhosis on follow-up. Multiple relapses have previously been associated with both development of cirrhosis and with liver-related death or transplantation. 10 In contrast to some reports, in which outcome was better in men 6 and worse in nonwhite patients, 16 we did not find that male sex or nonwhite ethnicity was associated with a worse outcome. Also, we found that patients with probable and definite AIH had a similar outcome. They also had similar initial response to prednisolone and relapse rates. Although some reports only include patients with definite AIH, 6,10 these similarities reinforce the view that probable AIH and definite AIH are the same disease. This study has some weaknesses. Because case capture was potentially incomplete in patients presenting before 1987, we may be underestimating overall mortality. However, SMR in the 192 patients recruited since 1987 was similar to SMR in the cohort as a whole. Also, Cox regression analysis on patients presenting since 1987 revealed essentially identical prognostic factors (Table 4) to those in the overall cohort. Our preferred management strategy was prednisolone plus azathioprine 1 mg kg 1 day 1 to induce remission followed by withdrawal of prednisolone and long-term maintenance azathioprine 2 mg kg 1 day 1. In practice, some patients had their dosages of azathioprine reduced because of concern about side effects; thus, some patients may have been undertreated with azathioprine. However, azathioprine has not been shown to reduce the risk of cirrhosis or liver-related death and long-term maintenance therapy is not routinely used in many centers. 3,5,11,13 We could not detect any relationship between outcome and either mean dose of azathioprine or percentage of time taking azathioprine. However, conclusions here must be provisional, given the retrospective analysis and the variable doses of azathioprine between individuals and at different times in the same individual. We do not believe these weaknesses alter our main conclusions: that the long-term outcome of AIH is less benign than has been widely assumed and that progression to cirrhosis and liver-related death (or need for a transplant) are common, despite immunosuppressive therapy, and may actually accelerate after 10 years of follow-up. Currently recommended treatment of AIH is based on immunosuppressive regimens evaluated in trials performed 20 to 40 years ago. These regimens, although dramatically reducing 5-year mortality rates, failed to induce histologic remission and/or prevent progressive fibrosis in many patients. More potent immunosuppressive regimens may be required for optimal management of patients with AIH. Supplementary Material Note: To access the supplementary material accompanying this article, visit the online version of

10 June 2011 OUTCOME OF AUTOIMMUNE HEPATITIS 1989 Gastroenterology at and at doi: /j.gastro References 1. Alvarez F, Berg PA, Bianchi FB, et al. International Autoimmune Hepatitis Group Report: review of criteria for diagnosis of autoimmune hepatitis. J Hepatol 1999;31: Soloway RD, Summerskill WHJ, Baggenstoss AH, et al. Clinical, biochemical, and histological remission of severe chronic active liver disease: a controlled study of treatments and early prognosis. Gastroenterology 1972;63: Roberts SK, Therneau TM, Czaja AJ. Prognosis of histological cirrhosis in type 1 AIH. Gastroenterology 1996;110: Kanzler S, Löhr H, Gerken G, et al. Longterm management and prognosis of autoimmune hepatitis (AIH): a single center experience. Z Gastroenterol 2001;39: Feld JJ, Dinh H, Arenovich T, et al. Autoimmune hepatitis: effects of symptoms and cirrhosis on natural history and outcome. Hepatology 2005;42: Al-Chalabi T, Underhill JA, Portmann BC, et al. Impact of gender on the long-term outcome and survival of patients with autoimmune hepatitis. J Hepatol 2008;48: Muratori P, Granito A, Quarneti C, et al. Autoimmune hepatitis in Italy: the Bologna experience. J Hepatol 2009;50: Czaja AJ, Menon N, Carpenter HA. Sustained remission after corticosteroid therapy for type 1 autoimmune hepatitis: a retrospective analysis. Hepatology 2002;35: Czaja AJ, Carpenter HA. Distinctive clinical phenotype and treatment outcome of type 1 autoimmune hepatitis in the elderly. Hepatology 2006;43: Montano-Loza A, Carpenter HA, Czaja AJ. Consequences of treatment withdrawal in type 1 autoimmune hepatitis. Liver Int 2007; 27: Verma S, Gunuwan B, Mendler M, et al. Factors predicting relapse and poor outcome in type I autoimmune hepatitis: role of cirrhosis development, patterns of transaminases during remission, and plasma cell activity in the liver biopsy. Am J Gastroenterol 2004;99: Seela S, Sheela H, Boyer JL. Autoimmune hepatitis type 1: safety and efficacy of prolonged medical therapy. Liver Int 2005;25: Miyaki Y, Iwasaki Y, Tereda R, et al. Persistent normalisation of serum alanine aminotransferase levels improves the prognosis of type 1 autoimmune hepatitis. J Hepatol 2005;43: Al-Chalabi T, Underhill JA, Portmann BC, et al. Effects of serum aspartate aminotransferase levels in patients with autoimmune hepatitis influence disease course and outcome. Clin Gastroenterol Hepatol 2008;6: Keating J, O Brien CJ, Stellon AJ, et al. Influence of aetiology, clinical and histological features on survival in chronic active hepatitis: an analysis of 204 patients. Q J Med 1987;62: Verma S, Torbenson M, Thulvath PJ. The impact of ethnicity on the natural history of autoimmune hepatitis. Hepatology 2007; 46: Kogan J, Safadi R, Ashur Y, et al. Prognosis of symptomatic versus asymptomatic autoimmune hepatitis: a study of 68 patients. J Clin Gastroenterol 2002;35: Montano-Loza A, Carpenter HA, Czaja AJ. Features associated with treatment failure in type 1 AIH and predictive value of the model of end-stage liver disease. Hepatology 2007;46: Hino T, Kumashiro R, Ide T, et al. Predictive factors for remission and death in 73 patients with autoimmune hepatitis in Japan. Int J Mol Med 2003;11: Montano-Loza AJ, Carpenter HA, Czaja AJ. Improving the endpoint of corticosteroid therapy in type I autoimmune hepatitis to reduce the frequency of relapse. Am J Gastroenterol 2007;102: Ishak K, Baptista A, Bianchi L, et al. Histological grading and staging of chronic hepatitis. J Hepatol 1995;22: Office for National Statistics. Mortality data and cause-specific mortality. Available at: Accessed December 29, Al-Chalabi T, Underhill JA, Portmann BC, et al. Autoimmune hepatitis (AIH) in the elderly: a systematic retrospective analysis of a large group of consecutive patients with definite AIH followed at a tertiary referral centre. J Hepatol 2006;45: Schramm C, Kanzler S, Meyer zum Büschenfelde KH, et al. Autoimmune hepatitis in the elderly. Am J Gastroenterol 2001; 96: Parker DR, Kingham JGC. Type 1 autoimmune hepatitis is primarily a disease of later life. Q J Med 1997;90: Poulsen LO, Thulstrup AM, Mellemkjaer L, et al. Mortality and causes of death in patients with lupoid hepatitis. A long-term follow-up study in Denmark. Dan Med Bull 2002;49: Sorensen HT, Thulstrup AM, Mellenkjaer L, et al. Longterm survival and cause-specific mortality in patients with cirrhosis of the liver: a nationwide cohort study in Denmark. J Clin Epidemiol 2003;56: Boberg KM. Prevalence and epidemiology of autoimmune hepatitis. Clin Liver Dis 2002;6: Ritland S. The incidence of chronic active hepatitis in Norway. Scand J Gastroenterol 1985;20(Suppl 107): Hoeroldt B, Farquharson N, Shepherd K, et al. Cancer in autoimmune hepatitis. Gut 2008;57(Suppl I):A Werner M, Almer S, Prytz H, et al. Hepatic and extrahepatic malignancies in autoimmune hepatitis. A long-term follow-up in 473 Swedish patients. J Hepatol 2009;50: Yuen MF, Yuan HJ, Wong DK, et al. Prognostic determinants for chronic hepatitis B in Asians: therapeutic implications. Gut 2005;54: Received January 21, Accepted February 25, Reprint requests Address requests for reprints to: Dermot Gleeson, MD, BSc, FRCP, P Floor, Royal Hallamshire Hospital, Sheffield S10 2JF, England. dermot.gleeson@sth.nhs.uk. Conflicts of interest The authors disclose no conflicts.