Potential conflict of interest: Nothing to report.
Hepatocellular carcinoma (HCC) has traditionally been considered a rare complication of cirrhosis secondary to autoimmune hepatitis (AIH), yet the true incidence remains unknown due to a lack of published data. Consequently, some professional guidelines do not mandate routine surveillance for HCC in this condition. Our aims were to evaluate the rate at which HCC develops among a large, prospectively obtained cohort of patients with AIH at a single center. Demographic, clinical, and laboratory indices associated with the development of HCC were also identified. HCC was discovered in 15 of 243 patients with AIH, all of whom had type 1 AIH equating to 1090 cases per 100,000 patient follow-up years. HCC occurred in the same proportion of females as males, 6.1% versus 6.4%, P = 0.95. HCC occurred more frequently in patients who had cirrhosis at presentation, 9.3% versus 3.4%, P = 0.048, or who had a variceal bleed as the index presentation of AIH, 20% versus 5.3%, P = 0.003. The median duration from time of confirmed cirrhosis to a diagnosis of HCC was 102.5 months, range 12-195 months. Median survival in patients whose HCC was diagnosed on surveillance was 19 months (range 6-36 months) compared with 2 months (range 0-14 months) for patients presenting symptomatically (P = 0.042). Conclusion: Cirrhosis in AIH is the sine qua non for HCC development, which subsequently occurs at a rate of 1.1% per year and affects men and women in equal proportions. (HEPATOLOGY 2008.)
Hepatocellular carcinoma (HCC) is the sixth leading type of cancer worldwide and represents a feared complication of cirrhosis, irrespective of etiology. Cognizant of the increased risk of HCC inherent upon the development of cirrhosis, the American Association for the Study of Liver Diseases (AASLD) and the British Society of Gastroenterology have issued recommendations for surveillance of HCC in certain patient subgroups.1, 2 Autoimmune hepatitis (AIH) with cirrhosis has, however, not been formally identified by either of these societies as a high-risk group mandating surveillance for HCC.
AIH is a chronic inflammatory disorder of unknown etiology that may proceed to cirrhosis and end-stage liver disease. In common with other patients with cirrhosis, such patients are at risk of developing HCC. However, the predominance of females in this condition might be expected to result in a lower incidence of HCC than other etiologies. Unlike other cohorts of patients with cirrhosis, the majority of patients with AIH respond to immunosuppression and even in the context of established cirrhosis, the potential exists for individuals to retain a near normal life expectancy. Thus, for patients with AIH, there may be a greater time interval during which HCC can develop as a complication. However, the interactions of disease activity, response to treatment, and other factors in relation to the risk of HCC development in AIH are unknown.
Although the development of HCC in patients with AIH and cirrhosis is considered a rare occurrence, the true incidence remains unknown due to the paucity of published data addressing this issue. Additionally, risk factors other than cirrhosis for the development of HCC in AIH are still to be determined.
In view of these unresolved issues, we set out to investigate the frequency with which HCC occurs in the context of AIH in a cohort of patients with standardized follow-up, to evaluate what, if any, demographic, clinical, biochemical, and treatment factors are associated with its development. Any such factors that are positively associated with the development of HCC could potentially be utilized in the future to target surveillance practices to groups at the highest risks of developing this complication.
AFP, alpha-fetoprotein; AIH, autoimmune hepatitis; AMA, antimitochondrial antibody; ANA, antinuclear antibody; BCLC, Barcelona clinic liver cancer; CLIP, cancer of the liver Italian programme; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; HLA, human leukocyte antigen; HR, hazard ratio; IAIHG, International Autoimmune Hepatitis Group; LKM-1, liver-kidney-microsomal 1; LT, liver transplantation; SMA, smooth muscle antibody.
Patients and Methods
The medical records of 243 patients (196 female, 47 male) with a diagnosis of definite AIH, based on the International Autoimmune Hepatitis Group (IAIHG)3, 4 score, presenting between 1971 and 2007 at King's College Hospital, London, UK, were examined from a prospectively recorded database. Patients were excluded from the study if they scored less than 17 points after treatment or less than 15 points before treatment, using the IAIHG scoring system, or if there was evidence of an overlap syndrome based on histological and/or cholangiographic findings. Viral hepatitis was excluded in all patients by testing for hepatitis B virus (HBV) surface antigen and hepatitis C virus (HCV) immunoglobulin G antibody. Additionally, for patients who presented before 1990, retrospective testing for HCV antibodies was performed on stored sera. Other causes of liver disease, such as excess alcohol, drugs, or herbal remedies had been excluded by appropriate history and investigations.
Histological material was available in 228 patients at index presentation. In 15 patients (1 male, 14 female), histological data were unavailable at accession either related to procedural risk, an inadequate specimen being obtained, or to patient reluctance to undergo biopsy. Despite the absence of histology analysis, all scored >17 points, using the IAIHG scoring system.3
In the absence of histology analysis, cirrhosis was diagnosed if individuals met radiological criteria for cirrhosis or had clinical evidence of portal hypertension. The presence of jaundice at presentation was determined clinically or biochemically, whereas ascites was determined either clinically or on ultrasound examination.
All patients were tested for the following antibodies: antinuclear antibody, smooth muscle antibody, antimitochondrial antibody, and liver-kidney microsomal-1 autoantibodies by indirect immunofluorescence on sections of rodent liver, kidney and stomach. Human leukocyte antigen (HLA) allotypes were tested in 147 patients (111 female and 36 male). Follow-up began from the time of diagnosis and terminated with the most recent outpatient appointment at King's College Hospital, at the time of death of the patient, or at the time of liver transplantation (LT). All patients with AIH had been initially treated according to a standard protocol with prednisolone at 0.5 mg/kg/day (20-40 mg/day) and azathioprine at 1 mg/kg/day (50-100 mg/day). The steroid dose was subsequently individually tapered to the lowest dose required to maintain biochemical remission (Fig. 1). In patients who sustained complete biochemical and clinical remission for at least 1 year on maintenance therapy, the azathioprine dose (in those who tolerated the drug) was increased to 2 mg/kg/day, the steroids were then gradually withdrawn, and (if remission was sustained) treatment was continued with azathioprine alone, as described in a previous study from this center.5 After this publication, it has not been our policy to completely withdraw immunosuppressive therapy from patients with AIH who remain in remission.
A complete response to treatment was defined according to the revised criteria of the IAIHG as either or both of the following: marked improvement of symptoms and return of serum aspartate aminotransferase (AST) or alanine aminotransferase (ALT), bilirubin and immunoglobulin values completely to normal within 1 year and sustained for at least a further 6 months on maintenance therapy, or a liver biopsy specimen at some time during this period showing at most minimal activity; or, either or both of the following: marked improvement of symptoms together with at least 50% improvement of all liver test results during the first month of treatment, with AST or ALT levels continuing to fall to less than twice the upper limit of normal within 6 months during any reductions toward maintenance therapy, or as liver biopsy within 1 year showing only minimal activity. Relapse was also defined using the IAIHG criteria as either or both of the following: an increase in serum AST or ALT levels of greater than twice the upper limit of normal or a liver biopsy showing active disease, with or without reappearance of symptoms, after a “complete” response as defined above; or reappearance of symptoms of sufficient severity to require increased (or reintroduction of) immunosuppression, accompanied by any increase in serum AST or ALT levels, after a “complete” response as defined above.3 A partial or no response to initial therapy were defined according to the original IAIHG criteria.4 Patients were defined as having achieved remission if they had normal serum biochemistry (globulins/immunoglobulin G/AST), had no symptoms indicative of a relapse, and (where available) liver biopsies showing only minimal activity with no necrosis.5
HCC was diagnosed, where possible, according to the AASLD guidelines,1 whereby a lesion greater than 2 cm on imaging with a raised alpha-fetoprotein (AFP) >200 ng/mL and arterialization on computed tomography or magnetic resonance imaging is considered diagnostic of HCC. Recently, however, the utility of contrast “washout” in the venous phase of a multiphased study has been appreciated. Therefore, for lesions less than 2 cm, the presence of this characteristic on one of the aforementioned imaging modalities can also be considered diagnostic of HCC in accordance with the AASLD consensus document. If radiology is nondiagnostic or the AFP is less than 200 ng/mL, then biopsy of the lesion is recommended.
The Barcelona Clinic Liver Cancer (BCLC) model,6 incorporates the Child-Pugh Score, Okuda stage, and physical performance status to derive not only prognostic estimates but suggested treatment strategies for each stage. Importantly, it identifies patients with early disease, who may be candidates for curative therapies. The AASLD1 consensus document therefore recommends the implementation of the BCLC model in the management of patients with HCC. Consequently, where clinical information is allowed, the BCLC stage was calculated for patients diagnosed with HCC among our cohort.
Continuous data are presented as median and range. For quantitative data, analyses were performed using the Mann-Whitney test for comparison of two independent groups. Differences in proportions were analyzed by the Fisher's exact test when the number of subjects was <5, and the chi-squared test for 2 × 2 tables when the number of subjects in all cells was >5. Crude incidence rates and univariate hazard ratios of HCC were calculated for relevant demographic, clinical, and laboratory variables.
All analyses were performed using the SPSS statistical software package, version 14 (SPSS Inc., Chicago, IL).
A total of 243 patients were identified with AIH (median IAIHG score 22, range 15-28) of whom 122 (50.2%) had definite histological or radiological evidence of cirrhosis or bridging fibrosis at any time during follow-up. There were 196 females (81%) and 213 (89%) of the patients were Caucasian. Median age at diagnosis of AIH was 42 years (range 5-80). At the time of study, median duration of follow-up was 11 years (range 1-36 years).
There were no statistically significant differences between males and females in terms of age at diagnosis of AIH or duration of follow-up, although there was a trend toward earlier diagnosis and therefore longer follow-up in men (Table 1). The frequency of cirrhosis, ascites, and variceal hemorrhage at presentation was also similar between males and females. However, males were more likely to be Caucasian than females (P = 0.03), whereas females had a higher median IAIHG score (P = 0.02) (Table 1).
Table 1. Demographics and Clinical Characteristics of Patients with AIH Attending Kings College Hospital
Fifteen patients were identified with HCC, comprising 12 women and three men. This figure represents 6.2% of the total cohort and 12.3% of all patients confirmed as being cirrhotic at any time during the period of follow-up. The crude HCC incidence rate was 1090 cases per every 100,000 patient years of follow-up or 1.1% per year. Fourteen patients were Caucasian and the other was of Afro-Caribbean descent. The incidence and univariate hazard ratios of HCC development by gender and race demonstrated no significant differences (Table 2).
Table 2. Crude Incidence Rates and Univariate Hazard Ratios of HCC Development in Patients with AIH
Median age at diagnosis of HCC was 63 years (range 25-89). In those who developed malignancy, the median age at diagnosis of AIH was 45 years (range 13-77 years), whereas in those who did not develop HCC, it was 45.5 years (range 5-80 years), P = 0.55.
There was no difference in the median duration of follow-up in patients who developed HCC, being 13 years in patients with HCC (range 4-32 years) and 11 years in those without HCC (range 1-36 years, P = 0.092). The median duration from time of confirmed cirrhosis to a diagnosis of HCC was approximately 9 years (102.5 months, range 12-195 months).
Clinical and Laboratory Features at Presentation and Risk of HCC.
All patients with HCC had cirrhosis on radiological and/or histological grounds. In the remaining cohort of patients who did not develop HCC, the prevalence of cirrhosis was 35% (P = < 0.001). Additionally, those with cirrhosis at presentation, or at any time, had a higher likelihood of developing HCC than those who did not (9.3% versus 3.4%, P = 0.048; 12.3% versus 0%, P = < 0.0001, respectively). Four patients who developed HCC did not havecirrhosis at accession biopsy, but progressed to histological cirrhosis during follow-up. In contrast, no patient with stable fibrosis developed HCC (P = < 0.0001). The crude incidence rates and univariate hazard ratios for these variables are presented in Table 2. Kaplan-Meier analysis of HCC development in patients who had AIH with or without cirrhosis was performed, and the result is demonstrated in Fig. 2.
A significantly lower proportion of patients (16.7%) who were jaundiced at presentation developed HCC, compared to those who were not (71.4%), P = < 0.0001. Conversely, the presence of variceal hemorrhage as the initial presentation was associated with an increased risk of subsequent HCC development, with three of 15 patients with HCC presenting in this way. Thus, the risk of development of HCC following a variceal bleed at presentation was 20% compared to 5.25% in patients who did not present with hemorrhage (P = 0.003). In contrast, the presence of ascites at presentation was not associated with the development of HCC at a later date (6.2% versus 4.5%, P = 0.67).
HCC occurred in similar proportions of patients who were positive for ANA or SMA. One hundred and fifty patients (113 female and 37 male) underwent assessment of HLA allotypes, and there was a nonstatistical trend toward greater DR3 positivity in males with a similar sex distribution of HLA DR4 (Table 1). Neither DR3 nor DR4 positivity were, however, associated with the development of HCC compared to individuals without these allotypes. Additionally, the median IAIHG score was not significantly different between the HCC and non-HCC groups, being 23 and 22, respectively (P = 0.85).
Response to Treatment of AIH and Risk of HCC.
The majority of patients (60%) received prednisolone monotherapy as induction therapy after diagnosis and a further 34% were treated with prednisolone and azathioprine in combination. Maintenance therapy predominantly utilized azathioprine (69%) either as monotherapy (30%) or in combination with prednisolone (39%).
HCC occurred in 6.2% of patients who were deemed relapsers and in 6.4% of complete responders to treatment with no relationship identified between the presence of relapses and the subsequent development of HCC (Table 2). Additionally, the presence, or otherwise, of multiple relapses (more than two) did not predict HCC development (4% versus 6.7%, P = 0.739). Furthermore, an incomplete or nonresponse to standard induction treatment did not predict subsequent development of HCC when compared to complete responders (5.7% compared to 14.3%, P = 0.19). Among patients treated with azathioprine at any time during follow-up, 7.2% developed HCC compared to 4.7% who did not receive this drug (P = 0.46). In addition, the median duration of therapy with azathioprine was not significantly greater for those patients with HCC, compared to those not diagnosed with HCC being 83 months (range 18-261 months) and 70 months (range 7-25 months), respectively, P = 0.42.
Diagnosis and Outcomes of HCC.
Eleven patients met the AASLD criteria for the diagnosis of HCC; two were diagnosed on examination of explanted liver, whereas one was diagnosed at postmortem examination. The remaining patient was diagnosed prior to any published guidelines regarding diagnosis of HCC1, 2, 7 and prior to the introduction of surveillance at our institution. This patient presented with a large tumor on ultrasound in association with a raised AFP level and died within 2 months of diagnosis.
Seven patients had HCC diagnosed as part of a surveillance program. In a further two patients, the diagnosis was made incidentally on histological evaluation of the explanted liver at the time of liver transplantation. Five patients presented with nonspecific symptoms, the investigation of which led to a diagnosis of HCC; in one individual, the diagnosis was made on postmortem examination (Table 3). Four patients underwent transplantation in the study group. However, because two of these individuals were not known to have HCC prior to transplantation, only two patients underwent transplantation with HCC as their primary indication. Two patients were initially listed for orthotopic liver transplantation and treated with transarterial chemoembolization as a temporizing measure but progressed beyond transplant criteria while awaiting a liver graft; another patient was treated with transarterial chemoembolization as a primary treatment modality. One patient was outside the Milan criteria, based on the total number of lesions, from the outset of diagnosis of HCC and was treated with radiofrequency ablation as a primary therapy. The remaining eight patients with AIH and HCC were not given any antitumor therapy due to a combination of advanced age in four patients, futility related to advanced disease in three patients, and a lesion only being identified on postmortem examination in another (Table 3).
Table 3. Individual Patient Characteristics of Patients Who Developed HCC
Tumor Size (cm)
Patients diagnosed prior to surveillance programme at KCH.
Patients in whom surveillance was considered inappropriate.
Patients still alive at the time of study/last follow-up.
Of the five lesions detected outside our surveillance program, three occurred in older patients in whom definitive cancer therapy was precluded due to their age and comorbidity and therefore surveillance was not undertaken, whereas HCC developed in two patients prior to surveillance being instituted at King's College Hospital. For the remaining patient, HCC was identified at postmortem examination.
Sixty percent of patients with HCC were noted to have an elevated AFP level (median 64 ng/mL, range 2-57,200 ng/mL). Tumor histology in the four patients who underwent transplantation demonstrated all lesions to be well-differentiated HCC, whereas two were associated with localized intrahepatic portal vein thrombosis. Percutaneous biopsy of a lesion less than 1 cm in a patient with indeterminate radiology and a normal AFP level also revealed a well-differentiated HCC. Eight patients were diagnosed with HCC prior to publication of the BCLC staging system; in another patient, the diagnosis was made on explant examination. Of the six patients diagnosed subsequent to introduction of the BCLC staging, two were stage 0, three were stage B, and one was stage C.
Ten patients diagnosed with HCC (67%) have died with five patients still alive at the time of study or last follow-up (Table 3). Median survival time was greater when HCC was identified by surveillance rather than presenting symptomatically (19 months versus 2 months, P = 0.042). Patients diagnosed on explant or on postmortem examination were excluded from this analysis.
Although HCC is traditionally thought to be a rare occurrence in AIH, we have demonstrated in a large, prospectively followed cohort of patients that it develops in approximately 12% of those with confirmed cirrhosis, when followed up for a median of 10 years. Analysis of only those patients with cirrhosis reveals a crude incidence of HCC of 1090 cases per every 100,000 patient years of follow-up, or 1.1% per year.
Direct comparison of HCC incidence with other etiologies is difficult due to heterogenous patient populations in these studies and the utilization of varying durations of follow-up. However, comparable published data from Europe8 suggests an incidence of HCC in HBV-related and HCV-related cirrhosis of 2000 and 3700 cases per 100,000 patient follow-up years, respectively. Meanwhile, Chen et al.9 reported higher HCC incidence rates of 6482 cases per 100,000 patient follow-up years in patients with cirrhosis and HBV from Southeast Asia. Other authors from the United Kingdom have reported the incidence of HCC in stage III or IV primary biliary cirrhosis at 6% or 796 cases per 100,000 follow-up years.10
Although the yearly incidence of HCC in patients with cirrhosis with AIH falls just below the AASLD recommended figure of 1.5% per year to meet cost effectiveness, this has been extrapolated predominantly from patients with HCV and has not been validated in other specific etiologies of cirrhosis. Furthermore, the yearly incidence of HCC among our cohort appears similar to that reported from the United Kingdom for primary biliary cirrhosis, an etiology for which surveillance is considered mandatory. In addition, other, albeit older, methods of determining effectiveness of an intervention have suggested that a survival improvement of 3 months or more as a result of surveillance suggests effectiveness.11 The median survival among our HCC surveillance cohort was 17 months longer than those symptomatically detected, which far exceeds this reported threshold. Whether this translates into HCC surveillance in AIH being a cost-effective strategy remains uncertain. However, the proximity of our yearly incidence of HCC in AIH to the AASLD recommended cutoff, its comparability to that reported in primary biliary cirrhosis among a similar ethnic population, and an apparent survival benefit for those undergoing surveillance suggests it may be worthwhile.
Cirrhosis in patients with AIH is the sine qua non for the development of HCC, and the presence of cirrhosis at presentation or at any time during follow-up is associated with HCC formation. The other dominant factor associated with subsequent development of HCC in this cohort was a variceal bleed at the time of the index presentation. Severe portal hypertension manifested by the presence of bleeding varices implies an inherent severity of underlying liver disease. Thus, bleeding from varices represents a surrogate marker of cirrhosis in this population. In contrast, the presence of ascites at index presentation was not associated with subsequent HCC development. Indeed, ascites as a clinical sign has been well described in AIH, in patients with disease of acute onset with severe hepatitis even in the absence of cirrhosis on histologic examination.
Regarding sex as a risk factor for HCC development, men are more likely to develop HCC than women especially in HBV, HCV, and hemochromatosis.12 In this study, we did not replicate this observation. This was despite men with AIH having a diagnosis made at an earlier age, a longer duration of follow-up, and equivalent rates of cirrhosis to women. There is no obvious explanation for this, although the relatively small numbers of patients followed could predispose our findings toward a type-2 error. Therefore, examination of larger cohorts of patients with AIH is required to clarify and explain this observed phenomenon. Interestingly, an examination of the long-term outcome of men with AIH attending King's College Hospital revealed more favorable outcomes generally among men when compared to women.13
Although previous concerns have been raised regarding occult viral hepatitis (particularly HCV) infection as being instrumental in the etiology of HCC in AIH, all patients in this study were serologically negative for HBV and HCV infection. We also excluded all patients who had been described in a previous study from this unit by Ryder et al. to avoid confusion with the current cohort.14 The impact of autoimmune disease activity on the incidence of hepatic malignancy is unknown. In patients with viral hepatitis, control of viral activity with subsequent improvement in hepatic necroinflammation may reduce the risk of HCC.15, 16
A further example of dual causality in the development of HCC is provided by hereditary hemochromatosis. Here, although the effects of iron and cirrhosis maybe as such, dual cofactors, studies to date have not examined the relative importance of these two factors,17, 18 and it has not been possible to separate them as causative events.19 Such observations suggest that control of hepatic inflammation could significantly reduce the incidence of HCC in AIH. Despite this, among our cohort of patients with AIH, neither the presence of nor the number of relapses correlated with the subsequent risk of developing HCC. That disease activity did not affect HCC development among our cohort in contrast to the effect seen in chronic viral hepatitis likely reflects the inherent carcinogenic nature of both HBV and HCV infection.20
The relationship between immunosuppressive therapy and the development of HCC is of theoretical interest. On one hand, control of the necroinflammatory activity might be expected to reduce the oncological risk. However, this benefit might be offset by the permissive effects of immunosuppression. In addition, several studies have raised the concern regarding the oncogenic potential of azathioprine when used for any indication.21–23 This issue becomes of paramount importance when using such agents in patients with a predisposition to malignancy, such as patients with cirrhosis and AIH. However, this study did not demonstrate an association between the use of azathioprine and development of HCC.
Our findings suggest that the development of HCC is purely dependent on cirrhotic transformation which, in some individuals, occurs without apparent clinical or biochemical flares in activity.24 Among our cohort, jaundice at presentation was inversely associated with the development of HCC, and this clinical sign appears to identify a proportion of patients who present with very active disease but who are not yet cirrhotic. This again is consistent with observations from our unit.25 Because such individuals are usually profoundly symptomatic, they are more likely to be diagnosed and treated sooner. Thus, earlier intervention in symptomatic patients may prevent progression to cirrhosis and thus reduce the risk of HCC development.
In summary, only the presence of cirrhosis (or its surrogate marker of variceal bleeding) is significantly associated with an increased risk of a diagnosis of HCC during follow-up of patients with AIH. The majority of patients presented with HCC having had cirrhosis for a period of approximately 9 years. Although the incidence of HCC is less common than in other chronic liver diseases,8, 10, 16, 26, 27 we believe that the risk remains sufficient to undertake surveillance in all patients with cirrhosis with AIH who are fit enough to undergo curative therapies.