Chronic rejection (CR) is one of the causes of late liver

Transcription

1 Accuracy of Bile Duct Changes for the Diagsis of Chronic Liver Allograft Rejection: Reliability of the 1999 Banff Schema Mylène Sebagh, 1 Karin Blakolmer, 1 Bru Falissard, 3 Bru Roche, 2 Jean-Francois Emile, 1 Henri Bismuth, 2 Didier Samuel, 2 and Michel Reynès 1 Chronic rejection (CR) after liver transplantation is thought to be a dynamic and potentially reversible process. The Banff working group has developed recommendations for its histopathologic staging. The 1999 Banff classification of CR (i.e., bile duct dystrophy >50% and/or bile duct loss >20%) was applied to: 1) biopsies from patients retransplanted for CR (N 19) and pathologies other than CR (N 21) to evaluate its specificity and sensitivity, especially of the early stage lesions of CR; and 2) biopsies from nretransplanted patients (N 21) to evaluate the evolution of CR lesions. Atypical forms of CR were also described. Including an early stage into the definition of CR has resulted in a much higher sensitivity for its diagsis, as compared with the former classification (i.e., bile duct loss >50%) (89% vs. 33%; P.0001), while keeping an acceptable specificity (74% vs. 100%; P.03). In 55% of the nretransplanted patients, CR lesions were reversible. No histologic feature reliably predicted CR outcome. Transient lobular hepatitis, unrelated to viral infection, and veocclusive disease were seen significantly more often in the CR group (P.04 and P.03, respectively). We conclude that the application of the 1999 Banff classification is superior to the previous classification for the diagsis of CR. However, limited information can be drawn regarding the outcome of CR based on histology alone. Transient lobular hepatitis, unrelated to viral infection and ve-occlusive disease, may be an unusual expression of CR. (HEPATOLOGY 2002;35: ) Chronic rejection (CR) is one of the causes of late liver allograft dysfunction and failure, although its incidence is relatively low in contrast to other vascularized allografts. 1 The mechanisms of CR are complex. Its natural history and possible evolution from acute cellular rejection (ACR) are t well understood. In an effort to better understand chronic liver allograft rejection, the Banff working group, which has previously reported a schema for grading acute liver allograft rejection, 2 has developed recommendations for the histopathologic staging and reporting of CR. 3 Because CR is thought to be a dynamic process, this classification recognizes an early stage of CR. The histologic diagsis of CR is based on two principal features 4 : obliterative arteriopathy, affecting the medium-sized arteries, which are almost never present in biopsies, and loss of small bile ducts. CR was previously diagsed on a biopsy, when bile ducts were absent in more than 50% of the portal tracts. 5,6 The Banff 1999 classification broadens the Abbreviations: CR, chronic rejection; ACR, acute cellular rejection. From the 1 Service d Anatomie Pathologique, 2 Centre Hépato-biliaire, and 3 Unité de Santé Publique, Hôpital Paul Brousse, Villejuif, France. Received April 18, 2001; accepted October 5, Address reprint requests to: Mylène Sebagh, M.D., Service d Anatomie Pathologique, 14, avenue Paul Vaillant Couturier, Villejuif Cedex, France. mylene.sebagh@pbr.ap-hop-paris.fr; fax: (33) Copyright 2002 by the American Association for the Study of Liver Diseases /02/ $35.00/0 doi: /jhep definition of CR by including an early stage defined as bile duct loss between 20% and 50% of portal tracts. It also introduces bile duct dystrophy into the definition of (early) CR, if a majority of ducts are affected. Bile duct loss in more than 50% of portal tracts defines late-stage CR, which corresponds to the previous definition of CR. Liver CR has been shown to be potentially reversible. It reverses more often in its early, than its late, stage and may be amenable to therapy. 7,8 Early recognition is therefore important. However, the morphologic changes of early CR can also occur in n CR-related complications, such as bile duct obstruction and arterial thrombosis, thus potentially decreasing the specificity of these lesions. In this study, the 1999 Banff schema for CR was applied to cases from our transplantation center. The first aim was to evaluate the specificity and sensitivity, especially of the early-stage lesions of CR on liver biopsies, by comparing the histologic findings of patients retransplanted for CR with those of a control group of patients retransplanted for other indications. Atypical forms of CR were also described. The second aim was to examine biopsies from patients who were t retransplanted for histologic changes associated with progression and reversibility of CR. Materials and Methods Patient Population. Between 1990 and 1999, 911 patients underwent transplantation at the Hepatobiliary Center of Paul Brousse Hospital, Villejuif, France. For the first purpose of the 117

2 118 SEBAGH ET AL. HEPATOLOGY, January 2002 study, the population was drawn from the patients undergoing a first retransplantation. Only patients retransplanted after at least 2 weeks posttransplantation (a total of 40) were included in the study group. Patients with early graft failure caused by surgical complications were excluded. The 2-week interval was chosen because graft failure caused by CR has been reported as early as 2 weeks after transplantation. 9 Of the 40 retransplanted patients, 19 patients were because of CR, verified on the failed graft, and composed the CR group. The 21 other patients retransplanted for causes other than CR composed the control group. For the second purpose of the study, the patient transplanted before and the patient transplanted after each patient of the CR group were selected because they presumably underwent the same immusuppressive policy. Patients who had a posttransplantation course of less than 2 weeks (6 patients), patients who had only one biopsy available (1 patient), and patients who were retransplanted (10 patients) were excluded. Thus, 21 patients were included and composed the nretransplanted group. Histologic Evaluation. Liver biopsies were routinely performed at 1, 2, 5, and 10 years after transplantation, and when clinically indicated. Tissue was fixed, embedded in paraffin, and stained with hematein-eosin-safran, Picrosirius for collagen, and Perls for hemosiderin. All liver biopsies and the failed allografts were reviewed by two pathologists (K.B. and M.S.), together, without kwledge of the indication of retransplantation. Portal tracts, lobular parenchyma, and centrilobular veins were evaluated in a systematic fashion, using a pre-established format. Liver biopsies were given one or several diagses, including chronic rejection and its stage. From the failed grafts, sections taken from the right and left lobes were reviewed using the same criteria as for biopsies. To evaluate the percentage of bile duct loss and bile duct dystrophy in the failed grafts, 50 portal tracts were counted in randomly chosen fields. Sections taken from the hilum allowed for analysis of the main bile duct for the presence of fibrosis and inflammation, and the large and medium arteries for obliteration of the lumen of more or less than 50%, presence of foam cells, and fibrosis and thrombosis. Histologic Definition of Chronic Rejection. The diagsis of CR was confirmed by histologic examination of the failed graft, based on the presence of bile duct loss and obliterative arteriopathy, which fulfilled the established criteria. 4 For the review of liver biopsies, we applied the classification of CR defined at the 5th Banff Conference. 3 Briefly, this classification distinguishes early and late CR. Early CR is identified by dystrophic changes of the biliary epithelium affecting a majority of the bile ducts, with or without bile duct loss. Dystrophic changes of the biliary epithelium included uneven spacing of biliary cells, eosiphilic transformation of the cytoplasm, nuclear enlargement and hyperchromasia, or reduction of a duct to 2 or 3 cells (Fig.1). Late CR is defined by bile duct loss in more than 50% of the portal tracts. The late stage only corresponds to the previous definition of CR. Radiologic Assessment. Because histologically, a pattern of duct damage similar to that of CR can occur as a result of n rejection-related complications, such as obstructive cholangiopathy and hepatic artery stricture, data of all available radiologic material were collected. Cholangiograms were obtained routinely through the T-tube between the 2nd and the 4th postoperative Fig. 1. Bile duct dystrophy. This bile duct shows uneven spacing of biliary cells, densification of the cytoplasm, nuclear irregularity, and hyperchromasia. Note the presence of 3 arterial sections near the bile duct. week. Additional cholangiograms were performed when clinically indicated. Ultrasound for the evaluation of vascular patency was performed on days 1, 7, 21, and at 3 months after transplantation. Arteriograms were performed when vascular patency could t be assessed by ultrasound. Statistical Methods. The t test for quantitative variables, the Mann-Whitney test for semiquantitative variables, and the 2 test (or Fisher test for small samples) for qualitative values were applied. The McNemar test was used for paired comparison of percentages. All tests were 2-sided. The type 1 error was fixed to 5%. Results Retransplanted Patients (CR Group and Control Group) Demographic Data. Table 1 summarizes the demographic characteristics of the retransplanted patients. The CR group (N 19) and the control group (N 21) were statistically comparable for sex (13M/6F vs. 9M/12F), age at the time of the transplantation (mean age, 43 vs. 41 years), and primary disease. In contrast, the interval between transplantation and retransplantation was significantly longer for the CR group. Patients in the CR group were retransplanted after an average period of 566 days (range, 109-2,240 days), whereas the interval between the first transplantation and the retransplantation in the control group was 258 days (range, days; P.04). The initial immusuppressive treatment was comparable for the 2 groups and was essentially based on corticosteroids, cyclosporine, and azathioprine. Most patients in the CR group, but t in the control group, were later switched to tacrolimus. Biopsies. Of the 178 biopsies available for review, 6 biopsies were inadequate. Eighteen of the 19 patients in the CR group and 19 of the 21 patients in the control group had biopsies, with an average of 4.3 biopsies per patient. The average number of portal tracts per biopsy was 18.8 (range, 3-71). Figure 2 shows the repartition of liver biopsies by number of portal tracts. The mean number of portal tract per millimeter was 1.1 ( ). In the CR group, all but 2 patients had a diagsis of CR on the biopsies preceding graft failure. This corresponds to an overall

3 HEPATOLOGY, Vol. 35, No. 1, 2002 SEBAGH ET AL. 119 Table 1. Demographic Data of Patients of the CR Group, the Control Group, and the Nonretransplanted Group CR Group Control Group Nonretransplanted Group N Sex (M/F) 13/6 9/12 14/7 Mean age at the time of OLT Primary disease Chronic hepatitis B (HDV, HCC) 3 (2, 0) 6 (2, 1) 4 (2, 1) Chronic hepatitis C (HCC) 6 (3) 5 (0) 8 (3) HCC Alcoholic cirrhosis Cryptogenic cirrhosis PBC PSC AIH Fulminant hepatitis (drug, undetermined, HBV) 3 (1, 0, 2) 2 (2, 0, 0) 2 (1, 0, 1) Budd-Chiari syndrome Amyloidosis Protoporphyria Mean graft survival in days (range) 566 (109-2,240) 258 (15-870) 2,751 (439-4,143) Immusuppressive regimen Corticosteroids: cyclosporine, azathioprine Corticosteroids: tacrolimus ALS-corticosteroids Switch cyclosporine/tacrolimus Abbreviations: OLT, orthotopic liver transplantation; reolt, orthotopic liver retransplantation; HDV, hepatitis delta virus; HBV, hepatitis B virus; HCC, hepatocellular carcima; PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis; HAI, autoimmune hepatitis; ALS, anti-lymphocyte serum. sensitivity of 89%. The interval between transplantation and the first diagsis of CR ranged from 17 to 1,163 days (median, days; mean, days). CR was of early type only in 10 patients (55%). Late-type lesions without preceding early-type lesions were seen in 4 patients (22%). Two patients (11%) had early-type, then late-type, lesions of CR. Of the 105 biopsies of the CR group, 34 showed features of CR (Table 2). Early CR was seen in 22 biopsies. The diagsis was based on bile duct loss only in 54% of biopsies, based on bile duct dystrophy only in 31%, and on both in 13% of biopsies. Late CR was present in 12 biopsies. Bile duct loss was associated with dystrophy of more than 50% of bile ducts, in 75% of biopsies. Obliterative arteriopathy was never seen. Figures 3A and 3B show the evolution of bile duct loss and bile duct dystrophy, respectively, in successive liver biopsies. As expected, the pattern was progressive. Fig. 2. Repartition of liver biopsies by a number of portal tracts in the retransplanted patients (CR group and control group). In the control group, 14 patients had CR-related changes. The 5 remaining patients displayed early-type CR lesions. This yields an overall specificity of 74%. These lesions occurred between 10 and 57 days after transplantation (median, 36 days; mean, 34.2 days). All 5 patients had bile duct loss ranging from 22% to 43% of portal tracts. Only 1 patient also had bile duct dystrophy (71% of bile ducts). These lesions were attributed to extrahepatic biliary obstruction in 2 patients and to arterial thrombosis in 1 patient. In the other 2 patients, the lesions were unexplained. When only the liver biopsy with the most bile duct loss was considered, there was, as expected, a significant difference between the 2 groups. The mean percentage of bile duct loss was less than Table 2. Number of Biopsies With CR Features in Patients With CR and Controls CR Group (N 19) Control Group (N 21) No. of patients with biopsies 18 (100%) 19 (100%) No. of patients with biopsies with CR changes 16 (89%) 5 (26%) No. of patients with biopsies showing early CR changes 12 (67%) 5 (26%) No. of patients with biopsies showing late CR changes 6 (33%) 0 No. of biopsies 105 (100%) 73 (100%) No. of biopsies with CR changes 34 (32%) 13 (18%) No. of biopsies with early CR changes 22 (21%) 13 (18%) BDD only ( 50%) 7 (32%) 1 (8%) BDL only ( 50%) 12 (54%) 11 (84%) Both 3 (14%) 1 (8%) No. of biopsies with late CR changes 12 (11%) 0 BDD 50% 3 (25%) 0 BDD 50% 9 (75%) 0 Abbreviations: BDD, bile duct dystrophy; BDL, bile duct loss.

4 120 SEBAGH ET AL. HEPATOLOGY, January 2002 Fig. 3. Evolution of bile duct loss and bile duct dystrophy, respectively, in successive liver biopsies of the patients retransplanted for CR. The course was essentially progressive. 50% in both groups (39% vs. 11%; P.002). The percentage of bile duct dystrophy was significantly higher in the CR group (mean, 59% vs. 17%; P.01). Table 3 compares the sensitivity and the specificity for the diagsis of CR according to different percentages of bile duct loss and bile duct dystrophy and their combinations. Bile duct loss of 50% and bile duct dystrophy of 50% had a similarly high specificity for CR, but the sensitivity was low for bile duct dystrophy and was even lower for bile duct loss. Bile duct loss of 20% and bile duct dystrophy of 20% had almost the same specificity and sensitivity. The Banff 1999 classification (bile duct dystrophy of 50% and bile duct loss of 20%) was superior to the previous definition of CR (bile duct loss of 50%) with respect to sensitivity (89% vs. 33%; P.0001), but the specificity was significantly higher using the previous classification (74% vs. 100%; P.03). When bile duct dystrophy of more than 20% and bile duct loss of more than 20% were combined, sensitivity for the diagsis of CR was 100% and specificity was 57%. This was t significantly different from the Banff classification (P.16). Among the other histologic features recorded, only the presence of foam cell clusters, which although seen in biopsies of both groups, was significantly more frequent in the CR group (P.045). This was independent of the presence of cholestasis. There was difference in the occurrence of ductular proliferation between the groups. Neither perivenular inflammatory infiltrate r perivenular fibrosis r necrosis was significantly different between the groups.

5 HEPATOLOGY, Vol. 35, No. 1, 2002 SEBAGH ET AL. 121 Table 3. Comparison of Sensitivity and Specificity of the Percentage of Bile Duct Loss and Bile Duct Dystrophy for the Diagsis of CR Banff Classification BDD > 50% and/or BDL > 20% Previous Classification BDD > 20% CR Non-CR Sensitivity *P Specificity *P % BDL > 20% BDD > 50% % % BDL > 50% % % % 74% 95% 100% 74%.03 BDD > 20% and/or BDL > 20% %.16 57%.16 Abbreviations: BDD, bile duct dystrophy; BDL, bile duct loss. *As calculated with the McNemar test for paired comparison of percentages. There was difference in the incidence, the number, and the severity of episodes of ACR between the 2 groups. Seven patients of the CR group experienced at least one episode of ACR. Three patients had more than one episode. ACR was never severe. In the control group, 10 patients had ACR; 7 patients had one episode only. One patient had severe ACR. There was case of persistent ACR in either group. Because some authors suggested the possible role of a lobular inflammatory reaction in the developing phase of CR,10 this feature was investigated for in biopsies of all patients transplanted for reasons other than viral or autoimmune hepatitis. The histologic features included spotty hepatocyte necrosis and an inflammatory infiltrate composed of clusters of lymphocytes. These features were distributed evenly throughout the lobule (Fig. 4). This lobular hepatitis was found in 5 of 8 patients of the CR group (2 of 3 patients transplanted for PBC, 2 of 3 patients transplanted for CSP, and 1 of 1 patient transplanted for drug-induced fulminant hepatitis), and in 1 patient of the control group, originally transplanted for alcoholic liver disease (62.5% vs. 11%; P.04). Hepatitis caused by a hepatotropic virus was seen in 10 patients of the Fig. 4. Lobular hepatitis characterized by spotty hepatocyte necrosis and a cluster composed of lymphocytes and hystiocytes (*). (Hematein-eosinsafran stain.) CR group (recurrent or de vo infection) and in 8 patients of the control group (only recurrent infection). Three patients of the CR group and 10 patients of the control group never developed hepatitis. Features of ve-occlusive disease were present in 6 patients of the CR group and in 2 patients of the control group. These histologic features included complete or incomplete occlusion of lumen of the central vein by connective tissue and pericentral hemorrhagic necrosis (Fig. 5). Portal veins were never affected. Four patients of the CR group had concomitant ACR with central venulitis. Failed Allografts. This diagsis was made based on the combination of histologic and radiologic findings (data t shown). In the CR group, CR exclusively was present in 6 failed allografts. Concomitant pathology was present in 13 patients: 5 patients had ve-occlusive disease, 7 patients had chronic viral hepatitis, and 1 patient had both. In the control group, patients were retransplanted for arterial thrombosis with cholangitis-type related lesions (N 8), chronic viral hepatitis with or without cirrhosis (N 5), ve-occlusive disease (N 1), recurrent disease (N 2; protoporphyria, primary sclerosing cholangitis), primary nfunc- Fig. 5. Ve-occlusive disease. Incomplete obliteration of the centrilobular vein surrounded by hemorrhagic necrosis. (Picrosirius stain.)

6 122 SEBAGH ET AL. HEPATOLOGY, January 2002 Table 4. Histologic Findings in the Hilum of Failed Allografts in the CR and Control Groups CR Group Hilar Bile Ducts Control Group Normal 0 3 Inflammation 1 0 Fibrosis 9 8 Inflammation and fibrosis 9 10 Hilar Arteries Septal Arteries CR Group Control Group CR Group Control Group Normal Foam cell infiltration* 50% % Fibrosis* 50% % Foam cell infiltration and fibrosis* 50% % Arterial thrombosis *Degree of occlusion of the arterial lumen. tion (N 3), and posttransplantation lymphoproliferative disorder (N 1). One patient had cryptogenic cirrhosis of the graft, which was also her primary diagsis. Ve-occlusive disease in the failed grafts was seen significantly more often in the CR (P.03). Table 4 summarizes the histologic changes in the hilum of the failed allografts. Changes of large perihilar bile ducts were similar in both groups. Most of them showed mural fibrosis, accompanied by nspecific inflammation. In contrast, arterial changes differed in the 2 groups. In the hilar arteries, the main change was foam cell infiltration in 15 of 19 patients of the CR group (79%). Arterial thrombosis occurred in 8 of 21 patients of the control group (38%). Foam cell infiltration with or without fibrosis of the septal arteries was present in 11 of 19 failed allografts with CR (58%), while the majority of septal arteries were rmal in the control group. In total, 18 of 19 patients of the CR group had foam cell arteriopathy affecting only large arteries (N 5), only septal arteries (N 3), or both arteries (N 10). The only patient without arteriopathy had bile duct loss of 62% and bile duct dystrophy of 26%. There was difference in the degree of occlusion of the large arteries between the 2 groups in contrast to the septal arteries that were significantly more often occluded in the CR group than in the control group (P.001). Interestingly, bile duct loss of 50% was present only in 12 of 19 allografts with CR (63%; mean, 56%), which was, however, significantly higher than that found in the liver biopsy with the most bile duct loss (P.024). There was correlation between the degree of bile duct loss and the degree of occlusion of large or septal arteries. Bile duct dystrophy of 50% (mean, 69%) was present in 14 of 19 allografts (74%). The degree of bile duct dystrophy correlated with the degree of occlusion of septal, but t large, arteries (P.048). Nonretransplanted Patients Demographic Characteristics and Clinical Course. The demographic characteristics of the nretransplanted patients are given in Table 1. In particular, initial immusuppressive treatment was based on corticosteroids, cyclosporine, and azathioprine in all but 1 patient. Only 1 patient was later switched to tacrolimus. At the time of writing, 17 patients were alive with rmal or near-rmal liver function tests. Three of them developed an extrahepatic neoplasia (pancreatic carcima, bronchial carcima, and ocular melama). Four patients died. Causes of death were recurrent and metastatic hepatocellular carcima in 2 patients, cirrhosis caused by recurrent hepatitis C virus infection in 1 patient, and hepatic failure resulting from CR in 1 patient (patient 10), who was awaiting retransplantation. The mean graft survival (at the time of last clinical follow-up or at the time of death) was 2,751 days posttransplantation (range, 439-4,143 days). Biopsies. Ninety-eight biopsies, with an average of 4.4 biopsies per patient, were available for review. The average number of portal tracts per biopsy was 17 (range, 2-42). Figure 6 gives the repartition of liver biopsies by number of portal tracts. A review of liver biopsies showed that 10 patients had CRrelated changes after transplantation. In the 11 remaining patients, a diagsis of CR was made on at least one biopsy. The interval between transplantation and the first diagsis of CR ranged from 87 to 1,983 days (median, 364 days; mean, 555 days) and was t significantly different from that of the CR group. The diagsis of CR was always based on bile duct loss and/or bile duct dystrophy, and never on the presence of obliterative arteriopathy. According to the 1999 Banff classification, CR was of early type only in 8 patients and of late type only in 1 patient. Both early- and late-type CR lesions were present in 2 patients. Figures 7A and 7B illustrate the evolution of bile duct loss and bile duct dystrophy, respectively, in successive liver biopsies. There was single pattern with regard to the course of bile duct damage: In the 8 patients with only early CR, 3 had progressive bile duct loss and/or dystrophy (patients 10, 16, 20). In 3 patients, bile duct loss and/or dystrophy decreased over time (patients 11, 14, 17). In the 2 remaining patients, the diagsis of CR was made on the last follow-up biopsy (patients 19, 21). In the 3 patients with late CR, Fig. 6. Repartition of liver biopsies by number of portal tracts in the nretransplanted patients.

7 HEPATOLOGY, Vol. 35, No. 1, 2002 SEBAGH ET AL. 123 Fig. 7. Evolution of bile duct loss and bile duct dystrophy, respectively, in successive liver biopsies of the nretransplanted patients. There was single pattern. Note that patient 6 had complete reversion of the CR lesions after conversion to tacrolimus. 2 (patients 6, 15) had complete reversion of the CR lesions. Thishappened after conversion to tacrolimus in patient 6. In the remaining patient (patient 8), alternating biopsies showed early and late CR, with recognizable trend. The last follow-up biopsy showed late CR changes. In the 10 patients without CR, 2 (patients 2 and 13) had bile duct damage attributed to bile duct obstruction. Discussion In this study, the 1999 Banff classification of CR has been applied to biopsies from patients retransplanted for CR and pathologies other than CR, and from nretransplanted patients. Including an early stage into the definition of CR has resulted in a much higher sensitivity for its diagsis, as compared with the old

8 124 SEBAGH ET AL. HEPATOLOGY, January 2002 classification, while keeping an acceptable specificity. However, limited information can be drawn regarding the outcome of CR based on histology alone. This study allowed the identification of atypical, previously poorly recognized forms of CR. Histologically proven CR requiring retransplantation occurred in 2% of patients transplanted in our center after This is in keeping with the results of other groups. 1 In the first part of this study, the presence or absence of CR was proven by the gold standard, the failed graft, removed at the time of retransplantation. When the liver biopsies from patients with CR and patients retransplanted for reasons other than CR were tested for CR features described in the new Banff classification, the sensitivity for the diagsis of CR was 89%. Specificity was still within acceptable limits (74%). These findings compare with previous reports describing similar rates. 11 The false-positive diagses of CR were made in patients with hepatic artery thrombosis, bile duct obstruction, and in 2 unexplained cases. All false-positive cases showed changes in keeping with early CR. It is possible that the 2 cases with bile duct dystrophy and bile duct loss, seen on several biopsies, had, in fact, early CR, which was reversible in 1 patient. The second patient was retransplanted for ve-occlusive disease. 9 When the old classification was applied, and only biopsies with 50% bile duct loss were diagsed as CR, the specificity was 100%. However, the vast majority of biopsies would t have been diagsed correctly before retransplantation, because the sensitivity was only 33%. Furthermore, more than half of our patients never had signs of late CR on biopsies preceding retransplantation, and would never have been detected with the previous classification. In contrast to other allografts, liver CR is reversible 6,9,11,12 and may be amenable to treatment. 8 It may therefore be important to recognize CR as early as possible. In the 1999 Banff classification, the diagstic criteria for CR in a biopsy specimen are bile duct lesions and foam cell arteriopathy. In this study, foam cell arteriopathy was never seen in biopsies, and the diagsis of CR on biopsies was based only on bile duct lesions. We therefore aimed to define the minimal criteria for the diagsis of CR, taking into account only bile duct dystrophy and bile duct loss. In the Banff 1999 classification, the minimal criteria for the diagsis of early CR are 50% bile duct dystrophy and/or 20% bile duct loss. In our patients, bile duct dystrophy of 50% had a specificity for the detection of CR, which was almost as high as that for bile duct loss of 50%. Sensitivity as compared with the previous classification increased; however, almost half of the patients with CR would t have been detected if only bile duct dystrophy of 50% had been taken into account. Bile duct loss of 20% had a sensitivity and specificity around 75%. The combined values of bile duct loss of 20% and bile duct dystrophy of 50% resulted in a higher sensitivity of 89%, without increasing the specificity. Surprisingly, bile duct loss of 20%, which is one of the Banff criteria for the diagsis of CR, and bile duct dystrophy of 20%, which is below the Banff value of 50%, had the same specificity of 74%, whereas bile duct dystrophy of 20% had a slightly higher sensitivity of 89%. These values corresponded exactly to the sensitivity and specificity of the combination of bile duct loss of 20% and bile duct dystrophy of 50%. When bile duct loss of 20% and bile duct dystrophy of 20% were combined, the sensitivity was higher and the specificity was lower than the Banff classification. Although the specificities were t significantly different, a specificity of 57% may t be acceptable for the proposition of lowering bile duct dystrophy from 50% to 20% for the diagsis of CR. However, the number of patients in this study, as in all studies of chronic liver rejection, was small, and this difference may become statistically significant if a larger patient population is studied. This should be verified by additional studies. Among the other features described in CR, the presence of foam cell clusters was the only relevant sign in this study. This was independent of cholestasis. Perivenular fibrosis and/or necrosis were t significantly different between the 2 groups. 13 This finding was t surprising because of the great number of patients with arterial thrombosis in the control group. There was difference in the occurrence of bile duct proliferation between the patients retransplanted for CR and the control group. Both showed ductular proliferation: In the control group, prominent ductular proliferation was caused by arterial thrombosis and secondary bile duct obstruction. In the CR group, it could represent an element of obstruction secondary to foam cell arteriopathy. Some authors suggested that ductular reaction at the interface may be associated with reversible CR and may be caused by bile duct regeneration. 6,12 This hypothesis was t supported by results obtained in our patients with irreversible CR. This study has also shown that the distinction between the two stages of CR, which is made with regard to a potential reversibility of the early stage, is often inaccurate on a biopsy. The distinction between the two stages rests on both bile duct and arterial changes. In a biopsy, medium-sized arteries are almost never present. In our study, there was dissociation between the severity of bile duct changes and the degree of artery obstruction in one third of the failed grafts, in which bile duct morphology remained within the limits of early changes while there was significant arterial obstruction. In addition, when liver biopsies were evaluated prospectively, the course of CR changes in the nretransplanted patients did t follow a single pattern. Reversibility of CR lesions was seen in 55% (5 of 9) of patients with early and late CR. We were unable to identify histologic parameters that could predict a favorable course. There was also significant difference in the onset of CR lesions between the CR group and the nretransplanted group. Thus, the course (rather than a given percentage at one time posttransplantation), combined with clinical information, liver function tests, and the status of immune suppression, should be taken into account to make a treatment decision. An objective was also to describe atypical features of CR. Two particular histologic findings were associated with CR. Firstly, transient lobular hepatitis, unrelated to viral infection, was seen significantly more often in the CR group. This finding was described previously and may correspond to an unrecognized form of CR, characterized by a parenchymal component. 10 Secondly, ve-occlusive disease was present more often in the liver biopsies and in the failed grafts of patients with CR, as has recently been reported. 14 We have previously reported ve-occlusive disease in association with acute rejection. It is possible that ve-occlusive disease reflects a particularly severe central venulitis caused by either persistent ACR or CR. 15,16 The only patient with veocclusive disease in the control group had bile duct changes con-

9 HEPATOLOGY, Vol. 35, No. 1, 2002 SEBAGH ET AL. 125 sistent with early CR in several biopsies and the failed graft. There was foam cell arteriopathy. Therefore, he did t fulfill the gold-standard criteria for graft failure resulting from CR. Graft failure in this patient was caused by ve-occlusive disease. The presence of early CR changes in the graft supports the theory that ve-occlusive disease is part of the immulogic process in the context of liver disease. In conclusion, the application of the 1999 Banff classification, which includes an early stage of CR, is superior to the old classification for the diagsis of CR. It allows earlier recognition of CR and identifies a larger number of patients. However, limited information can be drawn regarding the outcome of CR based on histology alone. Transient lobular hepatitis, unrelated to viral disease and ve-occlusive disease, may be an unusual expression of CR and should alert pathologists and clinicians to the possibility of an immulogic process. References 1. Demetris AJ, Murase N, Lee RG, Randhawa P, Zeevi A, Pham S, Duquesy R, et al. Chronic rejection. A general overview of histopathology and pathophysiology with emphasis on liver, heart and intestinal allografts. Ann Transplant 1997;2: Banff schema for grading liver allograft rejection: an international consensus document. HEPATOLOGY 1997;25: Demetris A, Adams D, Bellamy C, Blakolmer K, Clouston A, Dhillon AP, Fung J, et al. Update of the International Banff Schema for Liver Allograft Rejection: working recommendations for the histopathologic staging and reporting of chronic rejection. An International Panel. HEPATOLOGY 2000;31: Anymous. Termilogy for hepatic allograft rejection. International Working Party. HEPATOLOGY 1995;22: Ludwig J, Wiesner RH, Batts KP, Perkins JD, Krom RA. The acute vanishing bile duct syndrome (acute irreversible rejection) after orthotopic liver transplantation. HEPATOLOGY 1987;7: Hubscher SG, Buckels JA, Elias E, McMaster P, Neuberger J. Vanishing bile-duct syndrome following liver transplantation is it reversible? Transplantation 1991;51: Demetris AJ, Fung, JJ, Todo S, McCauley J, Jain A, Takaya S, Alessiani M, et al. Conversion of liver allograft recipients from cyclosporine to FK506 immusuppressive therapy a clinicopathologic study of 96 patients. Transplantation 1992;53: Sher LS, Cosenza CA, Petrovic LM, Rojter S, Hoffman A, Lopez RR, Meehan M, et al. Tacrolimus (FK 506) for rescue of chronic rejection following orthotopic liver transplantation. Transplant Proc 1996;28: Blakolmer K, Seaberg EC, Batts, K, Ferrell L, Markin R, Wiesner R, Detre K, et al. Analysis of the reversibility of chronic liver allograft rejection implications for a staging schema [published erratum appears in Am J Surg Pathol 2000 Jan;24(1):164]. Am J Surg Pathol 1999;23: Quaglia AF, Del Vecchio Blanco G, Greaves R, Burroughs AK, Dhillon AP. Development of ductopaenic liver allograft rejection includes a hepatitic phase prior to duct loss [In Process Citation]. J Hepatol 2000;33: Demetris AJ, Seaberg EC, Batts KP, Ferrell L, Lee RG, Markin R, Detre KM. Chronic liver allograft rejection: a National Institute of Diabetes and Digestive and Kidney Diseases interinstitutional study analyzing the reliability of current criteria and proposal of an expanded definition. National Institute of Diabetes and Digestive and Kidney Diseases Liver Transplantation Database. Am J Surg Pathol 1998;22: Freese DK, Sver DC, Sharp HL, Gross CR, Savick SK, Payne WD. Chronic rejection after liver transplantation: a study of clinical, histopathological and immulogical features. HEPATOLOGY 1991;13: Nakazawa Y, Walker NI, Kerlin P, Steadman C, Lynch SV, Strong RW, Clouston AD. Clinicopathological analysis of liver allograft biopsies with late centrilobular necrosis: a comparative study in 54 patients. Transplantation 2000;69: Nakazawa Y, Jonsson JR, Walker NI, Kerlin P, Steadman C, Lynch SV, Strong RW, et al. Fibrous obliterative lesions of veins contribute to progressive fibrosis in chronic liver allograft rejection. HEPATOL- OGY 2000;32: Dhillon AP, Burroughs AK, Hudson M, Shah N, Rolles K, Scheuer PJ. Hepatic venular stesis after orthotopic liver transplantation. HEPATOLOGY 1994;19: Sebagh M, Debette M, Samuel D, Emile JF, Falissard B, Cailliez V, Shouval D, et al. Silent presentation of ve-occlusive disease after liver transplantation as part of the process of cellular rejection with endothelial predilection. HEPATOLOGY 1999;30: