Nonalcoholic fatty liver disease (NAFLD) is a

Transcription

1 Utility and Appropriateness of the Fatty Liver Inhibition of Progression (FLIP) Algorithm and Steatosis, Activity, and Fibrosis (SAF) Score in the Evaluation of Biopsies of Nonalcoholic Fatty Liver Disease Pierre Bedossa and the FLIP Pathology Consortium* Biopsy is still the gold standard for the diagnosis of nonalcoholic steatohepatitis but the definition may vary among pathologists, a drawback especially in evaluation of biopsies for clinical trials. We previously developed a scoring system (steatosis, activity, fibrosis [SAF]) allowing the use of an algorithm (fatty liver inhibition of progression [FLIP]) for the classification of liver injury in morbid obesity. The aim of this study was to determine whether the use of the SAF score and FLIP algorithm can decrease interobserver variations among pathologists. In a first session, pathologists categorized 40 liver biopsies of patients with nonalcoholic fatty liver disease (NAFLD) according to their own experience. In a second reading session, each pathologist reclassified the same slides by using the FLIP algorithm and SAF score, blinded to their first evaluation. The experiment was repeated with two different groups of pathologists at varying levels of training in liver pathology. The percentage of biopsy interpretation concordant with reference evaluation increased from 77% to 97% in Group 1 and from 42% to 75% in Group 2 after the use of the SAF score and FLIP algorithm. The strength of concordance in classification increased in Group 1 from moderate (j ) to substantial (j ) and from fair (j ) to substantial (j ) in Group 2 with application of the algorithm. With regard to the SAF score, concordance was substantial in Group 1 for steatosis (j ), activity (j ), and almost perfect for fibrosis (j after pooling 1a, 1b, and 1c together into a single score F1). Similar trends were observed in Group 2 (j for S, j for A, and j for F). Conclusion: The FLIP algorithm based on the SAF score should decrease interobserver variations among pathologists and are likely to be implemented in pathology practice. (HEPATOLOGY 2014;60: ) Nonalcoholic fatty liver disease (NAFLD) is a generic term encompassing a disease spectrum related to metabolic risk factors, spanning steatosis, nonalcoholic steatohepatitis (NASH) with/without cirrhosis, and hepatocellular carcinoma in the absence of significant alcohol consumption. 1 It is a highly prevalent disease worldwide and a major health concern due to both hepatic and extrahepatic morbidity/mortality. 2-5 Since the seminal descriptions of Adler and Schaffner and Ludwig et al. and the pioneering study of Matteoni et al., 6-9 NASH has been considered a particular entity within the spectrum of NAFLD which is associated with a risk of evolution to fibrosis and cirrhosis. Therefore, characterizing patients with NASH is a major issue in the clinical setting because of its potential risk for liver fibrosis progression and increased liver-related mortality. 10 Indeed, thus far, most clinical trials in NAFLD have targeted patients Abbreviations: FLIP, fatty liver inhibition of progression; NAFLD, nonalcoholic fatty liver disease; NAS, nonalcoholic steatohepatitis activity score; NASH, nonalcoholic steatohepatitis; SAF, steatosis, activity, fibrosis. From the FLIP Pathology Consortium. Received November 17, 2013; accepted April 11,

2 566 BEDOSSA ET AL. HEPATOLOGY, August 2014 with NASH Although a combination of clinical factors and liver function tests abnormalities may help to predict the presence of NASH, the definitive diagnosis continues to be based on histology. 1,2,15-18 It is therefore of utmost importance to use universally adopted criteria for the histological definition of NASH Although elementary histological items that may be present in NASH are well accepted among pathologists, some inconsistencies exist in the definition of the disease and may cause interobserver variations between pathologists. 22,23 Indeed, in a review analysis of the prevalence of NASH in a fairly homogeneous group of patients, i.e., in morbidly obese patients submitted to bariatric surgery, the prevalence of NASH based on histopathological analysis of surgical biopsies obtained at the time of surgery varied from 24 to 98%. 24 As underlined by the authors, this wide variation was partly related to the different histopathological criteria used by the different groups. Recently, a definition endorsed by renowned hepatologists and liver pathologists has progressively gained acceptance in the liver community. 10 NASH is defined as the necessary combination of three histological features: steatosis, ballooning/clarification of hepatocytes, and lobular inflammation. Based on this definition, the FLIP consortium (fatty liver inhibition of progression) has created a simple histological algorithm (FLIP algorithm) based on a scoring system, the SAF score (steatosis, activity, fibrosis) intended for pathologists to reliably diagnose NASH and limit interobserver variation; the algorithm has been validated for assessment of liver injuries in patients with morbid obesity undergoing bariatric surgery. 25 However, most patients seen in hepatology clinics with suspected NAFLD/NASH are not morbidly obese, but rather have one or more features of the metabolic syndrome and, arguably, a liver injury that might differ from that of morbidly obese patients. The first aim of the present study was to assess the applicability of the FLIP algorithm in biopsies of patients with metabolic syndrome and NAFLD who are not morbidly obese. We tested a systematic algorithmic approach and determined whether it can decrease interobserver variations among liver pathologists in the diagnosis of NASH. Because the entire range of histological lesions may be observed in patients with NAFLD, it would be artificial to segregate patients into those with or without NASH based on predetermined criteria that might not always accurately reflect long-term outcomes. Indeed, in a retrospective study some patients with steatosis and associated lesions, but lacking all features of NASH on index biopsy, showed NASH or significant fibrosis a few years later. 26 Although this study has significant limitations, it underlines that some cases may not be easily dichotomized into steatosis or NASH. Another approach would be to consider NAFLD as a continuous histological spectrum, with probably a fluctuation in disease severity, as is the case for other chronic liver diseases. In this perspective, we recently proposed the SAF score that separately assesses steatosis, activity (the sum of hepatocyte ballooning and lobular inflammation), and fibrosis through semiquantitative ordinal scales. 25 The second aim of this study was to assess the reproducibility of the SAF score among pathologists and its applicability to biopsies of patients with liver injuries related to metabolic syndrome. Finally, since liver biopsies are reported by either highly specialized liver pathologists at academic centers or by less specialized pathologists, the *The FLIP Pathology Consortium: Pierre Bedossa, Assistance Publique-H^opitaux de Paris, h^opital Beaujon, University Paris-Diderot, Paris, France. Alastair D. Burt, School of Medicine, University of Adelaide, Adelaide, Australia. Annette S.H. Gouw, Dept. of Pathology & Medical Biology, University Medical Center Groningen, Groningen, The Netherlands. Carolin Lackner, Institute of Pathology, Medical University of Graz, Graz, Austria. Peter Schirmacher, Institut f ur Pathologie, Universit atsklinikum Heidelberg, Heidelberg, Germany. Luigi Terracciano, Institute of Pathology, University Hospital, Basel, Switzerland. D. Tiniakos, Medical School, National and Kapodistrian University of Athens, Greece and Medical School, Newcastle University, Newcastle-upon-Tyne, United Kingdom. J. Brain, Medical School, Newcastle University, Newcastle-upon-Tyne, United Kingdom. Yvonne Bury, Royal Victoria Infirmary, Hospitals Foundation Trust, Newcastleupon-Tyne, United Kingdom. Daniela Cabibi, University of Palermo, Palermo, Italy. Frederic Charlotte, Assistance Publique H^opitaux de Paris, Groupe Hospitalier Pitie-Salpetrière, Universite Pierre et Marie Curie, Paris, France. Ezio David, University of Torino, Ospedale Molinette, Torino, Italy. Luisa Losi, Modena University Hospital, Modena, Italy. Matteo Montani, Institute of Pathology, University of Bern, CH-3010 Bern, Switzerland. Marıa Jesus Pareja, Hospital Universitario de Valme, Sevilla, Spain. Dominique Wendum, Assistance publique-h^opitaux de Paris, H^opital St Antoine Universite Pierre et Marie Curie. Fritz Wrba, Medical University of Vienna, Vienna, Austria. Marianne Ziol, Assistance Publique-H^opitaux de Paris, H^opital jean Verdier, Universite Paris 13, Bobigny, France. Vlad Ratziu, Assistance Publique H^opitaux de Paris, H^opital Pitie Salp^etrière, Universite Pierre et Marie Curie, Paris, France. Address reprint requests to: Pierre Bedossa, M.D., Ph.D., Department of Pathology, INSERM U773, H^opital Beaujon, 100 Boulevard General Leclerc, Clichy, France. pierre.bedossa@bjn.aphp.fr; fax: Copyright VC 2014 by the American Association for the Study of Liver Diseases. View this article online at wileyonlinelibrary.com. DOI /hep Potential conflict of interest: Nothing to report.

3 HEPATOLOGY, Vol. 60, No. 2, 2014 BEDOSSA ET AL. 567 usefulness of both the SAF score and the FLIP algorithm was assessed and compared among pathologists with different degrees of training and experience in hepatopathology. Materials and Methods Liver Biopsy. Forty liver biopsies from nonmorbidly obesepatientswithfeaturesofmetabolicsyndromeas the only clinical risk factor for chronic liver diseases were retrospectively selected for this study. Selection of biopsies was based on: 1) a well-documented clinical history of metabolic syndrome and absence of other causes of chronic liver diseases using clinical, biological, and virological testing; 2) adequacy of the biopsy, especially in terms of its length, the absence of fragmentation, and the quality of staining; 3) the presence of at least 5% steatosis in the biopsy specimen. Liver biopsies were routinely formalin-fixed and paraffin-embedded. For review, hematoxylin and eosin (H&E) and picrosirius hemalun staining were available for each case. Initial interpretation was performed by one liver pathologist (P.B.). This was considered the reference diagnosis. Study Design. The study was carried out initially by a group of expert liver pathologists from different countries. As a first step, each pathologist individually reviewed 40 biopsies and rated the biopsy as steatosis without NASH or steatosis with NASH, based on criteria used in his/her everyday practice (baseline diagnosis). The evaluation of concordance between pathologists according to weighted kappa statistics, the percentage of overall agreement, and the number of biopsies with full agreement between all pathologists were calculated. 27 In a second session of slide reviewing, the same pathologists were given a briefing describing the component of the SAF score and the FLIP algorithm (see below). Based on such tutorial guidelines, they were asked to reassess each of the slides according to the SAF score and FLIP algorithm (algorithmic diagnosis). Slides were recoded between each round so that pathologists were blinded to their initial evaluation. As previously done, the percentage of overall agreement, kappa score concordance, and number of cases with full agreement between all pathologists were calculated for algorithmic diagnosis. During this reading step, pathologists were asked to provide an SAF score (see below) for each biopsy. The impact of FLIP algorithm use was also evaluated by comparing reference diagnosis to baseline and to algorithmic diagnosis (the one provided by the majority of pathologists for each case). This two-step process was repeated by two different groups of pathologists, first among six expert liver pathologists from different countries (A.D.B., D.T., P.S., L.T., K.L., A.S.H.G.) and then in another group comprising 10 pathologists who also have received training in liver pathology, but who are also involved in a more general pathology practice (E.D., Y.B., M.Z., F.C., D.W., M.J.P., D.C., M.M., F.W., L.L.). The only difference between the two experiments was that the first experiment was performed with glass slides viewed under a light microscope independently by each pathologist, while the second used digitalized images of the same biopsies read remotely through the Internet and the Aperio system. SAF Score and Algorithmic Classification. For each biopsy an SAF score summarizing the main histological lesions was defined. This assesses both and separately the grade of steatosis (S), the grade of activity (A), and the stage of fibrosis (F), the latter according to the NASH Clinical Research Network (CRN). 19 Steatosis was assessed by the percentage of hepatocytes containing large and medium-sized intracytoplasmic lipid droplets (but not foamy microvesicles), on a scale of 0 to 3 (S0: <5%; S1: 5%-33%, S2: 34%-66%, S3: >67%). Ballooning of hepatocytes was graded from 0 to 2 (0: normal hepatocytes with cuboidal shape, sharp angles and pink eosinophilic cytoplasm; 1: presence of clusters of hepatocytes with a rounded shape and pale cytoplasm, usually reticulated where, although the shape is different, the size is similar to that of normal hepatocytes; 2, as for grade 1, but where there was also at least one enlarged ballooned hepatocyte (at least 2-fold size compared with that of normal cells within a cluster of hepatocytes with grade 1 ballooning). Lobular inflammation was defined as a focus of two or more inflammatory cells within the lobule organized either as microgranulomas or located within the sinusoids. Foci were counted at 203 magnification (grade 0: none; 1: < 2 foci per lobule; 2: >2 foci per lobule). The grade of activity (A from A0 to A4) was calculated by addition of grades of ballooning and lobular inflammation. Aspects of grading are illustrated in Fig. 2. Then the FLIP algorithm that enables dichotomous classification of a biopsy in either the steatosiswithout-nash or the steatosis-with-nash groups was used. It is based on the semiquantification of three elementary features, steatosis, clarification/ballooning of hepatocytes, and lobular inflammation as evaluated according to the SAF score (Fig. 1). For classification using the FLIP algorithm, steatosis was used as the criterion for entry into the algorithm weighted by hepatocellular ballooning and lobular inflammation. For

4 568 BEDOSSA ET AL. HEPATOLOGY, August 2014 reading, 22 biopsies were classified as steatosis without NASH and 18 as steatosis with NASH. Fig. 1. Diagnostic algorithm for NASH. any biopsy with at least grade 1 steatosis, the algorithm included nine possible means of arriving at a decision (Fig. 1). A case presenting with at least grade 1 of each of the three features (steatosis, ballooning, lobular inflammation) was classified as NASH. Other cases were diagnosed as steatosis (without NASH). Other histologic features possibly present in NASH were not considered for categorization. In addition, overall histological severity of disease was defined according to activity and fibrosis independently of a diagnosis of NASH. Biopsies were split at data management into two subgroups, mild disease (A<2 and F<2) and significant disease (A>2 and/or F>2). Statistical Analysis. Quantitative variables were expressed as means (6SD) or medians, interquartile range (IQR), as appropriate, and categorical variables as frequency and percentage. Interobserver agreement in qualitative variables was evaluated using the kappa index (j), which excludes chance-expected agreement. Agreement of ordered categorical variables was evaluated by weighted k with linear weighting, which takes into account the proximity of categories. Strength of agreement was considered either poor (j <0.19), fair (j: ), moderate (j: ), substantial (j: ), or almost perfect (j > 0.80). 28 The proportion of overall agreement for diagnosis of NASH was expressed as the mean of the agreement rate calculated for all pairs of observers. Results Patient Characteristics The main clinical, biological, and histological characteristics of the 40 patients whose biopsies were selected for this study are shown in Table 1. The mean biopsy length was mm. According to reference Appropriateness of the FLIP Algorithm and SAF Scores to Biopsies of Patients With Metabolic Syndrome Since the SAF scores and the FLIP algorithm were initially set up in a group of biopsies of morbidly obese patients, we sought to determine their applicability to biopsies of patients with metabolic syndrome in the absence of morbid obesity. 25 No limitation, such as the lack of some relevant specific features or a major difficulty to assess histopathologic features, was reported by any of the observers in the use of the FLIP algorithm or SAF score. Furthermore, the use of the FLIP algorithm increased the agreement with the reference diagnosis in both panels of pathologists. Indeed, agreement with reference diagnosis increased from 31 to 39 out of 40 biopsies after using the FLIP algorithm (respectively 77% to 97%) in Group 1 and from 17 to 30 out of 40 biopsies (42% to 75%) in Group 2 (Table 2). These results support the appropriateness of the FLIP algorithm in the diagnostic classification of biopsies of patients with metabolic syndrome. Concordance Between Expert Liver Pathologists Baseline and Algorithmic Classification. The first reading session (baseline evaluation) was based on every participating pathologists individual reading criteria in her/his everyday clinical practice. Table 3 shows the most frequent diagnosis among the six pathologists. Thirty-one out of 40 (77%) baseline diagnosis (the most frequently proposed diagnosis by the group) were concordant with the reference classification. The strength of agreement in diagnosis among pathologists was moderate (j ), with 77% overall agreement. Full agreement between all pathologists was found for 26 out of 40 biopsies (65%), including 15 biopsies with a diagnosis of steatosis, no NASH and 11 with a diagnosis of NASH (Table 2). In the second reading session the use of the FLIP algorithm allowed the group to consensually reclassify eight biopsies (20%), including five with a split diagnosis at the first reading (Table 3). Thirty-nine out of 40 (97%) algorithmic diagnoses became concordant with the reference diagnosis. The strength of agreement between pathologists increased with the use of the FLIP algorithm. Concordance became substantial (j ), with the percent of overall agreement increasing to 83%. The number of biopsies with agreement between all pathologists increased to 34 out of 40 (85%). In the six biopsies without full agreement

5 HEPATOLOGY, Vol. 60, No. 2, 2014 BEDOSSA ET AL. 569 Fig. 2. Histological grading used in SAF score. Grading of steatosis. (A) Grade 1, lipid droplets in 30% of hepatocytes (H&E, 32.5). (B) Grade 2, steatosis in 60% of hepatocytes, (H&E, 32.5). (C) Grade 3: steatosis in >80% of hepatocytes (H&E, 32.5). Hepatocyte ballooning and clarification. (D) Ballooning grade 1: rounded hepatocyte with pale cytoplasm and size quite similar to that of normal hepatocytes (H&E, 320). (E) Ballooning grade 2: presence of enlarged hepatocyte at least 2-fold that of normal cells in a background of clear and rounded hepatocytes, arrow (H&E, 320). Lobular inflammation. (F) Lobular inflammation grade 1: two small foci of inflammatory cells, arrow (H&E 320). (G) Lobular inflammation grade 2: several large inflammatory foci (arrow) (H&E, 320). between all pathologists, only one of the observers did not agree with the diagnosis of the other five and none of the observers consistently provided a different diagnosis, which suggests that there was no observer bias. SAF Score. For the different components of the SAF score, concordance was substantial for steatosis (j ) and activity (j ). Concordance for clarification/ballooning and lobular inflammation, the two components of activity were, respectively, 0.8 and Concordance was moderate for fibrosis (j ). Major disagreement in fibrosis staging was observed among stage 1 subgroups (1a vs. 1b vs. 1c). Pooling of these substages by a data manager into a single F1 score greatly increased the concordance of fibrosis (j ). When biopsies were split into two subgroups, i.e., mild histological disease (A<2 and F<2) and significant disease (A>2 and/or F>2) based on individual SAF score evaluation, concordance was almost perfect (j ). Interestingly, two biopsies (5%) which were classified as steatosis without NASH fell into the group of disease with significant histological severity (A>2 or/and F >2) becauseofascoreoffibrosis>2.

6 570 BEDOSSA ET AL. HEPATOLOGY, August 2014 Table 1. Study Group Baseline Characteristics (n 5 40) Mean, % Median [IQR] Age (years) ( ) Gender F/M (% female) 16/24 (40%) Body mass index (kg/m 2 ) ( ) Diabetes mellitus, n (%) 29 (72.5%) Hypertension, n (%) 27 (67.5%) Insulin (lmol/l) ( ) HOMA-IR (%) ( ) Total cholesterol (mmol/l) ( ) Triglycerides (mmol/l) ( ) ALT (IU/L) ( ) Length of liver biopsy (mm) (20-30) Number of fragments (1-3) NASH, (%)* 18 (45%) *According to the reference reading. Concordance Between the Second Group of Pathologists Algorithmic Classification. Table 4 shows the most frequent diagnosis among the 10 pathologists at baseline evaluation. Seventeen out of 40 cases (42%) were concordant with the reference diagnosis. Concordance in classification of NAFLD was fair among the 10 pathologists (j ). The percentage of overall agreement was 55% (Table 2). The number of biopsies with full agreement between the 10 pathologists was 18 out of 40 (45%), including 13 biopsies with a diagnosis of steatosis, no NASH and five with a diagnosis of NASH. In the second reading session, the use of the FLIP algorithm led the group to consensually reclassify 16 biopsies (40%), 15 of which were equally split at first reading (between steatosis and NASH each by five pathologists) or categorized as the opposite of the final algorithmic diagnosis (Table 4). Thirty out of 40 cases (75%) became concordant with the reference diagnosis. The FLIP algorithm increased both the strength of agreement between pathologists, which became substantial (j ), and the percentage of overall agreement (81%). The number of biopsies showing full agreement between pathologists reached 28 out of 40 (70%). The results are summarized in Table 4. Among biopsies causing disagreement, one observer disagreed with the consensual diagnosis in 6 out of the 12 biopsies, suggesting possible observer bias. SAF Score. With regard to the different components of the SAF score, concordance was moderate for steatosis (j ), and substantial for activity (j ), and for fibrosis (j ) after pooling of F1 substages. When biopsies were classified into histologically mild disease and significant disease (A>2 and/or F>2) based on individual SAF scores, concordance was almost perfect (j ). Four biopsies with steatosis but no NASH fell into the group of significant severity because of fibrosis >2. Discussion Due to the lack of a definitive, conclusive noninvasive test, liver biopsy continues to play a critical role in the management of patients with NAFLD. However, liver biopsy is prone to limitations such as sampling error and observer variability. 29,30 The degree of interobserver variability is mainly determined by the accuracy of the definition of a pathological feature and its different semiquantitative grades and the applicability to common practice. While a single lesion (steatosis, fibrosis, inflammation) might be highly reproducible, lower reproducibility is expected for a composite diagnosis based on a combination of features such as NASH, the definition of which is based on the association of several histopathological features. 10,19,31 Yet a reliable histological identification of NASH is mandatory, as this diagnosis carries prognostic implications for fibrosis development and liver- and nonliver-related mortality. 3,4 Although NAFLD is one of the most common forms of chronic liver disease, the present study shows that concordance between pathologists for the diagnosis of NASH is limited when each pathologist uses his/ her individual criteria based on everyday practice, a discordance that has been previously pointed out. 24 In order to increase the reliability of NASH diagnosis, we previously developed a simple algorithm that takes into consideration the actual definition of the disease based on a semiquantitative evaluation of steatosis, hepatocyte ballooning, and lobular inflammation. This FLIP algorithm was previously validated in biopsies of patients with liver injury related to morbid obesity. 25 Table 2. Agreement for Diagnosis Before and After Use of Algorithm in the Two Groups of Pathologists Group 1 Group 2 Baseline classification Algorithmic classification Baseline classification Algorithmic classification j Score Biopsy With Full Agreement Between Pathologists (%) Biopsy With Agreement With Reference Diagnosis (%) /40 (65%) 31/40 (77%) /40 (85%) 39/40 (97%) /40 (45%) 17/40 (42%) /40 (70%) 30/40 (75%)

7 HEPATOLOGY, Vol. 60, No. 2, 2014 BEDOSSA ET AL. 571 Table 3. Individual Evaluation of Liver Biopsies by the First Group of Expert Liver Pathologists Case Reference Classification*,k Baseline Classification y,z Algorithmic Classification y, SAF y Disease Severity k 1. NASH NASH NASH S2A4F1 Significant 2. NASH NASH NASH S1 A2-3F2 Significant 3. Steatosis Steatosis Steatosis S1A1F1 Mild 4. Steatosis NASH NASH S2A3F4 Significant 5. Steatosis Steatosis/NASH Steatosis S1A1F1 Mild 6. NASH NASH NASH S1A4F3 Significant 7. Steatosis Steatosis Steatosis S2A1F1 Mild 8. NASH NASH NASH S1A4F3 Significant 9. NASH NASH NASH S2-3A3F1-2 Significant 10. Steatosis Steatosis Steatosis S1A1F1 Mild 11. NASH NASH NASH S3A4F3 Significant 12. Steatosis Steatosis Steatosis S1A1F3 Significant 13. Steatosis Steatosis Steatosis S1-2A1F1 Mild 14. NASH NASH NASH S3A4F3 Significant 15. Steatosis Steatosis Steatosis S3A1F0 Mild 16. NASH NASH NASH S1A3F3 Significant 17. Steatosis Steatosis/NASH Steatosis S1A1F1 Mild 18. Steatosis Steatosis Steatosis S3A1F2 Significant 19. Steatosis Steatosis Steatosis S1A0F0 Mild 20. NASH NASH NASH S1A3F1 Significant 21. Steatosis Steatosis Steatosis S2A1F1 Mild 22. Steatosis Steatosis Steatosis S2A1F0 Mild 23. NASH Steatosis/NASH NASH S1-2A2F1 Significant 24. NASH Steatosis NASH S1A3F1 Significant 25. Steatosis Steatosis Steatosis S3A1F0 Mild 26. Steatosis Steatosis Steatosis S1A0F1 Mild 27. NASH NASH NASH S2-3A4F2 Significant 28. Steatosis Steatosis/NASH Steatosis S2A1F1 Mild 29. NASH NASH NASH S3A4F3 Significant 30. Steatosis Steatosis Steatosis S2A1F1 Mild 31. Steatosis Steatosis Steatosis S2-3A1F0 Mild 32. NASH Steatosis/NASH NASH S1A2F3 Significant 33. Steatosis Steatosis Steatosis S1A1F0 Mild 34. Steatosis Steatosis Steatosis S2A1F0 Mild 35. NASH NASH NASH S3A3F2 Significant 36. Steatosis Steatosis Steatosis S1A1F1 Mild 37. NASH Steatosis NASH S1A2F2 Significant 38. NASH NASH NASH S1A3F3 Significant 39. NASH Steatosis NASH S2A2F1 Significant 40. Steatosis Steatosis Steatosis S2A1F1 Mild In bold: biopsies where algorithm allows final classification different from baseline evaluation. *Reference interpretation is the initial evaluation done by one pathologist using accepted criteria, algorithm and SAF score. Based on the most frequent diagnosis among the 6 pathologists. When the diagnosis is equally split between pathologists, both propositions are reported. Classification (Steatosis vs. NASH) based on everyone own criteria used in its practice. Classification (Steatosis vs. NASH) after using the algorithm. k Based on SAF score (mild for A < 2 and F < 2 or significant for A > 2 and/or F > 2). In the present study, we observed that this algorithm was also applicable to patients with NAFLD related to metabolic syndrome in the absence of morbid obesity. The higher concordance with the reference diagnosis after the use of the SAF score and FLIP algorithm supports the relevance of this approach in these patients. We also demonstrated that when using the FLIP algorithm the concordance between observers increased from fair or moderate to substantial after use of the FLIP algorithm. This suggests that this simple tool may help to standardize patient histological stratification at different centers, a process that is essential, especially in multicenter clinical trials. While a dichotomous classification of NAFLD into NASH and no-nash conveniently simplifies the diagnosis and prediction of long-term outcomes, NAFLD, in fact, covers a continuous histological spectrum akin to chronic viral hepatitis. In that regard a binary classification might be too simplistic, if not artificial. In an attempt to take into account this complexity we previously developed the SAF score, a semiquantitative score that separately grades the main histopathological

8 572 BEDOSSA ET AL. HEPATOLOGY, August 2014 Table 4. Individual Evaluation of Liver Biopsies by Group 2 of General Pathologists Case Reference Classification *,k Baseline Classification y,z Algorithmic Classification y, SAF y Severity y,k 1. NASH NASH NASH S2A4F1 Significant 2. NASH NASH NASH S1A2F2-3 Significant 3. Steatosis NASH Steatosis S1A1F1 Mild 4. Steatosis NASH NASH S2A2F4 Significant 5. Steatosis NASH Steatosis S1A1F0 Mild 6. NASH NASH NASH S2A4F3 Significant 7. Steatosis NASH Steatosis S2A1F1 Mild 8. NASH NASH NASH S1A4F3 Significant 9. NASH NASH NASH S3A2-3F2 Significant 10. Steatosis Steatosis Steatosis S1A1F1 Mild 11. NASH NASH NASH S2A4F4 Significant 12. Steatosis NASH NASH S1-2A2F3 Significant 13. Steatosis NASH Steatosis S2A1F1 Mild 14. NASH NASH NASH S3A4F3 Significant 15. Steatosis NASH Steatosis S3A1F0 Mild 16. NASH NASH/Steatosis NASH S1A2F3 Significant 17. Steatosis NASH Steatosis S1A1F2 Significant 18. Steatosis Steatosis Steatosis S3A1F2 Significant 19. Steatosis NASH Steatosis S1A1F1 Mild 20. NASH Steatosis NASH S1A2F1 Significant 21. Steatosis Steatosis Steatosis S2A1F1 Mild 22. Steatosis NASH NASH S2A3F1 Significant 23. NASH NASH NASH S1-2A2F1 Significant 24. NASH NASH NASH S1A3F1 Significant 25. Steatosis NASH Steatosis S3A1F1 Mild 26. Steatosis Steatosis Steatosis S1A0F1 Mild 27. NASH NASH/Steatosis NASH S2A2-3F2 Significant 28. Steatosis NASH NASH S2A2F2 Significant 29. NASH Steatosis Steatosis S3A1F4 Significant 30. Steatosis NASH Steatosis S2A1F1 Mild 31. Steatosis NASH Steatosis S3A1F0 Mild 32. NASH NASH NASH S1A3F3 Significant 33. Steatosis NASH NASH S1A2F0 Significant 34. Steatosis NASH NASH S2A2-3F0 Significant 35. NASH NASH NASH S3A3F2 Significant 36. Steatosis NASH NASH S1A2F1 Significant 37. NASH NASH Steatosis S1A1F2 Significant 38. NASH NASH/Steatosis NASH S1A3F3 Significant 39. NASH NASH NASH S2A2F1 Significant 40. Steatosis NASH/Steatosis NASH S2A2F1 Significant In bold: biopsies where algorithm allows final classification different from baseline evaluation. *Reference reading is the initial reading done by one pathologist using criteria, algorithm and SAF score. Based on the most frequent diagnosis among the 6 pathologists. When the diagnosis is equally split between pathologists, both propositions are reported. Classification (Steatosis vs. NASH) based on everyone own criteria used in its practice. Classification (Steatosis vs. NASH) after using the algorithm. k Based on SAF score (mild for A < 2 and F < 2 or significant for A > 2 and/or F > 2). patterns of NAFLD: steatosis, activity, and fibrosis. Like the METAVIR classification for chronic viral hepatitis, the SAF score describes in a snapshot the main histological features. 32 In contrast to the widely used NAFLD activity score (NAS), it dissociates steatosis from necroinflammation, two lesions with distinct pathogenesis and prognostic information. Indeed, while necroinflammation may favor fibrosis, the prognosis of steatosis remains controversial. 34,35 Nevertheless, biopsy interpretation should not be limited to scoring, since patterns of injury are not all included into any scoring systems, while these features can provide important information. The SAF score usefully complements the histopathological evaluation. For instance, it may enable the identification of several specific patterns, such as steatofibrosis without necroinflammation, which is impossible to categorize using a dichotomist (NASH/non- NASH) approach. In the context of metabolic syndrome, some advanced liver diseases may display no or rare fat vacuoles but still retain the hepatocyte ballooning and lobular

9 HEPATOLOGY, Vol. 60, No. 2, 2014 BEDOSSA ET AL. 573 inflammation characteristic of NASH. Classification of these cases raises not only a semantic question (fatty liver disease with no fat) but also points out one of the limitations of the actual definition, which requires 5% of steatosis as a criteria for NAFLD. 10 Interestingly, the use of the SAF score should allow describing this subgroup satisfactorily (as for example S0A2F4 for cirrhosis with clarification and ballooning grade 1, lobular inflammation grade 1, and no steatosis). Our study also shows a satisfactory level of reproducibility of the three items of the SAF score, a result mainly related to the increased accuracy of the description of various lesions, including ballooning provided here. Here we added some references regarding the shape and size in reference to normal hepatocyte that may have helped to recognize this feature consistently. As previously shown, staging of fibrosis is highly reproducible. 19,33 This was also the case in our study providing the three substages of 1 (1a, 1b, 1c) pooled together. However, it needs to be underlined that stage 1c refers to a different location of fibrosis (portal) than stage 1a and 1b (acinar zone 3), an important difference that needs to be explicitly mentioned in the liver biopsy report. Interestingly, previous studies showed a substantial reproducibility in assessing these different substages. Noteworthy, they were performed with Masson s trichrome staining, while the present study used picrosirius red. Whether this might explain the discrepancy with the present study is not known, although both are reference stains for fibrosis. Improvement in NAS and resolution of NASH are widely used as eligibility criteria and primary endpoints for clinical trials in NASH (Clintrials.Gov). However, several studies have raised concern about the lack of correlation between NAS and liver-related mortality. 18,22 Rather than subclassifying NAFLD into patients with or without NASH or stratifying patients according to NAS, we suggest separating the disease histologically into mild (A < 2and/orF< 2) and significant (A > 2and/orF> 2) histological lesions based on the SAF score evaluation. Although activity and fibrosis must be assessed separately, these two features have prognostic significance and therefore should be considered together when defining patients at risk of unfavorable evolution for inclusion in clinical trials. This is particularly relevant, since a recent study demonstrated that advanced fibrosis and the presence of NASH are often dissociated, especially in elderly patients. 34 Interestingly, our study shows that this classification attained very good to almost perfect reliability among pathologists whatever their involvement in liver pathology. Furthermore, it allowed including patients with steatosis, significant fibrosis, but no or mild activity into the group of patients with significant histological disease, an adjudication that would not have been possible using the NAS or the NASH dichotomist approach. In the present study we duplicated the experiments by way of two groups of pathologists in order to test the external validity of the classifications. Both groups comprised pathologists with experience in liver pathology, although the second group was more heterogeneous in terms of hepatopathology as a subspecialty, along with academic involvement. The results show similar trends of improvement of diagnostic accuracy and agreement between observers, thus confirming the exportability of the classification beyond the initial group of pathologists that validated them. Thus, both the FLIP algorithm and SAF score proved highly reliable tools for the standardization of biopsy reports in patients with NAFLD whatever the specialization and academic level of the reading pathologist. This study has some limitations. First, the biopsies were retrospectively selected partly based on significant length and absence of fragmentation. Indeed, concordance between observers may also be influenced by technical factors and previous studies have shown the significant influence of biopsy length on diagnosis reliability Second, while the first experiment was performed by reading slides under a light microscope, the second was done using digitalized images and remote interpretation of the same set of slides through the Internet. It is not known whether this influences concordance, although the high quality of scanned images and similar trends observed between interpretations in the two groups suggest that it does not. Of note, a, previous study suggested that virtual microscopy has a same diagnostic accuracy as conventional microscopy. 39 Third, the two groups were composed of 6 and 10 pathologists, respectively, rendering comparison of the kappa score difficult between the two groups. Nevertheless, the similar trend toward improved agreement after use of the SAF score and FLIP algorithm by the two groups strongly suggests that the improvement brought by these two histological tools is real. Histological scoring systems have been very successful in evaluation of chronic liver diseases. Although they are reliable and useful for communication between hepatologists and liver pathologists, they are shortcuts that do not retain all information conveyed by a liver biopsy. Indeed, patterns of histological injury, which are not part of the proposed scoring systems but which are essential for biopsy interpretation, have major importance since they may be useful to characterize borderline cases and subgroups of NAFLDs, provide insight into

10 574 BEDOSSA ET AL. HEPATOLOGY, August 2014 pathophysiology, and may help the differential diagnosis in case of comorbidities. Therefore, the description of patterns of injuries must remain the core of the liver biopsy report and the SAF score and FLIP algorithm considered only as additional tools. In conclusion, the use of the histological algorithm and the SAF score provides a reliable and reproducible diagnosis and grading/staging of NAFLD with reduced interobserver variability. It can be applied to patients with NAFLD and features of the metabolic syndrome and therefore to the wider population of patients at risk of NAFLD seen in liver clinics in addition to those with morbid obesity undergoing bariatric surgery. It can be successfully used by a wide variety of liver pathologists as well as among more general pathologists, which should facilitate its implementation outside of leading tertiary care centers with a specific interest in NAFLD research. Nonetheless, these simple tools should complement and not replace the textual description of a liver biopsy report. Acknowledgment: The research leading to these results has received funding from the European Community s Seventh Framework Program (FP7/ ) under grant agreement n HEALTH-F for the project FLIP. References 1. Chalasani N, Younossi Z, Lavine JE, Diehl AM, Brunt EM, Cusi K, et al. The diagnosis and management of non-alcoholic fatty liver disease: practice guideline by the American Gastroenterological Association, American Association for the Study of Liver Diseases, and American College of Gastroenterology. Gastroenterology 2012;142: Musso G, Gambino R, Cassader M, Pagano G. Meta-analysis: natural history of non-alcoholic fatty liver disease (NAFLD) and diagnostic accuracy of non-invasive tests for liver disease severity. Ann Med 2011;43: Ekstedt M, Franzen LE, Mathiesen UL, Thorelius L, Holmqvist M, Bodemar G, et al. Long-term follow-up of patients with NAFLD and elevated liver enzymes. HEPATOLOGY 2006;44: Soderberg C, Stal P, Askling J, Glaumann H, Lindberg G, Marmur J, et al. Decreased survival of subjects with elevated liver function tests during a 28-year follow-up. HEPATOLOGY 2010;51: Targher G, Day CP, Bonora E. Risk of cardiovascular disease in patients with nonalcoholic fatty liver disease. N Engl J Med 2010;363: Matteoni CA, Younossi ZM, Gramlich T, Boparai N, Liu YC, McCullough AJ. Nonalcoholic fatty liver disease: a spectrum of clinical and pathological severity. Gastroenterology 1999;116: Adler M, Schaffner F. Fatty liver hepatitis and cirrhosis in obese patients. Am J Med 1979;67: Ludwig J, Viggiano TR, McGill DB, Oh BJ. Nonalcoholic steatohepatitis: Mayo Clinic experiences with a hitherto unnamed disease. Mayo Clin Proc 1980;55: Vernon G, Baranova A, Younossi ZM. Systematic review: the epidemiology and natural history of non-alcoholic fatty liver disease and nonalcoholic steatohepatitis in adults. Aliment Pharmacol Ther 2011;34: Sanyal AJ, Brunt EM, Kleiner DE, Kowdley KV, Chalasani N, Lavine JE, et al. Endpoints and clinical trial design for nonalcoholic steatohepatitis. HEPATOLOGY 2011;54: Sanyal AJ, Chalasani N, Kowdley KV, McCullough A, Diehl AM, Bass NM, et al. Pioglitazone, vitamin E, or placebo for nonalcoholic steatohepatitis. N Engl J Med 2010;362: Torres DM, Jones FJ, Shaw JC, Williams CD, Ward JA, Harrison SA. Rosiglitazone versus rosiglitazone and metformin versus rosiglitazone and losartan in the treatment of nonalcoholic steatohepatitis in humans: a 12-month randomized, prospective, open-label trial. HEPATO- LOGY 2011;54: Belfort R, Harrison SA, Brown K, Darland C, Finch J, Hardies J, et al. A placebo-controlled trial of pioglitazone in subjects with nonalcoholic steatohepatitis. N Engl J Med 2006;355: Aithal GP, Thomas JA, Kaye PV, Lawson A, Ryder SD, Spendlove I, et al. Randomized, placebo-controlled trial of pioglitazone in nondiabetic subjects with nonalcoholic steatohepatitis. Gastroenterology 2008; 135: Wieckowska A, Zein NN, Yerian LM, Lopez AR, McCullough AJ, Feldstein AE. In vivo assessment of liver cell apoptosis as a novel biomarker of disease severity in nonalcoholic fatty liver disease. HEPATO- LOGY 2006;44: Angulo P, Hui JM, Marchesini G, Bugianesi E, George J, Farrell GC, et al. The NAFLD fibrosis score: a noninvasive system that identifies liver fibrosis in patients with NAFLD. HEPATOLOGY 2007;45: Gawrieh S, Chalasani N. NAFLD fibrosis score: is it ready for wider use in clinical practice and for clinical trials? Gastroenterology Angulo P, Bugianesi E, Bjornsson ES, Charatcharoenwitthaya P, Mills PR, Barrera F, et al. Simple noninvasive systems predict long-term outcomes of patients with nonalcoholic fatty liver disease. Gastroenterology 2013;145: Kleiner DE, Brunt EM, Van Natta M, Behling C, Contos MJ, Cummings OW, et al. Design and validation of a histological scoring system for nonalcoholic fatty liver disease. HEPATOLOGY 2005;41: Tiniakos DG, Vos MB, Brunt EM. Nonalcoholic fatty liver disease: pathology and pathogenesis. Annu Rev Pathol 2010;5: Brunt EM. Pathology of nonalcoholic fatty liver disease. Nat Rev Gastroenterol Hepatol 2010;7: Younossi ZM, Stepanova M, Rafiq N, Makhlouf H, Younoszai Z, Agrawal R, et al. Pathologic criteria for nonalcoholic steatohepatitis: interprotocol agreement and ability to predict liver-related mortality. HEPATOLOGY 2011;53: Younossi ZM, Gramlich T, Liu YC, Matteoni C, Petrelli M, Goldblum J, et al. Nonalcoholic fatty liver disease: assessment of variability in pathologic interpretations. Mod Pathol 1998;11: Machado M, Marques-Vidal P, Cortez-Pinto H. Hepatic histology in obese patients undergoing bariatric surgery. J Hepatol 2006;45: Bedossa P, Poitou C, Veyrie N, Bouillot JL, Basdevant A, Paradis V, et al. Histopathological algorithm and scoring system for evaluation of liver lesions in morbidly obese patients. HEPATOLOGY 2012;56: Pais R, Charlotte F, Fedchuk L, Bedossa P, Lebray P, Poynard T, et al. A systematic review of follow-up biopsies reveals disease progression in patients with non-alcoholic fatty liver. J Hepatol 2013;59: Cohen J. Weighted kappa: nominal scale agreement with provision for scaled disagreement or partial credit. Psychol Bull 1968;70: Fleiss JL, Levin B, Paik MC. Statistical methods for rates and proportions. Hoboken, NJ: John Wiley & Sons; Merriman RB, Ferrell LD, Patti MG, Weston SR, Pabst MS, Aouizerat BE, et al. Correlation of paired liver biopsies in morbidly obese patients with suspected nonalcoholic fatty liver disease. HEPATOLOGY 2006;44: Ratziu V, Charlotte F, Heurtier A, Gombert S, Giral P, Bruckert E, et al. Sampling variability of liver biopsy in nonalcoholic fatty liver disease. Gastroenterology 2005;128:

11 HEPATOLOGY, Vol. 60, No. 2, 2014 BEDOSSA ET AL Brunt EM, Janney CG, Di Bisceglie AM, Neuschwander-Tetri BA, Bacon BR. Nonalcoholic steatohepatitis: a proposal for grading and staging the histological lesions. Am J Gastroenterol 1999;94: Bedossa P, Poynard T. An algorithm for the grading of activity in chronic hepatitis C. The METAVIR Cooperative Study Group. HEPA- TOLOGY 1996;24: Kleiner DE, Brunt EM. Nonalcoholic fatty liver disease: pathologic patterns and biopsy evaluation in clinical research. Semin Liver Dis 2012;32: Noureddin M, Yates KP, Vaughn IA, Neuschwander-Tetri BA, Sanyal AJ, McCullough A, et al. Clinical and histological determinants of nonalcoholic steatohepatitis and advanced fibrosis in elderly patients. HEPATOLOGY 2013;58: Rousselet MC, Michalak S, Dupre F, Croue A, Bedossa P, Saint-Andre JP, et al. Sources of variability in histological scoring of chronic viral hepatitis. HEPATOLOGY 2005;41: Vuppalanchi R, Unalp A, Van Natta ML, Cummings OW, Sandrasegaran KE, Hameed T, et al. Effects of liver biopsy sample length and number of readings on sampling variability in nonalcoholic Fatty liver disease. Clin Gastroenterol Hepatol 2009;7: Bedossa P, Dargere D, Paradis V. Sampling variability of liver fibrosis in chronic hepatitis C. HEPATOLOGY 2003;38: Cholongitas E, Senzolo M, Standish R, Marelli L, Quaglia A, Patch D, et al. A systematic review of the quality of liver biopsy specimens. Am J Clin Pathol 2006;125: Kayser K. Introduction of virtual microscopy in routine surgical pathology a hypothesis and personal view from Europe. Diagn Pathol 2012;7:48.