Histopathological algorithm and scoring system for evaluation of liver lesions in morbidly obese patients


  • Pierre Bedossa,

    Corresponding author
    1. Assistance Publique-Hôpitaux de Paris, Beaujon Hospital, Pathology Department, Clichy, France; Centre de Recherche Bichat-Beaujon, INSERM U773, University Paris-Diderot, Paris, France
    • Department of Pathology, Hôpital Beaujon, 100 Bd general Leclerc, 92100—Clichy, France
    Search for more papers by this author
    • fax: +33140870077

  • Christine Poitou,

    1. Institute of Cardiometabolism and Nutrition, (ICAN), Assistance Publique-Hôpitaux de Paris Pitié-Salpêtrière Hospital, Paris, France
    2. INSERM, U872, Nutriomique, Paris, France; Université Pierre et Marie Curie-Paris6, Centre de Recherche des Cordeliers, UMR S 872, Paris, France
    Search for more papers by this author
  • Nicolas Veyrie,

    1. Assistance Publique-Hôpitaux de Paris, Ambroise Paré Hospital, Surgery Department, Boulogne-Billancourt, France
    Search for more papers by this author
  • Jean-Luc Bouillot,

    1. Assistance Publique-Hôpitaux de Paris, Ambroise Paré Hospital, Surgery Department, Boulogne-Billancourt, France
    Search for more papers by this author
  • Arnaud Basdevant,

    1. Institute of Cardiometabolism and Nutrition, (ICAN), Assistance Publique-Hôpitaux de Paris Pitié-Salpêtrière Hospital, Paris, France
    2. INSERM, U872, Nutriomique, Paris, France; Université Pierre et Marie Curie-Paris6, Centre de Recherche des Cordeliers, UMR S 872, Paris, France
    Search for more papers by this author
  • Valerie Paradis,

    1. Assistance Publique-Hôpitaux de Paris, Beaujon Hospital, Pathology Department, Clichy, France; Centre de Recherche Bichat-Beaujon, INSERM U773, University Paris-Diderot, Paris, France
    Search for more papers by this author
  • Joan Tordjman,

    1. Institute of Cardiometabolism and Nutrition, (ICAN), Assistance Publique-Hôpitaux de Paris Pitié-Salpêtrière Hospital, Paris, France
    2. INSERM, U872, Nutriomique, Paris, France; Université Pierre et Marie Curie-Paris6, Centre de Recherche des Cordeliers, UMR S 872, Paris, France
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Karine Clement

    1. Institute of Cardiometabolism and Nutrition, (ICAN), Assistance Publique-Hôpitaux de Paris Pitié-Salpêtrière Hospital, Paris, France
    2. INSERM, U872, Nutriomique, Paris, France; Université Pierre et Marie Curie-Paris6, Centre de Recherche des Cordeliers, UMR S 872, Paris, France
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Potential conflict of interest: Nothing to report.

  • We thank the Assistance Publique - Hôpitaux de Paris (APHP) and the Direction of Clinical Research, which promoted and supported clinical investigations, the Programme Hospitalier de Recherche Clinique (AOR 02076 to K.C.) and the Contrat de Recherche Clinique (CRC P050318 to C.P.). This work was also supported by the Commission of European Communities (Collaborative Project “Hepatic and adiposetissue and functions in the metabolic syndrome” HEPADIP, seehttp://www.hepadip.org/, contract LSHM-CT-2005-018734). The research leading to these results received funding from the European Union Seventh Framework Program (FP7/2007-2013) under grant agreement no. Health-F2-2009-241762, for the FLIP project.


Nonalcoholic fatty liver disease (NAFLD) is highly prevalent and being overweight is a significant risk factor. The aim was to build an algorithm along with a scoring system for histopathologic classification of liver lesions that covers the entire spectrum of lesions in morbidly obese patients. A cohort of 679 obese patients undergoing liver biopsy at the time of bariatric surgery was studied. An algorithm for segregating lesions into normal liver, NAFLD, or nonalcoholic steatohepatitis (NASH) was built based on semiquantitative evaluation of steatosis, hepatocellular ballooning, and lobular inflammation. For each case, the SAF score was created including the semiquantitative scoring of steatosis (S), activity (A), and fibrosis (F). Based on the algorithm, 230 obese patients (34%) were categorized as NASH, 291 (43%) as NAFLD without NASH, and 158 (23%) as not NAFLD. The activity score (ballooning + lobular inflammation) enabled discriminating NASH because all patients with NASH had A ≥ 2, whereas no patients with A < 2 had NASH. This score was closely correlated with both alanine aminotransferase (ALT) and aspartate aminotransferase (AST) (P < 0.0001, analysis of variance [ANOVA]). Comparison of transaminase levels between patients with normal liver and pure steatosis did not reveal significant differences, thus lending support to the proposal not to include steatosis in the activity score but to report it separately in the SAF score. In the validation series, the interobserver agreement for the diagnosis of NASH was excellent (κ = 0.80) between liver pathologists. There was no discrepancy between the initial diagnosis and the diagnosis proposed using the algorithm. Conclusion: We propose a simple but robust algorithm for categorizing liver lesions in NAFLD patients. Because liver lesions in obese patients may display a continuous spectrum of histologic lesions, we suggest describing liver lesions using the SAF score. (HEPATOLOGY 2012;56:1751–1759)

Nonalcoholic fatty liver disease (NAFLD) is a highly prevalent disease occurring in 20%-25% of the general population and is currently the main chronic liver disease in the Western world.1-3 Obesity and insulin resistance (IR) are believed to exert the major pathophysiological roles in development of NAFLD.4-6 Due to the tremendously increasing prevalence of obesity in the last half-century, accurate characterization of obesity-associated liver diseases is of crucial importance.7, 8

NAFLD is characterized by macrovesicular steatosis; indeed, the presence of steatosis in at least 5% of hepatocytes is generally accepted as a definition.9, 10 Nonalcoholic steatohepatitis (NASH), a subgroup of NAFLD, deserves particular attention because of the risk for of evolution to cirrhosis.11 There is general agreement that diagnosis of NASH relies on liver biopsy. The diagnosis relies on an association of histological features and a consensus exists on the main elementary features that are commonly observed in NASH (ballooning/clarification, inflammation, perisinusoidal fibrosis, predominance of lesions in zone 3).10, 12. However, and despite significant efforts at clarification by pathologists, such as the NASH-CRN (Clinical Research Network) group, the histological definition of NASH as well as the NAFLD activity score (NAS) were subject to recent controversy.8, 13, 14 Indeed, the list of histological features as well as their amount required for diagnosis were not precisely defined until recently, when a pragmatic definition of NASH was proposed: the association of >5% macrovesicular steatosis, lobular inflammation of any degree, and liver cell ballooning of any amount.9 This definition, which has been endorsed by several recognized hepatologists and pathologists, established the list of necessary criteria, as well as minimal semiquantitative required thresholds for each of these three criteria. Although this definition provides firm grounds for multicentric epidemiologic studies and clinical trials, it has not yet been tested in large series of morbidly obese patients.

Histological scoring systems have been successively developed in most of the chronic liver diseases.15-17 They have been widely used and some of them belong now to everyday practice. The NAFLD activity score (NAS) has been proposed and has been repeatedly used both in clinical trials and for evaluation of noninvasive markers in morbidly obese patients.18-22 It is presumed to mirror disease activity that is more adequate than the simple distinction between NASH versus NAFLD without NASH. NAS is the unweighted summation of semiquantitative evaluation of steatosis, ballooning, and lobular inflammation. It had repeatedly been stated that NAS was not considered a diagnostic tool, but rather a continuous scale for activity assessment.13, 23 Indeed, by removing perisinusoidal fibrosis and zone 3 predominance of lesions from the actual definition, NASH is differentiated from pure NAFLD simply by histological markers of ongoing liver damage (i.e., activity), the presence of ballooning, and lobular inflammation. Therefore, there is a rationale for expecting a close correlation between a relevant activity score and the presence of NASH. Although the likelihood of NASH increases with NAS, there exists a wide gray zone (NAS 3-4) where NASH may or may not be present.10

Therefore, the aims of our study were to devise an algorithm for categorization of obesity-associated liver diseases, to apply this algorithm to a large population of morbidly obese patients so as to assess the prevalence of liver lesions in a well-defined group of obese undergoing bariatric surgery, to validate the algorithm in a series of biopsies in patients with metabolic syndrome without morbid obesity, and to make new proposals regarding the scoring of obesity-associated liver diseases.


ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; GGT, γ-glutamyltransferase; NAFLD, nonalcoholic fatty liver disease; NAS, NAFLD activity score; NASH, nonalcoholic steatohepatitis; SAF, steatosis-activity-fibrosis.

Patients and Methods


The study enrolled 705 obese subjects included in a bariatric surgery program, prospectively and consecutively recruited by the Department of Nutrition of Hotel-Dieu and Pitié-Salpêtrière Hospitals (reference center for medical and surgical care of obesity in Paris, France).24 All patients had undergone liver biopsy at the time of the surgical procedure. Subjects were between 13 and 69 years old and the criteria for obesity surgery, i.e., body mass index (BMI) ≥40 kg/m2 or ≥35 kg/m2 and at least one comorbidity (hypertension, type II diabetes, dyslipidemia, or obstructive sleep apnea syndrome). Of note, only three patients were operated on before the age of 18 due to morbid forms of obesity with negative impact on health. The subjects were weight-stable (i.e., variation of less than ±3 kg) during the 3 months prior to surgery. Preoperative evaluation included detailed medical history, physical, nutritional, metabolic, cardiopulmonary, and psychological assessments. Subjects were excluded if they showed evidence of autoimmune, inflammatory, or infectious diseases including viral hepatitis, cancer, hepatotoxic treatment, or known alcohol consumption (>20 g/day). Subjects were classified as diabetic according to the criteria of fasting glycemia over 7 mM or use of an antidiabetic drug. Blood samples were collected after overnight fast. Liver tests, including alanine aminotransferase (ALT), aspartate aminotransferase (AST), γ-glutamyltransferase (GGT), total cholesterol, and HDL-cholesterol were immediately measured enzymatically. Plasma glucose was measured using enzymatic methods. Serum insulin concentrations were determined with an immunoradiometric assay kit (Bi-Insulin IRMA; CisBio International, Saclay, France). The Ethics Committee (CPP Ile-de-France 1) approved the clinical investigations. All subjects gave written informed consent prior to inclusion in the study.

For the validation study, an independent series of 60 needle biopsies (mean length 22 mm ± 6) from patients with metabolic syndrome in the absence of morbid obesity or other known liver diseases were retrospectively selected from the files of the Department of Pathology of Beaujon Hospital.

Liver Biopsy.

The liver biopsy was formalin-fixed and paraffin-embedded. Serial sections were stained. The minimal set of staining included hematoxylin and eosin (H&E), picroSirius red, and Perls staining. All biopsies were reviewed by the same liver pathologist (P.B.). For each biopsy, a preestablished form for evaluation of the main histologic patterns was filled out by the pathologist blinded to clinical and biological data except for the fact that biopsies were obtained at the time of bariatric surgery. The histological check-list was from the NASH-CRN with few additional comments.10 Among major features, ballooning was graded from 0 to 2 (0: normal hepatocytes with cuboidal shape and pink eosinophilic cytoplasm; 1: presence of clusters of hepatocytes with a rounded shape and pale cytoplasm usually reticulated. Although shape is different, size is quite similar to that of normal hepatocytes; 2 same as grade 1 with some enlarged hepatocytes, at least 2-fold that of normal cells). Lobular inflammation was defined as a focus of two or more inflammatory cells within the lobule. Foci were counted at 20× magnification (0: none; 1: ≤2 foci per 20×; 2: >2 foci per 20×). Although we did not change the definition of hepatocellular ballooning proposed by the NASH-CRN, we added reference to the size and shape of hepatocyte for clarity and the same number of scales (0-2) was used for both lesions. These features are illustrated in Fig. 1.

Figure 1.

Ballooning and lobular inflammation grading. (A) Liver biopsy showing in the upper part, normal hepatocytes, and, in the lower part, ballooning, grade 1. H&E, ×20. (B) Normal hepatocytes, ballooning, grade 0. Cytoplasm is pink and granular and liver cells have sharp angles. H&E, ×40. (C) Ballooning, grade 1. Hepatocytes have rounded contours with clear reticular cytoplasm. Size is quite similar to that of normal hepatocytes, H&E, ×40. (D) Ballooning, grade 2. Cells are rounded with clear cytoplasm and twice as large as normal hepatocytes (star), H&E, ×40 (E) Lobular inflammation, grade 1. There is one focus of inflammatory cells in a background liver with steatosis. H&E, ×20. (F) Lobular inflammation, grade 2. Several inflammatory foci within the lobule. H&E, ×20.

Histopathological Algorithm.

NAFLD is defined by the presence of steatosis in >5% of hepatocytes, and NASH by the presence, in addition, of hepatocellular ballooning of any degree and lobular inflammatory infiltrates of any amount.9 Therefore, steatosis was used as the criteria for entry into the algorithm weighted by hepatocellular ballooning and lobular inflammation. In this algorithm tree, all cases with at least grade 1 steatosis were diagnosed NAFLD independently of other criteria, whereas when each of the three features (steatosis, ballooning, lobular inflammation) was classified as at least grade 1, then the lesion was categorized as NASH. Therefore, the algorithm included 10 possible means of arriving at a decision and is shown in Fig. 2. The algorithm was applied to the whole series of biopsies in order to assess prevalence of the different categories.

Figure 2.

Diagnostic algorithm for NASH.

Validation Study.

To assess the strength of the algorithm and its usefulness in NAFLD other than morbid obesity-associated liver diseases, two liver pathologists (P.B. and V.P.) independently evaluated a series of biopsies of patients with metabolic syndrome in the absence of morbid obesity and classified the lesions using the algorithm previously described. Inter- and intraobserver variations were assessed using kappa statistics. In this series, the diagnosis obtained with algorithmic classification was compared to the initial diagnosis of the biopsy report that was based on the classical definition of NASH.

SAF Score.

For each patient an SAF score (steatosis, activity, fibrosis) summarizing the main histological lesions was defined. The steatosis score (S) assessed the quantities of large or medium-sized lipid droplets, but not foamy microvesicules, from 0 to 3 (S0: <5%; S1: 5%-33%, mild; S2: 34%-66%, moderate; S3: >67%, marked). Activity grade (A, from 0-4) was the unweighted addition of hepatocyte ballooning (0-2) and lobular inflammation (0-2) as described previously. Cases with A0 (A = 0) had no activity, A1 (A = 1), mild activity, A2 (A = 2), moderate activity, A3 (A ≥ 3) severe activity.

Stage of fibrosis (F) was assessed using the score described by NASH-CRN as follows; stage 0 (F0) none); stage 1 (F1): 1a or 1b perisinusoidal zone 3 or 1c portal fibrosis, stage 2 (F2): perisinusoidal and periportal fibrosis without bridging, stage 3 (F3): bridging fibrosis and stage 4 (F4): cirrhosis.10

Figure 3 shows a biopsy of a morbidly obese patient with SAF score S3A2F1 (marked steatosis, moderate activity, and mild fibrosis) and a biopsy scored S1A3F4 (cirrhosis with severe activity and mild steatosis).

Figure 3.

Liver biopsy of obese patients. (A) SAF score S3A2F1; marked steatosis, moderate activity, and mild fibrosis. (B) SAF score S1A3F4; cirrhosis with severe activity and mild steatosis.

Statistical Analyses.

Statistical analyses were performed using SAS 9.1 software. Continuous data were expressed as means (± standard deviation [SD]) and noncontinuous data in percentages. Comparison between groups used Student t test, analysis of variance (ANOVA) followed by P for linear trend posttest when appropriate, and the chi-square test for noncontinuous variables. Weighted kappa scores were used to measure the degree of interobserver and intraobserver agreement between pathologists and for a given pathologist. Strength of agreement was considered slight for values between 0 and 0.19, fair for values between 0.20 and 0.39, moderate for values between 0.40 and 0.59, substantial for values between 0.60 and 0.79, and almost perfect if κ values were greater than 0.80.


Among the 705 obese patients who underwent liver biopsy at the time of bariatric surgery, 26 patients dropped out because biopsy material was insufficient or not representative enough (surgical biopsy less than 1 cm depth into the liver), leaving 679 patients for the study, including 539 females (80%) with a mean age of 43 years (range: 13-69). The mean body weight was 132 ± 25 kg and the BMI 48 ± 8 kg/m2. Prevalence of diabetes, dyslipidemia and hypertension were respectively 33%, 49% and 48%, respectively. Main clinical and biological data are shown in Table 1.

Table 1. Main Clinical and Biological Criteria in the Cohort of 679 Patients Undergoing Bariatric Surgery
 All PatientsNo NAFLDNAFLD (Including NASH)P
  1. *Values are N (%), means ± SD.

  2. P values derived from chi-square tests for categorical variables (Fisher's exact test when expected numbers were small), from Student t test for quantitative measures.

Age at surgery (yrs)43 ± 1136.9 ± 1144.3 ± 11<0.0001
Females (%)539 (80%)143 (90%)396 (76%)<0.0001
Weight (kg)131.9 ± 25.4127.3 ± 22133.3 ± 260.01
BMI (kg/m2)48 ± 845.6 ± 848.3 ± 7.70.0001
AST (U/L)28 ± 1323.9 ± 9.329.8 ± 13.6<0.0001
ALT (U/L)35 ± 2724.4 ± 17.138.1 ± 29<0.0001
GGT50 ± 9638.9 ± 4754.0 ± 106NS
Total bilirubin6.6 ± 36.5 ± 2.56.6 ± 3.1NS
Diabetes (%)218 (33%)12 (8%)206 (43%)<0.0001
Dyslipidemia (%)321 (49%)51 (34%)270 (54%)<0.0001
Hypertension (%)319 (48%)49 (33%)270 (54%)<0.0001
Glucose (mmol/l)6.2 ± 2.35.3 ± 1.16.5 ± 2.4<0.0001
Insulin (μU/ml)22.4 ± 3714.5 ± 10.124.8 ± 41.8< 0.01
Total cholesterol (mmol/l)4.9 ± 14.9 ± 1.14.9 ± 1.0NS
Triglycerides (mmol/l)1.6 ± 1.31.3 ± 1.61.7 ± 1.2< 0.001
HDL (mmol/l)1.2 ± 0.31.3 ± 0.41.2 ± 0.3<0.0001

Figure 4 shows the distribution of patients according to pathways followed in the algorithmic tree. According to the algorithm, 158 (23%) patients displayed no steatosis (grade 0), whereas the remaining 521 patients (77%) had steatosis, of at least grade 1, and were included in the NAFLD group. In the group of patients without NAFLD, 2 had ballooning (grade 1) without lobular inflammation and 9 had lobular inflammation without ballooning (8 with grade 1). In addition, 36 patients without steatosis displayed some degree of fibrosis (mainly stage 1). Altogether, these 47 cases (7%) were considered as minimal obesity-associated liver disease, leaving 111 patients (16%) with definitive normal liver.

Figure 4.

Number of patients (%) in the different algorithmic pathways. In bold, pathways leading to diagnosis of NASH.

Among the group of 521 patients with NAFLD, 230 patients (34%) meet the criteria for NASH (steatosis + ballooning, any amount + lobular inflammation, any degree). The remaining 291 patients (43%) were categorized as NAFLD (without NASH) including 141 patients with pure steatosis and 150 with steatosis and either inflammation or ballooning. Table 1 shows the main clinical and biological data of patients with and without NAFLD. Patients with NAFLD were significantly older (43 ± 11 versus 37 ± 11, P < 0.0001), with a higher BMI (48.3 ± 7.7 versus 45.6 ± 8, P < 0.0001) and more frequently had diabetes (43% versus 8%, P < 0.0001), dyslipidemia (54% versus 34%, P < 0.0001), or hypertension (54% versus 33%, P < 0.0001).

The reproducibility and usefulness of the algorithm were assessed in a series of liver biopsies of patients with metabolic syndrome. Using the algorithm, the interobserver agreement for the final diagnosis (no NAFLD, NAFLD without NASH, or NASH) was very good (κ = 0.8), as was intraobserver agreement (κ = 0.82). Agreement between pathologists for the three features used in the algorithm was substantial with a kappa value of 0.8, 0.69, and 0.60 for steatosis, ballooning injury, and lobular inflammation, respectively. Intraobserver agreement was higher in all categories than interobserver agreement (0.82, 0.72, and 0.70, respectively). Interestingly, there was no discrepancy between the initial diagnosis established at time of the biopsy and the classification obtained after re-reading the slides using the algorithm, blinded to the initial diagnosis.

Although the probability of having NASH or not was closely associated with extreme values of NAS, among the 204 patients with a NAS score from 3 to 4, 116 (57%) had no NASH, whereas 88 (43%) had NASH. Detailed data are shown in Fig. 5. In contrast, and stemming from the definition of NASH, the activity score (ballooning + lobular inflammation) was strongly correlated with diagnosis of NASH. When setting the threshold at A ≥ 2, 230 out of 249 patients (92%) were correctly classified as NASH, whereas all cases with A < 2 were correctly classified as having no NASH (Fig. 5).

Figure 5.

(A) Distribution of NAS according to presence of NASH (algorithmic definition, in green, % of cases without NASH, in red, % of cases with NASH). (B) Distribution of NASH according to activity grade. (C) Correlation between AST and activity grade. (D) Correlation between ALT and activity grade.

As expected, there was a strong correlation between score of activity and ALT (ANOVA, P < 0.0001) and AST (P < 0.0001). Data are shown in Fig. 5. Thus, this activity score provides a robust histological approach that matches transaminase levels and clearly distinguishes most patients with NASH.

The NAS includes steatosis as a component of its activity score. We sought to determine whether steatosis could truly be a marker of activity. For this purpose we compared transaminase levels of patients with strictly normal liver (S0A0F0, n = 111) with those patients with pure steatosis (SanyA0F0, n = 74). Interestingly, there were no significant differences between these groups, either for ALT (24 ± 17 IU in patient with normal liver versus 27 ± 19 IU in those with pure steatosis, n.s.) or for AST (24 ± 8 IU in patient with normal liver versus 25 ± 9 IU in those with pure steatosis, n.s.), supporting the decision not to consider steatosis as a marker of activity and to dissociate grading of steatosis from the activity score.

According to the fibrosis score, 256 patients (38%) had no fibrosis (F0), 253 (37%) were F1 (96 scored 1a, 80 scored 1b, 77 scored 1c), 122 (18%) were F2, 42 (6%) were F3, and 6 (1%) had cirrhosis (F4). The prevalence of significant fibrosis (F ≥ 2) was 4 (3%) out of 147 patients with normal liver, 51 (18%) out of 291 patients with NAFLD/no NASH, and 114 (50%) out of 230 patients with NASH.

Finally, considering significant disease as A ≥ 2 and/or F ≥ 2, 303 of the 679 patients (45%) met this criterion. Among them, 54 patients had F ≥ 2 but A < 2, whereas 133 had A ≥ 2 but F < 2. Detailed data are shown in Fig. 6.

Figure 6.

(A) Correlation between activity grade and fibrosis stage. (B) Between fibrosis stage and activity grade.


Although clinical variables and noninvasive tests are able to predict advanced fibrosis, the diagnosis of NASH, a subgroup of NAFLD, continues to rely on liver biopsy.21, 25, 26 A recent review of the literature found major discrepancies in reports of the prevalence of NASH in individuals with severe obesity, ranging from 24% to 98%, with a plausible explanation being that the definition of NASH varies among pathologists.8, 27 Indeed, although the elementary features usually present in NASH are well known, the histopathological definition of NASH as an entity has changed slightly over time. Based on a seminal article, the definition included steatosis, ballooning, mixed acute and chronic lobular inflammation, and zone 3 perisinusoidal fibrosis.12 More recently, the NASH-CRN stated that NASH was “established by the presence of a characteristic pattern of steatosis, inflammation and hepatocellular ballooning on liver biopsies in the absence of significant alcohol consumption,” thereby eliminating sinusoidal fibrosis from the definition.23 A distinction between no NASH, borderline, and definite NASH has also been proposed, underlining current difficulty in establishing a clear distinction between what constitutes NASH and what does not.10, 23 These ambiguities were recently highlighted by intense debate between several pathologists.13 In this context, it is therefore not surprising that diagnosis may vary according to pathologists, justifying the need for clarifying this issue. A step toward a more practical definition was recently taken, based on conclusions of an American Association for the Study of Liver Diseases (AASLD) workshop, and this definition has been endorsed by renowned pathologists.9 It provides a list of minimal criteria with quantitative information on the necessary diagnostic features: i.e., steatosis >5%, hepatocyte ballooning of any degree, and lobular inflammation of any grade. Although this might be considered a simplified approach, it is a step toward creating a more pragmatic and convenient definition for this entity. It is expected that this clear statement will create greater coherence among pathologists, essential for carrying out multicentric epidemiologic studies and clinical trials.

Therefore, we used this definition to draw an algorithm for histopathological classification of liver biopsies of patients with morbid obesity and we applied it to a large series of liver biopsies of adult patients at the time of bariatric surgery. According to this definition, steatosis >5% was considered the minimal required criterion for inclusion in the NAFLD group, and we found that, at the time of bariatric surgery, a significant group of obese patients (23%) did not reach this threshold. It must be emphasized that 5% steatosis is an arbitrary threshold that is difficult to accurately assess. Indeed, cases with rare steatotic vacuoles, but characteristic liver cell injuries and inflammation, are not uncommon in morbidly obese patients, especially at a stage of advanced fibrosis. In our series, 11 patients (2%) had liver cell ballooning or lobular inflammation without steatosis. This suggests that liver cell damage may occur in the absence of steatosis or that steatosis has vanished. Although biopsies were clearly abnormal, it remains unclear as to whether, in the absence of steatosis, these cases should be designated as NAFLD. In the algorithm, steatosis was then modulated by cardinal features of hepatocyte ballooning and lobular inflammation using, for both, a zero-to-two grading scale. According to the definition of NASH, a minimum of grade 1 hepatocyte ballooning in association with a minimum of grade 1 lobular inflammation needs to be present in order for the lesion to be classified as NASH. In the present study, we provide a precise description of the semiquantitative scoring of hepatocyte ballooning in order to minimize interobserver variation, one of the major limitations of liver biopsy. Indeed, liver cell ballooning/clarification is among the most difficult lesions to consistently identify and grade among pathologists.10 In the present study, we added references to the shape and size of adjacent normal hepatocytes in order to complement the current definition; this should help to better recognize and semiquantify this feature. Immunohistochemical markers such as loss of cytokeratin 8/18 expression in clarified/ballooned hepatocytes might also be of use.28

It might be argued that the algorithmic approach is an oversimplification. Indeed, only three histopathologic patterns (clarification/ballooning, lobular inflammation, and steatosis) are taken into account, whereas NASH usually displays a more complex pattern, sometimes with Mallory-Denk hyalines or apoptotic bodies. Nevertheless, comparison between the initial diagnosis and the algorithmic classification blinded from initial diagnosis in the validation cohort did not show any discrepancies, suggesting the robustness of the algorithmic approach and indicating that most of the information for diagnosing NASH is contained in the three cardinal features. The slight interobserver variation in the validation study also lends support to the usefulness of this algorithm.

NAS has been proposed as a semiquantitative scoring system similar to the activity score in chronic hepatitis.10 It has been repeatedly cautioned that, like any scoring system, it should not be used for diagnostic purposes (i.e., diagnosis of NASH), although clinical trials have often selected patients with NASH based on an NAS value ≥5.23, 27 Consistent with previous studies, our study shows that, in obese patients, NAS does not match closely with the diagnosis of NASH. Although extreme values are closely linked to the absence (NAS < 3) or presence of NASH (NAS > 4), there is a gray zone concerning intermediate values. Indeed, the discrepancy between NAS and the diagnosis of NASH is linked to the inclusion of steatosis in NAS. Taking chronic viral hepatitis as a paradigm, activity grade is related to histopathologic features that are presumed to favor development of liver fibrosis, the main histologic factor predictive of an unfavorable outcome in NAFLD.14 The rationale for including steatosis in an activity score is questionable. Experimental studies have shown that triglycerides themselves are not toxic and may in fact protect the liver from lipotoxicity by buffering the accumulation of fatty acids.29, 30 Indeed, patients with rare genetic diseases display massive steatosis but will not develop inflammation or fibrosis, as is the case for most animal models of NAFLD. This is also emphasized by results of the present study showing that liver function tests, including transaminases, did not differ when pure steatosis was compared to normal liver. Although transaminase is not a sensitive marker, the close correlation between transaminases and the score of lobular inflammation and cell injury, on the one hand, and the absence of a correlation with steatosis, on the other hand, encourage us to clearly dissociate grade of steatosis from grade of activity, thus defining a new activity score for NAFLD that combines only lobular inflammation (0-2) and hepatocellular ballooning (0-2). Stemming directly from the definition, the setting of a threshold at 2 for this activity score enables clear-cut recognition of most cases of NASH when A ≥ 2. Whether liver cell injury and lobular inflammation have the same prognostic value and should be associated warrants further studies; however, in light of the strong association between these two features when examining case-by-case, we have combined them, with similar weights for each (from 0 to 2).

In the present study, 18 cases (2%) displayed either isolated severe ballooning (16 cases) or severe lobular inflammation (2 cases). They were not included in the NASH group so as to remain consistent with the proposed definition; however, these cases underline the fact that creating a boundary between what does and does not constitute NASH is somewhat artificial for a disease that may evolve spontaneously or following treatment along the entire course of the histopathologic spectrum, sometimes with dissociation between evolution of hepatocytic, inflammatory, steatosis, and fibrous lesions. Therefore, we recommend that liver lesions be described using the SAF score that dissociates grade of steatosis, grade of activity, and stage of fibrosis, a simple scoring system that seems more relevant than simply dichotomizing cases according to the presence or absence of NASH. By separately describing the main histopathologic features in a practical manner, the SAF score enables easy comparison between biopsies and changes observed in paired biopsies during clinical trials. The SAF score is able to cope with any case and would eventually permit retrieval of the biopsy with NAFLD (S≥1AanyFany) or with NASH (S≥1A≥2Fany) according to the given definition. In addition, the SAF score can identify special subgroups of patients difficult to classify, such as those with steatosis and perisinusoidal fibrosis, but no cell injuries or lobular inflammation (S≥1A0F1) and cases with inflammation and cell injuries but no steatosis (S0A≥2Fany). Whether or not this scoring is relevant in other clinical contexts, such as NASH associated with other liver diseases or pediatric NASH, warrants further study.

Finally, no scoring system is meant to replace the analytical description of the biopsy, which enables introducing refinements, but both the algorithm and the SAF score might provide practical tools for pathologists that would simplify comprehension of liver lesions by hepatologists.


We thank Mme. Christine Baudouin, Dr. Florence Marchelli, and Mme. Patricia Ancel (Center for Research on Human Nutrition, Pitié-Salpêtrière Hospital, Paris, France), who were involved in patient recruitment, data collection, and sampling. We thank Patricia Bonjour for help in histochemistry. Dr. M. Aissat also contributed to bariatric surgery intervention in the patients. We thank Mrs. Nathalie Colnot who provided technical assistance for tissue processing.