With the epidemic of overweight and obesity in the LIVER, PANCREAS, AND BILIARY TRACT

Transcription

1 GASTROENTEROLOGY 2001;121: LIVER, PANCREAS, AND BILIARY TRACT Nonalcoholic Fatty Liver Disease: Predictors of Nonalcoholic Steatohepatitis and Liver Fibrosis in the Severely Obese JOHN B. DIXON,* PRITHI S. BHATHAL, and PAUL E. O BRIEN* *Monash University Department of Surgery, Alfred Hospital, and Department of Pathology, University of Melbourne, Melbourne, Victoria, Australia Background & Aims: Nonalcoholic fatty liver disease (NAFLD) is common in severely obese subjects and can progress to cirrhosis and liver failure. Predicting advanced or progressive disease may help in selecting patients for liver biopsy and assist the development of therapeutic options. Methods: Liver biopsies were taken at laparoscopic obesity surgery in 105 consecutive patients. The clinical and biochemical variables were analyzed for correlation with specific histologic features. Results: Twenty-six patients (25%) were found to have nonalcoholic steatohepatitis (NASH), and 11 (42%) of these had advanced fibrosis. A raised index of insulin resistance (odds ratio [OR] 9.3, 95% confidence interval [CI] ), systemic hypertension (OR 5.2, 95% CI ), and raised alanine aminotransferase (OR 8.6, 95% CI ) were independent predictors of NASH. A combination of 2 or 3 of these predictors allows a sensitivity of 0.8 and specificity of 0.89 for NASH. Alcohol consumption was associated with a reduction in NASH (OR 0.35, 95% CI ) and diabetes (OR 0.18, 95% CI ). Conclusion: Insulin resistance and systemic hypertension, features of the metabolic syndrome, are independently associated with advanced forms of NAFLD. Moderate alcohol consumption seems to reduce the risk of NAFLD in the severely obese, possibly by reducing insulin resistance. With the epidemic of overweight and obesity in the developed and developing world, 1,2 the clinician is increasingly confronted with the problems and dilemmas associated with nonalcoholic fatty liver disease (NAFLD). The spectrum of this disease is broad from steatosis and nonalcoholic steatohepatitis (NASH) through to cirrhosis and liver failure. After exclusion of other forms of liver disease, the clinician needs to be able to select patients at risk of progressive disease, investigate them appropriately, and develop management strategies. NAFLD is commonly associated with severe obesity, defined as a body mass index (BMI) higher than 35 kg/m 2, with 74% to 90% of liver biopsies showing fatty change. 3,4 NASH, first named by Ludwig et al., 5 is probably a less aggressive condition than alcoholic hepatitis, but nevertheless may progress through necroinflammatory change and early fibrosis to cirrhosis. 6,7 In contrast, nonalcoholic steatosis without necro-inflammatory change is generally a benign condition. 8 A strict definition of NASH is important. Lee 9 argued against the broad definition including all with macrovesicular steatosis and parenchymal inflammation as NASH. He proposed that, in addition, either hepatocyte ballooning ( Mallory bodies) or fibrosis in a zone 3 pericellular or perivenular pattern be required. Indeed, the presence of either zone 3 fibrosis or ballooning degeneration Mallory bodies on index biopsy has been shown to predict more aggressive disease, cirrhosis, and higher mortality within 10 years. 10 A postmortem study by Wanless et al. 11 indicates that 12% of those dying with cirrhosis show the features of NASH. The challenge for the clinician in assessing the obese patient with abnormal liver function tests is to be able to separate those with steatosis alone from those with progressive disease. A number of studies have clearly associated the more severe liver changes with features of the metabolic syndrome or syndrome X. 3,12 14 Central features of this syndrome include insulin resistance, central or visceral obesity, dyslipidemia, impaired glucose tolerance, and hypertension. 15 Kral et al. 12 found that the severity of steatosis was related to the level of central Abbreviations used in this paper: BAAT, body mass index, age, ALT, and serum triglyceride; BMI, body mass index; CI, confidence interval; HAIR, hypertension, ALT, insulin resistance; HbA1c, glycosylated hemoglobin A1c; HOMA, homeostasis model assessment; IR, insulin resistance; NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis; OR, odds ratio; QUICKI, quantitative insulin sensitivity check index; ROC, receiver operating characteristic; TGF 1, transforming growth factor by the American Gastroenterological Association /01/$35.00 doi: /gast

2 92 DIXON ET AL. GASTROENTEROLOGY Vol. 121, No. 1 obesity and was more common in men. These findings were subsequently supported by 2 retrospective studies of severely obese patients that have shown steatosis to be related to male gender, central obesity, impaired glucose tolerance, and the dyslipidemia of obesity. 3,13 Features of the metabolic syndrome have also been linked to hepatic fibrosis. 13,14,16,17 Between 15% and 20% of the adult population of the Western world is obese (BMI 30 kg/m 2 ). 18,19 The clinician needs to be able to select from this group those at highest risk of NASH who may warrant a liver biopsy. The aim of this study was to identify, in a severely obese population, the clinical and/or biochemical risk factors that would predict significantly abnormal liver histology. The study excludes patients in which liver changes were not likely to be primarily related to obesity or obesity-associated disease (such as high alcohol intake or hepatitis C). Patients and Methods Study Design This prospective study of 105 consecutive patients was a part of an extensive preoperative and perioperative data collection on patients with severe obesity (BMI 35 kg/m 2 ) having an adjustable gastric band (Lap-Band System; Bioenterics, Carpenteria, CA) inserted laparoscopically for weight reduction. A liver biopsy taken as a routine part of the operative procedure was assessed for specific abnormalities, and preoperative predictors of these abnormalities were sought. The study sample size was determined to detect a 33% difference between groups, calculated using a power of 0.8, with a significant P Calculations using known characteristics of an obese population showed that a sample size of 50 to 100 subjects would be required for key study variables. The study design planned analysis after approximately 100 consecutive liver biopsies had been performed. Patients and Preoperative Assessment Patients with a BMI greater than 35 kg/m 2, suffering significant medical, physical, or psychosocial disabilities, and who had attempted weight reduction by other means for at least 5 years, were considered for entry into the program. After 2 consultations with the surgeon when clinical evaluation was made and the procedure discussed extensively, patients who elected to proceed underwent an extensive preoperative assessment. This includes clinical assessment by a consultant physician, anthropometric measurements, and laboratory tests. An assessment of alcohol consumption was obtained at the surgical consultation, 2 medical consultations, and separately in a questionnaire completed by the patient. Laboratory tests included liver function tests, iron studies, fasting lipid profile, fasting plasma glucose, fasting insulin, C-peptide (marker of endogenous insulin production), glycosylated hemoglobin A1c (HbA1c), and hepatitis B and C serology. Studies were performed to exclude hemachromatosis, 1 -antitrypsin deficiency, Wilson s disease, or autoimmune liver disease if indicated on the liver biopsy. Patients were excluded from the NAFLD study if they had a history of alcoholism, had an alcohol consumption greater than 200 g/week, had evidence of hepatitis B or C, were taking known hepatotoxic medications, or had a history of or finding consistent with another specific liver disease. Diagnosis of type 2 diabetes was based on the American Diabetes Association criteria. 20 Hypertension was diagnosed if the patient had a past history of hypertension and was on antihypertensive medication or if the patient had a resting recumbent blood pressure of greater than or equal to 140/90 on at least 2 occasions. Alcohol consumption was classified into 4 groups: (0) no alcohol, (1) less than 20 g per week, (3) 20 to 100 g per week, and (4) greater than 100 g per week. Liver biopsies were performed as a routine part of the operative procedure, which is performed in the left upper quadrant. The liver biopsy was taken under laparoscopic view using a 14 gauge 20 mm Temno Biopsy (Allegiance; Health Care Corp., McGraw Park, IL) needle introduced percutaneously into the left lobe of the liver. Adequacy of the biopsy was assessed macroscopically, and an additional core was taken if a specimen of less than 8 mm in length was obtained. At laparoscopy, the appearance of the liver was scored on a scale of 0 to 3 for each of size, fatty color change, and surface nodularity. Informed written consent was obtained from all patients, and the study was conducted in conformance with the Helsinki Declaration. Histologic Assessment All liver biopsy specimens were examined using H&E, silver reticulin, Masson trichrome and Sirius red for collagen, 21 and Perls stain. The presence of Mallory bodies was confirmed by immunostaining for expression of ubiquitin. 22 One pathologist (P.S.B.), blinded to the patient s clinical condition and biochemical data, scored each biopsy on a scale of 0 to 4 for steatosis, injury/inflammation, and fibrosis, according to criteria set out in Table 1. The scoring of fibrosis (Table 1) was based on criteria developed by Brunt et al. 23 and Lee. 24 The criteria for injury/inflammation are those proposed by Lee. 24 Steatosis was graded on a scale 0 to 4. Liver biopsy specimens were included if there were 6 or more portal tracts; none were excluded for being inadequate. Most specimens contained 12 to 24 portal tracts. A diagnosis of NASH was based on steatosis and 2 of the following 3, zone 3-centric features: (1) necro-inflammatory foci with mononuclear cells and/or neutrophils, (2) ballooning degeneration of hepatocytes with or without Mallory bodies, and (3) pericellular fibrosis. 9 Portal fibrosis with or without portal inflammation in the absence of zone 3 pericellular fibrosis was disregarded in the staging of NASH. This feature was analyzed separately for those biopsies that did not have NASH.

3 July 2001 PREDICTORS OF NASH AND FIBROSIS IN OBESITY 93 Table 1. Criteria Used for Histologic Scoring Steatosis 0 No steatosis 1 5% of lobular parenchyma involved % of lobular parenchyma involved % of lobular parenchyma involved 4 75% of lobular parenchyma involved Inflammation 0 No hepatocyte injury or inflammation 1 Sparse zone 3 inflammation 2 Mild focal zone 3 hepatocyte injury/inflammation 3 Noticeable zone 3 hepatocyte injury/inflammation 4 Severe zone 3 hepatocyte injury/inflammation Fibrosis a 0 Normal connective tissue 1 Focal pericellular fibrosis in zone 3 2 Perivenular and pericellular fibrosis confined to zone 2 and 3 with or without portal/periportal fibrosis 3 Bridging or extensive fibrosis with architectural distortion; no obvious cirrhosis 4 Cirrhosis NOTE. Histologic criteria used in this study are modifications of those published by Lee. 24 a Portal fibrosis in the absence of zone 3 pericellular fibrosis was disregarded in scoring of fibrosis in NASH. Specimens showing this feature were placed in a separate category and analyzed. Insulin Resistance Insulin levels were measured using an immunoenzymometric assay (MEIA; Abbott Diagnostics, Abbott Park, IL) with inter- and intra-assay coefficients of variation less than 3%. C-peptide was measured using a double antibody competitive radioimmunoassay (Diagnostic Products Corp., Los Angeles, CA) with inter- and intra-assay coefficients of variation less than 7%. Two methods were used to assess the combined effect of hyperglycemia and hyperinsulinemia on liver histology, both indirect measures of insulin resistance (IR). The commonly used term insulin resistance is simply the inverse of insulin sensitivity. First, a validated method of estimating IR and islet-cell function using the fasting plasma glucose and C-peptide measures was used to calculate percentage insulin sensitivity (HOMA%S) and the percentage of islet-cell function (HOMA%B). This homeostasis model assessment (HOMA) was originally developed by Matthews and was later modified. 25,26 Secondly, a quantitative insulin sensitivity check index (QUICKI), recently described 27 and based on the log transformed insulin-glucose product, was used. This index has been shown to correlate well with the hyperinsulinemic euglycemic clamp technique, and therefore, its inverse is a good indirect measure of IR. IR index {(log insulin) (log fasting plasma glucose)} 1/QUICKI Data Analysis Histologic scoring and features were treated as ordinal categorical variables. For example, fibrosis was scored 0 to 4, and NASH was scored 0 or 1. Continuous variables were assessed for correlation with histologic scoring using nonparametric bivariate analysis. Two samples of quantitative variables were tested by 2-sided, Student t test (mean [SD]) or by Mann-Whitney U test (median [interquartile range]). Some quantitative laboratory variables, e.g., fasting plasma insulin, required log transformation before parametric analysis (geometric mean [interquartile range]). 2 method (Fisher exact) was used to test the significance of differences between proportions and categorical variables. Odds ratio (OR) was calculated for some 2 2 tables and expressed with 95% confidence interval (CI). Multivariate analysis was tested using binary logistic regression (forward and backward) and linear regression analysis. Receivers operating characteristic curves (ROC) were used for assessing an appropriate cutoff for the continuous variable, IR index, and for a score combining the independent predictors of NASH. The SPSS 28 statistical software was used for statistical analysis. A P value of less than 0.05 was considered statistically significant. No correction was used for assessing for correlation with multiple variables. Results One hundred eight liver biopsies were performed at laparoscopy during Three exclusions, 1 for hepatitis C and 2 with known history of alcoholism, allowed for 105 patients to be included in the study. The M:F ratio was 23:82. This reflected the sex distribution of patients operated on in a previously reported larger group. 29 The mean age was years, and the mean BMI was 47 7 kg/m 2. In this group of 105 subjects, there were 40 patients with hypertension, 19 with type 2 diabetes, and 1 type 1 diabetic. The aspartate aminotransferase (AST) level was raised ( 34 IU/L) in 19 patients, and the alanine aminotransferase (ALT) level was raised ( 40 IU/L) in 32 patients. The predictive effects of AST and ALT were similar on histology and had a high correlation with one another (r 0.85). As ALT is an enzyme that is more specific for liver cell injury, this enzyme was assessed in the multivariate analysis. 30 A summary of the major histologic findings is shown in Table 2. Twenty-six patients (25%) had NASH. Of these, 11 had advanced fibrosis (stage 3 & 4), including Table 2. Liver Histology in the Severely Obese n n Normal 4 Steatosis a 75 Without inflammation 40 With nonspecific inflammation 35 Steatohepatitis 26 With zone 3 fibrosis 26 (100%) With Mallory bodies 16 (62%) With advanced fibrosis (stage 3) 10 (38%) With cirrhosis (stage 4) 1 (4%) Total 105 a Portal fibrosis without zone 3 fibrosis was considered separately.

4 94 DIXON ET AL. GASTROENTEROLOGY Vol. 121, No. 1 1 with cirrhosis. Portal fibrosis in the absence of zone 3 fibrosis and NASH occurred in 19 biopsies, and in 16 was associated with portal inflammation. Steatohepatitis (NASH) The predictors for the 26 patients with NASH are shown in Table 3. Using logistic regression, waist-to-hip ratio was the most significant predictor from the range of patient physical measurements. Raised fasting plasma glucose and fasting insulin levels had independent predictive value for NASH, with the IR index an indicator of their combined effect. Of the enzymes measured in the liver function panel, raised ALT and AST levels correlated best with a diagnosis of NASH, and alkaline phosphatase levels were of no significance. No biopsies were found to have significant iron loading, and plasma levels of iron, transferrin, and iron saturation were no different in those with NASH. Ferritin levels were significantly raised in those with NASH. Table 3. Comparison Between Those With NASH (n 26) With Those Without NASH (n 79) No NASH (mean, median, %) NASH (mean, median, %) P value Number Age (yr) 40.4 (11) 44.0 (11) NS BMI (kg/m 2 ) 44 (7) 47.2 (7) NS % in group male (%) Waist (cm) 125 (14) 137 (17) W:H Ratio 0.87 (0.08) 0.97 (0.08) Neck (cm) 41.5 (4.4) 44.0 (4.4) Diabetic (%) Hypertensive (%) Fasting P glucose (mmol/l) 5.2 (1) 7.2 (4.2) HbAlc (%) 5.4 (0.5) 6.7 (2.4) F P Insulin (mu/l) 17.2 (12) 25 (12) C Peptide (pmol/ml) 1.2 (0.7) 1.63 (0.7) Insulin sensitivity HOMA%S 38 (25) 27 (11) Beta cell function HOMA%B 162 (68) 121 (151) 0.06 Insulin resistance index (1/QUICKI) 4.5 (0.6) 5.3 (0.04) Total cholesterol (mmol/l) 5.8 (1.5) 5.8 (1.2) NS Fasting triglycerides (mmol/l) 1.7 (0.9) 2.3 (1.3) HDL-C (mmol/l) 1.20 (0.36) 1.15 (0.44) NS LDL-C (mmol/l) 3.7 (1.3) 3.55 (1.0) NS AST (IU/L) 20 (7.5) 34 (20) ALT (IU/L) 24 (15) 48 (23) AST/ALT ratio 0.86 (0.3) 0.81 (0.3) NS GGT (IU/L) 27 (25) 46 (34) Alk Phos (IU/L) 53.7 (26) 51 (27) NS Ferritin ( g/l) 73 (136) 138 (228) NOTE. For all laboratory measures: median (interquartile range), P value Mann-Whitney U (2-tail). For continuous demographics and anthropometric: mean (SD), P value unpaired t test (2-tail). Proportions: percentage, P value chi-square. HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; GGT, -glutamyltransferase. Figure 1. ROC curve for the independent predictors of NASH: IR index 5.0, ALT 40, and hypertension. A cutoff HAIR score of 2 gives the best combination of sensitivity (0.8) and specificity (0.89). Three factors were found to have significant independent predictive effects. These were IR index, ALT level, and hypertension. Cutoff values for the 2 continuous variables were 5.0 for IR index (based on ROC) and 40 IU/L for ALT (upper limit of laboratory reference range). A HAIR score for each patient (0 3) was calculated by adding Hypertension 1, ALT 40 IU 1, and IR index The ROC curve (Figure 1) shows the respective sensitivities and specificities for any combination of 1 to 3 of the HAIR predictors. ORs for the 3 independent predictors of NASH were hypertension 5.2, 95% CI , P 0.001; ALT 40 IU, 8.6, 95% CI , P 0.001; and IR 5.0, 9.3, 95% CI , P All 11 patients with a score of 3 had NASH on liver biopsy. The presence of at least 2 of the 3 provides the best combination of sensitivity (0.8) and specificity (0.89) for predicting NASH. The area under the curve for the combination of these 3 predictors was Figure 2. The association between NASH and type 2 diabetes, and hypertension. The percentage of patients with NASH in each group are indicated.

5 July 2001 PREDICTORS OF NASH AND FIBROSIS IN OBESITY 95 The association between type 2 diabetes and hypertension with NASH is shown in Figure 2. If a patient has neither condition, the prevalence of NASH is 7%. With both of these conditions, the prevalence is 75%. The appearance of the liver at laparoscopy added no predictive value for NASH when examined in a model with the 3 independent predictors above. Histology Grading Clinical and laboratory correlations with scored histologic features on liver biopsy are shown in Table 4. Of the 105 liver biopsies examined, 101 had some steatosis, 61 some inflammation, and 27 some zone 3 fibrosis. Steatosis Predictors of the degree of hepatic steatosis are shown in Table 4. The indirect measures of IR, IR index, and 1/HOMA%S showed the best correlation with steatosis (r 0.47, P 0.001). Binary logistic regression for predictors of those with high level steatosis (score 3 and 4) showed that a diagnosis of diabetes and a high waist-to-hip ratio were the only independent clinical predictors. IR index and ALT were the only independent laboratory predictors. When modeled together, only the laboratory predictors remained significant. A diagnosis of hypertension was not of independent predictive value. The size of the liver at laparoscopy was an additional independent predictor of the level of steatosis. The steatosis score on biopsy was not predictive of NASH when modeled with the 3 independent predictors. Inflammation The predictors of inflammation associated with NAFLD are understandably similar to NASH (Table 4). For the 26 with a histologic diagnosis of NASH, only a Table 4. Analysis of the Association Between Clinical and Laboratory Data and Histologic Features on Liver Biopsy (n 105) Steatosis (0 4) Inflammation (0 4) Fibrosis (0 4) Age (yr) NS NS NS Sex (male) d NS P P BMI (kg/m 2 ; all 35 kg/m 2 ) NS NS NS Weight (kg) 0.20 a NS 0.22 a Waist circumference (cm) 0.27 a 0.24 a 0.34 c Waist:hip ratio 0.30 b 0.39 c 0.35 c Neck circumference (cm) 0.42 c 0.28 a 0.28 a Hypertension d P P P Diabetes type 2 d P P P Fasting glucose (mmol/l) 0.38 c 0.44 c 0.41 c HbA1c (%) 0.36 c 0.34 c 0.35 c Fasting insulin (mu/l) 0.35 c 0.41 b 0.38 b C Peptide (pmol/ml) 0.38 c 0.38 c 0.37 c Insulin resistance index (1/QUICKI) 0.47 c 0.49 c 0.47 c HOMA%S 0.47 c 0.44 c 0.43 c HOMA%B NS NS NS Total cholesterol (mmol/l) NS NS NS Fasting triglyceride (mmol/l) 0.27 b 0.22 a NS HDL cholesterol (mmol/l) 0.22 a NS NS LDL cholesterol (mmol/l) NS NS NS Total protein (g/l) 0.20 a 0.23 a 0.22 b Albumin (g/l) NS NS NS Globulin (g/l) NS NS NS ALT (IU/L) 0.42 c 0.48 c 0.50 c AST (IU/L) 0.40 c 0.52 c 0.54 c GGT (IU/L) 0.27 b 0.34 b 0.38 c Alkaline phosphatase (IU/L) NS NS NS Bilirubin ( mol/l) NS NS NS Ferritin ( g/l) NS 0.20 a 0.25 a BAAT Score 0 4 e P P P 0.08 NOTE. Spearman correlation coefficients used for continuous clinical and laboratory variables. HDL, high-density lipoprotein; LDL, low-density lipoprotein; GGT, -glutamyltransferase. a Correlation is significant at 0.05 level (2-tail). b Correlation is significant at 0.01 level (2-tail). c Correlation is significant at level (2-tail). d P value Mann-Whitney U (2-tail). e BAAT score (BMI, age, ALT, and triglyceride) as described by Ratziu et al 14 (P value Kruskal-Wallis).

6 96 DIXON ET AL. GASTROENTEROLOGY Vol. 121, No. 1 raised ALT level was associated with higher levels of inflammation. Inflammation with a score of 1 (n 30) or 2 (n 5) was found in 35 of 75 subjects with steatosis who did not meet the criteria for NASH on liver biopsy. This inflammatory change was therefore not specific for NASH. When this group of 35 were compared with the 44 with no inflammatory changes (40 with steatosis and 4 normal), significant differences were found. This group had higher ALT levels (mean [SD]; 38 [29] IU/L vs. 25 [16] IU/L, P 0.018), and lower high-density lipoprotein cholesterol levels (1.13 [0.24] vs [0.27], P 0.009). When compared with those with NASH, significantly less disturbance of glucose metabolism, lipid metabolism, and lower aminotransferase levels were found. This group had significantly less type 2 diabetes (OR 0.08, 95% CI , P 0.001) and hypertension (OR 0.23, 95% CI , P 0.007) and were more likely to drink alcohol regularly (OR 3.0, 95% CI , P 0.042). Ten of the biopsies showed epithelioid granulomas (not lipogranulomas), and of these, only 2 had NASH. The remainder had steatosis. Fibrosis There were 2 patterns of fibrosis. First there was zone 3 fibrosis, a key feature of NASH, and secondly, fibrosis confined to the portal tracts with no zone 3 abnormality. Zone 3 pericellular fibrosis. There were 27 patients with at least a score of 1, on a scale of 1 to 4, for this type of fibrosis. All 26 with NASH had fibrosis, and one other biopsy scored 1 for this form of fibrosis but did not meet the criteria for NASH. The predictors of fibrosis on liver biopsy are similar to those of NASH and are shown in Table 4. There were 11 subjects with advanced (stage 3 or 4) fibrosis. The most significant clinical risk factors were hypertension (OR 8.7, 95% CI , P 0.002), male gender (OR 8.5, 95% CI , P 0.001), and type 2 diabetes (OR 4.4, 95% CI , P 0.018). All patients with advanced fibrosis had diabetes or hypertension or both (Figure 3). Logistic regression identified 3 independent clinical predictors of advanced fibrosis. These were hypertension, high ALT level, and high C-peptide level. All patients with advanced fibrosis had a C-peptide level above the laboratory reference value of 1324 pmol/l. A comparison of those with mild to moderate (stage 1 & 2) fibrosis (n 16) and with advanced (stage 3 & 4) fibrosis (n 11) is shown in Table 5. The group with more advanced fibrosis were more likely to be Figure 3. The percentage of subjects with type 2 diabetes and hypertension in relation to the stage of fibrosis: no zone 3 fibrosis, mild to moderate fibrosis, and advanced fibrosis. male and have significantly better islet-cell function (HOMA%B), as indicated by higher mean C-peptide levels and lower fasting plasma glucose levels. Fibrosis confined to the portal tract. After excluding those with pericellular fibrosis in zone 3 (n 27), a group of 19 patients had isolated portal fibrosis on liver biopsy. In 16 of 19, this was associated with portal inflammation regarded as not specific for NASH. This group of 19 were compared with the remaining 59 subjects. Hypertension (OR 7.6, 95% CI , P 0.001) was the only condition that was predictive of portal fibrosis. Twelve of 19 with portal fibrosis had hypertension, with 11 of the remaining 59 having hypertension. There were no other clinical or biochemical predictors. Of importance, there was no increase in any measure of glucose metabolism or IR, no difference in any lipid measures, and no difference in aminotransferase levels. For 10 patients with stage 1 zone 3 fibrosis and NASH, 9 had portal fibrosis, and 7 of these were hypertensive. Mallory Bodies Mallory bodies were found in 16 (62%) of the 26 biopsies showing NASH. All with Mallory bodies had NASH, and all had at least a score of 1 for zone 3 fibrosis. In addition to the 3 independent predictors of NASH, no alcohol consumption was a significant predictor of Mallory bodies: 12 of 48 with zero alcohol consumption had Mallory bodies, whereas they were present in only 2 of 40 with an alcohol consumption of greater than 20 g/week (P 0.01). This is shown in Figure 4. Alcohol Consumption Forty-eight subjects consumed no alcohol. There was minimal consumption ( 20 g/week) by 17, consumption of greater than 20 g/week and less than 100 g/week by 31, and greater than 100 g by 9. There

7 July 2001 PREDICTORS OF NASH AND FIBROSIS IN OBESITY 97 Table 5. Zone 3 Fibrosis Score (n 27): A Comparison of Those With Mild-Moderate vs Advanced Fibrosis Mild-mod fibrosis (1&2; n 16) Advanced fibrosis (3&4; n 11) P value Age (yr) 42.3 (13) 44.6 (11) NS BMI (kg/m 2 ) 47 (7) 48 (7) NS % male % with hypertension NS % with type 2 Diabetes NS F P Glucose (mmol/l) 9.1 (3) 6.4 (1.2) HbA1c (%) 7.34 (1.9) 6.23 (0.8) 0.06 C Peptide (pmol/ml) 1.44 (0.4) 1.9 (0.5) F Insulin (mu/l) a 25.5 (13) 29.7 (12) NS Insulin resistance index 5.21 (0.54) 5.29 (0.31) NS HOMA%S a 26.5 (12.7) 23.7 (9.0) NS HOMA%B a 93.2 (132) (102) Waist:hip ratio 0.94 (0.11) 0.99 (0.06) NS NOTE. Continuous variables mean (SD), P value unpaired t test. Proportions: percentages, P value chi-square test. a Required Ln transformation. Geometric mean (interquartile range). was an OR for diabetes of 0.18 (95% CI , P 0.005), for NASH of 0.35 (95% CI , P 0.040), and for Mallory bodies of 0.16 (95% CI , P 0.010) in alcohol consumers vs. nonconsumers. Figure 4 shows the percentage of subjects in each alcohol consumption category and the prevalence of NASH, Mallory bodies, and type 2 diabetes. The effect of alcohol on NASH was not significant after controlling for diabetes or IR index. Alcohol consumption is not independently predictive of NASH; its effects are reflected in components of the HAIR score. Global Histologic Score For statistical purposes, the sum of each patient s score for steatosis, inflammation, and fibrosis provided a global histologic score. The scores were normally distributed with a mean of 4.3 (SD 2.7). This allowed a forward stepwise regression analysis for independent factors influencing this score. Three factors were independently predictive of the global score: IR index (r ), hypertension (r ), and ALT (r ). When combined, these factors predicted 55% (r ) of the variance in the global histologic score. Discussion This study found that 26% of severely obese subjects presenting for weight loss surgery have NASH and 42% of these have advanced fibrosis on liver biopsy, and has confirmed the significance of central weight distribution and some features of the metabolic syndrome as important predictors of NASH and hepatic fibrosis. The study was confined to those with a BMI 35 kg/m 2, and the findings cannot necessarily be generalized to those with a lower BMI. However, BMI, in this study (BMI 35 70), was not correlated with histologic change. Of the anthropometric measurements, weight distribution as measured by waist-to-hip ratio was highly significant. This is the first study of NAFLD to measure fasting plasma insulin and C-peptide levels in addition to fasting plasma glucose, and first to show that they are independent predictors of NASH. As hypothesized by others, 3,13 hyperinsulinemia and increased IR are indeed associated with adverse histologic findings. The relationship between IR and steatosis as determined by ultrasound examination has been recently reported. 31 The IR index is the best predictor of zone 3-centric steatosis, inflammation, and fibrosis with the exception that C-peptide alone is the best predictor of advanced fibrosis (stage 3 & 4). Patients with advanced fibrosis when compared with those with mild to moderate fibrosis (stage 1&2)were found to have significantly lower fasting plasma glucose with significantly higher C-peptide levels reflecting better islet-cell function. These findings suggest that more extensive fibrosis is seen in patients with hyperinsulinemia and better islet-cell function, and warrants further exploration. We found systemic hypertension to be a significant independent predictor of NASH, zone 3 fibrosis of any stage and advanced (stage 3& 4) fibrosis, and is the only predictor of the nonspecific portal fibrosis occurring in association with NAFLD. Angiotensin II levels may contribute to hepatic fibrosis seen in NAFLD patients with hypertension. A possible mechanism is via angiotensin II enhancing the production of the profibrogenic cytokine transforming growth factor 1 (TGF- 1), which con- Figure 4. The prevalence of NASH, Mallory bodies, and type 2 diabetes for groups based on alcohol consumption. The effect of alcohol reducing the likelihood of NASH is probably related to its effect on IR and reduced risk of type 2 diabetes.

8 98 DIXON ET AL. GASTROENTEROLOGY Vol. 121, No. 1 tributes to hepatic stellate cell activation Blocking angiotensin II may be a potential method of reducing expression of TGF- 1, 33 thereby reducing hepatic fibrosis. Powell et al. 32 have shown a statistically significant relationship between the inheritance of high TGF- 1 and angiotensinogen-producing genotypes and the development of progressive hepatic fibrosis in patients with chronic hepatitis C. In the current study, light to moderate alcohol consumption in the severely obese seems to reduce the risk of steatosis, NASH, and Mallory bodies seen on liver biopsy. Diabetics seemed less likely to consume alcohol, and there was a negative relationship between blood HbA1c levels and alcohol consumption. Alcohol consumption has been associated with lower insulin levels 35 and reduced IR, 36 and reduces the risk of developing type 2 diabetes. 37 Alcohol intake has a significant effect on lowering post-load glucose levels, but not fasting levels. 38 This may explain why HbA1c levels correlated negatively with alcohol intake, but not fasting glucose levels. Modest alcohol consumption has been associated with reduced mortality in both men and women, 39,40 and part of the cardiovascular beneficial effects of alcohol may be mediated by reduced IR. 36,41 Reduced IR with light to moderate alcohol consumption provides a possible explanation for the beneficial effect it seems to have on liver histology in the obese, often insulin-resistant, group. Rigorous exclusion of subjects with regular alcohol consumption from other studies of NAFLD may have obscured this effect in the past. The effect of alcohol in apparently reducing the risk of NASH and type 2 diabetes in the severely obese needs careful evaluation as it clearly has important management implications. Combining any 2 of the 3 independent predictors of NASH (HAIR score 2 or 3) in the severely obese provides a method of selecting groups that are more at risk of progressive liver damage. This method may assist in selection of patients for liver biopsy in the future. Using the data of the present study liver biopsies on 30 patients with a HAIR score of 2 or 3 would have identified 21 of the 26 with NASH and 10 of 11 with advanced fibrosis. The many variables that show correlation with histologic changes and NASH, for example, waist-to-hip ratio, are correlated with the components of the HAIR score, and their influence is reflected in this score. They are not independently predictive of NASH. The finding that clinical and laboratory features were better predictors of NASH and advanced fibrosis than was the appearance of the liver at laparoscopy or the steatosis score on liver biopsy, suggests imaging the liver to assess its size or fat content may be of limited value. Dyslipidemia as a predictor of the type of NAFLD is unclear from our study. We found that in the severely obese, abnormal glucose metabolism and hypertension are the features of the metabolic syndrome that have the greatest value in predicting both NASH and fibrosis. The aspect of histology that correlates best with dyslipidemia and least with hypertension is steatosis (Table 4). The main focus of some studies has been an assessment of the degree of steatosis, and this may have biased the role serum lipids may play in the prediction NAFLD. 3,13 There is also the possibility that dyslipidemia has more influence on those with mild to moderate obesity. The current study was confined to those with severe obesity (BMI 35 kg/m 2 ). In a recent study 14 of overweight patients, a BMI of 28 kg/m 2 or greater, age over 50 years, an ALT level greater than twice the upper limit of the laboratory reference range, and a triglyceride level of 1.7 mmol/l or greater (BAAT score), were found to be significant predictors of septal fibrosis. These patients were being investigated for abnormal liver function tests. However, their patients differ from ours, in that a significant proportion were referred with abnormal liver function tests discovered during the exploration of dyslipidemia. 14 Triglycerides and age were not significant factors in the current study. All our patients had a BMI greater than 28 kg/m 2, and levels of ALT above 2N were of little predictive value. The combination BAAT score (0 4) 14 was only weakly predictive of NASH and zone 3 fibrosis in our study. Clearly, the nature of selection of the patient cohort group may alter which predictors are associated with histologic change. Obesity and IR have been associated with fibrosis and cirrhosis in other liver disease. High BMI has been associated with steatosis and fibrosis in hepatitis C 42 and with cirrhosis in alcoholic liver disease. 43 In addition to weight loss, therapeutic options for NASH should include optimal control of diabetes, hypertension, and dyslipidemia. Exercise may be of great value as it reduces IR 44 and reduces visceral obesity preferentially, even in the absence of weight loss. 45 Medications that either specifically reduce the impact of angiotensin II or reduce IR may be of therapeutic value. In summary, this study of severely obese subjects has identified risk factors that are predictive of NASH and advanced fibrosis. We have shown that IR, a central feature of the metabolic syndrome, is also strongly associated with the more advanced forms of NAFLD. In addition to the other features of the metabolic syndrome,

9 July 2001 PREDICTORS OF NASH AND FIBROSIS IN OBESITY 99 we have found systemic hypertension to be a significant independent predictor of advanced fibrosis (stage 3&4). Three independent predictors of NASH have been identified, and a HAIR score of 0 to 3 has been calculated based on these predictors. A score of 2 or 3 is highly predictive of NASH, but further prospective validation is necessary. Low to moderate alcohol consumption seems to be associated with a lower risk of NASH. This is possibly because of its beneficial effect of reduced IR and reducing the risk of type 2 diabetes. References 1. Mokdad AH, Serdula MK, Dietz WH, Bowman BA, Marks JS, Koplan JP. The spread of the obesity epidemic in the United States, JAMA 1999;282: Obesity: preventing and managing the global epidemic. Report of the WHO Consultation on Obesity. Geneva, Switzerland: WHO, Luyckx FH, Desaive C, Thiry A, Dewe W, Scheen AJ, Gielen JE, Lefebvre PJ. Liver abnormalities in severely obese subjects: effect of drastic weight loss after gastroplasty. Int J Obes Relat Metab Disord 1998;22: Clain D, Lefkowitch J. Fatty liver disease in morbid obesity. Gastroenterol Clin North Am 1987;16: Ludwig J, Viggiano TR, McGill DB, Oh BJ. Nonalcoholic steatohepatitis: Mayo Clinic experiences with a hitherto unnamed disease. Mayo Clin Proc 1980;55: Sheth S. Nonalcoholic steatohepatitis. Ann Intern Med 1997; 126: Powell E, Cooksley W, Hanson R, Searle J, Halliday J, Powell L. The natural history of nonalcoholic steatohepatitis: a follow up study of 42 patients for up to 21 years. Hepatology 1990;11: Teli MR, James OF, Burt AD, Bennett MK, Day CP. The natural history of nonalcoholic fatty liver: a follow-up study. Hepatology 1995;22: Lee RG. Nonalcoholic steatohepatitis: tightening the morphological screws on a hepatic rambler. Hepatology 1995;21: Younossi ZM, Matteoni C, Gramlisch T, Liu Y, Boparai N, Wohl D, McCullough AJ. Patient characteristics predicting cirrhosis and death in non-alcoholic liver disease (abstr). Hepatology 1998;28: 303A. 11. Wanless I, Lentz J. Fatty liver hepatitis (steatohepatitis) and obesity: an autopsy study with analysis of risk factors. Hepatology 1990;12: Kral J, Schaffner F, Pierson R, Wang J. Body fat topography as an independent predictor of fatty liver. Metabolism 1993;42: Marceau P, Biron S, Hould F, Marceau S, Simard S, Thung S, Kral J. Liver pathology and the metabolic syndrome X in severe obesity. J Clin Endocrinol Metab 1999;84: Ratziu V, Giral P, Charlotte F, Bruckert E, Thibault V, Theodorou I, Khalil L, Turpin G, Opolon P, Poynard T. Liver fibrosis in overweight patients. Gastroenterology 2000;118: Timar O, Sestier F, Levy E. Metabolic syndrome X: a review. Can J Cardiol 2000;16: Angulo P, Keach JC, Batts KP, Lindor KD. Independent predictors of liver fibrosis in patients with nonalcoholic steatohepatitis. Hepatology 1999;30: Silverman EM, Sapala JA, Appelman HD. Regression of hepatic steatosis in morbidly obese persons after gastric bypass. Am J Clin Pathol 1995;104: Australian Institute of Health and Welfare. Australia s health 1998: the sixth biennial health report of the Australian Institute of Health and Welfare. Australia s Health 1998: Flegal KM, Carroll MD, Kuczmarski RJ, Johnson CL. Overweight and obesity in the United States: prevalence and trends, Int J Obes Relat Metab Disord 1998;22: Colman P, Thomas D, Zimmet P, Welborne T, Garcia-Webb P, Moore M. New classification and criteria for diagnosis of diabetes mellitis. MJA 1999;170: James J, Bosch KS, Zuyderhoudt FM, Houtkooper JM, van Gool J. Histophotometric estimation of volume density of collagen as an indication of fibrosis in rat liver. Histochemistry 1986;85: Vyberg M, Leth P. Ubiquitin: an immunohistochemical marker of Mallory bodies and alcoholic liver disease. APMIS Suppl 1991; 23: 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: Lee RG. Alcoholic and nonalcoholic steatohepatitis pathology. In: Bloomer JR, Goodman ZD, Ishak KG, eds. Clinical and pathological correlations in liver disease: approaching the next millenium. Washington, DC: The Armed Forces Institute of Pathology, The American Registry of Pathology, The American Association for the Study of Liver Diseases, 1998: Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC. Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia 1985;28: Levy JC, Matthews DR, Hermans MP. Correct homeostasis model assessment (HOMA) evaluation uses the computer program [letter; comment]. Diabetes Care 1998;21: Katz A, Nambi SS, Mather K, Baron AD, Follmann DA, Sullivan G, Quon MJ. Quantitative insulin sensitivity check index: a simple, accurate method for assessing insulin sensitivity in humans. J Clin Endocrinol Metab 2000;85: SPSS for Windows ed. Chicago, IL: SPSS. Inc., O Brien P, Brown W, Smith A, McMurrick P, Stephens M. Prospective study of a laparoscopically placed, adjustable gastric band in the treatment of morbid obesity. Br J Surg 1999;86: Kew MC. Serum aminotransferase concentration as evidence of hepatocellular damage. Lancet 2000;355: Marchesini G, Brizi M, Morselli-Labate AM, Bianchi G, Bugianesi E, McCullough AJ, Forlani G, Melchionda N. Association of nonalcoholic fatty liver disease with insulin resistance. Am J Med 1999;107: Powell EE, Edwards-Smith CJ, Hay JL, Clouston AD, Crawford DH, Shorthouse C, Purdie DM, Jonsson JR. Host genetic factors influence disease progression in chronic hepatitis C. Hepatology 2000;31: Kim S, Ohta K, Hamaguchi A, Omura T, Yukimura T, Miura K, Inada Y, Ishimura Y, Chatani F, Iwao H. Angiotensin II type I receptor antagonist inhibits the gene expression of transforming growth factor-beta 1 and extracellular matrix in cardiac and vascular tissues of hypertensive rats. J Pharmacol Exp Ther 1995; 273: Friedman SL. Cytokines and fibrogenesis. Semin Liver Dis 1999; 19: Lazarus R, Sparrow D, Weiss ST. Alcohol intake and insulin levels. The Normative Aging Study. Am J Epidemiol 1997;145: Facchini F, Chen YD, Reaven GM. Light-to-moderate alcohol intake is associated with enhanced insulin sensitivity. Diabetes Care 1994;17: Rimm EB, Chan J, Stampfer MJ, Colditz GA, Willett WC. Prospec-

10 100 DIXON ET AL. GASTROENTEROLOGY Vol. 121, No. 1 tive study of cigarette smoking, alcohol use, and the risk of diabetes in men. BMJ 1995;310: Mayer EJ, Newman B, Quesenberry CP Jr., Friedman GD, Selby JV. Alcohol consumption and insulin concentrations. Role of insulin in associations of alcohol intake with high-density lipoprotein cholesterol and triglycerides. Circulation 1993;88: Thun MJ, Peto R, Lopez AD, Monaco JH, Henley SJ, Heath CW Jr., Doll R. Alcohol consumption and mortality among middle-aged and elderly U.S. adults. N Engl J Med 1997;337: Fuchs CS, Stampfer MJ, Colditz GA, Giovannucci EL, Manson JE, Kawachi I, Hunter DJ, Hankinson SE, Hennekens CH, Rosner B. Alcohol consumption and mortality among women. N Engl J Med 1995;332: Flanagan DE, Moore VM, Godsland IF, Cockington RA, Robinson JS, Phillips DI. Alcohol consumption and insulin resistance in young adults. Eur J Clin Invest 2000;30: Hourigan LF, Macdonald GA, Purdie D, Whitehall VH, Shorthouse C, Clouston A, Powell EE. Fibrosis in chronic hepatitis C correlates significantly with body mass index and steatosis. Hepatology 1999;29: Naveau S, Giraud V, Borotto E, Aubert A, Capron F, Chaput JC. Excess weight risk factor for alcoholic liver disease. Hepatology 1997;25: Borghouts LB, Keizer HA. Exercise and insulin sensitivity: a review. Int J Sports Med 2000;21: Thomas EL, Brynes AE, McCarthy J, Goldstone AP, Hajnal JV, Saeed N, Frost G, Bell JD. Preferential loss of visceral fat following aerobic exercise, measured by magnetic resonance imaging. Lipids 2000;35: Received November 6, Accepted March 14, Address requests for reprints to: John Dixon, MBBS, FRACGP, Monash University Department of Surgery, Alfred Hospital, Melbourne 3181, Australia. john.dixon@med.monash.edu.au; fax: (61)