Clinical Studies

Useful parameters for distinguishingnonalcoholic steatohepatitis with mildsteatosis from cryptogenic chronichepatitis in the Japanese population

Naoki Tanaka1, Eiji Tanaka1, YoSheena2, Michiharu Komatsu1,Wataru Okiyama1, Noriko Misawa1,Hidetomo Muto1, Takeji Umemura1,Tetsuya Ichijo1, AkihiroMatsumoto1, Kaname Yoshizawa1,Akira Horiuchi3 and KendoKiyosawa1

1Division of Gastroenterology, Department of

Internal Medicine, Shinshu University School of

Medicine, Matsumoto, Japan, 2Department of

Economics, Shinshu University, Matsumoto,

Japan, 3Department of Gastroenterology, Showa

Inan General Hospital, Komagane, Nagano,

Japan

Tanaka N, Tanaka E, Sheena Y, Komatsu M, OkiyamaW, Misawa N, MutoH, Umemura T, Ichijo T, Matsumoto A, Yoshizawa K, Horiuchi A,Kiyosawa K. Useful parameters for distinguishing nonalcoholicsteatohepatitis with mild steatosis from cryptogenic chronic hepatitis in theJapanese population.Liver International 2006: 26: 956–963.r 2006 The Authors. Journal compilation r 2006 Blackwell Munksgaard

Abstract: Background/Aims: As detecting mild steatosis is difficult byabdominal ultrasonography (US), nonalcoholic steatohepatitis (NASH) withmild steatosis may sometimes be confused with cryptogenic chronic hepatitis.We aimed to test this possibility and to isolate factors that may indicateNASH. Methods: First, 53 Japanese patients diagnosed as havingcryptogenic chronic hepatitis by laboratory examination and US wereenrolled. These patients were histologically divided into NASH and non-NASH groups, and their clinical features were compared. Second, thediagnostic accuracy of predictors of NASH was examined prospectively.Results: Fifteen patients (28%) were histologically diagnosed as havingNASH with mild steatosis. Multivariable analysis revealed that body massindex (BMI) and serum ferritin level were independent predictors of NASH.The best cutoff values to detect NASH were assessed by using receiver-operating characteristic curves: BMI425.2 kg/m2 and serum ferritin level4142 ng/ml. When both markers were concomitantly negative, the negativepredictive value to detect NASH was 100%. Conclusions: In cases of mildsteatosis, US is not a perfect tool for the accurate diagnosis of NASH. BMIand serum ferritin level are useful discriminators of NASH from cryptogenicchronic hepatitis, and might be helpful markers for diagnosing NASH moreaccurately in Japanese patients.

Key words: abdominal ultrasonography – body

mass index – cryptogenic chronic hepatitis –

ferritin – nonalcoholic steatohepatitis

Naoki Tanaka, MD, PhD, Division of Gastroen-

terology, Department of Internal Medicine,

Shinshu University School of Medicine,

Matsumoto, Japan

e-mail: naopi@hsp.md.shinshu-u.ac.jp

Received 31 May 2006,

accepted 18 June 2006

�����������������������������������������������

�����������������������������������������������

Chronic hepatitis is histologically characterizedby sustained hepatocyte injury with apparentinflammation, and is clinically defined as thepersistent elevation of serum aminotransferaselevels for more than 6 months. The commoncauses of chronic hepatitis include persistent viralinfection such as that of hepatitis B virus andhepatitis C virus (HCV), autoimmune disorders,and intake of alcohol, drugs, or chemicals. How-ever, patients with cryptogenic chronic hepatitisor unexplained elevation of serum aminotransfer-ase levels still exist whose exact pathogenesis hasnot been verified. Because cryptogenic chronic

hepatitis may progress to cirrhosis, accurate di-agnosis and treatment are needed.

Nonalcoholic steatohepatitis (NASH) is de-fined as a disease entity showing characteristicpathological findings common to alcoholic liverdisease, including hepatic steatosis, hepatocellu-lar ballooning, and perisinusoidal/pericellular fi-brosis, despite no alcohol consumption. Skellyet al. (1) have reported that 34% of Britishpatients with unexplained abnormal liver func-tion tests are later diagnosed as NASH by liverbiopsy, and in the United States, most patientswith unexplained aminotransferase elevation are

Liver International 2006; 26: 956–963 r 2006 The AuthorsJournal compilation r 2006 Blackwell Munksgaard

DOI: 10.1111/j.1478-3231.2006.01338.x

956

considered to have nonalcoholic fatty liver dis-ease (NAFLD) (2, 3). Recently, NASH has alsobeen recognized as one of the major causes ofchronic hepatitis in Japan, though its prevalencein patients with unexplained aminotransferaseelevation or cryptogenic chronic hepatitis re-mains unclear.Radiological imaging devices such as ultraso-

nography (US), computed tomography, andmagnetic resonance are indispensable in evaluat-ing hepatic steatosis. When lipid accumulation isobserved in more than 33% of hepatocytes, thesemodalities have a strong ability to accuratelydiagnose hepatic steatosis (4). Of these, US isthe least invasive and thus most preferredmethod. If fatty infiltration is evident by USexamination, it is simple to conclude that unex-plained persistent elevation of aminotransferaselevels may be caused by NAFLD or NASH.However, mild (accumulation of triglycerides inless than 33% of hepatocytes) or focal steatosiscan be underestimated by imaging modalitiessuch as US (5). Additionally, in advanced stagesof NASH, it is increasingly difficult to detecthepatic steatosis by imaging modalities as stea-tosis regresses and becomes focal as fibrosisprogresses. Therefore, NASH with mild steatosisor advanced fibrosis may be confused with cryp-togenic chronic hepatitis.Based on these premises, we hypothesized that

NASH presenting atypical features such as mildsteatosis may be erroneously included with cryp-togenic chronic hepatitis, and we retrospectivelyexamined histological findings in patients whowere diagnosed as having cryptogenic chronichepatitis based on biochemical data and abdom-inal US to reevaluate the prevalence of NASH.We also compared the clinical features betweenNASH and non-NASH (intrinsically cryptogenicchronic hepatitis) groups and sought to finduseful markers to differentiate NASH from cryp-togenic chronic hepatitis that can be used inaddition to US.

Patients and methods

Study 1

PatientsChronic hepatitis was clinically defined by thefollowing criteria: (1) elevation of serum amino-transferase levels (440 IU/l) on two or moreoccasions during a period of at least 6 monthsand (2) exclusion of extrahepatic-origin elevationof serum aminotransferase levels such as myo-pathy and thyroid diseases by measuring serumlevels of lactate dehydrogenase, creatine kinase,

thyroid hormones, and thyroid-stimulating hor-mone. Patients who showed evidence as havingliver cirrhosis were excluded. One thousand sixhundred and ninety-one patients with chronichepatitis who underwent liver biopsy at ShinshuUniversity Hospital or affiliated hospitals be-tween April 1, 1990 and September 30, 2004were enrolled in this study. The diagnosis ofcryptogenic chronic hepatitis was made accordingto the exclusion criteria shown in Fig. 1: (1) noconsumption of alcohol; (2) negative results forhepatitis B surface antigen (HBsAg), high titer ofhepatitis B core antibody (anti-HBc), and anti-HCV antibody (anti-HCV); (3) exclusion of otherliver diseases such as drug-induced liver injury,autoimmune hepatitis, primary biliary cirrhosis,primary sclerosing cholangitis, Wilson’s disease,hereditary hemochromatosis, and a1-antitrypsindeficiency; and (4) exclusion of obvious hepaticsteatosis in abdominal US, that is, positive hepa-torenal contrast and blurring of vascular walland/or profound attenuation of the diaphragm.Patients presenting obvious hepatic steatosis wereexcluded from this study, as typical NAFLD orNASH is easily distinguishable using US. In total,53 Japanese patients were clinically diagnosed ashaving cryptogenic chronic hepatitis (Fig. 1).At the time of admission for liver biopsy, body

mass index (BMI) was calculated. Patients wereconsidered to have hypertension if their systolic/diastolic pressure was greater than 140/90mmHg,

Fig. 1. The number of patients with chronic hepatitis enrolled inStudy 1. The number of patients is indicated in parentheses.ALD, alcoholic liver disease; CH-C, chronic hepatitis C; CH-B,chronic hepatitis B; AIH, autoimmune hepatitis; PSC, primarysclerosing cholangitis; NAFLD, nonalcoholic fatty liver disease;US, ultrasonography.

957

Predictors of NASH with mild steatosis

or if they were taking anti-hypertensive drugs.Patients were considered to be diabetic if they hada fasting glucose level equal to or higher than126mg/dl, or if they were taking insulin or oralhypoglycemic drugs. Patients were considered tohave hyperlipidemia if their fasting serum levelsof cholesterol and triglycerides were equal to orhigher than 220 and 150mg/dl, respectively, or ifthey were taking lipid-lowering drugs.

Laboratory examinationAll data were obtained in a fasting state andmeasured by standard methods. The homeostasismodel assessment for insulin resistance (HOMA-IR) was calculated using the following equation:[fasting glucose (mg/dl)� fasting insulin (mU/ml)]/405. A HOMA-IR greater than 2.0 is con-sidered to indicate the presence of insulin resis-tance. If the patient had a fasting glucose levelequal to or higher than 140mg/dl, or was takinginsulin, HOMA-IR was not calculated.

Histological diagnosis of NASHBefore liver biopsy, informed consent was ob-tained from each patient. Liver biopsy specimenswere immediately fixed in 10% neutral formalin.Sections were cut at 4mm thickness and stainedwith hematoxylin and eosin and Azan–Mallorymethods. Histological diagnosis of NASH wasmade according to the following criteria: macro-vesicular steatosis mainly present in zone 3, hepa-tocellular ballooning and/or perisinusoidal/pericellular fibrosis. Histological findings wereclassified using the grading/staging system pro-posed by Brunt et al. (6) with minor modifications.The degree of hepatic steatosis was expressed asthe percentage of steatotic hepatocytes in biopsiedspecimens. The appearance frequency of hepato-cellular ballooning, glycogenated nuclei, and eosi-nophilic intracytoplasmic inclusion bodies wasgraded as absent, few, or many based on thenumber of hepatocytes showing the respectivechanges. The activity of lobular and portal inflam-mation was graded as absent, mild, moderate, orsevere. Perisinusoidal/pericellular fibrosis wasscored as absent, mild, or severe based on theproportion of zone 3 area involved, and portalfibrosis was assessed as absent, periportal, brid-ging, or cirrhosis. Histological diagnosis was madebe three experts (E. T., N. T., K. Y.).

Detection of hepatic steatosis by USEach patient underwent abdominal US (Hitachimodel EUB-525 equipped with a 3.5MHz con-vex-type transducer, Hitachi, Medical Corp., To-kyo, Japan) in a fasting state. The presence ofhepatic steatosis was assessed independently by

three hepatologists according to findings such ashepatorenal contrast, blurring of the vascularwall, and profound attenuation of the diaphragm.

Study 2

To estimate the diagnostic accuracy of the mar-kers found in Study 1, 256 patients with elevatedserum aminotransferase levels who underwentliver biopsy between October 1, 2004 and March31, 2006 were eligible for entry into this prospec-tive study. Exclusion criteria were positive forHBsAg, anti-HBc, anti-HCV, anti-mitochondrialantibody, the presence of the history of alcoholcomsumption or hepatotoxic drug intake, and thepresence of hepatic steatosis, which was easilydetectable by US. The remaining 22 patients weredivided into three groups according to the num-ber of positive markers, and the final diagnosiswas performed histologically. Diagnostic accu-racy was calculated by sensitivity, specificity, andpositive and negative predictive values (PPV andNPV, respectively).

Statistical analysis

Statistical analyses were performed using SPSSsoftware 11.5J for Windows (SPSS Inc., Chicago,IL). Comparison between the groups was madeusing Fisher’s exact probability test for the cate-gorical variables, w2 test for the histologicalfindings, and Mann–Whitney U-test for the con-tinuous variables, respectively. To assess the useof clinical parameters in differentiating NASHfrom cryptogenic chronic hepatitis, we con-structed receiver-operating characteristic (ROC)curves by plotting the sensitivity against thereverse specificity (1 minus specificity) for eachvalue. In this assessment, a larger area under theROC curve (AUC) corresponds to a more usefulmarker for diagnosing NASH. The most appro-priate cutoff point for the diagnosis of NASHwas the point at which the sum of the sensitivityand the specificity was maximized. To identifyindependent predictors of NASH, multivariablelogistic regression analysis was conducted. Aprobability value of o0.05 was considered statis-tically significant.

Results

Study 1

NASH in cryptogenic chronic hepatitisIn this study, we addressed the specific group ofsubjects with unexplained aminotransferase ele-vation and normal US findings. Of the 53 patientswho were clinically diagnosed as having crypto-

958

Tanaka et al.

genic chronic hepatitis, 15 (28%) fulfilled thehistological diagnostic criteria of NASH (Fig.1). The histological features of these 15 patients,who were not diagnosed as having steatosis byUS, but were histologically confirmed as havingNASH, are shown in Table 1. The degree ofhepatic steatosis was generally mild (o33% ofhepatocytes in the biopsy involved), so steatosiswas hard to detect by US (Figs 2 and 3). Bal-looned hepatocytes, glycogenated nuclei, andperisinusoidal/pericellular fibrosis in zone 3 wereobserved in all NASH patients, and eosinophilicintracytoplasmic inclusion bodies were detectedin 60% of these patients. Lobular and portalinflammation was relatively mild, and portalfibrosis was variable. These results suggest thatdetection of NASH with mild steatosis is clini-cally unreliable using US only.

Comparison of histological findings betweenbiopsy-proven NASH patients and non-NASHpatientsWe compared the histological findings betweenthe biopsy-proven NASH patients, i.e. patientshaving NASH with mild steatosis (NASH group,n5 15), and non-NASH patients, i.e. patientshaving intrinsically cryptogenic chronic hepatitis(non-NASH group, n5 38). As shown in Table 1,the prevalence of macrovesicular steatosis, hepa-

Table 1. Comparison of histological findings between NASH and non-

NASH groups

NASH(n 5 15)

nonNASH(n 5 38) P

Macrovesicular steatosis (%) 0.000o5 0 (0%) 35 (92%)5–20 3 (20%) 2 (5%)21–33 12 (80%) 1 (3%)433 0 (0%) 0 (0%)

Ballooning 0.000Absent 0 (0%) 37 (97%)Few 4 (27%) 1 (3%)Many 11 (73%) 0 (0%)

Glycogenated nuclei 0.000Absent 0 (0%) 33 (87%)Few 9 (60%) 5 (13%)Many 6 (40%) 0 (0%)

Eosinophilic intracytoplasmicinclusion bodies

0.000

Absent 6 (40%) 38 (100%)Few 6 (40%) 0 (0%)Many 3 (20%) 0 (0%)

Lobular inflammation 0.627Absent 2 (13%) 10 (26%)Mild 10 (67%) 19 (50%)Moderate 3 (20%) 8 (21%)Severe 0 (0%) 1 (3%)

Portal inflammation 0.021Absent 4 (27%) 2 (5%)Mild 9 (60%) 16 (42%)Moderate 2 (13%) 11 (29%)Severe 0 (0%) 9 (24%)

Perisinusoidal/pericellular fibrosis 0.000Absent 0 (0%) 37 (97%)Mild 11 (73%) 1 (3%)Severe 4 (27%) 0 (0%)

Portal fibrosis 0.508Absent 8 (53%) 25 (66%)Periportal 2 (13%) 7 (18%)Bridging 4 (27%) 4 (11%)Cirrhosis 1 (7%) 2 (5%)

Data are the number positive and prevalence (in parentheses). A P valuewas calculated using the w2 test. NASH, nonalcoholic steatohepatitis.

Fig. 2. Ultrasonographic and histological findings of the non-alcoholic steatohepatitis patient, who were clinically diagnosedas having cryptogenic hepatitis. (A) A 61-year-old woman withobesity and hyperlipidemia was diagnosed as having crypto-genic chronic hepatitis because of no obvious steatosis byultrasonography. (B) Histological evaluation confirmed thepresence of mild macrovesicular steatosis, ballooned hepato-cytes with eosiniphilic intracytoplasmic inclusion bodies, glyco-genated nuclei, and mild perisinusoidal/pericellular fibrosis(upper photograph, Azan–Mallory staining, � 100; lowerphotograph, hematoxylin and eosin staining, � 400).

959

Predictors of NASH with mild steatosis

tocellular ballooning, glycogenated nuclei, eosi-nophilic intracytoplasmic inclusion bodies, andperisinusoidal/pericellular fibrosis was signifi-cantly higher in the NASH group. Three patientsin the non-NASH group (three in 38 patients,8%) exhibited mild steatosis but could not bediagnosed as having NASH because of the ab-sence of ballooned hepatocytes or perisinusoidal/pericellular fibrosis. Although the activity oflobular inflammation was similar in both groups,portal inflammation tended to be more severe inthe non-NASH group. In NASH livers, lympho-

cytes and/or polymorphonuclear leukocytes werepresent in the inflammatory foci, whereas in non-NASH livers, infiltration of plasma cells and/oreosinophils, as well as lymphocytes, was ob-served. These results demonstrate clear histologi-cal differences between the two groups.

Comparison of clinical features between the NASHgroup and the non-NASH groupTo explore other helpful markers for differentiat-ing NASH with mild steatosis from cryptogenicchronic hepatitis, we compared the clinical fea-tures and laboratory findings between the twogroups. As shown in Table 2, the prevalence ofhyperlipidemia (P5 0.002), BMI (P5 0.001),fasting glucose level (P5 0.021), HOMA-IR(P5 0.009), and serum ferritin level (P5 0.001)were all significantly higher in the NASH group.The prevalence of diabetes and hypertension,serum levels of high-sensitivity C-reactive pro-tein, aminotransferases, and g-glutamyltransfer-ase, immunoglobulin G and A concentrations,and hemoglobin A1c value were not significantlydifferent between the two groups.

Multivariable analysisMultivariable logistic regression analysis revealedthat BMI and serum ferritin level were indepen-dent factors associated with NASH. The oddsratio for BMI was 1.836 [95% confidence interval(CI), 1.063–3.173; P5 0.029], and that for serumferritin level was 1.014 (95% CI, 1.000–1.027;P5 0.048).

ROC curve analysisROC curves were constructed for these twoparameters. The AUCs for BMI and serumferritin level were as great as 0.791 (95% CI,0.644–0.938; P5 0.001) and 0.782 (95% CI,0.651–0.914; P5 0.001), respectively. We nextdetermined the cutoff values of these parametersfor the discrimination between NASH with mildsteatosis and cryptogenic chronic hepatitis byusing the ROC curve. The most appropriate cut-off values were identified as BMI425.2 kg/m2

and serum ferritin level 4142 ng/ml, respectively.

Study 2

We prospectively examined the diagnostic accu-racy of the predictors of NASH found in Study 1.In 22 patients with persistent unexplained eleva-tion of serum aminotransferase levels, eight werehistologically diagnosed as having NASH (Fig.4). Most of these NASH patients exhibited severefibrosis or cirrhosis. When both parameters wereconcomitantly positive, the sensitivity, specificity,and PPV and NPV to detect NASH were 87.5%,85.7%, 77.8%, and 92.3%, respectively. On the

Fig. 3. Ultrasonographic and histological findings of anothernonalcoholic steatohepatitis patient. (A) A 72-year-old womanwith diabetes was diagnosed as having cryptogenic chronichepatitis because obvious fatty infiltration was not confirmedby ultrasonography. (B) Histologically, mild and focal macro-vesicular steatosis, hepatocellular ballooning, and advancedfibrosis were confirmed (upper photograph, Azan–Mallorystaining, � 50; lower photograph, hematoxylin and eosinstaining, � 400).

960

Tanaka et al.

other hand, when both parameters were conco-mitantly negative, the sensitivity, specificity, andPPV and NPV were 100%, 71.4%, 71.4%, and100%, respectively. Although the number ofpatients enrolled in Study 2 was very limited,these results might suggest the relevance of theseparameters as supporting diagnostic markers ofNASH for patients with unexplained persistentaminotransferase elevation.

Discussion

The present study demonstrates that sole relianceon abdominal US might overlook NASH with

mild steatosis, and that BMI and serum ferritinlevel are helpful for differentiating NASH withmild steatosis from cryptogenic chronic hepatitisand for diagnosing NASH more accurately.In NASH, appropriate correction of life style

and pharmacological interventions (e.g. insulinsensitizers) can reduce disease activity and pre-vent the progression (7, 8). These therapeuticstrategies are fundamentally distinct from othertypes of chronic hepatitis, so accurate diagnosis isvery important for NASH. US is the most com-mon diagnostic tool for detecting hepatic steato-sis, which is an essential component of NASH.Generally speaking, US has a relatively highsensitivity and specificity for the detection ofhepatic steatosis. In cases with more than 33%fatty infiltration, the sensitivity and PPV of USare 100% and 62%, respectively (4). However,the diagnostic accuracy of US declines sharply incases of less than 30% fatty infiltration. Thischange may be associated not only with mild orfocal fatty deposition in livers, but also withinterobserver differences in image readings.Although US has been used for the diagnosis ofNAFLD in several studies (9–11), there is apossibility that NASH with mild steatosis hasbeen overlooked. Indeed, Hamaguchi et al. (11)have described that US may lead to an incorrectdiagnosis of NAFLD in 10–30% of cases ana-lyzed. We found that, regardless of no detection

Table 2. Comparison of clinical features and biochemical markers between NASH and non-NASH groups

NASH (n 5 15) non-NASH (n 5 38) P

Age (years) 59 (39–77) 56 (14–72) 0.431Female 57% 85% 0.532Diabetes 27% 16% 0.143Hyperlipidemia 47% 5% 0.002Hypertension 27% 18% 0.275BMI (kg/m2) 26.9 (19.7–33.0) 23.1 (17.9–31.6) 0.001Platelet (�104/ml) 18.2 (12.5–35.0) 19.9 (7.9–35.7) 0.552hsCRP (mg/dl) 0.099 (0.017–0.496) 0.078 (0.004–0.500) 0.103Albumin (g/dl) 4.5 (3.7–4.9) 4.3 (3.3–4.9) 0.232AST (IU/l) 64 (19–161) 57 (16–639) 0.809ALT (IU/l) 75 (30–238) 79 (22–604) 0.713gGT (IU/l) 51 (37–213) 71 (9–269) 0.742Total cholesterol (mg/dl) 213 (155–278) 178 (129–353) 0.054Triglycerides (mg/dl) 122 (66–398) 97 (49–266) 0.088HDL-cholesterol (mg/dl) 40 (20–61) 40 (21–74) 0.826Immunoglobulin G (mg/dl) 1368 (677–3012) 1341 (987–2612) 0.750Immunoglobulin A (mg/dl) 250 (99–487) 255 (80–713) 0.951Glucose (mg/dl) 96 (77–164) 89 (56–113) 0.021Hemoglobin A1c (%) 5.8 (5.0–7.8) 5.3 (4.4–7.2) 0.064HOMA-IRn 2.4 (0.3–8.9) 1.2 (0.2–7.6) 0.009Iron (mg/dl) 145 (67–325) 102 (20–266) 0.104Transferrin saturation (%) 44 (29–94) 34 (5–90) 0.076Ferritin (ng/ml) 229 (62–776) 120 (3–376) 0.001

Qualitative data are expressed as percentages, and quantitative data are written as medians and ranges (in parentheses). A P value for qualitative andquantitative data was calculated using Fisher’s exact probability test and Mann–Whitney U-test, respectively. nHOMA-IR was not evaluated in patientsreceiving insulin therapy or having a fasting glucose level equal to or higher than 140 mg/dl. BMI, body mass index; gGT, g-glutamyltransferase; HDL,high density lipoprotein; HOMA-IR, homeostasis model assessment for insulin resistance.

Fig. 4. Flow chart in Study 2. Twenty-two patients with persis-tent unexplained elevation of aminotransferase levels wereenrolled in this prospective study. The number of patients isindicated in parentheses. 21, body mass index (BMI) 425.2kg/m2 and serum ferritin level 4142 ng/ml; 11, BMI425.2kg/m2 or serum ferritin level 4142 ng/ml; 0, BMIo25.2 kg/m2

and serum ferritin levelo142 ng/ml.

961

Predictors of NASH with mild steatosis

of steatosis by US, 28% of Japanese patients withunexplained elevation of serum aminotransferaselevels have NASH with mild steatosis, suggestingthe inadequacy of US to detect certain types ofNASH. Therefore, US cannot be considered as agold standard test for the accurate diagnosis ofNASH with mild steatosis, and novel diagnosticmarkers of NASH, which may compensate forthis imperfection, are clearly needed.We found that BMI and serum ferritin level are

highly predictive of NASH. It is well known thatNASH is strongly associated with obesity andvisceral fat accumulation (11–13). It has also beenreported that serum ferritin level, a major determi-nant of NAFLD in apparently healthy obeseindividuals (10), is significantly correlated withthe amount of visceral fat mass and hepatic stea-tosis (14). Furthermore, serum ferritin level hasbeen reported to be significantly higher in NASHthan that in simple steatosis, which may reflectincreased hepatic iron overload and enhancedoxidative stress (15). Therefore, assessment of thesemarkers may be useful not only for discriminatingNASH from chronic hepatitis caused by unknownhepatotoxic factors, but also for diagnosing NASHmore accurately and efficiently.As shown in Fig. 3, in patients with NASH in

advanced fibrosis or cirrhosis, the accuracy ofhepatic steatosis detection is markedly reduced,so the correct diagnosis of NASH becomes in-creasingly difficult. The results obtained fromStudy 2 (Fig. 4) suggest that a combination ofBMI and serum ferritin level might be helpfulparameters for distinguishing NASH from cryp-togenic chronic hepatitis, even in advancedstages. Thus, it might be possible to use thesemarkers for the purpose to isolate NASH-derivedcirrhosis (burned-out NASH) from intrinsicallycryptogenic cirrhosis. Further study is needed todetermine whether these markers would be reallyhelpful for identifying the etiology in patientswith cryptogenic hepatitis with advanced fibrosisor cirrhosis.In this study, we could not assess the etiology

of the difference between NASH and NASHwith mild steatosis. The development of hepaticsteatosis depends primarily on the changes ofthree pathways in fatty acid metabolism: in-creased influx of circulating nonesterified fattyacids into hepatocytes, increased de novo lipogen-esis in hepatocytes, and decreased degradationthrough the mitochondrial b-oxidation system(16). Thus, an imbalance of these pathways mightcontribute to the difference in the severity ofsteatosis in NASH.This study has several limitations. First, the

number of patients analyzed was limited in this

retrospective study, so a large-scale prospectiveanalysis is needed to confirm these results. Sec-ond, although it has been reported that severalbiomarkers such as adipocytokines (e.g. adipo-nectin) (17), proinflammatory cytokines (e.g. tu-mor necrosis factor-a) (18), and oxidative stressmarkers (e.g. thioredoxin) (15) might be usefulpredictors of NASH, we could not examine theseparameters in this study. In addition to BMI andserum ferritin level, measuring these biomarkersmay enable us to more accurately diagnoseNASH with mild steatosis. Finally, it is possiblethat some NASH patients might have been mis-placed into the non-NASH group because ofbiopsy sampling errors. NASH livers show moreheterogeneous histological findings than thosewith chronic hepatitis C (19). Moreover, thesampling variability is more significant in liversof burned-out NASH, so patients with advancedstages of NASH would be classified into the non-NASH group. Indeed, in this study, one obesefemale patient with hyperlipidemia, hypertension,diabetes, and hyperferritinemia was histologicallydiagnosed as having cryptogenic cirrhosis be-cause of a lack of histological findings specificto NASH, but her aminotransferase levels nor-malized only by weight reduction, suggesting astrong likelihood of burned-out NASH. Re-peated US-guided biopsy or laparoscopy-assistedbiopsy (20) may minimize the possibility of sam-pling error in patients strongly suspicious ofhaving NASH. To overcome these limitations, itwill be mandatory to establish novel biochemicalmarkers of NASH, which are available even inadvanced stages of NASH and are independent ofthe amount of hepatic steatosis.

In conclusion, US is not a perfect tool for theaccurate diagnosis of NASH with mild steatosis.Additionally, BMI425.2 kg/m2 and serum ferri-tin level 4142 ng/ml may be good non-US dis-criminators of NASH from chronic hepatitis ofunknown etiology.

References

1. Skelly M M, James P D, Ryder S D. Findings on liverbiopsy to investigate abnormal liver function tests in theabsence of diagnostic serology. J Hepatol 2001; 35: 195–9.

2. Daniel S, Ben-Menachem T, Vasudevan G, et al. Pro-spective evaluation of unexplained chronic liver transami-nase abnormalities in asymptomatic and symptomaticpatients. Am J Gastroenterol 1999; 94: 3010–4.

3. Clark J M, Brancati F L, Diehl A M. The prevalenceand etiology of elevated aminotransferase levels in theUnited States. Am J Gastroenterol 2003; 98: 960–7.

4. Saaden S, Younossi ZM, Remer EM, et al. The utility ofradiological imaging in nonalcoholic fatty liver disease.Gastroenterology 2002; 123: 745–50.

962

Tanaka et al.

5. Ataseven H, Yildirim M H, Yalniz M, et al. The valueof ultrasonography and computerized tomography in esti-mating the histopathological severity of nonalcoholic stea-tohepatitis. Acta Gastroenterol Belg 2005; 68: 221–5.

6. Brunt E M, Janney C G, Di Bisceglie A M, et al.Nonalcoholic steatohepatitis: a proposal for grading andstaging the histological lesions. Am J Gastroenterol 1999;94: 2467–74.

7. Huang M A, Greenson J K, Chao C, et al. One-yearintense nutritional counseling results in histological im-provement in patients with non-alcoholic steatohepatitis: apilot study. Am J Gastroenterol 2005; 100: 1072–81.

8. Promrat K, Lutchman G, Uwaifo G I, et al. A pilotstudy of pioglitazone treatment for nonalcoholic steatohe-patitis. Hepatology 2004; 39: 188–96.

9. Bellentani S, Saccoccio G,Masutti F, et al. Prevalenceof and risk factors for hepatic steatosis in Northern Italy.Ann Intern Med 2000; 132: 112–7.

10. Hsiao T J, Chen J C, Wang J D. Insulin resistance andferritin as major determinants of nonalcoholic fatty liverdisease in apparently healthy obese patients. Int J ObesRelat Metab Disord 2004; 28: 167–72.

11. Hamaguchi M, KojimaT, Takeda N, et al. The metabolicsyndrome as a predictor of nonalcoholic fatty liver disease.Ann Intern Med 2005; 143: 722–8.

12. Ludwig J, Viggiano T R, McGill D B, et al. Nonalco-holic steatohepatitis: Mayo Clinic experiences with a

hitherto unnamed disease. Mayo Clin Proc 1980; 55:434–8.

13. Marchesini G, Bugianesi E, Forlani G, et al. Nonalco-holic fatty liver, steatohepatitis, and the metabolic syn-drome. Hepatology 2003; 37: 917–23.

14. Iwasaki T, Nakajima A,Yoneda M, et al. Serum ferritinis associated with visceral fat area and subcutaneous fatarea. Diabetes Care 2005; 28: 2486–91.

15. Sumida Y, Nakashima T, Yoh T, et al. Serum thiored-oxin levels as a predictor of steatohepatitis in patientswith nonalcoholic fatty liver disease. J Hepatol 2003; 38:32–8.

16. Tamura S, Shimomura I. Contribution of adipose tissueand de novo lipogenesis to nonalcoholic fatty liver disease.J Clin Invest 2005; 115: 1139–42.

17. Hui J M, Hodge A, Farrell G C, et al. Beyond insulinresistance in NASH: TNF-a or adiponectin? Hepatology2004; 40: 46–54.

18. Abiru S, Migita K, Maeda Y, et al. Serum cytokine andsoluble cytokine receptor levels in patients with non-alco-holic steatohepatitis. Liver Int 2006; 26: 39–45.

19. Ratziu V, Charlotte F, Heurtier A, et al. Samplingvariability of liver biopsy in nonalcoholic fatty liver disease.Gastroenterology 2005; 128: 1898–906.

20. Tanaka N, Ichijo T, Okiyama W, et al. Laparoscopicfindings in patients with nonalcoholic steatohepatitis. LiverInt 2006; 26: 32–8.

963

Predictors of NASH with mild steatosis