- Email: [email protected]
Development and validation of a new histological score for pediatric non-alcoholic fatty liver disease
Research Article
Development and validation of a new histological score for pediatric non-alcoholic fatty liver disease Naim Alkhouri1,2,⇑, Rita De Vito4, Anna Alisi4, Lisa Yerian2, Rocio Lopez3, Ariel E. Feldstein5, Valerio Nobili4 1 Department of Pediatric Gastroenterology, Cleveland Clinic, Cleveland, OH 44195, United States; 2Digestive Disease Institute, Cleveland Clinic, Cleveland, OH 44195, United States; 3Quantitative Health Sciences, Cleveland Clinic, Cleveland, OH 44195, United States; 4Liver Unit, ‘‘Bambino Gesù’’ Children’s Hospital and Research Institute, Rome, Italy; 5Department of Pediatric Gastroenterology, Rady Children’s Hosptial, University of California San Diego, San Diego, CA 92123, United States
Background & Aims: Pediatric non-alcoholic fatty liver disease (NAFLD) may present with a distinct histopathological pattern characterized by the presence of predominant portal-based injury and portal inflammation (PI). We aimed at developing a new grading score for pediatric NAFLD to be used in clinical trials that takes into account the presence of PI and the weight of histological features. Methods: Our training set consisted of 203 children with biopsyproven NAFLD. The diagnosis of non-alcoholic steatohepatitis (NASH) was based on Brunt’s criteria. Histological features were scored: steatosis (0–3), lobular inflammation (0–3), ballooning (0–2), and PI (0–2). Logistic regression analysis was performed to apply weight to each feature. The new score was called the Pediatric NAFLD Histological Score or PNHS. The validation set consisted of 100 children with NAFLD. Results: The mean age of the initial cohort was 12.4 ± 3.4 years and significant fibrosis (fibrosis stage P2) was present in 26 patients (12.8%). NASH was diagnosed in 135 patients with a mean NAS of 4.5 ± 1.4. The mean PNHS in the NASH group was 89 ± 20.5 compared to 21.9 ± 24.5 in the ‘‘not NASH’’ group, p <0.001. PNHS correlated with the presence of NASH according to the pathologist’s diagnosis, better than the NAFLD activity score (NAS), p = 0.011. The area under the ROC curve (AUC) for the diagnosis of NASH was 0.96 for PNHS. Similar findings were observed in the validation set with an AUC of 0.94.
Keywords: Non-alcoholic fatty liver disease; Non-alcoholic steatohepatitis; Histological score; Children; Portal inflammation. Received 5 March 2012; received in revised form 13 June 2012; accepted 16 July 2012; available online 4 August 2012 ⇑ Corresponding author. Address: Department of Pediatric Gastroenterology, Cleveland Clinic Foundation, 9500 Euclid Avenue, Cleveland, OH 44195, United States. Tel.: +1 216 444 9000; fax: +1 216 444 2974. E-mail address: [email protected] (N. Alkhouri). Abbreviations: NAFLD, non-alcoholic fatty liver disease; PI, portal inflammation; NASH, non-alcoholic steatohepatitis; PNHS, pediatric NAFLD histological score; NAS, NAFLD activity score; TG, triglycerides; BMI, body mass index; MetS, metabolic syndrome; WC, waist circumference; BP, blood pressure; IGT, impaired glucose tolerance; IR, insulin resistance; HOMA, homeostatic model assessment; GGT, gamma glutamyl transferase; INR, international normalization ratio; ALT, alanine aminotransferase; AST, aspartate aminotransferase; NPV, negative predictive value; PPV, positive predictive value.
Conclusions: PNHS may be used for histological grading of pediatric NAFLD with excellent correlation with the presence of NASH. Ó 2012 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved.
Introduction Pediatric NAFLD encompasses a wide spectrum of disease ranging from hepatic steatosis characterized by hepatic lipid accumulation in the form of triglyceride (TG), to non-alcoholic steatohepatitis (NASH) characterized by the association of lipid accumulation with evidence of hepatocyte injury, inflammation, and various degrees of fibrosis to advanced fibrosis and cirrhosis [1–3]. Currently, a liver biopsy remains the gold standard to distinguish NASH from simple steatosis and establish the extent of liver damage and fibrosis [4]. Since the initial description by Ludwig et al., the histological criteria for diagnosing NASH have been evolving and several grading and staging systems have been proposed to assess the histologic severity of the disease [5–7]. The NAFLD activity score (NAS) was developed by the National Institute of Diabetes and Digestive and Kidney Diseases NASH Clinical Research Network (CRN) [8]. The score is based on the classification proposed earlier by Brunt et al. [5] and consists of the unweighted sum of scores for each of the following lesions: steatosis, hepatocyte ballooning, and lobular inflammation. This semiquantitative score was specifically developed for the assessment of response to therapeutic intervention. In children, it was based on a total of 18 cases that were each read once by nine pathologists for a total of 162 sets of observations [8]. Since the development of the original NAS, it has become evident that pediatric NASH may present with a distinct histopathological pattern characterized mainly by the presence of portal-based disease including portal inflammation (PI) [9–11]. Indeed, we and others have demonstrated that PI may represent an important feature of more severe disease in children with NASH [9,11,12]. In addition, some histological features are more indicative of the presence of NASH and should carry more weight when developing a histological score to identify patients for enrollment in therapeutic trials. For example, hepatocyte
Journal of Hepatology 2012 vol. 57 j 1312–1318
JOURNAL OF HEPATOLOGY ballooning is a major distinguishing feature of NASH that confers an increased risk of disease progression [13]. The importance of ballooning has been emphasized by the results of the two largest randomized trials for NAFLD treatment in adults and children, namely the PIVENS and TONIC trials [14,15]. As a result of these issues, when NAS is applied to the pediatric population, typically, only about half of patients can be categorized into a clear-cut pattern while the other half fall into the ‘‘borderline’’ category, supporting the need for a more reproducible scoring system to interpret liver histology in pediatric cases of NAFLD. This score should allow for a better separation of cases into clear-cut meaningful categories that would greatly facilitate inclusion and risk stratification of patients in therapeutic and diagnostic trials. Thus, the aim of this study was to develop and validate a new grading score for pediatric NAFLD that takes into account the presence of PI and the weight of each histological feature and to examine the level of agreement between this score and the pathologist diagnostic determination of NASH.
Materials and methods Study population A total of 303 consecutive children with biopsy-proven NAFLD were included in this study. The study was approved by the Ethics Committee of the Bambino Gesù Children’s Hospital and Research Institute, Rome, Italy. Informed consent was obtained from each patient or responsible guardian. Inclusion criteria were persistently elevated serum aminotransferase levels, diffusely hyperechogenic liver on ultrasonography suggestive of fatty liver, and biopsy consistent with the diagnosis of NAFLD. Exclusion criteria were hepatic virus infections (hepatitis A, B, C, D, and E, cytomegalovirus and Epstein–Barr virus), alcohol consumption, history of parenteral nutrition, and use of drugs known to induce steatosis (e.g., valproate, amiodarone or prednisone) or to affect body weight and carbohydrate metabolism. Autoimmune liver disease, metabolic liver disease, Wilson’s disease, and a-1-antitrypsin-associated liver disease were ruled out using standard clinical, laboratory and histological criteria. The body mass index (BMI) and its standard deviation score (Z-score) were calculated [16,17]. The metabolic syndrome (MetS) was defined as the presence of P3 of the following five criteria [18]: abdominal obesity as defined by a waist circumference (WC) P90th percentile for age [19]; hypertriglyceridemia as defined by TG >95th percentile for age and sex [20]; low HDL cholesterol as defined by <5th percentile for age and sex [20]; elevated blood pressure (BP) as defined by systolic or diastolic BP >95th percentile for age and sex [21]; and impaired fasting glucose, impaired glucose tolerance (IGT) or known type 2 diabetes mellitus as described in detail elsewhere [22]. The degree of insulin resistance (IR) was determined by the homeostatic model assessment (HOMA-IR) using the formula: IR = (insulin glucose)/22.5. Liver histology The clinical indication for biopsy was either to assess the presence of NASH and the degree of fibrosis or other likely independent or competing liver diseases. A liver biopsy was performed in all children, after an overnight fast, using an automatic core biopsy 18 gauge needle (Biopince, Amedic, Sweden) under general anesthesia and ultrasound guidance. The length of liver specimen (in millimeters) was recorded. Only samples that were not fragmented with a length of 15 mm and including at least six complete portal tracts were considered adequate for the purpose of the study. Biopsies were routinely processed (i.e., formalin-fixed and paraffin-embedded) and sections of liver tissue, 5-mm thick, were stained with hematoxylin–eosin, Van Gieson, periodic acid-Schiff diastase, and Prussian blue stain. Biopsies were evaluated by a single expert pediatric hepatopathologist (RDV) who established the histopathological diagnosis of NASH. Based on this categorization, patients were divided into two groups: ‘‘NASH’’ and diagnosis not compatible with NASH or ‘‘not NASH’’. Liver biopsy features for each case were also graded according to the NAFLD activity scoring (NAS) system proposed by Kleiner et al. [8]. Furthermore, individual histological features of NAFLD were scored as follows: steatosis (0–3), lobular inflammation (0–3), ballooning (0–2), and portal
inflammation (0–2; 0 = no PI, 1 = mild PI, 2 = more than mild). Mild PI was defined as a few mononuclear cells, usually, in more than one portal tract. More than mild PI was defined as one portal area showing moderate to marked density of inflammation, and/or the presence of lymphoid aggregates as proposed by Brunt et al. [23]. Fibrosis was scored as 0 – none; 1 – periportal or perisinusoidal fibrosis; 2 – perisinusoidal and portal/periportal fibrosis; 3 – bridging fibrosis; and 4 – cirrhosis. To assess for intra-observer variation, 40 liver biopsies were read and scored again by the same hepatopathologist (R.D.V.) in a blinded manner. For external validation of our histological score, we included liver biopsies from a group of children with NAFLD recruited at the Cleveland Clinic Children’s Hospital for a previous study. All biopsies were read and scored by a second hepatopathologist (L.Y.). Statistical analysis The data set was randomly split; two-thirds of the subjects were used for the main analysis (training set) and the remaining one-third were used to validate the proposed cut points for diagnosis of NASH (validation set). Descriptive statistics were computed for all variables. These include means, standard deviations and percentiles for continuous variables and frequencies and percentages for categorical factors. A univariable analysis was done to assess differences between subjects with and without NASH; Student’s t-tests or non-parametric Wilcoxon rank sum tests were used to compare continuous and ordinal factors and Pearson’s Chi-square tests were used to compare categorical variables. Logistic regression analysis was performed to assess whether applying weights to the individual histological features improved the prediction of NASH based on pathologist’s diagnosis. Receiver operating characteristics (ROC) analysis was done to assess the utility of histological scores for diagnosis of NASH. Areas under the ROC curve (AUC) and corresponding 95% confidence intervals are reported. Cut points that maximized the sensitivity and specificity were chosen. The validation group was used to assess the validity of the proposed cut points; sensitivity, specificity, negative and positive predictive values are presented. The level of intra-observer variation was evaluated using the kappa index, called j, which excludes chanceexpected agreement. A p <0.05 was considered statistically significant. All analyses were performed using SAS (version 9.2, The SAS Institute, Cary, NC) and R (version 2.13.1, The R Foundation for Statistical Computing, Vienna, Austria).
Results Patient characteristics The main clinical and laboratory characteristics of the initial set of patients (n = 203) used to develop the new histological score are summarized in Table 1. The median age of the patients at the time of diagnosis was around 12 years and was not different between patients with or without NASH. Subjects with NASH were found to have significantly greater WC, higher WC percentile, WC/height ratio, total cholesterol, triglyceride levels and gamma glutamyl transferase (GGT), lower total bilirubin and international normalization rate (INR) and were more likely to have MetS. The prevalence of insulin resistance as measured by HOMA-IR and IGT/diabetes was not different between the two groups. Interestingly, although alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels were elevated in both groups, there was no difference between ALT and AST levels in patients with or without NASH. Development of the pediatric NAFLD histological score (PNHS) A description of the main histological findings for the initial set of patients is shown in Table 2. As expected, subjects with NASH had significantly higher histologic scores than those without NASH. Lobular inflammation was of mild severity (grade 1) in most patients; however, 26% of patients with NASH had grade 2 or 3 lobular inflammation. Ballooning was present in 66% of patients with NASH with 41% showing many ballooned hepatocytes (grade
Journal of Hepatology 2012 vol. 57 j 1312–1318
1313
Research Article Table 1. Demographic and clinical characteristics of subjects.
Factor
p value
Male (%) Age at diagnosis (yr) Obese (%) BMI (kg/m2) BMI percentile Waist circumference (cm)
NASH (n = 135) 42 (31.1) 12.4 ± 3.2 123 (91.1) 26.4 ± 4.6 97.0 [94.0, 98.0] 92.0 ± 12.1
Not NASH (n = 68) 29 (42.6) 12.2 ± 3.2 59 (86.8) 25.7 ± 3.3 97.0 [92.0, 98.0] 88.6 ± 10.0
0.1 0.65 0.34 0.27 0.35 0.046
WC percentile WC/height ratio Cholesterol (mg/dl) Triglycerides (mg/dl) Low HDL (mg/dl) High triglycerides (%) Hypertension (%) HOMA-IR IGT/diabetes (%) MetS (%) ALT (IU/L) AST (IU/L) GGT (IU/L) Total bilirubin (mg/dl) Albumin (g/dl) INR
97.0 [90.0, 97.0] 0.60 ± 0.04 159.8 ± 31.7 108.9 ± 59.3 77 (57.0) 87 (64.4) 39 (28.9) 2.8 ± 1.8 59 (43.7) 86 (63.7) 70.0 [48.0, 94.0] 56.7 ± 28.5 27.0 [20.0, 40.0] 0.64 ± 0.24 4.6 ± 0.60 1.10 ± 0.20
90.0 [89.5, 97.0] 0.58 ± 0.05 150.6 ± 24.8 80.9 ± 30.2 30 (44.1) 35 (51.5) 19 (27.9) 3.0 ± 2.9 29 (42.6) 26 (38.2) 65.0 [49.5, 82.5] 50.6 ± 17.3 23.5 [17.0, 32.5] 0.72 ± 0.21 4.6 ± 0.45 1.2 ± 0.25
<0.001 0.012 0.037 <0.001 0.082 0.075 0.89 0.62 0.89 <0.001 0.24 0.11 0.038 0.019 0.57 0.006
Values are presented as mean ± SD with t test; median [P25, P75] with Wilcoxon rank sum test, or N (%) with Pearson’s Chi-square test.
2 ballooning). In patients in the ‘‘not NASH’’ group, only 4% had grade 1 ballooning and none had grade 2. As expected, PI was a common finding in children with NAFLD with 75% of those with NASH and 44% of those without NASH having some degree of PI on biopsy. Fibrosis was prevalent in NASH patients (84%) but it was mostly stage 1 and none of our patients had cirrhosis or stage 4 fibrosis. Interface hepatitis or other features suggestive of autoimmune hepatitis were not present in any of our biopsies. Consistent with previous reports, the median NAS in those with NASH diagnosis was 4.5 ± 1.4 with 45% of patients falling in the indeterminate ‘‘borderline’’ category (NAS = 3–4). Logistic regression analysis revealed that applying weights to the individual histological features improved the prediction of NASH based on pathologist’s diagnosis. PNHS was calculated by using the weighted sum of steatosis, ballooning, lobular inflammation and portal inflammation according to the following formula: PNHS ¼ 100 expðzPNHS Þ=½1 þ expðzPNHS Þ where zPNHS = 8.4 + 2.5 steatosis + 3.5 ballooning + 3.4 lobular inflammation + 0.87 portal inflammation. This new score can be easily calculated by entering the individual histological features on liver biopsy [steatosis (1–3), ballooning (0–2), lobular inflammation (0–3), and PI (0–2)] into the following website: http://rcc.simpal.com/RCEval.cgi?RCID= RPCxtv#Result. PNHS was remarkably higher in the NASH group compared to the ‘‘not NASH’’ group (89 ± 20.5 vs. 21.9 ± 24.5, p <0.001).
1314
Correlation between PNHS and pathologist’s diagnosis of NASH Based on the ROC analysis, we selected the cut-off value with the best sensitivity, specificity, as well as negative predictive value (NPV) and positive predictive value (PPV), for predicting the presence of NASH according to the pathologist’s diagnosis. The PNHS cut-off value of 85 had a sensitivity of 77%, a specificity of 97%, an NPV of 68%, and a PPV of 98% for the prediction of NASH (compared to a sensitivity of 47%, specificity of 100%, NPV of 49%, PPV of 100% for NAS of 5). Fig. 1 shows the relationship between PNHS and the diagnostic category (NASH and ‘‘not NASH’’). Cases with PNHS >85 were almost all consistent with a diagnosis of NASH. The overall accuracy of PNHS P85 for diagnosing NASH is excellent with an AUC of 0.96 (Fig. 2A). Fig. 3 shows liver biopsy samples from three children with NAFLD: patient 1 from the ‘‘not NASH’’ group is diagnosed with simple steatosis with a steatosis grade of 3, NAS of 3 and PNHS of 28.5; patient 2 is diagnosed with NASH (steatosis grade 3, ballooning grade 2, lobular inflammation grade 2, and PI grade 2), NAS of 7 and PNHS of 100; patient 3 is diagnosed with NASH according to the pathologist’s diagnosis (steatosis grade 2, ballooning grade 1, no lobular inflammation, PI grade 2), a NAS of 3, and a PNHS of 86.1. The last example represents a common scenario in pediatric NAFLD with the absence of lobular inflammation, yet the presence of significant PI and a pathologist’s diagnosis of NASH. This clearly demonstrates the additional benefit of using PNHS over NAS to predict the presence of NASH.
Journal of Hepatology 2012 vol. 57 j 1312–1318
JOURNAL OF HEPATOLOGY Not NASH (n = 68)
Steatosis
<0.001
<5%
0 (0.0)
1 (1.5)
5-33%
21 (15.6)
49 (72.1)
34-65%
61 (45.2)
17 (25.0)
≥66%
53 (39.3)
1 (1.5)
3 (2.2)
10 (14.7)
<2 under 20x
97 (71.9)
58 (85.3)
2-4 under 20x
34 (25.2)
0 (0.0)
>4 under 20x
1 (0.74)
0 (0.0)
None
33 (24.4) 86 (63.7)
29 (42.6)
More than mild
16 (11.9)
1 (1.5)
Ballooning 46 (34.1)
65 (95.6)
Few
34 (25.2)
3 (4.4)
Many
55 (40.7)
0 (0.0)
Fibrosis 0
22 (16.3)
52 (76.5)
91 (67.4)
12 (17.6)
2
10 (7.4)
3 (4.4)
3
20 0 20
40 60 PNHS
80
B
0.6 0.4 0.2
PNHS (0.960) 0
0.2 0.4 0.6 0.8 1.0 1-Specificity
1 (1.5)
4.5 ± 1.4
2.2 ± 0.59
<0.001
PNHS
89.0 ± 20.5
21.9 ± 24.5
<0.001
0.6 0.4 0.2
PNHS (0.941) 0
0.2 0.4 0.6 0.8 1.0 1-Specificity
Fig. 2. PNHS for diagnosis of NASH. ROC analysis for (A) the training set and (B) the validation set.
A
12 (8.9)
1.0
0.0
0.0
NAS
100
0.8
0.8
<0.001
1
NASH Not NASH
40
1.0
<0.001
None
60
A
38 (55.9)
Mild
80
Fig. 1. Fraction of biopsies with NASH and ‘‘not NASH’’ according to PNHS. Liver biopsies with PNHS of >85 were almost all consistent with NASH according to the pathologist’s diagnosis.
<0.001
Portal inflammation
100
0
<0.001
Lobular inflammation None
p value
Sensitivity
NASH (n = 135)
Sensitivity
Factor
Predicted probability of NASH
Table 2. Histological features.
B
C
Values are presented as mean ± SD with t test or N (%) with Wilcoxon rank sum test for histology scores. NAS, NAFLD activity score; PNHS, pediatric NAFLD histology score; PNFHS, pediatric NAFLD fibrosis histology score.
Correlation between PNHS and clinical features and the presence of fibrosis PNHS did not correlate with the age of the patient at diagnosis or the degree of insulin resistance as assessed by HOMA-IR (Table 3). However, there was a positive and statistically significant correlation between PNHS and WC, WC percentile, triglycerides levels, and the presence of MetS. More importantly, there was a strong correlation between PNHS and the presence of fibrosis on liver biopsy (rho = 0.54, confidence interval of 0.44–0.63, p <0.001). Interestingly, there was no correlation between ALT values and PNHS (rho = 0.06, confidence interval of 0.06, 0.17, p = 0.31). Validation of PNHS There were no significant differences in any of the histologic features or the prevalence of NASH between the training and validation sets (Table 4). The median PNHS in the validation set (both NASH and ‘‘not NASH’’ patients) was similar to the median PNHS in the initial training set (69.3 ± 37 in the validation set and 66.5 ± 38.5 in the training set, p = 0.54). A PNHS of 85 gave a
400x
400x
400x
Fig. 3. Liver biopsy samples from three children with NAFLD. Biopsy specimens were formalin-fixed, paraffin embedded, and stained with H&E. (A) Represents a child with simple steatosis without evidence of hepatocyte ballooning or inflammation, his PNHS was calculated to be 28.5. (B) Represents a child with NASH (steatosis grade 3, ballooning grade 2, lobular inflammation grade 2, and portal inflammation grade 2) and his PNHS was calculated to be 100. (C) Represents a child with NASH according to the pathologist’s diagnosis with portal-based inflammation, no lobular inflammation and PNHS of 86.1. (This figure appears in color on the web.)
sensitivity of 76% and a specificity of 91% for the diagnosis of steatohepatitis compared to a sensitivity of 50% and a specificity of 100% for a NAS of 5. The AUC for PNHS P85 in the validation set was 0.94 (Fig. 2B). The level of intra-observer variation was evaluated in a subset of 40 liver biopsies using the kappa index (j). The intra-observer agreement was found to be excellent with j of 0.92 (0.8, 1) for steatosis, j of 0.91 (0.78, 1) for lobular inflammation, j of 0.96 (0.89, 1) for PI, and j of 1 for ballooning and diagnosis of NASH.
Journal of Hepatology 2012 vol. 57 j 1312–1318
1315
Research Article Table 3. Correlation between PNHS and clinical features and the presence of fibrosis.
Factor Age at diagnosis Waist circumference WC percentile HOMA-IR Triglycerides MetS ALT Fibrosis
Weighted PNHS rho (95% CI) p value 0.04 (-0.07, 0.15) 0.47 0.21 (0.10, 0.32) <0.001 0.34 (0.23, 0.44) <0.001 0.07 (-0.04, 0.18) 0.24 0.34 (0.23, 0.45) <0.001 0.25 (0.14, 0.36) <0.001 0.06 (-0.06, 0.17) 0.31 0.54 (0.44, 0.63) <0.001
rho, Spearman’s correlation coefficient; CI, confidence interval.
Table 4. Comparison of training and validation sets.
Factor
Training set (n = 203)
Validation (n = 100)
1 (0.49)
1 (1.0)
Steatosis <5%
0.47
5-33%
70 (34.5)
32 (32.0)
34-65%
78 (38.4)
35 (35.0)
≥66%
54 (26.6)
32 (32.0)
None
13 (6.4)
11 (11.0)
<2 under 20x
155 (76.4)
75 (75.0)
2-4 under 20x
34 (16.7)
14 (14.0)
>4 under 20x
1 (0.49)
0 (0.0)
71 (35.0)
34 (34.0)
Inflammation
0.19
Portal inflammation None
0.75
Mild
115 (56.7)
62 (62.0)
More than mild
17 (8.4)
4 (4.0)
111 (54.7)
50 (50.0)
Ballooning None
p value
0.65
Few
37 (18.2)
24 (24.0)
Many
55 (27.1)
26 (26.0)
0
74 (36.5)
36 (36.0)
1
103 (50.7)
51 (51.0)
2
13 (6.4)
6 (6.0)
3
13 (6.4)
7 (7.0)
Fibrosis
0.93
NAS
3.7 ± 1.6
3.8 ± 1.6
0.91
PNHS
66.5 ± 38.5
69.3 ± 37.0
0.54
NASH
135 (66.5)
67 (67.0)
0.93
We also performed external validation of PNHS in a small group of children with biopsy-proven NAFLD (n = 22) recruited at the Cleveland Clinic Children’s Hospital. In this group, 10 children had NASH and 12 were classified as ‘‘not NASH’’ according to the pathologist’s diagnosis. The PNHS had an excellent accuracy in predicting the presence of NASH, with 10/10 children correctly classified as NASH and 12/12 children correctly classified as ‘‘not
1316
NASH’’. Mild PI was present in 18 biopsies, more than mild PI in one biopsy and no PI in three biopsies. Interestingly, when we applied NAS to this group, 9/22 children (41%) were classified as borderline NASH.
Discussion The main finding of our study relates to the development and validation of a new histological score, the PNHS, to categorize NAFLD into NASH and ‘‘not NASH’’ for inclusion in pediatric clinical trials. Avoiding the ‘‘borderline NASH’’ category will allow for a better classification of children into groups with or without NASH. Our results demonstrated that: (1) there is a high level of agreement between categorization of NAFLD cases using PNHS and the pathologist’s diagnosis, (2) a PNHS of P85 can be considered as a cut-off value for inclusion in trials with sensitivity of 76–77% and specificity of 91–97%, (3) PNHS correlates with the presence of metabolic syndrome and of fibrosis on liver biopsy. NAFLD has become a growing epidemic in the pediatric population. Within the spectrum of NAFLD, hepatic steatosis appears to follow a benign non-progressive course, whereas NASH may progress to cirrhosis in 15–20% of patients [24–26], which makes the differentiation between the two entities of high clinical importance. Furthermore, a diagnosis of NASH warrants a more aggressive approach to lifestyle modifications and is a prerequisite for inclusion in most clinical trials. Although recent studies that have attempted to non-invasively identify patients with NASH have yielded encouraging results [27,28], liver biopsy is currently the only reliable way of differentiating hepatic steatosis from NASH and is the recommended method for assessment of disease activity for phase 2 and 3 clinical trials [29]. The main histological features of NASH include the presence of >5% large droplet steatosis, ballooning degeneration of hepatocytes, and mixed lobular inflammatory infiltrates [30]. Atypical features, such as predominantly portal inflammation and fibrosis, have been described and are common in the pediatric population [9,11,12]. A recent study has demonstrated that portal chronic inflammation was also seen in adult patients with concurrent chronic liver disease or following successful treatment intervention [23]. NAS includes all the main histological feature of NASH (steatosis, ballooning and lobular inflammation); however, NAS as numeric value is not intended to replace the pathologist’s diagnosis of NASH [8]. Results of recent large randomized clinical trials for treatment of NAFLD in both adults and children [14,15] have made it clear that NAS has some serious limitations that needed to be addressed with a new scoring system. First of all, the histologic spectrum of NAFLD in children is more variable than their adult counterparts and PI has been consistently shown to be a common finding on liver biopsy from pediatric patients, which makes it reasonable to include PI as part of any new scoring system for this population. Indeed, the TONIC trial, which is the largest pediatric treatment intervention trial to date, did not report the effect of therapy with vitamin E or metformin on portal inflammation [14]. Second, not all histological features of NASH are equal in terms of their relative significance for the diagnosis of steatohepatitis. Ballooning is more indicative of the presence of NASH and in the adult PIVENS trial, a score of at least one for hepatocellular ballooning was required in all cases for inclusion. Third, the significance of improvement of various features of NAS on the risk of
Journal of Hepatology 2012 vol. 57 j 1312–1318
JOURNAL OF HEPATOLOGY developing liver-related mortality and cirrhosis is unknown. For example, it has been known for a while that as the disease progresses to advanced fibrosis and cirrhosis, the amount of steatosis tends to decrease [31]. Therefore, an improvement in steatosis grade, which can influence NAS in a significant manner, can be a sign of disease progression instead of improvement. Furthermore, in the TONIC trial, ballooning was the only histological feature that significantly improved in both the vitamin E and metformin arms and it seems that this was the reason behind the improvement in NAS and the highest percentage of NASH resolution in the vitamin E arm. Perhaps ballooning should be given higher significance in a disease activity score, which was provided by the weighing process that we used to develop PNHS. Most experts agree that resolution of NASH with no worsening of fibrosis is a desirable primary end point for treatment trials [29]. Our study has several strengths including the inclusion of the largest number to date of consecutive children with biopsy-proven NAFLD with the full spectrum of disease ranging from simple steatosis to NASH and advanced fibrosis; all biopsies were reviewed by a single expert liver pathologist; and sampling issues associated with the quality of liver biopsies were excluded by not including any biopsy that was fragmented, <1.5 cm in length or had <6 portal tracts. Limitations to our study include referral bias as our patients were seen at a large tertiary care medical center and the biopsies were evaluated by an expert hepatopathologist, which may limit the applicability of our results in communitybased practices with less experienced pathologists. Second, most of our children were white, thus, PNHS should be validated in other ethnic groups such as Hispanic and black children. In addition, we did not evaluate the effect of inter-observer variability of histopathologic interpretation of liver biopsy on our results. We cannot emphasize enough that PNHS was developed to be used for clinical trials and is not intended to replace pathologist’s diagnosis. Until more data are available about its reproducibility and correlation with disease progression, the presence of NASH should rely on overall assessment of the individual histologic features by a pathologist. In conclusion, the results of our study demonstrated a high level of agreement between the categorization of NAFLD cases using PNHS and the pathologist’s diagnostic determination and thus support the use of PNHS in future therapeutic trials in children. In the situation where NAS and the score for each of its individual components are available, PNHS is simply calculated by giving a score from 0 to 2 for PI and then using the formula provided in this study. Therefore, validating this score in previously performed intervention trials, such as TONIC, can be easily done and may reveal some interesting data.
Financial support Supported by grants from ‘‘Bambino Gesù’’ Children’s Hospital and Research Institute, Rome, Italy.
Conflict of interest The authors who have taken part in this study declared that they do not have anything to disclose regarding funding or conflict of interest with respect to this manuscript.
References [1] Angulo P. Non-alcoholic fatty liver disease. N Eng J Med 2002;346: 1221–1231. [2] Brunt EM, Tiniakos DG. Pathological features of NASH. Front Biosci 2005;10:1475–1484. [3] Feldstein AE, Charatcharoenwitthaya P, Treeprasertsuk S, Benson JT, Enders FB, Angulo P. The natural history of non-alcoholic fatty liver disease in children: a follow-up study for up to 20 years. Gut 2009;58:1538–1544. [4] Wieckowska A, Feldstein AE. Diagnosis of non-alcoholic fatty liver disease: invasive versus noninvasive. Semin Liver Dis 2008;28:386–395. [5] Brunt EM, Janney CG, Di Bisceglie AM, Neuschwander-Tetri BA, Bacon BR. Non-alcoholic steatohepatitis: a proposal for grading and staging the histological lesions. Am J Gastroenterol 1999;94:2467–2474. [6] Ludwig J, Viggiano TR, McGill DB, Oh BJ. Non-alcoholic steatohepatitis: Mayo Clinic experiences with a hitherto unnamed disease. Mayo Clin Proc 1980;55:434–438. [7] Mendler MH, Kanel G, Govindarajan S. Proposal for a histological scoring and grading system for non-alcoholic fatty liver disease. Liver Int 2005;25: 294–304. [8] Kleiner DE, Brunt EM, Van Natta M, Behling C, Contos MJ, Cummings OW, et al. Design and validation of a histological scoring system for non-alcoholic fatty liver disease. Hepatology (Baltimore, MD) 2005;41:1313–1321. [9] Carter-Kent C, Yerian LM, Brunt EM, Angulo P, Kohli R, Ling SC, et al. Nonalcoholic steatohepatitis in children: a multicenter clinicopathological study. Hepatology (Baltimore, MD) 2009;50:1113–1120. [10] Patton H, Lavine JE, Van Natta ML, Schwimmer JB, Kleiner D, Molleston J. Clinical correlates of histopathology in pediatric non-alcoholic steatohepatitis. Gastroenterology 2008. [11] Schwimmer JB, Behling C, Newbury R, Deutsch R, Nievergelt C, Schork NJ, et al. Histopathology of pediatric non-alcoholic fatty liver disease. Hepatology (Baltimore, MD) 2005;42:641–649. [12] Nobili V, Marcellini M, Devito R, Ciampalini P, Piemonte F, Comparcola D, et al. NAFLD in children: a prospective clinical–pathological study and effect of lifestyle advice. Hepatology (Baltimore, MD) 2006;44:458–465. [13] Caldwell S, Ikura Y, Dias D, Isomoto K, Yabu A, Moskaluk C, et al. Hepatocellular ballooning in NASH. J Hepatol 2010;53:719–723. [14] Lavine JE, Schwimmer JB, Van Natta ML, Molleston JP, Murray KF, Rosenthal P, et al. Effect of vitamin E or metformin for treatment of non-alcoholic fatty liver disease in children and adolescents: the TONIC randomized controlled trial. JAMA 2011;305:1659–1668. [15] Sanyal AJ, Chalasani N, Kowdley KV, McCullough A, Diehl AM, Bass NM, et al. Pioglitazone, vitamin E, or placebo for non-alcoholic steatohepatitis. N Engl J Med 2010;362:1675–1685. [16] Kuczmarski RJ, Ogden CL, Grummer-Strawn LM, Flegal KM, Guo SS, Wei R, et al. CDC growth charts: United States. Adv Data 2000:1–27. [17] Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ (Clinical Research Ed) 2000;320:1240–1243. [18] Boney CM, Verma A, Tucker R, Vohr BR. Metabolic syndrome in childhood: association with birth weight, maternal obesity, and gestational diabetes mellitus. Pediatrics 2005;115:e290–e296. [19] Fernandez JR, Redden DT, Pietrobelli A, Allison DB. Waist circumference percentiles in nationally representative samples of African–American, European–American, and Mexican–American children and adolescents. J Pediatr 2004;145:439–444. [20] American Academy of Pediatrics. National cholesterol education program: report of the expert panel on blood cholesterol levels in children and adolescents. Pediatrics 1992;89:525–584. [21] Report of the Second Task Force on Blood Pressure Control in Children – 1987. Task Force on Blood Pressure Control in Children. National Heart, Lung, and Blood Institute, Bethesda, Maryland. Pediatrics 1987;79:1–25. [22] Genuth S, Alberti KG, Bennett P, Buse J, Defronzo R, Kahn R, et al. Follow-up report on the diagnosis of diabetes mellitus. Diabetes Care 2003;26: 3160–3167. [23] Brunt EM, Kleiner DE, Wilson LA, Unalp A, Behling CE, Lavine JE, et al. Portal chronic inflammation in non-alcoholic fatty liver disease (NAFLD): a histologic marker of advanced NAFLD-clinicopathologic correlations from the non-alcoholic steatohepatitis clinical research network. Hepatology (Baltimore, MD) 2009;49:809–820. [24] Adams LA, Lymp JF, St Sauver J, Sanderson SO, Lindor KD, Feldstein A, et al. The natural history of non-alcoholic fatty liver disease: a population-based cohort study. Gastroenterology 2005;129:113–121.
Journal of Hepatology 2012 vol. 57 j 1312–1318
1317
Research Article [25] Ekstedt M, Franzen LE, Mathiesen UL, Thorelius L, Holmqvist M, Bodemar G, et al. Long-term follow-up of patients with NAFLD and elevated liver enzymes. Hepatology (Baltimore, MD) 2006;44:865–873. [26] Matteoni CA, Younossi ZM, Gramlich T, Boparai N, Liu YC, McCullough AJ. Non-alcoholic fatty liver disease: a spectrum of clinical and pathological severity. Gastroenterology 1999;116:1413–1419. [27] Wieckowska A, Zein NN, Yerian LM, Lopez AR, McCullough AJ, Feldstein AE, et al. In vivo assessment of liver cell apoptosis as a novel biomarker of disease severity in non-alcoholic fatty liver disease. Hepatology (Baltimore, MD) 2006;44:27–33.
1318
[28] Tamimi TA, Berk MP, Alkhouri N, Yerian LM, Lopez R, Zein NN, et al. An apoptosis panel for non-alcoholic steatohepatitis diagnosis. Gastroenterology 2009;136:A89. [29] Sanyal AJ, Brunt EM, Kleiner DE, Kowdley KV, Chalasani N, Lavine JE, et al. Endpoints and clinical trial design for non-alcoholic steatohepatitis. Hepatology (Baltimore, MD) 2011;54:344–353. [30] Yeh MM, Brunt EM. Pathology of non-alcoholic fatty liver disease. Am J Clin Pathol 2007;128:837–847. [31] Caldwell SH, Lee VD, Kleiner DE, Al-Osaimi AM, Argo CK, Northup PG, et al. NASH and cryptogenic cirrhosis: a histological analysis. Ann Hepatol 2009;8:346–352.
Journal of Hepatology 2012 vol. 57 j 1312–1318
Development and validation of a new histological score for pediatric non-alcoholic fatty liver disease Naim Alkhouri1,2,⇑, Rita De Vito4, Anna Alisi4, Lisa Yerian2, Rocio Lopez3, Ariel E. Feldstein5, Valerio Nobili4 1 Department of Pediatric Gastroenterology, Cleveland Clinic, Cleveland, OH 44195, United States; 2Digestive Disease Institute, Cleveland Clinic, Cleveland, OH 44195, United States; 3Quantitative Health Sciences, Cleveland Clinic, Cleveland, OH 44195, United States; 4Liver Unit, ‘‘Bambino Gesù’’ Children’s Hospital and Research Institute, Rome, Italy; 5Department of Pediatric Gastroenterology, Rady Children’s Hosptial, University of California San Diego, San Diego, CA 92123, United States
Background & Aims: Pediatric non-alcoholic fatty liver disease (NAFLD) may present with a distinct histopathological pattern characterized by the presence of predominant portal-based injury and portal inflammation (PI). We aimed at developing a new grading score for pediatric NAFLD to be used in clinical trials that takes into account the presence of PI and the weight of histological features. Methods: Our training set consisted of 203 children with biopsyproven NAFLD. The diagnosis of non-alcoholic steatohepatitis (NASH) was based on Brunt’s criteria. Histological features were scored: steatosis (0–3), lobular inflammation (0–3), ballooning (0–2), and PI (0–2). Logistic regression analysis was performed to apply weight to each feature. The new score was called the Pediatric NAFLD Histological Score or PNHS. The validation set consisted of 100 children with NAFLD. Results: The mean age of the initial cohort was 12.4 ± 3.4 years and significant fibrosis (fibrosis stage P2) was present in 26 patients (12.8%). NASH was diagnosed in 135 patients with a mean NAS of 4.5 ± 1.4. The mean PNHS in the NASH group was 89 ± 20.5 compared to 21.9 ± 24.5 in the ‘‘not NASH’’ group, p <0.001. PNHS correlated with the presence of NASH according to the pathologist’s diagnosis, better than the NAFLD activity score (NAS), p = 0.011. The area under the ROC curve (AUC) for the diagnosis of NASH was 0.96 for PNHS. Similar findings were observed in the validation set with an AUC of 0.94.
Keywords: Non-alcoholic fatty liver disease; Non-alcoholic steatohepatitis; Histological score; Children; Portal inflammation. Received 5 March 2012; received in revised form 13 June 2012; accepted 16 July 2012; available online 4 August 2012 ⇑ Corresponding author. Address: Department of Pediatric Gastroenterology, Cleveland Clinic Foundation, 9500 Euclid Avenue, Cleveland, OH 44195, United States. Tel.: +1 216 444 9000; fax: +1 216 444 2974. E-mail address: [email protected] (N. Alkhouri). Abbreviations: NAFLD, non-alcoholic fatty liver disease; PI, portal inflammation; NASH, non-alcoholic steatohepatitis; PNHS, pediatric NAFLD histological score; NAS, NAFLD activity score; TG, triglycerides; BMI, body mass index; MetS, metabolic syndrome; WC, waist circumference; BP, blood pressure; IGT, impaired glucose tolerance; IR, insulin resistance; HOMA, homeostatic model assessment; GGT, gamma glutamyl transferase; INR, international normalization ratio; ALT, alanine aminotransferase; AST, aspartate aminotransferase; NPV, negative predictive value; PPV, positive predictive value.
Conclusions: PNHS may be used for histological grading of pediatric NAFLD with excellent correlation with the presence of NASH. Ó 2012 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved.
Introduction Pediatric NAFLD encompasses a wide spectrum of disease ranging from hepatic steatosis characterized by hepatic lipid accumulation in the form of triglyceride (TG), to non-alcoholic steatohepatitis (NASH) characterized by the association of lipid accumulation with evidence of hepatocyte injury, inflammation, and various degrees of fibrosis to advanced fibrosis and cirrhosis [1–3]. Currently, a liver biopsy remains the gold standard to distinguish NASH from simple steatosis and establish the extent of liver damage and fibrosis [4]. Since the initial description by Ludwig et al., the histological criteria for diagnosing NASH have been evolving and several grading and staging systems have been proposed to assess the histologic severity of the disease [5–7]. The NAFLD activity score (NAS) was developed by the National Institute of Diabetes and Digestive and Kidney Diseases NASH Clinical Research Network (CRN) [8]. The score is based on the classification proposed earlier by Brunt et al. [5] and consists of the unweighted sum of scores for each of the following lesions: steatosis, hepatocyte ballooning, and lobular inflammation. This semiquantitative score was specifically developed for the assessment of response to therapeutic intervention. In children, it was based on a total of 18 cases that were each read once by nine pathologists for a total of 162 sets of observations [8]. Since the development of the original NAS, it has become evident that pediatric NASH may present with a distinct histopathological pattern characterized mainly by the presence of portal-based disease including portal inflammation (PI) [9–11]. Indeed, we and others have demonstrated that PI may represent an important feature of more severe disease in children with NASH [9,11,12]. In addition, some histological features are more indicative of the presence of NASH and should carry more weight when developing a histological score to identify patients for enrollment in therapeutic trials. For example, hepatocyte
Journal of Hepatology 2012 vol. 57 j 1312–1318
JOURNAL OF HEPATOLOGY ballooning is a major distinguishing feature of NASH that confers an increased risk of disease progression [13]. The importance of ballooning has been emphasized by the results of the two largest randomized trials for NAFLD treatment in adults and children, namely the PIVENS and TONIC trials [14,15]. As a result of these issues, when NAS is applied to the pediatric population, typically, only about half of patients can be categorized into a clear-cut pattern while the other half fall into the ‘‘borderline’’ category, supporting the need for a more reproducible scoring system to interpret liver histology in pediatric cases of NAFLD. This score should allow for a better separation of cases into clear-cut meaningful categories that would greatly facilitate inclusion and risk stratification of patients in therapeutic and diagnostic trials. Thus, the aim of this study was to develop and validate a new grading score for pediatric NAFLD that takes into account the presence of PI and the weight of each histological feature and to examine the level of agreement between this score and the pathologist diagnostic determination of NASH.
Materials and methods Study population A total of 303 consecutive children with biopsy-proven NAFLD were included in this study. The study was approved by the Ethics Committee of the Bambino Gesù Children’s Hospital and Research Institute, Rome, Italy. Informed consent was obtained from each patient or responsible guardian. Inclusion criteria were persistently elevated serum aminotransferase levels, diffusely hyperechogenic liver on ultrasonography suggestive of fatty liver, and biopsy consistent with the diagnosis of NAFLD. Exclusion criteria were hepatic virus infections (hepatitis A, B, C, D, and E, cytomegalovirus and Epstein–Barr virus), alcohol consumption, history of parenteral nutrition, and use of drugs known to induce steatosis (e.g., valproate, amiodarone or prednisone) or to affect body weight and carbohydrate metabolism. Autoimmune liver disease, metabolic liver disease, Wilson’s disease, and a-1-antitrypsin-associated liver disease were ruled out using standard clinical, laboratory and histological criteria. The body mass index (BMI) and its standard deviation score (Z-score) were calculated [16,17]. The metabolic syndrome (MetS) was defined as the presence of P3 of the following five criteria [18]: abdominal obesity as defined by a waist circumference (WC) P90th percentile for age [19]; hypertriglyceridemia as defined by TG >95th percentile for age and sex [20]; low HDL cholesterol as defined by <5th percentile for age and sex [20]; elevated blood pressure (BP) as defined by systolic or diastolic BP >95th percentile for age and sex [21]; and impaired fasting glucose, impaired glucose tolerance (IGT) or known type 2 diabetes mellitus as described in detail elsewhere [22]. The degree of insulin resistance (IR) was determined by the homeostatic model assessment (HOMA-IR) using the formula: IR = (insulin glucose)/22.5. Liver histology The clinical indication for biopsy was either to assess the presence of NASH and the degree of fibrosis or other likely independent or competing liver diseases. A liver biopsy was performed in all children, after an overnight fast, using an automatic core biopsy 18 gauge needle (Biopince, Amedic, Sweden) under general anesthesia and ultrasound guidance. The length of liver specimen (in millimeters) was recorded. Only samples that were not fragmented with a length of 15 mm and including at least six complete portal tracts were considered adequate for the purpose of the study. Biopsies were routinely processed (i.e., formalin-fixed and paraffin-embedded) and sections of liver tissue, 5-mm thick, were stained with hematoxylin–eosin, Van Gieson, periodic acid-Schiff diastase, and Prussian blue stain. Biopsies were evaluated by a single expert pediatric hepatopathologist (RDV) who established the histopathological diagnosis of NASH. Based on this categorization, patients were divided into two groups: ‘‘NASH’’ and diagnosis not compatible with NASH or ‘‘not NASH’’. Liver biopsy features for each case were also graded according to the NAFLD activity scoring (NAS) system proposed by Kleiner et al. [8]. Furthermore, individual histological features of NAFLD were scored as follows: steatosis (0–3), lobular inflammation (0–3), ballooning (0–2), and portal
inflammation (0–2; 0 = no PI, 1 = mild PI, 2 = more than mild). Mild PI was defined as a few mononuclear cells, usually, in more than one portal tract. More than mild PI was defined as one portal area showing moderate to marked density of inflammation, and/or the presence of lymphoid aggregates as proposed by Brunt et al. [23]. Fibrosis was scored as 0 – none; 1 – periportal or perisinusoidal fibrosis; 2 – perisinusoidal and portal/periportal fibrosis; 3 – bridging fibrosis; and 4 – cirrhosis. To assess for intra-observer variation, 40 liver biopsies were read and scored again by the same hepatopathologist (R.D.V.) in a blinded manner. For external validation of our histological score, we included liver biopsies from a group of children with NAFLD recruited at the Cleveland Clinic Children’s Hospital for a previous study. All biopsies were read and scored by a second hepatopathologist (L.Y.). Statistical analysis The data set was randomly split; two-thirds of the subjects were used for the main analysis (training set) and the remaining one-third were used to validate the proposed cut points for diagnosis of NASH (validation set). Descriptive statistics were computed for all variables. These include means, standard deviations and percentiles for continuous variables and frequencies and percentages for categorical factors. A univariable analysis was done to assess differences between subjects with and without NASH; Student’s t-tests or non-parametric Wilcoxon rank sum tests were used to compare continuous and ordinal factors and Pearson’s Chi-square tests were used to compare categorical variables. Logistic regression analysis was performed to assess whether applying weights to the individual histological features improved the prediction of NASH based on pathologist’s diagnosis. Receiver operating characteristics (ROC) analysis was done to assess the utility of histological scores for diagnosis of NASH. Areas under the ROC curve (AUC) and corresponding 95% confidence intervals are reported. Cut points that maximized the sensitivity and specificity were chosen. The validation group was used to assess the validity of the proposed cut points; sensitivity, specificity, negative and positive predictive values are presented. The level of intra-observer variation was evaluated using the kappa index, called j, which excludes chanceexpected agreement. A p <0.05 was considered statistically significant. All analyses were performed using SAS (version 9.2, The SAS Institute, Cary, NC) and R (version 2.13.1, The R Foundation for Statistical Computing, Vienna, Austria).
Results Patient characteristics The main clinical and laboratory characteristics of the initial set of patients (n = 203) used to develop the new histological score are summarized in Table 1. The median age of the patients at the time of diagnosis was around 12 years and was not different between patients with or without NASH. Subjects with NASH were found to have significantly greater WC, higher WC percentile, WC/height ratio, total cholesterol, triglyceride levels and gamma glutamyl transferase (GGT), lower total bilirubin and international normalization rate (INR) and were more likely to have MetS. The prevalence of insulin resistance as measured by HOMA-IR and IGT/diabetes was not different between the two groups. Interestingly, although alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels were elevated in both groups, there was no difference between ALT and AST levels in patients with or without NASH. Development of the pediatric NAFLD histological score (PNHS) A description of the main histological findings for the initial set of patients is shown in Table 2. As expected, subjects with NASH had significantly higher histologic scores than those without NASH. Lobular inflammation was of mild severity (grade 1) in most patients; however, 26% of patients with NASH had grade 2 or 3 lobular inflammation. Ballooning was present in 66% of patients with NASH with 41% showing many ballooned hepatocytes (grade
Journal of Hepatology 2012 vol. 57 j 1312–1318
1313
Research Article Table 1. Demographic and clinical characteristics of subjects.
Factor
p value
Male (%) Age at diagnosis (yr) Obese (%) BMI (kg/m2) BMI percentile Waist circumference (cm)
NASH (n = 135) 42 (31.1) 12.4 ± 3.2 123 (91.1) 26.4 ± 4.6 97.0 [94.0, 98.0] 92.0 ± 12.1
Not NASH (n = 68) 29 (42.6) 12.2 ± 3.2 59 (86.8) 25.7 ± 3.3 97.0 [92.0, 98.0] 88.6 ± 10.0
0.1 0.65 0.34 0.27 0.35 0.046
WC percentile WC/height ratio Cholesterol (mg/dl) Triglycerides (mg/dl) Low HDL (mg/dl) High triglycerides (%) Hypertension (%) HOMA-IR IGT/diabetes (%) MetS (%) ALT (IU/L) AST (IU/L) GGT (IU/L) Total bilirubin (mg/dl) Albumin (g/dl) INR
97.0 [90.0, 97.0] 0.60 ± 0.04 159.8 ± 31.7 108.9 ± 59.3 77 (57.0) 87 (64.4) 39 (28.9) 2.8 ± 1.8 59 (43.7) 86 (63.7) 70.0 [48.0, 94.0] 56.7 ± 28.5 27.0 [20.0, 40.0] 0.64 ± 0.24 4.6 ± 0.60 1.10 ± 0.20
90.0 [89.5, 97.0] 0.58 ± 0.05 150.6 ± 24.8 80.9 ± 30.2 30 (44.1) 35 (51.5) 19 (27.9) 3.0 ± 2.9 29 (42.6) 26 (38.2) 65.0 [49.5, 82.5] 50.6 ± 17.3 23.5 [17.0, 32.5] 0.72 ± 0.21 4.6 ± 0.45 1.2 ± 0.25
<0.001 0.012 0.037 <0.001 0.082 0.075 0.89 0.62 0.89 <0.001 0.24 0.11 0.038 0.019 0.57 0.006
Values are presented as mean ± SD with t test; median [P25, P75] with Wilcoxon rank sum test, or N (%) with Pearson’s Chi-square test.
2 ballooning). In patients in the ‘‘not NASH’’ group, only 4% had grade 1 ballooning and none had grade 2. As expected, PI was a common finding in children with NAFLD with 75% of those with NASH and 44% of those without NASH having some degree of PI on biopsy. Fibrosis was prevalent in NASH patients (84%) but it was mostly stage 1 and none of our patients had cirrhosis or stage 4 fibrosis. Interface hepatitis or other features suggestive of autoimmune hepatitis were not present in any of our biopsies. Consistent with previous reports, the median NAS in those with NASH diagnosis was 4.5 ± 1.4 with 45% of patients falling in the indeterminate ‘‘borderline’’ category (NAS = 3–4). Logistic regression analysis revealed that applying weights to the individual histological features improved the prediction of NASH based on pathologist’s diagnosis. PNHS was calculated by using the weighted sum of steatosis, ballooning, lobular inflammation and portal inflammation according to the following formula: PNHS ¼ 100 expðzPNHS Þ=½1 þ expðzPNHS Þ where zPNHS = 8.4 + 2.5 steatosis + 3.5 ballooning + 3.4 lobular inflammation + 0.87 portal inflammation. This new score can be easily calculated by entering the individual histological features on liver biopsy [steatosis (1–3), ballooning (0–2), lobular inflammation (0–3), and PI (0–2)] into the following website: http://rcc.simpal.com/RCEval.cgi?RCID= RPCxtv#Result. PNHS was remarkably higher in the NASH group compared to the ‘‘not NASH’’ group (89 ± 20.5 vs. 21.9 ± 24.5, p <0.001).
1314
Correlation between PNHS and pathologist’s diagnosis of NASH Based on the ROC analysis, we selected the cut-off value with the best sensitivity, specificity, as well as negative predictive value (NPV) and positive predictive value (PPV), for predicting the presence of NASH according to the pathologist’s diagnosis. The PNHS cut-off value of 85 had a sensitivity of 77%, a specificity of 97%, an NPV of 68%, and a PPV of 98% for the prediction of NASH (compared to a sensitivity of 47%, specificity of 100%, NPV of 49%, PPV of 100% for NAS of 5). Fig. 1 shows the relationship between PNHS and the diagnostic category (NASH and ‘‘not NASH’’). Cases with PNHS >85 were almost all consistent with a diagnosis of NASH. The overall accuracy of PNHS P85 for diagnosing NASH is excellent with an AUC of 0.96 (Fig. 2A). Fig. 3 shows liver biopsy samples from three children with NAFLD: patient 1 from the ‘‘not NASH’’ group is diagnosed with simple steatosis with a steatosis grade of 3, NAS of 3 and PNHS of 28.5; patient 2 is diagnosed with NASH (steatosis grade 3, ballooning grade 2, lobular inflammation grade 2, and PI grade 2), NAS of 7 and PNHS of 100; patient 3 is diagnosed with NASH according to the pathologist’s diagnosis (steatosis grade 2, ballooning grade 1, no lobular inflammation, PI grade 2), a NAS of 3, and a PNHS of 86.1. The last example represents a common scenario in pediatric NAFLD with the absence of lobular inflammation, yet the presence of significant PI and a pathologist’s diagnosis of NASH. This clearly demonstrates the additional benefit of using PNHS over NAS to predict the presence of NASH.
Journal of Hepatology 2012 vol. 57 j 1312–1318
JOURNAL OF HEPATOLOGY Not NASH (n = 68)
Steatosis
<0.001
<5%
0 (0.0)
1 (1.5)
5-33%
21 (15.6)
49 (72.1)
34-65%
61 (45.2)
17 (25.0)
≥66%
53 (39.3)
1 (1.5)
3 (2.2)
10 (14.7)
<2 under 20x
97 (71.9)
58 (85.3)
2-4 under 20x
34 (25.2)
0 (0.0)
>4 under 20x
1 (0.74)
0 (0.0)
None
33 (24.4) 86 (63.7)
29 (42.6)
More than mild
16 (11.9)
1 (1.5)
Ballooning 46 (34.1)
65 (95.6)
Few
34 (25.2)
3 (4.4)
Many
55 (40.7)
0 (0.0)
Fibrosis 0
22 (16.3)
52 (76.5)
91 (67.4)
12 (17.6)
2
10 (7.4)
3 (4.4)
3
20 0 20
40 60 PNHS
80
B
0.6 0.4 0.2
PNHS (0.960) 0
0.2 0.4 0.6 0.8 1.0 1-Specificity
1 (1.5)
4.5 ± 1.4
2.2 ± 0.59
<0.001
PNHS
89.0 ± 20.5
21.9 ± 24.5
<0.001
0.6 0.4 0.2
PNHS (0.941) 0
0.2 0.4 0.6 0.8 1.0 1-Specificity
Fig. 2. PNHS for diagnosis of NASH. ROC analysis for (A) the training set and (B) the validation set.
A
12 (8.9)
1.0
0.0
0.0
NAS
100
0.8
0.8
<0.001
1
NASH Not NASH
40
1.0
<0.001
None
60
A
38 (55.9)
Mild
80
Fig. 1. Fraction of biopsies with NASH and ‘‘not NASH’’ according to PNHS. Liver biopsies with PNHS of >85 were almost all consistent with NASH according to the pathologist’s diagnosis.
<0.001
Portal inflammation
100
0
<0.001
Lobular inflammation None
p value
Sensitivity
NASH (n = 135)
Sensitivity
Factor
Predicted probability of NASH
Table 2. Histological features.
B
C
Values are presented as mean ± SD with t test or N (%) with Wilcoxon rank sum test for histology scores. NAS, NAFLD activity score; PNHS, pediatric NAFLD histology score; PNFHS, pediatric NAFLD fibrosis histology score.
Correlation between PNHS and clinical features and the presence of fibrosis PNHS did not correlate with the age of the patient at diagnosis or the degree of insulin resistance as assessed by HOMA-IR (Table 3). However, there was a positive and statistically significant correlation between PNHS and WC, WC percentile, triglycerides levels, and the presence of MetS. More importantly, there was a strong correlation between PNHS and the presence of fibrosis on liver biopsy (rho = 0.54, confidence interval of 0.44–0.63, p <0.001). Interestingly, there was no correlation between ALT values and PNHS (rho = 0.06, confidence interval of 0.06, 0.17, p = 0.31). Validation of PNHS There were no significant differences in any of the histologic features or the prevalence of NASH between the training and validation sets (Table 4). The median PNHS in the validation set (both NASH and ‘‘not NASH’’ patients) was similar to the median PNHS in the initial training set (69.3 ± 37 in the validation set and 66.5 ± 38.5 in the training set, p = 0.54). A PNHS of 85 gave a
400x
400x
400x
Fig. 3. Liver biopsy samples from three children with NAFLD. Biopsy specimens were formalin-fixed, paraffin embedded, and stained with H&E. (A) Represents a child with simple steatosis without evidence of hepatocyte ballooning or inflammation, his PNHS was calculated to be 28.5. (B) Represents a child with NASH (steatosis grade 3, ballooning grade 2, lobular inflammation grade 2, and portal inflammation grade 2) and his PNHS was calculated to be 100. (C) Represents a child with NASH according to the pathologist’s diagnosis with portal-based inflammation, no lobular inflammation and PNHS of 86.1. (This figure appears in color on the web.)
sensitivity of 76% and a specificity of 91% for the diagnosis of steatohepatitis compared to a sensitivity of 50% and a specificity of 100% for a NAS of 5. The AUC for PNHS P85 in the validation set was 0.94 (Fig. 2B). The level of intra-observer variation was evaluated in a subset of 40 liver biopsies using the kappa index (j). The intra-observer agreement was found to be excellent with j of 0.92 (0.8, 1) for steatosis, j of 0.91 (0.78, 1) for lobular inflammation, j of 0.96 (0.89, 1) for PI, and j of 1 for ballooning and diagnosis of NASH.
Journal of Hepatology 2012 vol. 57 j 1312–1318
1315
Research Article Table 3. Correlation between PNHS and clinical features and the presence of fibrosis.
Factor Age at diagnosis Waist circumference WC percentile HOMA-IR Triglycerides MetS ALT Fibrosis
Weighted PNHS rho (95% CI) p value 0.04 (-0.07, 0.15) 0.47 0.21 (0.10, 0.32) <0.001 0.34 (0.23, 0.44) <0.001 0.07 (-0.04, 0.18) 0.24 0.34 (0.23, 0.45) <0.001 0.25 (0.14, 0.36) <0.001 0.06 (-0.06, 0.17) 0.31 0.54 (0.44, 0.63) <0.001
rho, Spearman’s correlation coefficient; CI, confidence interval.
Table 4. Comparison of training and validation sets.
Factor
Training set (n = 203)
Validation (n = 100)
1 (0.49)
1 (1.0)
Steatosis <5%
0.47
5-33%
70 (34.5)
32 (32.0)
34-65%
78 (38.4)
35 (35.0)
≥66%
54 (26.6)
32 (32.0)
None
13 (6.4)
11 (11.0)
<2 under 20x
155 (76.4)
75 (75.0)
2-4 under 20x
34 (16.7)
14 (14.0)
>4 under 20x
1 (0.49)
0 (0.0)
71 (35.0)
34 (34.0)
Inflammation
0.19
Portal inflammation None
0.75
Mild
115 (56.7)
62 (62.0)
More than mild
17 (8.4)
4 (4.0)
111 (54.7)
50 (50.0)
Ballooning None
p value
0.65
Few
37 (18.2)
24 (24.0)
Many
55 (27.1)
26 (26.0)
0
74 (36.5)
36 (36.0)
1
103 (50.7)
51 (51.0)
2
13 (6.4)
6 (6.0)
3
13 (6.4)
7 (7.0)
Fibrosis
0.93
NAS
3.7 ± 1.6
3.8 ± 1.6
0.91
PNHS
66.5 ± 38.5
69.3 ± 37.0
0.54
NASH
135 (66.5)
67 (67.0)
0.93
We also performed external validation of PNHS in a small group of children with biopsy-proven NAFLD (n = 22) recruited at the Cleveland Clinic Children’s Hospital. In this group, 10 children had NASH and 12 were classified as ‘‘not NASH’’ according to the pathologist’s diagnosis. The PNHS had an excellent accuracy in predicting the presence of NASH, with 10/10 children correctly classified as NASH and 12/12 children correctly classified as ‘‘not
1316
NASH’’. Mild PI was present in 18 biopsies, more than mild PI in one biopsy and no PI in three biopsies. Interestingly, when we applied NAS to this group, 9/22 children (41%) were classified as borderline NASH.
Discussion The main finding of our study relates to the development and validation of a new histological score, the PNHS, to categorize NAFLD into NASH and ‘‘not NASH’’ for inclusion in pediatric clinical trials. Avoiding the ‘‘borderline NASH’’ category will allow for a better classification of children into groups with or without NASH. Our results demonstrated that: (1) there is a high level of agreement between categorization of NAFLD cases using PNHS and the pathologist’s diagnosis, (2) a PNHS of P85 can be considered as a cut-off value for inclusion in trials with sensitivity of 76–77% and specificity of 91–97%, (3) PNHS correlates with the presence of metabolic syndrome and of fibrosis on liver biopsy. NAFLD has become a growing epidemic in the pediatric population. Within the spectrum of NAFLD, hepatic steatosis appears to follow a benign non-progressive course, whereas NASH may progress to cirrhosis in 15–20% of patients [24–26], which makes the differentiation between the two entities of high clinical importance. Furthermore, a diagnosis of NASH warrants a more aggressive approach to lifestyle modifications and is a prerequisite for inclusion in most clinical trials. Although recent studies that have attempted to non-invasively identify patients with NASH have yielded encouraging results [27,28], liver biopsy is currently the only reliable way of differentiating hepatic steatosis from NASH and is the recommended method for assessment of disease activity for phase 2 and 3 clinical trials [29]. The main histological features of NASH include the presence of >5% large droplet steatosis, ballooning degeneration of hepatocytes, and mixed lobular inflammatory infiltrates [30]. Atypical features, such as predominantly portal inflammation and fibrosis, have been described and are common in the pediatric population [9,11,12]. A recent study has demonstrated that portal chronic inflammation was also seen in adult patients with concurrent chronic liver disease or following successful treatment intervention [23]. NAS includes all the main histological feature of NASH (steatosis, ballooning and lobular inflammation); however, NAS as numeric value is not intended to replace the pathologist’s diagnosis of NASH [8]. Results of recent large randomized clinical trials for treatment of NAFLD in both adults and children [14,15] have made it clear that NAS has some serious limitations that needed to be addressed with a new scoring system. First of all, the histologic spectrum of NAFLD in children is more variable than their adult counterparts and PI has been consistently shown to be a common finding on liver biopsy from pediatric patients, which makes it reasonable to include PI as part of any new scoring system for this population. Indeed, the TONIC trial, which is the largest pediatric treatment intervention trial to date, did not report the effect of therapy with vitamin E or metformin on portal inflammation [14]. Second, not all histological features of NASH are equal in terms of their relative significance for the diagnosis of steatohepatitis. Ballooning is more indicative of the presence of NASH and in the adult PIVENS trial, a score of at least one for hepatocellular ballooning was required in all cases for inclusion. Third, the significance of improvement of various features of NAS on the risk of
Journal of Hepatology 2012 vol. 57 j 1312–1318
JOURNAL OF HEPATOLOGY developing liver-related mortality and cirrhosis is unknown. For example, it has been known for a while that as the disease progresses to advanced fibrosis and cirrhosis, the amount of steatosis tends to decrease [31]. Therefore, an improvement in steatosis grade, which can influence NAS in a significant manner, can be a sign of disease progression instead of improvement. Furthermore, in the TONIC trial, ballooning was the only histological feature that significantly improved in both the vitamin E and metformin arms and it seems that this was the reason behind the improvement in NAS and the highest percentage of NASH resolution in the vitamin E arm. Perhaps ballooning should be given higher significance in a disease activity score, which was provided by the weighing process that we used to develop PNHS. Most experts agree that resolution of NASH with no worsening of fibrosis is a desirable primary end point for treatment trials [29]. Our study has several strengths including the inclusion of the largest number to date of consecutive children with biopsy-proven NAFLD with the full spectrum of disease ranging from simple steatosis to NASH and advanced fibrosis; all biopsies were reviewed by a single expert liver pathologist; and sampling issues associated with the quality of liver biopsies were excluded by not including any biopsy that was fragmented, <1.5 cm in length or had <6 portal tracts. Limitations to our study include referral bias as our patients were seen at a large tertiary care medical center and the biopsies were evaluated by an expert hepatopathologist, which may limit the applicability of our results in communitybased practices with less experienced pathologists. Second, most of our children were white, thus, PNHS should be validated in other ethnic groups such as Hispanic and black children. In addition, we did not evaluate the effect of inter-observer variability of histopathologic interpretation of liver biopsy on our results. We cannot emphasize enough that PNHS was developed to be used for clinical trials and is not intended to replace pathologist’s diagnosis. Until more data are available about its reproducibility and correlation with disease progression, the presence of NASH should rely on overall assessment of the individual histologic features by a pathologist. In conclusion, the results of our study demonstrated a high level of agreement between the categorization of NAFLD cases using PNHS and the pathologist’s diagnostic determination and thus support the use of PNHS in future therapeutic trials in children. In the situation where NAS and the score for each of its individual components are available, PNHS is simply calculated by giving a score from 0 to 2 for PI and then using the formula provided in this study. Therefore, validating this score in previously performed intervention trials, such as TONIC, can be easily done and may reveal some interesting data.
Financial support Supported by grants from ‘‘Bambino Gesù’’ Children’s Hospital and Research Institute, Rome, Italy.
Conflict of interest The authors who have taken part in this study declared that they do not have anything to disclose regarding funding or conflict of interest with respect to this manuscript.
References [1] Angulo P. Non-alcoholic fatty liver disease. N Eng J Med 2002;346: 1221–1231. [2] Brunt EM, Tiniakos DG. Pathological features of NASH. Front Biosci 2005;10:1475–1484. [3] Feldstein AE, Charatcharoenwitthaya P, Treeprasertsuk S, Benson JT, Enders FB, Angulo P. The natural history of non-alcoholic fatty liver disease in children: a follow-up study for up to 20 years. Gut 2009;58:1538–1544. [4] Wieckowska A, Feldstein AE. Diagnosis of non-alcoholic fatty liver disease: invasive versus noninvasive. Semin Liver Dis 2008;28:386–395. [5] Brunt EM, Janney CG, Di Bisceglie AM, Neuschwander-Tetri BA, Bacon BR. Non-alcoholic steatohepatitis: a proposal for grading and staging the histological lesions. Am J Gastroenterol 1999;94:2467–2474. [6] Ludwig J, Viggiano TR, McGill DB, Oh BJ. Non-alcoholic steatohepatitis: Mayo Clinic experiences with a hitherto unnamed disease. Mayo Clin Proc 1980;55:434–438. [7] Mendler MH, Kanel G, Govindarajan S. Proposal for a histological scoring and grading system for non-alcoholic fatty liver disease. Liver Int 2005;25: 294–304. [8] Kleiner DE, Brunt EM, Van Natta M, Behling C, Contos MJ, Cummings OW, et al. Design and validation of a histological scoring system for non-alcoholic fatty liver disease. Hepatology (Baltimore, MD) 2005;41:1313–1321. [9] Carter-Kent C, Yerian LM, Brunt EM, Angulo P, Kohli R, Ling SC, et al. Nonalcoholic steatohepatitis in children: a multicenter clinicopathological study. Hepatology (Baltimore, MD) 2009;50:1113–1120. [10] Patton H, Lavine JE, Van Natta ML, Schwimmer JB, Kleiner D, Molleston J. Clinical correlates of histopathology in pediatric non-alcoholic steatohepatitis. Gastroenterology 2008. [11] Schwimmer JB, Behling C, Newbury R, Deutsch R, Nievergelt C, Schork NJ, et al. Histopathology of pediatric non-alcoholic fatty liver disease. Hepatology (Baltimore, MD) 2005;42:641–649. [12] Nobili V, Marcellini M, Devito R, Ciampalini P, Piemonte F, Comparcola D, et al. NAFLD in children: a prospective clinical–pathological study and effect of lifestyle advice. Hepatology (Baltimore, MD) 2006;44:458–465. [13] Caldwell S, Ikura Y, Dias D, Isomoto K, Yabu A, Moskaluk C, et al. Hepatocellular ballooning in NASH. J Hepatol 2010;53:719–723. [14] Lavine JE, Schwimmer JB, Van Natta ML, Molleston JP, Murray KF, Rosenthal P, et al. Effect of vitamin E or metformin for treatment of non-alcoholic fatty liver disease in children and adolescents: the TONIC randomized controlled trial. JAMA 2011;305:1659–1668. [15] Sanyal AJ, Chalasani N, Kowdley KV, McCullough A, Diehl AM, Bass NM, et al. Pioglitazone, vitamin E, or placebo for non-alcoholic steatohepatitis. N Engl J Med 2010;362:1675–1685. [16] Kuczmarski RJ, Ogden CL, Grummer-Strawn LM, Flegal KM, Guo SS, Wei R, et al. CDC growth charts: United States. Adv Data 2000:1–27. [17] Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ (Clinical Research Ed) 2000;320:1240–1243. [18] Boney CM, Verma A, Tucker R, Vohr BR. Metabolic syndrome in childhood: association with birth weight, maternal obesity, and gestational diabetes mellitus. Pediatrics 2005;115:e290–e296. [19] Fernandez JR, Redden DT, Pietrobelli A, Allison DB. Waist circumference percentiles in nationally representative samples of African–American, European–American, and Mexican–American children and adolescents. J Pediatr 2004;145:439–444. [20] American Academy of Pediatrics. National cholesterol education program: report of the expert panel on blood cholesterol levels in children and adolescents. Pediatrics 1992;89:525–584. [21] Report of the Second Task Force on Blood Pressure Control in Children – 1987. Task Force on Blood Pressure Control in Children. National Heart, Lung, and Blood Institute, Bethesda, Maryland. Pediatrics 1987;79:1–25. [22] Genuth S, Alberti KG, Bennett P, Buse J, Defronzo R, Kahn R, et al. Follow-up report on the diagnosis of diabetes mellitus. Diabetes Care 2003;26: 3160–3167. [23] Brunt EM, Kleiner DE, Wilson LA, Unalp A, Behling CE, Lavine JE, et al. Portal chronic inflammation in non-alcoholic fatty liver disease (NAFLD): a histologic marker of advanced NAFLD-clinicopathologic correlations from the non-alcoholic steatohepatitis clinical research network. Hepatology (Baltimore, MD) 2009;49:809–820. [24] Adams LA, Lymp JF, St Sauver J, Sanderson SO, Lindor KD, Feldstein A, et al. The natural history of non-alcoholic fatty liver disease: a population-based cohort study. Gastroenterology 2005;129:113–121.
Journal of Hepatology 2012 vol. 57 j 1312–1318
1317
Research Article [25] Ekstedt M, Franzen LE, Mathiesen UL, Thorelius L, Holmqvist M, Bodemar G, et al. Long-term follow-up of patients with NAFLD and elevated liver enzymes. Hepatology (Baltimore, MD) 2006;44:865–873. [26] Matteoni CA, Younossi ZM, Gramlich T, Boparai N, Liu YC, McCullough AJ. Non-alcoholic fatty liver disease: a spectrum of clinical and pathological severity. Gastroenterology 1999;116:1413–1419. [27] Wieckowska A, Zein NN, Yerian LM, Lopez AR, McCullough AJ, Feldstein AE, et al. In vivo assessment of liver cell apoptosis as a novel biomarker of disease severity in non-alcoholic fatty liver disease. Hepatology (Baltimore, MD) 2006;44:27–33.
1318
[28] Tamimi TA, Berk MP, Alkhouri N, Yerian LM, Lopez R, Zein NN, et al. An apoptosis panel for non-alcoholic steatohepatitis diagnosis. Gastroenterology 2009;136:A89. [29] Sanyal AJ, Brunt EM, Kleiner DE, Kowdley KV, Chalasani N, Lavine JE, et al. Endpoints and clinical trial design for non-alcoholic steatohepatitis. Hepatology (Baltimore, MD) 2011;54:344–353. [30] Yeh MM, Brunt EM. Pathology of non-alcoholic fatty liver disease. Am J Clin Pathol 2007;128:837–847. [31] Caldwell SH, Lee VD, Kleiner DE, Al-Osaimi AM, Argo CK, Northup PG, et al. NASH and cryptogenic cirrhosis: a histological analysis. Ann Hepatol 2009;8:346–352.
Journal of Hepatology 2012 vol. 57 j 1312–1318