OBESITY, INSULIN RESISTANCE, AND OTHER CLINICOPATHOLOGICAL
CORRELATES OF PEDIATRIC NONALCOHOLIC FATTY LIVER DISEASE
JEFFREY B. SCHWIMMER,MD,REENA DEUTSCH,PHD, JEFFREY B. RAUCH,BA,CYNTHIA BEHLING,MD,ROBERT NEWBURY,MD,
AND JOEL E. LAVINE,MD,PHD
Objective To describe the clinical characteristics of nonalcoholic fatty liver disease (NAFLD) in children, including insulin
resistance, and to test for correlation with liver pathology.
Study design A retrospective review of children with biopsy-proven NAFLD at Children’s Hospital San Diego from 1999 to
2002. Liver biopsy specimens were independently reviewed by two pathologists.
Results Children with NAFLD (n = 43) were mostly male (70%), Hispanic American (53%) and obese (88%). The criteria
for insulin resistance were met by 95% of subjects. Steatosis was predicted by the combination of quantitative insulin sensitivity
check index, age, and ethnicity (P < .0001). Portal inflammation was predicted by the combination of ALT and fasting insulin
(P = .0009). Perisinusoidal fibrosis was predicted by the combination of AST, fasting insulin, and BMI Z score (P < .0001).
Portal fibrosis was predicted by the combination of right upper quadrant pain and homeostasis model assessment of insulin
resistance (P = .0028).
Conclusions We identified significant predictors of liver pathology in children with NAFLD. Children being evaluated for
NAFLD should be screened for insulin resistance, which is nearly universal and correlates with liver histology. (J Pediatr
2003;143:500-5)
T
hroughout the past 30 years, the prevalence of childhood obesity in the United
States has tripled from 5% to 15%.
1
In children, obesity is the major risk for the
features of Syndrome X, the metabolic link between obesity, insulin resistance, and
health-related consequences.
2
Furthermore, in adults, nonalcoholic fatty liver disease
(NAFLD) is considered to be the hepatic manifestation of Syndrome X.
3,4
NAFLD in
children is recognized globally, with reports from Asia,
5
Europe,
6
Australia,
7
and North
America.
8,9
Pediatric NAFLD is likely increasing in prevalence, at least in the United
States, in parallel with the prevalence of obesity.
10
NAFLD represents a spectrum of
conditions characterized by macrovesicular hepatic steatosis and little or no exposure
to ethanol. The liver pathology encompasses a range from isolated fatty liver to
steatohepatitis, advanced fibrosis, and cirrhosis. Nonalcoholic steatohepatitis (NASH)
may progress to cirrhosis even in children,
11
whereas isolated fatty liver is believed to have
a low likelihood for progression.
12
The distinction between fatty liver and NASH currently
can be made only by liver biopsy.
13,14
In adults, insulin resistance is regarded as an essential factor for the development of
NAFLD.
15-17
In children, NAFLD occurs most commonly in conditions associated with
insulin resistance, including obesity and diabetes mellitus.
8,9
Insulin resistance in children
with biopsy-proven NAFLD has not been assessed. In addition, clinical correlates of liver
histology features in pediatric NAFLD have not been tested. In adults with NAFLD,
clinical correlates related to histology include age, sex, obesity, diabetes mellitus, elevated
From the Joint Program in Pediatric
Gastroenterology, Hepatology, and
Nutrition, Department of Family and
Preventive Medicine, Department of
Pathology, the School of Medicine,
University of California-San Diego,
and Children’s Hospital and Health
Center, San Diego, California.
Supported in part by grant M01
RR00827 from the National Center
for Research Resources of the Na-
tional Institutes of Health for the
General Clinical Research Center at
University of California-San Diego.
Presented in abstract form at the 53rd
Annual Meeting of the American
Association for the Study of Liver
Diseases, Boston, Massachusetts,
November 5, 2002.
Submitted for publication Oct 16, 2002;
revisions received Mar 5, 2003, and May
2, 2003; accepted May 29, 2003.
Reprint requests: Joel E. Lavine, MD,
PhD, University of California, San
Diego, 200 West Arbor Dr, San Diego,
CA 92103-8450. E-mail: jolavine@ALT Alanine aminotransferase NAFLD Nonalcoholic fatty liver disease
AST Aspartate aminotransferase NASH Nonalcoholic steatohepatitis
BMI Body mass index QUICKI Quantitative insulin sensitivity check index
HOMA-IR Homeostasis model assessment of insulin resistance S
I
Insulin sensitivity
I
0
Fasting insulin level
ucsd.edu.
Copyrightª 2003 Mosby, Inc. All rights
reserved.
0022-3476/2003/$30.00 + 0
10.1067/S0022-3476(03)00325-1
500

triglycerides, and insulin resistance.
18-22
No consensus exists investigations including the fasting insulin level (I
0
), theas to whether serum aminotransferase values correlate with
histologic features. Furthermore, the validity of applying adult
data to pediatric NAFLD has not been demonstrated. We
examined the clinical features of pediatric NAFLD in the
context of histologic findings to explore predictors of liver
histology.
METHODS
Definition of NAFLD
The diagnosis of NAFLD is first suspected because of
a persistent elevation of serum alanine aminotransferase
(ALT) (>35 U/L) or in the setting of an echogenic liver
detected by ultrasound. Diagnosis requires exclusion of other
causes of chronic hepatitis including hepatitis B, hepatitis C,
a-1 antitrypsin deficiency, autoimmune hepatitis, Wilson’s
disease, drug toxicity (valproate, tetracycline, methotrexate,
amiodorone, or prednisone) and total parenteral nutrition.
Alcohol intake is excluded on the basis of reported history.
Definitive diagnosis requires confirmation with liver bi-
opsy.
13,14
Along with the clinical components as detailed,
the minimum criterion for the diagnosis of NAFLD is
macrovesicular steatosis filling >5% of the hepatocytes.
Subjects and Clinical Data Collection
A computerized database identified consecutive subjects
<18 years of age with biopsy-proven NAFLD, as defined
above, at Children’s Hospital San Diego, from January 1999 to
March 2002. Charts for all subjects meeting inclusion criteria
were reviewed retrospectively for age, sex, race, weight, height,
and the presence of right upper quadrant pain, acanthosis
nigricans, hepatomegaly, or diabetes mellitus. Results of
laboratory tests obtained <3 months before liver biopsy were
recorded (serum aspartate aminotransferase [AST], ALT,
alkaline phosphatase, gamma-glutamyl transferase, fasting
insulin and glucose, triglyceride, cholesterol, high-density
lipoprotein cholesterol, low-density lipoprotein cholesterol).
Body mass index (BMI) was calculated as the weight (kg)
divided by the height (m
2
). To compare BMI across different
ages and both sexes, BMI Z score was calculated. The Z score
represents the number of standard deviations above or below
the United States population mean value based on standard-
ized tables for children. The protocol was approved by the
institutional review boards of the University of California San
Diego, and Children’s Hospital and Health Center, San
Diego.
Insulin Sensitivity
Insulin sensitivity (S
I
) may vary widely and is influenced
by age, ethnicity, and obesity. Impaired S
I,
termed insulin
resistance, has been defined as the S
I
of the lowest 10% of
a nonobese, nondiabetic population.
23
The most accurate
technique for measuring S
I
, the euglycemic-hyperinsulinemic
clamp study, is invasive and labor-intensive. Several less
cumbersome models have been used in numerous clinicalObesity, Insulin Resistance, and Other Clinicopathological
Correlates of Pediatric Nonalcoholic Fatty Liver Diseasehomeostasis model assessment of insulin resistance (HOMA-
IR), and the quantitative insulin sensitivity check index
(QUICKI).
24,25
All three have been shown to correlate with
euglycemic-hyperinsulinemic clamp studies and have been
used previously in children and adolescents.
26,27
HOMA-IR,
used in previous studies of adult NAFLD,
28
was calculated as
follows: HOMA IR = (fasting insulin [lU/mL])(fasting
glucose [mmol/L]/22.5); insulin resistance is defined as
HOMA-IR >2.
25
QUICKI was calculated as: QUICKI = 1/
[(log[fasting insulin (lU/mL) + log(fasting glucose [mg/
dL])]; impaired S
I
is defined as QUICKI <0.339.
29
Whereas
HOMA-IR and QUICKI are algebraically related, the
information each provides may be unique.
Pathology
Two experienced pathologists unaware of the clinical
data independently evaluated the liver biopsies for features of
steatohepatitis. The formalin-fixed, paraffin-embedded tissue
was stained with hematoxylin and eosin, Masson’s trichrome,
and periodic-acid Schiff reagent. Steatosis was quantified as
a percentage of hepatocytes containing macrovesicular fat and
expressed categorically as mild (5-33%), moderate (33-66%),
or severe (>66%). The presence or absence of perisinusoidal
fibrosis was noted. Portal inflammation and fibrosis were
quantified using a modified Knodell score.
30
Data Analysis
Means, standard deviations, and percentages were
calculated for descriptive summary statistics. Univariate
relations between the clinical and pathologic variables were
explored using Pearson v
2
tests, Fisher exact tests, and logistic
regression.
Distributional assumptions were verified, and non-
parametric methods were used where appropriate. Patterns
of association were identified from the preliminary analyses,
and potential combinations were developed in models using
variables from the univariate analyses with observed signifi-
cance levels of <0.10. Recognizing that multiple testing may
create an increased probability of falsely finding significant
results, exact P values are used instead of making adjustments
for multiple comparisons. Reported P values must therefore
be interpreted with caution. Only one variable from com-
binations of correlated or related variables, using Spearman
correlation tests, analysis of variance, independent Student t
tests, or Wilcoxon rank-sum tests, were considered when
applying rigorous statistical significance testing. Logistic
regression was used for testing the final models. Statistical
significance was defined as a P value <.05.
RESULTS
Study Population
Clinical and laboratory details for the 43 subjects
identified with NAFLD are shown in Table I. The median501