S ps
Li
VL EUR
ER C
LID
*S que,
Ho , Fran
Ba
(N
he
the
stu
an
Me
cu
ag
se
the
ag
ste
ba
for
ma
dis
ing
24
ne
no
or
fib
on
sta
systematically lower than sampling variability and there-
fore could not account for most of the sampling error.
Conclusions: Histologic lesions of NASH are unevenly dis-
trib
pli
no
T
cau
con
or
me
eit
un
de
inf
NA
ati
lec
tiv
ser
all
liv
sho
tol
sam
ica
bio
of
the
NA
de
wi
liv
sta
Materials and Methods
Selection Criteria
GASTROENTEROLOGY 2005;128:1898–1906uted throughout the liver parenchyma; therefore, sam-
ng error of liver biopsy can result in substantial misdiag-
sis and staging inaccuracies.
he diagnosis and staging of nonalcoholic fatty liver
disease (NAFLD) rely heavily on liver biopsy
1,2
be-
se no reliable serum or genetic diagnostic tests for
ditions such as nonalcoholic steatohepatitis (NASH)
bland steatosis are yet available. Alternative diagnostic
thods such as imaging techniques or serum markers
her do not have enough discriminating value,
3
are
Consecutive patients with a suspected diagnosis of
NAFLD were prospectively enrolled in this study. The diag-
nosis of NAFLD was based on the following: (1) a bright
pattern of the liver on abdominal ultrasound, (2) elevated
serum transaminase levels, and (3) no other identifiable causes
of liver disease. Other causes of liver disease were ruled out by
appropriate work-up: these included daily alcohol consump-
tion greater than 30 g for men and 20 g for women; drug-
induced hepatotoxicity; infection with HBV or HCV viruses;
Abbreviations used in this paper: NAFLD, nonalcoholic fatty liver
disease; NASH, nonalcoholic steatohepatitis.ampling Variability of Liver Bio
ver Disease
AD RATZIU,* FRÉDÉRIC CHARLOTTE,
‡
AGNÈS H
IC BRUCKERT,


ANDRÉ GRIMALDI,
§
FRÉDÉRIQUE
O Study Group
ervice d’Hépatogastroentérologie,
‡
Laboratoire d’Anatomie Pathologi
spitalier Pitié Salpêtrière, and Université Pierre et Marie Curie, Paris
ckground & Aims: In nonalcoholic fatty liver disease
AFLD), the distinction between steatosis and steato-
patitis (NASH) and the assessment of the severity of
disease rely on liver histology alone. The aim of this
dy was to assess the sampling error of liver biopsy
d its impact on the diagnosis and staging of NASH.
thods: Fifty-one patients with NAFLD underwent per-
taneous liver biopsy with 2 samples collected. The
reement between paired biopsy specimens was as-
ssed by the percentage of discordant results and by

reliability test. Results: No features displayed high
reement; substantial agreement was only seen for
atosis grade; moderate agreement for hepatocyte
llooning and perisinusoidal fibrosis; fair agreement
Mallory bodies; acidophilic bodies and lobular inflam-
tion displayed only slight agreement. Overall, the
cordance rate for the presence of hepatocyte balloon-
was 18%, and ballooning would have been missed in
% of patients had only 1 biopsy been performed. The
gative predictive value of a single biopsy for the diag-
sis of NASH was at best 0.74. Discordance of 1 stage
more was 41%. Six of 17 patients with bridging
rosis (35%) on 1 sample had only mild or no fibrosis
the other and therefore could have been under
ged with only 1 biopsy. Intraobserver variability wasable to stage the disease process,
3
or await further
velopment.
4
Liver biopsy provides not only criticaly in Nonalcoholic Fatty
TIER,
§
SOPHIE GOMBERT,


PHILIPPE GIRAL,


APRON,
‡
and THIERRY POYNARD* for the
§
Service de Diabétologie, and


Service d’Endocrinologie, Groupe
ce
ormation regarding the diagnosis and the prognosis of
FLD but also patient selection and end-point evalu-
on of therapeutic clinical trials.
5,6
The basic assumption that the small fragment col-
ted through percutaneous liver biopsy is representa-
e of overall hepatic involvement has been, however,
iously challenged. First, a needle biopsy sample usu-
y represents around 1/50,000 of the total mass of the
er.
7,8
Second, multiple studies have been published
wing considerable sampling variability for most his-
ogic features including cirrhosis when more than 1
ple is analyzed.
9–15
This variability can have signif-
nt influence in the diagnostic performance of liver
psy specimens,
12,16
as well as in the staging or grading
hepatic disease.
13
So far, there have been no studies in the literature on
sampling variability of percutaneous liver biopsy in
FLD using comprehensive and up-to-date pathologic
scriptions. Therefore, we undertook the present study
th the aim of assessing the variability of percutaneous
er biopsy findings and its impact on diagnosis and
ging of the disease.? 2005 by the American Gastroenterological Association
0016-5085/05/$30.00
doi:10.1053/j.gastro.2005.03.084

gen
aut
ros
liv
ma
ste
ult
tie
we
(Br
per
the
anc
sec
to
con
usi
fib
thi
1 t
the
on
wh
the
slid
for
of
his
inv
In
in
sys
tor
toc
ma
sca
cyt
ma
mi
ma
ter
to
low
foc
ext
bri
we
sco
ste
ph
(ly
tis
eve
bu
sam
of
ass
ter
ste
wi
2 d
all
as
per
0.6
era
bel
bet
sam
the
the
in
and
A
Ta
Ste
S
He
Ma
Fib
E
D
Por
Inte
Lob
Aci
June 2005 SAMPLING VARIABILITY OF LIVER BIOPSY IN NAFLD 1899etic hemochromatosis; autoimmune liver disease including
oimmune hepatitis, primary biliary cirrhosis, primary scle-
ing cholangitis; biliary gallstones; vascular diseases of the
er;

1
-antitrypsin deficiency; Wilson’s disease; and abnor-
l serum thyroid hormone levels. One patient did not have
atosis on liver biopsy despite a bright hepatic pattern on
rasound and was excluded from this study.
Liver Biopsy Procedure and Pathologic
Examination
After providing written informed consent, every pa-
nt underwent a liver biopsy, during which 2 liver samples
re extracted using a 16-gauge Hepafix Luer Lock needle
aun Melsungen, Melsungen, Germany). Both biopsies were
formed by the same investigator (V.R.) in the right lobe of
liver through an intercostal route using ultrasound guid-
e. The first passage aimed at the tip of the xyphoid, and the
ond was dorsally reclined by an angle of approximately 30°
45°. The liver specimens were each placed in a separate
tainer, fixed in formalin, paraffin-embedded, and stained
ng hematein-eosin saffron, and picro-sirius red-hematein for
rosis.
17
Only samples longer than 15 mm were retained for
s study. Once all biopsy samples were collected, a code from
o 102 was given randomly to each slide. The 102 slides were
n read by an experienced liver pathologist (F.C.), with the
ly information available being the random code. To assess
ether intraobserver variability might account for some of
variability between different samples, a random set of 50
es was resubmitted to the same pathologist 1 month later
pathologic assessment.
Pathologic Assessment
The size in millimeters of each biopsy and the number
portal tracts were recorded. Although a consensus for the
tologic diagnosis of NASH is currently lacking,
2
most
estigators use the scoring system proposed by Brunt et al.
1
the present study, the pathologic assessment was performed
2 ways. First, we scored all biopsy samples using the scoring
tem proposed by Brunt et al.
1
Briefly, the necroinflamma-
y activity was graded as follows: grade 0: absent, no hepa-
yte ballooning, no lobular inflammation, no portal inflam-
tion; grade 1: mild, occasional ballooned hepatocytes,
ttered and mild lobular inflammation (whether lympho-
es or polymorphonuclear cells), no or mild portal inflam-
tion; grade 2: moderate, obvious hepatocyte ballooning,
ld lobular inflammation, mild to moderate portal inflam-
tion; grade 3: severe, marked hepatocyte ballooning, scat-
ed lobular inflammation and polymorphonuclear cells, mild
moderate portal inflammation. Fibrosis was staged as fol-
s: stage 0: no fibrosis; stage 1: zone 3 perisinusoidal fibrosis,
al or extensive; stage 2: perisinusoidal fibrosis, focal or
ensive, and periportal fibrosis, focal or extensive; stage 3:
dging fibrosis, focal or extensive; grade 4: cirrhosis. Second,analyzed elementary histologic lesions that compose the
ring system proposed by Brunt et al,
1
which are as follows:
atosis, perisinusoidal fibrosis, hepatocyte ballooning, acido-
sta
for
Coilic bodies (ie, intralobular necrosis), lobular inflammation
mphocytes and polymorphonuclear cells), interface hepati-
, portal inflammation, and Mallory bodies (Table 1). When-
r appropriate, the extent of involvement and/or the distri-
tion were noted (Table 1). Finally, to study the impact of
pling variability on the diagnosis of NASH, 2 definitions
NASH that are available in the literature were used: (1) the
ociation of steatosis and hepatocyte ballooning with a pat-
n of centrilobular accentuation, and (2) the association of
atosis, hepatocyte ballooning, and perisinusoidal fibrosis
th a pattern of centrilobular accentuation.
Statistical Methods
The agreement between the pathologic assessments of
istinct samples from the same individual was calculated for
paired sample biopsy specimens using the  reliability test,
already performed by others,
13
in which values of 1 indicate
fect agreement, greater than 0.8 high agreement, between
and 0.8 substantial agreement, between 0.4 and 0.6 mod-
te agreement, between 0.4 and 0.2 fair agreement, and
ow 0.2 slight agreement.
18
The percentage of discordance
ween 2 samples was calculated from the number of pairs of
ples differing by more than 1 grade/stage. To underscore
variability of sampling error, the information provided by
first sample was compared with the second and expressed
terms of diagnostic value as sensitivity, specificity, positive
negative predictive values, and area under the ROC curves.
2-sided P value of .05 was considered to indicate a
ble 1. Elementary Histologic Lesions Assessed in This
Study
Lesion Description
atosis Quantification: % of involved acinus at low
magnification
teatosis grade 0: 0%–29%; 1: 30%–59%; 2: 60%–100%
patocyte ballooning 0: Absent; 1: occasional; 2 numerous
llory bodies 0: Absent; 1: occasional; 2: numerous
rosis, perisinusoidal
xtent of involvement 0: Absent; 1: present in some lobules
(focal); 2: present in most lobules
(extensive)
istribution zone 1, zone 2, zone 3 or panacinar
tal inflammation 0: Absent; 1: mild; 2: moderate; 3:
marked
rface hepatitis 0: Absent; 1: focal alteration of the
periportal or periseptal plate in some
portal tracts
2: Diffuse alteration of the periportal or
periseptal plate in some portal tracts or
focal lesions involving all portal tracts
or septa
3: Diffuse alteration of the periportal
plate in all portal tracts or septa
ular inflammation 0: Absent; 1: mild; 2: moderate; 3:
marked
dophilic bodies 0: Absent; 1: rare; 2: numeroustistically significant difference. Statistical analyses were per-
med using the NCSS 2003 software (Jerry Hintze, Visual
mponents; Sybase, Kaysville, UT).