Original article
A multicenter prospective study of the real-time
use of narrow-band imaging in the diagnosis of
premalignant gastric conditions and lesions
Authors
Institutions
submitted
10. January 2016
accepted after revision
17. April 2016
Bibliography
DOI http://dx.doi.org/
10.1055/s-0042-108435
Published online: 2016
Endoscopy
© Georg Thieme Verlag KG
Stuttgart · New York
ISSN 0013-726X
Corresponding author
Pedro Pimentel-Nunes, MD
Gastroenterology Department
Portuguese Oncology Institute
of Porto
Rua Dr. Bernardino de Almeida
4200-072 Porto
Portugal
Fax: + 351-22-5513646
pedronunesml@gmail.com
Pedro Pimentel-Nunes1, 2, Diogo Libânio1, Jorge Lage1, Diogo Abrantes3, Miguel Coimbra3, Gianluca Esposito4,
David Hormozdi5, Mike Pepper5, Silvia Drasovean6, Jonathan R. White7, Daniela Dobru6, James Buxbaum5,
Krish Ragunath7, Bruno Annibale4, Mário Dinis-Ribeiro1, 2
Institutions are listed at end of article.
Background and aim: Some studies suggest that
narrow-band imaging (NBI) can be more accurate
at diagnosing gastric intestinal metaplasia and
dysplasia than white-light endoscopy (WLE)
alone. We aimed to assess the real-time diagnos-
tic validity of high resolution endoscopy with and
without NBI in the diagnosis of gastric premalig-
nant conditions and to derive a classification for
endoscopic grading of gastric intestinal metapla-
sia (EGGIM).
Methods: A multicenter prospective study (five
centers: Portugal, Italy, Romania, UK, USA) was
performed involving the systematic use of high
resolution gastroscopes with image registry with
and without NBI in a centralized informatics plat-
form (available online). All users used the same
NBI classification. Histologic result was consid-
ered the diagnostic gold standard.
Results: A total of 238 patients and 1123 endo-
scopic biopsies were included. NBI globally in-
creased diagnostic accuracy by 11 percentage
Introduction
! ple and reproducible NBI classification was vali-
Gastric cancer remains a major problem in Wes-
tern society with significant mortality rates [1].
However, unlike in some Asian countries, screen-
ing is not recommended for the early detection of
this cancer mainly because it has not yet proven
to be cost-effective [2]. In fact, even though
endoscopy is considered the best tool for the de-
tection of gastric premalignant conditions and
early cancer, the correlation between convention-
al endoscopy and histology is considered inade-
quate [3].
Recent studies have shown that under ideal cir-
cumstances virtual chromoendoscopy with nar-
row-band imaging (NBI) is highly accurate for the
diagnosis of these conditions [4, 5]. However, the
classifications and patterns that have been used
for the description and detection of these lesions
previously have been heterogeneous and not re-
Pimentel-Nunes Pedro et al. NBI for the diagnosis of gastric lesions … Endoscopy
Downloaded by: University of Massachusetts. Copyrighted material.
points (NBI 94 % vs. WLE 83 %; P < 0.001) with no
difference in the identification of Helicobacter
pylori gastritis (73 % vs. 74 %). NBI increased sensi-
tivity for the diagnosis of intestinal metaplasia
significantly (87 % vs. 53 %; P < 0.001) and for the
diagnosis of dysplasia (92 % vs. 74 %). The added
benefit of NBI in terms of diagnostic accuracy
was greater in OLGIM III/IV than in OLGIM I/II
(25 percentage points vs. 15 percentage points,
respectively; P < 0.001). The area under the curve
(AUC) of the receiver operating characteristic
(ROC) curve for EGGIM in the identification of
extensive metaplasia was 0.98.
Conclusions: In a real-time scenario, NBI demon-
strates a high concordance with gastric histology,
superior to WLE. Diagnostic accuracy higher than
90 % suggests that routine use of NBI allows tar-
geted instead of random biopsy samples. EGGIM
also permits immediate grading of intestinal
metaplasia without biopsies and merits further
investigation.
producible in clinical practice [5]. In 2012, a sim-
dated and demonstrated to have accuracy rates
higher than 85 % – 90 % for the diagnosis of intes-
tinal metaplasia and dysplasia [4]. However, the
application of this simple NBI classification was
shown to be dependent on training, so in real-
time clinical practice the results may be affected
by this [6].
To our knowledge no study has described the dai-
ly clinical use of high resolution endoscopy (HRE)
with or without NBI in the diagnosis of gastric
premalignant conditions and lesions, or in the
estimation of the presence of advanced stage
intestinal metaplasia. Our study aim was to be
the first to describe the real-time diagnostic accu-
racy of HRE with and without NBI in the diagnosis
of gastric intestinal metaplasia and dysplasia. Sec-
ondly, we aimed to develop an endoscopic classi-
fication of gastric intestinal metaplasia and assess
 Original article

Table 1 Characteristics of the 238 patients who underwent high resolution endoscopy with white-light endoscopy and narrow-band imaging, their indications
for endoscopy, types of endoscope used, and final histology findings.

Portugal Italy USA UK Romania Total

Patients, n (% of total) 42 (18) 80 (33) 50 (21) 35 (15) 31 (13)  238
Biopsies, n (% of total) 200 (19) 411 (36) 181 (16) 170 (15) 161 (14) 1123
Male, % 41 42 33 49 48   42
Age, mean (SD), years 59 (11) 61 (14) 52 (12) 72 (12) 61 (10)   60 (14)
PPI use, % 26 14 46 23 74   32
Indication for endoscopy, %
Symptoms 57 82 96 71 61   77
Surveillance 43 18 4 29 39   23
Endoscope, %
H180 17 89 100 – –   54
H190 83 11 – – 100   31
GIF-260 – – – 100 –   15
Histology, %
Normal 31 52 60 37 10   42
Focal intestinal metaplasia 24 36 32 40 51   35
Extensive intestinal metaplasia 45 12 8 23 39   21
Dysplasia biopsies, n 7 4 2 1 9   23

Downloaded by: University of Massachusetts. Copyrighted material.

Helicobacter pylori infection, % 32 15 26 29 23   23
SD, standard deviation; PPI, proton pump inhibitor.

its diagnostic ability in the assessment of the advanced stages of Endoscopic procedures and biopsies
this disease. In each center, one or more endoscopists with NBI experience
(more than 100 HRE-NBI per year) performed the endoscopy
with the patient under pharyngeal anesthesia or deep sedation.
Methods The type of HRE gastroscope and NBI varied between centers
! (●" Table 1).

Study design and selection of patients
A multicenter prospective study was proposed and approved by
the Ethics committee of the Portuguese Oncology Institute of
Porto, Portugal. An invitation to participate in the study was
sent to several gastroenterology departments around the world
during the year 2013. The requirements for center participation
were access to HRE gastroscopes with NBI and the capability to
upload images. Five tertiary gastroenterology centers from differ-
ent countries agreed to participate in the study: Portuguese
Oncology Institute of Porto (Portugal); Hospital Sant’Andrea, Uni-
versity Sapienza Roma (Italy); Los Angeles County Hospital, Keck
School of Medicine, University of Southern California, Los Angeles
First, detailed observations of the gastric mucosa with white-
light endoscopy (WLE) without NBI were made. In each pro-
cedure, five different images were advised (the incisura and the
lesser and greater curvature of the antrum and corpus) but a
minimum of one representative image (three in total) from each
gastric area (antrum, incisura, and corpus) was required. The
images were recorded and from each area the endoscopist pre-
dicted a diagnosis of normal, metaplasia, or dysplasia. Next, NBI
observation of the entire mucosa was performed and images
from the same areas were recorded with the endoscopist again
assigning a diagnosis of normal, metaplasia, or dysplasia, accord-
ing to the classification of Pimentel-Nunes et al. [4] (●" Fig. 1).

(USA); University of Medicine and Pharmacy TG., Mures (Roma-
nia); Nottingham University Hospitals NHS Trust and the Univer-
sity of Nottingham, Queen’s Medical Centre, Nottingham (United
Kingdom).
Once they had registered in the centralized web-based platform,
the centers obtained access to the study protocol and were able to
start registering patients. After a pilot period of 3 months
(September– December 2013), consecutive patients undergoing
upper gastrointestinal (GI) endoscopy with an HRE gastroscope
because of symptoms, surveillance of gastritis, or screening in
each of these centers from January 2014 to March 2015 were
considered and included in this study after giving informed con-
sent. Exclusion criteria were: not having an indication for biopsy;
patients with significant comorbidities; anticoagulant therapy or
coagulation disorders; previous gastric neoplasia or surgery; not
being able to perform at least three biopsies during the endos-
copy.
Guided biopsies were then taken from the areas represented in
the images and sent for histopathologic evaluation in separate
jars (i. e. 5 WLE images, 5 NBI images, and 5 separated biopsies
from the areas represented in the images). This would later allow
for site-specific and patient-specific evaluation.
Centralized web-based informatics platform
A centralized web-based informatics platform that was accessible
online was created and logon details were sent to the different
centers. During a period of 3 – 4 months (September– December
2013), the platform was tested by each center and corrections
and suggestions for improvement were made. A final version of
the platform for the inclusion of patient data was made available
from January 2014.
When the platform was accessed, patient demographic and clin-
ical data were recorded first; once this had been done, it was then
possible to enter endoscopic data into the registry. The platform
required information on three areas (antrum, incisura, and cor-
pus) and three different questions were posed for each area: (A)
“Did you suspect/diagnose any superficial lesion (e. g. suspicious

Pimentel-Nunes Pedro et al. NBI for the diagnosis of gastric lesions … Endoscopy
 Original article

Downloaded by: University of Massachusetts. Copyrighted material.

Fig. 1 Endoscopic narrow-band imaging (NBI) classification. a Pattern Aa (normal antrum), regular oval (or circular) antral mucosa surrounding regular vessels
in the center of the gland. b Pattern Ab (normal gastric body), regular circular corpus mucosa, surrounded by regular vessels. c Pattern A+ (H. pylori gastritis in
normal mucosa), variable vascular density but with a normal corpus mucosa (histologic outcome was H. pylori-positive gastritis). d Pattern B (intestinal meta-
plasia) in gastric antrum, regular ridge/tubulovillous mucosa with regular vessels; slight aspects of light blue crest are also seen. e Pattern B (intestinal meta-
plasia) in corpus, two foci of tubulovillous mucosa (note the normal pattern Ab is also present). f Pattern C (dysplasia), irregular mucosa and vessels and loss of
mucosal architecture.

of dysplasia/carcinoma)?”, (B) “Did you perform any targeted NBI endoscopic grading of gastric intestinal metaplasia
biopsies according to endoscopic features other than suspicion (EGGIM)
of carcinoma?”, (C) “Did you perform any random biopsies?”. For A scale for endoscopic grading of gastric intestinal metaplasia
each of these questions the endoscopist had to answer “yes” or (EGGIM) using NBI was created (● " Table 2). Briefly, five different

“no” for WLE and for NBI. The images (WLE and NBI) were then
uploaded alongside the endoscopic diagnosis (normal, metapla-
sia, or dysplasia/carcinoma) and Helicobacter pylori endoscopic
diagnosis (yes/no).
The histology results were added at a later date. The platform
would only allow the histologic diagnosis to be inserted after
the endoscopic diagnosis section had been completed to avoid
bias. When all the data had been collected, the registry was
blocked and the data were converted to SPPS and Excel files.
Histopathologic evaluation
In each center, specimens were fixed in buffered formalin, pro-
cessed for paraffin embedding, sectioned, and stained with
hematoxylin and eosin (H&E). Gastric specimens were also eval-
uated for H. pylori infection using modified Giemsa (2 %) stain.
Two expert GI pathologists in each center, who were blind to the
WLE and NBI features, made the final histologic diagnosis accord-
ing to the Sydney–Vienna classification. Whenever possible, the
OLGIM (Operative Link on Gastric Intestinal Metaplasia) grading
for intestinal metaplasia (0 – IV) was calculated, but in 25 patients
(11 %) it was not possible to calculate OLGIM because there were
less than five biopsies) [7].
areas were considered (two areas in the antrum, two in the
corpus, and one in the incisura). Each area was scored 0 (no intes-
tinal metaplasia), 1 (focal intestinal metaplasia, ≤ 30 % of the
area), or 2 points (extensive intestinal metaplasia in that area,
> 30 % of the area), giving a possible total of 10 points.
When all of the registries were completed (March 2015), a single
observer (P.P.N.), who was blind to the final histology, observed
all the images and applied the EGGIM classification. The correla-
tion between EGGIM and OLGIM was then assessed. When the
images or histology did not allow total grading, the patient was
excluded from the analysis.
Statistical analysis
The Statistical Package for Social Sciences (SPSS 20.0 Package
Facility, SPSS Inc., Chicago, Illinois, USA) and Excel from Office
2010 (Microsoft Corporation, Redmond, Washington, USA) were
used for data support and analysis.
For estimation of sample size, the prevalence of intestinal meta-
plasia (main outcome) in our population was estimated to vary
between 10 % and 20 %. Assuming random biopsies as the gold
standard, previous reports suggested a sensitivity of 90 % for NBI
[4]. Given the very low accuracy of conventional endoscopy for
the diagnosis of intestinal metaplasia and that the sensitivity of
HRE is unknown, we hypothesized that NBI would have to have
more than 90 % sensitivity for intestinal metaplasia (as the mark-
er of risk) and more than 90 % global accuracy per biopsy, with a

Pimentel-Nunes Pedro et al. NBI for the diagnosis of gastric lesions … Endoscopy
 Original article

Endoscopic intestinal Antrum Incisura Corpus Table 2 Proposed classification

metaplasia scale for endoscopic grading of gastric
Lesser Greater Lesser Greater intestinal metaplasia (EGGIM)
curvature curvature curvature curvature with narrow-band imaging (NBI).
No intestinal metaplasia 0 0 0 0 0
Focal (≤ 30 % intestinal metaplasia) 1 1 1 1 1
Diffuse (> 30 % intestinal metaplasia) 2 2 2 2 2
Intestinal metaplasia score 0–4 0–2 0–4
The total score will vary from 0 (normal endoscopy with no areas suggestive of intestinal metaplasia) to 10 (diffuse metaplasia in all gastric
areas). We suggest a letter a or c is added to the score if metaplasia is more evident in the antrum (a) or in the corpus (c). For instance,
EGGIM 4a would represent intestinal metaplasia mostly in the antrum (suggestive of environmental gastritis), while EGGIM 4c would represent
intestinal metaplasia mostly in the corpus (suggestive of autoimmune gastritis).

10 % difference for those diagnostic measures being clinically

significant. Assuming a normal distribution, we calculated that
100
218 patients (corresponding to 1090 biopsies) would be required
to show a 10 % difference in sensitivity and accuracy between
WLE and NBI, with a power of 80 % and a probability of type I
error of 0.05. 80
Sensitivity, specificity, and global accuracy were estimated sep-

Downloaded by: University of Massachusetts. Copyrighted material.

arately and for the five centers combined, along with the 95 %
confidence intervals (CIs). Likelihood ratios were estimated on
60
the basis of mean sensitivity and specificity estimates. Histo- Accuracy
pathologic diagnosis was considered to be the gold standard or
reference test for accuracy estimates.
Comparison of patient demographics, clinical, and pathologic 40
features, as well as differences in the diagnostic accuracy of NBI/
WLE in each center and between centers was performed using
the chi-squared test, McNemar chi-squared test for the compari-
20
son of diagnostic accuracies, Student’s t test, or the correspond-
ing nonparametric test as appropriate. A receiver operating char-
acteristic (ROC) curve was used to assess the diagnostic accuracy
of the proposed EGGIM scale and to determine the optimal cut- 0
Portugal Romania Italy UK USA Global
off value for the detection of OLGIM III/IV patients (extensive
Center
metaplasia) on this scale. A value of P < 0.05 was considered to
Accuracy_WLE Accuracy_NBI
be statistically significant.
Fig. 2 Diagnostic accuracy of white-light endoscopy (WLE) and narrow-
band imaging (NBI) globally and for each center. Accuracy was significantly
Results increased globally with NBI (P < 0.01) and was increased in all centers (sta-
! tistically in 3 out of 5 centers).
Patients and clinicopathologic features
A total of 238 patients and 1123 endoscopic biopsies were in-
cluded in the study. Clinical and histologic characteristics of the was highest in Romania (19 percentage points) and lowest in the
patients are shown in ● " Table 1. The main indication for endos- USA (5 percentage points) (● " Fig. 2).

copy was symptoms (dyspepsia, reflux, abdominal pain, or ane- When levels of accuracy for the different stages of intestinal
mia) in all centers. However, in the Portuguese and Romanian metaplasia were analyzed, the added benefit of NBI was greater
centers the proportion of surveillance endoscopies was higher in OLGIM III/IV than in OLGIM I/II (25 percentage points vs. 15
than in the other centers, which resulted in more advanced histo- percentage points, respectively; P < 0.001) (● " Fig. 3). Moreover,

logic findings (P < 0.001). The proportion of patients with gastric rates of correct diagnosis according to gastric location were also
superficial lesions (dysplasia/carcinoma) was 8 % and the propor- higher with NBI in all areas (● " Table 3).

tion of biopsies that showed dysplasia was 2 %.
Diagnostic accuracy of HRE with WLE and NBI
Global diagnostic accuracy of high resolution WLE (HR-WLE)
was 83 %, varying between 69 % (Romania, the center with the
most advanced histology) and 89 % (USA, the center with the
least advanced histology). Global diagnostic accuracy of HRE-
NBI was 94 %, varying between 88 % (Romania) and 97 % (Italy).
In total, 23 lesions with dysplasia were identified (according to
the Paris classification 13 lesions were 0-IIa, 5 were 0-IIa + c, 2
were 0-IIc, and 3 lesions were 0-IIb). WLE “missed” two 0-IIa
lesions, one 0-IIc lesion, and two 0-IIb lesions, with NBI only
“missing” one 0-IIa lesion (NBI diagnosis of metaplasia with his-
tology showing low grade dysplasia).
Global and center-specific sensitivity, specificity, and positive
and negative likelihood ratios are shown in ● " Table 4. NBI

NBI globally increased diagnostic accuracy by 11 percentage increased sensitivity for the diagnosis of intestinal metaplasia
points (P < 0.001). This increase in diagnostic accuracy occurred significantly (87 % vs. 53 %; P < 0.001) and increased sensitivity
in all centers and was statistically significant in three centers for dysplasia (92 % vs. 74 %), even though the study is underpow-
(Portugal, Romania, and Italy). The increase in accuracy with NBI ered for detecting statistical differences in dysplasia. There were

Pimentel-Nunes Pedro et al. NBI for the diagnosis of gastric lesions … Endoscopy
 Original article

1.0
100

90 0.8

80 0.6

Sensitivity
Accuracy

70
0.4

60
0.2
Accuracy_WLE
Accuracy_NBI
50
0.0
0 I–II III–IV 0.0 0.2 0.4 0.6 0.8 1.0

Downloaded by: University of Massachusetts. Copyrighted material.

OLGIM stage 1-Specitivity

Fig. 3 Diagnostic accuracy with white-light endoscopy (WLE) and narrow-
band imaging (NBI) according to the extent of intestinal metaplasia. The
increase in diagnostic accuracy using NBI compared with WLE is higher in
the more advanced stages of intestinal metaplasia (P < 0.001). OLGIM,
operative link on gastric intestinal metaplasia. OLGIM I/II means focal
intestinal metaplasia; OLGIM III/IV means extensive metaplasia.
Fig. 4 Receiver operating characteristic (ROC) curve for the diagnostic
accuracy of the endoscopic grading of gastric intestinal metaplasia (EG-
GIM) classification in the detection of extensive metaplasia (OLGIM III/I).
The area under the curve (AUC) was 0.98. EGGIM scores of ≥ 4 and ≥ 5 were
identified as the best cut-off points, resulting in sensitivities of 98.1 % and
94.2 % and specificities of 86 % and 95.2 %, respectively. OLGIM, operative
link on gastric intestinal metaplasia.

Table 3 Diagnostic accuracy for white-light endoscopy (WLE) and narrow-

is advised) was investigated using a ROC curve (●" Fig. 4). The area
band imaging (NBI) according to gastric location.
under the curve (AUC) of the ROC curve in the identification of
Complete concordance with the histology result for that extensive intestinal metaplasia was 0.983 (95 %CI 0.969 – 0.997).
area with An EGGIM score of 5 was identified as the optimal cut-off value
WLE NBI
to identify patients with OLGIM III/IV, with a sensitivity of 94.2 %
and a specificity of 95.2 %. An EGGIM score of 4 (the cut-off for
n (%) 95 %CI n (%) 95 %CI moderate metaplasia) also performed well, achieving a sensitiv-
Body 187 (78.6) 72.7 % – 83.5 % 216 (90.8) 86.2 % – 93.9 % ity of 98.1 %, although with a lower specificity (86 %). ● " Table 5

Incisura 206 (86.6) 81.4 % – 90.5 % 225 (94.5) 90.6 % – 96.9 %
Antrum 171 (71.8) 65.6 % – 77.4 % 211 (88.7) 83.8 % – 92.3 %
n, number of patients where the technique showed complete concordance in that
area; %, proportion of patients where the technique showed complete concordance
in that area; CI, confidence interval.
NBI significantly increased the concordance with histology in all gastric areas.
no differences in the specificity for intestinal metaplasia and dys-
plasia between WLE and NBI, with results of more than 95 % for
both techniques. With regard to the diagnosis of H. pylori gastri-
tis, there was no difference in accuracy (74 % NBI vs. 73 % WLE),
although NBI demonstrated slightly higher sensitivity but re-
duced specificity (non-statistically significant; ●
" Table 4).
Relationship between histology (OLGIM) and NBI
endoscopy (EGGIM)
In 25 patients the number of biopsies and pictures were not
enough to calculate the OLGIM and EGGIM; in an additional 12
patients the observer was not able to calculate the EGGIM
because the pictures lacked quality, so there were 201 patients
included for the correlation between OLGIM and EGGIM.
The diagnostic accuracy of the proposed EGGIM classification
(●
" Table 2) in the identification of patients with extensive intes-
tinal metaplasia (OLGIM III/IV, for whom endoscopic surveillance
shows the correspondence between the OLGIM and EGGIM clas-
sifications.
Discussion
!
This study is the first prospective multicenter study of the real-
time application of HRE both with WLE and NBI for the diagnosis
of gastric intestinal metaplasia and dysplasia. We showed that
even though the results with HR-WLE were acceptable, NBI im-
proved the diagnostic yield in more than 10 %, achieving a 94 %
diagnostic accuracy rate. Moreover, we have proposed an endo-
scopic classification for grading of intestinal metaplasia that pre-
sented an excellent correlation with histology and OLGIM stages.
These results suggest that NBI should be used for direct guidance
of endoscopic gastric biopsies instead of random biopsies in a
first endoscopy and that NBI may even obviate the need for biop-
sies in patients under surveillance.
Our study has some limitations. First, even though we compared
the diagnostic accuracy of WLE with NBI, the design of the study
is not ideal to compare NBI to WLE as, in most cases, the endo-
scopists evaluated the gastric mucosa firstly with WLE and subse-
quently with NBI. Therefore, the study design meant that the ob-
server was not blinded to the WLE evaluation and this could have

Pimentel-Nunes Pedro et al. NBI for the diagnosis of gastric lesions … Endoscopy
 Original article

Table 4 Global and center-specific sensitivity, specificity, positive and negative likelihood ratios for the diagnosis of intestinal metaplasia, dysplasia, and
Helicobacter pylori gastritis.

Portugal Italy USA UK Romania Total (95 %CI)

WLE NBI WLE NBI WLE NBI WLE NBI WLE NBI WLE NBI
Intestinal metaplasia
Sensitivity, % 64 97 47 91 37 81 59 76 51 84 53 (47 – 58) 87 (83 – 91)
Specificity, % 96 90 99 98 99 95 98 99 96 92 98 (97 – 99) 97 (95 – 98)
LR+ 17.8 9.3 48.9 52.5 28.5 17.9 32.8 82.2 13.1 10.9 28.8 (17 – 49) 27.8 (19 – 41)
LR− 0.37 0.04 0.54 0.09 0.64 0.19 0.42 0.25 0.5 0.17 0.48 (0.43 – 0.53) 0.13 (0.1 – 0.17)
Dysplasia
Sensitivity, % 66 100 100 100 100 100 100 100 67 89 74 (52 – 90) 92 (73 – 99)
Specificity, % 100 100 99 99 100 100 99 100 100 99 99 (98 – 100) 99 (98 – 100)
LR+ – – 407 407 – – 169 – – 135 407 (100 – 1658) 512 (128 – 2058)
LR− 0.33 0 0 0 0 0 0 0 0.33 0.11 0.26 (0.13 – 0.52) 0.08 (0.02 – 0.31)
Helicobacter pylori gastritis
Sensitivity, % 42 55 67 71 67 92 37 46 71 86 57 (43 – 69) 69 (55 – 81)
Specificity, % 73 71 79 81 84 66 88 88 71 75 79 (72 – 85) 67 (60 – 74)
LR+ 1.6 1.9 3.2 3.7 4.3 2.7 2.9 3.6 2.5 3.4 2.7 (1.9 – 3.9) 2.1 (1.6 – 2.7)
LR− 0.8 0.64 0.42 0.36 0.4 0.13 0.73 0.62 0.4 0.19 0.55 (0.41 – 0.74) 0.46 (0.31 – 0.69)
CI, confidence interval; WLE, white-light endoscopy; NBI, narrow-band imaging; LR+ positive likelihood ratio; LR− negative likelihood ratio.

Downloaded by: University of Massachusetts. Copyrighted material.

% of the total EGGIM score Table 5 Correspondence
(% within each OLGIM grade) between the operative link on
0 1–2 3–4 5–7 8 – 10 gastric intestinal metaplasia
OLGIM 0 37 (88) 5 (12) 0 0 0 (OLGIM) grade and endoscopic
I 3 (15) 10 (59) 4 (22) 1 (4) 0 grading of gastric intestinal
II 1 (5) 4 (20) 11 (59) 3 (16) 0 metaplasia (EGGIM) score.
III 0 1 (4) 1 (8) 9 (88) 0
IV 0 0 0 2 (19) 9 (81)
Extent of intestinal metaplasia Absent Focal to moderate Extensive
There was correspondence between OLGIM and EGGIM.

influenced the subsequent NBI diagnosis. However, the study bet-
ter imitates the reality in which the endoscopist has access to both
modalities at the touch of a button. We believe that the design of
the current study is the best for showing the advantage of NBI in a
real-life scenario and that our results show the best possible diag-
nostic accuracy in clinical practice when using both techniques si-
multaneously. Moreover, the great importance of WLE observa-
tion implies that NBI can only be seen as an adjunct to WLE and
not as a technology to be used in isolation.
The second and most important limitation of this study is that
gastric atrophy, an important premalignant condition, is not con-
sidered. However, to our knowledge, there is no validated NBI
endoscopic classification for the diagnosis of gastric atrophy.
Moreover, reproducibility for the diagnosis of atrophy amongst
pathologists is also poor in comparison to intestinal metaplasia,
which is reproducible both histologically and endoscopically
[7 – 10]. Furthermore, there is greater consensus that intestinal
metaplasia is a marker for high risk in gastric cancer than there
is for isolated gastric atrophy [11].
The third limitation of our multicenter study is the absence of
centralized pathologic evaluation for all of the biopsy samples.
However, we believe that this did not influence our results as in
every center the samples were evaluated by two expert GI
pathologists who are able to make this reproducible histologic
diagnosis [3, 7].
Fourth, even though EGGIM correlated very well with histology,
only one experienced observer applied the endoscopic NBI classi-
fication. Indeed, the identification of these patterns has been
shown to be dependent on training, so with less experienced ob-
servers this correlation may not be as strong. Nevertheless, our
purpose was not to validate this classification but to show its fea-
sibility in clinical practice. Future studies are needed to apply and
validate this classification further.
Finally, all of the participant endoscopists had significant NBI
experience (more than 100 NBI upper GI endoscopies per year).
We have previously shown that the correct identification of NBI
patterns requires a learning curve [6]. So, our results can be gener-
alized only to reference centers that commonly use NBI for the di-
agnosis and management of these situations and probably not to
the daily routine of an endoscopist who is not experienced at NBI.
Other studies have compared WLE with NBI for the diagnosis of
gastric lesions. Xirouchakis et al. [12] failed to show a clear advan-
tage of NBI; however, their population was non-high risk and they
used only dark-NBI without previous knowledge of the WLE find-
ings. On the other hand, Capelle et al. [13], with a similar design
but with patients under surveillance because of metaplasia or dys-
plasia (high-risk population), were able to show that dark-NBI was
better than WLE for the detection of metaplasia and dysplasia,
even though their results were somewhat modest. Ang et al. [14]
with a more robust design (multicenter study of Asiatic centers)
and using only light-NBI clearly showed that NBI is better than
WLE for the detection of metaplasia (10 % more detection of meta-
plasia). Our own group showed that in fact light-NBI is better than
WLE for the detection of advanced intestinal metaplasia, with
more than 90 % accuracy in the diagnosis of extensive disease [15].
The advantages of the present study are its prospective multicen-
ter design involving centers with different populations (both
average and high risk), the real-time diagnosis with WLE and

Pimentel-Nunes Pedro et al. NBI for the diagnosis of gastric lesions … Endoscopy

then with NBI (representing real-life practice), and the systema-
tic recording of WLE and NBI photographs that allowed the crea-
tion of a classification for staging of gastric intestinal metaplasia
(EGGIM).
For the detection of H. pylori gastritis, both WLE and NBI appear
limited (73 % – 74 % global accuracy). These results are in accord
with our previous studies and mean that we still have to rely on
histology (or non-invasive tests) for a final H. pylori diagnosis [4,
6].
With regards to dysplasia, although the study was underpowered
to detect differences between WLE and NBI, the difference in sen-
sitivity (92 % NBI vs. 74 % WLE) suggests that NBI might be better
for the diagnosis of superficial gastric lesions. However, when we
analyzed the data of the five lesions that were “missed” by WLE
but were correctly diagnosed by NBI, three of them were seen by
WLE but were interpreted as only intestinal metaplasia (two 0-IIa
lesions and one 0-IIc lesion). Probably, even without NBI, biopsies
would have been taken from these areas and the final diagnosis
of a superficial lesion would therefore not have been missed.
The two lesions that were seen only with NBI (normal mucosa
with WLE) were two 0-IIb lesions, suggesting that the greatest
diagnostic advantage of NBI would be in the detection of this
kind of lesion.
In accordance with our results, Ezoe et al. [16] also showed that
magnification NBI was better than WLE for the diagnosis of small
depressed cancers (accuracy 90 % vs. 65 %; P < 0.001) but that the
real advantage of NBI would be in combination with WLE, which
increased global accuracy to 97 %. Overall, NBI may help in the de-
tection of gastric superficial lesions but is most useful in delineat-
ing lesions, in giving the endoscopist confidence in the endoscopic
diagnosis, and to guide sampling of the most suspicious areas.
HR-WLE achieved a global diagnostic accuracy of 83 % (center
variation of 69 % – 89 %). This means that by itself WLE without
biopsies would have incorrectly evaluated 11 % – 31 % of the pa-
tients. Moreover, when we consider only the patients with exten-
sive metaplasia (OLGIM III and IV), this accuracy drops to 68 %,
meaning that more than one-third of the patients requiring sur-
veillance would have been missed without biopsy samples. NBI
increased this accuracy to 94 % (87 % – 98 %), but the benefit was
greater in OLGIM III/IV patients particularly, with a global accura-
cy of 97 %. Indeed, the main advantage of NBI is its sensitivity for
intestinal metaplasia, which is much higher than with WLE (87 %
vs. 53 %), without a significant difference in specificity. Moreover,
the cases where NBI might not detect metaplasia appear to be
small/focal areas of intestinal metaplasia that are probably clini-
cally irrelevant, as suggested by the analysis by OLGIM stage.
Our results strongly suggest that, after an HR-WLE without NBI,
biopsies are needed to correctly stage our patients. It is also im-
portant to note that, when observing the extent of the gastric
mucosa with NBI, the areas of intestinal metaplasia are dispersed
and often alternate with areas of normal mucosa, even in cases of
OLGIM III/IV. Therefore, it is possible that purely random biopsies
may be normal, even in cases of diffuse/extensive metaplasia, and
consequently the stage of intestinal metaplasia may vary be-
tween endoscopies, even though the extent of metaplasia will
probably be the same [17]. For these reasons we believe that NBI
observation of all of the mucosa gives more detailed information
about the true extent of metaplasia and should be applied in clin-
ical practice.
In that context, we have proposed a numeric classification for
staging of gastric intestinal metaplasia (EGGIM; ● " Table 2) that
can easily be applied in clinical practice and that has proven to
Original article
correlate strongly with OLGIM and with the extent of the intes-
tinal metaplasia. We predicted that a score of 6 would represent
the lowest score for extensive metaplasia but found in fact that
the best cut-off to suggest surveillance appears to be a score of
5, which identified almost all of the patients with OLGIM III/IV.
The application of this classification and consequently NBI in
clinical practice may have several advantages, namely: detailed
observation of the total area of the mucosa; guided instead of
purely random biopsies; an accurate view of the spread of meta-
plasia in the gastric mucosa as biopsies represent only a fraction
of the area of the mucosa; the ability to calculate the grade of
metaplasia immediately after the endoscopy and make a propo-
sal for surveillance; follow-up of patients without biopsies, con-
sequently reducing the cost of endoscopies [17].
Nevertheless, it should be noted that the application of EGGIM in
clinical practice will probably only be feasible routinely with light-
NBI and not with dark-NBI. In fact, even though there were no sta-
tistically significant differences between the two types of NBI, all
of the 12 cases that were excluded from the EGGIM analysis were
examined with CV 180 scopes. In these cases the observer was not
Downloaded by: University of Massachusetts. Copyrighted material.
able to calculate the EGGIM given that the dark images did not
allow the extent of the metaplasia to be calculated.
From a research point of view also, EGGIM may be a powerful tool
as it can be used to help to solve some important questions that
histology alone cannot answer. For example, “Is intestinal meta-
plasia reversible after H. pylori eradication as there is no consen-
sus in the literature?” [18 – 23]. However, as we can see with NBI,
the areas of metaplasia alternate with normal mucosa, so random
biopsies can both overestimate and underestimate metaplasia at
different times depending on the areas that are biopsied. In our
opinion, only with observation of the total area of intestinal
metaplasia can we conclusively answer this question and others
regarding the effect of an intervention in the regression or pro-
gression of this disease. Future studies should apply the EGGIM
classification (as well as NBI-guided biopsies) in order to achieve
conclusive answers to these unanswered problems.
In conclusion, this is the first study to use real-time application of
advanced HRE for the diagnosis of gastric premalignant and
superficial neoplastic conditions. We have demonstrated the
ability to diagnose gastric lesions with a high degree of certainty
and have been able to grade intestinal metaplasia endoscopically
to a point where we are able to assign advanced disease on sur-
veillance endoscopies without relying on biopsy results. We feel
this study will change clinical practice by introducing a strategy
of NBI-targeted biopsies and, with further development, could
potentially remove the need for biopsies in patients under sur-
veillance in the future.
Competing interests: None
Institutions
1
Gastroenterology Department, Portuguese Oncology Institute of Porto,
Porto, Portugal
2
Center for Research in Health Technologies and Information Systems
(CINTESIS), Faculty of Medicine, Porto, Portugal
3
Instituto de Telecomunicações, Faculty of Sciences of the University of Porto,
Porto, Portugal
4
Department of Medicine, Surgery and Translational Medicine, University
Hospital Sant'Andrea, University Sapienza Roma, Rome, Italy
5
Los Angeles County Hospital, Keck School of Medicine, University of Southern
California, Los Angeles, California, USA
6
University of Medicine and Pharmacy TG., Mures, Romania
7
NIHR Biomedical Research Unit in Gastrointestinal and Liver Diseases at
Nottingham University Hospitals NHS Trust and the University of Notting-
ham, Nottingham, United Kingdom
Pimentel-Nunes Pedro et al. NBI for the diagnosis of gastric lesions … Endoscopy
 Original article
Acknowledgments
! of gastric atrophy. Histopathology 1999; 34: 320 – 325
None of the authors have any disclosures.
This article presents independent research supported by the Na-
tional Institute for Health Research (NIHR). The views expressed
are those of the author(s) and not necessarily those of the NHS,
the NIHR, or the Department of Health.
References
1 Jemal A, Bray F, Center MM et al. Global cancer statistics. CA Cancer J
Clin 2011; 61: 69 – 90
2 Leja M, You W, Camargo MC et al. Implementation of gastric cancer
screening – the global experience. Best Pract Res Clin Gastroenterol
2014; 28: 1093 – 1106
3 Dinis-Ribeiro M, Areia M, de Vries AC et al. Management of precancer-
ous conditions and lesions in the stomach (MAPS): guideline from the
European Society of Gastrointestinal Endoscopy (ESGE), European
Helicobacter Study Group (EHSG), European Society of Pathology
(ESP), and the Sociedade Portuguesa de Endoscopia Digestiva (SPED).
Endoscopy 2012; 44: 74 – 94
4 Pimentel-Nunes P, Dinis-Ribeiro M, Soares JB et al. A multicenter valida-
tion of an endoscopic classification with narrow band imaging for gas-
tric precancerous and cancerous lesions. Endoscopy 2012; 44: 236 –
246
5 Kikuste I, Marques-Pereira R, Monteiro-Soares M et al. Systematic re-
view of the diagnosis of gastric premalignant conditions and neoplasia
with high-resolution endoscopic technologies. Scand J Gastroenterol
2013; 48: 1108 – 1117
6 Dias-Silva D, Pimentel-Nunes P, Magalhaes J et al. The learning curve for
narrow-band imaging in the diagnosis of precancerous gastric lesions
by using Web-based video. Gastrointest Endosc 2014; 79: 910 – 920 ;
quiz 983 e911, 983 e914
7 Capelle LG, de Vries AC, Haringsma J et al. The staging of gastritis with
the OLGA system by using intestinal metaplasia as an accurate alterna-
tive for atrophic gastritis. Gastrointest Endosc 2010; 71: 1150 – 1158
8 Chen XY, van der Hulst RW, Bruno MJ et al. Interobserver variation in
the histopathological scoring of Helicobacter pylori related gastritis. J
Clin Pathol 1999; 52: 612 – 615
9 el-Zimaity HM, Graham DY, al-Assi MT et al. Interobserver variation in
the histopathological assessment of Helicobacter pylori gastritis. Hum
Pathol 1996; 27: 35 – 41
Pimentel-Nunes Pedro et al. NBI for the diagnosis of gastric lesions … Endoscopy
10 Offerhaus GJ, Price AB, Haot J et al. Observer agreement on the grading
11 den Hoed CM, Holster IL, Capelle LG et al. Follow-up of premalignant le-
sions in patients at risk for progression to gastric cancer. Endoscopy
2013; 45: 249 – 256
12 Xirouchakis E, Laoudi F, Tsartsali L et al. Screening for gastric premalig-
nant lesions with narrow band imaging, white light and updated Syd-
ney protocol or both? Dig Dis Sci 2013; 58: 1084 – 1090
13 Capelle LG, Haringsma J, de Vries AC et al. Narrow band imaging for the
detection of gastric intestinal metaplasia and dysplasia during surveil-
lance endoscopy. Dig Dis Sci 2010; 55: 3442 – 3448
14 Ang TL, Pittayanon R, Lau JY et al. A multicenter randomized compari-
son between high-definition white light endoscopy and narrow band
imaging for detection of gastric lesions. Eur J Gastroenterol Hepatol
2015; 27: 1473 – 1478
15 Lage J, Pimentel-Nunes P, Figueiredo PC et al. Light-NBI to identify high-
risk phenotypes for gastric adenocarcinoma: do we still need biopsies?
Scand J Gastroenterol 2016; 51: 501 – 506
16 Ezoe Y, Muto M, Uedo N et al. Magnifying narrowband imaging is more
accurate than conventional white-light imaging in diagnosis of gastric
mucosal cancer. Gastroenterology 2011; 141: 2017 – 2025 e2013
17 Dinis-Ribeiro M, Kuipers EJ. Identification of gastric atrophic changes:
from histopathology to endoscopy. Endoscopy 2015; 47: 533 – 537
18 Annibale B, Aprile MR, D’Ambra G et al. Cure of Helicobacter pylori in-
fection in atrophic body gastritis patients does not improve mucosal
Downloaded by: University of Massachusetts. Copyrighted material.
atrophy but reduces hypergastrinemia and its related effects on body
ECL-cell hyperplasia. Aliment Pharmacol Ther 2000; 14: 625 – 634
19 Correa P, Fontham ET, Bravo JC et al. Chemoprevention of gastric dys-
plasia: randomized trial of antioxidant supplements and anti-Helico-
bacter pylori therapy. J Natl Cancer Inst 2000; 92: 1881 – 1888
20 De Vries AC, Kuipers EJ. Review article: Helicobacter pylori eradication
for the prevention of gastric cancer. Aliment Pharmacol Ther 2007; 26:
0225 – 35
21 Mera R, Fontham ET, Bravo LE et al. Long term follow up of patients
treated for Helicobacter pylori infection. Gut 2005; 54: 1536 – 1540
22 Satoh K, Kimura K, Takimoto T et al. A follow-up study of atrophic gas-
tritis and intestinal metaplasia after eradication of Helicobacter pylori.
Helicobacter 1998; 3: 236 – 240
23 Wang J, Xu L, Shi R et al. Gastric atrophy and intestinal metaplasia be-
fore and after Helicobacter pylori eradication: a meta-analysis. Diges-
tion 2011; 83: 253 – 260