ORIGINAL ARTICLE: Clinical Endoscopy

Pit pattern analysis with high-definition chromoendoscopy and

narrow-band imaging for optical diagnosis of dysplasia in
patients with ulcerative colitis
Raf Bisschops, MD, PhD,1 Talat Bessissow, MD, PhD,2 Evelien Dekker, MD, PhD,3
James E. East, MD, PhD, FRCP,4 Adolfo Para-Blanco, PD, PhD,5 Krish Ragunath, MD, PhD,5
Pradeep Bhandari, MD, PhD,6 Matt Rutter, MD, PhD,7 Erik Schoon, MD, PhD,8 Ana Wilson, MD, PhD,9
Jestinah Mahachie John, PhD,10 Kristel Van Steen, PhD,10 Filip Baert, MD, PhD,11 Marc Ferrante, MD, PhD1
Leuven, Liège, Roeselare, Belgium; Montreal, Quebec, Canada; Amsterdam, Eindhoven, The Netherlands; Oxford,
Nottingham, Portsmouth, Cleveland, Durham, London, United Kingdom

Background and Aims: Patients with longstanding ulcerative colitis (UC) are at increased risk of developing colo-
rectal neoplasia. Chromoendoscopy (CE) increases detection of lesions, and Kudo pit pattern classification I and II
have been suggested to be predictive of benign polyps in UC. Little is known on the use of this classification in non-
magnified high-definition (HD) (virtual) CE and narrow-band Imaging (NBI) or on the interobserver agreement. The
aim of this pilot study was to assess the diagnostic accuracy and the interobserver agreement of the Kudo pit pattern
classification in UC patients undergoing surveillance with methylene blue CE or NBI in a multicenter study.
Methods: Fifty images of lesions identified in 27 UC patients (13 neoplastic) either with classical CE (methylene
blue .1%; n Z 24) or NBI (n Z 26) were selected by an independent investigator. Images were selected from a
randomized controlled trial to compare CE and NBI. All nonmagnified images were obtained with a processor and
mounted in a PowerPoint file in a standardized way (same size; black background). Ten endoscopists with exten-
sive experience in NBI/CE were asked to assess the lesions for the predominant Kudo pit pattern (I, II, IIIL, IIIS, IV,
and V) to indicate if they believed the lesion was neoplastic and how confident they were about the diagnosis.
Histology was used as the criterion standard.
Results: Median sensitivity, specificity, negative predictive value, and positive predictive value for diagnosing neoplasia
based on the presence of pit pattern other than I or II was 77%, 68%, 88%, and 46%, respectively. Diagnostic accuracy was
significantly higher when a diagnosis was made with a high level of confidence (77% vs 21%, P < .001). The overall inter-
observer agreement for any pit pattern was only fair (k Z .282), with CE being significantly better than NBI (.322 vs .224,
P < .001). From a clinical viewpoint the difference between neoplastic and non-neoplastic lesions is important. The agree-
ment for differentiation between non-neoplastic patterns (I, II) and neoplastic patterns (IIIL, IIIS, IV, or V) was moderate
(k Z .587) and even significantly better for NBI in comparison with CE (k Z .653 vs .495, P < .001).
Conclusions: Differentiation between non-neoplastic and neoplastic pit patterns in UC lesions shows a moderate
to substantial agreement among expert endoscopists. The agreement for differentiating neoplastic from non-
neoplastic lesions is significantly better for NBI in comparison with HD CE. The assessment of pit pattern I or
II with nonmagnified HD CE or NBI has a high negative predictive value to rule out neoplasia. (Clinical trial regis-
tration number: NCT01882205.) (Gastrointest Endosc 2017;86:1100-6.)

(footnotes appear on last page of article)

Patients with longstanding ulcerative colitis (UC) are at disease duration and greater disease activity and extent.4 In
increased risk of developing colorectal cancer (CRC).1,2 addition, better endoscopic surveillance allows early
Epidemiologic data suggest that the risk of CRC in those pa- detection of dysplasia before CRC development and allows
tients is decreasing but is still higher than an average-risk pop- curative endoscopic or surgical resection.
ulation without inflammatory bowel disease.3 In comparison Chromoendoscopy (CE) increases the detection of
with older data, this decrease in CRC is often attributed to neoplasia in patients with longstanding UC compared with
both a better medical treatment and better surveillance. regular white-light endoscopy with random 4-quadrant
Indeed, the risk of UC-associated CRC increases with a longer biopsy specimens taken every 10 cm.5-7 Because of the

1100 GASTROINTESTINAL ENDOSCOPY Volume 86, No. 6 : 2017 [Link]

Bisschops et al Diagnostic accuracy of Kudo pit pattern classification in UC patients

significantly increased diagnostic yield, CE is now the 13 (historically and per protocol defined as 6 adenoma
preferred surveillance method in several guidelines.8,9 In a with low-grade dysplasia, 5 adenoma-like mass with low-
randomized controlled trial comparing methylene blue CE grade dysplasia, 1 dysplasia-associated lesion or mass
and 4-quadrant random biopsy sampling during white- with low-grade dysplasia, and 1 sessile serrated adenoma
light endoscopy, a pragmatic Kudo pit pattern classification with low-grade dysplasia). Lesions were classified accord-
was put forward that allowed the differentiation in a dichot- ing to the Vienna classification.14 All biopsy specimens
omous way between non-neoplastic and neoplastic lesions were reviewed by an pathologist with specific expertise
in patients with longstanding colitis undergoing surveil- in GI pathology. In case of dysplasia the diagnosis was
lance.5 As such, Kudo I and II have been suggested to be confirmed by a second expert pathologist.
predictive of benign lesions. However, the diagnostic The selected images originated from 27 different pa-
accuracy and interobserver variability of nonmagnified tients with longstanding UC undergoing surveillance
high-definition (HD) CE is unknown. endoscopy (9/27 [33% female]; median disease duration
The role of narrow-band imaging (NBI) in surveillance at index endoscopy, 16 years [interquartile range, 10-
of longstanding colitis is still controversial, mainly because 22]). The maintenance therapy consisted of mesalamine
NBI has not been shown to detect more dysplasia in com- in 18 (67%), a thiopurine in 7 (26%), and/or a biologic in
parison with HD CE or HD white-light endoscopy, and NBI 11 (41%) patients. Twenty-four images (44%) were
use is not supported but international consensus guide- retrieved from 12 patients who underwent surveillance
lines.10-13 Furthermore, it is unclear whether the pit endoscopy with classical CE, using methylene blue .1%.
pattern classification can also be applied to HD NBI. The The other 26 images (56%) were derived from 15 patients
aim of the current pilot study was to assess diagnostic ac- undergoing surveillance colonoscopy with NBI.
curacy and the interobserver and intraobserver agreement Images were obtained with an Exera II processor
of Kudo pit pattern in UC surveillance endoscopy with (Olympus, Tokyo, Japan) in HD format and HD colono-
nonmagnified HD CE or HD NBI. scopes (180Q series; Olympus) and downloaded from
the server in a joint photographic experts group or JPEG
format with a size of about 1000 kilobyte and a pixel array
METHODS of 1008  1280 and 72 dots per inch. Next, all 50 images
were mounted in a PowerPoint file in 2 different sets
Selection of endoscopic images with a random order but in a standardized way. All images
Stored images from endoscopic procedures performed be- received a black background in the PowerPoint file and
tween July 2008 and March 2012 were retrieved from a ran- were further displayed in their original size, without chang-
domized controlled trial comparing HD CE versus NBI for ing the brightness, contrast, or color balance with any soft-
the detection of neoplasia in patients with longstanding UC ware. Figure 1 shows an example of 2 of these images.
([Link] NCT01882205).12 Written informed consent
was obtained from all patients participating in this trial.
Images were stored on a computerized database of the Assessment of endoscopic images
Endoscopy Unit of the University Hospitals Leuven Ten endoscopists (R.B., T.B., A.P.-B., P.B., M.R., K.R.,
(Leuven, Belgium). One investigator (M.F.), who was not E.D., J.E.E., A.W., and E.S.) with previous experience in
involved in the endoscopic procedures or in the assess- NBI and HD CE and in Kudo pit pattern classification
ment of the pit patterns in a later phase of the study, through participation in clinical trials were invited for this
screened all 267 available images and selected 50 of study. They received twice a PowerPoint file with the
them to be used for the study purpose. The images were same 50 images and a score sheet to indicate the most
selected to ascertain the image was sharp with a clearly advanced Kudo pit pattern, the neoplastic nature, and
visible lesions with visible pits and crypts. In addition, how confident they were about their assessment on a scale
these images had previously been scored for Kudo pit from 1 (not sure at all) to 5 (very sure) (see Supplementary
pattern by a single experienced gastroenterologist (R.B.) Figure 1). Scores of 4 and 5 were considered as a high-level
during the endoscopic procedure (so blinded to the final confidence diagnosis. Pit pattern diagnosis was made ac-
pathology), and histologic samples of matching biopsy cording to the Kudo classification as types I, II, IIIL, IIIS,
specimens had been analyzed by an experienced patholo- IV, or V.15 The interval between set 1 and 2 was at least
gist. The intent was to have an equal distribution between 10 weeks to minimize recall bias. Furthermore, the
CE and NBI pictures and a selection comprising a wide endoscopists (except for R.B. and T.B.) were not
range of the Kudo pit pattern (I, 8/50; II, 22/50; III, 9/50; informed that they were going to assess the same images
IIIs, 5/50; IV, 6/50) and morphology (Table 1). a second time (in a different order). The images were
The original histologic diagnosis was non-neoplastic in assessed independently by each investigator, who was
37 (2 mild architectural abnormalities, 4 inactive chronic blinded to the clinical profiles of the patients, including
colitis, 6 UC inflammatory changes, 5 inflammatory pseu- the endoscopic management of the lesions and the
dopolyps, and 20 hyperplastic polyps) and neoplastic in histologic analysis.

[Link] Volume 86, No. 6 : 2017 GASTROINTESTINAL ENDOSCOPY 1101

Diagnostic accuracy of Kudo pit pattern classification in UC patients Bisschops et al

TABLE 1. Distribution of morphology of the lesions according to the RESULTS

SCENIC consensus modified Paris classification
Accuracy to predict neoplastic lesions
Neoplastic Non-neoplastic
lesions lesions
Thirteen of 50 lesions were neoplastic on histology. As
shown in Table 2, median (range) sensitivity, specificity,
Polypoid NPV, and PPV for diagnosing neoplasia based on the
Pedunculated (Ip) 0 2 presence of pit pattern IIIL, IIIS, IV, or V were 77% (54%-
Sessile (Is) 2 9 85%), 68% (51%-84%), 88% (84%-94%), and 46% (36%-
Nonpolypoid 61%), respectively. Similarly, based on the overall
Flat endoscopic appearance of the lesion, sensitivity (range),
specificity, NPV, and PPV were 77% (31%-100%), 69%
IIa 11 17
(43%-92%), 90% (79%-100%), and 48% (37%-67%),
IIb 0 8
respectively. The NPV was comparable for lesions detected
Depressed (IIc) 0 1 by CE or NBI (P Z .971 for I and II vs IIIL, IIIS, IV or V;
P Z .739 for non-neoplastic vs neoplastic appearance).

Statistics
Level of certainty
All data were collected with predesigned forms. Both
During the first assessment endoscopists were sure or
SPSS 20.0 (SPSS Inc., Chicago, Ill) and R (version 2.15.1,
very sure about the neoplastic nature of the lesion of interest
[Link] were used for statistical analyses.
in 62% of cases (range, 34%-90%). The overall diagnostic ac-
Reproducibility of interpretations was tested by Cohen’s
curacy was significantly better when endoscopists had a high
kappa coefficient for interobserver variability (between 2 ob-
level of confidence (77% vs 21%, P < .001). This was mainly
servers, described as median kappa and range) and by
because of an improvement of the PPV (35% vs 58%). No in-
Fleiss’s kappa coefficient for general interobserver vari-
fluence was seen on the NPV (Table 3).
ability (among 10 independent observers, described as
kappa and P value). In general, k > .81 was regarded excel-
lent, .80 > k > .61 as substantial, .60 > k > .41 as moderate, Interobserver and intraobserver agreement
.40 < k < .00 as average, and k < .00 as poor.16 Interobserver The overall interobserver agreement for the first assess-
agreement was assessed as agreement between readings of ment of the images is depicted in Table 4. Interobserver
2 or more observers, whereas intraobserver agreement was agreement was significantly worse for the complete Kudo
assessed as agreement between first and second readings of pit pattern assessment with 6 different options compared
the same observer. For the interobserver agreement only with all other endoscopic groups (Kudo I and II vs IIIL
readings from the first round of image assessment were and IV vs IIIS and V; Kudo I, II vs IIIL, IIIS, IV, and V; and
used. Interobserver and intraobserver agreement were overall assessment of the neoplastic character of the
analyzed for all Kudo pit pattern possibilities (6 options), lesion; all P < .001). The overall interobserver agreement
for 3 groups of Kudo pit pattern possibilities (I and II vs for any pit pattern assessment with 6 options was only
IIIL and IV vs IIIS and V, 3 options), for 2 groups of Kudo fair (k Z .282) but was significantly better with CE (k Z
pit pattern possibilities (I, II vs IIIL, IIIS, IV and V, 2 .322 for CE vs k Z .224 for NBI, P Z .001). The
options), and finally for the overall assessment of the agreement for differentiation between non-neoplastic (I,
neoplastic character of the lesion (non-neoplastic vs II) and neoplastic pit patterns (IIIL, IIIS, IV, V) was moder-
neoplastic, 2 options). ate (k Z .587) but significantly better and substantial for
Inter- and intraobserver agreement were first analyzed for NBI (k Z .653 for NBI vs k Z .495 for CE, P < .001). Simi-
all images together and then separately for the different endo- larly, the overall endoscopic assessment of the neoplastic
scopic techniques (CE vs NBI). Mann-Whitney U test was per- character of the lesion was better for NBI (k Z .564) in
formed to assess differences in interobserver and comparison with HD CE (k Z .493, P < .001).
intraobserver agreement between different endoscopic tech- The median (range) intraobserver agreement between the
niques. Wilcoxon signed rank test was performed to assess dif- 2 assessments is depicted in Table 5. Although the range in
ferences between groups of endoscopic scoring. intraobserver agreement among the different observers was
Finally, agreement between endoscopic (Kudo pit pattern, wide, intraobserver agreement was moderate for the overall
overall endoscopic assessment) and histologic assessment pit pattern assessment and substantial for differentiating
(non-neoplastic vs neoplastic) was assessed using c2 statistics. neoplastic pit patterns from non-neoplastic pit patterns. In-
Sensitivity, specificity, positive predictive value (PPV), negative traobserver agreement was significantly worse for the com-
predictive value (NPV), and accuracy, described as median and plete Kudo pit pattern assessment with 6 different options
interquartile range, were also computed. All hypotheses were compared with all other combinations (Kudo I and II vs IIIL
tested at the 5% level of significance, and P < .05 was used for and IV vs IIIS and V; and Kudo I, II vs IIIL, IIIS, IV and V;
significance threshold. both P Z .005). The intraobserver agreement was similar

1102 GASTROINTESTINAL ENDOSCOPY Volume 86, No. 6 : 2017 [Link]

Bisschops et al Diagnostic accuracy of Kudo pit pattern classification in UC patients

Figure 1. Two images originating from the PowerPoint file for the first round. A, Image 1.22 was initially regarded as a Kudo I pit pattern during endos-
copy with narrow-band imaging, whereas final histology showed a hyperplastic polyp. B, Image 1.43 was initially regarded as a Kudo IIIL pit pattern during
endoscopy with chromoendoscopy, whereas final histology demonstrated low-grade dysplasia.

for images derived from endoscopic procedures with HD CE realize this criterion standard had 81.5% correlation with
and NBI (all P > .05). stereomicroscopic assessment.15 Interestingly, no
hyperplastic polyps were included in the Kudo study, but
only 2.8% of the adenomas exhibited a type I/II non-
DISCUSSION neoplastic pattern. A more recent image-based study assessed
the diagnostic accuracy of low-grade dysplasia in patients with
This is the first pilot study assessing the diagnostic accuracy inflammatory bowel disease based on nonmagnified HD CE.17
and interobserver agreement of both nonmagnified HD CE These authors had 30 lesions (13 low-grade dysplasia) assessed
and NBI of visible lesions detected during UC surveillance. by 17 endoscopists. In contrast to our findings, the interob-
This study assessed the applicability of Kudo’s pit pattern in server agreement to differentiate between neoplastic and
UC lesions and proposes a simplified dichotomous classifica- non-neoplastic lesions was only fair with a k Z .24. This may
tion of neoplastic versus non-neoplastic patterns. The pres- seem contradictory, but the endpoints of assessment were
ence of a pit pattern I or II has a good NPV to rule out different from our study. Indeed, in the study from Wanders
dysplasia, with an acceptable overall sensitivity of 77%. We et al,17 endoscopists had to indicate if the lesion was
have shown that the interobserver agreement for differenti- neoplastic or not. In our study we first asked to assess the
ating neoplastic from non-neoplastic pit patterns in patients pit pattern, keeping in mind that pattern I and II were non-
with longstanding UC is moderate to substantial for endoscop- neoplastic, and at the end the lesions needed to be classified
ists with extensive experience in CE and NBI. We also found as neoplastic or non-neoplastic. The diagnostic accuracy of
that the interobserver agreement for differentiation between identifying neoplastic lesions and pit pattern I/II versus others
non-neoplastic and neoplastic pit patterns is significantly bet- was almost identical, indicating that in our study the latter was
ter for NBI in comparison with HD CE, although the agree- the main driving force to classify a lesion and may possibly in-
ment for the individual pit patterns is better for HD CE. We crease the interobserver agreement. In addition, in our trial we
believe our data support the development of a prospective trial also included more advanced lesions.
in consecutive patients referred for UC surveillance to validate It is important to realize that pit pattern assessment may
the application of this simplified dichotomous classification of be challenging in the setting of UC with NBI. Matsumoto
the pit pattern using NBI and nonmagnified HD CE. et al18 already described that unlike the fact that type IIIL
In general, little is known about the diagnostic accuracy of and IV pit pattern are predictive of neoplasia in screening
nonmagnified HD CE in predicting histology of lesions de- colonoscopy, they could not find that a villous pattern in UC
tected during UC surveillance endoscopy. One study showed patients predicted neoplasia. When using NBI it seems to be
a high diagnostic accuracy of magnified standard resolution that protruding lesions and a tortuous pattern are suggestive
CE in 87 patients undergoing CE with methylene blue. The of neoplasia.18 Case reports also show the challenge of
prediction of neoplasia was based on the assessment of a differentiating inflammatory polyps exhibiting a type IV
type I or II pit pattern for non-neoplastic changes and type pattern.19 All studies assessing the diagnostic yield of CE for
III-V for the prediction of neoplastic changes. In the latter study colitis-associated neoplasia were performed in the absence
the single-operator sensitivity and specificity were 93% both, of active inflammation. This is a prerequisite for good-
with a PPV of 83% and NPV of 98%.5 Similarly, the original quality screening and to enable optical diagnosis based on
Kudo pit pattern was designed for magnifying endoscopy the pit pattern.
using a combination of indigo carmine and cresyl violet. It is Experience is an important issue in optical diagnosis and pit
interesting to look back at this original publication and pattern assessment. For standard polyp pit pattern assessment

[Link] Volume 86, No. 6 : 2017 GASTROINTESTINAL ENDOSCOPY 1103

Diagnostic accuracy of Kudo pit pattern classification in UC patients Bisschops et al

TABLE 2. Accuracy to predict histologic-proven neoplastic lesion

Sensitivity Specificity NPV PPV Accuracy

Overall (n Z 50)
I-II vs IIIL-IV-IIIS-V 77% (54%-85%) 68% (51%-84%) 88% (84%-94%) 46% (36%-61%) 70% (58%-82%)
Non-neoplastic vs neoplastic 77% (31%-100%) 69% (43%-92%) 90% (79%-100%) 48% (37%-67%) 72% (58%-84%)
CE (n Z 24)
I-II vs IIIL-IV-IIIS-V 88% (63%-100%) 63% (44%-81%) 89% (78%-100%) 57% (40%-73%) 73% (54%-88%)
Non-neoplastic vs neoplastic 88% (25%-100%) 59% (38%-94%) 91% (71%-100%) 55% (40%-78%) 71% (54%-88%)
NBI (n Z 26)
I-II vs IIIL-IV-IIIS-V 60% (40%-80%) 74% (57%-86%) 89% (86%-92%) 35% (25%-43%) 71% (58%-77%)
P vs CE P < .001 P Z .280 P Z .739 P < .001 P Z .481
Non-neoplastic vs neoplastic 60% (40%-100%) 76% (48%-90%) 89% (86%-100%) 38% (27%-78%) 73% (58%-81%)
P vs CE P Z .132 P Z .315 P Z .971 P Z .015 P Z .481
Values are medians with ranges in parentheses. The diagnostic accuracy is first calculated based on the dichotomous model pit pattern I-II vs other pit patterns and then based
on the overall impression of the lesion as assessed by the endoscopist (non-neoplastic vs neoplastic).
CE, Chromoendoscopy with methylene blue; NBI, narrow-band imaging; NPV, negative predictive value; PPV, positive predictive value.

TABLE 3. Diagnostic accuracy for neoplasia per level of confidence of the observer

Not sure (1-3) Sure (4-5) P value

Median sensitivity (range) 73% (0%-100%) 80% (43%-100%) .436

Median specificity (range) 66% (25%-100%) 73% (56%-89%) .353
Median NPV (range) 90% (50%-100%) 89% (81%-100%) .684
Median PPV (range) 35% (0%-100%) 58% (38%-70%) .052
Median accuracy (range) 21% (4%-50%) 77% (62%-84%) <.001
Values are medians with ranges in parentheses.
NPV, Negative predictive value; PPV, positive predictive value.

with magnifying endoscopy, the interobserver and intraob-

TABLE 4. Interobserver agreement
server kappa values were .716 and .810, respectively, for expe-
rienced Japanese endoscopists for perfectly selected images. Overall CE NBI
In the latter study the interobserver and intraobserver agree- (n [ 50) (n [ 24) (n [ 26)
ment for predicting histology was better than for pit pattern Kudo pit pattern k Z .282 k Z .322 k Z .224
analysis alone, which is similar to our findings.20 The P < .001 P < .001 P < .001
learning curve for optical diagnosis in UC is unknown. A
I-II vs. IIIL-IV vs. IIIS-V k Z .497 k Z .462 k Z .492
learning effect after formal training has been described for
P < .001 P < .001 P < .001
pit and vessel pattern analysis using CE and NBI with low
and high magnification. Pretraining assessment of 220 high- I-II vs. IIIL-IV-IIIS-V k Z .587 k Z .495 k Z .653
quality images and a formal 1-hour training on pit pattern P < .001 P < .001 P < .001
and vessel pattern classification improves the interobserver Non-neoplastic vs. neoplastic k Z .493 k Z .394 k Z .564
kappa value for endoscopists with some experience but does P < .001 P < .001 P < .001
not bring this to the level of highly experienced endoscopists. The P-value was generated by calculating the variance of kappa and deriving a z
Moreover, endoscopists without any experience in optical statistic.
diagnosis do not perform to the level of more experienced en- CE, Chromoendoscopy with methylene blue; NBI, narrow-band imaging.

doscopists, even after a formal training.21 This is also echoed

by Wanders et al,17 who showed a lower diagnostic accuracy with less experience assessing nonmagnified images. None-
for nonexperts in predicting low-grade dysplasia. theless, the findings of our pilot study are valuable because
In this study all endoscopists had been involved in previous the use of nonmagnifying endoscopy represents better real-
imaging studies, but with different levels of experience. In life endoscopy in general secondary care units. In addition,
addition, we used nonmagnified images and the pit pattern we have shown that using a dichotomous assessment of the
assessment in the setting of UC. These factors explain why pit pattern (types I and II vs the rest for prediction of neoplasia)
the overall kappa value was comparable with endoscopists obtained a relatively good sensitivity for diagnosing neoplastic

1104 GASTROINTESTINAL ENDOSCOPY Volume 86, No. 6 : 2017 [Link]

Bisschops et al Diagnostic accuracy of Kudo pit pattern classification in UC patients

TABLE 5. Intraobserver agreement

Overall (n [ 50) CE (n [ 24) NBI (n [ 26)

Kudo pit pattern k Z .507 (.270-.565) k Z .464 (.249-.671) k Z .412 (.152-.776)

I-II vs. IIIL-IV vs. IIIS-V k Z .642 (.373-.862) k Z .527 (.249-.934) k Z .686 (.226-.891)
I-II vs. IIIL-IV-IIIS-V k Z .685 (.436-.960) k Z .587 (.249-1.000) k Z .756 (.350-.906)
Non-neoplastic vs. neoplastic k Z .663 (.160-.960) k Z .577 (.216-1.000) k Z .738 (.025-1.000)
Values are median kappa with range in parentheses.
CE, Chromoendoscopy with methylene blue; NBI, narrow-band imaging.

lesions, even without magnifying endoscopy, which is compa- setting of nonmagnifying HD CE or NBI for assessment of
rable with the diagnostic accuracy obtained by highly experi- UC associated neoplasia is not perfect in a group of endoscop-
enced endoscopists for assessing standard polyps.21 We also ists with mixed levels of experience. It can be anticipated,
included a clinically realistic number of neoplastic lesions in based on available literature, that endoscopists without any
the image data set (26%), which is comparable with experience in optical diagnosis are unable to make the right
previously reported neoplasia detection rate in UC.5,10,11 call about pit pattern and final histologic diagnosis.17 This is
The level of certainty about the optical diagnosis needs to already the case for standard polyp assessment21 and most
be considered. For instance, for the Narrow-Band Imaging In- likely pertains in a setting of UC. Training programs should
ternational Colorectal Endoscopic classification, a high level of be provided, and in the absence of any experience,
confidence is necessary to make an optical NBI-based endoscopists performing UC surveillance should biopsy
diagnosis.22 Similarly, we found a significant improvement in sample any visible lesion for proper histologic diagnosis and
diagnostic accuracy when endoscopists were confident to get feedback on their own findings. As for more-
about the diagnosis. Although we did not assess this in the experienced endoscopists, it is reassuring to realize that the
study, one can speculate that the level of confidence is interobserver agreement is moderate to substantial to differen-
mainly determined by the quality of the image and to some tiate a non-neoplastic from a neoplastic pit pattern even
extent the presence of inflammation. We opted not to use without magnification and that the diagnostic accuracy for
the Narrow-Band Imaging International Colorectal Endo- this dichotomous assessment approaches expert levels. Never-
scopic classification because it has not been validated for UC theless, the system is not perfect, and therefore it is advised to
lesions. Unlike the fact that NBI has mainly been attributed take biopsy samples of any lesion unless the diagnosis of a very
to highlight vascular structures, we did show, however, that typical pseudopolyp with pit pattern type I or a clearly identifi-
nonmagnifying NBI of high-quality still images allows one to able small hyperplastic lesion is made with a high level of con-
distinguish between neoplastic and non-neoplastic pit pat- fidence. In cases of larger lesions and especially when the level
terns with a high interobserver and intraobserver agreement. of confidence is low, biopsy specimens should be taken for
The limitation of our study is that all image assessment was pathologic confirmation.
performed off-site and on selected still images by experts. It is In conclusion, a dichotomous pit pattern assessment after
not clear how this translates to assessment of videos or assess- nonmagnifying CE or NBI has a moderate to substantial inter-
ment during endoscopy by less-experienced endoscopists. observer agreement and a good diagnostic accuracy for differ-
Although the investigator who selected the images for this entiating neoplastic from non-neoplastic lesions in UC
study was otherwise not involved in either the original surveillance. Diagnostic accuracy improves significantly with
study or in the interobserver study, a selection bias is still the level of confidence. However, even with experience in op-
possible. Recall bias was probably limited because of the inter- tical diagnosis, taking biopsy specimens of lesions that are not
val of at least 10 weeks in between the 2 assessments. In addi- typical pseudopolyps or hyperplastic polyps remains neces-
tion, 1 of the assessors (R.B.) performed the original sary. The findings of this pilot study need to be confirmed in
endoscopy, and 2 assessors (R.B. and T.B.) were aware that a prospective study with a real-time scenario during ongoing
the same set of images were being used during the UC surveillance.
second evaluation. However, a sensitivity analysis excluding
these 2 assessors did not change the overall findings of the ACKNOWLEDGMENT
study. Also, size of the lesions was not taken into account as
a possible confounding factor, but in the case of UC even in- R.B. and M.F. are Senior Clinical Investigators for the
flammatory polyps can be very large. Finally, because of the Research Foundation–Flanders (FWO), Belgium.
relatively small sample size of the images, our findings need
to be confirmed prospectively, preferably in a real-time sce-
nario during ongoing surveillance.
REFERENCES
What can we learn from this pilot study for general endos-
copy practice? We clearly showed that interobserver agree- 1. Eaden JA, Abrams KR, Mayberry JF. The risk of colorectal cancer in ul-
ment and diagnostic accuracy of optical diagnosis in the cerative colitis: a meta-analysis. Gut 2001;48:526-35.

[Link] Volume 86, No. 6 : 2017 GASTROINTESTINAL ENDOSCOPY 1105

Diagnostic accuracy of Kudo pit pattern classification in UC patients Bisschops et al

2. Jess T, Loftus EV, Velayos FS, et al. Risk of intestinal cancer in inflamma- 21. Higashi R, Uraoka T, Kato J, et al. Diagnostic accuracy of narrow-band
tory bowel disease: a population-based study from olmsted county, imaging and pit pattern analysis significantly improved for less-
Minnesota. Gastroenterology 2006;130:1039-46. experienced endoscopists after an expanded training program. Gastro-
3. Jess T, Simonsen J, Jørgensen KT, et al. Decreasing risk of colorectal intest Endosc 2010;72:127-35.
cancer in patients with inflammatory bowel disease over 30 years. 22. Hayashi N, Tanaka S, Hewett DG, et al. Endoscopic prediction of deep
Gastroenterology 2012;143:375-81. submucosal invasive carcinoma: validation of the Narrow-Band Imag-
4. Rutter M, Saunders B, Wilkinson K, et al. Severity of inflammation is a ing International Colorectal Endoscopic (NICE) classification. Gastroint-
risk factor for colorectal neoplasia in ulcerative colitis. Gastroenter- est Endosc 2013;78:625-32.
ology 2004;126:451-9.
5. Kiesslich R, Fritsch J, Holtmann M, et al. Methylene blue-aided
Abbreviations: CE, chromoendoscopy; CRC, colorectal cancer; HD, high
chromoendoscopy for the detection of intraepithelial neoplasia
definition; NBI, narrow-band imaging; NPV, negative predictive value;
and colon cancer in ulcerative colitis. Gastroenterology
PPV, positive predictive value; UC, ulcerative colitis.
2003;124:880-8.
6. Marion J, Waye J, Present D. Chromoendoscopy-targeted biopsies are DISCLOSURE: The following authors disclosed financial relationships
superior to standard colonoscopic surveillance for detecting dysplasia relevant to this publication: R. Bisschops: Speaker for Pentax, Fujifilm,
in inflammatory bowel disease patients: a prospective. Am J Gastroen- Olympus, Covidien, Boston Scientific, Ferring, and Norgine; consultant
terol 2008;103:2342-9. for Pentax, Fujifilm, CDx Diagnostics, Covidien, Boston Scientific, and
7. Subramanian V, Mannath J, Ragunath K, et al. Meta-analysis: the diag- Norgine; research support recipient from Pentax, Fujifilm, Covidien,
nostic yield of chromoendoscopy for detecting dysplasia in patients and Cook. T. Bessissow: Speaker and consultant for Janssen, Abbvie,
with colonic inflammatory bowel disease. Aliment Pharmacol Ther Shire, Takeda, and Ferring. M. Ferrante: Research grant recipient
2011;33:304-12. from Takeda; speaker for Takeda, Abbvie, Boehringer-Ingelheim,
8. Cairns SR, Scholefield JH, Steele RJ, et al. Guidelines for colorectal can- Chiesi, Falk, Ferring, Janssen, Mitsubishi Tanabe, MSD, Tilotts, and
cer screening and surveillance in moderate and high risk groups (up- Zeria; consultant for Abbvie, Boehringer-Ingelheim, Ferring, Janssen,
date from 2002). Gut 2010;59:666-89. and MDS. E. Dekker: Research support recipient from Olympus and
9. Kami nski MF, Hassan C, Bisschops R, et al. Advanced imaging for detec- Fujifilm. J. E. East: Research support and advisory board member for
tion and differentiation of colorectal neoplasia: European Society of Olympus, Cosmo Pharmaceuticals, and Lumendi. K. Ragunath:
Gastrointestinal Endoscopy (ESGE) Guideline. Endoscopy 2014;46: Consultant and research support recipient from Olympus, Pentax,
435-49. Cook, Boston Scientific, Erbe, Astra Zeneca, and Ferring. P. Bhandari:
10. Pellisé M, López-Cerón M, Rodríguez de Miguel C, et al. Narrow- Research support recipient from Olympus, Pentax, and Fujifilm. M.
band imaging as an alternative to chromoendoscopy for the detec- Rutter: Honoraria from Pentax; research support recipient from
tion of dysplasia in long-standing inflammatory bowel disease: a Olympus. E. Schoon: Speaker for Medtronic and Boston Scientific;
prospective, randomized, crossover study. Gastrointest Endosc equipment loan from Fujifilm. F. Baert: Research support recipient
2011;74:840-8. from Abbott, Chiesi, Roche, and MSD; speaker and consultant for
11. Ignjatovic A, East JE, Subramanian V, et al. Narrow band imaging for Abbvie, Boehringer-Ingelheim, MSD, Mundipharma, Falk, Ferring,
detection of dysplasia in colitis: a randomized controlled trial. Am J Takeda, and Vifor. All other authors disclosed no financial
Gastroenterol 2012;107:885-90. relationships relevant to this publication.
12. Bisschops R, Bessissow T, Jospeh JA, et al. Chromo-endoscopy versus
Copyright ª 2017 by the American Society for Gastrointestinal Endoscopy
narrow band imaging in ulcerative colitis: a prospective randomized
0016-5107/$36.00
controlled trial. Gut. Epub 2017 Jul 11.
[Link]
13. Laine L, Kaltenbach T, Barkun A, et al. SCENIC international
consensus statement on surveillance and management of Received June 1, 2017. Accepted September 22, 2017.
dysplasia in inflammatory bowel disease. Gastroenterology
Current affiliations: Department of Gastroenterology and Hepatology,
2015;148:639-51.
University Hospital Leuven, Leuven, Belgium (1), Division of
14. Schlemper RJ, Riddell RH, Kato Y, et al. The Vienna classification of
Gastroenterology, Royal Victoria Hospital, McGill University Health Center,
gastrointestinal epithelial neoplasia. Gut 2000;47:251-5.
Montreal, Quebec, Canada (2), Department of Gastroenterology and
15. Kudo S, Tamura S, Nakajima T, et al. Diagnosis of colorectal tumorous
Hepatology, Amsterdam Medical Center, Amsterdam, The Netherlands (3),
lesions by magnifying endoscopy. Gastrointest Endosc 1996;44:8-14.
Translational Gastroenterology Unit, Nuffield Department of Medicine,
16. Altman D. Practical statistics for medical research. London, UK:
University of Oxford, Oxford, United Kingdom (4), NIHR Nottingham
Chapman & Hall; 1990.
Digestive Diseases Biomedical Research Unit City Hospital Campus,
17. Wanders LK, Mooiweer E, Wang J, et al. Low interobserver agreement
Nottingham University Hospitals NHS Trust, Nottingham, United Kingdom
among endoscopists in differentiating dysplastic from non-dysplastic
(5), Department of Gastroenterology, Spire Portsmouth Hospital,
lesions during inflammatory bowel disease colitis surveillance. Scand
Portsmouth, United Kingdom (6), Department of Gastroenterology,
J Gastroenterol 2015;50:1011-7.
University Hospital of North Tees, Stockton-on-Tees, Cleveland, United
18. Matsumoto T, Kudo T, Jo Y, et al. Magnifying colonoscopy with narrow
Kingdom, and School of Medicine, Durham University, Durham, United
band imaging system for the diagnosis of dysplasia in ulcerative colitis:
Kingdom (7), Department of Gastroenterology, Catherina Hospital,
a pilot study. Gastrointest Endosc 2007;66:957-65.
Eindhoven, The Netherlands (8), Wolfson Unit for Endoscopy, St. Mark’s
19. East J, Suzuki N, Herbay A Von, et al. Narrow band imaging with magni-
Hospital, London, United Kingdom (9), Medical Genomics Unit, BIO3,
fication for dysplasia detection and pit pattern assessment in ulcera-
GIGA-R, University of Liège, Liège, Belgium (10), and Department of
tive colitis surveillance: a case with multiple dysplasia associated.
Gastroenterology, AZ Delta, Roeselare, Belgium (11).
Gut 2016;55:1432-5.
20. Huang Q, Fukami N, Kashida H, et al. Interobserver and intra-observer Reprint requests: Raf Bisschops, MD, PhD, Department of Gastroenterology
consistency in the endoscopic assessment of colonic pit patterns. Gas- and Hepatology, University Hospital Leuven (UZ Leuven), KU Leuven,
trointest Endosc 2004;60:520-6. Herestraat 49, B-3000 Leuven, Belgium.

1106 GASTROINTESTINAL ENDOSCOPY Volume 86, No. 6 : 2017 [Link]

Bisschops et al Diagnostic accuracy of Kudo pit pattern classification in UC patients

Supplementary Figure 1. This figure shows the case report forms used
by the different investigators for assessing the Kudo Pit pattern. Kudo Pit
pattern I and II was regarded as non-neoplastic for the analysis. However,
observers also had to indicate if the lesion was neoplastic or not indepen-
dent from the Kudo pattern and the degree of certainty.

[Link] Volume 86, No. 6 : 2017 GASTROINTESTINAL ENDOSCOPY 1106.e1