Introduction
Patients with long-standing ulcerative colitis (UC) have an increased risk of colorectal
cancer (CRC), which can develop through the dysplasia-carcinoma sequence. Therefore,
periodic colonoscopies are performed to reduce CRC-related morbidity and mortality
by detecting and removing neoplasia at the earliest stage [1].
Neoplastic lesions of colorectal mucosa in UC were formerly described as flat (endoscopically
invisible) or elevated (endoscopically visible). More recently, gross morphology of
neoplastic lesions in the colon and rectum has been better described as polypoid and
non-polypoid, according to the Paris classification [2], but it has been used in very few studies in inflammatory bowel disease (IBD). On
the other hand, similar macroscopic aspects can be seen in inflammatory, non-neoplastic,
lesions, which can be even more frequent in IBD and often have been excluded from
previous studies assessing diagnostic accuracy of surveillance endoscopy in UC [3].
Early international guidelines for colonoscopic surveillance in IBD used white light
endoscopy (WLE) and recommended performing targeted biopsies of any visible suspected
neoplastic lesion as well as multiple random biopsies along the colorectal mucosa
to detect those neoplastic lesions arising in otherwise apparently normal mucosa,
which were more difficult to see with standard-resolution endoscopes [1].
Recent availability of powered endoscopy, such as high-definition equipment and dye-based
chromoendoscopy, has improved detection of neoplastic lesions in IBD, including the
flat and non-polypoid forms [4]
[5]. Moreover, dye-based chromoendoscopy, when added to magnification endoscopy to better
define the mucosal crypt architecture according to the pit-pattern classification
of Kudo [6], has been used to differentiate neoplastic and non-neoplastic lesions, although
the accuracy of this classification in the specific setting of IBD is still controversial
due to inflammatory and regenerative artifacts [3].
International guidelines suggest methylene blue or indigo carmine chromoendoscopy
as the preferred method for surveillance colonoscopy [1]
[7]. On the other hand, so-called “virtual” chromoendoscopy, which includes narrow band
imaging (NBI, Olympus), i-SCAN (Pentax), Fuji Intelligent Color Enhancement (FICE,
Fujifilm) and Blue-laser imaging (BLI, Fujifilm), has only recently been included
for routine use in the latest European Society of Gastrointestinal Endoscopy guidelines
[8], despite the paucity of ad hoc controlled studies in IBD [9] and the failure of early controlled studies with unmagnified NBI compared to standard
WLE or dye-based chromoendoscopy [8].
Recently, we published the first experience with FICE in IBD, which focused only on
visible polypoid and non-polypoid lesions found during surveillance in UC [3]. Through a post-hoc, retrospective analysis of endoscopic predictors of neoplasia,
that study showed the need to develop and validate a more accurate endoscopic classification
system for characterization of raised lesions in IBD, based not only on pit-pattern
visualization but also on inflammatory markers. This new classification, specific
for FICE in IBD, was prospectively used for the first time in this parallel study,
in which FICE was compared to WLE for differentiation of all visible lesions found
in consecutive patients with UC scheduled for surveillance colonoscopy.
Patients and methods
Study design and patients
This was a prospective, observational, parallel study in which consecutive outpatients
with long-lasting UC who were scheduled for surveillance colonoscopy at our center,
were submitted to a withdrawal colonoscopy with WLE or FICE ([Fig. 1]), depending on the availability of study instruments (only one FICE was available)
at the time of the scheduled endoscopy.
Fig. 1 Study design and allocation of patients according to the endoscopic method and the
random versus targeted protocol of sample collection.
The study was approved by our Ethical Committee and patients signed an informed consent
form to undergo endoscopic examination.
Inclusion criteria were a previous confirmed diagnosis of UC according to clinical,
endoscopic, and histological data; disease duration of at least 8 years since onset
of symptoms; no or mild clinical activity according to a Mayo score ≤ 4; and presence
of at least one visible polypoid or non-polypoid lesion during the surveillance colonoscopy,
according to the Paris classification [2].
Exclusion criteria were proctitis, coagulopathy, pregnancy, melanosis coli, previous
colorectal resection, previous colonoscopy in the last 3 months with unresected neoplasia,
massive pseudopolyposis (> 30 polyps), poor bowel preparation (Boston Bowel Preparation
Score < 2 in any segment), and use of dye-based chromoendoscopy or virtual chromoendoscopy
other than FICE.
The following demographic and clinical data were collected at baseline: sex, date
of birth, disease onset, disease diagnosis, disease extent, personal/familiar history
of colorectal neoplasia, clinical activity according to the Mayo score, extra-intestinal
manifestations including primary sclerosing cholangitis (PSC), and previous and concomitant
medications.
According to the parallel design of our study, patient enrolment stopped after two
comparable groups of patients were obtained according to all significant demographic,
clinical, and endoscopic variables, including the rate of neoplastic lesions, in order
to decrease the risk of selection and detection bias in a study focused on differential
diagnosis rather than detection rate of neoplasia.
Endoscopic protocol
In the WLE arm, standard white light colonoscopy was performed using the Olympus CV-180
Evis Exera II system (Olympus Corp., Tokyo, Japan). At the time of study conception
(2012), this was our standard of care for endoscopic surveillance of IBD, with random
biopsies performed every 10 cm plus targeted biopsies of suspected neoplastic lesions.
FICE was then acquired by our endoscopic unit and tested in the context of a plan
to validate its use in IBD.
In the FICE arm, virtual chromoendoscopy was performed with the high-definition, magnified,
Fujinon EG-590ZW colonoscope (Fujinon Corp, Saitama, Japan), equipped with the EPX4400
processor. The FICE set number 4 was used exclusively, according to previous validation
studies [3]
[10]. Magnification was activated according to the endoscopist’s need, based on his confidence
to assess the mucosal and vascular patterns of each lesion in supporting his differential
diagnosis.
In both arms, assessment of the colon to search for visible lesions was performed
systematically during withdrawal of the instrument with a minimum diagnostic extubation
time of 10 minutes.
Classification of lesions
Morphology of each lesion was determined in accordance with the Paris classification
of macroscopic aspects as polypoid or non-polypoid [2].
In the WLE arm, areas of suspected neoplasia were defined as any mucosal irregularity,
ulceration, polypoid or non-polypoid lesion that was not entirely consistent with
chronic or active UC according to the usual clinical practice [11].
In the FICE arm, neoplasia was first suspected according to the conventional Kudo
classification (FICE-KUDO), and then according to the modified Kudo classification
for FICE in IBD (FICE-KUDO/IBD) which was developed in our previous pilot study adding
three specific endoscopic modifiers for risk of neoplasia [3]. In particular, as shown in [Fig. 2], Kudo pit-patterns I and II, which are usually associated with non-neoplastic lesions
in non-IBD studies, were considered neoplastic lesions if associated with the presence
of at least one of two further endoscopic factors:
Fig. 2 Schematic classification of lesions suspected and not suspected for neoplasia according
to the modified Kudo classification, based on type and heterogeneity of pit-patterns
and presence of a fibrin cap or visible microvessels (see text).
(1) irregular brown/bluish visible capillary vessels, defined according to a Teixeira
III-V vascular pattern, which were previously described in studies with FICE in the
general population and were related to higher risk of neoplasia [3]
[12]; and (2) pits heterogeneity, defined as variable density or size of pits in the
context of low-risk (type I and II) patterns, which was previously associated with
higher risk of neoplasia [3]
[13]
[14].
In contrast, Kudo pit-patterns III-IV, usually related to neoplasia in non-IBD studies
[6], were considered non-neoplastic lesions if associated with a fibrin cap, defined
as a non-removable, circumscribed, white exudate covering at least 25 % of the surface
of the lesions, which has been associated with inflammatory non-neoplastic lesions
in previous studies in IBD [3]
[15]
[16].
Kudo V and III-s pit-patterns, usually associated with higher risk of advanced lesions,
including invasive cancer, were always considered suspicious for neoplasia, independent
of presence of the three endoscopic modifiers.
Finally, lesions not classified with enough confidence by the endoscopist according
to the conventional Kudo classification (which accounted up to 15 % of visible lesions
in our previous study [3]) were considered neoplastic, except in the presence of fibrin cap, which have been
associated with non-neoplastic inflammatory lesions.
Biopsy protocol
In both arms, three types of histological samples were obtained: 1) every 10 cm, random
biopsies of otherwise normal flat mucosa, from the cecum to the rectum; 2) targeted
biopsies or full endoscopic resection of visible suspected neoplastic lesions (polypoid
or non-polypoid), as appropriate; and 3) targeted biopsies of unsuspected neoplastic
lesions (polypoid or non-polypoid).
Neoplastic changes were classified according to the new Vienna classification as low-grade
intramucosal, high-grade intramucosal, or invasive neoplasia [17].
Non-neoplastic lesions, including hyperplasic and inflammatory polyps, were described
according to current classifications.
Statistical methods
At the time this study was conceived (2012), we did not find studies comparing FICE
with WLE or other types of chromoendoscopy in IBD, which precluded rigorous sample
size calculation. Moreover, previous controlled studies with other chromoendoscopy
techniques focused on dysplasia detection rather than differential diagnosis of visible
lesions [4]. Therefore, according to our parallel study design, we decided a priori to collect
at least 100 visible lesions in each arm, stopping the enrollment when all significant
demographic, clinical, and endoscopic variables between the two groups were comparable;
this included, by the way, a not significantly different rate of neoplastic visible
lesions.
The diagnostic performance of FICE (according to the two classifications) and WLE
was evaluated by comparing the endoscopic evaluation (suspected or unsuspected neoplasia)
with the histological diagnosis (reference test).
The analyses were primarily performed per lesion, since sample size and enrollment
of patients were based on the number of visible lesions. A per patient analysis was
also included to permit the reader to compare our results to those from previous studies
that used the same approach.
Sensitivity and specificity 95 % confidence intervals (Cis) were estimated by applying
the binomial exact method. Positive and negative likelihood ratio (LR) and positive
(PPVs) and negative predictive values (NPVs) were also calculated. For NPV, the Preservation
and Incorporation of Valuable Endoscopic Innovations (PIVI) threshold set at 90 %
or higher was used for adopting real-time endoscopic assessment of histology, although
our sample had no limitations on polyp size [18].
Data were described using means and standard deviation, or medians and interquartile
range, when appropriate. To test the difference between WLE and FICE, the z test was applied and the two tails probability reported.
Statistical analyses were conducted using IBM-SPSS for Windows 24th version and Excel
2013.
Results
Patients and lesions
Between September 2012 and October 2014, 329 patients with UC underwent surveillance
colonoscopy for long-standing UC. Eighty-four patients were excluded according to
various exclusion criteria ([Fig. 3]). Among the remaining 245 patients, 145 had no visible lesions while 100 patients
had at least one visible lesion (total number of lesions 275, mean number per patient
2.7) during their colonoscopy with WLE (n = 54; 139 lesions) or FICE (n = 46; 136
lesions).
Fig. 3 Study population overview.
Both groups of patients had comparable baseline characteristics ([Table 1]), including similar percentages of neoplastic lesions (n = 14 in both arms, 10 %)
and patients with neoplasia (n = 11 in both arms; 20 % vs 24 %), as required by our
study design.
Table 1
Characteristics of patients included in the two groups of surveillance endoscopy.
|
WLE
(54 patients, 139 lesions)
|
FICE
(46 patients, 136 lesions)
|
|
Age – mean (range), years
|
56 (26 – 84)
|
54 (34 – 73)
|
|
Male/female – ratio
|
42/12
|
32/14
|
|
Disease duration – mean (range), years
|
19 (8 – 36)
|
19 (8 – 44)
|
|
Extensive colitis versus left colitis – rate
|
67 %
|
78 %
|
|
Endoscopic activity (Mayo ≥ 1) – rate
|
54 %
|
61 %
|
|
|
17 %
|
26 %
|
|
|
22 %
|
20 %
|
|
|
15 %
|
15 %
|
|
Familiarity for CRC – rate
|
17 %
|
18 %
|
|
Primary sclerosing cholangitis – rate
|
0 %
|
0 %
|
|
Site of lesions – number (rate)
|
|
|
11 (8 %)
|
11 (8 %)
|
|
|
12 (9 %)
|
22 (16 %)
|
|
|
23 (16 %)
|
24 (18 %)
|
|
|
35 (25 %)
|
22 (16 %)
|
|
|
46 (33 %)
|
48 (35 %)
|
|
|
12 (9 %)
|
9 (7 %)
|
|
Size of lesions – median (range), mm
|
5 (2 – 20)
|
5 (2 – 30)
|
|
Neoplastic lesions – number (rate)
|
14 (10 %)
|
14 (10 %)
|
|
Site of neoplastic lesions – number (rate)
|
|
|
4 (29 %)
|
2 (14 %)
|
|
|
2 (14 %)
|
2 (14 %)
|
|
|
1 (7 %)
|
0 (0 %)
|
|
|
3 (21 %)
|
7 (50 %)
|
|
|
3 (21 %)
|
2 (14 %)
|
|
|
1 (7 %)
|
1 (7 %)
|
|
Size of neoplastic lesions – median (range), mm
|
5 (3 – 12)
|
7 (3 – 30)
|
|
Inflammatory lesions – number (rate)
|
117 (84 %)
|
104 (76 %)
|
|
Patients with neoplasia – number (rate)
|
11 (20 %)
|
11 (24 %)
|
P = not significant for all variables.
All neoplastic lesions had low-grade dysplasia, except for one adenocarcinoma in the
FICE group. Non-neoplastic lesions were inflammatory lesions except for 26 hyperplasic
lesions (FICE n = 18, WLE n = 8).
Two-hundred and sixty-six lesions were polypoid (WLE n = 138, FICE n = 128) and nine
were non-polypoid (WLE n = 1, FICE n = 8; P = 0.016). All non-polypoid lesions in the FICE group were type IIa, while the only
non-polypoid lesion in the WLE arm (non-neoplastic) was type IIb.
Frequency of neoplasia in random biopsies of flat mucosa
In the WLE arm, 1011 random biopsies were taken (mean per patient 19 ± 4), while 992
were taken during withdrawal with FICE (mean per patient 22 ± 5). No neoplasia was
found in either arm ([Table 2]).
Table 2
Rate of neoplasia according to different types of visible lesions (polypoid vs non-polypoid,
suspected for or not suspected for neoplasia, flat or raised visible), and endoscopic
criteria for differential diagnosis.
|
Flat normal mucosa
|
Visible lesions
|
|
Polypoid
|
Non-polypoid
|
|
Suspected
|
Not suspected
|
Suspected
|
Not suspected
|
|
WLE
|
Number of lesions
|
1011
|
26
|
112
|
1
|
0
|
|
Dysplasia
|
0 (0 %)
|
9 (35 %)
|
5 (4.5 %)
|
0 (0 %)
|
0 (0 %)
|
|
FICE-KUDO
|
Number of lesions
|
992
0 (0 %)
|
29
|
99
|
7
|
1
|
|
Dysplasia
|
8 (28 %)
|
1 (1 %)
|
5 (71 %)
|
0 (0 %)
|
|
FICE-KUDO/IBD
|
Number of lesions
|
11
|
117
|
6
|
2
|
|
Dysplasia
|
8 (73 %)
|
1 (0.9 %)
|
5 (83 %)
|
0 (0 %)
|
WLE, white-light endoscopy; FICE, Fuji Intelligent Color Enhancement; KUDO, Kudo classification;
IBD, inflammatory bowel disease.
Differentiation of neoplastic and non-neoplastic lesions in targeted biopsies
Among the 275 visible lesions analyzed in the WLE (n = 139) and FICE (n = 136) arm,
27 (19 %), 17 (12.5 %) and 36 (26 %) were suspected for neoplasia by WLE, FICE-KUDO/IBD,
and FICE-KUDO. The number of lesions suspected for neoplasia was significantly higher
using FICE-KUDO than FICE-KUDO/IBD (P = 0.004) ([Table 2]).
Within all inflammatory lesions, the number of false positives was similar between
FICE-KUDO (18 /104) and WLE (17/117), but significantly less with FICE-KUDO/IBD (3/104)
than with the other two methods (P = 0.001 and P = 0.003, respectively). In contrast, no differences were found between the two FICE
arms and WLE in rates of false positives and negatives within hyperplasic and neoplastic
lesions, respectively.
As shown in [Table 3], the sensitivity of both FICE methods was higher than that for WLE (93 % vs 64 %),
but not significantly by per lesion analysis (P = 0.065). Nonetheless, the specificity, positive-LR and PPV of FICE-KUDO/IBD compared
with WLE were 97 % vs 86 % (P = 0.002), 28.3 vs 4.5 (P = 0.001) and 76 % vs 33 % (P = 0.005), respectively, while the negative-LR was 0.07 vs 0.42 (P = 0.092) and the NPV was 99 % vs 96 % (P = 0.083). FICE-KUDO/IBD had also higher specificity (97 % vs 81 %; P = 0.001), positive-LR (28.3 vs 4.9; P = 0.001) and PPV (76 % vs 36 %; P = 0.006) than FICE-KUDO, while having similar sensitivity and NPV. Notably, NPV was
above the PIVI threshold (96 % to 99 %) with all three methods, without significant
differences.
Table 3
Diagnostic performance of WLE versus FICE (according to the Kudo or the new classification)
in the protocol of targeted sample of visible lesions (per lesion and per patient
analysis).
|
Per Lesion
|
Per Patient
|
|
WLE
|
FICE-NEW
|
FICE-KUDO
|
WLE
|
FICE-NEW
|
FICE-KUDO
|
|
Sensitivity (95 %CI)
|
0.64
(0.35 – 0.87)
|
0.93
(0.66 – 0.99)
|
0.93
(0.66 – 0.99)
|
0.54
(0.23 – 0.83)
|
0.91
(0.59 – 0.99)
|
0.91
(0.59 – 0.99)
|
|
Specificity (95 %CI)
|
0.86
(0.79 – 0.92)
|
0.97
(0.94 – 0.99)[1]
,
[2]
|
0.81
(0.74 – 0.88)
|
0.65
(0.49 – 0.79)
|
0.91
(0.77 – 0.98)[1]
|
0.71
(0.54 – 0.85)
|
|
Positive likelihood ratio (95 %CI)
|
4.5
(2.5 – 8.0)
|
28.3
(10.7 – 75.1)[1]
,
[2]
|
4.9
(3.3 – 7.3)
|
1.6
(0.8 – 3.1)
|
10.6
(3.5 – 31.8)[1]
,
[2]
|
3.2
(1.8 – 5.5)
|
|
Negative likelihood ratio (95 %CI)
|
0.42
(0.2 – 0.8)
|
0.07
(0.01 – 0.49)
|
0.09
(0.01 – 0.96)
|
0.7
(0.35 – 1.4)
|
0.1
(0.01 – 0.65)
|
0.1
(0.02 – 0.83)
|
|
Positive predictive value (95 %CI)
|
0.33
(0.22 – 0.47)
|
0.76
(0.55 – 0.90)[1]
,
[2]
|
0.36
(0.28 – 0.46)
|
0.29
(0.17 – 0.44)
|
0.77
(0.53 – 0.91)[1]
|
0.50
(0.36 – 0.64)
|
|
Negative predictive value (95 %CI)
|
0.96
(0.91 – 0.98)
|
0.99
(0.95 – 0.99)
|
0.99
(0.94 – 0.99)
|
0.85
(0.74 – 0.92)
|
0.97
(0.83 – 0.99)
|
0.96
(0.79 – 0.99)
|
WLE, white-light endoscopy; FICE, Fuji Intelligent Color Enhancement; NEW, new classification;
KUDO, Kudo classification
1
P < 0.05 vs WLE
2
P < 0.05 vs FICE-KUDO
Similar results were obtained in the per patient analysis ([Table 3]). Notably, NPVs were not significantly different among the three methods but were
above the PIVI threshold only with FICE, irrespective of the modified Kudo classification
(96 – 97 % vs 85 %).
Discussion
Our study is the first prospective trial evaluating FICE in real-world endoscopic
surveillance for UC. Using a pragmatic study design that included multiple morphological
and histological types of lesions, our study demonstrates that FICE is more accurate
than standard WLE in differentiation of visible lesions only if it is used with a
modified Kudo classification adapted for IBD.
The most original feature of our study was use, for the first time in a prospective
study, of a modified Kudo classification, which we previously reported in a pilot
study based on systematic analysis of 205 raised lesions found in 59 consecutive patients
with long-standing UC [3]. In that study, accuracy of the conventional Kudo classification was unsatisfactory,
but a retrospective post-hoc analysis revealed that it would have been significantly
improved after the addition of further endoscopic markers: presence of a fibrin cap,
significantly associated with inflammatory lesions, reclassified as unsuspected for
neoplasia those lesions unclassified according to conventional Kudo or type IIIL-IV
suspected neoplastic lesions, while presence of irregular vessels or pits heterogeneity
reclassified a lesion as suspicious for neoplasia despite type-I or type-II Kudo pit
patterns.
The need to develop new classifications, specific for surveillance in IBD, is not
new, although no specific conclusions have been drawn. Previous studies using dye-based
chromoendoscopy [4]
[19]
[20] and NBI [21]
[22] showed that the conventional Kudo classification of mucosal pit-patterns used in
the general population is not accurate enough to differentiate between neoplastic
and non-neoplastic lesions in IBD. More recently, in the FACILE classification developed
for i-SCAN in IBD [23], the Kudo criteria were not integrated in the final multivariate analysis of endoscopic
criteria of neoplasia to avoid uncertainty in the absence of magnified endoscopy and
in the presence of regenerative changes of the mucosa. However, the Kudo classification
is often used by Eastern endoscopists and by many European endoscopists who translate
their usual approach for CRC screening in the general population to patients with
IBD.
The debate on accuracy and reproducibility of endoscopic markers of neoplasia has
an important clinical and scientific implication, because variability in diagnostic
performance of various endoscopic methods of surveillance in IBD (with positive results
reported mostly with dye-based chromoendoscopy rather than virtual chromoendoscopy)
can be explained, at least in part, by limitations in the classification system used
to identify lesions as suspicious or not suspicious for neoplasia, rather than by
the imaging technology itself. Notably, if our study had used the conventional Kudo
classification, it would have been a further failure for virtual chromoendoscopy in
IBD, as was the case in most previous trials with non-magnified NBI [8]
[21]: the accuracy of FICE with conventional Kudo was in fact similar to WLE.
The need to critically reanalyze the role of chromoendoscopy based on more specific
rules in IBD depends on the different epidemiology of endoscopically visible lesions
compared to the screening programs for CRC in the general population.
First, the most frequent visible lesions in IBD are inflammatory. This was confirmed
by our study, in which approximately 80 % of lesions found in our consecutive patients
were inflammatory. Second, inflammatory lesions are more likely to be false positives
for neoplasia, according to conventional criteria developed for WLE or the Kudo classification.
Therefore, their prevalence in the sample of any study on chromoendoscopy in IBD can
strongly influence the accuracy of the tested diagnostic methods. This was clearly
shown in our study, in which WLE and FICE-KUDO had an overall similar rate of false
positives, while FICE-KUDO/IBD was better than the other two methods only in the high
number of inflammatory lesions.
In our opinion, inclusion of inflammatory lesions is a crucial point for any study
testing the accuracy of advanced endoscopy in IBD. Unfortunately, one limitation of
previous studies on surveillance endoscopy was the frequent exclusion (or partial
inclusion) of inflammatory lesions. That could have led not only to simplification
of the real endoscopic scenario in IBD but also to overestimation of the accuracy
of endoscopy, because inflammatory lesions are often difficult to differentiate from
neoplastic lesions, based on a bias of pit patterns related to the effect of the inflammatory
infiltrate of the mucosa [3]
[24].
Thanks to our inclusion criteria, this study suggests a well-defined role for FICE
in differential diagnosis of any type of lesion in UC (polypoid or non-polypoid),
in that it correctly characterized 93 % of neoplasms and 97 % of non-neoplastic lesions.
With higher specificity than WLE and the best combination of positive and negative
likelihood ratios (with a positive LR > 10, a negative LR < 0.1 and a + LR/-LR ratio
around 400), FICE-KUDO/IBD appears to be the most accurate diagnostic method in our
sample.
In our study, the greatest contribution of FICE-KUDO/IBD was its specificity (91 %),
which was competitive with the variable rates reported in IBD with dye-based chromoendoscopy
(57 % to 97 %) [4]
[19]
[25]
[26] and quite higher than the range reported with not-magnified NBI (66 %-81 %) [22]
[27]
[28]. From a clinical point of view, such a rate can appear high enough to support endoscopists
in assessment of multiple lesions throughout the colon in patients with IBD. Moreover,
the very high (97 % to 99 %) NPV of FICE-KUDO/IBD in both per-patient and per-lesion
analysis addressed the 90 % threshold established by the ASGE for accuracy required
for a “diagnose and leave” strategy for more frequent non-neoplastic lesions, although
that strategy requires further study specifically in neoplasia associated with IBD
[18].
The high sensitivity (91 %) reported with FICE-KUDO/IBD in our study is higher than
the pooled sensitivity (83 %) shown in patients with IBD in a recent meta-analysis
on dye-chromoendoscopy [25], and clearly higher than the rate (75 % to 76 %) reported with non-magnified NBI
in IBD [22]
[27], although direct comparison is not possible due to different inclusion criteria
and study design. Nevertheless, the high sensitivity did not reach statistical significance
as compared with WLE. We believe that may be due to the experience of the endoscopists
at our IBD tertiary center even when using WLE. On the other hand, the absolute high
sensitivity of FICE may be due to the different resolutions of the two imaging systems.
High-definition plus magnification imaging used with FICE can better analyze the mucosal
surface than can the standard definition used in our WLE arm, as previously described
in patients who did not have IBD [29], but less clearly in patients with IBD [9]
[29]
[30]
[31]
[32]. Certainly, further studies are needed in independent cohorts comparing high-definition
endoscopy with or without FICE-chromoendoscopy.
Our study was not conceived to analyze the detection rate for the two endoscopic imaging
methods because of its parallel design. Rather, it was developed to study the accuracy
of differential diagnosis in two comparable populations. Therefore, the rate of neoplastic
lesions had to be similar at the end of the non-randomized, prospective, patient allocation
and the detection rate could not be an independent outcome. However, within visible
lesions, our study seems to suggest that FICE can more easily detect non-polypoid
lesions because a higher number of such lesions, among all consecutive spontaneous
visible lesions included, were found compared with the WLE group. Recently, non-polypoid
morphology was found to be an independent endoscopic marker of neoplasia in IBD in
a study using high-definition endoscopy with i-SCAN [23]. Globally, our data suggest higher accuracy of FICE than WLE for lesions that are
more difficult to visualize and clinically important.
Notably, this also was recently underscored for dye-based chromoendoscopy in a meta-analysis
comparing it with WLE. Dye-based chromoendoscopy was superior to WLE in detecting
non-polypoid (but not polypoid) dysplastic lesions [33]. Conversely, no additional role was found for random biopsies when systematic analysis
of the entire colorectal mucosa was supported by powered endoscopy. Despite the significant
number random biopsies done in our study, and their related costs and time, no false
negatives for neoplasia were found.
Our study has several limitations. First, it was performed in a single center using
a modified Kudo classification, which has never been validated before outside our
IBD center. Therefore, a multicenter study on intraobserver and interobserver agreement
is ongoing. Second, the study was not randomized and the timely patient allocation
was not homogeneous. The parallel study design may be associated with more potential
bias in low sample sizes compared with crossover studies in which testing is done
on the same lesions with both techniques. In our opinion, however, it would be impossible
to perform randomized crossover studies because of the many clinical, endoscopic,
and technical variables that would have to be reconciled in this complex clinical
setting. On the other hand, in our study, all efforts were made to ensure that the
case-control groups were comparable, including comparable lesions from a quantitative
and qualitative point of view.
Conclusion
In conclusion, our study suggests that a protocol based only on targeted biopsies
of suspected neoplastic lesions appears to be the best approach we can perform with
a virtual chromoendoscopy technique like FICE in differentiation of polypoid and non-polypoid
lesions in UC, if the adapted IBD specific classification, such as our modified Kudo
classification, is used. Further studies in independent cohorts are required to show
the reproducibility of such a classification and to confirm its diagnostic yield when
used with other chromoendoscopy techniques and/or high-definition instruments.
