Introduction
The development of endoscopy has greatly benefited diagnosis of colorectal neoplasia.
Chromoendoscopy has been used to establish the pit pattern classification, which is
effective for differentiation between non-neoplasia and neoplasia as well as between
adenoma and cancer [1]. Narrow-band imaging (NBI) was developed in 2006; this technique uses superficial
tissue structures and emphasizes the imaging of certain features [2], such as vascular and mucosal patterns. Endocytoscopy (EC) was more recently developed
and can magnify objects by 380-fold to 450-fold. EC enables on-site observation of
structural and nuclear irregularities [3]
[4], and therefore has the potential to allow for “optical biopsy” [5]. The qualitative and quantitative usefulness of superficial microvascular findings
in the examination of colorectal lesions with NBI magnification endoscopy (NBI-ME)
has been reported by many institutions [6]
[7]
[8]
[9], but the utility of EC with NBI (EC-NBI) has not been demonstrated.
Kudo et al. [10] classified the endocytoscopic vascular pattern (EC-V) into 3 types ([Fig. 1]): obscure surface microvessels (EC-V1), clearly observed surface microvessels of
a uniform caliber and arrangement (EC-V2), and dilated surface microvessels of a non-homogeneous
caliber or arrangement (EC-V3). EC-V1 mainly corresponds to non-neoplasia, EC-V2 to
neoplasia, and EC-V3 to invasive cancer. However, whether the vascular findings of
EC-V3 are irregularities remains unclear.
Fig. 1 Endocytoscopic vascular pattern (EC-V) classification.
The aims of this study were to examine the vascular findings of submucosal deep invasive
cancer by EC-NBI with a new video processor system (EVIS LUCERA ELITE SYSTEM; Olympus,
Tokyo, Japan) and to determine the clinical significance of these ultra-high magnification
findings.
Patients and methods
Patients
We retrospectively analyzed 98 patients who underwent endoscopic or surgical resection
after observation with EC-NBI from May 2013 to December 2014. We excluded patients
with inflammatory bowel disease and those with hyperplastic polyps (including sessile
serrated adenomas/polyps) and cancers deeper than T2. Before the examination, patients
underwent bowel preparation with 2 L to 3 L of polyethylene glycol solution. Diazepam
and butylscopolamine were used intravenously for sedation and prevention of peristalsis.
The study took place at the Digestive Disease Center of Showa University Northern
Yokohama Hospital. The protocol was approved by the Medical Ethics Committee of our
hospital (No. 1405-03; approved on June 6, 2014) and registered in the University
Hospital Medical Information Network Clinical Trials Registry (UMIN-CTR000014033;
approved on May 25, 2014). All participants gave written informed consent, and the
study was conducted according to the Declaration of Helsinki.
EC system
The endocytoscope (CF-Y0020; Olympus) had a magnification range of 80 to 380 × with
a standard video processor system (EVIS LUCERA ELITE SYSTEM; CV-290/CLV-290SL) and
a digital image filing system (Solemio; Olympus). The EC-NBI was set at enhancement
mode A8 and color mode 3. The endocytoscope had a working length of 133 cm, outer
diameter of 13.6 mm, and single lens. The instrument allowed gradual magnification
at the center of the monitor, thus pinpointing the target area being viewed. Our observation
focused on the area of interest, which showed irregularity after conventional NBI
magnification before staining. Next, we checked the superficial layers of polyps as
thoroughly as possible.
Evaluation of EC-NBI findings
We anticipated 5 different types of vascular findings in light of the detailed vascular
findings commonly seen in submucosal deep invasive cancer: 1) caliber dilatation;
2) abnormal tortuosity and branching of microvessels; 3) loss of the micro-network
(MN) pattern; 4) caliber change; and 5) beaded sign. First, we defined caliber dilation
as vessel thickening > 2-fold (> 30 µm) that seen in adenomas. We reported a mean
blood vessel caliber size measured by EC of 19.5 ± 4.2 µm in adenoma and 33.0 ± 7.2 µm
in T1 cancer [11]. We therefore set a threshold caliber of 30 µm, which was almost twice that for
adenoma vessels, to allow us to detect caliber dilation. As for abnormal tortuosity
and branching of microvessels, vessels are usually arranged along the crypt opening
in adenoma. Abnormal tortuosity and branching was defined as sharply angled microvessels
or thick, irregular vessels passing from a deep layer to join a surface vessel. We
defined the MN pattern as the presence of fine and tortuous vessels in the intervening
part around the crypt opening of adenomas [12] ([Fig. 2]). However, the MN pattern tended to disappear in invasive carcinomas; thus, we regarded
loss of the MN pattern as suggestive of invasive carcinoma. Caliber change was defined
as a maximum caliber size of at least double the minimum caliber size in 1 vessel.
The vascular thickness is relatively uniform in adenomas, but seemed to become less
uniform in cases of submucosal deep invasive cancer. The beaded sign was defined as
blood vessels with the appearance of a string of beads ([Fig. 3]). The blood vessel diameter was measured using Nitenkeisokuki version 1.1.1. and
was evaluated relative to an endoscopic screen width of 600 μm. Endoscopic images
were randomly allocated to 2 readers (H. N. and M. M.) for evaluation without a final
pathologic diagnosis.
Fig. 2 Loss of the micro-network (MN) pattern. a Image of typical adenoma. b Schema of MN pattern; typical adenoma has fine and tortuous vessels in the intervening
part (α in dotted line) around the crypt opening (β). We defined these vessels as
the MN pattern. c Image of submucosal deep invasive cancer; loss of the MN pattern.
Fig. 3 Microvascular findings and schema.
NBI-ME classification
We classified NBI-ME findings into 6 categories according to the vascular pattern
classification [9]. Sparse and irregular patterns were considered the index of submucosal deep invasive
cancer.
Outcome measure and statistical analysis
We calculated the sensitivity, specificity, and overall accuracy of each endoscopic
modality. Furthermore, we compared the diagnostic accuracy between the EC-V and NBI-ME.
We used SPSS for Windows version 20.0 (IBM Corp., Armonk, NY, USA) for data analysis.
Pearson’s chi-square test was applied to calculate the odds ratios (ORs) of the endocytoscopic
findings, and McNemar’s test was applied to compare the diagnostic accuracy between
EC-V and NBI-ME. κ values were calculated for interobserver agreement. A κ value of
0.00 indicated poor agreement; 0.00 – 0.20, slight agreement; 0.21 – 0.40, fair agreement;
0.41 – 0.60, moderate agreement; 0.61 – 0.80, substantial agreement; and 0.80 – 1.00,
almost perfect agreement. P-values of < 0.05 were considered statistically significant.
Pathologic diagnosis
All specimens were fixed in 10 % buffered formalin solution after retrieval. They
were stained with conventional hematoxylin and eosin and diagnosed using the JSCCR
guideline for the degree of submucosal invasion; each cancer was subclassified accordingly
[13].
Results
Overall, 163 lesions were included. Of these, 31 lesions were excluded for various
reasons ([Fig. 4]). We therefore assessed 132 lesions: 81 adenomas, 18 intramucosal cancers, 4 submucosal
slightly invasive cancers, and 29 submucosal deep invasive cancers. In total, 959
pictures of these lesions were evaluated. Patient demographics and lesion characteristics
are summarized in [Table 1]. Sensitivity, specificity, and diagnostic accuracy of the caliber dilation for differentiating
adenoma/intramucosal cancer/submucosal slightly invasive cancer or submucosal deep
invasive cancer were 75.9 %, 74.8 %, and 75.0 %, respectively (P < 0.01; OR 9.31). Furthermore, sensitivity, specificity, and diagnostic accuracy
of abnormal tortuosity and branching were 17.2 %, 92.2 %, and 75.8 %, respectively
(P = 0.153; OR 2.47). Sensitivity, specificity, and diagnostic accuracy of loss of the
MN pattern were 75.9 %, 95.1 %, and 90.9 %, respectively (P < 0.01; OR 61.64). Sensitivity, specificity, and diagnostic accuracy of the caliber
change were 35.7 %, 97.1 %, and 87.1 %, respectively (P < 0.01; OR 45.78). Sensitivity, specificity, and diagnostic accuracy of the beaded sign
were 31.0 %, 99.0 %, and 84.1 %, respectively (P < 0.01; OR 45.95) ([Table 2]).
Fig. 4 Patient flowchart.
Table 1
Patient and lesion characteristics.
|
Adenoma
(n = 81)
|
Intramucosal carcinoma
(n = 18)
|
Submucosal slightly invasive carcinoma
(n = 4)
|
Invasive carcinoma
(n = 29)
|
|
Age, y
|
63.9 ± 13.1 (32 – 91)
|
66.5 ± 8.7 (50 – 81)
|
71.8 ± 3.3 (68 – 75)
|
63.2 ± 13.5 (36 – 94)
|
|
Sex
|
|
Male
|
58
|
10
|
4
|
16
|
|
Female
|
23
|
8
|
0
|
13
|
|
Location
|
|
Left side of colon
|
41
|
5
|
1
|
7
|
|
Right side of colon
|
30
|
8
|
1
|
12
|
|
Rectum
|
10
|
5
|
2
|
10
|
|
Size of lesions, mm
|
10.7 ± 6.4 (3 – 35)
|
18.7 ± 10.0 (4 – 40)
|
18.5 ± 9.9 (4 – 25)
|
20.4 ± 7.1 (8 – 35)
|
|
Protruded and flat-elevated type
|
77
|
14
|
2
|
21
|
|
Depressed type
|
4
|
4
|
2
|
8
|
Data are presented as mean ± standard deviation (range) or n.
Table 2
Diagnostic accuracy of microvascular findings.
|
Vascular findings
|
Sensitivity
|
Specificity
|
Accuracy
|
OR
(95 % CI), %
|
P value
|
|
Caliber dilation
|
75.9 %
(22/29)
|
74.8 %
(77/103)
|
75.0 %
(99/132)
|
9.31
(3.57 – 24.3)
|
< 0.01
|
|
Abnormal tortuosity and branching
|
17.2 %
(5/29)
|
92.2 %
(95/103)
|
75.8 %
(100/132)
|
2.47
(0.74 – 8.25)
|
0.153
|
|
Loss of the micro-network pattern
|
75.9 %
(22/29)
|
95.1 %
(98/103)
|
90.9 %
(120/132)
|
61.64
(17.87 – 212.29)
|
< 0.01
|
|
Caliber change
|
35.7 %
(15/29)
|
97.1 %
(100/103)
|
87.1 %
(115/132)
|
45.78
(9.16 – 139.14)
|
< 0.01
|
|
Beaded sign
|
31.0 %
(9/29)
|
99.0 %
(102/103)
|
84.1 %
(111/132)
|
45.95
(5.50 – 382.73)
|
< 0.01
|
Total number of polyps: 132.
CI, confidence interval; OR, odds ratio.
The interobserver agreement κ value between experienced endoscopists for dilation
was 0.67, that for abnormal tortuosity and branching was 0.44, that for loss of the
MN pattern was 0.55, that for caliber change was 0.52, and that for the beaded sign
was 0.55.
Sensitivity, specificity, and overall accuracy of EC-NBI were 96.3 %, 97.1 %, and
97.0 %, respectively, among the cases included in this study. On the other hand, NBI-ME
had a sensitivity of 82.1 %, specificity of 94.2 %, and overall accuracy of 91.7 %
([Table 3]).
Table 3
Comparison of diagnostic accuracy between EC-V and NBI-ME.
|
Predicting invasive cancer (T1b)
|
EC-V3
|
NBI-ME
|
P value
|
|
Sensitivity
|
96.3 %
|
82.1 %
|
0.375
|
|
Specificity
|
97.1 %
|
94.2 %
|
0.219
|
|
Accuracy
|
97.0 %
|
91.7 %
|
0.065
|
McNemar’s test was applied.
EC-V, endocytoscopic vascular pattern; NBI-ME, NBI magnification endoscopy.
Discussion
Our results suggest that caliber dilation, loss of the MN pattern, caliber change,
and the beaded sign are highly correlated with submucosal deep invasive cancers. In
addition, the endoscopists showed moderate agreement, and the results obtained were
validated.
With recent advances in endoscopic techniques such as endoscopic submucosal dissection,
even a large lesion requiring endoscopic piecemeal mucosal resection can be resected
en bloc. It has also become possible to resect non-lifting lesions other than submucosal
deep invasive cancer, because of fibrosis, after biopsy or injection of liquid, and
other circumstances. For appropriate treatment, however, it is important to differentiate
whether the lesion is submucosal slightly invasive cancer or submucosal deep invasive
cancer. Therefore, high specificity is required to prevent excessive surgical treatment.
In a comparison of diagnostic accuracy between EC-V and NBI-ME, all of these diagnostic
parameters were higher with EC-V than with NBI-ME. However, there were no significant
differences between EC-NBI and NBI-ME in terms of these parameters. However, all cases
evaluated as EC-V3 and as having an irregular/sparse pattern of NBI-ME were classified
as submucosal deep invasive cancer. This finding seems to illustrate an advantage
of EC-NBI.
EC-NBI can allow for observation of vascular irregularity in vivo and might facilitate observation of angiogenesis of colorectal lesions [11]. Konerding et al. [14] described the normal colonic mucosa as having honeycomb-like vasculature. Colorectal
tumor capillaries become thick due to angiogenesis, the caliber changes as the grade
of atypia increases, the vessels become blocked, and changes in vascular density occur
[15]. Ultra-high magnification of EC-NBI may enable the blood vessels to be observed
in a manner close to these models. EC-NBI has shown that tumor capillaries are thicker
than non-tumor capillaries and that the capillaries are significantly thick in cases
of submucosal deep invasive cancer. Loss of the MN pattern is considered to reflect
a phenomenon of decreasing vascular density, although severely atypical blood vessels
remain as the lesion develops into submucosal deeply invasive cancer. Although significant
differences were observed in vascular dilation, the diagnostic accuracy was 75.6 %.
This result was lower than that in other cases. A previous study that utilized microvascular
corrosion casting revealed that invasive cancer has significantly larger microvessel
diameters [14]. The vascular dilation could therefore be the most important finding in submucosal
deep invasive cancer. Conversely, there were no significant differences in tortuosity.
In NBI-ME, tortuosity is an important abnormal vascular finding, and its significance
has been confirmed. However, at 380 × magnification with EC, even usual adenomas show
mild tortuosity. It becomes difficult to distinguish normal from abnormal findings
according to tortuosity.
When we evaluated the detailed vascular findings in this study, high-quality images
were needed. We used the EVIS LUCERA SPECTRUM SYSTEM until 2012. With this system,
the quality of images was usually poor because sufficient brightness was not obtained
with EC-NBI. With the advent of the EVIS LUCERA ELITE SYSTEM in October 2012, the
images obtained with EC-NBI were brighter and clearer ([Fig. 5]). This may have been attributable to increased light intensity, increased exposure
time of NBI, and improved noise reduction, thereby allowing clearer observation of
the vessels and detailed examination of EC-V. With regard to advantages of EC-NBI
over conventional EC, conventional EC always requires staining with crystal violet
and methylene blue [16]
[17]. However, EC-NBI does not require staining, and testing with EC-NBI therefore can
be performed safely and quickly. We recently reported on use of a computer-aided diagnosis
(CAD) system for EC-V [18]. The CAD system provided sufficient diagnostic ability for lesion characterization,
but it could not predict the invasion depth of colorectal cancer. Therefore, EC-V
still has an advantage over the CAD system.
Fig. 5 Comparison of endocytoscopy with narrow-band imaging using EVIS LUCERA ELITE SYSTEM
and EVIS LUCERA SPECTRUM SYSTEM. The (a) EVIS LUCERA ELITE SYSTEM images became brighter and clearer than the (b) EVIS LUCERA SPECTRUM SYSTEM images.
Future development of quantitative and qualitative diagnostic techniques will use
elements different from those of EC classification because EC-NBI does not focus on
the nucleus and gland formation of surface colorectal lesions but on the vascular
caliber and its irregularity. For example, when cancer is deep-seated and not exposed
to the surface, nuclear or structural atypia may not be seen by EC. In these cases,
vascular findings might help to diagnose the submucosal deeply invasive area. Furthermore,
it might be possible to predict potential metastasis, such as venous invasion, by
comparing vascular irregularities and thus select candidates for vascular endothelial
growth factor inhibitors (e. g., bevacizumab) or predict treatment efficacy [19]
[20].
This study had some limitations. First, it was retrospective study and conducted in
a single institution. We are planning to increase the number of patients for a further
prospective study. Second, although EC enables the living body to be observed, it
is difficult to compare the results with pathologic results. We considered that because
the focal depth of the endocytoscope is about 50 µm, comparison with dilated vessels
on the surface would be possible. However, it was actually difficult to evaluate deformed
blood vessels, including changes in vascular caliber, because of dehydration and staining
during specimen preparation. Therefore, it may be necessary to consider that pathologic
findings are new diagnostic findings in the living body.
Conclusion
In conclusion, dilation, loss of the MN pattern, caliber change, and a beaded sign
are important vascular findings observed by EC in patients with colorectal submucosal
deep invasion. These results appear to contribute to the definition of irregular vascular
findings of EC-V3.
