Introduction
Sessile serrated lesions (SSLs) are the major precursor lesions of serrated pathway
cancers and account for 20 % to 30 % of all sporadic cases of colorectal cancer [1]
[2]. Despite their importance, they are easily overlooked because they are difficult
to detect [3]
[4]. Furthermore, there is a need for not only the diagnosis but an appropriate endoscopic
treatment method for SSL.
Previous studies on snare endoscopic mucosal resection (EMR) for colonic polyps ≥ 20 mm
showed that there were significantly more local recurrences in the piecemeal resection
(11 % to 18 %) than in the en bloc resection (2 % to 3 %) [5]
[6]
[7]. In addition, in a previous large study on 5- to 20-mm colonic EMR, incomplete resection
was more probable in piecemeal resection compared to en bloc resection (20.4 % vs.
8.4 %; relative risk 1.41), and was more likely in lesions with 15- to 20-mm size
compared to 5- to 7-mm size (23.3 % vs. 5.3 %; relative risk 3.21) [8].
As Agarwal et al. reported, in an en bloc EMR, the residual lesions were significantly
higher at the resection margins in the SSL than in the adenoma group [9]. However, Klein et al. [10] reported the efficacy of thermal ablation of the post-EMR mucosal defect margin
in reducing polyp recurrence. In another report, cold piecemeal EMR was shown to treat
large SSLs adequately, safely, and effectively [11]. Furthermore, endoscopic resection using a snare is also gaining progress in Western
countries.
On the other hand, endoscopic submucosal dissection (ESD) for superficial colorectal
tumors is technically more difficult and results in a higher risk of adverse events
compared to EMR. However, colorectal ESD is currently recognized as a standard technique
for treating colorectal tumors because of the development of improved modalities and
specialized devices [12]. A number of studies have reported the clinical outcomes of ESD [7], its efficacy for en bloc resection, and the long-term low recurrence rates for
large superficial colorectal tumors [13]
[14].
There have been several studies on the choice of EMR for treating SSL; however, few
studies have reported successful use of ESD for SSL. Therefore, its efficacy in ESD
for SSL is still unknown. We hypothesized that ESD for SSL may be acceptable if it
were as safe and effective as ESD for conventional adenoma. In this study, we examined
the characteristics and outcomes of SSL and compared them with those of non-SSL in
ESD.
Patients and methods
Patients
We retrospectively reviewed patients who were treated between January 2016 and March
2020 at our hospital. Indications for treatment were confirmed using colonoscopy with
magnifying function (CF-HQ290ZI or CF-H260AZI; Olympus, Tokyo, Japan). This was determined
endoscopically in combination with pit pattern diagnosis using magnifying chromoendoscopy
or the Japan NBI Expert Team (JNET) classification by narrow band imaging (NBI) [15]
[16]
[17]. Endoscopic resection was planned for lesions diagnosed as serrated lesion or adenoma,
intramucosal carcinoma, submucosal slightly invasive carcinoma (< 1000 μm). The indications
for ESD were as follows: 1) lesions that were difficult to resect with a snare, and
that required endoscopic en bloc resection for accurate histological diagnosis, regardless
of size (e. g., large depressed-type tumors) [18]; and 2) lesions for which complete resection using snare EMR was difficult. Although
SSL is considered to have a lower malignant potential than that of adenomatous lesions,
the indication for ESD was decided as above.
We performed colorectal ESD on 370 consecutive cases in 322 patients during the study
period. We excluded 39 0–Is lesions classified according to the Paris classification
[19] from the analysis. The 0-I lesions of SSL are usually treated with EMR at our center,
and all 0-I lesions that underwent ESD were adenoma/intramucosal carcinoma cases.
The 0-I lesions of adenoma/intramucosal carcinoma often had severe fibrosis, and we
considered that the dissection was difficult. Therefore, we excluded 0-I lesions from
this study. In addition, we excluded four traditional serrated adenomas, 33 submucosal
invasive carcinomas (22 submucosal slight invasions and 11 submucosal deeper invasions),
20 submucosal tumors, four local recurrence lesions, one lesion < 10 mm (lesions < 10 mm
are usually treated using cold snare polypectomy in our center; this lesion was simultaneously
resected with multiple others), and two interrupted lesions (one severe fibrosis and
one muscular propria invasion) from the analysis. Finally, 267 0–IIa lesions were
included. Among these, 41 SSL cases (SSL group) and 226 adenoma/intramucosal carcinoma
cases (non-SSL group) ([Fig. 1]) were compared. Almost all intramucosal carcinomas detected arose from adenomas;
however, one of the cases was determined to arise from an SSL and was included in
the SSL group. Further, 16 SSLs, 19 adenomas, and 12 intramucosal carcinomas were
simultaneously resected with multiple lesions. This study was approved by our institutional
review board (F2020C34; 8 June 2020). The need for informed consent was waived via
the opt-out method on the hospital’s website.
Fig. 1 Study flow diagram. ESD, endoscopic submucosal dissection.
Endoscopic system and ESD procedure
We used an endoscope with a water-jet system (PCF-Q260J; Olympus). When scope operability
was poor because of the paradoxical movement or adhesion, a balloon-assisted endoscope
with a hydrophilic-coated silicone splinting tube (ST-CB 1; Olympus) was used [20]. A short-type small-caliber-tip transparent hood (Fujifilm; Tokyo, Japan) was placed
on the endoscopic tip. For all the cases, we used carbon dioxide for insufflation,
and the electrosurgical unit was ESG-100 (Olympus). We used a dual knife (KD650Q;
Olympus), and ESD was performed under conscious sedation, using intravenous midazolam
(1–3 mg), pethidine hydrochloride (35 mg), or both. Submucosal injection of hyaluronic
acid solution mixed with a small amount of indigo carmine and 0.1 % epinephrine (1 mL
of 0.1 % epinephrine in 9 mL of indigo carmine solution = 1:100,000 injectate) was
applied. After injection into the submucosal layer, we alternately performed the partial
circumferential incision (“Pulse cut slow" mode [30 W]) and the subsequent submucosal
dissection (“Forced coagulation” mode [30 W]). We performed all the ESD procedures
as described in the previous study [20].
Histological assessment
Each resected specimen was immersed in 10 % formalin, and a pathologist performed
the histological assessment. The histological diagnosis of the lesion and the involvement
of the resection margin were evaluated according to the Japanese Classification of
Colorectal Carcinoma [21]. The diagnostic category of “intramucosal carcinoma” in Japan is equivalent to “high-grade
dysplasia” in Western countries. Sessile serrated lesions were classified as SSL without
cytological dysplasia, SSL with cytological dysplasia, traditional serrated adenomas,
and hyperplastic polyps, in accordance with the classification provided by the World
Health Organization [22]. Histological diagnosis of the nature of intramucosal carcinoma depended on whether
there was concomitant detection of tumor belonging to one of the two categories in
the adjacent tissue.
Measured characteristics and outcomes
We drew comparisons between the groups using the following variables: sex and age,
tumor size and location, macroscopic type, operator, histology, procedure duration,
dissection speed, degree of fibrosis, en bloc resection rate, R0 resection rate, and
associated complications. The proximal colon was defined as the cecum, ascending colon,
and transverse colon, and the distal colon as the splenic flexure, descending colon,
sigmoid colon, and rectum. The 0-IIa lesion was categorized into laterally spreading
tumor, granular type (LST-G), and non-granular type (LST-NG) [16]. Considering the operator, Expert (two endoscopists) was defined as having experienced
with more than 100 cases of colorectal ESD, and Non-expert (six endoscopists) had
experience with fewer than 50 cases of colorectal ESD. There were no endoscopists
who had performed between 50 and 100 colorectal ESDs. We further analyzed the fibrosis
grade (F0–1: absence, F2–3: presence) [23]. R0 resection was defined as an en bloc resection with free vertical and horizontal
margins. R1 resection (positive resection margin) and RX resection (unclear/indeterminate
resection margin) were included in the category of non-R0 resection. Perforation was
considered as being intraoperative, delayed, or both. Postoperative bleeding was referred
to as overt bleeding within 14 days after ER, resulting in the need for endoscopic
hemostasis.
Statistical analysis
Group differences were evaluated using the Mann-Whitney U test for the continuous
variables and the χ2 test was used for the dichotomous variables. P < 0.05 was considered statistically significant. Four factors were considered potential
clinical variables for outcomes in ESD: tumor size, location (proximal/ distal), macroscopic
type (LST-G/ LST-NG), and operator (Expert/ Non-expert). The propensity score was
estimated using a logistic regression model for cases in which we performed ESD for
SSL. We performed a one-to-one matching analysis between the SSL and non-SSL groups
using the nearest neighbor method, with a caliper width equal to 0.2 of the standard
deviation of the logit of the propensity score. The C-statistic (0.876) for evaluating
the goodness of fit was calculated. All statistical analyses were performed using
EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), which is a
graphical user interface for R (The R Foundation for Statistical Computing, Vienna,
Austria). This is a modified version of the R-commander, which is designed to add
statistical functions frequently used in biostatistics [24].
Results
Baseline characteristics and outcomes
Baseline characteristics of the ESD patients are presented in [Table 1]. Patients in the SSL group were younger than those in the non-SSL group (median
age, 58 vs. 71 years; P < 0.001). The percentages of females and proximal colon involvement in the SSL group
were significantly higher than those in the non-SSL group (63.4 vs. 35.4 %; P < 0.001 for females and 87.8 vs. 57.1 %; P < 0.001 for proximal colon location). There were four and six cases with lesions
involving the appendiceal orifice in the SSL and the non-SSL groups, respectively.
There were no significant differences in the median tumor size (25 vs. 25 mm; P = 0.31) and percent expert operators (44 vs. 47 %; P = 0.73) between the groups. In terms of macroscopic type, the rate of LST-NG was
significantly higher in the SSL than that in the non-SSL group (97.6 vs. 40.3 %; P < 0.001).
Table 1
Clinical characteristics of ESD.
|
Baseline cases
|
Propensity-matched cases
|
|
SSL
|
non-SSL
|
P
|
SSL
|
non-SSL
|
P
|
|
N = 41
|
N = 226
|
|
N = 38
|
N = 38
|
|
|
Sex (female)
|
26 (63.4)
|
80 (35.4)
|
< 0.001[1]
|
23 (60.5)
|
11 (28.9)
|
0.011[1]
|
|
Age (years); median (range)
|
58 (38–78)
|
71 (31–88)
|
< 0.001[2]
|
58 (38–76)
|
69.5 (31–88)
|
< 0.001[2]
|
|
Location
|
|
|
< 0.001[1]
|
|
|
> 0.99[1]
|
|
Proximal colon (%)
|
36 (87.8)
|
129 (57.1)
|
|
33 (86.8)
|
33 (86.8)
|
|
|
Distal colon (%)
|
5 (12.2)
|
97 (42.9)
|
|
5 (13.2)
|
5 (13.2)
|
|
|
Tumor size, median mm (range)
|
25 (10–50)
|
25 (10–105)
|
0.31[1]
|
25 (10–40)
|
23.5 (12–40)
|
0.61†
|
|
Macroscopic type(LST-G/LST-NG)
|
1/40
|
135/91
|
< 0.001[1]
|
1/37
|
1/37
|
> 0.99[1]
|
|
Operator (Expert/Non-expert)
|
18/23
|
106/120
|
0.73[1]
|
15/23
|
17/21
|
0.82[1]
|
|
Histology
|
|
SSL without cytological dysplasia
|
39 (95.2)
|
|
|
37 (97.4)
|
|
|
|
SSL with cytological dysplasia
|
1 (2.4)
|
|
|
1 (2.6)
|
|
|
|
SSL with carcinoma
|
1 (2.4)
|
|
|
|
|
|
|
Adenoma
|
|
113 (50)
|
|
|
20 (52.6)
|
|
|
Intramucosal cancer
|
|
113 (50)
|
|
|
18 (47.4)
|
|
|
Fibrosis (presence) (%)
|
1 (2.4)
|
15 (6.6)
|
0.48[1]
|
1 (2.6)
|
3 (7.9)
|
0.61[1]
|
ESD, endoscopic submucosal dissection; SSL, Sessile serrated lesion; non-SSL, adenoma,
or intramucosal cancer; Proximal colon (i. e., cecum, ascending colon, and transverse
colon); Distal colon (i. e., splenic flexure, descending colon, sigmoid colon, rectum);
LST-G, laterally spreading tumor (granular type); LST-NG, laterally spreading tumor
(non-granular type); Expert, ≥ 100 cases of experience in colorectal ESD, Non-expert,
< 50 cases of experience in colorectal ESD; Fibrosis, F0–1: absence/F2–3: presence
1 Chi-square test
2 Mann-Whitney U test
Histological results in the SSL group included 39 SSL without cytological dysplasia
adenomas, one SSL with cytological dysplasia, and one SSL with carcinoma. The results
of the non-SSL group included 113 adenomas and 113 intramucosal carcinomas. Baseline
outcomes in the ESD patients are presented in [Table 2]. The median duration of the ESD procedure tended to be slightly shorter in the SSL
than in the non-SSL group (52 vs. 55 minutes; P = 0.076). There were no significant differences between the groups in terms of median
dissection speed (12 vs. 11.2 cm2/h; P = 0.44), presence of fibrosis (2.4 vs. 6.6 %; P = 0.48), en-bloc resection rate (97.6 vs. 99.6 %; P = 0.74), R0 resection rate (92.7 vs. 93.4 %; P = 0.74), rate of perforation (0 vs. 0.9 %; P > 0.99), and rate of postoperative bleeding (2.4 vs. 1.8 %; P = 0.56). In all cases of non-R0 resection, we could not clearly determine the horizontal
margin. There was no perforation in the SSL group. The two cases of perforation in
the non-SSL group were conservatively treated after performing endoscopic closure
using endoclips. Postoperative bleeding in both groups could be managed using endoclips.
We successfully managed all adverse events endoscopically.
Table 2
Clinical outcomes of ESD.
|
Baseline cases
|
Propensity-matched cases
|
|
SSL
|
non-SSL
|
P
|
SSL
|
non-SSL
|
P
|
|
N = 41
|
N = 226
|
|
N = 38
|
N = 38
|
|
|
En bloc resection rate (%)
|
40 (97.6)
|
225 (99.6)
|
0.28[1]
|
37 (97.4)
|
37 (97.4)
|
> 0.99[1]
|
|
R0 resection rate (%)
|
38 (92.7)
|
211 (93.4)
|
0.74[1]
|
36 (94.7)
|
35 (92.1)
|
> 0.99[1]
|
|
Duration of ESD procedure (median; min) (range)
|
52 (5–160)
|
55 (6–300)
|
0.076[2]
|
52.5 (5–160)
|
55 (10–210)
|
0.17[2]
|
|
Dissection speed (median; cm2/h) (range)
|
12 (2.8–48)
|
11.2 (1.1–43)
|
0.44[2]
|
12 (2.9–48)
|
7.7 (2.7–25)
|
0.0095[2]
|
|
Perforation rate (%)
|
0 (0)
|
2 (0.9)
|
> 0.99[1]
|
0 (0)
|
1 (2.6)
|
> 0.99[1]
|
|
Postoperative bleeding rate (%)
|
1 (2.4)
|
4 (1.8)
|
0.56[1]
|
0 (0)
|
1 (2.6)
|
> 0.99[1]
|
ESD, endoscopic submucosal dissection; SSL, sessile serrated lesion; non-SSL, adenoma,
or intramucosal cancer; R0 resection, en bloc resection with free vertical and horizontal
margins
1 Chi-square test
2 Mann-Whitney U test
Of the 41 cases in the SSL group, one lesion was not resected en bloc because of involvement
of the appendiceal orifice with severe fibrosis and was divided into two parts with
a snare. Similarly, of the 226 cases in the non-SSL group, one lesion could not be
resected en bloc because of the involvement of the ileocecal valve and was divided
into two parts with a snare. Histological evaluation of the non-en bloc resection
cases revealed a positive horizontal margin and a negative vertical margin.
Characteristics and outcomes after propensity score matching
The matched four factors and the characteristics between the SSL and non-SSL groups
after the propensity score matching are shown in [Table 1]. Finally, 38 pairs were matched in the study, and the differences in outcomes of
the ESD treatment were compared between the two groups ([Table 2]). Similar to the baseline results, there were statistically significant differences
in age (P < 0.001) and sex (P = 0.011) between the groups. The rate of en bloc resection (97.4 %) was similar in
both groups. There were no significant differences in the existence of fibrosis (2.6
vs. 7.9 %; P = 0.61), the rate of R0 resection (94.7 vs. 92.1 %; P > 0.99), and adverse events between the groups. The median ESD procedure duration
was not significantly different between the groups (52.5 vs. 55 minutes; P = 0.17); however, the median dissection speed (12 vs. 7.7 cm2/h; P = 0.0095) was significantly faster in the SSL than in the non-SSL group.
Discussion
Several studies described the efficacy of ESD for superficial colorectal neoplasia
for large lesions, and rates of en bloc and complete resections were higher in ESD
than in EMR [7]
[12]. Rates of en bloc resection (99 %) and R0 resection (93 %) reflected favorable outcomes
in baseline cases in this study.
As Agarwal et al. reported, in en bloc resection EMR, incidence of residual lesions
at the resection margin was significantly higher in the SSL compared to the adenoma
group [9]. This suggests that it may have been more difficult to recognize the border in the
SSL than in the adenoma group. In the present study, in the baseline cases, there
was no difference in the rates of R0 resection between both groups (SSL 92.7 % vs.
non-SSL 93.4 %; P = 0.74), which was a favorable result. In large lesions with excessive folds, it
may be more difficult to identify the border. Some SSLs were covered with a mucus
cap. A mucus cap is effective in detecting SSL and it assists in the identification
of the borderline. However, a mucus cap may shift from the lesion due to contact during
endoscopic resection which may affect the identification of SSLs. In such cases, to
avoid misidentifying the borderline, mucosal incision under direct vision in ESD was
beneficial. Marking of the incision is not usually necessary in the non-SSL group
because the border between the tumor and the normal tissues is quite clear after spraying
with indigo carmine. However, SSL borders are usually inconspicuous and when they
remain unclear on chromoendoscopy, marking of the borders prior to incision can be
useful ([Fig. 2]). After propensity score matching, there were no differences in rates of en bloc
and R0 resections between the SSL and the non-SSL groups.
Fig. 2 Sessile serrated lesion at the ascending colon. a Conventional image by white light. b Indigo carmine dye (0.4 %) was sprayed over the lesion, and the borderline was partially
unclear. c A marking under NBI mode was useful for incision. d Submucosal dissection using a dual knife. More fatty tissue in the submucosa was observed
in some cases at deep parts of the colon. e Ulcer bed after en bloc resection. f Resected specimen. The tumor measured 40 × 25 mm, and histological evaluation revealed
a negative-margin sessile serrated lesion without cytological dysplasia.
The median dissection speed (12 vs. 7.7 cm2/h; P = 0.0095) was significantly faster in the SSL than in the non-SSL group. Only one
SSL with submucosal fibrosis was observed in this study. We believe that the absence
of fibrosis in SSL was one of the reasons for faster dissection speed. Similarly,
Pellise et al. described that SSL was lifted easily after injection and was not associated
with submucosal fibrosis [6].
Although dissection speed was significantly higher in the SSL group, the tumor size
and procedure time were not different between the two groups. We believe this is because
the SSL group had larger horizontal margins than did the non-SSL group, possibly because
we marked the borders further outside the lesions in the SSL group. The dissection
speed for the non-SSL group after the propensity score matching (7.7 cm2/h) was lower than that before matching (11.2 cm2/h). This is probably because the number of LST-Gs decreased after matching. In this
study, SSL was significantly more common in younger patients, females, and proximal
colon location, which was consistent with the findings of previous reports [6]
[25]. The areas of the proximal colon, especially the ascending colon, have rich fat
in the colonic submucosa, and endoscopists must be careful to identify the dissection
line in ESD. After a clear submucosal layer under the fat tissue is directly observed,
submucosal dissection is relatively easy.
In the SSL group, the only non-en bloc resection was a lesion covering the appendiceal
orifice. Jacob et al. reported that the area near the appendiceal orifice was often
fibrotic, which increased the difficulty of ESD [26]; the case in the present study had fibrosis of similar severity. SSLs can commonly
occur in the appendix [27]; therefore, care must be taken. The traction system may be effective in the ESD
lesion involving the appendiceal orifice [28].
Moss et al. reported that in large lesions ( ≥ 20 mm), if the initial EMR was deemed
successful, the recurrent lesion was usually unifocal and diminutive, and could be
managed. Overall, local recurrence was managed endoscopically in 93 % of the cases
[5]. Large residual tumors and local recurrences caused by the incomplete resections
are very difficult to resect with additional endoscopic treatment; therefore, should
be avoided [29]. Moreover, patients with piecemeal resection must undergo intensive endoscopic follow-up,
highlighting the importance of en bloc resection. However, in endoscopic treatment
for large SSLs, Klein et al. [10] reported a 4-fold reduction in adenoma recurrence after systematic ablation of the
post-EMR mucosal defect margins. Tutticci et al. reported that the cold piecemeal
EMR adequately treats large SSLs with low local recurrence and low postoperative bleeding
[11]. Furthermore, endoscopic resection using a snare is widely acceptable in Western
countries. However, there is a scope for consideration of ESD in the future. Both
ESD and EMR have their advantages and disadvantages, and it is difficult to decide
which is useful at this stage. In addition, it is necessary to consider their financial
burden. Some studies have been conducted on EMR; nonetheless, no study has demonstrated
the indication of ESD in SSL. It is difficult to determine the indication of ESD for
SSL depending on only the lesion size. Although ESD has higher rates of en bloc and
complete resection than does EMR, it requires experience and long procedure times.
ESD is not required in all cases. Alternatively, a large EMR study that categorized
lesions ( ≥ 20 mm) by size demonstrated a statistically significant eight-fold increase
in the risk of recurrence for adenomas compared with the SSLs in 20–25 mm lesions.
This persisted after adjusting for the potential confounding variables (adjusted hazard
ratio 7.6; P = 0.047); nevertheless, this was not maintained in the larger lesions [6]. These findings suggest that SSL ≥ 25 mm may be good indications for ESD. Further
appropriate indications of ESD for large SSLs should be identified and validated using
randomized trials and multicenter studies, or with a direct comparison of EMR and
ESD.
It is difficult to achieve complete resection using the snare technique in large lesions
that exceed the folds and those that involve the appendiceal orifice or the ileocecal
valve as with adenomas. Such SSLs may be good indications for ESD because it is difficult
to detect the borderline characteristics of SSL.
Considering lesions (e. g., lesions that require accurate histological diagnosis such
as an SSL with carcinoma) requiring endoscopic en bloc resection for which the use
of the snare technique is difficult, ESD should be listed as an option for endoscopic
treatment. The skill and knowledge of an accurate preoperative endoscopic diagnosis
for dysplasia or cancer within SSL is indispensable [30]
[31]
[32].
As reported by Byeon et al. [33], the rate of en bloc resection was higher for ESD than for hybrid ESD (snaring resection
technique during the ESD procedure with submucosal dissection) (87 vs. 64 %, P < .01). The rate of the histologically complete resection was also higher for ESD.
However, the rate of resection of lesions less than 20 mm was not different between
the ESD and the hybrid ESD groups; further, the objects of this study were colorectal
adenomatous lesions. According to another report, hybrid ESD for SSL was easy [34]; therefore, in the case of SSL of 20 mm and under, hybrid ESD may be considered.
Complete removal of SSL may ultimately prevent interval colorectal cancer. However,
Yang et al. reported that there was no decrease in the frequency of right-sided colorectal
cancer as might be expected with improved recognition and management of predominantly
right-sided SSLs [35]. They have discussed the possibility that the diagnostic ability of SSLs or interval
in the SSLs-inducing colorectal cancer occurrence was not appropriate in this study.
Further validation will require future large-scale, long-term studies.
There are several limitations to our study. First, it was a retrospective cohort study
performed at a single institution, and the number of SSLs was small. Second, this
study did not consider local recurrence during follow-up colonoscopy. This study was
based on previous studies that demonstrated that en bloc resection in ESD was associated
with less local recurrence. Differentiation of hyperplastic polyps from SSLs was not
possible because of the limitations of the JNET classification. Use of NBI classifications
that differentiate between serrated lesions and hyperplastic polyps was not routine
in our center during the study period. The macroscopic type (e. g., Paris classification,
LST sub-type) was a subjective judgment.
Conclusion
In conclusion, ESD for SSL was safely performed and the tumors were more easily removed
than with use of ESD for non-SSL. ESD might be an acceptable endoscopic treatment
option for SSL that may be difficult to completely resect even with piecemeal resection
by EMR using a snare, or for SSL judged, during preoperative assessment, to require
en bloc resection.
