Introduction
Endoscopic resection of adenomatous polyps reduces the incidence of colorectal cancer
(CRC) and prevents CRC-related mortality [1 ]
[2 ]
[3 ]. Multiple interventions have been proposed to improve complete resection and mitigate
the rate of adverse events (AEs) associated with endoscopic resection of colorectal
polyps.
Cold snare polypectomy (CSP) has been used as the preferred method for the removal
of
nonpedunculated colorectal polyps <10 mm because of its safety, effectiveness, and
lower
costs ([Fig. 1 ]
a ) [4 ]
[5 ]; however, complete histologic resection rates with CSP vary widely (44%–96%) [6 ]. For the removal of sessile polyps of 10–19 mm in size, current guidelines recommend
hot snare polypectomy (HSP), with or without submucosal injection [7 ]
[8 ]. For pedunculated polyps, HSP is also the recommended first-line modality [7 ]
[8 ].
Fig. 1 Endoscopic images of: a cold snare polypectomy; b cold endoscopic mucosal resection with submucosal injection of a saline solution
with epinephrine and methylene blue.
Endoscopic mucosal resection (EMR) includes a submucosal injection of a given solution
to lift the lesion away from the muscular layer, decreasing the risk of thermal transmural
injury [9 ]. Some recent studies have reported that the addition of a submucosal injection to
CSP, referred to as cold endoscopic mucosal resection (C-EMR), may be useful and represent
a valuable adaptation of the standard cold snare technique ([Fig. 1 ]
b ) [10 ]
[11 ]
[12 ]
[13 ]
[14 ]
[15 ]
[16 ]. Data comparing CSP with C-EMR are however still scarce [17 ].
Two recent systematic reviews, both based mainly on observational studies, have evaluated
C-EMR [18 ]
[19 ]. Liang et al. [18 ] demonstrated that, although C-EMR is efficient for polyps measuring<20 mm, it did
not improve the efficiency and safety of CSP and hot snare EMR as expected. Abdallah
et al. [19 ] showed that C-EMR has a high technical success rate and an excellent safety profile,
with variable recurrence rates based on polyp size and histology.
The aim of this systematic review and meta-analysis was to compare the rates of complete
resection, en bloc resection, and AEs, as well as the procedure time between CSP and
C-EMR for the resection of colorectal lesions, analyzing only randomized controlled
trials (RCTs). Prespecified subgroup analyses were performed based on the size of
the lesions and their morphology.
Methods
This study was registered in the International Prospective Register of Systematic
Reviews (PROSPERO) under protocol CRD42023439605.
Study outcomes and subgroup analysis
The primary outcome was complete lesion resection. Secondary outcomes included procedure
time, en bloc resection, and AEs. Prespecified subgroup analyses were performed based
on the size of the lesions (<10 mm and ≥10 mm), and the morphology (pedunculated and
nonpedunculated).
Eligibility criteria
We selected studies based on the following inclusion criteria: (i) RCTs; (ii) comparing
C-EMR with CSP; (iii) for resection of colorectal lesions in adult patients undergoing
colonoscopy; and (iv) reporting at least one of the outcomes of interest. Only studies
published as original full-text, peer-reviewed articles were included. Editorials,
letters, reviews, systematic reviews, and meta-analyses were excluded.
Search strategy
We searched PubMed, Embase, and the Cochrane Library to identify studies meeting the
inclusion criteria, published until 1 June 2023. The search strategy is detailed in
Appendix 1s , see online-only Supplementary material. Two reviewers (P.A.E.S. and S.B.) conducted
the search, imported results into Zotero 6.0, and triaged the studies. After the exclusion
of duplicates and titles/abstracts clearly not related to the clinical question, the
eligibility of each remaining study was assessed based on review of the full-text
articles. We also searched for any additional studies from the references of the included
RCTs, as well as from previous systematic reviews and meta-analyses. Disagreements
were solved by a third author (G.C.M.).
Data extraction
Two authors (P.A.E.S. and S.B.) extracted the data into a standardized format, including
study characteristics (first author, year of publication); study population (number
of patients, age, sex); polyp characteristics (number, size, location, and morphology);
complete resection rate; en bloc resection rate; AE rate; and procedure time.
Definitions
Complete resection was defined as either no residual lesion in biopsy samples performed
after polypectomy or when muscularis mucosae was present under the tumor along more
than 80% of the tumor’s horizontal dimension (Table 1s ). En bloc resection was defined as visual total polyp eradication judged by the endoscopist's
experience.
Immediate bleeding was defined by all studies as clinically significant bleeding requiring
endoscopic intervention during the procedure. Wei et al. [20 ] also included bleeding within 24 hours of the procedure as immediate bleeding.
Delayed bleeding was defined as clinically significant bleeding requiring hospitalization,
blood transfusion, or endoscopic or surgical intervention within 14 days of polypectomy
in three studies [6 ]
[21 ]
[22 ], whereas two others defined this outcome as bleeding within 30 days after polypectomy
[5 ]
[20 ]. Other studies did not specify their definition of delayed bleeding [4 ]
[23 ].
Perforation was defined as visualization of an extraintestinal structure during the
endoscopic examination or patients presenting with symptoms and signs of peritonitis
within several hours after completion of the colonoscopy.
Assessment of the risk of bias
Quality assessment of RCTs was performed using the Cochrane tool for assessing the
risk of bias in randomized trials (RoB 2) [24 ]. Two authors (P.A.E.S. and E.C.B.) independently evaluated each included study for
its risk of bias in five domains, and assigned a rating of high, low, or some concerns
for each. Publication bias in the primary outcome was investigated by funnel plot
analysis.
Data analysis and reporting
The systematic review and meta-analysis were designed and reported according to the
Cochrane Collaboration Handbook for Systematic Review of Interventions and the Preferred
Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement guidelines
[25 ]
[26 ]. We calculated the risk ratio (RR) and 95%CI for dichotomous outcomes, and mean
difference and 95%CI for continuous outcomes. The pooled estimates of RR or mean difference
were calculated using a random-effects model, in which both within-study and between-study
variations were considered. Not all studies provided the mean and SD of the sample.
In this case, the data was calculated on the basis of the sample’s reported median
and range, according to the methods devised by Luo et al. and Wan et al. [27 ]
[28 ].
Between-study heterogeneity was assessed using the Cochran’s Q test and I ² statistics; P
values <0.10 and I ²>25% were considered to be significant
for heterogeneity. For each end point, a DerSimonian and Laird random-effects model
was
used. We also performed sensitivity analyses removing each individual study from the
outcome
assessment. Statistical analysis was performed using Review Manager 5.4 (Nordic Cochrane
Centre, The Cochrane Collaboration, Copenhagen, Denmark) and OpenMeta (Analyst)
Software.
Results
Characteristics of included studies
The literature search yielded 2552 results. As shown in [Fig. 2 ], after study triage, we included seven studies, with 1556 patients and 2287 polyps
[4 ]
[5 ]
[6 ]
[20 ]
[21 ]
[22 ]
[23 ]. A total of 1176 (51.4%) lesions were removed using C-EMR and 1111 (48.5%) with
CSP.
Fig. 2 PRISMA flow diagram of study screening and selection.
Most studies used saline solution with epinephrine, along with methylene blue or indigo
carmine for submucosal injection in the C-EMR procedure. Wei et al.
[20 ] used EverLift (a solution composed of water, glycerin, hydroxyethyl cellulose, benzyl
alcohol, sodium and potassium phosphate, and methylene blue) and Rex et al. [23 ] reported the use of saline solution, hydroxyethyl starch, or any commercially available
injection fluid.
Four studies included only nonpedunculated polyps [5 ]
[20 ]
[22 ]
[23 ]. Li et al. [21 ] included pedunculated polyps in their analysis, constituting 16% of the evaluated
colorectal lesions in the study. Katagiri et al. [4 ] and Kim et al. [6 ] did not provide specific information on the morphology of the polyps included.
[Table 1 ] summarizes the main characteristics of the included studies.
Table 1 Characteristics of the included randomized studies comparing cold endoscopic mucosal
resection (C-EMR) and cold snare polypectomy (CSP).
Study
Katagiri, 4 ]
Kim, 6 ]
Li, 21 ]
Mou, 5 ]
Rex, 23 ]
Shimodate, 22 ]
Patient characteristics
118
327
261
300
122
214
65.8
62.7
51.7
53.5
65.6
66.5
69 (58.5)
213 (65.1)
150 (57.4)
182 (60.6)
69 (56.5)
121 (56.5)
Polyp characteristics
Number of polyps, n
149
210
252
234
82
107
112
215
244
216
68
107
Polyp size, mean (SD) or median, mm
4.4 (1.8)
7.2 (1.4)
12.0 (3.4)
5
9.5 (2.8)
5
4.2 (1.8)
7.0 (1.2)
11.9 (3.3)
5
9.4 (3.1)
5
Proximal location, n (%)
108 (72)
117 (56)
133 (53)
111 (47)
62 (76)
50 (51)
73 (65)
136 (63)
149 (61)
89 (41)
55 (81)
44 (44)
Paris IIa/IIb, n (%)
98 (66)
N/A
33 (13)
127 (54)
47 (57)
30 (31)
82 (73)
N/A
27 (11)
110 (51)
32 (47)
41 (41)
Paris Is/Isp, n (%)
51 (34)
N/A
171 (68)
107 (46)
35 (43)
67 (69)
30 (27)
N/A
186 (76)
106 (49)
36 (53)
59 (59)
Pathologic diagnosis,
Adenoma,
Adenoma,
Adenoma,
Adenoma,
Adenoma,
Adenoma,
Adenoma,
Adenoma,
Adenoma, 223 (91.4)
Adenoma,
Adenoma,
Adenoma,
Quality assessment of the included studies
We conducted the quality assessment analysis using the RoB 2 tool [24 ]. As shown in Fig. 1s , because the endoscopists were aware of each patient's allocation group, there was
a potential source of bias in each study. Publication bias was assessed by analyzing
the funnel plot of complete resection. Owing to some asymmetry in the distribution,
it is possible that small trials with negative results have not been published (Fig. 2s ).
Primary outcome
The complete resection rate was reported in six studies, including 2026 polyps. The
pooled estimated rate in the C-EMR group was 89.4% (95%CI 83.5%–95.4%), and in CSP
group was 88.4% (95%CI 82.1%–94.6%). The risk of complete resection was similar for
the C-EMR and CSP techniques (RR 1.02, 95%CI 0.98–1.05; I ² = 72%) ([Fig. 3 ]
a ). Leave-one-out sensitivity analyses confirmed no significant difference between
C-EMR and CSP for the outcome of complete resection rate after the exclusion of each
individual study and recalculation of results. Moreover, after excluding Li et al.
[21 ], between-study heterogeneity dropped substantially, while still maintaining a similar
risk for the groups (RR 1.00, 95%CI 0.98–1.02; I ² = 8%) (Fig. 3s ).
Fig. 3 Forest plots showing no significant difference between the cold endoscopic mucosal
resection (C-EMR) and cold snare polypectomy (CSP) techniques in terms of: a complete resection rate; b en bloc resection rate.
Secondary outcomes
En bloc resection
Five studies reported en bloc resection rates, including 1601 polyps. The pooled estimated
rate in the C-EMR group was 87.5% (95%CI 79.3%–95.8%), and in the CSP group was 82.3%
(95%CI 70.9%–93.6%). The risk was not significantly different for the C-EMR and CSP
techniques (RR 1.08, 95%CI 0.82–1.41; I ² = 99%) ([Fig. 3 ]
b ). The findings were also confirmed in the leave-one-out sensitivity analysis. Similarly
to the complete resection rate, heterogeneity derived predominantly from Li et al.
[21 ]. After the exclusion of this trial, the results remained similar, and there was
no evidence of between-study heterogeneity (RR 1.00, 95%CI 0.99–1.01; I ² = 0%) (Fig. 4s ).
Adverse events
All studies reported the overall rate of AEs, including 2287 polyps. The pooled estimated
rate in the C-EMR group was 3.4% (95%CI 1.3%–5.5%), and in CSP group was 4.8% (95%CI
2.0%–7.6%). The risk was similar between C-EMR and CSP (RR 0.74, 95%CI 0.41–1.32;
I ² = 38%) ([Fig. 4 ]). In the leave-one-out analysis, the greatest heterogeneity was found to arise from
the study of Kim et al. [6 ]. After the exclusion of this study, we found a statistically significant reduction
in the risk of AEs in the C-EMR group, with nil between-study heterogeneity (RR 0.56,
95%CI 0.36–0.87; I ² = 0%) (Fig. 5s ).
Fig. 4 Forest plot of adverse events rate showing no significant difference between the cold
endoscopic mucosal resection (C-EMR) and cold snare polypectomy (CSP) techniques.
Bleeding end points
Immediate bleeding was reported in all studies, including a total of 2287 polyps.
The pooled estimated rate in the C-EMR group was 1.4% (95%CI 0.4%–2.3%), and in CSP
group was 1.2% (95%CI 0.1%–2.3%). No significant difference in the risk of immediate
bleeding was observed between the C-EMR and CSP groups (RR 1.18, 95%CI 0.36–3.86;
I ² = 56%) (Fig. 6s ). These results were confirmed on leave-one-out sensitivity analysis. Exclusion of
Li et al. [21 ] removed all the between-study heterogeneity, and the results remained similar between
the groups (RR 2.18, 95%CI 0.82–5.77; I ² = 0%) (Fig. 7s ).
Delayed bleeding was reported in six studies, including 2137 polyps. The pooled estimated
rate in the C-EMR group was 0.6% (95%CI 0.01%–1.1%), and in CSP group was 0.5% (95%CI
0.1%–0.9%). The risk was not significantly different between the C-EMR and CSP groups
(RR 0.98, 95%CI 0.39–2.44; I ² = 0%) (Fig. 8s ). Leave-one-out sensitivity analyses confirmed these results after the exclusion
of each trial.
Perforation
Six studies described the perforation rate in the procedures. A meta-analysis of this
outcome was not possible owing to the absence of events. Only Li et al.
[21 ] reported one event in the CSP group. Nevertheless, our meta-analysis shows that
perforations are rare in both C-EMR and CSP.
Procedure time
Six studies reported procedure time, including 2073 polyps. The mean procedure time
in the C-EMR group was 128.5 seconds (95%CI 86.6–170.3 seconds), and in the CSP group
was 85.6 seconds (95%CI 38.8–132.3 seconds). The procedure time was longer with C-EMR
than with CSP (mean difference 42.1 seconds, 95%CI 14.5–69.7 seconds; I ² = 97%) ([Fig. 5 ]). In the leave-one-out analysis, procedural duration remained longer in the C-EMR
group with the removal of each individual study. No single study was solely responsible
for the high heterogeneity in this outcome.
Fig. 5 Forest plot of procedure time showing cold snare polypectomy (CSP) was associated
with a shorter procedure time than cold endoscopic mucosal resection (C-EMR), but
with significant heterogeneity among the studies.
Subgroup analysis
In a stratified subgroup analysis based on lesion size, the risk of complete resection
was similar between C-EMR and CSP for polyps<10 mm (RR 1.01, 95%CI 0.99–1.04; I ² = 46%) and polyps>10 mm (RR 1.09, 95%CI 0.82–1.44; I
2 = 94%) (Fig. 9s ).
In a stratified subgroup analysis based on nonpedunculated morphology, there was no
statistically significant difference in the risk of complete resection (RR 1.0, 95%CI
0.98–1.02; I ² = 22%) (Fig. 10s ); en bloc resection (RR 1.0, 95%CI 0.99–1.01; I ² = 0%) (Fig. 11s ); or AEs (RR 0.62, 95%CI 0.30–1.31; I ² = 18%) (Fig. 12s ) between C-EMR and CSP. C-EMR was also associated with a longer procedure time when
analyzing only nonpedunculated polyps (mean difference 73.2 seconds, 95%CI 63.3–83.1
seconds; I ² = 0%) (Fig. 13s ).
A summary of the findings with their grades of evidence is shown in Table 2s.
Discussion
In this systematic review and meta-analysis of seven RCTs, we compared C-EMR with
CSP for the resection of colorectal polyps. The main findings were as follows. (i)
No significant difference was observed in the likelihood of complete resection between
the C-EMR and CSP techniques, including on stratified analyses by size and morphology
of the polyps. (ii) There was no significant difference in the risk of either en bloc
resection or AEs between the groups, even on stratified analyses by morphology of
the polyps. (iii) C-EMR was associated with a statistically significant longer procedure
time, even when considering stratified analysis by nonpedunculated morphology of the
polyps.
Most of the available evidence regarding the efficacy of C-EMR for the resection of
colorectal polyps is derived primarily from observational studies [13 ]
[14 ]
[15 ]
[16 ]
[29 ]. A recent meta-analysis of observational studies describing C-EMR reported a complete
resection rate of 96.3% for the resection of colorectal polyps ≤20 mm and a single-arm
meta-analysis showed a complete resection rate of 99.1% when analyzing C-EMR in polyps
≥10 mm [17 ]
[18 ]
[30 ]. The incomplete resection rate was also very low (3.8%) with C-EMR for 4–20 mm polyps
in a recent observational study [31 ].
Other studies have focused primarily on comparing C-EMR with EMR, as the former could
potentially prevent thermal injury and other disadvantages of electrocautery use.
A meta-analysis of sessile serrated polyps>10 mm reported lower delayed bleeding rates
for C-EMR compared with conventional EMR [32 ]; however, there was still uncertainty as to whether C-EMR could reduce AEs relative
to CSP.
C-EMR could theoretically bring more safety to endoscopic resection and protect against
perforations through the submucosal cushion it provides [17 ]. But our series showed only one occurrence of perforation in the control group (1111
CSP procedures), raising doubts about the need to perform routine C-EMR to prevent
such an infrequent event. Importantly, major AEs were infrequent in both groups.
Previous studies showed AE rates lower than 5.4% for C-EMR [10 ]
[11 ]
[12 ]
[13 ]
[14 ]
[15 ]
[21 ]
[29 ]
[33 ]. Tziatzios et al. showed in their network meta-analysis [34 ] that, in terms of preventing AEs, there is a benefit to choosing C-EMR over EMR;
however, when comparing C-EMR to CSP, there was no statistically significant difference
[5 ]
[20 ]
[23 ]
[34 ]. Our results, which included a larger number of RCTs, also showed equivalent safety
for C-EMR and CSP.
After conducting a sensitivity analysis for overall AE rate and excluding Kim et al.
[6 ], we found a significant benefit of C-EMR compared with CSP, and the heterogeneity
was eliminated (RR 0.56, 95%CI 0.36–0.87; I ² = 0%). The high between-study heterogeneity and the change in the overall AE rate
after sensitivity analysis could be attributed to the fact that Kim et al. did not
include minor AEs in their analyses, nor did they provide specific data about polyp
morphology, so the number of pedunculated polyps was unknown.
The longer procedure time associated with C-EMR (mean difference 42.1 seconds, 95%CI
14.5–69.7; I ² = 97%) can be explained by the fact that C-EMR requires additional time for submucosal
injection. The elevated heterogeneity among studies for this outcome is likely due
to differences between endoscopic experience and other procedural variations between
different institutions. As there is no evidence demonstrating the superiority of C-EMR
over CSP in terms of the efficacy and safety outcomes, we conclude that prolonging
the procedure time with C-EMR is not justified in this population. Further studies
with stratified analyses (i.e. by polyp histology and location) could however potentially
identify the benefits of this technique for specific populations.
Concerning the moderate-to-high heterogeneity identified in several outcomes, the
inclusion of pedunculated polyps in one RCT [21 ] and the lack of morphology information in another two [4 ]
[6 ], which raises the question of whether the latter also included pedunculated polyps,
may be contributing factors. The subgroup analysis of nonpedunculated polyps showed
similar results with low heterogeneity across all outcomes. Additionally, sensitivity
analyses excluding the study by Li et al. [21 ], in which 16% of the evaluated colorectal lesions were pedunculated, resulted in
a reduction in the heterogeneity in the outcomes of complete resection, en bloc resection,
and immediate bleeding. These findings suggest that the inclusion of pedunculated
polyps may be the primary cause of the heterogeneity identified.
There are also some peculiarities regarding the histology and location of the polyps
that
may also contribute to the heterogeneity. It is postulated that the soft consistency
of some
histologic types of polyps, such as sessile serrated polyps, maximizes the effectiveness
of a
cold snaring approach, making them potentially more susceptible to C-EMR [32 ]. The residual adenoma rate of sessile serrated polyps <20 mm after C-EMR was 1%
in
the study by Thoguluva Chandrasekar et al. [32 ]. Furthermore, there is growing evidence that this modality would be more appropriate
for large and right-sided sessile serrated polyps, with the aim of minimizing AEs
[32 ]. In our meta-analysis, the commonest histologic polyp type was adenoma in both groups
(C-EMR, 86.5%; CSP, 87.8%), but we were not able to provide a subgroup analysis based
on
histology owing to the lack of stratified analysis by polyp histology in the included
RCTs.
Moreover, it is important to highlight that variations in patient characteristics
among the studies might have influenced the heterogeneity. For instance, although
all studies shared the common inclusion criteria of adult patients undergoing colonoscopy,
Wei et al. [20 ] specifically focused on a male population, which comprised 97.6% of their study
cohort.
There are several strengths in our study, primarily the randomized nature of the included
studies, with these not being subject to the risk of confounding factors that are
present in observational data. In addition, the overall number of polyps included
in the meta-analysis was large (2287 polyps) and we used rigorous methods that were
prespecified in the meta-analysis protocol.
This systematic review also has several limitations. First, we could not stratify
all outcomes by lesion characteristics including location (proximal, distal, rectal)
and histology owing to the lack of individual patient-level data. Second, we found
moderate-to-high levels of heterogeneity for some outcomes, which may be related to
differences in patient and polyp characteristics as discussed previously. In addition,
interendoscopist variability in performance likely exists in most centers, including
those with esteemed experts, as highlighted in a recently published editorial [35 ]. Importantly, despite the heterogeneity, our results were consistent on leave-one-out
sensitivity analysis for most results, with the exception of AEs, as discussed previously.
Finally, it should be noted that analyzing multiple polyps from the same patient may
lead to statistical dependencies.
In conclusion, this systematic review and meta-analysis of RCTs demonstrated that
C-EMR has similar efficacy and safety to CSP, but significantly increases the procedure
time.
