Cold snare endoscopic resection for large colon polyps: a randomised trial

Abstract

Background

Complications of endoscopic mucosal resection (EMR) of large colorectal polyps remain a concern.

Objective

We aimed to compare safety and efficacy of cold EMR (without electrocautery) to hot EMR (with electrocautery) of large colorectal polyps.

Design

In this multicentre randomised trial, patients with any large (≥20 mm) non-pedunculated colon polyp were assigned to cold or hot EMR (primary intervention), and to submucosal injection with a viscous or non-viscous solution (secondary intervention) following a 2×2 design. The primary outcome was the rate of severe adverse events (SAEs). The secondary outcome was polyp recurrence. In this study, we report results of the primary intervention.

Results

660 patients were randomised and analysed. An SAE was observed in 2.1% of patients in the cold EMR group and in 4.3% in the hot EMR group (p=0.10) (per protocol analysis 1.4 vs 5.0%, p=0.017) with fewer perforations following cold EMR (0%) compared with hot EMR (1.6%, p=0.028). Postprocedure bleeding did not differ (1.5% vs 2.2%, p=0.57). The effect of cold resection was independent of the type of submucosal injection solution, polyp size or antithrombotic medications. Recurrence was detected in 27.6% and 13.6% in the cold and hot EMR groups, respectively (p<0.001). Recurrence was not significantly different for 20–29 mm polyps (18.6% vs 13.4%, p=0.24) and for sessile serrated polyps (14.1% vs 8.5%, p=0.33).

Conclusion

Universal application of cold EMR did not significantly lower SAEs (unless cold EMR could be completed) and doubled the recurrence rate compared with hot EMR.

INTRODUCTION

The efficacy of colorectal cancer screening relies on complete resection of precancerous lesions. This is particularly relevant for large ≥20 mm polyps with a greater risk of prevalent and incident cancer. Endoscopic resection has evolved to become the primary approach for removing large colorectal lesions. Despite an improvement in technique, complications occur in 5–8% of patients, with postprocedure bleeding being the most common.^{1 2}

Traditionally, resection of large non-pedunculated colorectal polyps has been performed by electrocautery or hot snare endoscopic mucosal resection (EMR). However, electrocautery increases the risk of deep mural injury, perforation, postpolypectomy syndrome and postprocedure bleeding. While cold snare resection without electrocautery has become the standard approach for small polyps up to 10 mm,^{1 3} recent studies also support cold snare resection for 10–19 mm polyps with excellent safety and variable efficacy results.^4–6 Few case series have reported on the use of cold snare EMR for polyps 20 mm or larger with favourable safety results, however, with in part high recurrence rates.^7–9 Two recent randomised trials from Australia and Germany compared hot with cold snare EMR of ≥15 mm and ≥20 mm colorectal polyps, respectively, showed a significant safety benefit for cold EMR, but high recurrence rates of 18% and 24%.^{10 11} These studies were limited by preselection of polyps for cold EMR, including polyps<20 mm in size,¹⁰ and lack of consistent clip closure and margin ablation after hot EMR,¹¹ the currently recommended standards.^{1 3}

It is unknown how a cold and hot EMR strategy for large polyps compare that applies current best hot EMR practice and limits polyp selection bias compare. We therefore conducted a randomised controlled trial (RCT) to examine the safety and efficacy of a universal application of cold EMR compared with hot EMR for 20 mm or larger non-pedunculated colorectal polyps. We hypothesised that cold EMR would be safer but less effective than hot EMR.

METHODS

Patient selection and study design

This multicentre randomised trial enrolled patients across 15 medical centres (14 in the USA and 1 in Canada) between October 2019 and January 2023. Participants were assigned following a 2×2 factorial design to cold snare EMR or hot snare EMR of large (≥20 mm) colorectal polyps, and to one of two submucosal injection solutions, either viscous solution (Eleview) or normal saline (control). Type of resection (cold vs hot EMR) was the primary intervention and type of submucosal injectate was the secondary and exploratory intervention to examine its effect on safety and risk of recurrence. Participants were assigned to one of four study groups (hot EMR+viscous solution, hot EMR+saline, cold EMR+viscous solution, cold EMR+saline) based on a computer-generated randomisation list in blocks of eight stratified by centre using sequentially numbered concealed envelopes. Envelopes with the assigned group were opened after polyp inclusion criteria were met and before polyp resection was initiated to minimise the potential for polyp selection bias. If the patient had more than one eligible study polyp, all lesions were resected according to the randomisation assignment (randomisation by patient). In this study, we report results related to the primary intervention.¹²

Eligible patients included all those who presented for resection of a large colorectal polyp and were at least 18 years of age. All non-pedunculated large polyps were eligible. This decision was driven by the intent to minimise selection bias of enrolling polyps that could more easily be removed by cold EMR. In addition, the specific polyp characteristics that would preclude cold EMR were unclear. We therefore expected some cross-over from cold to hot EMR and anticipated that the obtained data would provide relevant information for selecting lesions that would be amenable to cold EMR. Patients were excluded if they had inflammatory bowel disease, severe comorbidities (American Society of Anesthesiologists score>3), or a coagulopathy (international normalised ratio (INR) >1.5 or platelets<50). Pedunculated polyps (Paris 1P), ulcerated polyps (Paris 3) or those with suspected deep submucosal invasive cancer based on morphological assessment were excluded.

The study was approved by the institutional ethical review boards and registered at ClinicalTrials.gov (NCT 03865537). All participants gave written informed consent.

Procedure

Colonoscopies were performed according to local standard of practice. Once a potential study polyp was identified it was assessed for eligibility. If exclusion criteria were absent and polyp inclusion criteria were met, the randomisation envelope was opened, and the patient was assigned to one of the four possible study groups. Patients were not informed about the allocation; however, the colonoscopy reports contained information on the mode of resection.

Polyp size was measured by comparison to an open snare with known diameter. All polyps were removed by EMR, consisting of submucosal injection followed by transection of the polyp with a snare. Submucosal injection was performed according to randomisation. By protocol, all submucosal injection solution contained epinephrine at a concentration between 1:100 000 to 1:500 000.

In the cold snare group, resection was performed with a dedicated cold snare at the discretion of the endoscopist with the goal to obtain a large (at least 3 mm) healthy margin (figure 1). We anticipated that some large bulky polyps would not be amenable to cold snare resection and permitted endoscopists to crossover to hot snare resection if completeness of cold resection was deemed not possible. If any portion of the polyp was resected with electrocautery, then the patient was considered to have crossed over to the hot snare group. Endoscopists were asked to remove any residual polyp tissue by mechanical means if possible (eg, forceps) and only use cautery or ablative methods if mechanical means were not successful. According to the study protocol defects after cold snare resection were not to be closed. To ensure adequate quality of cold EMR, all participating endoscopists were required to submit a video of them performing a cold EMR of a large polyp. Endoscopists were only allowed to enrol patients if the quality of cold EMR was considered adequate as judged by two assessors with prior extensive experience with cold EMR.

Figure 1.

Hot EMR of Paris 2a, LST-G lesion in the sigmoid colon (A), after margin ablation using snare tip soft coagulation (B) and scar at surveillance colonoscopy (C). Cold EMR of Paris 2a, LST-G lesion (D) with wide normal resection margin (E), and scar at surveillance colonoscopy (F). EMR, endoscopic mucosal resection; LST-G, lateral spreading tumour of granular type.

In the hot EMR group resection was performed per usual care (figure 1). Snare choice was at the discretion of the endoscopist. Per guidelines, the study protocol required complete ablation of the resection margin after resection, either using snare tip soft coagulation (STSC) or argon plasma coagulation (APC). Preresection marking of the resection margin was also permitted.¹³ Crossover was not anticipated; however, if, per endoscopist discretion a polyp was resected cold (entire resection without electrocautery), it was considered a crossover. Further, the study protocol also required clip closure of resection defects of all lesions proximal to the splenic flexure.¹ Complete clip closure was defined as approximation of opposite resection margins with clips that were less than 1 cm apart with none or minimal visible submucosal tissue.

Periprocedural antithrombotics were managed according to current guidelines.¹⁴ We considered active antithrombotic medication use if the patient was on an antiplatelet agent within 7 days of resection or on an anticoagulant (warfarin or direct oral anticoagulant) within 5 days of resection. Histopathology examination of polyps was performed at the local pathology department.

Outcomes

The primary outcome was the rate of severe adverse events (SAEs). The reason for choosing this composite endpoint is related to the underlying pathophysiology of adverse events with hot EMR. Cautery may not only be associated with postprocedure bleeding, but can cause injury to the muscularis propria, which may result in abdominal pain, postpolypectomy syndrome or perforation. An SAE was defined as a clinically significant event that occurred during the procedure or up to 30 days following the procedure and posed a threat of permanent disability or death, requiring hospitalisation, blood transfusion, a colonoscopy or surgery. This definition is in line with prior studies and in accordance with the definition of the US Food and Drug Administration.¹⁵ Postprocedure adverse events were ascertained by phone call or during a clinical visit at least 30 days following the procedure, and by review of medical records. All SAEs were reviewed by a data safety monitoring board.

Secondary outcomes of interest included the rate of recurrence at first surveillance colonoscopy at all polyp resection scars that were identified. Recurrence was defined as a resection site that harboured any visible or biopsy-proven neoplastic polyp tissue at first surveillance colonoscopy following prior complete polyp resection. The protocol directed endoscopists to examine resection scars with high-definition white light and digital chromoendoscopy. Biopsies were not required if the absence of recurrent polyp tissue was determined visually with high confidence.^{16 17} We further examined SAEs by type, including bleeding, perforation,¹⁸ postpolypectomy syndrome and abdominal pain, as defined in online supplemental table 1. SAEs were also assessed by polyp location (proximal vs distal) and by use of antithrombotic medications. In addition, we assessed whether recurrence was affected by polyp size (20–29 mm versus >30 mm), polyp histology (serrated vs adenomatous), polyp morphology (flat or Paris 2a vs sessile or any Paris 1s), and polyp height as a marker for polyp bulkiness (<10 versus ≥10 mm).

Analysis

The primary analysis followed an intention-to-treat (ITT) principle. SAE and recurrence rate are expressed as absolute risks and absolute risk differences (ARD) with 95% CIs.

In general, continuous variables are expressed as means with SD or as medians with IQR and compared using the Student’s t-test or the Mann-Whitney-Wilcoxon test, respectively. Proportions are presented as percentages and were compared using the χ² test or the Fisher’s exact test as appropriate. Multivariable logistic regression was applied to compare the primary outcome between the study arms controlling for differences in baseline characteristics between hot and cold EMR groups (anticoagulant medications, presence of more than one study polyp, prior resection attempts). When polyps were the unit of analysis, we accounted for clustering of multiple study polyps within the same patient using generalised estimating equations. We also used this approach to assess variation in recurrence rates between participating sites. With respect to crossover, we examined polyp-associated characteristics using multivariable stepwise backwards logistic regression analysis. We further performed per protocol analyses using the conventional per protocol estimator that included all patients who completed the assigned intervention. Because this approach is known to be biased, the analysis was supplemented by the complier average causal effect (CACE) to estimate the treatment effect (cold EMR) on SAE and on recurrence rate. This approach provides an unbiased estimation of the treatment effect in the subgroup of patients for which the endoscopist would comply with the randomised intervention.¹⁹

The sample size calculation was based on the primary intervention. We assumed an overall SAE incidence following hot snare EMR of 7%.^{2 20} We considered a decreased SAE rate in the cold EMR group to 1.5% as clinically important. We further assumed that 10% of patients in the cold EMR group would cross over to hot EMR, which would increase the SAE rate in the cold EMR group in an ITT analysis to 2%. To show such a difference (7% vs 2%) as significant and assuming a two-sided alpha of 0.05 and a power of 0.80, 650 subjects needed to be randomised. An interim analysis was performed after 50% of patients were randomised to assess the rate of cross-over from cold to hot EMR and to modify polyp inclusion criteria should the cross-over rate exceed 15%. Because this threshold was not reached, inclusion criteria were not changed.

To facilitate reading, numbers were rounded in the text. Tables provide more precision. All coauthors had access to the study data and have reviewed and approved the final manuscript.

Patient and public involvement

Patients and public were not involved in the design, conduct or reporting of the study.

RESULTS

660 patients were randomised to cold EMR (n=336) or hot EMR (n=324) (figure 2). The mean age was 65 and 52% were men. Patient baseline characteristics were generally similar in the two groups (table 1). In the cold EMR group significantly more patients were on anticoagulants (11% vs 6.5%) and had more than one ≥20 mm polyp (8.9% vs 4.9%) compared with the hot EMR group. The median polyp size was 30 mm and 83% of polyps were located in the proximal colon (table 2). Most polyps were adenomatous, 19% were serrated lesions, and 2.5% harboured invasive cancer. Fewer polyps had prior resection attempts in the cold EMR group compared with the hot EMR group (7.5% vs 12%). In the cold EMR group, 14.6% of polyps were removed by hot EMR, and in the hot EMR group, 14.0% were removed by cold EMR. Reasons for crossover are shown in figure 2. Following an ITT analysis, visibly complete polyp removal was achieved in 98% in both groups. In the hot EMR group, complete margin treatment was accomplished for 82% of polyps (87% per protocol analysis), in the majority using STSC. In the cold EMR group adjunctive complete margin treatment was applied for 15% of polyps (4% per protocol analysis). Complete closure of the resection defect was achieved for 61% of proximal polyps in the hot EMR group (71% per protocol analysis). Defects were closed for 9% of polyps in the cold EMR group (3% per protocol analysis). Total resection time, including adjunctive interventions (ie, ablation or clipping), did not significantly differ between groups.

Figure 2.

Flow chart. EMR, endoscopic mucosal resection; SAEs, severe adverse events; SC1, first surveillance colonoscopy.

Table 1.

Characteristics of the patients at baseline

Characteristics	Cold snare group (n=336)	Hot snare group (n=324)	P value
Patients
Age, years, mean (SD)	65.6 (10.0)	65.0 (10.5)	0.45
Male sex, n (%)	179 (53.3)	166 (51.2)	0.60
ASA class, n (%)			0.10
I	25 (7.4)	15 (4.6)
II	193 (57.4)	210 (64.8)
III	118 (35.1)	99 (30.6)
Periprocedural antithrombotic medications, n (%)	109 (32.4)	94 (29.0)	0.34
Antiplatelet agents	82 (24.4)	76 (23.5)	0.78
Anticoagulants	37 (11.0)	21 (6.5)	0.040
Procedure
Sedation, n (%)			0.23
No sedation	3 (0.9)	0
Moderate sedation	42 (12.5)	40 (12.4)
Monitored anaesthesia care	291 (86.6)	283 (87.6)
Quality of bowel preparation, n (%)			0.326
Excellent	167 (49.7)	167 (51.5)
Good	131 (39.0)	124 (38.3)
Fair	37 (11.3)	33 (10.2)
Submucosal injection (randomisation), n (%)			0.70
Viscous solution	159 (49.1)	170 (50.6)
Normal saline	165 (50.9)	166 (49.4)
More than one≥20 mm study polyp, n (%)	30 (8.9)	16 (4.9)	0.044
ASA, American Society of Anesthesiologists.

Table 2.

Characteristics of study polyps and resection

Characteristics	Cold snare group (n=371)	Hot snare group (n=343)	P value
Size, median, mm (IQR)	30 (15)	30 (15)	0.19
Proximal location*, n (%)	306 (82.5)	286 (83.4)	0.75
Ileocecal valve involved, n (%)	20 (5.4)	27 (7.9)	0.18
Appendiceal orifice involved, n (%)	10 (2.7)	3 (0.9)	0.07
Flat morphology†, n (%)	207 (55.8)	166 (48.4)	0.11
Histology, n (%)			0.79
Tubular adenoma	164 (44.2)	161 (46.9)
Tubulovillous or villous adenoma	124 (33.4)	106 (30.9)
Serrated lesion‡	71 (19.1)	67 (19.5)
Cancer	10 (2.7)	8 (2.3)
Other	2 (0.5)	1 (0.3)
Grade of dysplasia, n (%)			0.94
Low grade dysplasia	240 (64.7)	219 (63.8)
High grade dysplasia	59 (15.9)	60 (17.5)
Prior resection attempts, n (%)	28 (7.5)	42 (12.2)	0.035
EMR difficulty§, n (%)			0.13
Easy/neutral	277 (74.7)	238 (69.6)
Difficult	94 (25.3)	104 (30.4)
Submucosal lifting§, n (%)			0.31
Complete	299 (80.8)	266 (77.8)
Partial	62 (16.8)	61 (17.8)
Non-lifting	9 (2.4)	15 (4.4)
Epinephrin within injection solution§, n (%)	346 (93.5)	305 (89.2)	0.06
En bloc resection, n (%)	23 (6.7)	4 (1.1)	<0.001
Complete resection, n (%)	364 (98.1)	336 (98.0)	0.99
Treatment of resection margin¶, n (%)			<0.001
Complete**	54 (14.8)	276 (82.1)
Partial	6 (1.6)	26 (7.7)
Type of margin treatment¶, n (%)			<0.001
STSC	49 (13.5)	230 (68.5)
APC	11 (3.0)	20 (6.0)
Hot snare trimming	0	4 (1.2)
Cold snare trimming	0	46 (13.7)
Margin marking only	0	2 (0.6)
Time of resection, median minutes (IQR)††	30 (22)	26 (21)	0.21
Crossover, n (%)	54 (14.6)	48 (14.0)	0.83
Complete defect closure, n (%)	34 (9.2)	188 (54.8)	<0.001
Proximal polyps, n (%)‡‡	28 (9.2)	175 (61.2)	<0.001
Distal polyps, n (%)	6 (9.2)	13 (22.8)	0.033

^*Proximal to splenic flexure.

^†Sessile was defined as a polyp with a Paris Is component, and flat as a polyp with Paris IIa, IIb or IIc.³⁵

^‡Includes 3 hyperplastic polyps, 131 sessile serrated adenomas/polyps and 5 traditional serrated adenomas.

^§Missing: EMR difficulty n=1, submucosal lifting n=2, margin treatment n=5 and epinephrine injection n=2.

^¶Among all 700 completely removed polyps.

^**Per protocol: cold EMR 12 (3.8%), hot EMR 251 (86.9%).

^††Defined as the time from starting the submucosal injection to completion of resection (does not include clipping).

^‡‡Per protocol: cold EMR 7 (2.6%), hot EMR 173 (70.9%).

APC, argon plasma coagulation; EMR, endoscopic mucosal resection; STSC, snare tip soft coagulation.

Primary outcome: severe adverse events (SAEs)

657 patients had complete 30-day follow-up data. SAEs were observed in 7 patients (2.1%) in the cold EMR group and in 14 patients (4.3%) in the hot EMR group, representing a non-significant difference (p=0.10, ARD 2.3%, 95% CI −0.4% to 5.0%) (table 3). There were five perforations (1.6%) with hot EMR and none with cold EMR (p=0.028, ARD 1.6%, 95% CI 0.2% to 2.9%). Three perforations (two deep mural injury type IV, one deep mural injury type III) were noticed during the resection and were successfully closed. Two patients with postprocedure pain had microperforations on CT and were treated with antibiotics. None required surgery. The rate of postprocedure bleeding (1.5% vs 2.2%), postpolypectomy syndrome (0% vs 0.6%) or abdominal pain (0.3% vs 0%) were similar in the cold and hot EMR groups, respectively. There was no significant difference in the need for colonoscopy or blood transfusions. One patient in the cold EMR group, who crossed over to hot EMR, was admitted 2 days after the procedure with myocardial infarction and congestive heart failure. The patient subsequently developed ventricular arrhythmias and renal failure and died on day 5 of admission. Need for colonoscopy or blood transfusions did not significantly differ between groups. Online supplemental table 2 provides outcomes for baseline characteristics that differed between both groups at baseline. While SAE were significantly different after excluding patients on anticoagulant medications (1.3% vs 4.3%, p=0.046), no factor remained significant after adjusting for baseline differences (unadjusted OR 0.47, 95% CI 0.19 to 1.18, adjusted OR 0.42, 95% CI 0.16 to 1.08). Similarly, there was no effect of differences in baseline characteristics on recurrence (data not shown).

Table 3.

Severe adverse events (SAE) in intention-to-treat and per protocol analysis

Characteristics and outcomes	Intention-to-treat analysis			Conventional per protocol analysis
	Cold snare group (n=336)	Hot snare group (n=324)	P value	Cold snare group (n=282)	Hot snare group (n=281)	P value
30-day follow-up complete, n	335	322		281	280
Any SAE, n (%)	7 (2.1)	14 (4.3)	0.10	4 (1.4)	14 (5.0)	0.017*
Postprocedure bleeding, n (%)	5 (1.5)	7 (2.2)	0.57	3 (1.1)	7 (2.5)	0.22
On antithrombotic medications	3 (2.8)	4 (4.3)	0.70	3 (3.1)	4 (4.7)	0.70
No antithrombotic medications	2 (0.9)	3 (1.3)	1.0	0	3 (1.5)	0.25
Proximal location	2 (0.7)	7 (2.6)	0.10	2 (0.8)	7 (3.0)	0.10
Distal location	3 (5.3)	0	0.24	1 (2.7)	0	0.44
Perforation, n (%)	0	5 (1.6)	0.028	0	5 (1.8)	0.030†
Postpolypectomy syndrome, n (%)	0	2 (0.6)	0.24	0	2 (0.7)	0.25
Abdominal pain, n (%)	1 (0.3)	0	1.0	1 (0.4)	0	1.0
Cardiovascular event, n (%)	1 (0.3)	0	1.0	0	0
SAE intervention, n (% of all SAE)
Colonoscopy	4 (1.2)	5 (1.5)	0.75	2 (0.7)	5 (1.8)	0.29
Blood transfusion	3 (0.9)	0	0.25	2 (0.7)	0	0.50
Surgery	0	0		0	0
Death	1 (0.3)	0	1.0	0	0

^*Complier average causal effect per protocol (CACE-PP) estimator (unbiased): cold EMR 1.7%, hot EMR 4.8%, p=0.10.

^†CACE-PP estimator: cold EMR 0%, hot EMR 2.2%, p=0.025. CACE-PP estimates remained non-significant for the remaining variables. EMR, endoscopic mucosal resection.

Subgroup analyses

There was no significant difference in SAE by type of submucosal injection agent (2.4% with normal saline vs 4.0% with viscous solution, p=0.26) and ARDs in the two groups were similar for both injection solutions. Fewer SAEs were observed following cold EMR for proximal polyps when compared with hot EMR (ARD 3.0%, 95% CI 0.2% to 5.8%). There was no difference between groups for polyps located in the distal colon (figure 3). When only analysing patients who were not on anticoagulant medications the risk for SAE was lower in the cold EMR group than in the hot EMR group (1.3% vs 4.3%, 95% CI 0.3% to 5.6%), but there was no difference when only analysing patients on anticoagulants. No differences in ARDs were seen for subgroups of polyp size.

Figure 3.

Severe adverse events, primary outcome and subgroup analyses (intention-to-treat analysis). EMR, endoscopic mucosal resection; NS, normal saline; VS, viscous solution.

Secondary outcome: recurrence at first surveillance colonoscopy

Of 615 patients who were eligible for follow-up (figure 2), a similar proportion underwent a surveillance colonoscopy in both groups after a median of 6.2 months (table 4). Overall, 99% of all resection scars were identified and examined. Among all examined resection sites, recurrence was observed in 28% in the cold EMR group and 14% in the hot EMR group (p<0.001, ARD 14%, 95% CI 9.7% to 18%). The ARD was very similar after inverse weighting to make the estimate representative of all patients, not just those completing surveillance. In subgroup analysis, a significant difference was seen for adenomatous polyps (31% vs 15%, p<0.001) and for polyps≥30 mm (36% vs 14%, p<0.001), but not for serrated lesions (14% vs 8.5%, p=0.33) and 20–29 mm polyps (19 vs 13%, p=0.24). Recurrence rates were lower for polyps with complete margin treatment in both groups when compared with those without margin treatment. Similarly, polyp morphology or height did not affect risk differences between groups. Four patients (10%) in the hot EMR group and none in the cold EMR group were referred for surgical resection because recurrent polyp tissue could not be managed endoscopically. Among these, three index polyps were 50 mm in size, two contained high-grade dysplasia and one was located at the ileocecal valve.

Table 4.

Recurrence at first surveillance colonoscopy, intention-to-treat and per protocol analysis

	Intention-to-treat analysis			Conventional per protocol analysis
Characteristics and outcomes	Cold snare group	Hot snare group	P value	Cold snare group	Hot snare group	P value
Eligible patients	309	306	0.34	267	266	0.94
Eligible polyps	342	324	0.36	299	277	0.92
First surveillance colonoscopy (SC1)
Patients undergoing SC1, n (%)	286 (92.6)	275 (89.9)	0.24	248 (92.9)	238 (89.5)	0.17
All resection sites at SC1, n (%)	314 (91.8)	289 (89.2)	0.25	275 (92.0)	248 (89.5)	0.31
Time, median months (IQR)	6.2 (2.9)	6.2 (3.2)	0.83	6.2 (2.6)	6.2 (3.2)	0.95
Resection sites identified, n (%)	308 (98.1)	286 (99.0)	0.51	269 (99.2)	246 (99.2)	0.29
Recurrence findings
Recurrence (among all examined resection sites), n (%)	85 (27.6)	39 (13.6)	<0.001	75 (27.9)	27 (11.0)	<0.001*
Size of recurrent polyp, median mm (IQR)	8.0 (6.0)	6.5 (9.5)	0.74	8.0 (5.0)	10.0 (16.5)	0.31
Histology of recurrent polyp
Adenoma	79 (92.9)	31 (79.5)	0.028	70 (93.3)	21 (77.8)	0.025
SSL	6 (7.1)	8 (20.5)	0.028	5 (6.7)	6 (22.2)	0.025
HGD	4 (4.7)	4 (10.3)	0.24	4 (5.3)	3 (11.1)	0.39
Cancer	0	1 (2.6)	0.14	0	1 (3.7)	0.09
Referral for surgical resection	0	4 (10.3)	0.007	0	4 (14.8)	0.005
Recurrence by index polyp characteristics
Size
Size 20–29 mm	27 (18.6)	17 (13.4)	0.24	25 (17.7)	10 (9.3)	0.057
Size ≥30 mm	58 (35.6)	22 (13.8)	<0.001	50 (39.1)	17 (12.3)	<0.001
Height <1 cm	70 (26.5)	36 (14.5)	0.001	64 (25.9)	24 (11.4)	<0.001
Height ≥1 cm	15 (34.1)	3 (7.9)	0.004	11 (50.0)	3 (8.3)	<0.001
Histology
Any adenoma	76 (31.1)	34 (15.0)	<0.001	68 (32.7)	24 (12.0)	<0.001
Any TVA or villous histology	35 (34.0)	13 (13.5)	0.001	35 (34.3)	13 (13.5)	0.001
Serrated lesion recurrence	9 (14.1)	5 (8.5)	0.33	7 (11.5)	3 (6.5)	0.38
HGD	22 (46.8)	10 (19.6)	0.004	19 (51.4)	6 (13.0)	<0.001
Morphology
Flat (only Paris II)	44 (24.2)	20 (13.4)	0.014	42 (24.3)	15 (11.4)	0.004
Sessile (any Paris Is)	41 (32.5)	19 (13.9)	<0.001	33 (34.4)	(10.5)	<0.001
Margin treatment
Complete	9 (21.4)	20 (8.7)	0.013	2 (20.0)	17 (8.0)	0.18
None or partial	76 (28.6)	19 (34.6)	0.38	73 (28.2)	10 (30.3)	0.80

^*Complier average causal effect per protocol estimator (unbiased): cold EMR 29.3%, hot EMR 8.1% p<0.001.

EMR, endoscopic mucosal resection; HGD, high-grade dysplasia; HGD, high grade dysplasia; IQR, interquartile range; SSL, sessile serrated lesion.

Crossover and per protocol analysis

In multivariable analysis, polyps with crossover from cold to hot EMR were larger, taller, located in the rectum and non-lifting, while polyps that crossed over from hot to cold EMR were less likely to be tall (online supplemental tables 3 and 4). While serrated appearance appeared to be a common reason for endoscopists to use cold EMR, after accounting for interaction with lesion height it was no longer significant. Results of the conventional per protocol analysis are shown in tables 3 and 4. SAEs occurred in 4 patients (1.4%) in the cold EMR group and in 14 patients (5.0%) in the hot EMR group (p=0.017. Of these, postprocedure bleeding was observed in 3 patients (1.1%) following cold EMR and in 7 patients (2.5%) following hot EMR. As in the ITT analysis, perforation (n=5, 1.8%) and postpolypectomy syndrome (n=2, 0.7%) only occurred with hot EMR. One patient (0.4%) had postprocedure abdominal pain following cold EMR. Recurrence rates were similar to the ITT analysis. Across participating centres recurrence varied between 0% and 20% following hot EMR and between 18% and 50% following cold EMR (online supplemental figure 1). Accounting for differences between groups after crossover, the unbiased per protocol CACE analysis estimated an SAE risk of 1.7% after cold EMR and 4.8% after hot EMR (p=0.10, ARD 3.1%, 95% CI −0.6% to 6.9%). The respective estimates for recurrence were 29.3% and 8.1% (p<0.001, ARD 21.2%, 95% CI 12.0% to 30.3%).

DISCUSSION

In this multicentre RCT, a universal application of cold EMR for all large polyps resulted in a non-significant lower SAE rate of 2.1% compared with 4.3% following hot EMR. When accounting for a high crossover rate in both directions, SAEs occurred in 1.5% with cold EMR as compared with 5.0% with hot EMR (p=0.017, conventional per protocol analysis). Any upfront benefit in safety was offset by an overall doubling in the polyp recurrence rate between groups, which was observed in 28% following cold EMR and 14% following hot EMR.

The results confirm those of two recent smaller non-US randomised trials. A German study reported SAEs in 1% and 8% in the cold and hot EMR groups, respectively.¹¹ An Australian study that included smaller polyps (≥15 mm) observed SAEs in 1% and 9%, respectively.¹⁰ Recurrence rates in the German trial were very similar to ours’ (24% and 14%), but were lower in the Australian study (11% and 1%). In contrast to our study, both studies excluded polyps that were considered less likely to be removed by cold EMR (eg, larger size, location at the ileocaecal valve location or presence of a nodule).

Several results are noteworthy. First, cold EMR does not equate to zero risk resection, and SAEs, in particular postprocedure bleeding, do occur. Of the three patients with postprocedure bleeding in the cold EMR group, all were on antithrombotic medications (two on anticoagulants), and one had three large polyps removed, of which two were in the cecum. The greater proportion of patients on anticoagulants at baseline may have had an unfavourable effect on the bleeding risk; however, the adjusted analysis did not identify this imbalance as significant. Clip closure after cold EMR might further reduce the bleeding risk, but it is unlikely that this would be cost-effective (assuming a 0% risk, the number needed to treat would be 67). No perforation or postpolypectomy syndrome was observed with cold EMR. In fact, perforations were significantly more frequent following hot EMR.

Second, the rate of SAE following hot EMR in our study was lower (5.0% per protocol) than reported in the two previous trials. In particular postprocedure bleeding occurred in only 2.5% (4.4% in the German study and 7.8% in the Australian study), despite similar definitions of clinically significant bleeding.^{10 11} In contrast to the other studies, clip closure of proximal defects after hot EMR was mandated in our study.²¹ Of all hot EMR defects, 37% were closed in the German study, 43% in the Australian study and 71% in our study. Our proximal closure rate is similar to previous studies^{2 20}; however, currently available closure techniques may now allow complete closure of most resection defects.^{22 23} The lower bleeding rate in the hot EMR group suggests that systematic defect closure of proximal hot EMR sites may reduce or minimise any bleeding differences between hot and cold EMR.

Third, we found a significant safety benefit of cold EMR for proximal polyps but not for distal polyps. This observation is likely related to a greater number of postprocedure bleeding events in the ITT analysis for distal polyps in the cold EMR group (n=3) when compared with hot EMR (n=0). However, in two of these three cases, polyps crossed over to hot EMR. These findings should therefore be viewed with caution. In other subgroup analyses, type of submucosal injection solution, use of anticoagulant medications and polyp size ≥40 mm did not modify the effect of the primary intervention on SAE. In particular, there was no interaction between the mode of resection (hot or cold EMR) and type of submucosal injection solution. Therefore, the presentation of results related to the type of submucosal injection (secondary explorative intervention) will be reported separately.

Fourth, the high recurrence rate of 28% following cold EMR occurred despite instructing endoscopists to obtain a wide resection margin (at least 3 mm). Residual tissue at a cold defect margin may be more easily recognisable than residual tissue at a cauterise margin after hot resection. In contrast to hot EMR, it is therefore plausible that regrowth occurs less likely at the margin, but at the defect base. This idea is supported by the observation that cold resection frequently does not cut through the muscularis mucosa to transect the submucosal plane.^{24 25} Yet, submucosal transection is likely required to ensure complete resection of neoplastic polyps. Techniques to optimise cold resection are the subject of ongoing studies, including the role of submucosal injection, the use of adjunctive ablation, or partial targeted hot resection.

Fifth, recurrence in the hot EMR group (11% per protocol) was moderately lower compared with our previous large-scale RCT (17%), wherein margin treatment was not performed.²⁶ If complete margin treatment was achieved, recurrence rates decreased to 8%. Despite some ablation benefit, these findings are in contrast with recent studies on margin ablation that reported recurrence rates between 1% and 5% with STSC and 9% with APC.^27–29 Margin treatment with either STSC, APC or margin marking before resection was mandated for all polyps randomised to hot EMR in our study,^{13 27–29} with the majority receiving STSC (76%). Different from other studies participating endoscopists did not receive formal training and margin ablation may not have been as effective.^{29 30}

The wide variation in recurrence rates among participating centres for both hot (0–20%) and cold (18–50%) EMR suggests that variation in resection skills and margin ablation skills are substantial even between groups of endoscopists considered experts. It should be acknowledged, however, that data on the benefit (and its performance) were only published in 2020, at the time when the trial had already started. These findings are important and suggest that extensive effort is needed to standardise techniques in multicentre trials of colorectal endoscopic resection. Further, these findings raise concerns that results of single expert centre trials could have reduced generalisability, and that considerable effort is needed to effectively transfer both hot and cold EMR skills to less experienced endoscopists.

Finally, the time of polyp resection was similar for cold EMR (30 min) and hot EMR (26 min), which accounted also for clip closure and margin ablation. This finding diminishes the argument that cold EMR may take too long.

While the results do not support a universal application of cold EMR for all ≥20 mm polyps, further studies may aim to identify subsets of polyps that are amenable to cold resection. With respect to histology, we found no difference in the efficacy between cold and hot EMR for serrated lesions. Recently published recurrence rates between 1% and 4%^{7 31} support the use of cold EMR for serrated lesions. Recurrence was also not significantly different for 20–29 mm polyps. While some studies show effective removal of up to 19 mm polyps by cold resection with <2% recurrence,^{5 32} our results encourage further study of cold EMR for polyps up to 29 mm. We also examined morphology, which had no effect on the observed difference between cold and hot EMR.

The study also provides first results on what polyps may not be amenable to cold EMR. Endoscopists were more likely to use hot EMR for polyps in the rectum and for larger and bulky polyps. Height ≥10 mm as a surrogate for ‘bulkiness’ increased the odds to require hot EMR almost nine times. Of concern are also high recurrence rates of histologically advanced lesions with villous components and high-grade dysplasia. Therefore, polyps with any visual suspicion for advanced histology (eg, JNET (Japanese NBI expert team) 2b, Kudo IV or V) should not undergo cold resection, but be removed by hot EMR or any technique that ensures en bloc resection.^{1 3}

Although our study did not show a significant safety benefit in the ITT analysis, the unexpected high crossover rate, the per protocol analyses, along with the results of other recent studies, indicated that cold EMR for large colorectal polyps is safer than hot EMR. If cold EMR is effective, avoids surgery and prevents cancer, it may be considered an alternative to hot EMR. Hot EMR studies have shown that recurrent polyp tissue can be easily removed.³³ Such data are not yet available for cold EMR. In our study, the size of recurrent polyp tissue was similar in both groups, and no patient in the cold EMR group was referred to surgery, suggesting that recurrence can be effectively treated. Furthermore, some patients may prefer a safer polyp resection at the expense of a higher recurrence rate. This may be especially relevant for patients on anticoagulation with a higher risk of postprocedure bleeding with hot EMR.³⁴ If cold EMR were the selected approach, adherence to follow-up colonoscopy is essential.

Strengths of our study include its sample size and its multicentre design, which is representative of current real-world practice. Furthermore, the comparison of resection strategies minimised selection bias. Finally, the high proportion of patients with complete follow-up data strengthens the validity of the findings.

Several limitations need to be acknowledged. While we expected crossover from cold to hot EMR, the crossover from hot to cold EMR occurred in 14% of polyps. This was only in part related to emerging data that sessile serrated lesion may be effectively removed by cold EMR (29% of crossovers). Other reasons for crossover (eg, concerns for bleeding and presence of fibrosis) raise concern for a lack of equipoise by participating endoscopists. The results of the conventional per protocol analysis showed a significant safety benefit for cold EMR and provided clinically relevant information; however, such analysis loses the benefit of randomisation and is known to be biased. We therefore added an unbiased per protocol estimator, which also showed a greater yet non-significant SAE risk difference. In contrast to the study protocol, margin ablation and defect closure were performed after cold EMR for some lesions, yet this proportion was small after accounting for crossover (3.8% and 2.8%, respectively). Although groups differed with respect to some baseline characteristics, these did not affect the results when adjusting for them. Finally, the quality of cold EMR may have been affected by limited experience. However, all participating endoscopists were required to submit a video case of them performing a cold EMR. Only after cold EMR skill was judged adequate by two blinded investigators with extensive cold EMR experience were endoscopists allowed to participate in the study.

In conclusion, this large multicentre randomised trial showed no significant safety benefit of a universal application of cold EMR for all ≥20 mm non-pedunculated colorectal polyps. If cold EMR could be completed, it was safer than hot EMR. However, cold EMR doubled the risk of recurrence compared with hot EMR. Future studies need to examine if cold EMR may be appropriate for a subset of polyps and identify the best cold EMR resection technique.

Supplementary Material

This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Cold endoscopic mucosal resection (EMR) of large non-pedunculated colorectal polyps may decrease risk of complications, but data on safety and efficacy are limited.

WHAT THIS STUDY ADDS

• This randomised study compared cold with hot EMR among 660 patients undergoing resection of large (≥20 mm) polyps. Universal application of cold EMR for all large polyps resulted in no significant difference in severe adverse events when compared with hot EMR (2.1 vs 4.3%, per protocol 1.4 vs 5.0%); while risk of recurrence was double (28 vs 14%).

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The findings argue against a universal use of cold EMR for ≥20 mm polyps, but it may be appropriate for selected polyps and when the clinical situation indicates the need to prioritise minimal risk of adverse events. Understanding reasons for incomplete resection of large lesions may provide insights on how to improve the efficacy of cold EMR.

Data availability statement

No data are available.