Introduction

The JNET (Japan NBI Expert Team) classification can be used to predict large non-pedunculated colorectal polyp (LNPCP) histology and presence/depth of submucosal invasion (SMI). Japanese endoscopist accuracies >84% have not been replicated in the West; for example, in a recent study of European experts the accuracy of JNET was 55%.  Furthermore, European guidelines suggest both low-grade dysplasia (LGD) and high-grade dysplasia (HGD) can be treated using endoscopic mucosal resection (EMR) thus there is no clinical consequence to interpreting HGD as JNET 2A (excluded in the original interpretation of JNET).

Aims & Methods

We aimed to validate a novel tool CRIS (Colorectal Regular-Irregular Score) by comparing it with JNET for LNPCP histology prediction amongst Western experts.
32 images of flat LNPCPs under narrow-band imaging were consecutively obtained with matching histopathology data. Each image was rated (anonymous online survey) using JNET and then all the images were rated again (order randomised) using CRIS, a tool which simply asks whether the most disordered vascular pattern visible in the image is ‘regular(=CRIS-R), regularly-irregular(=CRIS-RI) or completely-irregular(=CRIS-I)’. No further explanation of JNET or CRIS was offered. Accuracy of JNET and CRIS vs. histopathology was analysed using the original JNET interpretation (where JNET 1/2A and CRIS-R = LGD) and a clinically relevant approach (where JNET1/2A and CRIS-R can also = HGD). JNET2B & CRIS-RI were matched to HGD & superficial SMI and JNET3 & CRIS-I to deep SMI. Correct treatment was defined as pEMR for LGD or HGD, en-bloc EMR/ESD for HGD or superficial SMI and surgery for deep SMI.

Results

12 experts (median 150 colorectal EMRs and 13 ESDs/year) rated 32 images.
The overall accuracy of JNET vs. histopathology using the original JNET interpretation was 69.7% (95%CI) 46.0-89.3%] and CRIS 69.6% (95%CI 45.3-88.7%). Expert agreement was similar between JNET (91.3% (95%CI) 82.8-98.3%) and CRIS (88.5% [95%CI 78.2-97.3%]) with Κ=0.64 (moderate) for JNET and Κ=0.57 (moderate) for CRIS.
The accuracy of JNET 2B was 51.7% (95%CI 45.9-57.4%) and JNET 3 85.8% (95%CI 81.2-89.3%). Comparably CRIS-RI and CRIS-I showed accuracies of 51% (95%CI 45.2-56.7%) and 85.1% (95%CI 80.4-88.7%) respectively.  Correct treatment was determined in 144/288 (50.0%) observations using JNET and 144 (50.0%) using CRIS. Using the clinically relevant approach the accuracy of JNET was 73.9% (95%CI 46.0-89.3%) vs. CRIS 72.3% (95%CI 45.3-88.7%).
The accuracy, sensitivity and specificity of JNET 1/2A changed by +13.2%, -8.0% and +38.0% and CRIS-R by 8.3%, -8.2% and 35.5%. Correct treatment determination increased by 29.5% to 79.5% for JNET and by 27.1% to 77.1% for CRIS. 

	ORIGINAL JNET INTERPRETATION								CLINICALLY RELEVANT INTERPRETATION
Imaging	JNET 1/2A	CRIS-R	JNET 2B	CRIS-RI	JNET 3	CRIS-I	JNET Overall	CRIS Overall	JNET 1/2A	CRIS-R	JNET 2B	CRIS-RI	JNET 3	CRIS-I	JNET Overall	CRIS Overall
Histology	LGD		HGD, superficial SMI		Deep SMI				LGD, HGD		HGD, superficial SMI		Deep SMI
Accuracy % (95%CI)	62.5% (56.7 - 67.8%)	63.9% (58.0 – 69.0)	51.7% (45.9 - 57.4)	51.0% (45.2 - 56.7)	85.8% (81.2 - 89.3)	85.1% (80.4 - 88.7)	69.7% (46.0- 89.3)	69.6% (45.3- 88.7)	75.7% (70.4 - 80.2)	72.2% (66.7 - 77.0)	no change				73.9% (46.0-89.3)	72.3% (45.3- 88.7)
Sensitivity % (95%CI)	77.4% (67.3 - 85.0)	72.6% (62.0 – 81.0)	27.6% (21.1 - 35.0)	28.2% (21.7 - 35.7)	75.0% (61.2 - 85.0)	81.2% (68.0 - 89.8)	53.8% (21.2-85.1)	54.3% (21.7- 89.8)	69.4% (63.0 - 75.2)	64.4% (57.7 - 70.4)					58.4% (21.2-85.1)	56.4% (21.7- 89.8)
Specificity % (95% CI)	56.4% (49.5 - 62.9)	60.3% (53.0 – 67.0)	80.3% (72.7 - 86.1)	78.0% (70.2 - 84.2%)	87.9% (83.1 - 91.4)	85.8% (80.8- 89.6)	76.6% (49.5-91.5)	76.2% (53.4- 89.7)	94.4% (86.5 - 97.8)	95.8% (88.4 - 98.5)					87.4% (72.7- 97.8)	85.8% (70.2- 98.6)
Correct treatment n (%) /288	42.2% (34.3-50.4	43.0% (34.7-51.5)	62.3% (49.8-73.4)	60.3% (48.1-71.3)	55.4% (42.5-67.5)	53.4% (41.4-65.0)	50.0% (44.2-55.7)	50.0% (44.2-55.7)	97.4% (93.0-99.1)	97.9% (93.4-99.4)					79.5% (74.2-83.9)	77.1% (71.7-81.7)

Conclusion

Both JNET and CRIS, a novel score named and described by 3 self-explanatory words, can be used by Western experts to accurately determine LNPCP histology.This suggests JNET may be able to be simplified for non-experts. Importantly the accuracy of both JNET and CRIS can be increased by framing their interpretation versus. histology in a clinically relevant fashion. This has significant implications for signposting patients to correct treatment after detection of LNPCPs especially amongst non-experts.

References

• Ferlitsch M, et al. Colorectal polypectomy and endoscopic mucosal resection (EMR): European Society of Gastrointestinal Endoscopy (ESGE) Guideline (update 2024). In press.
  
• Kobayashi S, Yamada M, Takamaru H, et al. Diagnostic yield of the Japan NBI Expert Team (JNET) classification for endoscopic diagnosis of superficial colorectal neoplasms in a large-scale clinical practice database. United European Gastroenterol J. 2019;7(7):914-923. doi:10.1177/2050640619845987 
• Sano Y, Tanaka S, Kudo SE, et al. Narrow-band imaging (NBI) magnifying endoscopic classification of colorectal tumors proposed by the Japan NBI Expert Team. Dig Endosc 2016; 28: 526–533. 
• Saito Y, Sakamoto T, Dekker E, Pioche M, Probst A, Ponchon T, Messmann H, Dinis-Ribeiro M, Matsuda T, Ikematsu H, Saito S, Wada Y, Oka S, Sano Y, Fujishiro M, Murakami Y, Ishikawa H, Inoue H, Tanaka S, Tajiri H; IEE-JNET Group. First report from the International Evaluation of Endoscopic classification Japan NBI Expert Team: International multicenter web trial. Dig Endosc. 2023 Sep 13. doi: 10.1111/den.14682. Epub ahead of print. PMID: 37702082. 

Disclosure

DJT: Research support and consulting from Olympus, Pentax and Fujifilm.

Introduction

The detection of cancer within large non-pedunculated colorectal polyps (LNPCPs) is suboptimal in non-expert settings. Identification of 6 macroscopic (Blink) features (6BF) of deep submucosal invasive cancer (SMI) has shown significant improvement in the sensitivity of cancer detection by inexperienced endoscopists at the expense of decreased specificity. This study investigates the impact of endoscopic resection scars, which can resemble deep SMI, on the accuracy of non-expert endoscopists using the 6BF evaluation system known as Blink Impression.

Aims & Methods

An online survey was disseminated containing 20 overview images of LNPCPs (1 image/polyp). Images (randomized) were shown before, and after a 2-minute educational video (the intervention) introducing the 6BF - fold deformation, extra redness, depression, chicken skin mucosa, ulceration and spontaneous bleeding. Before and after the intervention participant Blink Impression was elicited (cancer/no cancer) and only after, the presence of the 6BF. Responses were analysed relative to histopathology (no vs superficial vs deep SMI) and expert opinion on the presence of the 6BF and endoscopic resection scars.

Results

7/20 LNPCPs contained cancer (3 superficial, 4 deep SMI), 3 contained scars. 191 participants from 21 countries completed 3,755 observations. 129/191 (67.5%) participants were inexperienced. Expert-identified scarred LNPCPs contained median 1.0 (IQR 2.0) BF vs 1.0 (IQR 2.0) for no cancer, 2.0 (IQR 2.0) for superficial SMI and 3.0 (IQR 2.0) for deep SMI (all P<.001). Presence of ≥2 BF therefore reliably identifies cancer within LNPCPs but not scars. Fold deformation was the only BF identified more commonly amongst scarred LNPCPs (44.6%) vs no cancer (22.5%, P<.001) and vs superficial SMI (33.0%, P<.001), but similar to deep SMI (44.6%, P=1.000). Depression, chicken skin mucosa and spontaneous bleeding were less commonly identified in scarred polyps (12.3%, 7.6% and 1.8%) than both no cancer (33.1%, 14.4% and 8.6%, all P<.001), superficially invasive (68.6%, 14.3% and 5.8%, all P<.001) or deeply invasive polyps (45.9%, 19.6% and 40.5%, all P<.001). Extra redness and ulceration were identified with similar frequency to no cancer in scars (26.4% vs 26.8%, P=0.891 and 10.2% vs 9%, P=0.441 respectively), but less commonly than superficial (64.2%, P<.001 and 15.7%, P<.001) or deep SMI (60.2%, P<.001 and 24.7%, P<.001) (Table 1).

Blink Feature	No SMI	Scar	Superficial SMI	Deep SMI	Comparison	p
Fold deformation	22.5 (20.6-24.4)	44.6 (40.6-48.7)	33.0 (29.1-36.9)	44.6 (41.2-48.2)	No SMI - Scar No SMI - sSMI No SMI - dSMI Scar - sSMI Scar - dSMI sSMI - dSMI	<.001 <.001 <.001 <.001 1.000 <.001
Extra redness	26.8 (24.8-28.8)	26.4 (22.7-30.0)	64.2 (60.3-68.1)	60.2 (56.7-63.7)	No SMI - Scar No SMI - sSMI No SMI - dSMI Scar - sSMI Scar - dSMI sSMI - dSMI	0.891 <.001 <.001 <.001 <.001 0.152
Depression	33.1 (31.0-35.2)	12.3 (9.6-15.0)	68.6 (64.8-72.4)	45.9 (42.3-49.4)	No SMI - Scar No SMI - sSMI No SMI - dSMI Scar - sSMI Scar - dSMI sSMI - dSMI	<.001 <.001 <.001 <.001 <.001 <.001
Chicken skin mucosa	14.4 (12.8-16.0)	7.6 (5.4-9.7)	14.3 (11.4-17.2)	19.6 (16.8-22.4)	No SMI - Scar No SMI - sSMI No SMI - dSMI Scar - sSMI Scar - dSMI sSMI - dSMI	<.001 1.000 .001 .001 <.001 .014
Ulceration	9.0 (7.7-10.3)	10.2 (7.7-12.7)	15.7 (12.7-16.7)	24.7 (21.6-27.8)	No SMI - Scar No SMI - sSMI No SMI - dSMI Scar - sSMI Scar - dSMI sSMI - dSMI	0.441 <.001 <.001 <.001 <.001 <.001
Spontaneous bleeding	8.6 (7.6-10.1)	1.8 (0.7-2.8)	5.8 (3.9-7.8)	40.5 (37.0-44.0)	No SMI - Scar No SMI - sSMI No SMI - dSMI Scar - sSMI Scar - dSMI sSMI - dSMI	<.001 .026 <.001 <.001 <.001 <.001
Number of features (median)	1.0 (2.0)	1.0 (2.0)	2.0 (2.0)	3.0 (2.0)	No SMI - Scar No SMI - sSMI No SMI - dSMI Scar - sSMI Scar - dSMI sSMI - dSMI	.651 <.001 <.001 <.001 <.001 <.001

Table 1. Comparison of the presence (% [95% confidence interval]) of the 6 different Blink features and the median number of Blink features per histopathological group (No SMI, Superficial SMI, Deep SMI) and resection scars with associated p-values. Submucosal invasive cancer, SMI; Superficial SMI, sSMI; Deep SMI, dSMI.

Conclusion

The presence of ≥2BF reliably discriminates LNPCPs with cancer from non-cancer even in the presence of scarred lesions amongst non-expert endoscopists. Scarred LNPCPs exhibit fold deformation at a similar rate to LNPCPs with deep SMI but other BF are not present.

References

Debels LK, Smeets S, Poortmans PJ, Argenziano ME, Desomer L, Anderson J, Valori R, David J Tate. Identification of 6 key features of colorectal polyps increases the sensitivity of cancer detection and ability to discriminate deep submucosal invasion The basis of the Blink (first) impression? United European Gastroenterology Journal 2023; Vol.11

Introduction

Cancer in large non-pedunculated colorectal polyps (LNPCPs) is poorly discriminated outside expert centres. It is unknown whether video- is comparable to live-assessment for determining submucosal invasion (SMI) in LNPCPs. Video-assessment has multiple advantages including collaborative discussion of imaging findings at multi-disciplinary meetings avoiding extra assessment procedures for patients and incorrect decision-making. We compared live LNPCP assessment with analysis of high-quality endoscopic videos produced by an expert.

Aims & Methods

Blinded live assessments of LNPCPs from consecutive, consenting patients were conducted by 15 endoscopists in a single center. After the live assessment, an expert recorded a standardized video of each LNPCP. The remaining endoscopists [who had not assessed the LNPCP live] assessed these videos using an online survey. Endoscopists were classified as experienced (>50 EMRs ever and >1000 colonoscopies) or inexperienced. Parameters such as Blink Impression(BI) [1], 6-Blink Features(BF) – namely fold deformation, extra redness, depression, chicken skin mucosa, ulceration and spontaneous bleeding - presence of a Demarcated Area(DA), Vascular Pattern(VP) of any DA, size, location, Paris classification, and morphology were recorded to form the endoscopist's impression of cancer presence in the LNPCP. The results were later compared to expert consensus including an external validator. The primary endpoint was to evaluate the accuracy of endoscopists’ final impression of cancer and choice of correct treatment between live and video assessments.

Results

64 live assessments (12 endoscopists) & 96 video assessments (9 endoscopists) of 18 LNPCPs were performed. 5 (41.7%) (live) vs 4 (44.4%) (video) endoscopists were inexperienced. The comparison of live versus video assessments revealed no significant difference in the final impression of cancer presence by the endoscopist (sens 80.0% vs 100%, spec 87.0% vs 85.5%), BI (sens 85.7% vs 92.3%, spec 88.6% vs 80.7%) and the presence of ≥2BF (sens 71.4% vs 84.6%, spec 81.8% vs 68.7%) vs histology (P=.77, .34 & .21 respectively). Further the ability of DA & JNET to predict SMI did not differ nor did the rate of correct treatment determination (71.9% live vs 60.4% video, P=.14). Versus expert consensus the determination of ≥2BF was equally accurate live (82.4%) vs video (76.0%,P=.38). Conversely the exact number of BF vs experts (live 60.4% vs video 37.5%,P=.01), the presence of ≥3 features (location, size, morphology or Paris classification) of COVERT (not visible on the surface) SMI live (81.3%) vs video (65.6%, P=.03) and the correct outcome of a structured algorithm[2] (live 96.9%, video 81.3%,P=.002) were determined differently between the groups.

	LIVE			Video			P
	Sens	Spec	Acc	Sens	Spec	Acc
Blink pre	85.7% [59.8%-100%]	88.6 % [79.3% - 98.0%]	88.2% [79.4% - 97.1%]	92.3% [77.8% - 100%]	80.7% [72.2% - 89.2%]	82.3% [74.7% - 89.9%]	0.345
Blink (vs SMI, ≥2 BF)	71.4% [378.0%-100%]	81.8% [70.4% - 93.2%]	80.4% [69.5% - 91.3%]	84.6% [65.0% - 100%]	68.7% [58.7% - 78.7%]	70.8% [61.7% - 79.9%]	0.207
Blink (vs expert, ≥2 BF)	70.0% [41.6%-98.4%]	85.4% [74.6% - 96.2%]	82.4% [71.9% - 92.8%]	63.5% [50.4% - 76.6%]	90.9% [82.4% - 99.4%]	76.0% [67.5% - 84.6%]	0.377
OVERT DA	100.0% [100% - 100%]	75.9% [64.5% - 87.3%]	79.7% [69.8% - 89.5%]	100.0% [100% - 100%]	77.1% [68.1% - 86.2%]	80.2% [72.2%-88.2%]	0.920
OVERT VascPattern JNET	100.0% [100% - 100%]	70.4% [58.2%-82.6%]	75.0% [64.4% - 85.6%]	100.0% [100% - 100%]	69.9% [60.0% - 79.8%]	74.0% [65.2% - 82.7%]	0.887
Cancer present	80.0% [55.2%-100%]	87.0% [78.1%-96.0%]	85.9% [77.4% - 94.5%]	100.0% [100% - 100%]	85.5% [78.0% - 93.1%]	87.5 [80.9% - 94.1%]	0.777

Conclusion

Final impression of the presence of cancer and determination of correct treatment in LNPCPs are predicted equally well from blinded video assessment versus live assessment in a group of similarly experienced endoscopists. This finding could minimise the need for patients to undergo a repeat procedure to assess their LNPCP prior to resection relying instead upon expert assessment of a video.

References

• Debels, L., Smeets, S., Poortmans, P. J., Jorissen, C., Lala, V., Lamiroy, T., Valori, R., Desomer, L., Anderson, J., & Tate, D. (2023). Identification of 6 key features of colorectal polyps increases the sensitivity of cancer detection and ability to discriminate deep submucosal invasion - The basis of the BLINK (first) impression? Gastrointestinal Endoscopy, 97(6, Supplement), AB190. https://doi.org/10.1016/j.gie.2023.04.318
  
• GIEQS. Year. "SMIC." Accessed [23/11/2023]. https://www.gieqs.com/smic.

Introduction

International guidelines recommend endoscopic submucosal dissection (ESD) for en-bloc resection of rectal lesions suspected of submucosal invasive cancer (SMI). ESD achieves favorable outcomes for benign lesions and early rectal cancers (ERC) with superficial (<1000μm) SMI. However, for ERCs with deep (≥1000μm) SMI, the rate of histologically radical (R0) resection remains suboptimal, at maximum 64% for SM2-3 lesions, even in expert centres. The muscle-retracting sign (MRS) has emerged as a potential endoscopic marker of deep SMI detected during endoscopic resection. Predicting deep SMI is particularly relevant in rectal lesions, as incorrect selection of resection modality may lead to non-radical endoscopic resections and the need for surgical salvage, which carries significant morbidity. To address some of these limitations, we applied the novel pocket-detection method (PDM) for MRS detection. This approach enhances visualization of the submucosal space and facilitates early identification of the MRS.

Aims & Methods

The aim of this study was to evaluate the diagnostic accuracy of the MRS detected using the PDM as an innovative approach for identifying deeply invasive ERC during endoscopic resection of rectal lesions.
In this prospective, observational, single-centre study consecutive rectal lesions that underwent endoscopic resection using the PDM between August 2023 and March 2025 were included. Directionality of initial submucosal dissection targeted any morphological or surface vascular pattern at risk of deep SMI to attempt MRS identification. Histopathology was the reference standard for MRS diagnostic accuracy in detecting deep SMI. Deep SMI was defined as ≥1000μm invasion from the muscularis mucosa at histopathology.

Results

38 rectal lesions, median size 30 mm (IQR 20-40), were included. PDM for MRS detection was successful in all cases, with MRS-positive in 15/38 (39.5%). Deep SMI was confirmed in 15/38 (39.5%). Accuracy of MRS-positive for deep SMI detection was 84.2%, (95% CI 69.6-92.6%), sensitivity 80% (95% CI 54.8-92.9%), specificity 87% (95% CI 67.9-95.5%), positive predictive value 80% (95% CI 54.8-92.9%), negative predictive value 87% (95% CI 67.9-95.5%), positive likelihood ratio 6.13, negative likelihood ratio 0.23, area under the curve 0.83 (95% CI: 0.71–0.96). The rate of R0 resection was 92.1% (35/38).

Patients, n	38
Male sex, n (%)	21 (55.3%)	Lesion size, mm, median (IQR)	30 (20-40)
Age, y, median (IQR)	68.5 (63.2-74.7)	JNET classification, n (%)
Rectal location, n (%)		JNET III	16 (42.1%)
Distal rectum	23 (60.5%)	Demarcated area, n (%)	20 (52.6%)
Rectal wall, n (%)		Submucosal fibrosis, n (%)
Postero-lateral	26 (68.4%)	Fibrosis, F1	7 (18.4%)
Paris classification, n (%)		Fibrosis, F2	3 (7.9%)
Paris 0-Is / 0-IIa+Is	28 (73.7%)	R0, n (%)	35 (92.1%)
Paris 0-IIa+c	4 (10.5%)

Conclusion

MRS detected using the PDM demonstrated moderate-to-high accuracy in predicting deep SMI in early rectal cancer. This finding may allow selection of endoscopic resection strategy based on invasion depth.

Introduction

Endoscopic submucosal dissection (ESD) is increasingly promoted as the optimal treatment strategy for all large non-pedunculated colorectal polyps (LNPCP). The effect of a universal en bloc strategy on oncological outcomes for the treatment of LNPCP in the right colon is unknown. We evaluated this in a large Western population.

Aims & Methods

We retrospectively analysed prospectively collected data from major tertiary centres where patients were referred for endoscopic resection (ER) of LNPCP (N=5084). Patients found to have cancer after ER and those referred directly to surgery were included. The primary outcome was to determine the proportion of right colon LNPCP with low-risk cancer.

Results

Over 180 months until June 2023, 3294 sporadic right colon LNPCP in 2956 patients were referred for ER at 7 sites (median size 30mm [IQR 15]). 63 (2.1%) patients were referred directly to surgery and cancer was proven in 56 (88.9%). 2851/2956 (96.4%) LNPCP underwent ER (median size 35mm [IQR 20]) of which 75 (2.6%) were cancers. The overall prevalence of cancer in the right colon was 4.4% (131/2956). Detailed histopathological analysis was possible in 115/131 (88%) cancers (71 after ER, 44 direct to surgery, Table 1). 23/2940 (0.78%) sporadic right colon LNPCP were low-risk cancers.

Table 1: Final Oncologic Outcomes; comparison between histopathologic variables of cases referred directly to surgery versus following endoscopic resection

	Direct to Surgery (n=44)	EMR Performed (n=71)	p
Depth of SM invasion (Kikuchi) N (%)			<0.001
SM1	4 (9.1)	22 (30.9)
SM2/3	8 (18.2)	39 (54.9)
Invasion beyond SM	32 (72.7)	10 (14.1)
High tumour grade N (%)	5 (11.4)	9 (12.7)	1
LVI N (%)	8 (18.2)	12 (16.9)	1
High grade TB N (%)	5 (20.8)	10 (14.9)	0.68
Any high-risk histology (high tumour grade, LVI, high grade TB, deep SM invasion) N (%)	40 (90.9)	52 (73.2)	0.04

Conclusion

The proportion of right colon LNPCP referred for ER containing low-risk cancer amenable to endoscopic cure was <1%, in a large, multicentre Western cohort. A universal ESD strategy for the management of right colon LNPCP is unlikely to yield improved patient outcomes given the minimal impact on oncological outcomes, significant resource demands and increased adverse events.

Introduction

Hot endoscopic mucosal resection (H-EMR) is the standard of care for large non-pedunculated colorectal polyps (LNPCPs) without suspicion of submucosal invasion, achieving low recurrence rates after en bloc resection or piecemeal resection with margin ablation in expert centers.^1,2 However, electrocautery is also responsible for complications which can lead to significant morbidity and necessitate additional invasive procedures.^3,4 Cold snare polypectomy (CSP) has proven to be preferable to H-EMR for adenomas ≤10mm and non-dysplastic sessile serrated lesions, providing comparable efficacy with improved safety, shorter procedural time and lower costs.⁵ These advantages have driven interest in expanding CSP to larger polyps.

Aims & Methods

This meta-analysis of randomized controlled trials (RCTs) aimed to compare the risk of recurrence and adverse events between cold snare polypectomy (CSP) and hot-endoscopic mucosal resection (H-EMR) for large non-pedunculated colorectal polyps (LNPCPs).
A systematic search of Medline, Embase and Cochrane Library databases was performed through August 2024 for studies comparing recurrence, bleeding and perforation rates between CSP and H-EMR for LNPCPs ≥15mm (PROSPERO ID: CRD42024568272). RCTs were included in the quantitative analysis. A random-effects meta-analysis, with heterogeneity measured with I², was conducted to generate pooled risk ratios (RR) with 95% confidence intervals.

Results

Four RCTs comprising 1516 LNPCPs (766 CSP and 750 H-EMR) in 1442 patients were included in the quantitative analysis. CSP demonstrated a higher recurrence risk at first surveillance colonoscopy than H-EMR in the pooled analysis (22.6% vs. 10.8%; RR=1.98; 95% CI: 1.22–3.21; p=0.02; moderate-certainty evidence), corresponding to a number needed to harm of 9. Regarding adverse events, CSP demonstrated a 67% reduced risk of delayed bleeding (1.2% vs. 3.9%; RR=0.33; 95% CI: 0.12-0.89; p=0.03; high-certainty evidence), corresponding to a number needed to treat of 37. Although CSP appeared to reduce the risk of intraprocedural bleeding (10.0% vs. 19.8%; RR=0.30, 95% CI: 0-52256.34, p=0.42), the wide confidence interval from the random-effects model included 1. There were no intraprocedural or delayed perforations in the CSP group.

Author	Study design	Number of patients		Age, mean (SD), years	Female, %	Number of LNPCPs		LNPCP size, mean/median (SD/IQR), mm	Right Location, %		Adenomatous histology, %
		CSP	H-EMR			CSP	H-EMR		CSP	H-EMR	CSP	H-EMR
O’ Sullivan T. et al., 2024	Single center, RCT	87	90	68.2 (10.0)	46.9	87	90	30 (25-35)	96.5	86.7	100	100
Steinbrück et al., 2024	Multicenter, RCT	184	192	65.9 (10.5)	48.2	193	203	30.1 (10.2)	88.1	86.2	60.7	68.5
Pohl H. et al., 2024	Multicenter, RCT	336	324	65.3 (10.2)	47.7	371	343	30 (25.4)	82.5	83.4	NR	NR
Nogales et al., 2023	Multicenter, RCT	115	114	67.9 (10.6)	43.3	115	114	25 (22-30)	80.0	78.1	71.9	77.2

Table 1. Patients, LNPCPs and procedure characteristics of the randomized controlled trials included in the quantitative analysis. (List of abbreviations: CSP, cold snare polypectomy; H-EMR, hot endoscopic mucosal resection; IQR, interquartile range; LNPCPs, large non-pedunculated colorectal polyps; NR, not reported; SD, standard deviation; RCT, randomized controlled trial)

Conclusion

CSP has nearly double the recurrence risk of H-EMR for LNPCPs. However, its superior safety profile may make it a preferable option for patients where procedural safety is prioritized over radicality, such as those with extensive comorbidities.

References

1. Ferlitsch M, Hassan C, Bisschops R et al. Colorectal polypectomy and endoscopic mucosal resection: European Society of Gastrointestinal Endoscopy (ESGE) Guideline - Update 2024. Endoscopy 2024; 56: 516-545. doi:10.1055/a-2304-3219
2. Kaltenbach T, Anderson JC, Burke CA et al. Endoscopic Removal of Colorectal Lesions: Recommendations by the US Multi-Society Task Force on Colorectal Cancer. Am J Gastroenterol 2020; 115: 435-464. doi:10.14309/ajg.0000000000000555
3. Sidhu M, Shahidi N, Gupta S et al. Outcomes of Thermal Ablation of the Mucosal Defect Margin After Endoscopic Mucosal Resection: A Prospective, International, Multicenter Trial of 1000 Large Nonpedunculated Colorectal Polyps. Gastroenterology 2021; 161: 163-170.e163. doi:10.1053/j.gastro.2021.03.044
4. Meulen LWT, Bogie RMM, Winkens B et al. Thermal ablation of mucosal defect margins to prevent local recurrence of large colorectal polyps: a systematic review and meta-analysis. Endosc Int Open 2022; 10: E1127-e1135. doi:10.1055/a-1869-2446
5. Rex DK, Anderson JC, Pohl H et al. Cold versus hot snare resection with or without submucosal injection of 6- to 15-mm colorectal polyps: a randomized controlled trial. Gastrointest Endosc 2022; 96: 330-338. doi:10.1016/j.gie.2022.03.006