Introduction
Endoscopic submucosal dissection (ESD) is a safe and effective technique to achieve R0 resection of superficial colorectal neoplasia, regardless of its size, and is associated with few recurrences [1 ]
[2 ]
[3 ]
[4 ]. Endoscopic mucosal resection (EMR) is still the usual technique to remove superficial colonic tumors in Western countries, despite a lower rate of effectiveness and a higher rate of recurrence (6 %–31 % of all patients) compared with ESD [5 ]
[6 ]
[7 ]
[8 ]. Furthermore, because of the lack of lifting due to submucosal fibrosis, residual or locally recurrent colonic lesions may be difficult to treat.
To avoid surgery, different techniques have been described to treat residual colorectal neoplasia (RCN), including endoscopic full-thickness resection (EFTR) with the full-thickness resection device (FTRD) system (Ovesco, Tübingen Germany) and ESD [9 ]
[10 ]
[11 ]. The FTRD system allows the resection of fibrous lesions of 5–30 mm, depending on the softness of the tissue. The quality of resection depends on the size of the lesion to be resected, with an R0 resection rate of 81 % for those < 20 mm vs. 58 % for those ≥ 20 mm, and rates of local recurrence of 15 %, adverse events of nearly 10 %, and surgery for complications of 2 % [11 ]
[12 ]
[13 ]. ESD is another option but few reports exist for this specific indication [14 ]. In addition, the accessibility of ESD is still a limitation of this technique and no direct comparisons are available in the literature.
To avoid an additional colonoscopy in the follow-up of these patients, R0 resection for the treatment of local recurrence or incomplete resection is mandatory. We therefore conducted a multicenter historical cohort study to compare the feasibility, efficacy, and safety of ESD and FTRD procedures for the treatment of these lesions, pairing lesions depending on their size and location.
Methods
Study population
We conducted a multicenter historical cohort study in 19 French hospital (15 academic hospitals, three general hospitals, one private hospital). Data were extracted from the prospectively maintained database for ESD cases (registered NCT 04592003) and from the database of FTRD procedures carried out in France for the post-market evaluation of the device. Consecutive patients with RCN from colorectal lesions referred for ESD from September 2019 to December 2021 and for FTRD procedures from January 2016 to December 2020 were included. All patients gave their approval for the use of their electronic medical records.
Objectives
The primary objective was the rate of curative (R0) resection for the procedure, confirmed histologically, meaning the demonstration of a resected specimen with healthy margins (the absence of tumor cells both at the basal and lateral resection margins).
Secondary end points were predictive criteria for the efficacy (R0 and curative resections) and safety (bleeding and perforation rates) of the resection.
Data collection
Clinical data and information on the previous type of resection were collected. The type of resection (ESD or FTRD), and the location, size, and classification according to Paris and Kudo for each lesion were extracted when available.
Definitions
RCN was defined as: (i) a residual lesion after a first incomplete endoscopic resection; or (ii) recurrent neoplasia on a site previously treated by ≥ 1 EMRs and considered to be complete. All consecutive cases of RCN referred for ESD or an FTRD procedure were included. The choice between the two techniques was left at the discretion of the endoscopist, depending on the local expertise in the two techniques. Lesions featuring deep invasive characteristics were excluded.
Procedure
Procedures were performed with the patient under general anesthesia (as is the case for all such procedures in France) and CO2 insufflation. Antibiotic prophylaxis was used routinely.
EFTR procedure
The EFTR procedure was performed using a conventional colonoscope. A first insertion was performed to analyze the lesion, to mark the margins, and possibly to leave in place a guidewire, when required, to facilitate subsequent access to the lesion with the FTRD system attached to the tip of the scope. The second insertion was then performed to access the lesion site and a dedicated grasper was used to pull the lesion into the dedicated cap. Once the lesion was drawn sufficiently into the cap, the clip was released, and the snare was closed to cut the lesion with cutting current ([Fig. 1 ]). Where this process initially failed, a second snare could be used to achieve the resection of the lesion trapped in the clip.
Fig. 1 Endoscopic image of the full-thickness resection technique being used for residual rectal neoplasia.
Endoscopic submucosal dissection with traction
ESD was performed using a therapeutic colonoscope (PCF 190 L, T, or TL; Olympus, Tokyo, Japan; or EC760 P/VM or R/VM; Fujifilm, Tokyo, Japan) and a distal transparent hood. The strategy for ESD and the model of knife used were left to the discretion of the endoscopist, as was the use of a double-clip traction system [14 ] ([Fig. 2 ]).
Fig. 2 Endoscopic images during endoscopic submucosal dissection showing: a a recurrent lesion; b the traction strategy being applied.
Outcomes
Resections were defined as en bloc if the lesion was resected in a single piece, and R0 if the en bloc specimen had tumor-free lateral and deep margins. R0 resection was classified as curative in all benign lesions (low grade dysplasia [LGD] and high grade dysplasia [HGD]) or if histopathological assessment showed adenocarcinoma with low risk criteria (depth of submucosal invasion ≤ 1000 μm; absence of lymphatic and vascular invasion; well or moderately differentiated).
The procedure duration (in minutes) was defined for ESD as the time between the first injection and specimen excision; for FTRD procedures, it was defined as the time between the detection of the lesion and specimen excision, and included the time for removal of the colonoscope to insert the FTRD system onto the colonoscope.
Specimen size was defined by the largest diameter, measured after stretching the resected specimen on cork.
Perforation was defined as the observation of a complete defect through the colonic muscle during treatment or clinical evidence of perforation (Sydney score > 3 [15 ]).
Endoscopic follow-up
Delayed bleeding was defined as clinical evidence of bleeding during the month following the resection leading to transfusion, new hospitalization, or a second-look endoscopy for hemostasis.
Where there was a non-R0 resection, patients had a second colonoscopy ≤ 6 months after the endoscopic procedure; biopsy or new resection of the recurrent tissue was then performed.
Statistical analysis
Quantitative variables were expressed as median (interquartile range [IQR]), and qualitative variables as count and percentages. Comparisons were performed with the Wilcoxon rank sum test for quantitative variables, and χ2 test or Fisher exact test for qualitative variables. Univariate analyses of factors associated with en bloc and R0 resection were performed by logistic regression, with associations quantified by the odds ratio (OR) and its associated 95 %CI. The endoscopic procedure, and factors with P values < 0.10 on univariate analyses were included in a multivariable analysis. A sensitivity analysis was performed using an inverse propensity weighting approach, including in the propensity score the initial histology, the size of the lesion, the age and sex of the patient, the location of the lesion, and the number of patients in the center (≤ 10 vs. > 10). Tests were two-sided, with an α level of 0.05.
Analyses were performed using R software version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria; http://www.r-projet.org ).
Results
Patient characteristics
A total of 275 patients (median age 70 years; 160 men) underwent 177 ESD and 98 FTRD procedures for RCN ([Fig. 3 ]). Lesions were mostly located in the colon and were statistically larger in the ESD group ([Table 1 ]). The numbers of lesions in each size category for the ESD and FTRD groups respectively were: 13 (7.5 %) and 71 (72.4 %) for lesions < 20 mm; 31 (17.9 %) and 21 (21.4 %) for lesions between 20 and < 30 mm; 48 (27.6 %) and 6 (6.1 %) for lesions between 30–40 mm; and 82 (47.1 %) and 0 for lesions ≥ 40 mm (missing data for three patients in the ESD group).
Fig. 3 Flowchart showing the patients who underwent either endoscopic submucosal dissection (ESD) or full-thickness resection device (FTRD) procedures and the outcomes for those treated for residual neoplasia.
Table 1
Comparison of the patient and lesion characteristics for the endoscopic submucosal dissection (ESD) and full-thickness resection device (FTRD) groups.
ESD (n = 177)
FTRD (n = 98)
P value
Age, median (IQR), years
69.5 (64–76)
71.0 (65–76)
0.74
Sex, female, n (%)
76 (42.9)
39 (39.8)
0.71
Previous procedure resulting in residual or recurrent lesion, n (%)
0.29
67/50 (91.4)
82 (88.2)
6 (4.7)
6 (6.5)
0 (0)
2 (2.2)
5 (3.9)
3 (2.2)
49
5
Size, median (range), mm
35 (5–140)
(3–30)
< 0.001
Lesion location, n (%)
0.08
54 (30.5)
22 (22.4)
35 (19.8)
26 (26.5)
12 (6.8)
15 (15.3)
45 (25.4)
22 (22.4)
31 (17.5)
13 (13.3)
Initial pathology report, n (%)[1 ]
0.07
85 (91.4)
79 (81.4)
8 (8 .6)
18 (18.6)
En bloc resection, n (%)
166 (94.3)
91 (92.9)
0.63
R0 resection, n (%)
145 (83.3)[2 ]
76 (77.6)
0.25
Procedure duration, median (IQR), minutes
45.0 (30.0–70.0)[3 ]
35.0 (21.3–54.8)[4 ]
0.20
7.5 (6.3–8.8)
NA
40 (12.5–55)
46 (37.3–55.5)
30 (20–42.5)
25 (20–45)
48 (34–80)
49 (42–56)
IQR, interquartile range; EMR, endoscopic mucosal resection; pEMR, piecemeal EMR; LGD, low grade dysplasia; HGD, high grade dysplasia; SSL, sessile serrated lesion; NET, neuroendocrine tumor; NA, not available.
1 Data available for 97 patients in the FTRD group and 93 patients in the ESD group.
2 Data available for 174 patients.
3 Data available for 166 patients.
4 Data available for 14 patients.
Outcomes
There were three (1.7 %) and six (6.1 %) failures of the planned procedure in the ESD and FTRD groups, respectively (P = 0.07). R0 resection was achieved in 83.3 % and 77.6 % for ESD and FTRD groups, respectively (P = 0.25). A traction strategy was used in 74 % of the ESD cases. The median (IQR) duration of the procedure was 45.0 (30.0–70.0) minutes for ESD vs. 35.0 (21.3–54.8) minutes for FTRD procedures (P = 0.20), but this median was affected by lesions ≥ 40 mm, with a median duration of 55 minutes for ESD, but no attempted procedures with the FTRD for lesions of this size.
Factor associated with en bloc and R0 resections
En bloc resection was achieved in 94.3 % of cases for ESD and 92.9 % for FTRD procedures (P = 0.63) ([Table 1 ]). R0 resection was achieved in 83.3 % and 77.6 % of the ESD and FTRD groups, respectively (P = 0.25). The lesions treated by ESD were however statistically larger than those treated by FTRD (P < 0.001), and small lesion size was the only factor associated with en bloc and R0 resection on univariate analysis ([Table 2 ]; Table 1 s, see online-only Supplementary material). On multivariable analysis adjusted for lesion size, there were statistically higher en bloc (OR = 5.51 [95 %CI 1.36–26.08]; P = 0.02) and R0 (OR = 4.73 [95 %CI 1.97–12.13]; P < 0.001) resection rates for ESD compared with FTRD procedures. Sensitivity analyses using the inverse probability weighting approach led to similar results for both en bloc resection (OR = 6.96 [95 %CI 1.69–28.59]; P = 0.007) and R0 resection (OR = 3.61 [95 %CI 1.40–9.30]; P = 0.008).
Table 2
Factors associated with R0 resection on univariate and multivariable analyses.
Variable
Level
OR (95 %CI)
P value
Univariate analysis
Endoscopic procedure
FTRD
1
0.25
ESD
1.45 (0.78–2.69)
Location of the lesion
Rectum
1
0.56
Sigmoid colon/left colon/left angle
1.11 (0.48–2.56)
Transverse colon
0.77 (0.28–2.15)
Right colon/right angle
1.36 (0.58–3.18)
Ileocecal valve/cecum
2.06 (0.70–6.09)
Lesion size
per 10-mm increase
0.83 (0.72–0.96)
0.01
Initial pathology
LGD, HGD, SSL
1
0.32
Adenocarcinoma, NET
0.61 (0.23–1.58)
Multivariable analysis
Endoscopic procedure
FTRD
1
< 0.001
ESD
4.73 (1.97–12.13)
Lesion size
per 10-mm increase
0.67 (0.53–0.82)
< 0.001
FTRD, full-thickness resection device; ESD, endoscopic submucosal dissection; LGD, low grade dysplasia; HGD, high grade dysplasia; SSL, sessile serrated lesion; NET, neuroendocrine tumor.
The R0 rates for lesions < 20 mm were 100 % and 87.3 % in the ESD and FTRD groups, respectively (P = 0.34). The R0 rate for lesions of 20–30 mm and 30–40 mm were 83.9 % and 57.1 % (P = 0.06) and 93.6 % and 33.3 % (P = 0.002) in the ESD and FTRD groups, respectively (Fig. 1 s ). The larger the lesion was, the more effective ESD seemed to be compared with the FTRD, but this trend could not be statistically confirmed owing to the moderate sample size of the study. The location of the lesion (rectum or colon) was not associated with either en bloc or R0 resection.
Follow-up
Of the 267 available histological assessments (97.1 %), 218 lesions were adenomas (Table 2 s ). There were 35 adenocarcinomas: intramucosal adenocarcinoma (n = 15); submucosal invasion < 1000 µm (SM < 1000; n = 8); submucosal invasion > 1000 µm (SM > 1000 µm; n = 8); and T2 (n = 4). In addition, there were three neuroendocrine tumors.
Although having been diagnosed as benign lesions on the initial histological diagnosis from the original resection, 16 lesions were eventually diagnosed as adenocarcinoma on the recurrence: intramucosal adenocarcinoma (n = 9); SM < 1000 µm (n = 2); SM > 1000 µm (n = 4); T2 (n = 1).
Finally, surgery for noncurative resection was required for four patients in the ESD group (2.3 %) and five patients in the FTRD group (5.1 %; P = 0.29) (Table 2 s ). Of the nine patients with a failure of the planned procedure (ESD or FTRD), seven had follow-up surgery, and two underwent a hybrid endoscopic resection technique.
Of patients with R1 resection without need of subsequent surgery, 10/22 (45.5 %) and 10/16 (62.5 %) in the ESD and FTRD groups, respectively, had a follow-up colonoscopy. One of the patients who underwent follow-up colonoscopy (2.1 %) in the ESD group experienced a recurrence (lesion of 35 mm with low grade dysplasia on a lateral margin after an initial R1 resection) and there were five recurrences (11.9 %) in the FTRD group (P = 0.09), including three on lesions with positive lateral margins (adenocarcinoma SM < 1000 µm [n = 2], and LGD [n = 1]) and two on lesions that had been considered to be R0 resections. All of the recurrent lesions in the FTRD group were initially lesions with a size of < 20 mm.
Safety
Complications related to the endoscopic procedure were more frequent in the ESD group (16.3 %) compared with the FTRD group (5.1 %; P = 0.01). However, according to the AGREE classification [16 ] (Tables 3 s and 4 s ), no statistically significant difference existed for severe adverse events (grades II–V) between the two groups.
Intraoperative perforation occurred in 20 patients (11.2 %) in the ESD group and in one patient (1.0 %) in the FTRD group because of the flow of current. The majority of intraoperative perforations in the ESD group were in lesions of ≤ 20 mm (Fig. 2 s ). Intraoperative perforations happened mostly in the colon. They were all managed endoscopically. In addition, one delayed perforation was noted in the ESD group and was managed surgically. One patient in the ESD group was operated on 15 days after the ESD for an enterocutaneous fistula. Delayed bleeding occurred in eight of the ESD group and two of the FTRD group; none of them needed surgical treatment. Finally, a rectal stricture was detected in one patient in the FTRD group, which was treated conservatively.
Discussion
Residual and recurrent colorectal lesions should always be referred for an attempt at endoscopic resection and not to surgery because the rate of patients cured without surgery is very high with ESD and the FTRD (> 90 %) [9 ]
[10 ]
[12 ]
[14 ].
Globally, our study demonstrated a higher R0 resection rate of 83.3 % in the ESD group compared with 77.6 % in the FTRD group, without a statistically significant difference (P = 0.25). Nevertheless, when looking at the results in detail, the two techniques are not equivalent because the FTRD was almost always reserved for small lesions of < 20 mm and its results decreased dramatically when used for larger lesions. For lesions of 20–30 mm, the difference was already large, with rates of 83.9 % and 57.1 % for R0 resections in the ESD and FTRD groups, respectively (P = 0.06), and this difference dramatically increased when the lesion size was > 30 mm (93.6 % vs. 33.3 %; P = 0.002). In those patients who had already undergone several colonoscopies (initial resection, follow-up detecting the recurrence, and new therapeutic procedure for management of RCN), achieving an R0 resection seems to be particularly important to avoid the need for another early follow-up procedure.
Although seducing because of its relative technical ease, use of the FTRD is in practice reserved for very selected lesions, because 89.8 % of the lesions were < 30 mm in the FTRD group. Furthermore, lesions < 20 mm represented < 10 % of the total number of residual or locally recurrent lesions referred to the centers that resected all the lesions referred by ESD without considering the size (Table 5 s ). Unlike the German register, in France the volume of ESDs is higher than FTRD procedures. In the German register, 1138 FTRD procedures were performed in 4 years and 458 ESDs in 3 years with only 78 colonic ESDs, whereas in our French multicentric prospective cohort 2087 colorectal ESDs were performed in 2 years ([Fig. 3 ]) [2 ]
[13 ]. This difference can be explained by the fact that the volume of FTRD procedures is directly inversely correlated with the volume of colorectal dissections. The systematic use of a traction strategy for ESD and intensive training have led to it being a comfortable technique, with a preferential choice for ESD instead of FTRD procedures, especially when lesions exceed 20 mm in size.
People will argue that ESD is more time-consuming and risky. We are aware that the small sample size available for the duration of the procedure limits conclusion; however, the time for FTRD procedures in our study correlates with the results of Meier et al. [13 ], who reported the median procedure time, measured from first introduction of the endoscope until extraction after reinspection of the resection site, was 35 minutes. In our study, the duration difference between the two techniques was not significant, with only 10 minutes separating the two types of procedure. The high speed of the ESD procedures can be explained by the use of traction-assisted ESD in 74 % of cases and by the degree of expertise [14 ]
[17 ].
The risk of perforation, which was significantly higher in the ESD group, did not lead to surgery for adverse events, irrespective of the location of the ESD, showing its feasibility in the colon. The risk of intraoperative perforation was higher in the ESD group, especially for lesions < 20 mm. This result could be explained by the heterogeneity of the centers practicing ESD with a different level of ESD expertise. In low volume ESD centers, ESD is more frequent in small lesions, but at the cost of more adverse events owing to the lower experience levels. Even if a colonic location is considered by some authors more difficult, in our study, we have shown the feasibility and effectiveness of colonic ESD in a Western country, without the occurrence of major adverse events.
Finally, the risk of residual or recurrent lesions after FTRD procedures was not trivial, at 11.9 % in our study, which is concordant with the literature (6.4 %–13.5 %) [9 ]
[10 ]
[12 ], and higher than in the ESD group, at 0 % for R0 resections and 2.1 % globally. Where recurrence occurs after ESD, a new ESD or an EFTR procedure are technically feasible, whereas the management is more challenging for recurrences into the clip at the FTRD site, requiring the use of the REMOVE (Ovesco, Tübingen, Germany) dedicated device [18 ] to ablate the macro clip and treat again the recurrence in the very fibrotic area.
We are aware that our study is a historical cohort, and that the heterogeneity of the population limits the interpretation of our results; however, we tried to use the most unbiased statistical analysis by focusing mainly on efficacy endpoints: R0 and en bloc resection. Our results showed a statistical difference between ESD and FTRD procedures in terms of both R0 and en bloc resections, even when using a Bonferroni approach to correct for multiple testing. As it is not a prospective randomized study, each center used is usual technique, with different levels of experience leading to a very high imbalance between ESD and FTRD procedures across the centers. To minimize the effect, we included the size of the center in a sensitivity analysis.
This study is a reflection of the current practice in France, showing that all centers have not yet converted to ESD, even if the number of gastroenterologists performing ESD is growing. The impact of experience level was not studied precisely, but it certainly impacted our results. Indeed, the majority of intraoperative perforations occurred on lesions < 20 mm, reflecting probably that the easiest lesions were resected by less experienced endoscopists. Furthermore, as this is a retrospective study, patient follow-up was short and data were missing, notably for the risk of recurrence after endoscopic resection.
We believe however that the efficacy of ESD for polyp recurrences of any size and its low complication rate place it as the treatment of choice in this situation, especially for lesions of > 20 mm. With regard to our results, the use of FTRD procedures for the treatment of recurrence should be reserved for lesions of < 20 mm. Another option for these lesions could be underwater EMR, which does not require any device or particular skill for its performance [19 ], with possibly a lower economic and environmental impact. The ESD technique is spreading all over the word and should not be a limitation. ESD learning and training will be the next challenge to correctly treat all our patients. The results observed in our study should be confirmed by a prospective study including a medico-economic evaluation of the two techniques, without forgetting their accessibility and their ecological impact, which seem to be similar for the two techniques in our evaluation on a simulated procedure (data not shown).
In conclusion, ESD seems to be the best option, when available, to achieve higher R0 resection rates, whatever the size and location of the lesion. On the other hand, FTRD procedures could be a very good option when ESD is not available or in nonexpert practice, but only for lesions of < 20 mm. For lesions above this threshold, the risk of recurrence is high and referring the patient for ESD appears to be a more effective option ([Fig. 4 ]).
Fig. 4 Suggested management algorithm for residual colorectal neoplasia. ESD, endoscopic submucosal dissection; FTRD, full-thickness resection device; UEMR, underwater endoscopic mucosal resection.