Endoscopic submucosal dissection (ESD) is an excellent technique for curative local en bloc resection of early digestive neoplastic lesions. However, ESD is technically demanding with a high adverse events’ rate and long procedure time especially among inexperienced trainees [1]. It has been reported that an estimate of 30 gastric cases are needed to learn the basic skills of ESD [2]. Recently, traction systems to facilitate ESD have been reported, but the technique was not widely used. Among these traction systems, the clip with line method is feasible and cost-effective in the hands of the experts [3–9], but no comparison with the conventional method was reported. We conducted a prospective study to compare the feasibility and efficacy of this traction method with the traditional ESD method among trainees using live pigs. The primary objective was to determine the benefits of the dental floss with endoclip traction method in facilitating trainees learning to perform gastric ESD.

Patients and methods

This is a prospective comparison using a live pig stomach model and was approved by ethics committee of the Beijing Military General Hospital. One trainee endoscopist without prior experience of ESD for gastric neoplasms performed all of the procedures. The trainee had previous experience with more than 1500 therapeutic endoscopies including 1000 polypectomies and 500 other types of endoscopic mucosal resection (EMR) for GI lesions.

Endoscopic procedures

We used a single-channel colonoscope, CF-Q260 (Olympus, Japan), with a transparent cap attached to the scope tip for ESD. A total of 20 lesions were created in the anesthetized porcine stomachs at four anatomical sites, each matched for location and size. The target lesions, ranging from 2 × 2 to 4 × 4 cm, were created on (1) greater curvature (n = 6), (2) lesser curvature (near the gastric angle) (n = 6), (3) anterior wall (n = 4), and (4) posterior wall (n = 4) of the corpus–antrum transition region of the stomach, by initially marking with the tip of dual knife (Olympus KD-650L) and a coagulation current (ESG100, Olympus Japan) with a setting of forced coagulation II, at 40 W. Subsequently, 0.9 % normal saline was injected into the submucosal layer using an injection needle (MTW Endoskopie, Wesel, Germany) to create a submucosal cushion. The Hook knife (Olympus) was used for a deeply circumferential mucosal incision. Submucosal dissection of the first five lesions was performed using traction provided by the dental floss with endoclip (Olympus, Japan) method. The subsequent five lesions (matched for similar lesion location and size) were performed using the conventional ESD method without traction. Ten more procedures including five traction assists and five conventional ESD were performed. Submucosal injections were repeated to maintain sufficient tissue elevation when necessary. The frequency of submucosal injections was documented. In conventional ESD, submucosal dissection was carried out with the dual knife after circumferential incision. In the clip group, the endoscope was removed after circumferential mucosal incision, and an endoclip was inserted into the accessory channel and then pushed out (Fig. 1A). A long dental floss was tied firmly over one arm of the clip as reported previously (Fig. 1B) [8, 9]. The scope was subsequently inserted into the stomach, and the clip was applied to grasp the proximal end of the exposed mucosa. The submucosal layer could be clearly visualized when traction was applied to the dental floss (Figs. 1C, 2A). Submucosal dissection was then carried out with the cap under the elevated mucosa. The traction was maintained until the submucosal dissection was completed (Fig. 2B) (video).

Fig. 1
figure 1

Dental floss with clip is easily equipped. A Endoclip is pushed out from the accessory channel. B Dental floss is tied firmly over one arm of the clip. C Traction can be easily applied

Full size image
Fig. 2
figure 2

Feasibility and efficacy of the dental floss with endoscopic clip method. A Clearly visualization of the submucosal layer. B Effective traction before the submucosal dissection complete

Full size image

Data collection

All data including type of knives, lesion locations, lesion size, size of the resected specimen, procedure time, frequency of submucosal injection, adverse events (bleeding, injury of muscular layer, or perforation), en bloc resection rate, circumferential mucosal incision time (time taken from marking to complete circumferential mucosal incision), and submucosal dissection time (time needed to complete the submucosal dissection) were recorded individually by an assistant. All procedures were video recorded. The resected specimen was retrieved and measured. The surface area of the resected specimen was calculated using the formula: (Area = π × radius2). The ESD speed was calculated by dividing the area of the resected specimen and the total dissection time (speed = area/time).

Statistical analysis

For statistical analysis, the description with mean values (standard deviation) for quantitative variables and frequency for qualitative variables are presented. The paired-sample t test was used (with statistical significance threshold of α = 0.05 or 0.1) after assumption of normal distribution. Categorical variables were compared by using the goodness-of-fit test (Chi-square test). The reusable two-factor analysis of variance was also used. The statistical significance level was set at P < 0.05.  Data were analyzed using statistical software (SPSS version 17.0; SPSS Inc., Chicago, IL).

Sample size calculation

According to a previous study [10], to detect a significant difference between experimental and control group in ESD time reduction, 10 cases were anticipated with a power of 93 % and α = 0.05.

Results

There was no significant difference in lesion location, lesion size, the size of resected specimens, and time taken for circumferential mucosal incision between the two groups. All lesions in the clip group were removed successfully using one clip to complete the submucosal dissection. There were no clip dislodgement or tear out before completion of the dissection. Furthermore, placement of the clip and dental floss was easily done within 120 s (video).

Comparison in the two groups

Compared with the controls, the procedure time was significantly shorter in clip group (5.6 ± 1.1 vs. 13.6 ± 2.5 min, respectively; paired t test, P = 0.003), and the speed of submucosal dissection was significantly faster (1.8 ± 0.3 vs. 0.6 ± 0.1 cm2/min, respectively; P = 0.001). The mean submucosal injection frequency was significantly less in the clip group (5.7 ± 0.6 vs. 8.5 ± 1.0 times, respectively; P = 0.024). There were no significant differences in the size of resected specimens and the time taken for circumferential mucosal incision (3.0 ± 0.3 vs. 2.3 ± 0.3 min, respectively; P > 0.05) between the two groups (Table 1). En bloc resection rate was 100 % in the clip group and 90 % in the conventional ESD group. There was no perforation in the two groups, but adverse events including bleeding and injury to the muscular layer were significantly less in the clip group (1 vs. 7, respectively; P = 0.068, α = 0.1, Chi-square test).

Table 1 Overall comparison—clip and control group (paired t test)
Full size table

Comparison according to anatomical location

We analyzed the mean circumferential dissecting time, submucosal injection frequency, and speed of submucosal resection of lesions in different anatomical locations (greater and lesser curvature, anterior and posterior wall) and with or without the use of clip/dental floss traction. The reusable two-factor analysis of variance indicated that different anatomical lesions (P = 0.854), or use of traction (P = 0.133), anatomical lesions and use of traction (P = 0.472) had no significant impact on time taken for circumferential mucosal incision. Different anatomical lesions (P = 0.895), and anatomical lesions and use of traction (P = 0.694) again had no significant effect on frequency of submucosal injection. With regard to anatomical location of the lesions, only traction significantly influenced the speed of submucosal dissection on lesions located on the anterior and posterior wall, (P = 0.011). For lesions located on the greater and lesser curvatures, different anatomical lesions (P = 0.024), use of traction (P = 0.004), different anatomical lesions and use of traction (P = 0.017), all had significant effect on speed of submucosal dissection (Table 2). For lesions located on the lesser curvature, the mean speed of submucosal dissection (0.6 ± 0.1 cm2/min) was shorter compared with other locations.

Table 2 Comparison according to anatomic location-reusable two-factor analysis
Full size table

Comparison in different sessions

In order to assess progress in learning to perform ESD, we divided the 20 lesions into four operating sessions with five cases in each session according to different methods used for dissection. The mean procedure time for traction-assisted submucosal dissection in the first session was 5 ± 1.1 min (range 2–8 min), compared with 6.2 ± 1.9 min (range 2–13 min) in the second session. For the control using conventional ESD, the mean procedure time in the first session was 15.0 ± 1.9 min (range 9.2–19.5 min) and 12.8 ± 4.8 min (range 5–30 min) for the second session (Fig. 3A). We did not observe an improvement in procedure time between the first and second sessions for the clip and traction and also the control group.

Fig. 3
figure 3

Median procedure time and speed of submucosal dissection based on the lesion location

Full size image

The speed for ESD taken in the four sessions was calculated and further compared. For the study group, the mean ± SD speed was 2.2 ± 0.3 cm2/min (range 1.6–3.4) in the first session and 1.3 ± 0.4 cm2/min (range 0.54–2.45) in the second session. For the control group, the speed was 0.5 ± 0.03 cm2/min (range 0.41–0.57) in the first session and 0.7 ± 0.2 cm2/min (range 0.27–0.98) in the second session (Figs. 3B, 4). Since the distance from the incisors to the stomach of live pigs was longer than that of human, it can be more difficult to dissect a lesion located on the lesser curvature. With the more difficult lesser curve lesions were excluded, the speed of submucosal dissection was significantly faster in first session with the traction-assisted method (P < 0.05) (Table 2). To assess any improvement with repeat practice, we also compared the ESD speed between the first and second sessions in the control group. An increase in dissection speed was observed (0.7 ± 0.2 vs. 0.5 ± 0.03 cm2/min) despite including the more difficult lesser curvature lesions in the second session group (3 vs. 0).

Fig. 4
figure 4

ESD speed in four sessions; the speed of submucosal cutting was significantly faster in first session of clip group (with no lesser curvature lesions included) than other sessions (P < 0.05); an up-regulation of ESD speed can be seen when compared with the ESD speed in first and second sessions in control group

Full size image

The en bloc resection rate for both study and control groups was 100 and 90 %, respectively. Adverse events including bleeding and injury of the muscularis propria were significantly less with the study group than the controls (1 vs. 7, respectively; P = 0.068, α = 0.1). Notably, perforation was avoided in all 20 lesions.

Discussion

ESD is a technically challenging and time-consuming procedure associated with a high adverse events’ rate. Previous reports of gastric ESD indicated en bloc resection rates ranging from 83–98 %, with perforation rates of 2.2–8.7 % [11–13]. In order to achieve faster and safer ESD, different traction systems have been developed to facilitate the dissection [3–10, 14–17]. Among these, the dental floss and clip method is feasible and cost-effective [3–9], because it does not require special equipment and can be easily applied to provide traction during dissection. The clip grasps and holds the proximal end of the exposed mucosa and with traction acts like a surgical hook to expose the submucosal tissue during dissection. The cap at the tip of the endoscope can be easily placed under the lifted mucosa to dissect the submucosal layer. This method provides a good endoscopic view to facilitate ESD.

The results of our study indicated that dental floss and clip traction-assisted ESD were safer and faster even in the hands of a trainee as shown by a significant decrease in procedure time for submucosal dissection and requirements for repeated submucosal injections as well as an increase in the dissection speed in the study group. This improvement is likely explained by adequate exposure of the submucosal tissue during dissection.

Overall dissection time was shorter at different locations for the study group. In general, lesions located on the lesser curvature were more difficult to dissect. As shown in Fig. 4, the ESD speed for the first session in the study group was faster than the second session. The difference is likely a result of different gastric locations, because all lesser curvature lesions were included in the second session. In the control group, despite all lesser curvature lesions were included in the second session, an increase in the dissection speed was observed compared with the first, suggested an improvement with practice of the basic skills.

Early reports on ESD training indicated a perforation rate of 20 % using the conventional method on the porcine stomach model [2], and a recommendation was made for practicing on at least 30 lesions in order to acquire the basic skills. Based on results of our study, we recommend including the dental floss and clip traction method as part of training for novice endoscopists learning ESD. With the trainee using this method to begin training, the en bloc resection rate was 95 %, and the perforation rate was 0 %.

One of the limitations of this study is the fact that few cases were performed in each arm. Some endoscopists may think the superiority of the clip method would be challenged if we included more cases in each arm. To avoid this potential bias, we compared submucosal injection and adverse events (bleeding and injury to the muscular layer) as secondary outcome between two groups. However, there are several differences between the stomach of pig and human that may contribute to technical differences. Firstly, bleeding and fibrosis in the pig are seldom and subtle. In contrast, bleeding and fibrosis are relatively frequent finding during ESD in humans, which makes the procedure more challenging. Secondly, the pig’s mucosal (the reason we used Hook knife to finish circumferential mucosal incision instead of dual knife) and muscular layers (perforation may be seldom happened during pig ESD) are thicker than human.

Some endoscopists may think the dental floss and clip traction method is more useful when the clip grasp the anal border of the lesion and conducts the SM dissection in retroflex view. When the oral side of the lesion was grasped and SM dissection was carried out in frontal view, the benefits of this method in such an ESD strategy may not so evident. But with our experience, it is useful to conduct the SM dissection in retroflex view with a lesion in the body and cardia in the stomach. In our study, all the lesions are in corpus–antrum transition region, which will be difficult to conduct the SM dissection in retroflex view. As we all know, that the most difficult parts of doing ESD are submucosal dissection after a complete circumferential mucosal incision and the last part of the lesion which could not be effectively elevated with additional submucosal injection. When we apply the clip in the proximal end, the most difficult parts of doing ESD could be resolved.

Finally, we suggest that ESD training should be conducted in four phases: (1) learning the basic knowledge and review literature on ESD; (2) acquire more basic experience in therapeutic endoscopy including polypectomy and EMR; (3) practice using in vivo animal model with the dental floss and clip traction method; and the (4) training in an in vivo animal model using the conventional method. In addition, both phase 3 and 4 practices should be supervised by an expert endoscopist.

In conclusion, the dental floss and clip traction method can simplify and shorten the procedure time in removing gastric neoplasms located in different locations. In addition, this method can improve the efficacy and safety of the procedure performed by trainees learning ESD.