Efficacy of a three-day training course in endoscopic submucosal dissection using a live porcine model: a prospective evaluation

Abstract

Background

Endoscopic submucosal dissection is currently regarded as the method of choice for the resection of superficial tumours. The objective of our study was to evaluate prospectively the efficiency of an endoscopic submucosal dissection training course using live anaesthetised pigs.

Methods

Fourteen novice endoscopists participated in three gastric endoscopic submucosal dissection training courses on anaesthetised pigs. Each trainee resected five ‘fake’ antral lesions. Resected specimen sizes, endoscopic submucosal dissection speeds and the complication rate were evaluated prospectively.

Results

Among the 70 procedures performed, 58 could be analysed. There was a strong increase in endoscopic submucosal dissection speed (from 9.7 mm²/minute to 30.4 mm²/minute) during the sessions, marked between the first two endoscopic submucosal dissections compared with the fourth and fifth. There was a significant relation between the surface area of the resected lesion and procedure speed (P < 0.0001). The complication rate was 8.6%.

Conclusion

There is a clear benefit from endoscopic submucosal dissection training courses on animal models. Improved endoscopist capability is evident from the third endoscopic submucosal dissection. These data validate the indispensable nature of dedicated training courses and echo the European Society of Gastrointestinal Endoscopy proposition for multistep learning, beginning on animal models.

Key summary

Endoscopic submucosal dissection is the best mini-invasive method to resect early neoplastic lesions such as high-grade dysplasia or intramucosal carcinoma in the stomach. This method is widely used in the eastern world but less commonly in western countries, at least in part to master the complexity of endoscopic submucosal dissection procedures. The benefit of training programmes on ex vivo then in vivo animal models has been demonstrated to decrease complication rates efficiently in humans. However, the number of sessions required as well as the speed of acquisition of endoscopic submucosal dissection skills remain largely unknown.

What are the new findings of this study?

• There is a strong increase in average speed dissection, with a cut-off from the third session.

• There is a significant correlation between the speed of dissection and the size of the resected specimen.

• Complications seem to occur when the endoscopist starts to become more confident and eager to perform larger resections.

Introduction

Endoscopic mucosal resection (EMR) was developed more than two decades ago as an alternative to surgery for the treatment of precancerous lesions and superficial cancers. Indeed, the efficiency of EMR is similar to that of surgery, with 99% disease-specific survival at 5 and 10 years, and it has lower morbidity and mortality rates than surgery.¹ However, EMR is recommended only for small lesions (less than 2 cm) because the treatment of larger lesions is associated with a high rate of recurrence, which may require several additional endoscopic sessions or a surgical cure.² A major step forward has been the introduction of endoscopic submucosal dissection (ESD), which was developed more than 10 years ago in Japan to allow ‘en bloc’ resections, and therefore to decrease dramatically the risk of recurrence, even in large lesions. Indeed, for selected indications and in expert hands, ESD is associated with low recurrence rates and high 5-year-specific survival rates (up to 97%), but at the cost of a longer procedure time.^3,4

A major limitation to the widespread use of ESD in clinical practice is that this technique requires new skills and a totally new, dedicated training programme. Also, wide differences exist in the incidence of early gastric cancer and diagnostic and management procedures between the eastern and western worlds. In the eastern world, the higher incidence of gastric cancer, plus the fact that most lesions are detected early through both screening programmes and an ‘imaging attitude’ of most endoscopists, is of pivotal importance in recruiting patients suitable for ESD. In contrast, in the western world, ESD is more difficult to learn as few gastric lesions are eligible for endoscopic resection, and the procedure thus remains confined to high-volume tertiary referral centres, where nonetheless similar results as in Asia are obtained.⁵ For instance, a recent population-based study showed that early gastric cancer is rare in France, and only 5% of cases were resected endoscopically.⁶ This explains the 30% complication rate found in a multicentre retrospective European study when ESD was first performed in human patients without preclinical training of endoscopists.⁷ Recently, the European Society of Gastrointestinal Endoscopy (ESGE) provided guidelines focusing on ESD, and proposing that at least five ESD procedures should be performed in an animal model (in vivo or ex vivo) before it is practised in humans.^8,9 Thus we decided to evaluate prospectively the ESD learning curve, in terms of efficiency, speed and safety of the procedure, using a waterjet knife in a porcine model, with procedure-naive endoscopists.

Materials and methods

Fourteen endoscopists, all practising interventional endoscopy but with no prior experience in ESD, were selected to participate in ESD training sessions (n = 3 per participant, one day per session). Each session was supervised by one or two senior endoscopists, all regularly practising ESD in humans (NM, MLR, EMC, DL, EC). The first session was introduced by a theoretical course on ESD (EC). Each trainee had to perform a total of five ESDs with en bloc resection of specimens progressively increasing in size.

ESD procedures were performed on Gall KO live pigs (mean weight 30 kg) in the dedicated surgical facility of a research laboratory of the University Hospital of Nantes, France (Laboratoire des Grands Animaux, Inserm U643). All procedures were performed under general anaesthesia, with induction by 5% isoflurane and 60% nitrous oxide and subsequent maintenance on 2% isoflurane (Abbvie, Rungis, France). General anaesthesia was induced and maintained with 2–5% isoflurane (Abbvie, Rungis, France). Procedures were approved by the local institutional animal research committee (agreement E. 44010; Inserm, Nantes, France).

All procedures were performed with standard endoscopes (EG 450, Fujifilm, Japan) with a distal attachment cap. Five gastric ESDs were performed consecutively by each endoscopist, using a specific ESD-designed knife equipped with a waterjet function at low pressure (FlushKnife; Fujifilm, Japan). Additional devices such as haemostatic forceps (Coagrasper; Olympus America Inc., USA) or clips (EZClip; Olympus America Inc., USA) were considered in the case of significant bleeding. Perforation was defined as a frank deep muscular injury (hole) seen during ESD. Minute perforation was defined as a serosal injury seen at the histological examination.

Procedure

The procedure time was measured from the first marking point using the tip of the knife to the complete removal of the target ‘gastric lesio’. Each endoscopist began the procedure by lifting the margins, followed by a circumferential incision of the ‘lesio’. After submucosal lifting, submucosal dissection was performed (using saline solution with indigo carmine). Additional injections were performed if necessary to ensure a sufficient lifting and exposition for safe dissection. Submucosal vessels were preventatively coagulated during dissection, or in the case of bleeding. When dissection was completed, the time of the entire ESD procedure was recorded. Resected specimens were removed from the stomach, pinned on to a cork board, measured and carefully examined. Finally, animals were killed and stomachs were removed for macroscopic examination to search for perforation.

Outcome parameters

En bloc resection rates and complications were recorded prospectively. The surface area of each specimen was calculated at the end of the procedure by reference to the two largest diameters using the formula for an ellipse, as used in other studies:¹⁰ area (mm²) = small diameter (mm)/2 × large diameter (mm)/2 × π when the specimen was elliptic in shape, and π × (diameter (mm)/2)² when the specimen was a circle. Procedure speed was calculated in mm² per minute using the formula speed = surface (mm²)/time (minutes).

Statistical analyses were performed with SAS version 9.4. Because of important outliers, speed results are presented in median (with upper and lower quartile Q1–Q3), and were compared using non-parametric Kruskall–Wallis and Wilcoxon’s tests.

Results

Of the 70 consecutive procedures performed, 58 ESDs performed in the fundus or the antro-fundic junction were included in the analysis (Table 1). It is known that not all parts of the pig stomach are suitable for ESD learning,¹¹ which is the reason why a variety of ESD procedures (n = 5) performed in other locations (oesophagus, distal antrum) were excluded from the analysis. In seven cases, data were missing.

Table 1.

Data during the ESD training course.

Endoscopist		1	2	3	4	5	6	7	8	9	10	11	12	13	14	Mean	Median
ESD 1
	Location	F	F	F	AFJ	F	F	F	AFJ	F	F	F	F	F	MD
	Time (min)	45	45	70	64	39	30	65	66	46	120	60	60	36	MD	57
	Surface (mm2)	79	79	177	314	314	471	628	314	1374	1178	785	942	687	MD	565
	Speed (mm2/min) Q1; Q3	1.8	1.8	2.5	4.9	8.1	15.7	9.7	4.8	29.9	9.8	13.1	15.7	19.1	MD	10.5	9.7 4.8; 15.7
	Complication	–	–	–	–	–	–	–	–	–	–	–	–	–	MD
ESD 2
	Location	F	F	F	F	AFJ	F	F	F	F	F	F	AFJ	MD	MD
	Time (min)	40	35	25	34	65	43	42	28	58	75	46	65	MD	MD	46
	Surface (mm2)	314	314	314	198	942	628	236	440	550	884	628	412	MD	MD	488
	Speed (mm2/min) Q1; Q3	7.9	9	12.6	5.8	14.5	14.6	5.6	15.7	9.5	11.8	13.7	6.3	MD	MD	10.6	10.6 7.1; 14.1
	Complication	–	–	–	–	–	–	–	–	–	–	–	–	MD	MD
ESD 3
	Location	AFJ	AFJ	F	AFJ	F	MD	F	F	F	F	F	OL	MD	F
	Time (min)	90	16	36	77	60	MD	24	23	68	90	150	OL	MD	41	61
	Surface (mm2)	2827	1964	236	942	707	MD	192	550	935	1374	4398	OL	MD	573	1336
	Speed (mm2/min) Q1; Q3	31.4	122.7	6.5	12.2	11.8	MD	8	23.9	13.7	15.3	29.3	OL	MD	14	26.3	14.0 11.8; 29.3
	Complication	–	–	–	B	–	MD	–	–	–	–	–	OL	MD	–
ESD 4
	Location	F	F	OL	OL	AFJ	F	F	F	F	F	OL	F	F	F
	Time (min)	60	55	OL	OL	66	56	45	26	176	120	OL	125	65	75	78
	Surface (mm2)	5027	5027	OL	OL	942	1100	424	825	6185	2827	OL	4241	1979	393	2634
	Speed (mm2/min) Q1; Q3	83.8	91.4	OL	OL	14.3	19.6	9.4	31.7	37	23.6	OL	33.9	30.5	5.2	34.6	30.4 14.3; 37.0
	Complication	–	MP	OL	OL	–	–	–	–	–	–	OL	–	P	–
ESD 5
	Location	F	F	OL	MD	F	OL	AFJ	F	OL	F	MD	AFJ	AFJ	F
	Time (min)	60	48	OL	MD	23	OL	43	39	OL	120	MD	46	120	62	64
	Surface (mm2)	5027	628	OL	MD	236	OL	877	1791	OL	3534	MD	1590	2081	1002	1652
	Speed (mm2/min) Q1; Q3	83.8	13.1	OL	MD	10.2	OL	20.4	45.9	OL	29.4	MD	34.6	17.3	16.2	39.3	20.4 16.2; 34.6
	Complication	–	–	OL	MD	B	OL	–	–	OL	–	MD	–	P	–

ESD: endoscopic submucosal dissection; B: bleeding; P: perforation; MP: minute perforation; F: fundus; AFJ: antrofundic junction; MD: missing data; OL: other locations (antrum, oesophagus): excluded data.

Efficiency

En bloc resection was achieved in 100% of procedures (i.e. all marking points were found in the resected specimen).

Overall, there was a strong difference in the median speed of dissection during the workshop ranging from 9.7 mm²/minute (Q1–Q3: 4.8; 15.7) to 30.4 mm²/minute (Q1–Q3: 7.1; 15.7). As shown in Figure 1, there was an increase in the median speed from the third ESD. We did a post hoc sample size calculation: the mean difference between speeds 1 and 5 was 10.5 ± 34.6 (mm²/minute). The number of procedures needed to highlight this difference with 80% power and 5% type I error rate was 14 (one-sample t-test for mean) while we could make statistics based on eight procedures on a point-by-point comparison (due to missing data and wrong locations).

Figure 1.

Median speed of dissection (mm²/minute) per session.

The surface areas of the resected specimens throughout the workshop varied significantly (P = 0.0001). The mean surface area of the resected specimen by ESD 4 was significantly greater than at all other procedures except ESD 5. The mean surface area of ESD 4 was significantly greater than ESD 1 (P < 0.0001), ESD 2 (P < 0.0001) and ESD 3 (P = 0.01), and the mean surface area of ESD 5 was significantly greater than ESD 1 and ESD 2 (P < 0.0001). There was a significant relation between the procedure speed and the surface area of the resected specimen (P < 0.0001), as shown in Figure 2.

Figure 2.

Correlation between speed of dissection and surface resected.

Complications

Overall, the complication rate was 8.6%, including two episodes of bleeding (in ESDs 3 and 5), both successfully treated with the Coagrasper, one minute perforation (ESD 4), and two macroscopic perforations successfully treated with endoscopic clips (ESDs 4 and 5).

Discussion

The aim of preclinical ESD learning is to achieve en bloc resections in more than 80% of procedures, at a cost of less than 10% complications.¹² A panel of eastern and western world endoscopic experts recently proposed a four-step ESD training strategy in the western world. In particular, they advised an experimental in vivo training course under expert supervision to be started early as a second phase after ex vivo training.¹³

Our study suggests that three supervised sessions in a live porcine model are a good way to train in ESD. It strongly supports the use of such in vivo training sessions as we showed a three-fold increase in terms of procedure speed between the first and the fourth procedure. The adjusted mean dissection speed found in our study (22.1 mm²/minute) was very close to that reported by Jacques et al. (23.3 mm²/minute) in a similar porcine model.¹⁴ We believe that, with experience, the junior endoscopist not only progresses in organising different strategies to resect ‘lesions’ of different sizes and locations, but also very practically in the use of ESD-associated devices. For instance, it is striking to note how trainees quickly learn how to use the cap to stabilise the tip of the endoscope, the waterjet knife to clean the working space and re-inject the submucosa with the correct amount of fluid, and to manipulate haemostatic forceps or clips when necessary. The most interesting finding of our study is the clear progression with a strong cut-off between the two first ESDs and the following procedures. This cut-off is in line with the ESGE recommendations on ESD to perform at least five procedures on animal models.⁸ However, as shown above with the post-hoc sample size calculation, we were not able to demonstrate statistically the differences between the median speeds, because of an insufficient number of procedures.

Surprisingly, no complication occurred within the two first dissections but only from the third. This might be due to a lack of attention during the procedure or, more likely, to an excess of confidence of the endoscopist. The complication rate reported in our study was much lower than that found in other pig models.¹⁴ This might reflect the fact that ‘junior’ endoscopists were carefully selected for ESD training and had mostly already performed endoscopic retrograde cholangiopancreatography or other interventional techniques. It might also be due, at least in part, to the fact that these training sessions were performed in very small groups, with one senior endoscopist per two trainees, thus allowing optimal supervision during ESD procedures. This close supervision is in concordance with previous recommendations by eastern world and Japanese experts.^15,16 This result concerning the ‘delayed’ complications appearing after several procedures is important. It shows that in the learning process of a new procedure not only efficiency but especially safety is of pivotal importance. Therefore, it is an important aim of such preclinical models to allow trainees to experience their own limits and make progress. In particular, these preclinical training courses are useful to learn gestures and positions that should be avoided (or adopted) to prevent complications. As early mistakes and complications are unavoidable, one major aim is to understand how to avoid them in future ESD procedures, and especially in patients.

The model used in this study has several limitations. First, anaesthetised porcine models are quite expansive and not widely accessible for regulatory reasons as compared to ex vivo models. Those ex vivo models are widely used, and are also very useful in the multistep learning process of ESD but are quite artificial. Some authors have recently proposed using a human gastric remnant from sleeve gastrectomy that allows them to surround some inconvenience of the porcine model (such as the thickness of the porcine gastric mucosa, as discussed below) and provide an accessible and cost-effective model.¹⁴ Indeed, ESD procedures performed in the stomach of healthy pigs do not reflect the complexity of ESD procedures in human diseases. This is probably why the complication rate (in particular, bleeding rates) observed in our study was still much lower than those observed in the first human cases.¹⁵ Also, it is known that the porcine stomach is less vascularised and thicker than in humans. Moreover, porcine coagulation and aggregation are significantly higher than human coagulation,¹⁶ thus lowering the risk of significant bleeding. Therefore, some authors have developed and validated novel porcine models using antiplatelet therapies and curative anticoagulation, in order to increase the bleeding risk and to deal with a ‘close to real life’ situation.¹⁷ This might be of particular interest in future sessions dedicated to ESD, or more generally to gain endoscopic skills in bleeding situations. Another limitation is the lack of fibrosis or submucosal invasion of healthy animal models because trainees had to remove ‘fake’ lesions (e.g. normal gastric mucosa) and not superficial cancers. Thus our 100% en bloc resection rate is quite uninformative about the capacity to replicate such high-quality resection in clinical practice. Our model would have to be improved. As already described, creating simulated ulcerated target lesions, closer to those found in real life, can be a good way to add difficulty and to realise a new step in preclinical training.¹⁸ After all, it is clear that this type of session is not sufficient in terms of efficiency but above all safety. A number of further procedures in preclinical models (ex vivo and in vivo) then in patients with close supervision is necessary to provide an optimal learning programme.

In conclusion, training with a preclinical model such as the live porcine model seems mandatory for optimally learning to use each endotherapy device required for ESD and to perform en bloc resection in conditions close to real life. Our study strongly supports the ESGE statements by showing a clear progression over the training sessions and a relatively short learning curve. This type of preclinical training course is a necessary but not sufficient step in ESD learning in order to lower the complication rates before practising in humans and ensure optimal en bloc resection rates.

Author contribution

NC and CV performed data analysis, wrote the first draft of the manuscript and participated in manuscript revision. NM, EMC, DL and MLR supervised training sessions and contributed to manuscript revision. EC, the principal investigator, worked on all phases of the study including study design, supervised training sessions, data analysis and interpretation, and manuscript preparation.

Declaration of conflicting interests

Emmanuel Coron has served as a speaker for Fujifilm, Olympus, Cook, Mayoly Spindler and Abbvie, and is a consultant for Medtronic and Cook. Nicolas Chapelle has served as a speaker for Abbvie. Dominique Luet has served as a speaker for Merck Serono and Sanofi. Elodie Metivier-Cesbron has served as a speaker for Cook and Abbvie. Nicolas Musquer, Christelle Volteau and Marc LeRhun have no conflicting interest.

Funding

This study was funded in part by Fujifilm (animal costs only).

Ethics approval

This animal study has been approved by the French ministry of research and the ethical commmity on animal research (CEEA N°6 Pays de la Loire).

Informed consent

Not applicable.