Article Text

Download PDFPDF

Original article
Wide-field endoscopic mucosal resection versus endoscopic submucosal dissection for laterally spreading colorectal lesions: a cost-effectiveness analysis
  1. Farzan F Bahin1,2,
  2. Steven J Heitman1,3,
  3. Khalid N Rasouli1,
  4. Hema Mahajan4,
  5. Duncan McLeod4,
  6. Eric Y T Lee1,
  7. Stephen J Williams3,
  8. Michael J Bourke1,2
  1. 1 Department of Gastroenterology and Hepatology, Westmead Hospital, Sydney, New South Wales, Australia
  2. 2 Westmead Clinical School, University of Sydney, Sydney, New South Wales, Australia
  3. 3 Department of Medicine, Cumming School of Medicine, University of Calgary, Calgary, Alberta, Canada
  4. 4 Department of Anatomical Pathology, Westmead Hospital, Sydney, New South Wales, Australia
  1. Correspondence to Professor Michael J Bourke, Department of Gastroenterology and Hepatology, Westmead Hospital, Sydney, New South Wales, 2145, Australia; michael{at}


Objective To compare the cost-effectiveness of endoscopic submucosal dissection (ESD) and wide-field endoscopic mucosal resection (WF-EMR) for removing large sessile and laterally spreading colorectal lesions (LSLs) >20 mm.

Design An incremental cost-effectiveness analysis using a decision tree model was performed over an 18-month time horizon. The following strategies were compared: WF-EMR, universal ESD (U-ESD) and selective ESD (S-ESD) for lesions highly suspicious for containing submucosal invasive cancer (SMIC), with WF-EMR used for the remainder. Data from a large Western cohort and the literature were used to inform the model. Effectiveness was defined as the number of surgeries avoided per 1000 cases. Incremental costs per surgery avoided are presented. Sensitivity and scenario analyses were performed.

Results 1723 lesions among 1765 patients were analysed. The prevalence of SMIC and low-risk-SMIC was 8.2% and 3.1%, respectively. Endoscopic lesion assessment for SMIC had a sensitivity and specificity of 34.9% and 98.4%, respectively. S-ESD was the least expensive strategy and was also more effective than WF-EMR by preventing 19 additional surgeries per 1000 cases. 43 ESD procedures would be required in an S-ESD strategy. U-ESD would prevent another 13 surgeries compared with S-ESD, at an incremental cost per surgery avoided of US$210 112. U-ESD was only cost-effective among higher risk rectal lesions.

Conclusion S-ESD is the preferred treatment strategy. However, only 43 ESDs are required per 1000 LSLs. U-ESD cannot be justified beyond high-risk rectal lesions. WF-EMR remains an effective and safe treatment option for most LSLs.

Trial registration number NCT02000141.

  • endoscopic polypectomy
  • colorectal adenomas
  • therapeutic endoscopy
  • colorectal cancer
  • cost-effectiveness

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Significance of this study

What is already known on this subject?

  • Wide-field endoscopic mucosal resection (WF-EMR) of large sessile and laterally spreading. lesions (LSLs) is safe, effective and durable; however, recurrence remains a limitation.

  • Endoscopic submucosal dissection (ESD) is advocated as a superior treatment for LSLs due to decreased disease recurrence and potential cure of low-risk submucosal invasive cancer (LR-SMIC).

  • Among patients with LSLs the benefit of potential cure of LR-SMIC by a universal or selective ESD strategy is unknown.

What are the new findings?

  • When applied to a large, prospective cohort of LSLs 20 mm, selective ESD prevents 19 additional surgeries per 1000 cases at slightly lower cost compared with WF-EMR.

  • Universal ESD could prevent an additional 13 surgeries per 1000 cases compared with selective ESD, but at substantial increased cost of over US$210 000 per surgery avoided.

  • Expanding selective ESD criteria according to contemporary Japanese guidelines provides little additional benefit.

  • Given the effectiveness of WF-EMR for benign lesions and the infrequency of LR-SMIC lesions, a universal ESD strategy in the colorectum is unjustified. 

How might it impact on clinical practice in the foreseeable future?

  • A universal ESD strategy cannot be justified for colorectal LSLs.

  • A selective ESD strategy may complement rather than replace WF-EMR, but widespread implementation is impractical.

  • Lesion assessment requires further advancement to identify those most likely to benefit from ESD.

  • Colorectal ESD practice and training should be adopted carefully in selected referral centres.


Large sessile and laterally spreading colorectal lesions (LSLs) are increasingly encountered during colonoscopy. The prevalence of flat lesions larger than 10 mm in screening populations is approximately 1%.1–3 LSLs have an increased risk of harbouring invasive cancer and can be challenging to excise endoscopically.4–7 The risk of submucosal invasive cancer (SMIC) can be stratified by lesion morphology and surface pattern, but the accuracy of endoscopic prediction of SMIC is low.4 6 8–10

In the West most LSLs are treated by wide-field endoscopic mucosal resection (WF-EMR) as it is documented to be a safe, effective and durable therapy.4 11 In addition it is less morbid and less costly than surgery with more than 97% of patients being discharged from hospital on the same day.12–15 In the East, endoscopic submucosal dissection (ESD) is the dominant technique due to its ability to achieve en bloc resection in over 80% of cases.16–19 This offers the advantage of potential cure of low-risk submucosal invasive cancer (LR-SMIC). However, ESD is technically challenging, more time-consuming and resource-consuming, and has greater frequency of serious adverse events compared with WF-EMR.20 For example, ESD had a rate of perforation of 5.7% compared with 1.4% in a recent meta-analysis comparing the two approaches.20 While colorectal ESD may generate clinical benefits, its applicability, relevance and health resource implications have not been evaluated in a Western context.

The purpose of this study was to assess the costs, expected health benefits and harms, and the cost-effectiveness of ESD-based strategies compared with WF-EMR for the treatment of colorectal LSLs>20 mm.



An incremental cost-effectiveness analysis using a decision analytical model was performed to determine the most technically efficient endoscopic strategy for the treatment of colorectal LSLs 20 mm among adult men and women. Lesions where invasive cancer was grossly evident were excluded. The analysis was conducted and the results reported according to contemporary guidelines.21 The following strategies were compared: (1) WF-EMR, (2) universal ESD (U-ESD) and (3) selective ESD (S-ESD). Data from the lead site of the Australian Colonic Endoscopic Resection (ACE) study were used to inform the model together with best available literature.4 Sensitivity and scenario analyses were performed, including criteria for ESD reflecting Japanese Gastroenterological Endoscopy Society Guidelines.22

The primary goal of endoscopic resection is effective and safe treatment without need for surgery. Therefore, the measure of effectiveness was the number of surgeries avoided per 1000 cases within each treatment strategy. We also determined the number of surgeries for SMIC (low and high risk), persistent recurrences, benign lesions incorrectly suspected of containing SMIC and for perforations. In the base case, costs were those relevant to the publicly funded healthcare system.

The time horizon of the analysis was set to 18 months because outcomes are well established after the index procedure and following the first and second surveillance colonoscopies (SC1 and SC2 at 4–6 months and 18 months, respectively).11 Furthermore, while recurrences can be effectively treated endoscopically at SC1, data beyond SC2 are limited and persistent recurrences may ultimately undergo surgery. Costs and health outcomes were not discounted.

Decision analytical model

The decision tree model was created using TreeAge Pro Suite 2016 software (TreeAge Software, Williamstown MA, USA). The model was reviewed by tissue resection experts with experience in both WF-EMR and ESD and deemed to have good face validity. Deliberate testing of the model using extreme input values to ensure they produced the expected outputs was performed to eliminate all syntactical errors. A basic overview of the model structure is depicted in figure 1.

Figure 1

Overview of decision tree analysis. ESD, endoscopic submucosal dissection; SMIC, submucosal invasive cancer; WF-EMR, wide-field endoscopic mucosal resection.

Within the WF-EMR and S-ESD strategies endoscopic lesion assessment guided initial therapy. Assessment encompassed the Paris classification of superficial neoplastic lesions,8 pit pattern based on the Kudo classification,23 granular versus non-granular surface topography and submucosal lifting. Lesions were classified as highly suspicious for containing SMIC when any of the following high-risk characteristics4 24 were present: Paris 0-IIc morphology (22.2% of SMIC lesions), pit pattern V (38.2% of SMIC lesions) and non-lifting (15.3% of SMIC lesions).

  1. WF-EMR strategy: Lesions classified as highly suspicious for SMIC were referred for surgical management. Suspicious rectal lesions underwent transanal excision with the aim of preventing invasive surgery for those containing LR-SMIC. The remainder were treated via WF-EMR. SC1 and SC2 were performed at 6 and 18 months, respectively. Benign recurrent lesions were treated endoscopically at SC1, but were referred for surgery when persistent at SC2. Thus, surgery was performed for all colonic lesions containing any SMIC, high-risk submucosal invasive cancer (HR-SMIC) rectal lesions, recurrent LR-SMIC rectal lesions, lesions incorrectly classified as highly suspicious for SMIC, persistent recurrences of benign lesions and for perforations not amenable to endoscopic management.

  2. U-ESD strategy: All lesions were treated initially by ESD regardless of SMIC suspicion. Lesions containing HR-SMIC were referred for surgery, whereas those containing LR-SMIC were initially classified as cured. Surgery was performed for all HR-SMIC lesions, early LR-SMIC recurrences, persistent benign recurrences at SC2 and for perforations not amenable to endoscopic management.

  3. S-ESD strategy: Lesions with high suspicion for SMIC were treated by ESD; all others were treated by WF-EMR. Lesions discovered to have HR-SMIC following ESD were referred for surgery. Lesions containing only LR-SMIC were initially considered cured when treated by ESD. Thus, surgery was performed for all lesions containing HR-SMIC or LR-SMIC when treated by WF-EMR, LR-SMIC recurrences treated by ESD, persistent recurrences of benign lesions at SC2 and for perforations not amenable to endoscopic management.

Study setting

Data were collected as part of the prospective, observational multicentre ACE study of patients referred for endoscopic resection of LSLs>20 mm ( NCT02000141). Written informed consent was obtained from each patient on the day of the procedure.

All LSLs were assessed by a senior endoscopist with extensive WF-EMR experience. Each endoscopist performs over 100 EMRs per year and the accumulated WF-EMR experience in the department exceeds 2000 cases. LSLs were assessed using high-definition white light and narrow band imaging using 180 or 190 (from mid-2012 onwards) series colonoscopes (Olympus, Tokyo, Japan). An electrosurgical generator with microprocessor-controlled current (VIO 300D Endo Cut Q Effect 3; ERBE Elektromedizin, Tuebingen, Germany) was used for tissue resection.

Model assumptions

WF-EMR was performed according to the standardised ACE protocol and ESD was performed as previously reported.4 25 26 For ESD, a 100% en bloc and R0 resection rate was assumed in the base case.

We assumed recurrences among lesions containing LR-SMIC manifest at SC1. Although adverse events were modelled, mortality was assumed to be zero. We assumed 100% adherence to endoscopic and surgical treatment.

For lesions in the sigmoid, descending to distal transverse colon and mid-transverse colon to caecum, laparoscopic anterior resection, laparoscopic left hemicolectomy and laparoscopic right hemicolectomy were assumed, respectively. Lesions less than 5 cm from the anal verge predicted to contain SMIC were treated by transanal excision in the WF-EMR strategy. A two-stage laparoscopic anterior resection was assumed for HR-SMIC or LR-SMIC recurrence. This comprised initial stoma formation and anastomosis subsequently. Lesions greater than 5 cm from the anal verge predicted to contain SMIC were treated by transanal minimally invasive surgery in the WF-EMR strategy. A one-stage laparoscopic anterior resection was assumed for HR-SMIC or LR-SMIC recurrence.


All cases of actual or suspected SMIC were initially reviewed by GI histopathologists. For the purpose of this study, all cases of SMIC were re-evaluated by two specialist GI histopathologists (HM, DM). This included reassessment of the endoscopic and/or surgical resection specimen. LR-SMIC was defined by the presence of the following: tumour depth <1000 μm below the muscularis mucosa and the absence of both lymphovascular invasion and tumour budding.22 27 To maximise agreement in the re-evaluation of histological specimens, only lesions from the lead ACE centre were included in this study.

Base case data inputs

The prevalence of LSLs with and without SMIC in the colon and rectum, the test performance characteristics of endoscopic lesion assessment and the effectiveness and risks associated with WF-EMR were obtained from the lead ACE Centre at Westmead Hospital in Sydney, Australia, between September 2008 and June 2017. The cohort included 1765 lesions among 1723 patients. Data for SC1 and SC2 were available in 1625 patients (94.3%) and 1396 patients (81.0%), respectively. SMIC was found in 8.2% of all LSLs, 63.1% of which contained HR-SMIC. The sensitivity and specificity of high suspicion SMIC following endoscopic lesions assessment were determined using the final histopathology as the reference standard. Sensitivity was true positives (SMIC assessed as SMIC) divided by the sum of true positives and false negatives (SMIC assessed as benign). Specificity was true negatives (benign assessed as benign) divided by the sum of true negatives and false positives (benign assessed as SMIC). Data on the effectiveness and safety of ESD and transanal excision including the risks of lesion recurrence with and without LR-SMIC and the proportion requiring surgery were obtained from systematic reviews or best available evidence. All parameter estimates along with the ranges considered are found in table 1.

Table 1

Reference case probabilities and ranges considered


All costs were determined locally and are reported in inflation-adjusted 2015 US dollars based on a conversion rate of $A1=US$0.75 (table 2). The non-physician costs of WF-EMR, ESD, transanal excision, surveillance colonoscopies and surgery were derived from the Australian Refined Diagnosis Related Group codes (V.7.0, 2013). Non-physician costs are the direct costs incurred by the hospital for treating a patient and are distinct from costs paid by the patient or insurer. We assumed that patients treated with WF-EMR were discharged same day, whereas those treated with ESD or transanal excision were observed in hospital for two nights. Physician costs were obtained from the Australian Medical Benefits Schedule. For WF-EMR we assumed physician-directed sedation, but for ESD and transanal excision we included the costs of monitored anaesthetic care.

Table 2

Cost estimates and ranges considered

Patient and caregiver time and travel costs were considered, but reported separately, whereas costs due to lost productivity were not incorporated.28 Time and travel costs were estimated from a study where these costs were determined for colonoscopy.29 We assumed patients could access care within driving distance.


Allowance for uncertainty in all base case estimates including costs was assessed through univariate sensitivity analyses. Scenario analyses were also performed. This included S-ESD in accordance with the 2015 Japanese Gastroenterological Endoscopy Society Guidelines for colorectal ESD,22 where lesions with pit pattern V, Paris c component or any lesion containing non-granular morphology were treated by ESD. We also evaluated the impact of much higher risks of lesion recurrence and need for surgery with WF-EMR and ESD. Lastly, given the potential greater impact of ESD in the rectum, rectal and colonic lesions were considered in a separate subgroup analysis.


Base case analysis

The least costly strategy was S-ESD followed by WF-EMR and U-ESD at US$4.22 million, US$4.33 million and US$6.91 million per 1000 cases, respectively (table 3). S-ESD was also slightly more effective than WF-EMR with 19 additional surgeries prevented per 1000 cases, 7 resulting from curative treatment of LR-SMIC and 12 due to imperfect lesion assessment specificity, whereby ESD was performed instead of surgery for benign lesions (table 4). U-ESD was more effective than S-ESD with 13 additional surgeries prevented per 1000 cases. However, this additional health benefit was realised at an incremental cost per surgery avoided of $210 112 (table 3).

Table 3

Base case costs, effectiveness and incremental cost-effectiveness ratios per 1000 cases treated

Table 4

ESD cases performed, surgeries and costs by strategy

Sensitivity and scenario analyses

Applying contemporary Japanese criteria within the S-ESD strategy improved lesion assessment sensitivity from 34.9% to 66%. This resulted in 26 fewer surgeries compared with WF-EMR (table 5), 16 among LSLs with LR-SMIC. However, with an associated decrease in lesion assessment specificity from 98.4% to 60.8%, 414 ESDs would be performed per 1000 cases (table 4). S-ESD using Japanese criteria was no longer the least expensive strategy costing $34 881 per surgery avoided compared with WF-EMR. U-ESD remained more effective than S-ESD with an additional six surgeries prevented at an incremental cost per surgery avoided of $274 271 (table 5).

Table 5

Sensitivity and scenario analyses

In a scenario analysis where the sensitivity and specificity of lesion assessment were changed to 83% and 70%,30 respectively, S-ESD became the cheapest strategy. One additional surgery could be prevented through U-ESD, but at a cost of $3 601 967 (table 5).

Increasing the risk of recurrence following WF-EMR to 25% in the colon and 35% in the rectum while increasing the risk of surgery to treat recurrences in the colon and rectum from 1% to 10% lowered the effectiveness of both WF-EMR and S-ESD. As a result, the cost per surgery avoided of U-ESD dropped to $66 768. In contrast, increasing the risk of recurrence following ESD to 10% in the colon and 15% in the rectum while increasing the risk of surgery to treat recurrences to 10% increased the incremental cost per surgery avoided of U-ESD to $822 607 (table 5). Lowering the R0 resection rate of ESD from 100% to 71%31 increased the incremental cost per surgery avoided of U-ESD to $379 204. Increasing the risk of perforation following ESD requiring surgery to 20% resulted in U-ESD preventing fewer surgeries at greater cost than both S-ESD and WF-EMR (table 5).

When considering only rectal LSLs, WF-EMR was as effective as S-ESD given transanal resection for lesions suspected of containing SMIC. However, S-ESD was still less costly. Given the higher prevalence of SMIC in the rectum, the incremental cost per surgery avoided of U-ESD dropped to $87 066 and dropped further to $32 132 among non-granular rectal lesions. U-ESD became the least costly and most effective strategy among higher risk non-granular Paris 0-Is rectal LSLs (table 5).

S-ESD remained the least costly strategy across plausible changes in the cost of ESD. The results were not affected by elimination of early surveillance following treatment with ESD or when considering non-medical costs (table 5).


This study demonstrates that S-ESD prevents more surgeries and is less costly than WF-EMR for treatment of colorectal LSLs>20 mm. WF-EMR provides definitive therapy without need for surgery in nearly 91% of case. S-ESD would prevent 19 additional surgeries per 1000 cases. U-ESD would prevent an additional 13 surgeries compared with S-ESD, but at an incremental cost per surgery avoided of $210 112.

We chose avoidance of surgery as our primary outcome since this is the goal of all endoscopic resection strategies. Irrespective of the potential to avoid surgery among LR-SMIC lesions, the findings of our study question the overall impact of ESD with a Western context. In our large tertiary referral cohort of 1765 LSLs >20 mm among 1723 patients, the prevalence of SMIC was only 8.2%. It is likely to be lower in less enriched cohorts. Furthermore, of the 144 lesions containing SMIC, only 54 (3.1% of the total cohort) were LR-SMIC. In large observational studies of ESD in Japan and Korea, the proportion of patients with LR-SMIC ranged from 4% to 10%.18 32–34 However, in European centres reporting on more carefully selected colorectal lesions treated by ESD, only 1%–5% of patients would have derived a clinically meaningful benefit through cure of LR-SMIC.35–38 In these studies the remainder of the patients with invasive disease required consideration of surgery. Thus the benefit of ESD as a surgery sparing curative treatment of LR-SMIC is applicable to no more than 10% and 5% of referral cohorts in East Asia and the West, respectively.

While S-ESD is an attractive concept, endoscopic prediction of SMIC is imperfect. When applying contemporary criteria for selecting highly suspicious lesions in our base case analysis, most LSLs containing SMIC were not triaged to ESD in the S-ESD strategy. Of the 19 additional surgeries avoided per 1000 cases, only seven were the result of treatment of LR-SMIC via ESD instead of surgery within the WF-EMR strategy. The findings of our study are supported by other recent publications comparing ESD with WF-EMR as a surgery sparing intervention. In a meta-analysis of eight studies comparing ESD and EMR, the rates of subsequent surgery were 9.9% and 5.8%, respectively.20 In a systematic review the rates were 7.9% for ESD and 8.4% with other types of endoscopic resection (predominantly WF-EMR).39 Lastly, in a prospective multicentre study additional surgery occurred in 7.5% after ESD and in 4.0% after EMR.18 Although these studies were heterogeneous and have methodological limitations, the best available evidence to date has not demonstrated superior outcomes for ESD.

Currently, tools for endoscopic prediction of SMIC have limited accuracy. Furthermore, the interobserver agreement is variable across studies. Several endoscopic features are used to derive an overall impression of the risk of SMIC. These include a Paris IIc component, Kudo pit pattern V or Sano pit pattern IIIA, non-lifting and non-granular surface morphology. In one study expert prediction of SMIC based on colour still images had a sensitivity of 83%, specificity of 70% and a positive and negative predictive value of 67% and 85%, respectively.30 Interobserver agreement for an overall SMIC prediction had a kappa value of 0.70.30 In a Japanese tertiary centre study on endoscopic resection of 271 lesions by experienced proceduralists, the non-lifting sign had a sensitivity, specificity, negative and positive predictive value of 61.5%, 98.4%, 80.0% and 96.0%, respectively.40 In two studies SMIC was present in 61% and 32% in lesions with a depressed component compared with 3% and 8%–11% in sessile non-depressed lesions, respectively.4 8 In our study the sensitivity of endoscopic prediction for SMIC assessment was only 34.9%. It could be increased to 66% by widening the criteria defining high suspicion of SMIC, but at the expense of lower lesion assessment specificity.

While our analysis supports S-ESD as the most cost-effective strategy, its introduction in the West poses considerable challenges. ESD is a technically demanding procedure requiring a high level of skill. Initial training should begin in animal models followed by clinical exposure in the more forgiving human stomach.27 Based on Japanese data, at least 40 gastric ESD procedures are required to achieve desirable outcomes while minimising complications, whereas 80 cases appear necessary in the colon.41 In Western countries where the incidence of early gastric cancer is low, opportunities for adequate training are thus limited.42 43 Compared with the colon, neoplastic lesions in the stomach confer greater benefit from removal by ESD.44–46 Maintenance of competency would be equally problematic. Our findings suggest that only 43 ESD procedures would be performed per 1000 cases given the low prevalence of SMIC and the limited ability to identify it. Thus, adoption of a S-ESD strategy broadly would be impractical within a Western context. Instead, it is more likely that a small proportion of existing EMR referral centres would be capable of developing and maintaining sufficient expertise to perform colorectal ESD.

Our results were robust to sensitivity analyses. Criteria for performing colorectal ESD are broader in most Eastern centres.22 This may not be justified given SMIC has been reported in only 15% of higher risk non-granular lesions and in just 3% of granular LSLs.4 Using contemporary Japanese S-ESD criteria, fewer surgical resections for lesions containing LR-SMIC would be necessary. However, of the 414 total ESD procedures performed per 1000 cases, 360 (87%) would be performed for benign lesions. While our model considers the higher cost of ESD compared with WF-EMR, it does not account for the opportunity cost of the longer procedure duration. A meta-analysis comparing WF-EMR with ESD showed the mean difference between the two modalities was 69 min.47 Several WF-EMR procedures could be performed in the time required to complete one ESD, an important consideration for busy tertiary referral centres operating within the confines of relatively static budgets. Instead of expanding ESD criteria to enhance the inclusion of SMIC, further innovation to facilitate the identification of high-risk lesions more suitable for ESD is preferred. When the test performance of endoscopic lesion assessment is substantially enhanced, S-ESD becomes less costly and more effective than U-ESD through curative treatment for the majority of LR-SMIC lesions while minimising the number unnecessary ESDs performed on lesions suitable for WF-EMR. In the future one might envision identifying nearly all lesions containing curable LR-SMIC, yet even then the overall clinical impact of ESD in the colorectum will be minimal.48

In addition to preventing surgery for lesions containing LR-SMIC, ESD may also provide more durable treatment of lesions without advanced histology.49 Although we modelled a higher risk of recurrence following WF-EMR, most EMRs were managed endoscopically.11 Yet, in a scenario where we nearly doubled the risk of recurrence following WF-EMR and increased the risk of surgery for recurrences tenfold, U-ESD remained costly at almost $67 000 per surgery avoided. However, endoscopic management of recurrences beyond 18 months is feasible, and it is conceivable that future innovations could lower recurrence rates following WF-EMR further.

Our model assumed 100% technical success and R0 resection rate for ESD. Most expert centres in East Asia have not achieved these results (table 6) and given the steep learning curve for ESD, success rates have been much lower in Western centres. A case series of 82 rectosigmoid ESDs from Germany showed en bloc and R0 resection rates of 82% and 70%, respectively.36 In another study from France involving 45 patients with rectal lesions, the R0 resection rate was 53% with an immediate perforation rate of 18%.37 As expected, lowering the R0 resection rate to 71% based on a recent meta-analysis31 resulted in fewer surgeries prevented by ESD and increased the incremental cost per surgery avoided of U-ESD to $379 204 compared with S-ESD.

Table 6

Key studies of colorectal ESD in East Asia and the West

Although avoidance of surgery is desirable, some patients with high-risk lesions may derive more definitive treatment following surgical resection with added information on regional lymph node status. This is particularly relevant for proximal colonic LSLs where endoscopic risks are greater and surgery is often technically easier with lower associated risks.50 In contrast, carefully selected lesions in the rectosigmoid colon where the prevalence of SMIC is higher, endoscopic risks are lower and surgical morbidity is higher, may be more suitable for ESD.51 A recent large multicentre cohort study of lesions referred for WF-EMR identified rectosigmoid location, combined Paris classification and surface morphology, and increased size with increased risk of covert SMIC.52 Our rectal scenario analyses provide support to this concept. When we considered only rectal lesions regardless of subtype, the incremental cost per surgery avoided of U-ESD fell from $210 112 to $87 066. Among non-granular rectal lesions of greater SMIC risk, the incremental cost per surgery avoided dropped further to $32 132. Finally, for Paris 0-Is non-granular lesions where the prevalence of SMIC was over 50%, U-ESD was the most effective strategy and also the cheapest.

Healthcare costs are rising rapidly despite relatively static healthcare budgets. As such, it is imperative that practitioners and policy makers support high value, efficient healthcare.53 The findings of our study indicate that routine ESD should be reserved for rectal lesion at high risk of containing SMIC. In contrast, WF-EMR complemented by ESD for highly selected lesions offers greater value for the majority of colorectal LSLs. Non-judicious use of ESD to treat lesions that are equally suitable for WF-EMR is extremely costly and of low net benefit.

Our study has several limitations. As with most economic evaluations our results are limited by available evidence. While the model was informed by a large Western observational data set of endoscopically resected LSLs, several inputs were obtained from a variety of literature sources. Nevertheless, attempts were made to select the highest quality data and all estimates were varied extensively in sensitivity analysis. Studies evaluating the efficacy of ESD compared with WF-EMR are eagerly awaited. At least one randomised control trial of rectal ESD versus EMR is currently underway ( NCT02198729). Finally, the short time horizon may be considered by some to be a limitation of our study design. However, data are not available to permit a longer time horizon without increasing the degree of speculation and amplification of errors over time.

A central principle of all forms of clinical care is that the treatment method offered should be appropriate for the disease state under consideration.54 An ideal intervention is one that most effectively treats the disease with the lowest level of invasiveness, inconvenience, risk and cost to the patient and the healthcare system. Compared with surgery, WF-EMR is safe, effective, ambulatory, organ-sparing and relatively inexpensive.4 11 13–15 ESD complements WF-EMR in select circumstances and most notably in the rectum. To further our understanding, carefully designed prospective comparative studies are warranted in referral centres with proven expertise in advanced endoscopic tissue resection.



  • Contributors FFB designed the study, collected data, analysed data, drafted the manuscript and revised the manuscript after review by the coauthors. SJH Designed the economic model, analysed data and codrafted the manuscript. KNR collected data. HM performed histological analysis. DML performed histological analysis. EYTL performed procedures and collected data. SJW performed procedures and collected data. MJB initiated, designed and led the study, performed procedures, collected data, cowrote the manuscript and critically reviewed the manuscript.

  • Funding FFB was supported by a grant from the National Health and Medical Research Council of Australia (NHMRC). There was no influence from the NHMRC on study design or conduct, data collection and management, analysis,interpretation, preparation and review or approval of the manuscript. SJH was supported by the N.B. Hershfield Professorship in therapeutic endoscopy.

  • Competing interests None declared.

  • Ethics approval Western Sydney Local Health District Human Ethics Research Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement There are no additional unpublished data from this study.

  • Correction notice This article has been corrected since it published Online First. Figure 1 has been updated.