- Email: [email protected]
Endoscopic mucosal resection: learning curve for large nonpolypoid colorectal neoplasia
Accepted Manuscript Endoscopic mucosal resection learning curve for large non-polypoid colorectal neoplasia Abhishek Bhurwal, MD, Michael J. Bartel, MD, Michael G. Heckman, MS, Nancy N. Diehl, Massimo Raimondo, MD, Michael B. Wallace, MD, MPH, Timothy A. Woodward, MD PII:
S0016-5107(16)30066-9
DOI:
10.1016/j.gie.2016.04.020
Reference:
YMGE 9979
To appear in:
Gastrointestinal Endoscopy
Received Date: 12 January 2016 Accepted Date: 12 April 2016
Please cite this article as: Bhurwal A, Bartel MJ, Heckman MG, Diehl NN, Raimondo M, Wallace MB, Woodward TA, Endoscopic mucosal resection learning curve for large non-polypoid colorectal neoplasia, Gastrointestinal Endoscopy (2016), doi: 10.1016/j.gie.2016.04.020. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Endoscopic mucosal resection learning curve for large non-polypoid colorectal neoplasia
Abhishek Bhurwal, MD1*, Michael J. Bartel, MD1*, Michael G. Heckman, MS2, Nancy N.
MD1
RI PT
Diehl2, Massimo Raimondo, MD1, Michael B. Wallace, MD, MPH1, Timothy A. Woodward,
Division of Gastroenterology, Mayo Clinic, Jacksonville, FL, USA
2
Division of Biomedical Statistics and Informatics, Mayo Clinic, Jacksonville, FL, USA
Corresponding author: Timothy A. Woodward, MD
M AN U
*authors should be considered co-first authors
SC
1
4500 San Pablo Road Jacksonville, FL 32224
EP
+1 (904) 953 2221
TE D
Mayo Clinic, Division of Gastroenterology
AC C
[email protected]
Author Contributions: Dr. Bhurwal contributed to the collection of data, generation of figures, and critical revision of the article. Dr. Bartel contributed to the analysis and interpretation of the data, generation of the figures, and drafting and critical revision of the article. Mr. Heckman and Ms. Diehl contributed to the analysis and interpretation of the data, generation of the figures, and critical revision of the article. Drs. Raimondo and Wallace contributed to the experiments and critical revision of the article. Dr. Woodward contributed to the conception and design of the
ACCEPTED MANUSCRIPT
study, experiments, analysis and interpretation of the data, and critical revision of the article. All authors approved the final draft of the article.
AC C
EP
TE D
M AN U
SC
RI PT
Keywords: Endoscopic mucosal resection; learning curve; colorectal neoplasia
ACCEPTED MANUSCRIPT -1-
Endoscopic mucosal resection learning curve for large non-polypoid colorectal neoplasia
ABSTRACT
RI PT
Background and Aims: Colorectal endoscopic mucosal resection (EMR) for non-polypoid neoplasia achieves better outcomes when performed by expert endoscopists. The time point when the endoscopist achieves expert level remains to be defined. The objective of this study
tissue on surveillance colonoscopy and adverse event rate.
SC
was to establish a learning curve of colorectal EMR for non-polypoid neoplasia based on residual
M AN U
Methods: Five hundred seventy-eight consecutive patients underwent EMR of colorectal neoplasia by 1 of 3 primary endoscopists between December 2004 and September 2013 in a tertiary academic center. Primary analyses focused on the largest lesion for patients with more than one lesion (median age 69 years, median polyp size 30 mm, 51% en bloc resection). Data on
TE D
surveillance colonoscopy were available in 74%. Learning curves were calculated for each of the 3 main outcome measurements: the presence of residual neoplasia on surveillance colonoscopy, endoscopic assessment of incomplete EMR, and the occurrence of an immediate bleeding
EP
adverse event.
Results: Residual neoplasia on surveillance colonoscopy was present for 23.2% of patients, the
AC C
rate of endoscopist-assessed incomplete EMR was 27.6%, and immediate bleeding adverse events occurred in 6.9% of patients. Although there was between-endoscopist variability, the overall rates of residual neoplasia and incomplete EMR decreased to below 20% to 25% after 100 EMRs; initial declines in both rates were observed for earlier EMRs. Immediate bleeding adverse events occurred at a low frequency for each endoscopists across all EMRs. Perforation requiring surgical intervention occurred in 1 patient (0.2%).
ACCEPTED MANUSCRIPT -2-
Conclusions: This study demonstrated that an unexpectedly high number of 100 colorectal EMR procedures for large non-polypoid colorectal neoplasia are required to achieve a plateau phase
RI PT
for crucial outcomes.
INTRODUCTION
Endoscopic resection of large (≥20 mm) colorectal non-polypoid neoplasia is commonly
SC
performed using the endoscopic mucosal resection (EMR) technique. The reported data on shortand long-term outcomes demonstrate substantial differences in the en bloc resection and residual
M AN U
neoplasia rate.1-4 These differences can be partly explained by included polyp sizes, polyp sites, and polyp histology. However, whether different endoscopist EMR skills contribute to these differences is not known.
Despite the broad application of colorectal EMR for neoplasia, surprisingly few data exist
TE D
on the required procedure volume per endoscopist to achieve satisfactory outcomes.5,6 In the recent past, attention was drawn to endoscopic competency and quality measures in order to provide optimal patient care. In this context, learning curves can be established to define
outcomes.7-11
EP
parameters when a sufficient procedure skill set is acquired to allow satisfactory procedure
AC C
Our primary goal was to define a learning curve for EMR of large colorectal non-
polypoid neoplasia, focusing on residual neoplasia at follow-up, endoscopist assessment of incomplete EMR, and immediate bleeding adverse events as quality-defining outcomes.
METHODS Patients
ACCEPTED MANUSCRIPT -3-
We included all patients who underwent EMR for sessile colorectal polyps, 20 mm or larger, from December 2004 to September 2013 at our academic endoscopy unit in this retrospective study. Institutional Review Board approval was obtained. All EMR procedures were performed
RI PT
by the advanced endoscopy group (M.R., M.B.W., and T.A.W.) at our institute since December 2004, with all endoscopists being more than 5 years out of formal gastroenterology training. Before this date, none of the physicians had performed appreciable numbers of EMR procedures.
SC
This resulted in a sample of 578 colorectal EMRs in 578 different patients; with all EMRs
n=223, Endoscopist no. 3: n=156).
M AN U
performed by the 3 aforementioned endoscopists (Endoscopist no. 1: n=199, Endoscopist no. 2:
A subset of patients (n=66) had more than one lesion for a given colonoscopy with EMR, ie, these patients underwent simultaneously 2 or more EMRs. Our primary analysis focused on the largest lesion for the given colonoscopy. However, in a secondary analysis, we also evaluated
TE D
presence of each of the 3 primary endpoints (residual neoplasia at follow-up, immediate bleeding
EMR.
EMR information
EP
adverse events, and endoscopist assessment of incomplete EMR) for any lesions undergoing
AC C
EMR materials and technique varied depending on the lesion and the preference of the operators. Colorectal lesions were removed by submucosal injection of a large volume of normal saline solution (normal saline solution; indigo carmine, 0.04%, Taylor Pharmaceuticals, Decatur, Ill; with or without 1:10,000 epinephrine) or a solution of hydroxypropyl methylcellulose (hydroxypropyl methylcellulose, normal saline solution, indigo carmine, 1:10,000 epinephrine). Lesions were marked with tattoo (Spot, GI supply, Camp Hill, Pa), per endoscopist discretion, to
ACCEPTED MANUSCRIPT -4-
facilitate site recognition at follow-up. Piecemeal or en bloc resection was depending on lesion size and morphology. At the initial endoscopy, supplemental methods were used per endoscopist discretion to achieve maximal macroscopic polyp resection, which included argon plasma
RI PT
coagulation (APC), EMR-cap, ablation cautery, snare, forceps avulsion, and endoloop. APC was used to treat residual macroscopic neoplasia, but not for prophylactic measures. An endoloop was used to facilitate wound closure after snare resection of particularly large neoplasia.
SC
All patients were scheduled for surveillance colonoscopy at 3 to 6 months. In light of limited compliance with this timeframe, all first surveillance colonoscopies were included in our
M AN U
evaluation and were performed within 24 months after the initial EMR. All follow-up colonoscopies were performed by the advanced endoscopy group.
Data collection and outcomes
TE D
A retrospective chart review was performed to extract the following data: baseline characteristics (age, gender, race, prior EMR or snare resection attempt [prior therapy], American Society of Anesthesiologists [ASA] class), EMR information (endoscopist, colon site, polyp size, Paris
EMR,
EP
classification, Kudo classification, Sano classification, injection type, lift sign, cap-assisted EMR snare methods, prophylactic APC, prophylactic clips, specimen [en bloc vs
AC C
piecemeal], optical methods (narrow-band imaging [NBI]), supplemental methods to remove residual neoplasia after EMR (ablation-APC, EMR-snare, Grasper assist [snare resection and simultaneous grasper assist with dual-channel endoscope], snare, EMR-cap, ablation-cautery, Endo loops), and outcomes (endoscopist assessment of complete resection with EMR, endoscopist assessment of complete resection with EMR using NBI, neoplasia histology according to pathology report, immediate adverse events, residual neoplasia at follow-up).
ACCEPTED MANUSCRIPT -5-
Residual neoplasia at follow-up was defined as presence of macroscopically visible neoplastic tissue at the polypectomy scar which was confirmed histologically in each case. Other quality measures like procedure times which include the withdrawal time, bowel preparation quality and
RI PT
pain score assessments were not available for data analysis. Description of neoplasia, including size and other classifications, was based on real-time endoscopist assessment without the use of additional tools (eg, for exact measurement of neoplasia size). There was a substantial amount of
SC
missing data for ASA class (n=78), Paris classification (n=380), Kudo classification (n=367), and Sano classification (n=446). The 3 primary outcome measures were residual macroscopic
M AN U
neoplasia at follow-up, endoscopist assessment of incomplete EMR, and immediate bleeding adverse events, defined as unexpected hemorrhage during EMR requiring additional treatment for hemostasis (epinephrine, clip, APC, cautery, endoloop). EMR specimen of 19 patients revealed T1 adenocarcinoma, of which 9 were referred to surgery, with the remaining patients (5
Statistical analysis
TE D
T1m and 5 T1sm) undergoing surveillance colonoscopy according to patient’s preference.
EP
Continuous variables were summarized using the sample median and range. Categorical variables were summarized using number and percentage. In order to address potential
AC C
confounding in learning curve evaluation, separately for each endoscopist and in the overall group, we examined changes in prior therapy, piecemeal specimen, and polyp size as endoscopist EMR experience increased using logistic regression models (prior therapy and specimen) and linear regression models (polyp size). Odds ratios (ORs) and 95% confidence intervals were estimated, and this analysis was performed separately for each endoscopist and overall. Linear and logistic regression models involving the overall group were adjusted for endoscopist. Polyp size was considered on the logarithm scale in all analysis due to its skewed distribution.
ACCEPTED MANUSCRIPT -6-
The learning curves of EMR regarding the 3 endpoints were assessed using a combination of descriptive summaries, graphical evaluation, and statistical tests. First, for each endoscopist, we divided EMRs into sequential groups of 20, and estimated the proportion of
RI PT
patients for whom each endpoint was experienced separately in each of these 20-EMR groups. These estimates were then combined across all 3 endoscopists. A similar strategy was followed for graphical display, except that a group size of 40 EMRs was used for easier presentation of
SC
graphical results. To evaluate how the frequency of each of the 3 binary categorical endpoints changed as EMR experience increased, we used logistic regression models. Separately for each
M AN U
endpoint and endoscopist, we included a variable in the model for endoscopist sequential EMR number in order to assess any systematic increases or decreases in the frequency of the given endpoint as EMR experience increased. Similar logistic regression models were used when combining across the 3 endoscopists in order to assess overall trends, with additional model
TE D
adjustment for endoscopist. Additional model adjustment for the potential confounding variables of prior therapy, piecemeal specimen, and polyp size was also considered. P-values of 0.05 or lower were considered as statistically significant. All statistical analysis was performed using
EP
SAS (version 9.2; SAS Institute, Inc., Cary, North Carolina) and R Statistical Software (version
AC C
2.14.0; R Foundation for Statistical Computing, Vienna, Austria).
RESULTS
Patient and EMR characteristics and outcomes A summary of baseline patient characteristics is provided in Table 1, whereas EMR information is shown in Table 2. Of the initial cohort of 578 patients, 427 (74%) had a follow-up visit with residual neoplasia information available; this visit occurred at a median of 4.2 months (range: 1–
ACCEPTED MANUSCRIPT -7-
23.7 months) after the initial EMR. When comparing the 427 patients with follow-up residual neoplasia information available to the 151 patients without it, characteristics were similar with the exception of ASA class (ASA 3-4: 35% vs 52%, p<0.001).
RI PT
Patient and EMR outcomes are displayed in Table 3. Endoscopists assessed EMR to be incomplete in 27.6% of patients, but this was reduced to 2.0% after use of supplemental methods. The most common pathologies were tubular adenoma (40%), tubulovillous adenoma
SC
(28%), sessile serrated adenoma (10%), and hyperplastic polyps (9%). Regarding immediate adverse events, bleeding occurred in 40 patients (7%) and confirmed perforation occurred in only
M AN U
1 patient (0.2%), who required surgical intervention. Only immediate bleeding adverse events were evaluated in the learning curve analysis given the rare or non-existent occurrence of other immediate adverse events. Residual neoplasia at follow-up was present in 99 of the 427 patients
TE D
(23%) for whom follow-up information was available.
Potential impact of prior therapy, piecemeal specimen, and polyp size on learning curve results
EP
Before examining the learning curves for residual neoplasia, immediate bleeding adverse events, and endoscopist assessment of incomplete EMR, we first investigated whether prior therapy at
AC C
the EMR site, specimen resection type (piecemeal or en bloc), or polyp size have the potential to impact learning curve results. Supplemental Tables and Figures 1, 2, and 3 display prior therapy, specimen, and polyp size according to endoscopist EMR experience, separately for each endoscopist, and overall. As endoscopist EMR experience increased, there was evidence of more frequent prior therapy for endoscopists no. 2 (p=0.006) and no. 3 (p=0.032), and less frequent piecemeal resection for all 3 endoscopists (all p≤0.022). Interestingly, polyp size did not
ACCEPTED MANUSCRIPT -8-
systematically change for Endoscopist no. 1 and no. 2; however, it decreased for Endoscopist no. 3 with gained experience (p = 0.0004). To better understand the impact that these 3 variables may have on learning curve results,
RI PT
we next evaluated the association between each variable and each learning curve outcome, and these results are shown in Supplemental Table 4. Overall, none of the 3 outcomes were associated with prior therapy at EMR site (all p ≥ 0.16). There was an association of piecemeal
SC
specimen with an increased risk of residual neoplasia at follow-up (OR, 3.19; p < 0.0001) and endoscopist assessment of incomplete EMR (OR, 3.84; p < 0.0001), and an association of polyp
M AN U
size with an increased risk of residual neoplasia at follow-up (OR, 2.02; p = 0.001), immediate bleeding adverse events (OR, 1.98; p = 0.011), and endoscopist assessment of incomplete EMR (OR, 1.63; p = 0.003) (Supplemental Table 4).
TE D
Residual neoplasia at follow-up learning curve
The proportion of patients who had residual neoplasia at follow-up is shown according to endoscopist EMR experience in Figure 1 and Table 4. Without accounting for potential
EP
confounding variables, there was evidence of a decrease in the proportion of patients with residual neoplasia at follow-up as EMR experience increased for endoscopists no. 1 (p=0.031)
AC C
and no. 3 (p=0.001), whereas there was no evidence of a systematic trend for Endoscopist no. 2 (p=0.14). For endoscopists no. 1 and no. 3, the proportion of patients with residual neoplasia at follow-up was consistently at or below 20% after EMR 100. For Endoscopist no. 2, though there was a notable decrease in the proportion of patients with residual neoplasia at follow-up during EMRs 100 to 140 (to between 10% and 20%), this was followed by a slight increase for EMRs 141 to 223 (to between 20% and 30%). When combining across the 3 endoscopists, there was
ACCEPTED MANUSCRIPT -9-
strong evidence of an overall decreasing trend (p=0.0004) with the proportion of patients with residual neoplasia being approximately 20% or lower after EMR 100, with the exception of EMRs 201 to 220, which only involved Endoscopist no. 2.
RI PT
In order to examine how prior therapy, piecemeal specimen, and polyp size may have affected trends in residual neoplasia at follow-up as endoscopist EMR experience increased, we adjusted our models for these 3 variables (Table 4). Results were similar for endoscopists no. 1
SC
and no. 2; however, for Endoscopist no. 3, the decreasing trend weakened noticeably, which may be due to the decreasing frequency of piecemeal specimen (Supplemental Tables and Figures 1-
M AN U
3). When examining all lesions for a given EMR, instead of only the largest lesion, results were similar (data not shown).
Immediate bleeding adverse events learning curve
TE D
The frequency of immediate bleeding adverse events according to endoscopist EMR experience is displayed in Figure 2 and Table 5. There was no evidence of a change in the proportion of patients with an immediate bleeding adverse event as EMR experience increased for Endoscopist
EP
no. 2 (p=0.82) and no. 3 (p=0.17), whereas there was actually evidence of a slight increasing trend for Endoscopist no. 1 (p=0.027). When combining across all 3 endoscopists, no significant
AC C
change in the proportion of patients with an immediate bleeding adverse event was observed as EMR experience increased (p=0.27). Despite any increasing or decreasing trends, the frequency of immediate bleeding adverse events was fairly low for all sequential 20-EMR groups. Results were similar when adjusting for prior therapy, piecemeal specimen, and polyp size (Table 5) and also when examining all lesions for a given EMR instead of only the largest lesion (data not shown).
ACCEPTED MANUSCRIPT - 10 -
Endoscopist assessment of incompletion learning curve Figure 3 and Table 6 display the proportion of EMRs that were assessed as incomplete by the
RI PT
given endoscopist according to endoscopist EMR experience. For both Endoscopist no. 1 and no. 2, there was evidence of a decreasing frequency of incomplete EMR as EMR experience increased (p=0.001 and <0.0001, respectively). For Endoscopist no. 1, the frequency of
SC
endoscopist-assessed incomplete EMR was consistently at or below 20% after EMR 140, whereas for Endoscopist no. 2 this frequency was below 20% after EMR 100. For Endoscopist
M AN U
no. 3, there was no evidence of a systematic increase or decrease in endoscopist-assessed incomplete EMR as EMR experience increased (p=0.26), with higher frequencies occurring in both earlier and later EMRs. When combining information for the 3 endoscopists, there was strong evidence of a decreasing trend in the frequency of endoscopist assessment of incomplete
TE D
EMR as EMR experience increased (p<0.0001); this frequency was approximately 20% to 25% or lower after EMR 100 in the combined group. These results were consistent when evaluating all lesions for a given EMR instead of only the largest lesion (data not shown) and when
AC C
DISCUSSION
EP
adjusting for prior therapy, specimen, and polyp size (Table 6).
This study addresses the learning curve for EMR of large non-polypoid colorectal neoplasia. The learning curve plateaued at 100 EMRs, reflecting a 20% or lower rate of residual neoplasia at follow-up, though there was inter-endoscopist variation for the endpoints. However, immediate bleeding adverse events occurred at a low and relatively stable frequency as EMR experience increased for each endoscopist, arguing that technical progress played no important role for immediate bleeding. Also, when controlling for the most likely confounders, including polyp
ACCEPTED MANUSCRIPT - 11 -
size, site, histology, and rate of piecemeal resection, the EMR number to achieve a learning curve plateau remained the same.
events rate (6.1%) are comparable with previous studies.2, 12-14
RI PT
The outcomes of EMR in this study regarding residual neoplasia (22%) and adverse
At least 90% of patients with large colorectal neoplasia can undergo colorectal EMR, and therefore, avoid surgical resection, which has a higher predicted mortality rate than EMR.1, 15
SC
Despite the variety of EMR outcomes (piecemeal resection rate 7-100%, residual or recurrent neoplasia rate at follow-up 0-55%1, 16), and the well-delineated predictors, which mainly focus on
achieve high EMR quality.17, 18
M AN U
polyp size, site, and histology, surprisingly few data are available on endoscopist competency to
It is evident that EMR in expert hands has better outcomes as implicated by Buchner et al12 and Swan et al19 who addressed EMR outcomes for patients particularly referred for EMR,
TE D
yielding en bloc resection rates of 34% to 53.5% and recurrent or residual neoplasia at follow-up of 10.5% to 27%. However, questions remain to be answered; when does the endoscopist reach a high EMR quality, and how is EMR quality defined? In this study, we defined our primary
EP
endpoints residual or recurrent neoplasia, complete macroscopic neoplasia resection during EMR, and early rebleeding as quality indicators.
AC C
A common presumption that procedure volume is associated with better outcome was
also shown by a British study. Endoscopic specialists had less residual polyp at follow-up than non-specialist.6 Also, Baxter et al20 demonstrated a lower interval colon cancer rate when patients underwent colonoscopy performed by an endoscopist with a high completion rate; however, the endoscopist’s procedure volume was not, per se, associated with interval colon cancer. Instruments to define endoscopist quality are well known and range from ADR to
ACCEPTED MANUSCRIPT - 12 -
competency assessments questionnaires like the Direct Observation of Polypectomy Skills (DOPyS) questionnaire; however, none of them can be used to define EMR quality, particularly as EMR key steps are not a part of the questionnaire, like choice of injection solution or
RI PT
particular snare choice.21, 22
To resolve this dilemma, important lessons can be learned from endoscopic submucosal dissection (ESD).7 The initial ESD studies report self-taught acquisition of skills, which reflects
SC
the learning method in our study. Hotta et al23 reported 40 ESDs to minimize perforations and 80 ESDs to master a maximum en bloc (97.5%) and R0 (92.5%) resection rate. Choi et al11 and
M AN U
Probst et al24 reported comparable numbers to perform high quality gastric and colorectal ESD, respectively. Lower ESD procedure volumes to achieve plateau phase were demonstrated when an ESD novice was supervised by ESD experts, achieving high quality gastric and colorectal ESDs after 30 supervised procedures.8, 9
TE D
EMR is a less complex procedure than ESD; however, our number of 100 EMRs to achieve a plateau phase is higher than expected. Based on our results, the plateau phase was followed by a phase with increased residual tissue and increased immediate bleeding rate.
EP
Although no significant change of residual neoplasia was found when adjusting for confounders, there still remained a down trending rate. We suspected that the higher rate of polypectomies for
AC C
polyps with prior therapy by 2 of the 3 endoscopists (p=0.006 and p=0.032) contributed to the increased rate of residual neoplasia after the plateau phase (Supplemental Tables and Figures 13). The evidence of prior therapy leading to increase in residual rates emphasizes that large polyps should be handled by experts at high-volume centers and that therapy should not be attempted until it is done by definitive resection.
ACCEPTED MANUSCRIPT - 13 -
We further speculate that the relatively high number of EMRs to achieve expertise is based on the self-teaching nature of this study, which is also reflected by including larger neoplastic lesions and lesions which previously failed to be resected in other facilities over the
RI PT
study period. Most instruments used for EMR were available throughout the study period, which argues that the endoscopists, not the new equipment, were measured. Extrapolating from other endoscopic and surgical techniques, and in order to minimize patient risk and maximize EMR
SC
outcomes for future EMR endoscopist generations, we believe that self-taught learning methods, as demonstrated in this study, should be avoided. We believe that specialty centers with well-
M AN U
established standards of care should provide training for EMR with direct expert supervision, as already offered in national gastroenterology society workshops. In this context, animal models are important as first training steps.25, 26
This study is limited by several aspects. First, a retrospective study performed in a
TE D
tertiary center raises concerns of a referral bias for particularly complicated EMRs, which consequently may have increased the EMR number to achieve a plateau phase in the learning curve. Missing or not reported data, which are reflected in the absence of a standardized post
EP
procedure recording and missing photo documentation, might have led to an underestimation of the prevalence of the main study outcomes, such as residual neoplasia, and in particular, adverse
AC C
events. For the same reason, scoring systems addressing the resection difficulty of a polyp could have not been used. However, such possible systematic errors would have been present over the entire study period and for all 3 endoscopists. Consequently, it would have had less effect on calculation of the plateau phase of the learning curve, which was our main goal. Second, this study included patients undergoing EMR over a time period of 9 years with a broad variety of neoplasia ranging from 20 to 150 mm. One can argue that techniques have
ACCEPTED MANUSCRIPT - 14 -
changed over 9 years which biased the learning curve. However, the endoscopic tools and instruments like snares and injection solutions were available over the entire study period. We did not assess changes in the utility of these methods during the study period, but it is very likely
RI PT
that, over 9 years, the endoscopists adjusted their techniques according to the variety of neoplastic lesions. From our perspective, this is the actual learning process, which includes subjective assessment of each neoplasm in order to choose the optimal EMR technique.
SC
Third, we cannot exclude additional selection bias due to patient drop out, although most patients were followed up (74%). In addition, the follow-up intervals varied widely between 3
M AN U
and 24 months, with most follow-up colonoscopies within 3 to 6 months, which is consistent with the multisociety guidelines. It is possible that a few patients had late residual neoplasia presenting after 1 year that could have been missed in our cohort. According to Khashab et al13 who reported late residual in 4.4% of all residual neoplasia cases, we suspect that missing late
TE D
residual neoplasia was of less importance.
Last, unexperienced endoscopists and trainees were not included. In what way the inclusion of trainees under expert supervision would have changed the results remains to be
EP
determined.
This study demonstrates an EMR learning curve for large non-polypoid colorectal
AC C
neoplasia. Our data indicate an unexpectedly high number (100 EMRs) to achieve a plateau phase of the learning curve, which reflects the complete polyp resection rate, rate of residual or recurrent neoplasia at resection site, and immediate adverse events. We recommend that training programs and credentialing bodies consider both volume and residual neoplasia as parameters for certification. Last, multicenter studies are necessary to determine what thresholds for residual neoplasia amongst other quality indicators should be used to determine proficiency in EMR.
ACCEPTED MANUSCRIPT - 15 -
REFERENCES 1.
Kaltenbach T, Friedland S, Maheshwari A, Ouyang D, Rouse RV, Wren S, et al. Short-
RI PT
and long-term outcomes of standardized EMR of nonpolypoid (flat and depressed) colorectal lesions > or = 1 cm (with video). Gastrointest Endosc. 2007 May;65:857-65. 2.
Belle S, Haase L, Pilz LR, Post S, Ebert M, Kaehler G. Recurrence after endoscopic
3.
SC
mucosal resection-therapy failure? Int J Colorectal Dis. 2014 Feb;29:209-15.
Su MY, Hsu CM, Ho YP, Lien JM, Lin CJ, Chiu CT, et al. Endoscopic mucosal resection
4.
M AN U
for colonic non-polypoid neoplasms. Am J Gastroenterol. 2005 Oct;100:2174-9. Wang AY, Ahmad NA, Zaidman JS, Brensinger CM, Lewis JD, Long WB, et al.
Endoluminal resection for sessile neoplasia in the GI tract is associated with a low recurrence rate and a high 5-year survival rate. Gastrointest Endosc. 2008 Jul;68:160-9. Heresbach D, Kornhauser R, Seyrig JA, Coumaros D, Claviere C, Bury A, et al. A
TE D
5.
national survey of endoscopic mucosal resection for superficial gastrointestinal neoplasia. Endoscopy. 2010 Oct;42:806-13.
Brooker JC, Saunders BP, Shah SG, Williams CB. Endoscopic resection of large sessile
EP
6.
colonic polyps by specialist and non-specialist endoscopists. Br J Surg. 2002 Aug;89:1020-4. Uraoka T, Parra-Blanco A, Yahagi N. Colorectal endoscopic submucosal dissection in
AC C
7.
Japan and Western countries. Dig Endosc. 2012 May;24 Suppl 1:80-3. 8.
Yamamoto S, Uedo N, Ishihara R, Kajimoto N, Ogiyama H, Fukushima Y, et al.
Endoscopic submucosal dissection for early gastric cancer performed by supervised residents: assessment of feasibility and learning curve. Endoscopy. 2009 Nov;41:923-8.
ACCEPTED MANUSCRIPT - 16 -
9.
Sakamoto T, Saito Y, Fukunaga S, Nakajima T, Matsuda T. Learning curve associated
with colorectal endoscopic submucosal dissection for endoscopists experienced in gastric endoscopic submucosal dissection. Dis Colon Rectum. 2011 Oct;54:1307-12. Iacopini F, Bella A, Costamagna G, Gotoda T, Saito Y, Elisei W, et al. Stepwise training
RI PT
10.
in rectal and colonic endoscopic submucosal dissection with differentiated learning curves. Gastrointest Endosc. 2012 Dec;76:1188-96.
Probst A, Golger D, Anthuber M, Markl B, Messmann H. Endoscopic submucosal
SC
11.
Endoscopy. 2012 Jul;44:660-7. 12.
M AN U
dissection in large sessile lesions of the rectosigmoid: learning curve in a European center.
Buchner AM, Guarner-Argente C, Ginsberg GG. Outcomes of EMR of defiant colorectal
lesions directed to an endoscopy referral center. Gastrointest Endosc. 2012 Aug;76:255-63. 13.
Khashab M, Eid E, Rusche M, Rex DK. Incidence and predictors of "late" recurrences
Aug;70:344-9. 14.
TE D
after endoscopic piecemeal resection of large sessile adenomas. Gastrointest Endosc. 2009
Moss A, Bourke MJ, Williams SJ, Hourigan LF, Brown G, Tam W, et al. Endoscopic
EP
mucosal resection outcomes and prediction of submucosal cancer from advanced colonic mucosal neoplasia. Gastroenterology. 2011 Jun;140:1909-18. Ahlenstiel G, Hourigan LF, Brown G, Zanati S, Williams SJ, Singh R, et al. Actual
AC C
15.
endoscopic versus predicted surgical mortality for treatment of advanced mucosal neoplasia of the colon. Gastrointest Endosc. 2014 Oct;80:668-76. 16.
Ah Soune P, Menard C, Salah E, Desjeux A, Grimaud JC, Barthet M. Large endoscopic
mucosal resection for colorectal tumors exceeding 4 cm. World J Gastroenterol. 2010 Feb 7;16:588-95.
ACCEPTED MANUSCRIPT - 17 -
17.
Woodward TA, Heckman MG, Cleveland P, De Melo S, Raimondo M, Wallace M.
Predictors of complete endoscopic mucosal resection of flat and depressed gastrointestinal neoplasia of the colon. Am J Gastroenterol. 2012 May;107:650-4. Kim HH, Kim JH, Park SJ, Park MI, Moon W. Risk factors for incomplete resection and
RI PT
18.
complications in endoscopic mucosal resection for lateral spreading tumors. Dig Endosc. 2012 Jul;24:259-66.
Swan MP, Bourke MJ, Alexander S, Moss A, Williams SJ. Large refractory colonic
SC
19.
polyps: is it time to change our practice? A prospective study of the clinical and economic
M AN U
impact of a tertiary referral colonic mucosal resection and polypectomy service (with videos). Gastrointest Endosc. 2009 Dec;70:1128-36. 20.
Baxter NN, Sutradhar R, Forbes SS, Paszat LF, Saskin R, Rabeneck L. Analysis of
administrative data finds endoscopist quality measures associated with postcolonoscopy
21.
TE D
colorectal cancer. Gastroenterology. 2011 Jan;140:65-72.
Gupta S, Bassett P, Man R, Suzuki N, Vance ME, Thomas-Gibson S. Validation of a
novel method for assessing competency in polypectomy. Gastrointest Endosc. 2012 Mar;75:568-
22.
EP
75.
Kaminski MF, Regula J, Kraszewska E, Polkowski M, Wojciechowska U, Didkowska J,
AC C
et al. Quality indicators for colonoscopy and the risk of interval cancer. N Engl J Med. 2010;362:1795-803. 23.
Hotta K, Oyama T, Shinohara T, Miyata Y, Takahashi A, Kitamura Y, et al. Learning
curve for endoscopic submucosal dissection of large colorectal tumors. Dig Endosc. 2010 Oct;22:302-6.
ACCEPTED MANUSCRIPT - 18 -
24.
Choi IJ, Kim CG, Chang HJ, Kim SG, Kook MC, Bae JM. The learning curve for EMR
with circumferential mucosal incision in treating intramucosal gastric neoplasm. Gastrointest Endosc. 2005 Dec;62:860-5. Parra-Blanco A, Arnau MR, Nicolas-Perez D, Gimeno-Garcia AZ, Gonzalez N, Diaz-
RI PT
25.
Acosta JA, et al. Endoscopic submucosal dissection training with pig models in a Western country. World J Gastroenterol. 2010 Jun 21;16:2895-900.
Herreros de Tejada A. ESD training: A challenging path to excellence. World J
M AN U
Gastrointest Endosc. 2014 Apr 16;6:112-20.
SC
26.
FIGURE LEGENDS
Figure 1: Frequency of residual neoplasia as endoscopic mucosal resection (EMR) experience increases for each of the 3 endoscopists and overall. Separately for each endoscopist, EMRs were
TE D
divided into sequential groups of 40, and the proportion of patients with residual neoplasia was estimated for each 40-EMR group. These endoscopist-specific estimates were then combined into one overall estimate across all 3 endoscopists; an overall estimate was not calculated for
EP
EMRs greater than 200 because only Endoscopist no. 2 performed more than 200 EMRs.
AC C
Figure 2: Frequency of immediate bleeding adverse events as endoscopic mucosal resection (EMR) experience increases for each of the 3 endoscopists and overall. Separately for each endoscopist, EMRs were divided into sequential groups of 40, and the proportion of patients with immediate bleeding adverse events was estimated for each 40-EMR group. These endoscopistspecific estimates were then combined into one overall estimate across all 3 endoscopists; an
ACCEPTED MANUSCRIPT - 19 -
overall estimate was not calculated for EMRs greater than 200 because only Endoscopist no. 2 performed more than 200 EMRs.
RI PT
Figure 3: Frequency of endoscopic mucosal resections (EMR) assessed as incomplete by
endoscopist as EMR experience increases for each of the 3 endoscopists and overall. Separately for each endoscopist, EMRs were divided into sequential groups of 40, and the proportion of
SC
patients with an incomplete assessment after EMR was estimated for each 40-EMR group. These endoscopist-specific estimates were then combined into one overall estimate across all 3
M AN U
endoscopists; an overall estimate was not calculated for EMRs greater than 200 because only Endoscopist no. 2 performed more than 200 EMRs.
SUPPLEMENTAL FIGURE LEGEND
TE D
Supplemental Figure 1: Frequency of prior therapy as endoscopic mucosal resection (EMR) experience increases for each of the 3 endoscopists and overall. Separately for each endoscopist, EMRs were divided into sequential groups of 40, and the proportion of patients with prior
EP
therapy was estimated for each 40-EMR group. These endoscopist-specific estimates were then combined into one overall estimate across all 3 endoscopists; an overall estimate was not
AC C
calculated for EMRs greater than 200 because only Endoscopist no. 2 performed more than 200 EMRs.
Supplemental Figure 2: Frequency of piecemeal specimen as endoscopic mucosal resection (EMR) experience increases for each of the 3 endoscopists and overall. Separately for each endoscopist, EMRs were divided into sequential groups of 40, and the proportion of patients with
ACCEPTED MANUSCRIPT - 20 -
piecemeal specimen was estimated for each 40-EMR group. These endoscopist-specific estimates were then combined into one overall estimate across all 3 endoscopists; an overall estimate was not calculated for EMRs greater than 200 because only Endoscopist no. 2
RI PT
performed more than 200 EMRs.
Supplemental Figure 3: Median polyp size as endoscopic mucosal resection (EMR) experience
SC
increases for each of the 3 endoscopists and overall. Separately for each endoscopist, EMRs were divided into sequential groups of 40, and median polyp size was estimated for each 40-EMR
M AN U
group. These endoscopist-specific estimates were then combined into one overall estimate across all 3 endoscopists; an overall estimate was not calculated for EMRs greater than 200 because
AC C
EP
TE D
only Endoscopist no. 2 performed more than 200 EMRs.
ACCEPTED MANUSCRIPT - 21 -
TABLES Table 1: Patient information Variable
All patients (N=578)
Patients with follow-up within 24 months of original EMR (N=427) 68.2 (22.0, 95.1) 232 (54.3%) 398 (93.6%) 65 (15.3%)
AC C
EP
TE D
M AN U
SC
RI PT
Age at EMR 68.6 (22.0, 95.1) Gender (Male) 306 (52.9%) Race (Caucasian) 539 (94.1%) Prior therapy 85 (14.8%) ASA class 1 13 (2.6%) 12 (3.3%) 2 291 (58.2%) 228 (62.1%) 3 190 (38.0%) 122 (33.2%) 4 6 (1.2%) 5 (1.4%) Continuous variables are summarized using the sample median (range). Information was unavailable regarding race (N=5 overall, N=2 in follow-up group), prior therapy (N=2 overall, N=1 in follow-up group), and ASA class (N=78 overall, N=60 in follow-up group). ASA: American Society of Anesthesiologists; EMR: endoscopic mucosal resection There was no significant difference for any variable.
ACCEPTED MANUSCRIPT - 22 -
Table 2: EMR information
199 (34.4%) 223 (38.6%) 156 (27.0%)
139 (32.6%) 165 (38.6%) 123 (28.8%)
RI PT
Patients with follow-up within 24 months of original EMR (N=427)
M AN U
174 (30.1%) 145 (25.1%) 69 (11.9%) 37 (6.4%) 36 (6.2%) 33 (5.7%) 29 (5.0%) 26 (4.5%) 29 (5.0%) 30 (20, 150)
134 (31.4%) 105 (24.6%) 49 (11.5%) 27 (6.3%) 29 (6.8%) 23 (5.4%) 19 (4.4%) 21 (4.9%) 20 (4.7%) 30 (20, 150)
23 (11.6%) 138 (69.7%) 20 (10.1%) 5 (2.5%) 12 (6.1%)
15 (9.8%) 110 (71.9%) 17 (11.1%) 3 (2.0%) 8 (5.2%)
2 (0.9%) 32 (15.2%) 19 (9.0%) 121 (57.3%) 29 (13.7%) 8 (3.8%)
2 (1.3%) 23 (14.6%) 15 (9.6%) 88 (56.1%) 25 (15.9%) 4 (2.5%)
22 (16.7%) 89 (67.4%) 21 (15.9%)
16 (17.4%) 64 (69.6%) 12 (13.0%)
340 (58.8%) 271 (46.9%) 227 (39.3%) 121 (20.9%) 75 (13.0%) 2 (0.3%) 550 (95.8%) 33 (5.7%)
246 (57.6%) 196 (45.9%) 165 (38.6%) 95 (22.2%) 54 (12.6%) 2 (0.5%) 409 (96.0%) 28 (6.6%)
137 (23.7%) 130 (22.5%) 102 (17.6%) 39 (6.7%) 58 (10.0%) 99 (17.1%)
106 (24.8%) 94 (22.0%) 75 (17.6%) 30 (7.0%) 43 (10.1%) 72 (16.9%)
293 (50.9%) 283 (49.1%)
222 (52.2%) 203 (47.8%)
TE D
AC C
EP
Endoscopist 1 2 3 Colon site Ascending Cecum Transverse Sigmoid Rectum Hepatic flexure Descending Ileocecal valve Other Polyp size (mm) Paris classification 1s 2a 2b 2c 2a/c Kudo classification I II III s III l IV V Sano classification I II III Injection type Indigo carmine Epinephrine Saline solution Gonak HPMC Other Lift sign Cap-assisted EMR snare methods Standard Braided Isnare Mini Prophylactic APC Prophylactic Clips Specimen En bloc Piecemeal
All patients (N=578)
SC
Variable
ACCEPTED MANUSCRIPT - 23 -
AC C
EP
TE D
M AN U
SC
RI PT
Optical methods Optical white light 526 (91.0%) 388 (90.9%) Optical NBI 143 (24.7%) 105 (24.6%) Supplemental methods Ablation-APC 241 (41.7%) 180 (42.2%) EMR-snare 31 (5.4%) 25 (5.9%) Grasper assist 22 (3.8%) 13 (3.0%) Snare 20 (3.5%) 12 (2.8%) EMR-cap 17 (2.9%) 14 (3.3%) Ablation-cautery 11 (1.9%) 8 (1.9%) Endo loops 8 (1.4%) 8 (1.9%) Other 108 (18.7%) 77 (18.0%) Continuous variables are summarized using the sample median (range). Information was unavailable regarding Paris classification (N=380 overall, N=274 in follow-up group), Kudo classification (N=367 overall, N=270 in follow-up grop), Sano classification (N=446 overall, N=335 in follow-up group), lift sign (N=4 overall, N=1 in follow-up group), and specimen (N=2 overall, N=2 in follow-up group). APC: argon plasma coagulation; EMR: endoscopic mucosal resection; HPMC: hydroxypropyl methylcellulose; NBI: narrow-band imaging There was no significant difference for any variable.
ACCEPTED MANUSCRIPT - 24 -
Table 3: Patient and EMR outcomes
AC C
EP
TE D
M AN U
SC
RI PT
Variable Summary Endoscopist assessment of incompletion after EMR 159/577 (27.6%) Endoscopist assessment of incompletion after supplemental methods 11/564 (2.0%) Pathology Tubular adenoma 228/576 (39.6%) Tubulovillous adenoma 163/576 (28.3%) Sessile serrated adenoma 59/576 (10.2%) Hyperplastic polyp 51/576 (8.9%) High-grade dysplasia 35/576 (6.1%) Adenocarcinoma 19/576 (3.3%) Other 21/576 (3.7%) Immediate adverse events Bleeding 40/578 (6.9%) Post-polypectomy pain 0/578 (0.0%) Confirmed perforation 1/578 (0.2%) Residual neoplasia at follow-up 99/427 (23.2%) Information was unavailable regarding endoscopist assessment of incompletion after EMR (N=1), endoscopist assessment of incompletion after supplemental methods (N=14), and pathology (N=2). Residual neoplasia at followup was assessed only in the 427 patients who had a follow-up visit within 24 months of the original EMR. EMR: endoscopic mucosal resection
ACCEPTED MANUSCRIPT - 25 -
Table 4: Learning curve for presence of residual neoplasia at follow-up
P value for linear trend
RI PT
0.031 (decreasing trend)
0.14
SC
1-20 21-40 41-60 61-80 81-100 101-120 121-140 141-160 161-180 181-200 201-220
0.001 (decreasing trend)
M AN U
EMR number
Fraction (%) with residual neoplasia for the largest lesion during a given EMR Endoscopist no. 1 Endoscopist no. 2 Endoscopist no. 3 Overall (N=199) (N=223) (N=156) (N=578) 5/15 (33.3%) 5/16 (31.3%) 7/17 (41.2%) 17/48 (35%) 3/17 (17.6%) 4/12 (33.3%) 2/15 (13.3%) 9/44 (20%) 2/15 (13.3%) 10/18 (55.6%) 8/16 (50%) 20/49 (41%) 4/13 (30.8%) 5/13 (38.5%) 4/17 (23.5%) 13/43 (30%) 4/9 (44.4%) 5/16 (31.3%) 0/16 (0%) 9/41 (22%) 3/15 (20%) 2/11 (18.2%) 1/18 (5.6%) 6/44 (14%) 3/18 (16.7%) 2/14 (14.3%) 2/16 (12.5%) 7/48 (15%) 0/14 (0%) 3/14 (21.4%) 0/8 (0%) 3/36 (8%) 0/12 (0%) 4/15 (26.7%) N/A 4/27 (15%) 1/11 (9.1%) 5/17 (29.4%) N/A 6/28 (21%) N/A 5/19 (26.3%) N/A 5/19 (26%) 0.0004 (decreasing trend)
AC C
EP
TE D
P value for linear trend – 0.040 (decreasing 0.026 (decreasing adjusting for prior therapy, 0.15 0.12 trend) trend) specimen, and polyp size The denominators for each 20-EMR do not equal 20 because in many instances there was no follow-up information available for a given EMR. N/A is given when there were no EMRs in the given EMR group for a given endoscopist. Endoscopist no. 2 performed a total of 223 EMRs – for better display of results, EMRs 221-223 were included in the 201-220 EMR group for this endoscopist. P-values result from logistic regression models. The model for the overall group combining across the 3 endoscopists was adjusted for endoscopist, and prior therapy, specimen, and polyp size were additionally adjusted for as described. EMR: endoscopic mucosal resection
ACCEPTED MANUSCRIPT - 26 -
Table 5: Learning curve for immediate bleeding adverse events
P value for linear trend
RI PT
0.027 (increasing trend)
0.82
SC
1-20 21-40 41-60 61-80 81-100 101-120 121-140 141-160 161-180 181-200 201-220
M AN U
EMR number
Fraction (%) with immediate bleeding adverse events for the largest lesion during a given EMR Endoscopist no. 1 Endoscopist no. 2 Endoscopist no. 3 Overall (N=199) (N=223) (N=156) (N=578) 0/20 (0%) 1/20 (5%) 2/20 (10%) 3/60 (5%) 0/20 (0%) 2/20 (10%) 2/20 (10%) 4/60 (7%) 3/20 (15%) 1/20 (5%) 2/20 (10%) 6/60 (10%) 1/20 (5%) 2/20 (10%) 1/20 (5%) 4/60 (7%) 0/20 (0%) 0/20 (0%) 1/20 (5%) 1/60 (2%) 2/20 (10%) 0/20 (0%) 0/20 (0%) 2/60 (3%) 3/20 (15%) 0/20 (0%) 0/20 (0%) 3/60 (5%) 3/20 (15%) 2/20 (10%) 2/16 (13%) 7/56 (13%) 4/20 (20%) 0/20 (0%) N/A 4/40 (10%) 2/19 (11%) 2/20 (10%) N/A 4/39 (10%) N/A 2/23 (9%) N/A 2/23 (9%) 0.17
0.27
AC C
EP
TE D
P value for linear trend – 0.055 0.007 (increasing adjusting for prior therapy, 0.58 0.62 (increasing trend) specimen, and polyp size trend) N/A is given when there were no EMRs in the given EMR group for a given endoscopist. Endoscopist no. 2 performed a total of 223 EMRs – for better display of results, EMRs 221-223 were included in the 201-220 EMR group for this endoscopist. P-values result from logistic regression models. The model for the overall group combining across the 3 endoscopists was adjusted for endoscopist, and prior therapy, specimen, and polyp size were additionally adjusted for as described. EMR: endoscopic mucosal resection
ACCEPTED MANUSCRIPT - 27 -
Table 6: Learning curve for endoscopist assessment of incomplete EMR
RI PT
1-20 21-40 41-60 61-80 81-100 101-120 121-140 141-160 161-180 181-200 201-220
SC
EMR number
Fraction (%) with endoscopist-assessed incomplete EMR for the largest lesion during a given EMR Endoscopist no. 1 Endoscopist no. 2 Endoscopist no. 3 Overall (N=199) (N=223) (N=156) (N=578) 7/20 (35%) 5/20 (25%) 3/20 (15%) 15/60 (25%) 12/20 (60%) 15/20 (75%) 12/20 (60%) 39/60 (65%) 9/20 (45%) 13/20 (65%) 4/20 (20%) 26/60 (43%) 6/20 (30%) 11/19 (58%) 2/20 (10%) 19/59 (32%) 6/20 (30%) 7/20 (35%) 4/20 (20%) 17/60 (28%) 6/20 (30%) 2/20 (10%) 4/20 (20%) 12/60 (20%) 6/20 (30%) 3/20 (15%) 6/20 (30%) 15/60 (25%) 4/20 (20%) 1/20 (5%) 2/16 (13%) 7/56 (13%) 3/20 (15%) 1/20 (5%) N/A 4/40 (10%) 3/19 (16%) 0/20 (0%) N/A 3/39 (8%) N/A 2/23 (9%) N/A 2/23 (9%)
<0.0001 (decreasing trend) P value for linear trend – <0.0001 0.005 (decreasing <0.0001 (decreasing adjusting for prior therapy, 0.15 (decreasing trend) trend) specimen, and polyp size trend) Information regarding endoscopist assessment of incompletion was unavailable for 1 EMR for Endoscopist no. 2 in the 61-80 EMR group. N/A is given when there were no EMRs in the given EMR group for a given endoscopist. Endoscopist no. 2 performed a total of 223 EMRs – for better display of results, EMRs 221-223 were included in the 201-220 EMR group for this endoscopist. P-values result from logistic regression models. The model for the overall group combining across the 3 endoscopists was adjusted for endoscopist, and prior therapy, specimen, and polyp size were additionally adjusted for as described. EMR: endoscopic mucosal resection <0.0001 (decreasing trend)
M AN U
0.001 (decreasing trend)
AC C
EP
TE D
P value for linear trend
0.26
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
Acronyms APC: argon plasma coagulation ASA: American Society of Anesthesiologists
RI PT
EMR: endoscopic mucosal resection ESD: endoscopic submucosal dissection NBI: narrow-band imaging
AC C
EP
TE D
M AN U
SC
OR: odds ratio
S0016-5107(16)30066-9
DOI:
10.1016/j.gie.2016.04.020
Reference:
YMGE 9979
To appear in:
Gastrointestinal Endoscopy
Received Date: 12 January 2016 Accepted Date: 12 April 2016
Please cite this article as: Bhurwal A, Bartel MJ, Heckman MG, Diehl NN, Raimondo M, Wallace MB, Woodward TA, Endoscopic mucosal resection learning curve for large non-polypoid colorectal neoplasia, Gastrointestinal Endoscopy (2016), doi: 10.1016/j.gie.2016.04.020. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Endoscopic mucosal resection learning curve for large non-polypoid colorectal neoplasia
Abhishek Bhurwal, MD1*, Michael J. Bartel, MD1*, Michael G. Heckman, MS2, Nancy N.
MD1
RI PT
Diehl2, Massimo Raimondo, MD1, Michael B. Wallace, MD, MPH1, Timothy A. Woodward,
Division of Gastroenterology, Mayo Clinic, Jacksonville, FL, USA
2
Division of Biomedical Statistics and Informatics, Mayo Clinic, Jacksonville, FL, USA
Corresponding author: Timothy A. Woodward, MD
M AN U
*authors should be considered co-first authors
SC
1
4500 San Pablo Road Jacksonville, FL 32224
EP
+1 (904) 953 2221
TE D
Mayo Clinic, Division of Gastroenterology
AC C
[email protected]
Author Contributions: Dr. Bhurwal contributed to the collection of data, generation of figures, and critical revision of the article. Dr. Bartel contributed to the analysis and interpretation of the data, generation of the figures, and drafting and critical revision of the article. Mr. Heckman and Ms. Diehl contributed to the analysis and interpretation of the data, generation of the figures, and critical revision of the article. Drs. Raimondo and Wallace contributed to the experiments and critical revision of the article. Dr. Woodward contributed to the conception and design of the
ACCEPTED MANUSCRIPT
study, experiments, analysis and interpretation of the data, and critical revision of the article. All authors approved the final draft of the article.
AC C
EP
TE D
M AN U
SC
RI PT
Keywords: Endoscopic mucosal resection; learning curve; colorectal neoplasia
ACCEPTED MANUSCRIPT -1-
Endoscopic mucosal resection learning curve for large non-polypoid colorectal neoplasia
ABSTRACT
RI PT
Background and Aims: Colorectal endoscopic mucosal resection (EMR) for non-polypoid neoplasia achieves better outcomes when performed by expert endoscopists. The time point when the endoscopist achieves expert level remains to be defined. The objective of this study
tissue on surveillance colonoscopy and adverse event rate.
SC
was to establish a learning curve of colorectal EMR for non-polypoid neoplasia based on residual
M AN U
Methods: Five hundred seventy-eight consecutive patients underwent EMR of colorectal neoplasia by 1 of 3 primary endoscopists between December 2004 and September 2013 in a tertiary academic center. Primary analyses focused on the largest lesion for patients with more than one lesion (median age 69 years, median polyp size 30 mm, 51% en bloc resection). Data on
TE D
surveillance colonoscopy were available in 74%. Learning curves were calculated for each of the 3 main outcome measurements: the presence of residual neoplasia on surveillance colonoscopy, endoscopic assessment of incomplete EMR, and the occurrence of an immediate bleeding
EP
adverse event.
Results: Residual neoplasia on surveillance colonoscopy was present for 23.2% of patients, the
AC C
rate of endoscopist-assessed incomplete EMR was 27.6%, and immediate bleeding adverse events occurred in 6.9% of patients. Although there was between-endoscopist variability, the overall rates of residual neoplasia and incomplete EMR decreased to below 20% to 25% after 100 EMRs; initial declines in both rates were observed for earlier EMRs. Immediate bleeding adverse events occurred at a low frequency for each endoscopists across all EMRs. Perforation requiring surgical intervention occurred in 1 patient (0.2%).
ACCEPTED MANUSCRIPT -2-
Conclusions: This study demonstrated that an unexpectedly high number of 100 colorectal EMR procedures for large non-polypoid colorectal neoplasia are required to achieve a plateau phase
RI PT
for crucial outcomes.
INTRODUCTION
Endoscopic resection of large (≥20 mm) colorectal non-polypoid neoplasia is commonly
SC
performed using the endoscopic mucosal resection (EMR) technique. The reported data on shortand long-term outcomes demonstrate substantial differences in the en bloc resection and residual
M AN U
neoplasia rate.1-4 These differences can be partly explained by included polyp sizes, polyp sites, and polyp histology. However, whether different endoscopist EMR skills contribute to these differences is not known.
Despite the broad application of colorectal EMR for neoplasia, surprisingly few data exist
TE D
on the required procedure volume per endoscopist to achieve satisfactory outcomes.5,6 In the recent past, attention was drawn to endoscopic competency and quality measures in order to provide optimal patient care. In this context, learning curves can be established to define
outcomes.7-11
EP
parameters when a sufficient procedure skill set is acquired to allow satisfactory procedure
AC C
Our primary goal was to define a learning curve for EMR of large colorectal non-
polypoid neoplasia, focusing on residual neoplasia at follow-up, endoscopist assessment of incomplete EMR, and immediate bleeding adverse events as quality-defining outcomes.
METHODS Patients
ACCEPTED MANUSCRIPT -3-
We included all patients who underwent EMR for sessile colorectal polyps, 20 mm or larger, from December 2004 to September 2013 at our academic endoscopy unit in this retrospective study. Institutional Review Board approval was obtained. All EMR procedures were performed
RI PT
by the advanced endoscopy group (M.R., M.B.W., and T.A.W.) at our institute since December 2004, with all endoscopists being more than 5 years out of formal gastroenterology training. Before this date, none of the physicians had performed appreciable numbers of EMR procedures.
SC
This resulted in a sample of 578 colorectal EMRs in 578 different patients; with all EMRs
n=223, Endoscopist no. 3: n=156).
M AN U
performed by the 3 aforementioned endoscopists (Endoscopist no. 1: n=199, Endoscopist no. 2:
A subset of patients (n=66) had more than one lesion for a given colonoscopy with EMR, ie, these patients underwent simultaneously 2 or more EMRs. Our primary analysis focused on the largest lesion for the given colonoscopy. However, in a secondary analysis, we also evaluated
TE D
presence of each of the 3 primary endpoints (residual neoplasia at follow-up, immediate bleeding
EMR.
EMR information
EP
adverse events, and endoscopist assessment of incomplete EMR) for any lesions undergoing
AC C
EMR materials and technique varied depending on the lesion and the preference of the operators. Colorectal lesions were removed by submucosal injection of a large volume of normal saline solution (normal saline solution; indigo carmine, 0.04%, Taylor Pharmaceuticals, Decatur, Ill; with or without 1:10,000 epinephrine) or a solution of hydroxypropyl methylcellulose (hydroxypropyl methylcellulose, normal saline solution, indigo carmine, 1:10,000 epinephrine). Lesions were marked with tattoo (Spot, GI supply, Camp Hill, Pa), per endoscopist discretion, to
ACCEPTED MANUSCRIPT -4-
facilitate site recognition at follow-up. Piecemeal or en bloc resection was depending on lesion size and morphology. At the initial endoscopy, supplemental methods were used per endoscopist discretion to achieve maximal macroscopic polyp resection, which included argon plasma
RI PT
coagulation (APC), EMR-cap, ablation cautery, snare, forceps avulsion, and endoloop. APC was used to treat residual macroscopic neoplasia, but not for prophylactic measures. An endoloop was used to facilitate wound closure after snare resection of particularly large neoplasia.
SC
All patients were scheduled for surveillance colonoscopy at 3 to 6 months. In light of limited compliance with this timeframe, all first surveillance colonoscopies were included in our
M AN U
evaluation and were performed within 24 months after the initial EMR. All follow-up colonoscopies were performed by the advanced endoscopy group.
Data collection and outcomes
TE D
A retrospective chart review was performed to extract the following data: baseline characteristics (age, gender, race, prior EMR or snare resection attempt [prior therapy], American Society of Anesthesiologists [ASA] class), EMR information (endoscopist, colon site, polyp size, Paris
EMR,
EP
classification, Kudo classification, Sano classification, injection type, lift sign, cap-assisted EMR snare methods, prophylactic APC, prophylactic clips, specimen [en bloc vs
AC C
piecemeal], optical methods (narrow-band imaging [NBI]), supplemental methods to remove residual neoplasia after EMR (ablation-APC, EMR-snare, Grasper assist [snare resection and simultaneous grasper assist with dual-channel endoscope], snare, EMR-cap, ablation-cautery, Endo loops), and outcomes (endoscopist assessment of complete resection with EMR, endoscopist assessment of complete resection with EMR using NBI, neoplasia histology according to pathology report, immediate adverse events, residual neoplasia at follow-up).
ACCEPTED MANUSCRIPT -5-
Residual neoplasia at follow-up was defined as presence of macroscopically visible neoplastic tissue at the polypectomy scar which was confirmed histologically in each case. Other quality measures like procedure times which include the withdrawal time, bowel preparation quality and
RI PT
pain score assessments were not available for data analysis. Description of neoplasia, including size and other classifications, was based on real-time endoscopist assessment without the use of additional tools (eg, for exact measurement of neoplasia size). There was a substantial amount of
SC
missing data for ASA class (n=78), Paris classification (n=380), Kudo classification (n=367), and Sano classification (n=446). The 3 primary outcome measures were residual macroscopic
M AN U
neoplasia at follow-up, endoscopist assessment of incomplete EMR, and immediate bleeding adverse events, defined as unexpected hemorrhage during EMR requiring additional treatment for hemostasis (epinephrine, clip, APC, cautery, endoloop). EMR specimen of 19 patients revealed T1 adenocarcinoma, of which 9 were referred to surgery, with the remaining patients (5
Statistical analysis
TE D
T1m and 5 T1sm) undergoing surveillance colonoscopy according to patient’s preference.
EP
Continuous variables were summarized using the sample median and range. Categorical variables were summarized using number and percentage. In order to address potential
AC C
confounding in learning curve evaluation, separately for each endoscopist and in the overall group, we examined changes in prior therapy, piecemeal specimen, and polyp size as endoscopist EMR experience increased using logistic regression models (prior therapy and specimen) and linear regression models (polyp size). Odds ratios (ORs) and 95% confidence intervals were estimated, and this analysis was performed separately for each endoscopist and overall. Linear and logistic regression models involving the overall group were adjusted for endoscopist. Polyp size was considered on the logarithm scale in all analysis due to its skewed distribution.
ACCEPTED MANUSCRIPT -6-
The learning curves of EMR regarding the 3 endpoints were assessed using a combination of descriptive summaries, graphical evaluation, and statistical tests. First, for each endoscopist, we divided EMRs into sequential groups of 20, and estimated the proportion of
RI PT
patients for whom each endpoint was experienced separately in each of these 20-EMR groups. These estimates were then combined across all 3 endoscopists. A similar strategy was followed for graphical display, except that a group size of 40 EMRs was used for easier presentation of
SC
graphical results. To evaluate how the frequency of each of the 3 binary categorical endpoints changed as EMR experience increased, we used logistic regression models. Separately for each
M AN U
endpoint and endoscopist, we included a variable in the model for endoscopist sequential EMR number in order to assess any systematic increases or decreases in the frequency of the given endpoint as EMR experience increased. Similar logistic regression models were used when combining across the 3 endoscopists in order to assess overall trends, with additional model
TE D
adjustment for endoscopist. Additional model adjustment for the potential confounding variables of prior therapy, piecemeal specimen, and polyp size was also considered. P-values of 0.05 or lower were considered as statistically significant. All statistical analysis was performed using
EP
SAS (version 9.2; SAS Institute, Inc., Cary, North Carolina) and R Statistical Software (version
AC C
2.14.0; R Foundation for Statistical Computing, Vienna, Austria).
RESULTS
Patient and EMR characteristics and outcomes A summary of baseline patient characteristics is provided in Table 1, whereas EMR information is shown in Table 2. Of the initial cohort of 578 patients, 427 (74%) had a follow-up visit with residual neoplasia information available; this visit occurred at a median of 4.2 months (range: 1–
ACCEPTED MANUSCRIPT -7-
23.7 months) after the initial EMR. When comparing the 427 patients with follow-up residual neoplasia information available to the 151 patients without it, characteristics were similar with the exception of ASA class (ASA 3-4: 35% vs 52%, p<0.001).
RI PT
Patient and EMR outcomes are displayed in Table 3. Endoscopists assessed EMR to be incomplete in 27.6% of patients, but this was reduced to 2.0% after use of supplemental methods. The most common pathologies were tubular adenoma (40%), tubulovillous adenoma
SC
(28%), sessile serrated adenoma (10%), and hyperplastic polyps (9%). Regarding immediate adverse events, bleeding occurred in 40 patients (7%) and confirmed perforation occurred in only
M AN U
1 patient (0.2%), who required surgical intervention. Only immediate bleeding adverse events were evaluated in the learning curve analysis given the rare or non-existent occurrence of other immediate adverse events. Residual neoplasia at follow-up was present in 99 of the 427 patients
TE D
(23%) for whom follow-up information was available.
Potential impact of prior therapy, piecemeal specimen, and polyp size on learning curve results
EP
Before examining the learning curves for residual neoplasia, immediate bleeding adverse events, and endoscopist assessment of incomplete EMR, we first investigated whether prior therapy at
AC C
the EMR site, specimen resection type (piecemeal or en bloc), or polyp size have the potential to impact learning curve results. Supplemental Tables and Figures 1, 2, and 3 display prior therapy, specimen, and polyp size according to endoscopist EMR experience, separately for each endoscopist, and overall. As endoscopist EMR experience increased, there was evidence of more frequent prior therapy for endoscopists no. 2 (p=0.006) and no. 3 (p=0.032), and less frequent piecemeal resection for all 3 endoscopists (all p≤0.022). Interestingly, polyp size did not
ACCEPTED MANUSCRIPT -8-
systematically change for Endoscopist no. 1 and no. 2; however, it decreased for Endoscopist no. 3 with gained experience (p = 0.0004). To better understand the impact that these 3 variables may have on learning curve results,
RI PT
we next evaluated the association between each variable and each learning curve outcome, and these results are shown in Supplemental Table 4. Overall, none of the 3 outcomes were associated with prior therapy at EMR site (all p ≥ 0.16). There was an association of piecemeal
SC
specimen with an increased risk of residual neoplasia at follow-up (OR, 3.19; p < 0.0001) and endoscopist assessment of incomplete EMR (OR, 3.84; p < 0.0001), and an association of polyp
M AN U
size with an increased risk of residual neoplasia at follow-up (OR, 2.02; p = 0.001), immediate bleeding adverse events (OR, 1.98; p = 0.011), and endoscopist assessment of incomplete EMR (OR, 1.63; p = 0.003) (Supplemental Table 4).
TE D
Residual neoplasia at follow-up learning curve
The proportion of patients who had residual neoplasia at follow-up is shown according to endoscopist EMR experience in Figure 1 and Table 4. Without accounting for potential
EP
confounding variables, there was evidence of a decrease in the proportion of patients with residual neoplasia at follow-up as EMR experience increased for endoscopists no. 1 (p=0.031)
AC C
and no. 3 (p=0.001), whereas there was no evidence of a systematic trend for Endoscopist no. 2 (p=0.14). For endoscopists no. 1 and no. 3, the proportion of patients with residual neoplasia at follow-up was consistently at or below 20% after EMR 100. For Endoscopist no. 2, though there was a notable decrease in the proportion of patients with residual neoplasia at follow-up during EMRs 100 to 140 (to between 10% and 20%), this was followed by a slight increase for EMRs 141 to 223 (to between 20% and 30%). When combining across the 3 endoscopists, there was
ACCEPTED MANUSCRIPT -9-
strong evidence of an overall decreasing trend (p=0.0004) with the proportion of patients with residual neoplasia being approximately 20% or lower after EMR 100, with the exception of EMRs 201 to 220, which only involved Endoscopist no. 2.
RI PT
In order to examine how prior therapy, piecemeal specimen, and polyp size may have affected trends in residual neoplasia at follow-up as endoscopist EMR experience increased, we adjusted our models for these 3 variables (Table 4). Results were similar for endoscopists no. 1
SC
and no. 2; however, for Endoscopist no. 3, the decreasing trend weakened noticeably, which may be due to the decreasing frequency of piecemeal specimen (Supplemental Tables and Figures 1-
M AN U
3). When examining all lesions for a given EMR, instead of only the largest lesion, results were similar (data not shown).
Immediate bleeding adverse events learning curve
TE D
The frequency of immediate bleeding adverse events according to endoscopist EMR experience is displayed in Figure 2 and Table 5. There was no evidence of a change in the proportion of patients with an immediate bleeding adverse event as EMR experience increased for Endoscopist
EP
no. 2 (p=0.82) and no. 3 (p=0.17), whereas there was actually evidence of a slight increasing trend for Endoscopist no. 1 (p=0.027). When combining across all 3 endoscopists, no significant
AC C
change in the proportion of patients with an immediate bleeding adverse event was observed as EMR experience increased (p=0.27). Despite any increasing or decreasing trends, the frequency of immediate bleeding adverse events was fairly low for all sequential 20-EMR groups. Results were similar when adjusting for prior therapy, piecemeal specimen, and polyp size (Table 5) and also when examining all lesions for a given EMR instead of only the largest lesion (data not shown).
ACCEPTED MANUSCRIPT - 10 -
Endoscopist assessment of incompletion learning curve Figure 3 and Table 6 display the proportion of EMRs that were assessed as incomplete by the
RI PT
given endoscopist according to endoscopist EMR experience. For both Endoscopist no. 1 and no. 2, there was evidence of a decreasing frequency of incomplete EMR as EMR experience increased (p=0.001 and <0.0001, respectively). For Endoscopist no. 1, the frequency of
SC
endoscopist-assessed incomplete EMR was consistently at or below 20% after EMR 140, whereas for Endoscopist no. 2 this frequency was below 20% after EMR 100. For Endoscopist
M AN U
no. 3, there was no evidence of a systematic increase or decrease in endoscopist-assessed incomplete EMR as EMR experience increased (p=0.26), with higher frequencies occurring in both earlier and later EMRs. When combining information for the 3 endoscopists, there was strong evidence of a decreasing trend in the frequency of endoscopist assessment of incomplete
TE D
EMR as EMR experience increased (p<0.0001); this frequency was approximately 20% to 25% or lower after EMR 100 in the combined group. These results were consistent when evaluating all lesions for a given EMR instead of only the largest lesion (data not shown) and when
AC C
DISCUSSION
EP
adjusting for prior therapy, specimen, and polyp size (Table 6).
This study addresses the learning curve for EMR of large non-polypoid colorectal neoplasia. The learning curve plateaued at 100 EMRs, reflecting a 20% or lower rate of residual neoplasia at follow-up, though there was inter-endoscopist variation for the endpoints. However, immediate bleeding adverse events occurred at a low and relatively stable frequency as EMR experience increased for each endoscopist, arguing that technical progress played no important role for immediate bleeding. Also, when controlling for the most likely confounders, including polyp
ACCEPTED MANUSCRIPT - 11 -
size, site, histology, and rate of piecemeal resection, the EMR number to achieve a learning curve plateau remained the same.
events rate (6.1%) are comparable with previous studies.2, 12-14
RI PT
The outcomes of EMR in this study regarding residual neoplasia (22%) and adverse
At least 90% of patients with large colorectal neoplasia can undergo colorectal EMR, and therefore, avoid surgical resection, which has a higher predicted mortality rate than EMR.1, 15
SC
Despite the variety of EMR outcomes (piecemeal resection rate 7-100%, residual or recurrent neoplasia rate at follow-up 0-55%1, 16), and the well-delineated predictors, which mainly focus on
achieve high EMR quality.17, 18
M AN U
polyp size, site, and histology, surprisingly few data are available on endoscopist competency to
It is evident that EMR in expert hands has better outcomes as implicated by Buchner et al12 and Swan et al19 who addressed EMR outcomes for patients particularly referred for EMR,
TE D
yielding en bloc resection rates of 34% to 53.5% and recurrent or residual neoplasia at follow-up of 10.5% to 27%. However, questions remain to be answered; when does the endoscopist reach a high EMR quality, and how is EMR quality defined? In this study, we defined our primary
EP
endpoints residual or recurrent neoplasia, complete macroscopic neoplasia resection during EMR, and early rebleeding as quality indicators.
AC C
A common presumption that procedure volume is associated with better outcome was
also shown by a British study. Endoscopic specialists had less residual polyp at follow-up than non-specialist.6 Also, Baxter et al20 demonstrated a lower interval colon cancer rate when patients underwent colonoscopy performed by an endoscopist with a high completion rate; however, the endoscopist’s procedure volume was not, per se, associated with interval colon cancer. Instruments to define endoscopist quality are well known and range from ADR to
ACCEPTED MANUSCRIPT - 12 -
competency assessments questionnaires like the Direct Observation of Polypectomy Skills (DOPyS) questionnaire; however, none of them can be used to define EMR quality, particularly as EMR key steps are not a part of the questionnaire, like choice of injection solution or
RI PT
particular snare choice.21, 22
To resolve this dilemma, important lessons can be learned from endoscopic submucosal dissection (ESD).7 The initial ESD studies report self-taught acquisition of skills, which reflects
SC
the learning method in our study. Hotta et al23 reported 40 ESDs to minimize perforations and 80 ESDs to master a maximum en bloc (97.5%) and R0 (92.5%) resection rate. Choi et al11 and
M AN U
Probst et al24 reported comparable numbers to perform high quality gastric and colorectal ESD, respectively. Lower ESD procedure volumes to achieve plateau phase were demonstrated when an ESD novice was supervised by ESD experts, achieving high quality gastric and colorectal ESDs after 30 supervised procedures.8, 9
TE D
EMR is a less complex procedure than ESD; however, our number of 100 EMRs to achieve a plateau phase is higher than expected. Based on our results, the plateau phase was followed by a phase with increased residual tissue and increased immediate bleeding rate.
EP
Although no significant change of residual neoplasia was found when adjusting for confounders, there still remained a down trending rate. We suspected that the higher rate of polypectomies for
AC C
polyps with prior therapy by 2 of the 3 endoscopists (p=0.006 and p=0.032) contributed to the increased rate of residual neoplasia after the plateau phase (Supplemental Tables and Figures 13). The evidence of prior therapy leading to increase in residual rates emphasizes that large polyps should be handled by experts at high-volume centers and that therapy should not be attempted until it is done by definitive resection.
ACCEPTED MANUSCRIPT - 13 -
We further speculate that the relatively high number of EMRs to achieve expertise is based on the self-teaching nature of this study, which is also reflected by including larger neoplastic lesions and lesions which previously failed to be resected in other facilities over the
RI PT
study period. Most instruments used for EMR were available throughout the study period, which argues that the endoscopists, not the new equipment, were measured. Extrapolating from other endoscopic and surgical techniques, and in order to minimize patient risk and maximize EMR
SC
outcomes for future EMR endoscopist generations, we believe that self-taught learning methods, as demonstrated in this study, should be avoided. We believe that specialty centers with well-
M AN U
established standards of care should provide training for EMR with direct expert supervision, as already offered in national gastroenterology society workshops. In this context, animal models are important as first training steps.25, 26
This study is limited by several aspects. First, a retrospective study performed in a
TE D
tertiary center raises concerns of a referral bias for particularly complicated EMRs, which consequently may have increased the EMR number to achieve a plateau phase in the learning curve. Missing or not reported data, which are reflected in the absence of a standardized post
EP
procedure recording and missing photo documentation, might have led to an underestimation of the prevalence of the main study outcomes, such as residual neoplasia, and in particular, adverse
AC C
events. For the same reason, scoring systems addressing the resection difficulty of a polyp could have not been used. However, such possible systematic errors would have been present over the entire study period and for all 3 endoscopists. Consequently, it would have had less effect on calculation of the plateau phase of the learning curve, which was our main goal. Second, this study included patients undergoing EMR over a time period of 9 years with a broad variety of neoplasia ranging from 20 to 150 mm. One can argue that techniques have
ACCEPTED MANUSCRIPT - 14 -
changed over 9 years which biased the learning curve. However, the endoscopic tools and instruments like snares and injection solutions were available over the entire study period. We did not assess changes in the utility of these methods during the study period, but it is very likely
RI PT
that, over 9 years, the endoscopists adjusted their techniques according to the variety of neoplastic lesions. From our perspective, this is the actual learning process, which includes subjective assessment of each neoplasm in order to choose the optimal EMR technique.
SC
Third, we cannot exclude additional selection bias due to patient drop out, although most patients were followed up (74%). In addition, the follow-up intervals varied widely between 3
M AN U
and 24 months, with most follow-up colonoscopies within 3 to 6 months, which is consistent with the multisociety guidelines. It is possible that a few patients had late residual neoplasia presenting after 1 year that could have been missed in our cohort. According to Khashab et al13 who reported late residual in 4.4% of all residual neoplasia cases, we suspect that missing late
TE D
residual neoplasia was of less importance.
Last, unexperienced endoscopists and trainees were not included. In what way the inclusion of trainees under expert supervision would have changed the results remains to be
EP
determined.
This study demonstrates an EMR learning curve for large non-polypoid colorectal
AC C
neoplasia. Our data indicate an unexpectedly high number (100 EMRs) to achieve a plateau phase of the learning curve, which reflects the complete polyp resection rate, rate of residual or recurrent neoplasia at resection site, and immediate adverse events. We recommend that training programs and credentialing bodies consider both volume and residual neoplasia as parameters for certification. Last, multicenter studies are necessary to determine what thresholds for residual neoplasia amongst other quality indicators should be used to determine proficiency in EMR.
ACCEPTED MANUSCRIPT - 15 -
REFERENCES 1.
Kaltenbach T, Friedland S, Maheshwari A, Ouyang D, Rouse RV, Wren S, et al. Short-
RI PT
and long-term outcomes of standardized EMR of nonpolypoid (flat and depressed) colorectal lesions > or = 1 cm (with video). Gastrointest Endosc. 2007 May;65:857-65. 2.
Belle S, Haase L, Pilz LR, Post S, Ebert M, Kaehler G. Recurrence after endoscopic
3.
SC
mucosal resection-therapy failure? Int J Colorectal Dis. 2014 Feb;29:209-15.
Su MY, Hsu CM, Ho YP, Lien JM, Lin CJ, Chiu CT, et al. Endoscopic mucosal resection
4.
M AN U
for colonic non-polypoid neoplasms. Am J Gastroenterol. 2005 Oct;100:2174-9. Wang AY, Ahmad NA, Zaidman JS, Brensinger CM, Lewis JD, Long WB, et al.
Endoluminal resection for sessile neoplasia in the GI tract is associated with a low recurrence rate and a high 5-year survival rate. Gastrointest Endosc. 2008 Jul;68:160-9. Heresbach D, Kornhauser R, Seyrig JA, Coumaros D, Claviere C, Bury A, et al. A
TE D
5.
national survey of endoscopic mucosal resection for superficial gastrointestinal neoplasia. Endoscopy. 2010 Oct;42:806-13.
Brooker JC, Saunders BP, Shah SG, Williams CB. Endoscopic resection of large sessile
EP
6.
colonic polyps by specialist and non-specialist endoscopists. Br J Surg. 2002 Aug;89:1020-4. Uraoka T, Parra-Blanco A, Yahagi N. Colorectal endoscopic submucosal dissection in
AC C
7.
Japan and Western countries. Dig Endosc. 2012 May;24 Suppl 1:80-3. 8.
Yamamoto S, Uedo N, Ishihara R, Kajimoto N, Ogiyama H, Fukushima Y, et al.
Endoscopic submucosal dissection for early gastric cancer performed by supervised residents: assessment of feasibility and learning curve. Endoscopy. 2009 Nov;41:923-8.
ACCEPTED MANUSCRIPT - 16 -
9.
Sakamoto T, Saito Y, Fukunaga S, Nakajima T, Matsuda T. Learning curve associated
with colorectal endoscopic submucosal dissection for endoscopists experienced in gastric endoscopic submucosal dissection. Dis Colon Rectum. 2011 Oct;54:1307-12. Iacopini F, Bella A, Costamagna G, Gotoda T, Saito Y, Elisei W, et al. Stepwise training
RI PT
10.
in rectal and colonic endoscopic submucosal dissection with differentiated learning curves. Gastrointest Endosc. 2012 Dec;76:1188-96.
Probst A, Golger D, Anthuber M, Markl B, Messmann H. Endoscopic submucosal
SC
11.
Endoscopy. 2012 Jul;44:660-7. 12.
M AN U
dissection in large sessile lesions of the rectosigmoid: learning curve in a European center.
Buchner AM, Guarner-Argente C, Ginsberg GG. Outcomes of EMR of defiant colorectal
lesions directed to an endoscopy referral center. Gastrointest Endosc. 2012 Aug;76:255-63. 13.
Khashab M, Eid E, Rusche M, Rex DK. Incidence and predictors of "late" recurrences
Aug;70:344-9. 14.
TE D
after endoscopic piecemeal resection of large sessile adenomas. Gastrointest Endosc. 2009
Moss A, Bourke MJ, Williams SJ, Hourigan LF, Brown G, Tam W, et al. Endoscopic
EP
mucosal resection outcomes and prediction of submucosal cancer from advanced colonic mucosal neoplasia. Gastroenterology. 2011 Jun;140:1909-18. Ahlenstiel G, Hourigan LF, Brown G, Zanati S, Williams SJ, Singh R, et al. Actual
AC C
15.
endoscopic versus predicted surgical mortality for treatment of advanced mucosal neoplasia of the colon. Gastrointest Endosc. 2014 Oct;80:668-76. 16.
Ah Soune P, Menard C, Salah E, Desjeux A, Grimaud JC, Barthet M. Large endoscopic
mucosal resection for colorectal tumors exceeding 4 cm. World J Gastroenterol. 2010 Feb 7;16:588-95.
ACCEPTED MANUSCRIPT - 17 -
17.
Woodward TA, Heckman MG, Cleveland P, De Melo S, Raimondo M, Wallace M.
Predictors of complete endoscopic mucosal resection of flat and depressed gastrointestinal neoplasia of the colon. Am J Gastroenterol. 2012 May;107:650-4. Kim HH, Kim JH, Park SJ, Park MI, Moon W. Risk factors for incomplete resection and
RI PT
18.
complications in endoscopic mucosal resection for lateral spreading tumors. Dig Endosc. 2012 Jul;24:259-66.
Swan MP, Bourke MJ, Alexander S, Moss A, Williams SJ. Large refractory colonic
SC
19.
polyps: is it time to change our practice? A prospective study of the clinical and economic
M AN U
impact of a tertiary referral colonic mucosal resection and polypectomy service (with videos). Gastrointest Endosc. 2009 Dec;70:1128-36. 20.
Baxter NN, Sutradhar R, Forbes SS, Paszat LF, Saskin R, Rabeneck L. Analysis of
administrative data finds endoscopist quality measures associated with postcolonoscopy
21.
TE D
colorectal cancer. Gastroenterology. 2011 Jan;140:65-72.
Gupta S, Bassett P, Man R, Suzuki N, Vance ME, Thomas-Gibson S. Validation of a
novel method for assessing competency in polypectomy. Gastrointest Endosc. 2012 Mar;75:568-
22.
EP
75.
Kaminski MF, Regula J, Kraszewska E, Polkowski M, Wojciechowska U, Didkowska J,
AC C
et al. Quality indicators for colonoscopy and the risk of interval cancer. N Engl J Med. 2010;362:1795-803. 23.
Hotta K, Oyama T, Shinohara T, Miyata Y, Takahashi A, Kitamura Y, et al. Learning
curve for endoscopic submucosal dissection of large colorectal tumors. Dig Endosc. 2010 Oct;22:302-6.
ACCEPTED MANUSCRIPT - 18 -
24.
Choi IJ, Kim CG, Chang HJ, Kim SG, Kook MC, Bae JM. The learning curve for EMR
with circumferential mucosal incision in treating intramucosal gastric neoplasm. Gastrointest Endosc. 2005 Dec;62:860-5. Parra-Blanco A, Arnau MR, Nicolas-Perez D, Gimeno-Garcia AZ, Gonzalez N, Diaz-
RI PT
25.
Acosta JA, et al. Endoscopic submucosal dissection training with pig models in a Western country. World J Gastroenterol. 2010 Jun 21;16:2895-900.
Herreros de Tejada A. ESD training: A challenging path to excellence. World J
M AN U
Gastrointest Endosc. 2014 Apr 16;6:112-20.
SC
26.
FIGURE LEGENDS
Figure 1: Frequency of residual neoplasia as endoscopic mucosal resection (EMR) experience increases for each of the 3 endoscopists and overall. Separately for each endoscopist, EMRs were
TE D
divided into sequential groups of 40, and the proportion of patients with residual neoplasia was estimated for each 40-EMR group. These endoscopist-specific estimates were then combined into one overall estimate across all 3 endoscopists; an overall estimate was not calculated for
EP
EMRs greater than 200 because only Endoscopist no. 2 performed more than 200 EMRs.
AC C
Figure 2: Frequency of immediate bleeding adverse events as endoscopic mucosal resection (EMR) experience increases for each of the 3 endoscopists and overall. Separately for each endoscopist, EMRs were divided into sequential groups of 40, and the proportion of patients with immediate bleeding adverse events was estimated for each 40-EMR group. These endoscopistspecific estimates were then combined into one overall estimate across all 3 endoscopists; an
ACCEPTED MANUSCRIPT - 19 -
overall estimate was not calculated for EMRs greater than 200 because only Endoscopist no. 2 performed more than 200 EMRs.
RI PT
Figure 3: Frequency of endoscopic mucosal resections (EMR) assessed as incomplete by
endoscopist as EMR experience increases for each of the 3 endoscopists and overall. Separately for each endoscopist, EMRs were divided into sequential groups of 40, and the proportion of
SC
patients with an incomplete assessment after EMR was estimated for each 40-EMR group. These endoscopist-specific estimates were then combined into one overall estimate across all 3
M AN U
endoscopists; an overall estimate was not calculated for EMRs greater than 200 because only Endoscopist no. 2 performed more than 200 EMRs.
SUPPLEMENTAL FIGURE LEGEND
TE D
Supplemental Figure 1: Frequency of prior therapy as endoscopic mucosal resection (EMR) experience increases for each of the 3 endoscopists and overall. Separately for each endoscopist, EMRs were divided into sequential groups of 40, and the proportion of patients with prior
EP
therapy was estimated for each 40-EMR group. These endoscopist-specific estimates were then combined into one overall estimate across all 3 endoscopists; an overall estimate was not
AC C
calculated for EMRs greater than 200 because only Endoscopist no. 2 performed more than 200 EMRs.
Supplemental Figure 2: Frequency of piecemeal specimen as endoscopic mucosal resection (EMR) experience increases for each of the 3 endoscopists and overall. Separately for each endoscopist, EMRs were divided into sequential groups of 40, and the proportion of patients with
ACCEPTED MANUSCRIPT - 20 -
piecemeal specimen was estimated for each 40-EMR group. These endoscopist-specific estimates were then combined into one overall estimate across all 3 endoscopists; an overall estimate was not calculated for EMRs greater than 200 because only Endoscopist no. 2
RI PT
performed more than 200 EMRs.
Supplemental Figure 3: Median polyp size as endoscopic mucosal resection (EMR) experience
SC
increases for each of the 3 endoscopists and overall. Separately for each endoscopist, EMRs were divided into sequential groups of 40, and median polyp size was estimated for each 40-EMR
M AN U
group. These endoscopist-specific estimates were then combined into one overall estimate across all 3 endoscopists; an overall estimate was not calculated for EMRs greater than 200 because
AC C
EP
TE D
only Endoscopist no. 2 performed more than 200 EMRs.
ACCEPTED MANUSCRIPT - 21 -
TABLES Table 1: Patient information Variable
All patients (N=578)
Patients with follow-up within 24 months of original EMR (N=427) 68.2 (22.0, 95.1) 232 (54.3%) 398 (93.6%) 65 (15.3%)
AC C
EP
TE D
M AN U
SC
RI PT
Age at EMR 68.6 (22.0, 95.1) Gender (Male) 306 (52.9%) Race (Caucasian) 539 (94.1%) Prior therapy 85 (14.8%) ASA class 1 13 (2.6%) 12 (3.3%) 2 291 (58.2%) 228 (62.1%) 3 190 (38.0%) 122 (33.2%) 4 6 (1.2%) 5 (1.4%) Continuous variables are summarized using the sample median (range). Information was unavailable regarding race (N=5 overall, N=2 in follow-up group), prior therapy (N=2 overall, N=1 in follow-up group), and ASA class (N=78 overall, N=60 in follow-up group). ASA: American Society of Anesthesiologists; EMR: endoscopic mucosal resection There was no significant difference for any variable.
ACCEPTED MANUSCRIPT - 22 -
Table 2: EMR information
199 (34.4%) 223 (38.6%) 156 (27.0%)
139 (32.6%) 165 (38.6%) 123 (28.8%)
RI PT
Patients with follow-up within 24 months of original EMR (N=427)
M AN U
174 (30.1%) 145 (25.1%) 69 (11.9%) 37 (6.4%) 36 (6.2%) 33 (5.7%) 29 (5.0%) 26 (4.5%) 29 (5.0%) 30 (20, 150)
134 (31.4%) 105 (24.6%) 49 (11.5%) 27 (6.3%) 29 (6.8%) 23 (5.4%) 19 (4.4%) 21 (4.9%) 20 (4.7%) 30 (20, 150)
23 (11.6%) 138 (69.7%) 20 (10.1%) 5 (2.5%) 12 (6.1%)
15 (9.8%) 110 (71.9%) 17 (11.1%) 3 (2.0%) 8 (5.2%)
2 (0.9%) 32 (15.2%) 19 (9.0%) 121 (57.3%) 29 (13.7%) 8 (3.8%)
2 (1.3%) 23 (14.6%) 15 (9.6%) 88 (56.1%) 25 (15.9%) 4 (2.5%)
22 (16.7%) 89 (67.4%) 21 (15.9%)
16 (17.4%) 64 (69.6%) 12 (13.0%)
340 (58.8%) 271 (46.9%) 227 (39.3%) 121 (20.9%) 75 (13.0%) 2 (0.3%) 550 (95.8%) 33 (5.7%)
246 (57.6%) 196 (45.9%) 165 (38.6%) 95 (22.2%) 54 (12.6%) 2 (0.5%) 409 (96.0%) 28 (6.6%)
137 (23.7%) 130 (22.5%) 102 (17.6%) 39 (6.7%) 58 (10.0%) 99 (17.1%)
106 (24.8%) 94 (22.0%) 75 (17.6%) 30 (7.0%) 43 (10.1%) 72 (16.9%)
293 (50.9%) 283 (49.1%)
222 (52.2%) 203 (47.8%)
TE D
AC C
EP
Endoscopist 1 2 3 Colon site Ascending Cecum Transverse Sigmoid Rectum Hepatic flexure Descending Ileocecal valve Other Polyp size (mm) Paris classification 1s 2a 2b 2c 2a/c Kudo classification I II III s III l IV V Sano classification I II III Injection type Indigo carmine Epinephrine Saline solution Gonak HPMC Other Lift sign Cap-assisted EMR snare methods Standard Braided Isnare Mini Prophylactic APC Prophylactic Clips Specimen En bloc Piecemeal
All patients (N=578)
SC
Variable
ACCEPTED MANUSCRIPT - 23 -
AC C
EP
TE D
M AN U
SC
RI PT
Optical methods Optical white light 526 (91.0%) 388 (90.9%) Optical NBI 143 (24.7%) 105 (24.6%) Supplemental methods Ablation-APC 241 (41.7%) 180 (42.2%) EMR-snare 31 (5.4%) 25 (5.9%) Grasper assist 22 (3.8%) 13 (3.0%) Snare 20 (3.5%) 12 (2.8%) EMR-cap 17 (2.9%) 14 (3.3%) Ablation-cautery 11 (1.9%) 8 (1.9%) Endo loops 8 (1.4%) 8 (1.9%) Other 108 (18.7%) 77 (18.0%) Continuous variables are summarized using the sample median (range). Information was unavailable regarding Paris classification (N=380 overall, N=274 in follow-up group), Kudo classification (N=367 overall, N=270 in follow-up grop), Sano classification (N=446 overall, N=335 in follow-up group), lift sign (N=4 overall, N=1 in follow-up group), and specimen (N=2 overall, N=2 in follow-up group). APC: argon plasma coagulation; EMR: endoscopic mucosal resection; HPMC: hydroxypropyl methylcellulose; NBI: narrow-band imaging There was no significant difference for any variable.
ACCEPTED MANUSCRIPT - 24 -
Table 3: Patient and EMR outcomes
AC C
EP
TE D
M AN U
SC
RI PT
Variable Summary Endoscopist assessment of incompletion after EMR 159/577 (27.6%) Endoscopist assessment of incompletion after supplemental methods 11/564 (2.0%) Pathology Tubular adenoma 228/576 (39.6%) Tubulovillous adenoma 163/576 (28.3%) Sessile serrated adenoma 59/576 (10.2%) Hyperplastic polyp 51/576 (8.9%) High-grade dysplasia 35/576 (6.1%) Adenocarcinoma 19/576 (3.3%) Other 21/576 (3.7%) Immediate adverse events Bleeding 40/578 (6.9%) Post-polypectomy pain 0/578 (0.0%) Confirmed perforation 1/578 (0.2%) Residual neoplasia at follow-up 99/427 (23.2%) Information was unavailable regarding endoscopist assessment of incompletion after EMR (N=1), endoscopist assessment of incompletion after supplemental methods (N=14), and pathology (N=2). Residual neoplasia at followup was assessed only in the 427 patients who had a follow-up visit within 24 months of the original EMR. EMR: endoscopic mucosal resection
ACCEPTED MANUSCRIPT - 25 -
Table 4: Learning curve for presence of residual neoplasia at follow-up
P value for linear trend
RI PT
0.031 (decreasing trend)
0.14
SC
1-20 21-40 41-60 61-80 81-100 101-120 121-140 141-160 161-180 181-200 201-220
0.001 (decreasing trend)
M AN U
EMR number
Fraction (%) with residual neoplasia for the largest lesion during a given EMR Endoscopist no. 1 Endoscopist no. 2 Endoscopist no. 3 Overall (N=199) (N=223) (N=156) (N=578) 5/15 (33.3%) 5/16 (31.3%) 7/17 (41.2%) 17/48 (35%) 3/17 (17.6%) 4/12 (33.3%) 2/15 (13.3%) 9/44 (20%) 2/15 (13.3%) 10/18 (55.6%) 8/16 (50%) 20/49 (41%) 4/13 (30.8%) 5/13 (38.5%) 4/17 (23.5%) 13/43 (30%) 4/9 (44.4%) 5/16 (31.3%) 0/16 (0%) 9/41 (22%) 3/15 (20%) 2/11 (18.2%) 1/18 (5.6%) 6/44 (14%) 3/18 (16.7%) 2/14 (14.3%) 2/16 (12.5%) 7/48 (15%) 0/14 (0%) 3/14 (21.4%) 0/8 (0%) 3/36 (8%) 0/12 (0%) 4/15 (26.7%) N/A 4/27 (15%) 1/11 (9.1%) 5/17 (29.4%) N/A 6/28 (21%) N/A 5/19 (26.3%) N/A 5/19 (26%) 0.0004 (decreasing trend)
AC C
EP
TE D
P value for linear trend – 0.040 (decreasing 0.026 (decreasing adjusting for prior therapy, 0.15 0.12 trend) trend) specimen, and polyp size The denominators for each 20-EMR do not equal 20 because in many instances there was no follow-up information available for a given EMR. N/A is given when there were no EMRs in the given EMR group for a given endoscopist. Endoscopist no. 2 performed a total of 223 EMRs – for better display of results, EMRs 221-223 were included in the 201-220 EMR group for this endoscopist. P-values result from logistic regression models. The model for the overall group combining across the 3 endoscopists was adjusted for endoscopist, and prior therapy, specimen, and polyp size were additionally adjusted for as described. EMR: endoscopic mucosal resection
ACCEPTED MANUSCRIPT - 26 -
Table 5: Learning curve for immediate bleeding adverse events
P value for linear trend
RI PT
0.027 (increasing trend)
0.82
SC
1-20 21-40 41-60 61-80 81-100 101-120 121-140 141-160 161-180 181-200 201-220
M AN U
EMR number
Fraction (%) with immediate bleeding adverse events for the largest lesion during a given EMR Endoscopist no. 1 Endoscopist no. 2 Endoscopist no. 3 Overall (N=199) (N=223) (N=156) (N=578) 0/20 (0%) 1/20 (5%) 2/20 (10%) 3/60 (5%) 0/20 (0%) 2/20 (10%) 2/20 (10%) 4/60 (7%) 3/20 (15%) 1/20 (5%) 2/20 (10%) 6/60 (10%) 1/20 (5%) 2/20 (10%) 1/20 (5%) 4/60 (7%) 0/20 (0%) 0/20 (0%) 1/20 (5%) 1/60 (2%) 2/20 (10%) 0/20 (0%) 0/20 (0%) 2/60 (3%) 3/20 (15%) 0/20 (0%) 0/20 (0%) 3/60 (5%) 3/20 (15%) 2/20 (10%) 2/16 (13%) 7/56 (13%) 4/20 (20%) 0/20 (0%) N/A 4/40 (10%) 2/19 (11%) 2/20 (10%) N/A 4/39 (10%) N/A 2/23 (9%) N/A 2/23 (9%) 0.17
0.27
AC C
EP
TE D
P value for linear trend – 0.055 0.007 (increasing adjusting for prior therapy, 0.58 0.62 (increasing trend) specimen, and polyp size trend) N/A is given when there were no EMRs in the given EMR group for a given endoscopist. Endoscopist no. 2 performed a total of 223 EMRs – for better display of results, EMRs 221-223 were included in the 201-220 EMR group for this endoscopist. P-values result from logistic regression models. The model for the overall group combining across the 3 endoscopists was adjusted for endoscopist, and prior therapy, specimen, and polyp size were additionally adjusted for as described. EMR: endoscopic mucosal resection
ACCEPTED MANUSCRIPT - 27 -
Table 6: Learning curve for endoscopist assessment of incomplete EMR
RI PT
1-20 21-40 41-60 61-80 81-100 101-120 121-140 141-160 161-180 181-200 201-220
SC
EMR number
Fraction (%) with endoscopist-assessed incomplete EMR for the largest lesion during a given EMR Endoscopist no. 1 Endoscopist no. 2 Endoscopist no. 3 Overall (N=199) (N=223) (N=156) (N=578) 7/20 (35%) 5/20 (25%) 3/20 (15%) 15/60 (25%) 12/20 (60%) 15/20 (75%) 12/20 (60%) 39/60 (65%) 9/20 (45%) 13/20 (65%) 4/20 (20%) 26/60 (43%) 6/20 (30%) 11/19 (58%) 2/20 (10%) 19/59 (32%) 6/20 (30%) 7/20 (35%) 4/20 (20%) 17/60 (28%) 6/20 (30%) 2/20 (10%) 4/20 (20%) 12/60 (20%) 6/20 (30%) 3/20 (15%) 6/20 (30%) 15/60 (25%) 4/20 (20%) 1/20 (5%) 2/16 (13%) 7/56 (13%) 3/20 (15%) 1/20 (5%) N/A 4/40 (10%) 3/19 (16%) 0/20 (0%) N/A 3/39 (8%) N/A 2/23 (9%) N/A 2/23 (9%)
<0.0001 (decreasing trend) P value for linear trend – <0.0001 0.005 (decreasing <0.0001 (decreasing adjusting for prior therapy, 0.15 (decreasing trend) trend) specimen, and polyp size trend) Information regarding endoscopist assessment of incompletion was unavailable for 1 EMR for Endoscopist no. 2 in the 61-80 EMR group. N/A is given when there were no EMRs in the given EMR group for a given endoscopist. Endoscopist no. 2 performed a total of 223 EMRs – for better display of results, EMRs 221-223 were included in the 201-220 EMR group for this endoscopist. P-values result from logistic regression models. The model for the overall group combining across the 3 endoscopists was adjusted for endoscopist, and prior therapy, specimen, and polyp size were additionally adjusted for as described. EMR: endoscopic mucosal resection <0.0001 (decreasing trend)
M AN U
0.001 (decreasing trend)
AC C
EP
TE D
P value for linear trend
0.26
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
Acronyms APC: argon plasma coagulation ASA: American Society of Anesthesiologists
RI PT
EMR: endoscopic mucosal resection ESD: endoscopic submucosal dissection NBI: narrow-band imaging
AC C
EP
TE D
M AN U
SC
OR: odds ratio