Article Text

Download PDFPDF

Original article
Endoscopic mucosal resection for large serrated lesions in comparison with adenomas: a prospective multicentre study of 2000 lesions
  1. Maria Pellise1,
  2. Nicholas G Burgess1,2,
  3. Nicholas Tutticci1,
  4. Luke F Hourigan3,4,
  5. Simon A Zanati5,6,
  6. Gregor J Brown5,7,
  7. Rajvinder Singh8,
  8. Stephen J Williams1,
  9. Spiro C Raftopoulos9,
  10. Donald Ormonde9,
  11. Alan Moss6,
  12. Karen Byth10,11,
  13. Heok P'Ng12,
  14. Hema Mahajan12,
  15. Duncan McLeod12,
  16. Michael J Bourke1,2
  1. 1Departments of Gastroenterology and Hepatology, Westmead Hospital, Sydney, New South Wales, Australia
  2. 2Westmead Clinical School, University of Sydney, Sydney, New South Wales, Australia
  3. 3Department of Gastroenterology and Hepatology, Princess Alexandra Hospital, Brisbane, Queensland, Australia
  4. 4Department of Gastroenterology and Hepatology, Greenslopes Private Hospital, Brisbane, Queensland, Australia
  5. 5Department of Gastroenterology and Hepatology, The Alfred Hospital, Melbourne, Victoria, Australia
  6. 6Department of Gastroenterology and Hepatology, Western Hospital, Melbourne, Victoria, Australia
  7. 7Department of Gastroenterology and Hepatology, Epworth Hospital, Melbourne, Victoria, Australia
  8. 8Department of Gastroenterology and Hepatology, Lyell McEwin Hospital, Adelaide, South Australia, Australia
  9. 9Department of Gastroenterology and Hepatology, Sir Charles Gairdner Hospital, Perth, Western Australia, Australia
  10. 10NHMRC Clinical Trials Centre, University of Sydney, Sydney, New South Wales, Australia.
  11. 11Research and Education Network, Westmead Hospital, Sydney, New South Wales, Australia
  12. 12Institute of Clinical Pathology and Medical Research, Westmead Hospital, Sydney, New South Wales, Australia
  1. Correspondence to Professor Michael J Bourke, Department of Gastroenterology and Hepatology, Westmead Hospital, c/- Suite 106a, 151-155 Hawkesbury Road, Westmead, Sydney, NSW 2143, Australia; michael{at}


Objective Endoscopic mucosal resection (EMR) is effective for large laterally spreading flat and sessile lesions (LSLs). Sessile serrated adenomas/polyps (SSA/Ps) are linked to the relative failure of colonoscopy to prevent proximal colorectal cancer. We aimed to examine the technical success, adverse events and recurrence following EMR for large SSA/Ps in comparison with large conventional adenomas.

Design Over 74 months till August 2014, prospective multicentre data of LSLs ≥20 mm were analysed. A standardised dye-based conventional EMR technique followed by scheduled surveillance colonoscopy was used.

Results From a total of 2000 lesions, 323 SSA/Ps in 246 patients and 1527 adenomas in 1425 patients were included for analysis. Technical success for EMR was superior in SSA/Ps compared with adenomas (99.1% vs 94.5%, p<0.001). Significant bleeding and perforation were similar in both cohorts. The cumulative recurrence rates for adenomas after 6, 12, 18 and 24 months were 16.1%, 20.4%, 23.4% and 28.4%, respectively. For SSA/Ps, they were 6.3% at 6 months and 7.0% from 12 months onwards (p<0.001). Following multivariable adjustment, the HR of recurrence for adenomas versus SSA/Ps was 1.7 (95% CI 0.9 to 3.0, p=0.097). Subgroup analysis by lesion size revealed an eightfold increased risk of recurrence for 20–25 mm adenomas versus SSA/Ps, but no significantly different risk between lesion types in larger lesion groups.

Conclusion Recurrence after EMR of 20–25 mm LSLs is significantly less frequent in SSA/Ps compared with adenomatous lesions. SSA/Ps can be more effectively removed than adenomatous LSLs with equivalent safety. Ensuring complete initial resection is imperative for avoiding recurrence.

Trial registration number NCT01368289.

View Full Text

Statistics from

Significance of this study

What is already known on this subject?

  • Wide field endoscopic mucosal resection (EMR) is the first-line therapy for large laterally spreading flat and sessile lesions (LSLs) including sessile serrated adenomas/polyps (SSA/Ps).

  • SSA/Ps are difficult to detect, often incompletely resected and are strongly associated with ‘interval’ colorectal cancer (CRC).

  • Outcomes after endoscopic treatment of large SSA/Ps are not well known.

What are the new findings?

  • In a multicentre study including 2000 LSLs ≥20 mm referred for tertiary endoscopic resection, large SSA/Ps were successfully removed by EMR in almost 100% of cases.

  • Large SSA/Ps were easier to remove, associated with less intraprocedural bleeding and similar rates of significant adverse events compared with adenomatous LSLs.

  • Recurrence after successful complete EMR of large SSA/Ps was low at 7.0% from 12 months onwards, whereas the cumulative recurrence rates for adenomas after 6, 12, 18 and 24 months were 16.1%, 20.4%, 23.4% and 28.4%, respectively.

  • Following multivariable adjustment, the HR of recurrence for adenomas versus SSA/Ps was 1.7 (95% CI 0.9 to 3.0, p=0.097). Subgroup analysis by lesion size revealed an eightfold increased risk of recurrence for 20–25 mm adenomas versus SSA/Ps, but no significantly different risk between lesion types in larger lesion groups. Half the SSA/Ps were 20–25 mm while nearly half the adenomas were >35 mm compared with only 20% of SSA/Ps.


Wide field endoscopic mucosal resection (EMR) for large (≥20 mm) laterally spreading flat and sessile lesions (LSLs) is safe and effective even for very large or complex lesions and is becoming the first-line therapy.1 ,2 However, the technique is not exempt from adverse events—perforation and clinically significant post-EMR bleeding (CSPEB) can occur in up to 1%–4% and 6%–11% of patients, respectively.1–7 In addition, following successful EMR, recurrence or presence of residual adenoma at the first follow-up examination can be present in 14%–32% of cases8–11

Sessile serrated adenomas/polyps (SSA/Ps) are now thought to be the major precursor lesion of serrated pathway cancers, which account for 20%–30% of all sporadic colorectal cancer (CRC) cases.12 ,13 They are likely to be a major contributor to the relative failure of colonoscopy in the prevention of proximal CRC. SSA/Ps are difficult to detect, variably recognised by endoscopists and pathologists alike and often incompletely resected.14–18 In a prospective study where 346 polyps (including 42 SSA/Ps) were examined for completeness of resection using standardised postpolypectomy margin biopsy, the rate of incomplete resection among SSA/Ps overall was 31% versus 7% in conventional adenomas, and almost half of serrated lesions 10–20 mm in size were incompletely resected.15 Incomplete resection results in a higher rate of recurrence during follow-up and may lead to interval cancer.19–22 Specific intrinsic characteristics of serrated lesions in comparison with adenomas include their flat morphology, indistinct edges, smooth surface and proximal location. The frequent absence of dysplasia may result in different EMR outcomes in comparison with conventional adenomas. To date, only two studies have specifically addressed the endoscopic treatment outcomes of large SSA/Ps.23 ,24 Both studies were retrospective and based on a unique centre and unique endoscopist experience. They both included a relatively limited number of SSA/Ps (112 and 46 patients, respectively) and a short follow-up (only early recurrence was assessed). Moreover, in the larger series, SSA/P resection was not performed using EMR technique.23 Albeit these limitations, both studies suggest that complications and recurrence rates after removal of large SSA/P are similar to those for similarly sized adenomas.

The aim of the present study is to determine the rates of technical success, adverse events within 14 days, and recurrence following EMR of large SSA/P compared with large conventional adenoma in a prospective multicentre cohort.



Consecutive patients referred to eight Australian academic hospitals for the management of large sessile and flat colorectal polyps ≥20 mm were enrolled in this prospective observational study (The Australian Colonic Endoscopic (ACE) Resection study; NCT01368289). All lesions had been initially identified and referred by a nationally accredited consultant endoscopist. Data were recorded in a comprehensive centralised database from September 2008 to September 2014. Institutional review board approval was obtained at each centre. Written informed consent was obtained from each patient on the day of the procedure. There were no exclusion criteria. All authors had access to the study data and reviewed and approved the final manuscript.

EMR and data collection

All EMR procedures were performed by a study investigator or a senior therapeutic endoscopy fellow under direct supervision. All clinical investigators were gastroenterologists with significant prior colonic EMR experience after training in high-volume tertiary referral centres in Australia or overseas. Colonoscopy was performed using Olympus 180 or 190 series high-definition variable-stiffness colonoscopes (180/190 PCF/CF; Olympus, Tokyo, Japan). The EMR technique is standard across all centres and has previously been described in detail.1 ,3 ,9 Data were prospectively collected at the time of patient admission, during and immediately after the procedure. Data included patient demographics and comorbidities, American Society of Anesthesiologists grade and medications. Lesions were carefully examined and classified according to Paris classification25 and Kudo's Pit Pattern.26 Variables related to technical issues such as constituent and dye used for lifting, type of snare and current and use of an additional modality for complete resection were also collected. Standardised advice was provided to patients on management of antithrombotic medications.

After EMR, patients remained in recovery for 4–6 h until medically cleared for discharge by the endoscopist. On discharge, dietary instructions were for clear fluids overnight and to resume a normal diet the following day. Written postprocedural instructions were provided including information on potential problems and contact details for advice.

Adverse events

Intraprocedural and postprocedural or delayed adverse events were recorded and short-term clinical follow-up was obtained at 14 days after the procedure by a non-blinded structured telephone interview performed by the research nurse. Where there was doubt about a specific event this was clarified with the investigators. The adverse outcomes for analysis were intraprocedural bleeding (IPB), CSPEB and deep mural injury or perforation. IPB was defined as any IPB that required endoscopic intervention. CSPEB was defined as any bleeding occurring after completion of the procedure and necessitating emergency department presentation, hospitalisation or reintervention (repeat endoscopy, angiography or surgery) within 14 days of EMR. Self-limited bleeding that was managed on an outpatient basis or did not result in the patient presenting for medical assessment was not included. Clinically significant perforation was defined as a mural injury that required long hospitalisation or reintervention.


Patients with successful endoscopic resection and without submucosal invasive cancer in the EMR specimens underwent colonoscopic surveillance at approximately 4–6 months and 16 months after the index EMR. To ensure at least 5 months potential follow-up for each patient, only those whose index EMR that occurred before 19 April 2014 were eligible for inclusion in the study of factors affecting recurrence.

Surveillance procedures were performed by the same endoscopists involved in the study. The exact surveillance time varied between patients for practical and logistic reasons. The EMR scar was interrogated in detail for evidence of residual lesions with high-definition white light followed by narrow band imaging and was photographed. Recurrent/residual polyp was documented and the area was treated as previously described.9 In cases where there remained any doubt as to the possibility of residual polyp at the EMR scar, this was biopsied for histology and then treated.

Recurrence was defined by the presence of endoscopic or histological evidence of residual polyp on the post-EMR scar site.

Pathological analysis

Histological specimens from EMR and from the surveillance procedure were analysed by gastrointestinal pathologists at their respective study centre. Results were then centrally collated on a prospectively maintained database. For this study, pathological specimens from serrated lesions removed by EMR, including lesions that contained any serrated morphology, were identified centrally and re-reviewed by an expert gastrointestinal pathologist at that centre. Any lesion with insufficient information in the histology report to adequately determine the correct classification was also re-examined.

Sessile lesions were classified according to the WHO classification system (2010)27 as SSA/Ps without cytological dysplasia (or ‘no dysplasia’ SSA/P-ND), SSA/Ps with cytological dysplasia (SSA/P-D), traditional serrated adenomas and hyperplastic polyps. Lesions with combined features of traditional serrated adenoma and conventional adenoma were separately classified. Twelve SSA/Ps-ND resected in 2011–2012 were unable to be formally reviewed so their diagnosis was clarified based on the original report alone. A total of 312 lesions underwent formal review. Forty-six serrated lesions included in the more recent period were not re-reviewed: 42 of them had a clear diagnosis of SSA/P and were included based on the original report alone; four of them presented mixed features and were excluded to avoid error.

For the present study, only lesions with final diagnosis of adenoma or SSA/P were included.

Data analysis

SPSS statistical software (IBM Corp 2012; IBM SPSS Statistics, (V.22) Armonk, New York, USA) was used to analyse the data. Continuous variables were summarised using mean (SD) or median (IQR) for skewed data. Categorical variables were summarised using frequencies (%) with 95% CIs when relevant. There were four outcomes of interest, namely EMR technical success, IPB, clinically significant adverse event within 14 days (defined as either CSPEB or deep mural injury or perforation) and recurrence. The main independent variable was lesion type (SSA/P vs adenoma). Other risk factors for each outcome of interest were considered only if they were potential confounders when estimating the effect of adenoma compared with SSA/P on the outcome. The potential confounders were those risk factors whose distribution differed substantially (p<0.1) between the adenoma and SSA/P groups and which also demonstrated univariable association with the outcome of interest.

EMR technical success was analysed on a per-lesion basis using generalised estimating equations (GEE) with a logit link, robust variance estimate and exchangeable correlation structure within patient. Adverse events, IPB and recurrence were analysed on a per-patient basis. If more than one lesion satisfying the inclusion criteria was resected from a patient, then only the largest lesion was included in the per-patient analyses (further details in Results section). Student's t test was used to compare the distribution of continuous variables between patient groups, and Pearson χ2 test was used to test for association between categorical variables. Two-tailed tests with a significance level of 5% were used throughout. There were insufficient adverse events in the SSA/P group to permit a meaningful multivariable analysis. The adverse events were described in detail and a χ2 test was used to test for differences between patient groups. Multivariable logistic regression was used to assess the effect of lesion type on IPB adjusted for the independently significant potential confounding variables identified using backward stepwise selection. ORs and associated 95% CIs were used to quantify the level of association.

Kaplan–Meier survival curves were used to illustrate the distribution of the months from EMR to first recurrence by lesion type and size. Cox proportional hazards models were used to estimate the associated HRs and their 95% CIs. Multivariable Cox regression was used to assess the effect of lesion type on recurrence adjusted for the independently significant potential confounding variables identified using stepwise selection. Since there was a large difference in the distribution of lesion size group by type, the two-way interaction between these factors was assessed in a multivariable Cox model of recurrence and a stratified analysis of recurrence by size group included.


A total of 2000 lesions ≥20 mm were assessed for resection in 1813 patients. About 105 lesions corresponding to 98 patients were not clearly identified as adenomas or SSA/Ps and were excluded from the analysis (23 traditional serrated adenomas, 35 tubulovillous adenomas with serrated component, 19 invasive cancers that occupied the full sample precluding identifying the precursor lesion, 28 other diagnoses such as neuroendocrine tumour, inflammatory polyp or hamartoma). EMR was not attempted in 45 polyps (40 adenomas and five SSA/Ps) because the appearance of the lesion was strongly suggestive of submucosal invasive carcinoma (32 adenomas and one SSA/P) or EMR was not technically possible (eight adenomas and four SSA/Ps). These patients were referred for surgery. Finally, 1527 adenomas ≥20 mm in 1425 patients and 323 SSA/Ps ≥20 mm in 246 patients were eligible for follow-up (figure 1).

Figure 1

Study enrolment and recurrence assessment flow diagram. EMR, endoscopic mucosal resection; NOS, carcinoma no otherwise specified; SSA, sessile serrated adenoma; TSA, traditional serrated adneoma; TVA, tubulo-villious adneoma.

EMR technical success—per-lesion analysis

Of the 1527 attempted adenomas, EMR was unsuccessful in 84 cases: in 18 the lesion did not elevate with submucosal injection, so these patients were directly referred for surgery; 66 lesions could not be completely resected (9 were attempted despite a non-lifting sign, 34 had submucosal fibrosis, 14 involved the ileocecal valve or appendiceal orifice, 11 were difficult to reach and position). Of the 323 attempted SSA/Ps, 3 lesions could not be completely resected because of (1) appendiceal orifice involvement or (2) submucosal fibrosis with non-lifting sign. In the per-protocol analysis, technical success was more likely in SSA/Ps compared with adenomas (99.1% vs 94.5% respectively; GEE OR 6.6, 95% CI (2.1 to 21), p=0.001) (figure 2).

Figure 2

Technical success flow diagram. EMR, endoscopic mucosal resection; SSA/P, sessile serrated adenoma/polyp.

Patient and their largest lesion characteristics

EMR was attempted in 1671 patients. Of these, 1571 patients (94%) had only a single lesion attempted. In the 100 patients (6%) with multiple lesions attempted, all lesions within the same patient were of the same type (SSA/P/adenoma) except for 10 patients (0.6%) in whom no adverse events or IPB were detected. The per-patient analysis of IPB and adverse events was based on the largest lesion attempted at EMR. Patient baseline characteristics are described in table 1 along with those details of their largest lesion, which showed substantial imbalance (p<0.1) between adenomas and SSA/Ps.

Table 1

Patient baseline characteristics and details of their largest lesion by adenoma and SSA/P groups

There are intrinsic differences between the SSA/P and adenoma groups. In particular, adenomas were larger than SSA/Ps: almost half the adenomas were >35 mm while half the SSA/Ps were ≤25 mm. Other potential risk factors exhibiting substantial imbalance between groups were age ≥70 years, gender, location, Paris 1s component, submucosal fibrosis, non-lifting sign, easy access and position, piecemeal resection, any additional modality, incomplete snare excision and EMR duration. Since EMR duration is largely determined by many of the other factors, it was not included as a potential confounding variable in the following per-patient analyses.

IPB and clinically significant adverse events—per-patient analyses

IPB occurred in 238 patients with adenoma (16.7%) and 17 patients with SSA/P (6.9%), (p<0.001). All cases of IPB were successfully controlled and no patient required transfusion or reintervention (colonoscopy, angiographic embolisation, or surgery). Multivariable logistic regression analysis identified adenoma histology, increasing largest lesion size and age <70 years as factors independently associated with IPB (table 2).

Table 2

Multivariable logistic regression model for IPB* significant potential confounding variables

The rate of CSPEB was similar in both cohorts: 6.3% (90 patients) and 5.7% (14 patients) for adenomas and SSA/Ps, respectively (p=0.708). Bleeding presentation was within 24 h in 44 of the 104 patients experiencing CSPEB (42%), 24–48 h in 24 (23%), 48 h–8 days in 20 (19%) and after 8 days in 15 patients (14%). Only two SSA/Ps bled after 48 h. Fifty-five patients (52.9%) were treated conservatively with admission and observation. Of the 48 patients who had repeat endoscopy, 14 (29.1%) had no active bleeding requiring treatment, 30 (62.5%) required endoscopic therapy, 3 had subsequent angioembolisation and 1 had surgery due to a perforation while treating the bleeding defect.

Deep mural injury was detected during the procedure in 33 patients: 26 (1.8%) adenomas and seven (2.8%) SSA/Ps (p=0.288). This comprised specimen and defect target sign in 17, only specimen target sign in 8, perforation with no contamination in 5 and perforation and contamination in 2. All injuries were closed with clips during the procedure. One patient had surgery following complete clip closure of an intraprocedural perforation where the lesion was in a difficult position with significant submucosal fibrosis, meaning that subsequent attempts at removing the residual adenoma would be difficult. There was one delayed perforation 12 days following the resection of a 40 mm transverse colon adenoma with a focus of invasive cancer. This patient had an extended right hemicolectomy. The SSA/P resection that resulted in a perforation was a 25 mm hepatic angle SSA/P-ND removed in three pieces.

There was no difference in perforation rates between adenomas and SSA/Ps (7 (0.4%) vs 1 (0.4%), respectively; p=0.974).

There were no deaths in the cohort.

Recurrence following successful EMR—per-patient analysis

Successful complete EMR was achieved for 1763 lesions in 1585 patients (see figure 1). Of these, 88 lesions in 78 patients were excluded due to surgery for histologically confirmed submucosal invasive carcinoma (SMIC), synchronous lesions or complications. Of the remaining 1675 lesions in 1507 patients eligible for follow-up, 80 lesions in 75 patients were not eligible for the recurrence study because the index EMR occurred after 9 April 2014. This left 1595 lesions in 1432 patients eligible for the recurrence study of which 70 lesions in 67 patients were lost to follow-up. Of the 1432 eligible patients, 1208 patients (84.4.0%) have so far been surveyed at least once: 1018 patients with adenomas and 190 with SSA/Ps.

A total of 516 patients with adenoma and 99 with SSA/Ps had more than one surveillance colonoscopy. In the 1208 patients with follow-up, recurrence occurred in only 5 patients who had multiple lesions (all of the same type). Recurrence occurred in the largest lesion except for one patient with multiple SSA/Ps and one with multiple adenomas. These two recurrences are not included in the following per-patient analyses based on the largest lesion resected.

Of those surveyed, 187 patients (18.4%) with adenoma ≥20 mm and 13 patients (6.8%) with SSA/Ps ≥20 mm had recurrent/residual lesion. The cumulative recurrence rates for adenoma after 6, 12, 18 and 24 months of follow-up were 16.1%, 20.4% 23.4% and 28.4%, respectively. For SSA/Ps they were 6.3% at 6 months and 7.0% from 12 months onwards (p<0.001). Of the 187 patients in the adenoma group identified as having a recurrence, 168 (89.8%) were diagnosed at the first surveillance colonoscopy. Twelve of the 13 patients (92.3%) with SSA/P recurrence were diagnosed at the first surveillance. All recurrences were macroscopically evident. Biopsies were taken in the 503 cases where there was uncertainty as to the possibility of recurrent polyp: residual/recurrent polyp was suspected in 124 and histologically confirmed in 122 (98.3%), whereas absence of recurrence/residual polyp was confirmed in 379 out of 379 cases (100%) where this condition was suspected.

Factors related to recurrence

Among SSA/Ps, factors related to recurrence on univariable analysis were: lesion size category (p<0.001), presence of any 0-Is component (p<0.001), submucosal fibrosis (p=0.051), non-lifting sign (p=0.044), incomplete snare resection (p=0.004), use of an additional modality (p=0.030) and presence of dysplasia (p=0.031).

Among the whole cohort, the HR of recurrence for adenoma versus SSA/P was 3.1 (95% CI (1.7 to 5.4), p<0.001). The potential confounding variables related to recurrence on univariable analysis were: age ≥70 years (p=0.018), size category (p<0.001), presence of 0-Is component (p<0.001), submucosal fibrosis (p=0.001), non-lifting sign (p=0.009), piecemeal resection (p<0.001), incomplete snare resection (p<0.001), use of additional modality (p<0.001), presence of IPB (p<0.001) and adenoma histology (p<0.001). The HR for adenoma versus SSA/P adjusted for the independently significant potential confounding variables identified using backward stepwise selection was 1.7 (95% CI (0.9 to 3.0), p=0.097; see table 3).

Table 3

Multivariable Cox proportional hazards model of recurrence*

Given the large difference in size by lesion type and the important role of size in recurrence, a subgroup analysis was performed by size. Table 4 shows the distribution of the variables of interest by lesion type within each size group. We note that the interaction between the effects of lesion type and size group on recurrence failed to reach statistical significance (p=0.099) but this test lacks power due to the small number of SSA/Ps >35 mm.

Table 4

Distribution of recurrence and potential confounding variables of interest by lesion type within each size group

Table 5 shows the unadjusted HR of recurrence for adenoma versus SSA/P and that adjusted for independently significant potential confounding variables in each size group. Among the smallest lesions (20–25 mm), there was a statistically significant eightfold increase in the risk of recurrence for adenomas compared with SSA/Ps, which persisted after adjusting for potential confounding variables. This significant difference in risk of recurrence was not evident among the larger lesions. Figure 3 illustrates the cumulative incidence of recurrence by lesion type for all patients, and within each size group. Of the 190 patients with SSA/P, 139, 98 and 46 remained at risk at 6, 12 and 18 months, respectively, but no recurrences were observed in this group after 7 months.

Table 5

HR of recurrence for adenoma versus SSA/P unadjusted and adjusted for independently significant potential confounding variables by size group

Figure 3

Illustration showing the cumulative incidence of recurrence by lesion type. EMR, endoscopic mucosal resection; SSA/P, sessile serrated adenoma/polyp.


In the past decade, sessile serrated lesions have emerged as an important factor in colorectal carcinogenesis. However, there remains considerable uncertainty about their natural history. This has resulted in a lack of scientific evidence to guide recommendations concerning their management. The present study is the first report comprehensively describing the endoscopic treatment outcomes of large SSA/Ps and comparing them with conventional adenomas.

In this multicentre, prospectively collected study of LSLs referred for EMR over 6 years, SSA/Ps comprised 16% (328/2000 of the total number of lesions). Previous reports in average-risk-screening patients have shown a much lower prevalence of serrated lesions ranging from 2% to 7%.17 ,28 Studies using ancillary techniques to optimise detection, including high-risk patients or using surrogate definitions for significant serrated lesions have reported prevalence as high as 14%–20%,15 ,16 ,29–31 in agreement with our results. In the present study, the protocol design has resulted in a selection bias in terms of lesion size and morphology but avoided any potential selection bias due to histology. Therefore, the proportion of SSA/Ps in this study is representative of typical large LSL populations. This considerable prevalence reinforces the importance of increasing our understanding of the specific behaviour of these lesions. This study is the largest series of LSLs referred for EMR with long-term follow-up and, with more than 300 SSA/Ps enrolled, provides a sufficient number of lesions to derive meaningful comparisons and conclusions.

Despite their proximal location, subtle demarcation and flat morphology, SSA/Ps were easier to remove than adenomas; in fact, EMR technical success approached 100%. There are several factors that contribute to this phenomenon. SSA/Ps were significantly smaller than adenomas—only 20.5% were larger than 35 mm in comparison with 44.2% of adenomas. This resulted in a higher rate of en-bloc resections. They usually lifted easily and were not associated with submucosal fibrosis. Despite their flat and subtle appearance, SSA/Ps seem to be relatively loosely attached to deeper mural structures and are easily detached.

SSA/Ps were associated with a 50% lower rate of IPB compared with adenomas, although they had a similar rate of clinically significant adverse events such as bleeding and perforation. Many endoscopists in clinical practice have been reluctant to systematically remove SSA/Ps, suggesting that the risk of adverse events was high and the clinical benefit was not clear.32 It is now well established that proximal serrated lesions have the potential for malignant degeneration and are associated with interval cancer. The results of this study are in accordance with previous results23 and demonstrate that SSA/Ps are not associated with a higher risk of adverse events.

Large SSA/Ps were associated with threefold less residual/recurrence than adenomas during follow-up. After multivariable adjustment for independently significant potential confounding variables, SSA/P histology was significantly associated with a better prognosis in 20–25 mm lesions, but this advantage was not maintained in larger lesion groups. Previously, lesion size >40 mm had been identified as an independent risk factor for early recurrence in LSLs.1 ,9 It is interesting to note that in the SSA/P cohort, as size increased, the relative protective effect against recurrence of serrated histology declined. Paradoxically, data from the CARE study have shown an inordinately high incomplete resection rate for SSA/Ps 1–2 cm in size.14 In light of these data, some experts have even suggested tattoo marking adjacent to SSA/P resection defects and repeating colonoscopy in 3–6 months to assess the site and remove any residual.33 Our results do not support this practice, and demonstrate that only 7.0% of SSA/Ps developed residual/recurrent tissue after a median of 13 months of follow-up. One could hypothesise that residual serrated tissue tends to spontaneously regress after partial polypectomy; however, this is unlikely since 19%–30% of interval cancers are thought to be secondary to incomplete polypectomy;19–22 and interval cancers are proximally located and often contain serrated pathway molecular features, indicating that they arise from a serrated lesion.19 ,34 ,35 Undoubtedly, technique is critical to ensure complete snare resection and low recurrence after EMR of SSA/Ps.24 Endoscopists must have the awareness and training to detect these inconspicuous lesions and accurately identify their margins.36 ,37 High-resolution scopes and ancillary techniques are especially helpful for this. Adding dye to the lifting solution clearly demarcates the lesion margins (figure 4). To ensure complete resection, the snare has to be placed carefully ensuring a rim of at least 1–2 mm of normal mucosa around the polyp. It is also mandatory to inspect in detail the margins of the defect to rule out any residual polyp; serrated mucosal changes can be extremely subtle and easily misinterpreted (figure 5), particularly after the effects of adjacent diathermic injury. If there is any doubt, suspected residual must be removed and the EMR defect extended without fear of adverse events.

Figure 4

Large flat serrated lesion in the ascending colon, best seen in retroflexion. The extent of the adherent stool debris acts as a guide to the subtle extent of the lesion (A). Narrow Band Imaging (NBI) confirms this impression (B). Submucosal injection shows the indistinct edges (C and D). Here the interface between the more obvious and flat indistinct region is seen after submucosal injection with dye.

Figure 5

Half the lesion has been resected (A). After complete resection the defect is inspected (B). The muscle layer has been exposed and is likely partly damaged by diathermy (C). Therefore 2 clips are placed.

This study is not without limitations. Biopsies from the post-polypectomy sites were not obtained systematically which could result in underestimation of recurrence. A previous prospective trial including 243 patients with standardised follow-up with chromoendoscopy and biopsies, after endoscopic removal of large non-pedunculated adenomas, showed that up to 7% of macroscopically inconspicuous polypectomy scars can present positive biopsies for residual/recurrent adenoma.11 In our series, biopsies were obtained in most cases and were always taken when there was doubt as to the presence of recurrence. Despite the extended follow-up in the present study, we cannot be sure that a few recurrences may have been missed in those patients surveyed. Despite every effort being made to follow-up all patients, a considerable number did not present for the surveillance procedure. We cannot rule out possible bias resulting from this incomplete follow-up which is very common in clinical practice. Finally, there is potential for interobserver variation among the pathologists. This has been reported in serrated lesion classification; however, almost all cases were reviewed by subspecialty GI pathologists and, with the release of the updated 2010 WHO guidelines, recent studies have shown very good reproducibility among pathologists.38 Moreover, to avoid confusion only adenomas and SSA/Ps were included in the study. Endoscopists were all highly experienced and although this may appear to limit applicability for general colonoscopy, these results emphasise the importance of specific training when facing LSLs. The CARE study has demonstrated that there is substantial variation in resection efficacy among endoscopists; this may in part relate to procedural volume and experience and this may have also influenced our results.14

In conclusion, we have shown that in comparison with adenomatous LSLs, recurrence after EMR of large SSA/Ps is about eight times less frequent for 20–25 mm lesions. This advantage disappears in larger lesions. SSA/Ps can be safely and effectively removed in comparable fashion with large adenomatous LSLs. We strongly advocate a standardised dye-based EMR technique including excision of at least a 1–2 mm rim of normal tissue at the margin. Careful technique is important in avoiding incomplete resection and the risk of interval cancer.


The authors would like to acknowledge the work of the following contributors who reviewed pathological cases: Dr Mark Bettington: ENVOI Pathology/QIMR Berghofer Medical Research Institute, Brisbane, Queensland, Australia. Associate Professor Andrew Clouston: ENVOI Pathology, Brisbane, Queensland, Australia. Dr Chris Dow: Dorevitch Pathology, Melbourne, Victoria, Australia. Professor Prithi Bhathal: Melbourne Pathology, Melbourne, Victoria, Australia. Dr Alan Pham Alfred Health Pathology, Melbourne, Victoria, Australia. Associate Professor Andrew Ruszkiewicz: SA Pathology, Adelaide, South Australia, Australia. Dr Andrina McGivern: Princess Alexandra Hospital, Brisbane, Queensland, Australia. Professor Priyanthi Kumarasinghe: Sir Charles Gairdner Hospital, Perth, Western Australia, Australia.


View Abstract


  • Meeting Presentation: Oral presentations at: Digestive Disease Week, Chicago 2014 Australian Gastroenterology Week, Gold Coast 2014

  • Contributors MP designed the study, collected data, analysed data, wrote the manuscript and revised the manuscript after review by the co-authors. NGB, NT, LFH, SAZ, GJB, RS, SJW, SCR, DO and AM identified and recruited patients, performed procedures, collected data and critically reviewed the manuscript. KB statistically analysed of data and critically reviewed the manuscript. HP'N, HM and DML collected and organised data, examined histological and pathological specimens and critically reviewed the manuscript. MJB initiated, designed and led the study, identified and recruited patients, performed procedures, collected data, co-wrote the manuscript and critically reviewed the manuscript.

  • Funding MP was supported by a funding grant from the Societat Catalana de Digestologia and by a grant from the Instituto de Salud Carlos III (BA14/00030). NGB was supported by a funding grant from the Westmead Medical Research Foundation (WMRF). The Cancer Institute New South Wales provided funding for a research nurse and data manager to assist with the administration of the study. There was no influence from the WMRF or the Cancer Institute on study design or conduct, data collection and management, analysis, interpretation, preparation and review or approval of the manuscript.

  • Competing interests None declared.

  • Ethics approval The study was approved by the Western Sydney Local Health District Human Research Ethics Committee. Ethical approval was obtained at each participating centre by the local ethics review board.

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

Request Permissions

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