Introduction
The most commonly used treatment modality for noninvasive large ( ≥ 20 mm) nonpedunculated
colorectal polyps (LNPCPs) is endoscopic mucosal resection (EMR). A first surveillance
colonoscopy after 6 months to check for local recurrence is advocated in several guidelines
[1 ]
[2 ]
[3 ]. Until recently, guidelines recommended routine biopsies of the post-EMR scar to
confirm the absence of recurrence, and placement of a tattoo for post-EMR scar identification
[2 ]
[4 ]. The recently updated European Society of Gastrointestinal Endoscopy (ESGE) guideline
stated that routine biopsies can be omitted if sufficiently trained endoscopists have
evaluated the scar tissue with enhanced imaging [1 ], using a standardized imaging protocol [5 ]
[6 ]. Importantly, detection of the post-EMR scar was possible with easy-to-use optical
evaluation criteria, without the need for universal tattoo placement [7 ].
However, it remains uncertain whether these results can be extrapolated to nonexpert
centers [8 ]. Therefore, in the setting of a prospective multicenter study [9 ], we investigated whether the diagnostic accuracy of optical assessment of post-EMR
scars for recurrence was high enough at a community level to refrain from standardized
biopsies and the need for universal tattoo placement.
Methods
In this post-hoc analysis of the STAR-LNPCP study (NTR7477), follow-up colonoscopies
performed after EMR of an LNPCP in 30 Dutch community hospitals between October 2019
and May 2022 were included. The STAR-LNPCP study was a multicenter, cluster randomized
trial, in which 59 endoscopists from 30 community hospitals included all consecutive
LNPCPs. Participating hospitals were randomly chosen and were asked to nominate 1
or 2 candidates from their endoscopists who were dedicated to large-sized EMR. Randomization
into the training or control group was performed on a cluster (center) level. Endoscopists
from 15 centers were additionally trained in EMR of LNPCPs, while the endoscopists
at the other 15 centers were not. Further study details are described in the original
article by Meulen et al. [9 ]. The study was approved by the Medical Ethical Review Committee of the Maastricht
University Medical Center (MEC 2017–0017). All patients provided written informed
consent prior to the study.
Patient selection
Consecutive patients undergoing follow-up colonoscopies after previous EMR of an LNPCP
were included. Exclusion criteria were: initially incomplete EMR, inflammatory bowel
disease, and poor bowel preparation (Boston Bowel Preparation Score < 2 for the segment
of concern). All patients who underwent a 6-month surveillance colonoscopy in the
STAR-LNPCP study were included in this post-hoc analysis ([Fig. 1 ]).
Fig. 1 Flowchart of patient inclusion and outcomes.
Baseline characteristics
Patient characteristics such as age, sex, American Society of Anesthesiologists (ASA)
classification, medication use, and medical history were obtained from case record
forms. Baseline lesion and treatment characteristics were obtained from endoscopy
and histology reports. Baseline lesion characteristics consisted of size, morphology,
location, accessibility, enhanced imaging, and initial histology. Baseline treatment
characteristics consisted of EMR type (en bloc or piecemeal), use of adjunctive treatment
(e. g. hot avulsion, cold avulsion, snare-tip soft coagulation, argon plasma coagulation),
use of adjuvant thermal ablation, use of clips, and tattoo placement. When a patient
had more than one LNPCP, only one LNPCP per patient was randomly included in the original
study [9 ].
E-module assessment of post-EMR scar
Before starting the study, participating endoscopists had the opportunity to watch
an e-module regarding the identification and assessment of post-EMR scars. In this
e-module, the criteria for scar identification, standardized scar assessment, and
biopsy protocol were explained. Furthermore, the difference between recurrence and
post-EMR scar clip artifact was demonstrated by discussing several examples. The e-module
topics and order, as well as some example images, are presented in Appendix 1 s , see online-only Supplementary Material.
Standardized assessment of the EMR scar and biopsy protocol
A standardized protocol was followed during the assessment of the EMR scar. This included
in vivo evaluation of the scar and potential recurrent neoplasia, with the combined
use of white-light imaging, advanced imaging, and zoom/near focus, and pictures taken
for every imaging modality. The scar was carefully assessed for recurrent neoplasia
and the following characteristics of the scar were documented: location of the scar,
size of the scar, presence of recurrence, certainty (yes/no) about the presence/absence
of recurrence, unifocal or multifocal recurrence, location of recurrence (at the edge,
in the center, or both), and the morphology of recurrence. When recurrence was present,
this was treated and documented in the endoscopy report.
When there were no signs of recurrence, biopsies of the EMR scar were taken according
to a standardized biopsy protocol: depending on the size and shape (e. g. straight
line or round) of the scar, 1–3 biopsies were taken from the center of the scar, and
in the periphery of the scar at least one biopsy per quadrant was performed. Biopsies
from the center and periphery of the EMR scar were presented separately for histological
evaluation in the pathology lab.
Tattoo placement
Placement of a tattoo on the contralateral side of the post-EMR defect was left at
the discretion of the endoscopist during the initial colonoscopy.
Outcomes
The primary outcome was optical assessment of recurrence. Furthermore, the diagnostic
accuracy, sensitivity, specificity, positive predictive value (PPV), and negative
predictive value (NPV) were calculated. The predictive value of tattoo placement for
identification of the scar was also evaluated.
In addition, differences in post-EMR scar identification between dedicated and nondedicated
endoscopists were evaluated. Dedicated endoscopists were defined as endoscopists participating
in the STAR-LNPCP study and performing large polypectomies in their center. Nondedicated
endoscopists were defined as the endoscopists not participating in the STAR-LNPCP
study and not primarily performing large polypectomies in their center. At the start
of the STAR-LNPCP study, participating centers were asked which doctors performed
large EMRs in their center.
Furthermore, the effect of prior clipping on the optical assessment of post-EMR recurrence
was evaluated, with the hypothesis being that clipping would lead to post-EMR scar
clip artifacts, which would be mistaken for recurrence. To evaluate this hypothesis,
the PPV of clipped EMR defects was compared with that of nonclipped EMR defects.
The optical recurrence miss rate was defined as all histologically confirmed recurrences
that were optically assessed as negative for recurrence, calculated as a proportion
of the total number of scar inspections performed. False positives were defined as
all optically assessed recurrences, that were not confirmed as a recurrence by histology,
calculated as a proportion of total scar assessments.
Statistical analysis
For descriptive statistics, categorical variables are presented as numbers and percentages,
and numerical variables are presented as mean (SD) or median with interquartile range
(IQR; 25th to 75th percentile). Pearson’s chi-squared or Fisher’s exact tests were
used to compare groups regarding categorical variables.
Risk regression with correction for clustering of patients within endoscopists (generalized
estimating equation [GEE] with exchangeable covariance structure and probit link)
was performed to evaluate which variables (dedicated or nondedicated endoscopist,
submucosal tattoo, size of initial LNPCP, location, accessibility, morphology) were
independently related to the identification of the post-EMR scar. Furthermore, the
same analysis was performed to evaluate the effect of clipping on optical recurrence
assessment, with correction for dedicated vs. nondedicated endoscopists. Alongside
the intraclass correlations obtained from these GEE analyses, we report the adjusted
risk ratios (RR) with corresponding 95 %CI, and P values for each risk factor, corrected for all other risk factors in the model. As
this included multiple testing, the Bonferroni corrected 95 %CIs and P values are also provided.
Cohen's kappa was used to determine the agreement between the optical assessment of
recurrence and histological evaluation of biopsies. According to the definition proposed
by Landis and Koch, a kappa of 0.61–0.80 was considered “moderate to good” and a kappa
of 0.81–1.00 was considered “(almost) perfect” [10 ]. Furthermore, we calculated the NPV, PPV, sensitivity, specificity, and overall
diagnostic accuracy with 95 %CIs. Analyses were performed in a per-protocol manner,
including only scars that were found, assessed, and biopsied. Additionally, intention-to-treat
analysis was performed, assuming that scars that were not found and scars that were
not biopsied would not have shown any signs of recurrence when histological evaluation
of biopsy had taken place.
Furthermore, sensitivity analysis with cluster bootstrapping was performed, where
correction for clustering of patients/scars within the same endoscopist was made.
A two-sided P value of ≤ 0.05 (after Bonferroni correction) was considered statistically significant.
Statistical analysis was performed using IBM SPSS Statistics v27.0.0, except for cluster
bootstrapping, which was performed using R (v4.3.1). CIs for proportions (with continuity
correction) were computed using http://vassarstats.net/prop1.html .
Results
Baseline characteristics
A total of 1277 patients (mean age 68 [SD 9]; 45 % women) who underwent 6-month surveillance
colonoscopy after complete EMR of an LNPCP (median size 30 mm, IQR 25–40 mm; 64 %
located in the proximal colon) were included ([Table 1 ]). Surveillance colonoscopy and assessment of the post-EMR scar was performed by
161 endoscopists in 30 community hospitals, of whom 59 (37 %) were dedicated endoscopists.
Table 1
Characteristics of the 1277 included patients and their index colonoscopies.
n (%) unless otherwise specified
Patient characteristics
Age, mean (SD), years
68 (9)
Sex, female
573 (45 %)
ASA classification
238 (19 %)
871 (68 %)
167 (13 %)
1 (0 %)
Index colonoscopy characteristics
Indication for colonoscopy
538 (42 %)
199 (16 %)
184 (14 %)
356 (28 %)
BBPS ≥ 2 per inspected segment
1253 (98 %)
Lesion characteristics
Polyp location
820 (64 %)
457 (36 %)
Polyp size, median (IQR), mm
30 (25–40)
Size groups, mm
399 (31 %)
399 (31 %)
479 (38 %)
Morphology
789 (62 %)
488 (38 %)
Accessibility
1111 (87 %)
166 (13 %)
SMSA score
64 (5 %)
586 (46 %)
627 (49 %)
Clip placement
258 (20 %)
Submucosal tattoo
488 (38 %)
ASA, American Society of Anesthesiologists; BBPS, Boston Bowel Preparation Scale;
IQR, interquartile range; SMSA: size, morphology, site, and access.
A total of 1215 scars were identified and assessed for the presence/absence of recurrence
([Fig. 1 ]). In 1060 /1215 cases (87 %, 95 %CI 85 %–89 %), a dedicated large polypectomy endoscopist
assessed the post-EMR scar, while in 155 cases (13 %, 95 %CI 11 %–15 %), assessment
of the post-EMR scar was performed by an endoscopist who was not specialized in large
polypectomy (nondedicated). The cohort of hospitals was representative of the Dutch
community hospital population.
Post-EMR scar identification
A tattoo was placed in 488 /1277 cases (38 %, 95 %CI 36 %–41 %). In 1215/1277 cases
(95 %, 95 %CI 94 %–96 %), the post-EMR scar was identified and assessed during surveillance
colonoscopy. The presence of a submucosal tattoo was not associated with a higher
identification rate of the post-EMR scar (95 % vs. 95 %, Bonferroni corrected P > 0.99) ([Table 2 ]). Performance of scar inspection by a dedicated endoscopist instead of nondedicated
endoscopist was independently associated with a higher scar identification rate (96 %
vs. 88 %, Bonferroni corrected P < 0.001). The scar identification rate increased with increasing LNPCP size, from
92 % (95 %CI 89 %–95 %) in lesions of 20–29 mm, to 95 % (95 %CI 92 %–97 %) in lesions
of 30–39 mm, and 98 % (95 %CI 96 %–99 %) in ≥ 40-mm lesions (overall Bonferroni corrected
P = 0.03). Other lesion characteristics (proximal location, difficult accessibility,
and flat morphology) did not significantly influence the scar identification rate.
Table 2
Factors associated with post-EMR scar identification[1 ].
Adjusted risk ratio
95 %CI
P value
Bonferroni corrected
95 %CI
P value
Dedicated endoscopist
1.82
1.35–2.47
< 0.001
1.20–2.77
< 0.001
Submucosal tattoo
1.04
0.82–1.32
0.72
0.75–1.45
0.99
Size, mm
Ref
0.005
0.03
1.19
0.91–1.54
0.20
0.83–1.69
> 0.99
1.68
1.23–2.29
0.001
1.09–2.57
0.008
Flat morphology of initial LNPCP
0.88
0.68–1.13
0.31
0.62–1.25
> 0.99
Proximal location
0.84
0.65–1.09
0.19
0.59–1.20
> 0.99
Difficult accessibility
0.88
0.64–1.20
0.42
0.57–1.35
> 0.99
EMR, endoscopic mucosal resection; LNPCP: large nonpedunculated colorectal polyp.
1 Multivariable generalized estimating equation (GEE) analysis with correction for
clustering of patients within endoscopist; intraclass correlation 0.014.
Post-EMR scar assessment
All post-EMR scars were assessed using high definition white-light and advanced imaging,
as validated by the presence of procedural images. In 1097 /1215 of the post-EMR scars
(90 %, 95 %CI 88 %–92 %), the use of zoom or near focus could be confirmed by inspection
of procedural images. Histology of the post-EMR scar was obtained from biopsies or
treatment of recurrence in 1050 cases. The median number of biopsies was four (IQR
3–5).
The outcomes of optical assessment compared with histological confirmation by biopsy
are presented in [Table 3 ]. The overall prevalence of recurrence in this prospective cohort was 200/1050 (19 %,
95 %CI 17 %–22 %). The optical diagnosis of post-EMR recurrence showed a high diagnostic
accuracy of 93 % (95 %CI 91 %–94 %), with a sensitivity of 93 % (95 %CI 88 %–96 %),
specificity of 92 % (95 %CI 90 %–94 %), PPV of 74 % (95 %CI 68 %–80 %), and NPV of
98 % (95 %CI 97 %–99 %). The agreement between the optical assessment of recurrence
and histological evaluation of biopsies was good, with a Cohen’s kappa of 0.78 (95 %CI
0.73–0.83). Intention-to-treat analysis, in which nonbiopsied post-EMR scars were
included, with the assumption made that optical assessment and histology would both
be negative for recurrence, showed similar results (Table 1 s ).
Table 3
Outcomes of optical assessment and biopsy of post-EMR scars.
Histology
Recurrence
No recurrence
Optical assessment
Recurrence
186
64
250
No recurrence
14
786
800
200
850
1050
Outcome
95 %CI
Prevalence
19 %
17 %–22 %
Sensitivity
93 %
88 %–96 %
Specificity
92 %
90 %–94 %
Positive predictive value
74 %
68 %–80 %
Negative predictive value
98 %
97 %–99 %
Diagnostic accuracy
93 %
91 %–94 %
Cohen’s kappa
0.78
0.73–0.83
EMR, endoscopic mucosal resection.
Sensitivity analysis, in which clustering of patients within endoscopists is taken
into account, also showed similar results: sensitivity 93 % (95 %CI 90 %–96 %), specificity
93 % (95 %CI 90 %–95 %), PPV 74 % (95 %CI 68 %–81 %), NPV 98 % (95 %CI 97 %–99 %),
and diagnostic accuracy 93 % (95 %CI 91 %–95 %).
The optical recurrence miss rate was 1 % (11/960; 95 %CI 0.6 %–2.0 %) for dedicated
endoscopists and 3 % (3/90; 95 %CI 1.1 %–9.3 %) for nondedicated endoscopists (P = 0.11).
Influence of clip placement on post-EMR scar assessment
In 223/1050 histologically evaluated scars (21 %), clips were used to close the initial
EMR defect. The PPV for optical diagnosis of recurrence in post-EMR scars decreased
after clipping, from 78 % (95 %CI 72 %–84 %) in the nonclipped group to 63 % (95 %CI
50 %–74 %) in the clipped group. Risk regression accounting for clustering of patients
within endoscopist (GEE intraclass correlation 0.004) on accuracy of optical recurrence
assessment with correction for dedicated versus nondedicated endoscopists showed an
RR for clipped versus nonclipped of 0.73 (95 %CI 0.58–0.90; P = 0.004). Furthermore, the proportion of false positives out of total assessments
was higher after clipping (11 % vs. 5 %; P = 0.02).
Certainty of post-EMR scar assessment
There was high certainty about post-EMR scar assessment performed by both dedicated
and nondedicated endoscopists (95 % vs. 94 %, respectively; P = 0.71). In the post-EMR scars where endoscopists identified recurrence with certainty
(n = 187), the PPV was 86 % (95 %CI 80 %–90 %), while this was only 41 % (95 %CI 29 %–54 %)
in the scars where endoscopists were uncertain about the presence of recurrence (n = 63).
False-positive cases in the uncertain group (n = 37) were more often biopsied and
less often endoscopically treated, compared with false-positive cases in the certain
group (n = 27). Biopsies were performed in 20/37 cases (54 %, 95 %CI 37 %–70 %) in
the uncertain group, compared with 9/27 cases (33 %, 95 %CI 17 %–54 %) in the certain
group. Endoscopic treatment was performed in 17/37 cases (46 %, 95 %CI 30 %–63 %)
in the uncertain group, compared with 18/27 cases (67 %, 95 %CI 46 %–83 %) in the
certain group (P = 0.09). Endoscopic treatment of false-positive cases was performed with re-EMR,
cold snare polypectomy, cold avulsion with snare-tip soft coagulation (CAST), argon
plasma coagulation, or hot snaring.
Discussion
In this post-hoc analysis of the STAR-LNPCP study [9 ], optical assessment of post-EMR scars for recurrence at 6 months was excellent,
with a sensitivity of 93 % (95 %CI 88 %–96 %), specificity of 92 % (95 %CI 90 %–94 %),
NPV of 98 % (95 %CI 97 %–99 %), and good agreement between optical assessment and
histological confirmation, represented by a Cohen’s kappa of 0.78 (95 %CI 0.73–0.83).
Dedicated endoscopists were more likely to identify the post-EMR scar (96 % vs. 88 %).
Tattoo placement was not significantly associated with scar identification. Clipping
of the post-EMR defect significantly complicated the correct optical assessment of
the scar, as demonstrated by a higher number of false positives after clipping (11 %
vs. 5 %).
In this study, scar identification was associated with the experience of the endoscopist,
but not significantly with the placement of a tattoo, which argues against universal
placement of a tattoo after EMR. A recent Delphi agreement report stated that a tattoo
should be placed for polyps > 20 mm resected piecemeal with additional predictors
of recurrence [11 ]; there was an 84 % level of consensus. The following additional predictors were
suggested: size > 40 mm, the use of adjunctive thermal techniques, size, morphology,
site, access (SMSA) score of 4, and a prior failed attempt. This advice is grounded
on the assumption that it is difficult to identify the scar in a significant number
of cases, although clear data on the magnitude are limited [12 ]. A submucosal marking would support correct identification of the scar. However,
a more recent study showed that application of easy-to-use criteria, such as a pale
area, convergence of folds, and disruption of the normal colonic surface microvasculature,
showed a scar identification rate of 99.7 % [5 ]
[7 ]. These results were obtained in high volume, experienced centers, which shows that
identification is related to experience. Our study confirms these findings in a real-life
practice setting. Scars of larger LNPCPs in particular were more easily identified
because of more clearly visible features.
Taken together, these results provide an argument for a practice wherein endoscopists
performing EMR evaluate the post-EMR scar themselves, instead of universally placing
a tattoo. Tattoo placement has been shown to interfere with successful resection if
the tattoo is aligned with the scar or a residual adenoma, can accidently be injected
in the peritoneal cavity or mesorectum, and adds unnecessary costs. It should therefore
be restricted to post-EMR scars that are anticipated to be difficult to identify.
On a national level, our outcomes are similar to those reported in tertiary centers.
This is demonstrated by an NPV of 98 % and a PPV of 74 % in this national cohort,
which is in line with the NPVs and PPVs in tertiary centers, as reported by the Australian
ACE cohort (NPV 99 %, PPV 76 %) and ESCAPE trial (NPV 97 %, PPV 81 %) [5 ]
[6 ]. Clipping significantly complicated the optical assessment of post-EMR scars. With
the increasing use of clips to prevent post-polypectomy bleeding, post-EMR scar clip
artifacts will be increasingly detected [13 ]
[14 ]. The presence of a scar clip artifact decreased the PPV and increased the rate of
false-positive cases, in which unnecessary endoscopic treatment was performed. Careful
inspection of the post-EMR scar with advanced imaging should however lead to a correct
differentiation between clip artifact and neoplastic recurrence.
Furthermore, the current study showed that uncertainty about the presence of recurrence
led to biopsies being more frequently taken instead of direct treatment. The latter
(i. e. optical assessment and treatment of a suspected recurrence in the same session)
is advised in the current guidelines [2 ]
[15 ]. Although recurrence was less often observed in uncertain cases, overtreatment of
non-neoplastic tissue outweighs postponing treatment of the recurrence to obtain histological
confirmation, as it results in unnecessary additional costs and burden for the patient.
The most used treatment modalities for recurrence (CAST, hot avulsion, cold/hot snare
polypectomy) are known to have few complications. Furthermore, most recurrences have
been shown to be small, unifocal, and easy to treat. While endoscopic overtreatment
is not desirable, it should be noted that the risk of missing neoplastic recurrence
is much more concerning than the overtreatment of non-neoplastic tissue. Therefore,
treatment of any inconclusive nodules or areas in a post-EMR scar should still be
performed. Additionally, when there is absolute certainty about the absence of recurrence,
biopsies can be omitted.
This study is important because it adds to the growing evidence that optical diagnosis
is highly accurate for the exclusion of post-EMR recurrence [16 ]
[17 ]. Data are obtained from a structured multicenter trial, on community level, with
a large number of post-EMR scars. Therefore, results are considered generalizable
to everyday colonoscopy practice. Obtaining an NPV of 98 % at a community level, clearly
surpassing the PIVI (preservation and incorporation of valuable endoscopic innovations)
threshold of 90 %, shows that, with high certainty optical diagnosis, standard biopsies
can be omitted, resulting in significant cost reductions.
Several limitations should also be emphasized. The first limitation concerns the observed
protocol violations in 13.5 % of cases in this study. According to the study protocol,
all post-EMR scars should have been biopsied in a standardized manner; however, in
165/1215, this was not performed. This could have led to an overestimation of the
diagnostic accuracy and NPV of optical assessment of the post-EMR scar, because of
possible false-negative cases not being histologically confirmed in this cohort. Given
the large number of cases in this cohort and the high NPV obtained with small CIs,
it is however unlikely that these protocol violations would have significantly changed
the outcomes.
A second limitation is that the follow-up is limited to the first surveillance colonoscopy
at 6 months. As a result, it is possible that late recurrences may have been missed.
It is known that approximately 4 % of patients still develop a recurrence despite
showing a scar without recurrence at 6 months [3 ]
[18 ]. Routine biopsies are however unlikely to have a significant impact on this number
of late recurrences. The biopsy protocol was extensive, with biopsies at the center
and at the periphery, with a median number of four biopsies. Sampling error could
have occurred but would be inherent to the implementation of routine scar biopsies,
and therefore does not dismiss another follow-up endoscopy at 18 or 36 months after
EMR.
A third limitation might be the effect of participating in a randomized controlled
study on the performance of the endoscopist after training. An e-learning package
was offered at the beginning of the study. Although the e-learning on scar identification
and recurrence detection was offered to all participants to increase the quality of
the study, and by itself was not part of the study intervention, it may have caused
a learning effect. This may limit the generalizability of our results to an untrained
group of community endoscopists. The uptake of the e-learning was 49 %. The NPV in
the group of trained dedicated endoscopists was similar compared with the group of
untrained dedicated endoscopists (99 % vs. 98 %). The effect of training is therefore
likely to be limited. Furthermore, participating in a study may have caused a Hawthorne
effect, increasing the performance of the participants, which would also limit the
generalizability to real-life practice.
A fourth limitation is a lack of power owing to low numbers in the difference in optical
recurrence miss rate between dedicated and nondedicated endoscopists. Furthermore,
the frequent use of zoom/near focus may limit the generalizability of our study, as
this may not be available in every hospital. Previous studies have shown that the
use of zoom/near focus increases the detection rate of recurrence [5 ]
[6 ]. Lastly, post-EMR scar identification was performed by a dedicated endoscopist in
the majority of cases in our study. This might also limit the generalizability of
our results to the population of endoscopists working in community practice settings;
however, this further underlines the importance of dedicated endoscopists performing
post-EMR scar assessment.
In conclusion, the quality of optical assessment for recurrence at the post-EMR scar
at a community level was found to be high. Identification of the post-EMR scar was
high and the optical recurrence miss rate was low, especially from dedicated endoscopists.
Therefore, routinely taking biopsies of the post-EMR scar could be omitted, as well
as universal tattoo placement after EMR.
