Introduction
Gastrointestinal neuroendocrine tumors (NETs) are increasingly being recognized due
to the widespread availability of flexible endoscopic procedures [1]
[2]
[3]
[4]. The rectum is a common site for diagnosis, comprising roughly 18 % of all NETs
and one-third of gastrointestinal NETs in the United States. Although the incidence
of rectal NETs has been steadily rising in recent decades [5] the majority of rectal NETs are detected at an early stage and carry a favorable
overall prognosis with 5-year survival approaching 88 % [1]
[6]
[8]. European and American guidelines recommend endoscopic resection for NETs under
10 mm that are confined to the mucosa or submucosa (T1) without high risk features,
and note endoscopic resection can be considered in lesions under 20 mm [1]
[2]
[9]
[10].
Endoscopic ultrasound (EUS) has typically been part of the staging workup when a rectal
NET has been identified. EUS can be applied to facilitate local staging and has been
shown to correlate well with depth of invasion and histopathology specimens in larger
tumors [11]
[12]. In addition, it can aid in the selection of an appropriate resection modality when
choosing between endoscopic and surgical resection techniques [11]. However, EUS may not be required for diminutive or smaller lesions due to their
low rate of locally advanced disease or distant metastases [13]
[14]
Despite increasing detection rates and improvements in local staging, the overall
incidence of rectal NETs remains low at 1.05 per 100,000 adults [29]. One study demonstrated a prevalence of 0.17 % for NETs on screening colonoscopy,
which is significantly lower than that of adenomatous and hyperplastic polyps [16]. Given their relative infrequency, rectal NETs may be misidentified as adenomatous
or hyperplastic polyps at the time of index colonoscopy and removed via standard polypectomy
techniques. Studies have demonstrated that standard polypectomy or biopsy forcep removal
is insufficient for complete excision due to frequent submucosal involvement of carcinoid
tissue, with rates of complete resection as low as 20 % to 30 % in some settings [3]
[7]
[18].
Endoscopic mucosal resection (EMR) is widely available and more effective at removing
submucosal tissue, however reported rates of complete resection for rectal NETs remain
variable (72 %-100 %) [19]. Furthermore, there can be significant discrepancy between endoscopic assessment
of complete resection and pathology findings [20]. EMR techniques including submucosal injection of saline (with or without dye),
band ligation, and cap-assisted, have shown promise to improve complete histologic
resection with limited additional procedural time and an acceptable complication profile
[21]
[22]
[23]
[24]. Endoscopic submucosal dissection (ESD) is another technique that has become more
widely available in western centers and has a role in en-bloc removal of rectal NETs,
particularly in a salvage therapy setting due to scarring from prior resection attempts
[25]
[26].
Guidelines on surveillance and salvage therapy for residual carcinoid tissue after
standard polypectomy and the effectiveness of EUS and surveillance biopsy in that
setting are limited. In this study we present a cohort of patients with rectal NETs
who were referred for EUS staging or follow-up endoscopy after an initial attempt
at endoscopic removal.
Patients and methods
Procedural data tracking software (Provation MD, Minneapolis, Minnesota, United States)
and gastrointestinal pathology records were queried for “neuroendocrine tumor” and
“carcinoid.” These search results were further stratified by location to include only
colonic specimens obtained during colonoscopy or flexible sigmoidoscopy. Patients
referred for staging or follow-up endoscopy after initial endoscopic removal of a
rectal NET were reviewed. Patients who did not complete follow-up endoscopy were excluded.
Two patients underwent follow-up endoscopy without tissue sampling due to inability
of the endoscopist to adequately identify the prior resection site. These cases were
excluded from analysis on NET recurrence, but still received full rectal EUS for staging
and these data were included. All cases occurred between February 2007 and May 2018.
Procedures were performed by three therapeutic endoscopists at a quaternary care,
urban academic medical center.
Patients that were referred for a follow-up endoscopy at our center and had tissue
sampling of the scar from prior resection attempt to assess for the presence of residual
neuroendocrine tissue were reviewed retrospectively. Baseline patient and lesion information
including age, gender, anatomic location, lesion size, initial resection technique
and histopathology were reviewed. Endoscopic appearance of the prior resection site,
size and description of any visible lesions, EUS findings (if performed), method of
tissue sampling and the presence of residual NET on histopathology in follow-up were
also noted. Tissue sampling occurred in the form of biopsy alone, EMR or ESD of the
previous resection site, and the sampling method was chosen by the endoscopist on
a case-by-case basis. EUS was performed on a number of patients at the discretion
of the endoscopist. Raw percentages were obtained for rate of residual NET across
a variety of procedural variables. A univariate analysis was then performed to identify
categorical variables that impacted the likelihood of residual rectal NET after initial
polypectomy. Categorical variables were analyzed for association with the outcome
of interest using the Chi-squared test.
Results
A total of 49 patients with colorectal NETs were initially identified based on pathology
specimens. From the initial cohort, 39 patients received follow-up endoscopy with
or without EUS to assess for residual neuroendocrine tumor. The average patient age
was 57.2 ± 13.4 years and 71 % were female. Lesion size was reported in 29 index procedures
with an average of 7.3 ± 4.2 mm (range 1 – 20 mm), including six lesions between 11
and 20 mm. All lesions were located in the rectum or rectosigmoid. Histologic examination
identified all NETs to be classified as low-grade lesions. Primary resection technique
was available in 27/39 cases and included cold biopsy polypectomy (n = 14, average
size 4 mm), snare removal with or without the application of electrocautery (n = 8,
average size 11.5 mm), EMR (n = 4, average size 8.75 mm) and ESD (n = 1, size 20 mm)
([Table 1]). The previous resection site was successfully identified in 37/39 cases during
follow-up endoscopy ([Fig. 1]).
Table 1
Baseline characteristics
|
Age
|
57 ± 13.4 yrs
|
|
Sex (%F/%M)
|
71/29
|
|
Lesion size
|
7.3 ± 4.2 mm
|
|
Lesions with histology
|
37/39
|
|
Initial resection technique
|
Cold biopsy – 14
Snare – 8
EMR – 4
ESD – 1
Not reported – 12
|
EMR, endoscopic mucosal resection; ESD, endoscopic submucosal dissection
Fig. 1 Study flowchart.
The resection site was described as a healthy-appearing scar in 27 cases and a scar
plus a visible lesion in 10 cases. All patients in whom the initial resection site
was identified underwent tissue sampling for histology with biopsy alone (n = 8),
EMR (n = 27) or ESD (n = 2) at the discretion of the endoscopist ([Table 2]). There was evidence of residual NET identified on histologic specimens in 38 %
of cases (n = 14), all of which were discovered through EMR or ESD of the initial
resection site ([Fig. 2]). These EMR and ESD samples harboring residual NET were all found to have negative
resection margins on final pathology.
Table 2
Results of scar evaluation with resection and EUS.
|
Residual NET
|
14/37
|
|
Appearance
|
Healthy Scar – 27
Scar plus lesion – 10
|
|
EUS
|
31 – No residual disease
|
|
Tissue acquisition
|
Biopsy – 8
EMR – 27
ESD – 2
|
|
Preventative hemostasis
|
Bicap – 2
Clip – 26
APC – 3
|
|
Complications
|
Bleeding – 1
Perforation – 1
|
EUS, endoscopic ultrasound; EMR, endoscopic mucosal resection; ESD, endoscopic submucosal
dissection; APC, argon plasma coagulation.
Fig. 2 Top left: Healthy-appearing scar after initial resection. Top right: Saline lift
for EMR of the scar. Bottom left: EMR resection site. Bottom right: Histopathology
showing residual carcinoid tissue.
Of these 14 cases of residual disease, six (43 %) had a normal-appearing scar endoscopically.
There were a total of 27 patients that had a normal appearing scar, therefore residual
NET in this group was six of 27 (22 %). No patients who received biopsy alone for
tissue sampling in follow-up were found to have evidence of residual disease. EUS
evaluation of the previous resection site was performed in 31 of 39 patients, with
all except for one case being sonographically normal. One patient had subtle thickening
of the lower mucosa and upper submucosal layers; however, EMR of the site was performed
with normal histology.
Resection margins from the initial procedure were involved in 20 of 39 lesions prior
to referral to our center for follow-up endoscopy or EUS, but this did not correlate
with residual disease (P = 0.39) found on subsequent tissue sampling. Initial cold biopsy polypectomy (P = 0.005) and endoscopically visible lesions aside from a normal scar (P = 0.001) correlated with residual NET on subsequent endoscopy in univariate analysis
([Table 3]). EMR or ESD of the scar during follow-up endoscopy as opposed to standard biopsy
sampling alone was more likely to discover residual NET (P = 0.012) in this cohort.
Table 3
Factors associated with residual neuroendocrine tumor (NET).
|
Factor
|
N
|
Odds ratio
|
P value
|
95 % confidence interval (CI)
|
|
Initial resection with positive margins
|
20
|
1.97
|
0.52
|
(0.4 – 9.5)
|
|
Initial cold biopsy polypectomy
|
14
|
7.2
|
0.006[1]
|
(1.6 – 31.2)
|
|
Visual appearance other than normal scar
|
10
|
14
|
0.001[1]
|
(2.3 – 84.3)
|
1 statistically significant
Preventative hemostasis was applied in 93 % of patients that underwent EMR or ESD
for sampling of the previous resection site, with 15 % of those receiving more than
one modality. Endoscopic clips (n = 26), argon plasma coagulation (n = 3) and bipolar
electrocautery (n = 2) were utilized. There was one perforation (3.4 %) and one episode
of delayed bleeding (3.4 %) after EMR in this cohort, both of which were treated endoscopically.
There were no other significant complications reported.
Discussion
Endoscopic resection is safe and feasible for localized rectal NETs under 20 mm in
diameter; however, greater awareness during screening colonoscopy and selection of
appropriate endoscopic resection modalities are needed. Current guidelines state there
are insufficient comparative data to recommend a specific endoscopic resection technique
but acknowledge a risk of positive margins after conventional polypectomy [1]
[2]. Due to the high rate of submucosal involvement, EMR or ESD should be considered
first line therapy if a localized rectal NET is suspected on initial endoscopic evaluation
to increase the chances of a complete and curative resection [24]
[25]
[26]
[27]
[28]. Often times multiple resections are required to achieve R0 resection if biopsy
or polypectomy alone are performed during the initial therapeutic endoscopy, especially
for larger lesions (10 – 20 mm) [29]. If a lesion is suspected to be a rectal NET (i. e. firm, yellow hued, well-circumscribed
lesion) during index colonoscopy and local expertise at EMR/ESD is not available,
an initial cold biopsy may be performed to confirm the diagnosis followed by prompt
referral for consideration of further staging and therapy [1]
[2]. When conventional resection techniques are applied to a suspected NET in the colon
or rectum, careful description of the location and/or tattoo placement adjacent to
the lesion may be helpful in future identification of the site for staging and salvage
therapy in the event that neuroendocrine tissue is identified [1].
If a rectal NET has been inadvertently identified as an adenomatous or hyperplastic
polyp and removed by conventional snare or cold biopsy polypectomy, referral for salvage
EMR, modified EMR, or ESD should be strongly considered to ensure complete resection
and obviate the need for close surveillance [24]
[26]
[27]. Our cohort demonstrates that regardless of deep margin findings on initial histopathology,
EMR or ESD of the scar during follow up identifies residual microscopic neuroendocrine
tissue in as many as one third of patients. Many of the patients with residual disease
(22 %) identified on pathologic specimens had a normal-appearing scar at time of follow-up
endoscopy, highlighting the need for salvage therapy due to the submucosal origin
of most NETs. Cold biopsy polypectomy on initial examination, in particular, was an
independent predictor of residual carcinoid tissue found on subsequent endoscopy in
univariate analysis. This raises concern that cold biopsy and possibly conventional
snare specimens may be too superficial to adequately excise submucosal neuroendocrine
tissue, as has been previously demonstrated for diminutive adenomatous and hyperplastic
polyps [30]
[31]
[32]. Specimens from piecemeal cold biopsy polypectomy may also be architecturally distorted,
and hot biopsy or snare removal often contains cautery artifact that can interfere
with accurate identification of the deep margin [33]
[34].
Routine long-term endoscopic surveillance is not recommended after complete resection
of a localized rectal NET given the low likelihood of residual disease; however, it
is recognized that a small percentage may experience late recurrence [1]
[2]. Follow-up endoscopy with EUS has been used for local staging in rectal NETs. However,
EUS may be insensitive for the detection of small residual NET confined to the mucosa
or superficial submucosa as demonstrated in this cohort. In fact, our data demonstrate
that EUS did not add any additional information and we suggest that it should not
be performed for surveillance of small rectal NETs < 10 mm in size.
Limitations of this study include its retrospective design and small sample size.
Furthermore, on follow-up endoscopy, not all patients underwent EMR or ESD of the
prior resection scar. Some patients only underwent biopsy sampling alone, and therefore
it is not certain that those patients did not have residual disease beyond the reach
of standard cold biopsy specimens. Follow-up for the purposes of this study was provider-dependent,
which may have introduced selection bias in regards to tissue sampling and residual
disease.
Conclusions
Management of rectal NETs that are confined to the mucosa or submucosa can be achieved
using predominately endoscopic methods, which offer the potential for curative resection
with minimally invasive techniques. EMR, ESD, or hybrid EMR/ESD techniques should
be considered first line in the treatment of localized disease and should ideally
be applied during the index procedure to ensure complete resection and potentially
obviate the need for further procedures. Increased provider awareness at the time
of screening procedures and appropriate referral for consideration of advanced endoscopic
modalities are needed. However, when a lesion is misidentified or removed via conventional
polypectomy techniques, such as snare or biopsy forceps, endoscopic follow-up with
salvage EMR or ESD of the scar should be performed due to the high rate of residual
disease due to submucosal involvement.
