Current status of endoscopic resection for small rectal neuroendocrine tumors

Abstract

Rectal neuroendocrine tumor (rNET) is an indolent malignancy often detected during colonoscopy screening. The incidence of rNET has increased approximately 10-fold over the past 30 years. Most rNETs detected during screening endoscopy are small, measuring < 10 mm. Current guidelines recommend endoscopic resection for small, well-differentiated rNET using modified endoscopic submucosal resection (mEMR) or endoscopic submucosal dissection. However, the optimal endoscopic treatment method remains uncertain. This paper summarizes the evidence on mEMR with submucosal stretching, mEMR without submucosal stretching, endoscopic submucosal dissection and endoscopic full-thickness resection. Given that rNETs often exhibit submucosal invasion, achieving adequate resection depth is crucial to ensure histological complete resection. mEMR with submucosal stretching appears favorable due to its high rate of histological complete resection, safety and convenience. Risk factors associated with lymph node and distant metastases are also discussed. A treatment algorithm is proposed to facilitate clinical decision-making.

Key Words: Rectal neuroendocrine tumor; Endoscopic resection; Endoscopic submucosal dissection; Modified endoscopic mucosal resection; Histological complete resection; Resection depth; Risk factor; Treatment algorithm

Core Tip: The initial evaluation of small rectal neuroendocrine tumors should thoroughly assess endoscopic features, including size, location, surface pit pattern and atypical changes. Imaging modalities should be used to rule out possible lymph node involvement and distant metastasis. Cold biopsy or polypectomy should be avoided, and modified endoscopic mucosal resection with submucosal stretching should be chosen as a priority. For small incompletely resected tumors without other risk factors, salvage resection can be carefully considered. Long-term follow-up is necessary if patients decline additional treatment. For patients with complete resection and risk factors, extended follow-up should be considered.

INTRODUCTION

Rectal neuroendocrine tumors (rNETs) are typically small and have low malignant potential, but the risk of metastasis exists[1]. Unlike neuroendocrine tumors (NETs) in other locations, carcinoid syndrome of rNETs is rare because tumors usually do not produce serotonin[2]. The majority of rNETs are < 10 mm, largely due to increased detection through screening colonoscopy[3]. Local excision, including endoscopic techniques and transanal surgery, is usually recommended for rNETs confined to the mucosal or submucosal layers. However, even tumors measuring ≤ 10 mm carry a 2%-3% risk of metastasis, underscoring the importance of developing robust strategies for optimal management of these tumors[4]. Currently, there are no clear guidelines for the optimal treatment or follow-up plan for small rNETs. This review aims to address these issues based on the most recent literature.

EPIDEMIOLOGY

The incidence of rNETs is about 1.04 per 100000 persons per year, accounting for about 1% of rectal neoplastic lesions[5,6]. Most patients are relatively young at diagnosis compared to those with mucosal malignancy, who are typically in their sixth decade with a slight male dominance[7]. rNETs are the third most common type of NET, representing about 15% of all gastroenteropancreatic-NETs[7]. There are global variations in the incidence of rNETs, with a high prevalence in Asian countries where they may account for up to 80% of all gastroenteropancreatic-NETs[8]. Racial disparities also exist, with higher rates observed in Black and Asian populations compared to White populations[9]. Risk factors for development of rNETs include low levels of high-density lipoprotein-cholesterol, elevated cholesterol and ferritin levels, and the presence of metabolic syndrome[10,11]. Other risk factors, including family history of cancer, smoking, alcohol consumption and body mass index, have not been consistently associated with the prevalence of rNETs in meta-analysis[12].

The incidence of NETs has been increasing across all sites, but rectal tumors have seen one of the most significant rises, likely due to increased recognition and the more frequent use of colonoscopy in screening[13]. In analysis of the Surveillance, Epidemiology, and End Results (SEER) database (SEER registry of the National Cancer Institute, United States), the age-adjusted incidence of rNETs has increased approximately 10-fold over the past 30 years[7]. In a large-scale observational study from Japan, colorectal NETs were identified in 67 of 82005 colonoscopies, yielding a detection rate of 82 per 100000 procedures[14]. However, age-adjusted incidence rate of rNETs has plateaued since 2015, possibly indicating the maximal benefit of colonoscopy screening[7]. According to the United States Cancer Statistics, rNET incidence rates during 2001 and 2020 have increased in younger adults (< 55 years) but not in older adults[15].

ENDOSCOPIC DIAGNOSIS

The presentation of rNETs ranges from asymptomatic, indolent tumors to aggressive metastatic disease. Resection can be either endoscopic or surgical, and the choice of technique depends on tumor characteristics such as size, grade, invasion into the muscularis propria (MP) layer, and the presence of lymph node involvement or distant metastases[8]. These factors should be thoroughly considered at the initial diagnosis.

Routine endoscopy

Recognizing rNETs during initial endoscopy is crucial for appropriate management[16,17]. Polypectomy or biopsy is commonly performed if the lesion is mistaken as a common polyp rather than an rNET. rNETs diagnosed or suspected before resection have a significantly higher complete resection rate compared to those treated as polyps and later diagnosed[18]. A retrospective analysis showed that it was possible to suspect a NET by macroscopic appearance of endoscopy in 96% of cases[19]. The diagnosis of rNETs could be possible just by routine endoscopy.

rNETs are typically small (≤ 10 mm in diameter), solitary lesions located 5-10 cm from the anal verge[1,2]. They usually appear as sessile lesions with a yellowish or whitish reflection and smooth, intact covering mucosa. Under narrow-band imaging, small round pits are typical features of rNETs, distinguishing them from starry-shaped pits in hyperplastic polyps (Figure 1A and B)[20]. These tumors are often described as hard and movable, and their consistency can be assessed using a biopsy forceps[1]. For those more significantly elevated, enlarged surface pit pattern may be found; however, the shape remains unchanged (Figure 1C and D). Larger lesions may exhibit surface irregularity with dilated microvessels (Figure 1E and F). Irregular surfaces usually show as a doughnut-shaped lesion with central depression, which can suggest a more aggressive clinical course[3]. Central depression, once thought to be risk factor for lymph node involvement, has not been confirmed as an independent risk factor, but may still be relevant for reducing the risk of incomplete resection[21,22]. Furthermore, dilated vessels around the lesion may be associated with tumor grade 2 and deep submucosal invasion (Figure 1E and F)[23].

Figure 1 Endoscopic diagnosis of rectal neuroendocrine tumors. A and B: The rectal neuroendocrine tumor typically appears as a small, yellowish, and subepithelial tumor-like lesion without surface change under white light or narrow band imaging as shown in case 1; C and D: For lesions with more significant elevation, there may be enlarged surface pit openings, but the overall shape remains unchanged, as seen in case 2; E and F: As the lesion grows larger, central depression may develop, accompanied by dilated micro-vessels, as illustrated in case 3.

Biopsy

Guidelines recommend taking a biopsy during initial endoscopy to confirm the diagnosis of rNETs[1]. However, evidence suggests that the prior biopsy may complicate complete endoscopic resection, presumably due to fibrosis and blurred tumor borders caused by inflammation and tissue repairing[19,24]. Biopsy should only be considered in doubtful cases with atypical features or in tumors > 20 mm, which should be referred for surgery[1]. The potential benefit of biopsy is to confirm histology before endoscopic resection, but given that most rectal subepithelial lesions (SELs) are NETs, and modified endoscopic mucosal resection (mEMR) is a safe and cost-effective procedure, biopsy may not be essential for lesions < 10 mm.

Endoscopic ultrasound

rNETs in endoscopic ultrasound (EUS) typically present as hypoechoic, homogeneous lesions originating from the submucosal layer[1]. EUS helps with predicting histology, estimating the size, evaluating the invasion depth and assessing possible lymph node metastasis in the perirectal area[8]. For histological prediction, it may have little significance for a detected rectal SEL due to the predominance of NETs in such lesions[25]. In a retrospective study for rectal SELs, the diagnostic accuracy of EUS in identifying rNETs was 89%[26]. While seemingly accurate, the prevalence of NETs was 79% in the cohort. Only a small portion of patients can benefit from the histological prediction. Furthermore, some rectal lesions may mimic the endoscopic appearance, even with similar EUS features, therefore it might be difficult for EUS to differentiate[27,28].

Regarding size estimation, a retrospective study including 120 rNETs cases reported no difference between endoscopy and EUS (r = 0.914/0.727, respectively)[29]. For depth estimation, it is well established on the high concordance in identifying the involvement of MP layer[30]. However, in lesions ≤ 10 mm, MP involvement is extremely rare[31]. There is a similar situation in assessing lymph node metastasis[32]. EUS does not have an adequate range to assess nodal involvement beyond the perirectal area[3]. Therefore, EUS may not be essential for rNETs < 10 mm[1,32]. Currently, EUS is generally recommend for lesions > 10 mm or those with atypical features, such as central depression[8,33].

METASTASIS

Over 80% of rNETs are < 10 mm and low-grade at diagnosis[8]. Although the risk of metastasis is low, rNETs ≤ 10 mm still carry a potential risk. In some cases, patients may even present with metastasis before the primary tumor is detected. Currently, no single parameter can precisely predict the presence of metastasis, so lymph node involvement and distant metastasis should be carefully evaluated using appropriate imaging modalities. A meta-analysis including 4575 rNET cases revealed 8% of patients with regional lymph node metastases and 4% with distant metastases[34]. The metastasis rate was higher than expected. It is important to note that most literature on lymph node involvement is based on radical resection studies with severe selection bias. Long-term surveillance with an endoscopically treated cohort is more likely to reflect the true metastatic rate of small rNETs.

Lymph node metastasis risk factors

Lymph node status is considered as a relevant prognostic factor for rNETs. Multiple factors have been linked to lymph node involvement. A previous meta-analysis identified tumor size > 1 cm, depth of invasion, venous invasion and central depression as factors of lymph node involvement, while lymphatic invasion was not[35].

Tumor size > 10 mm: This is the primary parameter for determining the risk for metastasis in rNETs[13,36]. In a population-based study from the United States, 20% of lymphatic metastases were found in 226 suspected cases out of 9000 rNETs < 10 mm, corresponding to 0.5% of all patients. In contrast, the metastasis rate was 8% for tumors 10-20 mm and 17% of those > 20 mm[37]. However, there are significant variations in metastasis rates across the published literature. The Niigatta Tumor Registry in Japan reported a metastasis rate of 10% for rNETs ≤ 10 mm[38].

Invasion of the MP: MP invasion significantly increases the risk of metastasis[39,40]. A meta-analysis found that MP invasion increased the risk of lateral pelvic lymph node metastasis with an odds ratio of 4.51[41].

Lymphovascular invasion: Lymphovascular invasion (LVI) is defined as the presence of tumor cells in blood vessels and/or lymphatic structures. Vascular invasion may have a stronger impact on lymph node metastasis than lymphatic invasion has[21]. The overall prevalence of LVI in small rNETs ≤ 10 mm was 26% according to a meta-analysis[4]. However, the prognosis of endoscopically treated rNETs appears to be complex and not directly proportional to the presence of LVI. No recurrence was found in 109 cases with identified LVI during follow-up of 30-76 months[4]. Multiple confounding factors further influence the significance of LVI in metastasis. Evaluation of LVI is subject to interobserver variations[16,42]. Additionally, the detection rates for LVI vary with diagnostic modalities. Immunohistochemical staining significantly increased the detection rate of LVI (up to 36%) compared to hematoxylin and eosin sections (13%)[4,43].

Tumor grade (G2/G3): Higher tumor grades (G2/G3) are associated with aggressive behavior in rNETs. A recent meta-analysis identified tumor grade as a significant risk factor for lateral pelvic lymph node metastasis, with an odds ratio of 7.76[41]. A retrospective study of 601 cases found the incidence of lymph node was 5% in G1 rNETs compared with 44% in G2 tumors[44]. However, when adjusted for tumor size, the difference in lymph node metastasis only persisted in tumors measuring 10-20 mm. This suggests that tumor grade may have little influence on metastasis risk for rNETs < 10 mm[45].

These risk factors are closely related, and it would be interesting to determine which is the most reliable predictor. Tumor size is the easiest and most widely accepted factor to assess before treatment. Predictive scores that incorporate both tumor size and LVI have been shown to provide an accurate assessment of the risk of metastatic lymph node involvement[46,47]. However, some predictive models rely solely on tumor size as a useful risk factor[48,49].

Distant metastasis risk factors

For distant metastasis in rNETs, the liver is the most common site (58%) followed by bone (9%), mesentery/peritoneum (8%) and lungs (8%)[13]. As in lymph node metastasis, tumor size is the major parameter for the risk of developing distant metastases. The optimal cutoff for tumor size in predicting distant metastasis is 11.5 mm[13]. Besides tumor size, a study based on the SEER database found distant metastasis in rNETs was associated with tumor grade, and regional lymph node metastasis, while T stage was not a significant factor[50].

Imaging modalities to evaluate metastasis

Our understanding of metastasis may be an under-reported finding as magnetic resonance imaging (MRI) of the pelvis and rectum is not routinely mandated for small rNETs. The indolent behavior of rNETs may lead to delayed detection of metastasis[51]. However, improvements in imaging quality, with MRI and more recently gallium-68-somatostatin receptor-positron emission tomography/computed tomography (68Ga-SSR-PET/CT), have enhanced the characterization of suspicious lymph nodes and the identification of small nodal metastases[3].

CT: CT has limited value for detection and characterization of regional metastatic lymph nodes in rNET patients[13]. However, for distant metastasis, abdominopelvic CT is the imaging modality of choice, particularly for patients with tumor ≥ 10 mm or grade G2/G3[13]. Contrast-enhanced CT with acquisitions at the delayed arterial (30 seconds) and then the portal venous (70-90 seconds) phases are recommended, as some highly vascularized NETs are only visible at one of these phases[17].

MRI: For localized rectal tumors, MRI is the best imaging modality, allowing precise disease staging. It should be regarded as mandatory before treatment, even in the absence of specific risk factors[13]. Similar to rectal adenocarcinoma, MRI with diffusion-weighted imaging (DWI) is considered the most sensitive imaging method for detecting regional lymph node and pelvic structure involvement in more advanced tumors[13]. Ga-enhanced MRI with DWI is more sensitive than CT for detecting liver and bone metastases. If liver metastases are only detectable by MRI, it should then be used as the primary procedure for follow-up[17]. The European Neuroendocrine Tumor Society recommends MRI for all tumors ≥ 10 mm, all G2/G3 tumors and all cases with suspected lymph node or liver involvement prior to resection[8,13].

Somatostatin-receptor imaging: 68Ga-SSR-PET/CT can assess the primary and metastatic sites. However, it is not routinely recommended for evaluating small rNETs before treatment, as it is not clear whether somatostatin receptors are highly expressed in these tumors, even though most well-to-moderately differentiated NETs do express them. If there is concern of lymph node involvement on MRI, functional imaging should be considered[8]. There is no strong recommendation favoring one modality (e.g., 68Ga-DOTA-PET/CT) over others (e.g. DOTATOC, DOTANOC, or DOTATATE)[17]. For PET/CT, it is important to understand the limitations, including its spatial resolution of about 5 mm and potential false positivity at sites of infection/inflammation. Imaging should be performed at an appropriate time after treatment with confirmed SSR expression, usually > 6 months[13]. Other lesions, such as leiomyomas with SSR expression, can also be a source of false positivity on PET/CT scans[52]. 68Ga-SSR-PET/CT is generally more sensitive than 18F-fluorodeoxyglucose-PET/CT[53]. Another promising somatostatin receptor imaging technique is 18F-AIF-NOTA-octreotide, which uses F-18 instead of Ga-68 for labeling. In a study of 26 patients with gastrointestinal NETs ≤ 10 mm, the patient-based sensitivity of 18F-AlF-octreotide PET/CT was 62%, compared to 38% for contrast-enhanced CT/MRI[54]. In rare cases, screening for distant metastasis in rNETs with PET/CT may help identify NETs in other location[55].

TREATMENT STRATEGY

There is ongoing debate regarding the management of rNETs. A key question is which tumors can be resected endoscopically, and which endoscopic technique should be used[8]. Unfortunately, due to the relatively low incidence of these tumors, most of current evidence comes from single-center and/or retrospective studies with inconsistent data. This inconsistency may stem from differences in endoscopic skills and training across centers, leading to a preference for one technique over another in specific situations. Meta-analyses have primarily included studies conducted in Asian countries, which may introduce bias due to regional variations in endoscopic expertise. Randomized clinical trials comparing the outcomes of specific techniques are scarce, and more studies are needed to support the existing evidence[13]. As a result, most guidelines do not strongly recommend specific techniques for endoscopic resection, or weakly recommend with low levels of evidence[40,56] (Table 1).

Table 1 Summary of resection techniques and recommendations for rectal neuroendocrine tumor less than 10 mm.

Ref.	Endoscopic option	Strategy after R1 resection	Strategy after R0 resection with risk factors	General follow-up recommendations
French intergroup, 2020[17]	EMRL/EMRC/ESD	Consider salvage resection	Surgical resection with lymphadenectomy	No follow-up required for rNETs that are G1, < 10 mm, T1 and R0 after the initial resection. For others, regular endoscopic examination and abdominal/pelvic MRI
JNETS, 2021[64]	Not specified	Surgery	Surgery for rNETs that are > 1 cm or G2; MP invasion; or suspected local LNM	Not specified
ESGE, 2022[56]	mEMR	Repeat endoscopy at 3-6 m. Salvage resection with confirmed residue disease in expert centers	Annual endoscopy as well as imaging modalities	No follow-up required for rNET that are < 10 mm, G1-G2, no MP invasion, and no LNM
ENETS, 2023[13]	mEMR/ESD/EFR	Watch and wait after discussion with patient if negative EUS, MRI and repeat biopsy. Salvage endoscopic resection or TAMIS	For rectal NET G1 L1 or V1 or G2/G3 ≤ 10 mm, 6 monthly abdominopelvic MRI and yearly sigmoidoscopy for at least 5 years. 68Ga-SSR-PET/CT initially and after 12 months	No follow-up for a rectal NET G1 L0 V0 ≤ 10 mm. After R1 resection without a second endoscopic resection, endoscopy and EUS or MRI 12 monthly for at least 5 years is recommended
Italian, 2024[33]	mEMR (EMRC preferred) or ESD	Watch and wait may be considered after patient consultation. Salvage resection with EMR > ESD > EFR or TAMIS	Not specified	Not specified
NCCN, 2025[122]	Not specified	Endoscopy at 6-12 m to assess for residue disease. For patients with residual disease, rectal MRI or EUS should be performed before TEM/ER	Not specified	No follow-up

R0 resection: Complete resection; R1 resection: Incomplete resection; EMRL: Endoscopic submucosal resection with a ligation device; EMRC: Cap-assisted endoscopic submucosal resection; ESD: Endoscopic submucosal dissection; rNET: Rectal neuroendocrine tumor; MRI: Magnetic resonance imaging; JNETS: Japanese Neuroendocrine Tumor Society; MP: Muscularis propria; LNM: Lymph node metastasis; ESGE: European Society of Gastrointestinal Endoscopy; mEMR: Modified endoscopic submucosal resection; EFR: Endoscopic full-thickness resection; EUS: Endoscopic ultrasound; TAMIS: Transanal minimally invasive surgery; NET: Neuroendocrine tumor; 68Ga-SSR-PET/CT: Gallium-68-somatostatin receptor-positron emission tomography/computed tomography; EMR: Endoscopic submucosal resection; NCCN: National Comprehensive Cancer Network; TEM: Transanal endoscopic microsurgery; ER: Endoscopic resection.

Benefit of histological complete resection

rNETs have metastatic potential, and resection offers the only curative option, reducing the risk of metastasis in most cases. Treatment should be carefully selected to ensure a histological complete (R0) resection[8]. R1 resection is defined when the tumor reaches the resection margin in the tissue sample[13]. Currently, there is no restriction for the extent of free margin; even a free margin < 1 mm is still considered R0[13]. However, measuring the vertical distance from the tumor border to the resection margin in the resected specimen may provide useful information about the adequacy of the resection margin[24]. Even with a 2-3-mm sampling interval thoroughly as in endoscopic submucosal dissection (ESD), an inadequate vertical margin distance still increases the risk of exposed tumor margins. However, there is no consensus on what constitutes an adequate margin. Theoretically, any endoscopic resection method that secures a longer vertical margin distance could achieve a more complete resection. This could potentially decrease surveillance burden, local recurrence, morbidity and mortality associated with rNET recurrence[24]. However, previous studies suggest that the vertical margin distance may be more related to invasion depth than to the endoscopic technique used[57]. Further studies are needed to validate the significance of vertical margin distance.

The slow progression of rNETs necessitates long-term follow-up, potentially up to 20 years. When properly treated, an R0 resection in grade 1 lesions ≤ 10 mm carries a low risk of recurrence and metastasis, allowing patients to be discharged from periodic follow-up[1,8]. Most incidentally diagnosed patients are relatively young and in good health, and they are likely to have only localized disease that can be cured with resection. Improper treatment can lead to a long-term burden for both the medical facilities and patients, increase the risk of metastasis, and result in repeated follow-up with radiological and endoscopic examinations, as well as the potential need for salvage therapy[1,3]. Therefore, it is crucial to balance the potential benefit of treatment against the risk of incomplete resection[8].

What not to do

Various endoscopic techniques have been attempted for the removal of rNETs, ranging from simple polypectomy to ESD and endoscopic full-thickness resection (EFR). Cold biopsy often results in an unevaluable specimen, making it difficult to guarantee complete resection on histological analysis and increasing the risk of incomplete resection in subsequent treatment[1,31]. This is commonly used when the lesion is not recognized as an rNET and assumed to be a common polyp[13,16]. Although the optimal endoscopic resection method remains unclear, it is clear that when an rNET is suspected, advanced resection techniques should be preferred over standard polypectomy or conventional EMR due to the high rate of positive margins with the latter[17]. Polypectomy is ineffective for rNETs, as they are mostly submucosal. In one study, the complete resection rate by polypectomy was only 31%[58]. Similarly, conventional EMR achieves R0 resection in only 50% of cases for rNETs ≤ 10 mm[59]. Multiple meta-analyses have shown the low R0 resection rate in conventional EMR, which is the reason why conventional EMR is generally recommended only for lesions < 5 mm[60-62].

Principles for rNETs ≤ 10 mm

rNETs originate from the lower crypts and infiltrate the submucosal layer, exhibiting a subepithelial tumor-like growth pattern[24]. Therefore, any diagnosed or suspected rNETs should be treated as a submucosal malignancy. Scheduled resection should be arranged rather than surveillance. According to SEER data, a delay of > 6 months was associated with an increased risk on survival with a hazard ratio of 4.5[63].

Guidelines recommend that choosing resection technique of rNETs depends on tumor size, grade and presence of lymph node involvement[8]. However, tumor size is the only factor that can be accurately evaluated before resection. For rNETs < 10 mm, local resection is recommended[2]. Although there are slight variations, current guidelines generally suggest resection either by mEMR for lesions ≤ 10 mm or by ESD/transanal endoscopic microsurgery (TEM) for lesions up to 20 mm[40,56] (Table 1). Oncological surgical resections are not routinely recommended for small rNETs without adverse histological or endoscopic features[3].

The most important technical aspect in resecting rNETs is to obtain a negative vertical margin[64]. Specimen handling should be given particular caution to facilitate the evaluation of margin status. A common safe practice is to align the fresh sample on cardboard to reduce fixation shrinking artifacts, allowing pathologists to properly assess the margins[13]. However, adherence to these practices is poor. According to a nationwide study in the Netherlands, resection margin was unsure or unknown in 48% of lesions. Tumor grade was unclassified in 50% of lesions, and only 33% and 13% of lesions were assessed for mitoses and Ki67 expression, respectively[65].

OPTIONS FOR ENDOSCOPIC RESECTION

mEMR techniques have been developed to improve the efficacy of conventional EMR. Some mEMR techniques enhance the resection depth by both a fluid cushion and submucosal stretching through suction, such as EMR with a ligation device (EMRL) and cap-assisted EMR (EMRC). Others do not include these steps, such as underwater EMR (EMRU) or circumferential incision EMR (CI-EMR).

mEMR with submucosal stretching

As mentioned earlier, mEMR with submucosal stretching combines submucosal injection and stretching to ensure adequate resection depth. EMRL is conducted using a ligation device attached to the endoscope. After submucosal injection with 3-5 mL normal saline, the lifted lesion is suctioned into the ligation device, allowing more submucosal tissue to be removed. An elastic band is then deployed, and the lesion is resected under the band using a snare[16]. In the EMRC procedure, a transparent cap with an inner groove is fitted to the scope. After submucosal injection with 3-5 mL normal saline, a crescent-shaped snare is looped along the inner groove, and the lesion is suctioned into the cap, grasped with the snare and resected[66]. These two mEMR techniques are the most commonly used methods. They provide undamaged round specimens with deeper and wider resection margins[24]. The key tip to obtain such specimens is to keep sustained suction to totally block the endoscopic view before releasing the rubber band or tightening the snare. The procedures are simple, have short procedure times and offer an excellent safety profile. Another advantage of EMRL or EMRC is that it requires less training. A Japanese retrospective study indicated EMRL performed by less-experienced endoscopists did not result in lower R0 resection[67]. Therefore, these techniques are recommended by guidelines for rNETs ≤ 10 mm (Table 1). There are also limitations for these methods. EMRL and EMRC rely on suction of submucosal tissue into the cap, and their performance dramatically compromises with significant submucosal fibrosis or larger lesions > 10 mm.

Several modifications of EMRL and EMRC have been proposed. Abe et al[68] used EUS to monitor the injection process, significantly improving the R0 resection rate. Li et al[69] performed EUS after deploying the rubber band to ensure completely ligation of the lesion. However, monitoring the injection process can be tricky, as it requires the simultaneous use of two devices. A linear EUS scope might be used instead of a regular endoscope. EUS after ligation may risk the rubber band becoming loose or falling off. In Li et al’s study[69], 12 of 48 lesions were considered as incomplete ligations requiring additional ligations, which seems less efficient when an 89% R0 resection rate was achieved in the control EMRL group. Similar to Li et al’s study[69], another study proposed routinely adding an extra band before snare resection in EMRL, but the additional benefit of this step remains debated[70]. Gao et al[71] suggested omitting the submucosal injection step in EMRC to aspirate more tissue and reduce device costs. They validated this method in a prospective trial, achieving a 97% R0 resection rate. However, the noninferior trial was designed to compare modified EMRC to ESD, not to standard EMRC with submucosal injection. Since mEMR relies on submucosal stretching after separating the tumor from underlying submucosal tissue with injection, it remains uncertain whether omitting injection compromises the vertical margin status.

Another form of mEMR with submucosal stretching is called strip biopsy or EMR using a dual-channel endoscope. Strip biopsy is performed with a dual-channel endoscope, usually without submucosal injection. The lesion is lifted with grasping forceps and then resected using a snare delivered through the other channel. There is limited data of strip biopsy, presumably less effective than ESD[72]. This technique is rarely used today due to the requirement for a multichannel endoscope. Lu et al[73] proposed using a prelooped snare on an attached transparent cap to substitute for the dual-channel endoscope. However, handling the snare can be tricky, and the efficacy of this method for achieving R0 resection remains to be validated.

mEMR without submucosal stretching

Several novel mEMR methods have been developed that omit the step of submucosal stretching. Many of them have been poorly evaluated for histological complete resection, with limited case numbers. A recent meta-analysis found that these techniques did not show superiority over convention EMR, unlike EMRL/EMRC[60].

The circumferential incision mEMR technique was first described as an option to facilitate en bloc resection for medium-sized mucosal lesions unsuitable for conventional EMR. It involves marking around the lesion, lifting the mucosal layer with injection, and then performing snare resection instead of dissection as in ESD. CI-EMR achieved an en bloc resection rate of 97% and a complete resection rate of 94%[74]. However, while this technique may improve lateral margin handling through mucosal circumferential incision, it does not enhance resection depth, as submucosal injection is the only measure to improve resection depth. A meta-analysis found that CI-EMR was less favorable than EMRL/EMRC in terms of positive vertical margin rates[62].

EMRU is inspired by the observation that the mucosa and submucosa separate from the MP when air is removed from the colorectal lumen and replaced with water. This reduces lumen stretching, thickens the submucosal layer, and facilitates resection with just a snare, allowing for the capture of a larger mucosal surface[16]. The main advantage of EMRU is that it requires no additional devices other than water immersion. However, there have been few studies with adequate case numbers and histological complete resection rate or resection depth for EMRU. Evidence from large sessile colorectal polyps suggests that a thickened submucosal layer facilitates obtaining an adequate vertical margin[75]. However, it is unclear whether EMRU is as effective as mEMR with submucosal stretching. EMRU has only been compared to ESD, with an R0 resection rate of 86%[76]. Additionally, the narrowed visual field due to changes in the refractive index of light and the shrunken intestinal lumen during water immersion can make securing the lateral margin challenging.

ESD

Most rNETs extend into the submucosal layer and require resection at a deeper plane than mucosal lesions, ideally close to the MP. The adjustable dissection plane in ESD makes it a seemingly ideal technique for rNETs[57]. Guidelines recommend ESD as an optimal option for rNETs < 20 mm. However, it is technically difficult to maintain an adequate and steady resection depth during submucosal dissection, often resulting in positive vertical margins or perforation if losing control of dissection plane. ESD may also be less ideal for securing lateral margins in small lesions like rNET, as the fluid cushion may not last till the end of the procedure. Additionally, ESD requires a long learning curve to achieve proficiency. A subgroup analysis by Kitagawa et al[66] found that trainees achieved a lower histological complete resection rate with ESD (65%) compared to mEMR (88%), while experts showed no difference, indicating that ESD should be performed by experienced endoscopists for rNETs.

The comparison of R0 resection rates between ESD and mEMR is controversial. Retrospective data comparing EMRC to ESD showed that ESD achieved a 100% en bloc resection rate, but the histological complete resection rate was only 80% for rNETs ≤ 10 mm, which was inferior to the 92% rate in the EMRC group[77]. In a similar study comparing EMRL to ESD, the R0 resection rate was 91% for EMRL and 82% for ESD[66]. Multiple meta-analyses have failed to showed a significant difference between mEMR and ESD[78]. A network meta-analysis ranked EMRL as the best treatment over ESD, although there was no significant difference in R0 resection rates. However, ESD showed more positive vertical margin cases than EMRL showed[62]. Similarly, another meta-analysis of 14 studies involving 823 patients found that mEMR with suction was superior to ESD in terms of R0 resection, although this result was achieved by excluding outliers[79]. A recent nonrandomized prospective study involving 50 institutes in Japan found no difference in R0 resection rates between mEMR (95%) and ESD (95%) for colorectal NETs ≤ 10 mm[80]. Further evidence from randomized clinical trials is needed to clarify the efficacy of ESD compared to mEMR[81].

Although mEMR may be superior to ESD, there are situations where ESD is the only option. For lesions > 10 mm, ESD is often necessary, as most mEMR caps cannot accommodate such sizes. If there is significant fibrosis after extensive biopsy without lymph node involvement, ESD may also be preferred over mEMR[8]. Therefore, a proper strategy should be used to avoid noncurative resection margins. Endoscopic intermuscular dissection, an emerging therapeutic modality that has gained attention for its efficacy in reducing positive vertical margins, particularly in managing NETs[82,83]. Underwater endoscopic intermuscular dissection has also been reported to enhance procedural efficacy[83]. However, intermuscular dissection is also technically challenging, as it requires maintaining a dissection plane between circular and longitudinal muscle fibers. Clips and loop traction may help achieve adequate depth by expanding and stabilizing the view of submucosal space (Figure 2)[84,85]. However, a recent retrospective study found no significant difference in vertical margin distance or R0 resection rates between traction-assisted ESD and conventional ESD[57]. This is possibly due to the small number of 24 cases in the traction-assisted ESD group. The tunnel method may also facilitate intermuscular dissection[82].

Figure 2 Endoscopic intermuscular dissection in a rectal neuroendocrine tumor. A: The lesion from case 2 (Figure 1C and D) was treated with intermuscular dissection. After circumferential incision, traction was applied with clips and a loop; B and C: During dissection, the circular muscle layer was carefully removed till the outer longitudinal muscle was fully exposed; D: Clips were used to close the mucosal defect.

EFR

When potential MP involvement is suspected, EFR should be utilized to ensure complete resection. Most EFR procedures are assisted by an over-the-scope clip (OTSC). The lesion is drawn into the cap of the endoscope by suction or grasping forceps, and OTSC is deployed over the lesion. A snare resection is performed above the clip, cutting through the entire bowel wall and enabling full-thickness resection[8]. A recent small study has reported a 100% R0 resection rate, with short intervention time and no major adverse events[86]. The short learning curve for OTSC makes it a user-friendly option for less-experienced endoscopists. However, these is a small chance of failing to capture the entire wall. Device malfunction, such as clip slippage, inadequate closure, and failed deployment, may occur, although they are uncommon[87]. Resection on the metal clip can also cause extensive thermal damage to the bowel wall and specimen, increasing the risk of indetermined margin status and postprocedural coagulation syndrome.

Traditional exposed EFR may also be performed. In other locations, such as the stomach, the defect produced by EFR is difficult to close due to compromised insufflation during the procedure[87]. In the rectum, however, the resection site is usually below the peritoneal reflection, and there is no free space if perforation occurs during the procedure. This allows the endoscopic view to be maintained, and closure of rectal wall defect may be easier (Figure 3). However, the risk of retroperitoneal infection remains, and closure with sufficient strength, such as using a detachable snare assisted clip closure, is recommended. Data are scarce on exposed EFR[88]. In small rNETs ≤ 10 mm, EFR is rarely needed, as ESD with intermuscular dissection should provide adequate resection depth to ensure clear margins.

Figure 3 Endoscopic full-thickness dissection in a 15-mm rectal neuroendocrine tumor. A: The lesion from case 3 (Figure 1E and F) was treated with endoscopic full-thickness dissection. Preprocedural positron emission tomography/computed tomography ruled out metastasis, and endoscopic ultrasound confirmed the involvement of the muscularis propria (MP). The lesion showed poor lifting with submucosal injection, indicating involvement of MP; B and C: After circumferential incision, traction was applied with clips and a loop; D: During dissection, the circular muscle was attempted for removing. However, extensive fibrosis made it impossible to preserve the outer longitudinal muscle, likely due to tumor invasion into MP; E and F: The entire rectal muscular wall was removed; G: Clips with a detachable snare were used to close the defect; H: Histological analysis revealed a complete structure of MP and a negative vertical margin, classified as pT2Nx. The patient opted to follow-up instead of additional radical surgery.

STRATEGY AFTER INCOMPLETE RESECTION

Histological complete resection of an rNET with no risk factors for recurrence is considered curative. Incomplete resection can be categorized as R1, where there is microscopic involvement of the margins, or R2, where visible tumor remains after resection[8]. For rNETs at low risk of metastasis with an initial R1 endoscopic resection, salvage resection by ESD or TEM can be proposed, but this should be restricted to expert centers[17]. If metastasis is suspected, oncological surgical resection should be considered. R1 resection is predominantly associated with endoscopic techniques[89]. It is crucial for endoscopists to properly recognized the lesion as an rNET and avoid routine biopsy[18,19]. Lesion characteristics, such as central depression and tumor size, may also play a role in R1 resection[22,90].

Detection of remnant tumor

The presence of remnant tumor after initial R1 resection varies significantly, ranging from 17% to 43%[13,91]. Many patients with R1 resection opt for follow-up rather than salvage resection, and only those with confirmed remnant tumors undergo additional treatment[1,13]. This may lead to biased patient selection in retrospective studies, with over-representation of cases with remnant tumors. In many studies addressing remnant tumors, a large number of cases were treated by biopsy or cold snaring rather than endoscopic resection techniques, sometimes accounting for up to 80% of cases[92]. For incidentally found lesion as small as 2-3 mm, complete resection may be achievable with biopsy alone[93]. However, for larger lesions, it is unlikely to achieve R0 resection with biopsy or cold snaring. Additionally, biopsied or cold snared specimens are rarely properly oriented, making margin status difficult to determine.

In cases where advanced techniques were used based on recommended treatment, the actual residual rate after additional treatment is < 20%. Therefore, adherence to guidelines is strongly recommended[40,94]. For lesions < 10 mm treated with endoscopic resection methods including EMR, the tumors with R1 resection are often resected at or close to the edge, which is technically an R1 resection by pathological standards. This does not necessarily mean that tumor tissue remains at the resection site, as the cautery effect may destroy residual tumor cells. However, the indolent nature of rNETs means that recurrence may take decades to manifest, as evidenced by a case report of recurrence 16 years after initial polypectomy[24]. The only independent factor related to residual disease appears to be the size of the lesion[92].

EUS has been recommended to assess the remaining submucosal tumors[17]. However, there is debate about whether EUS features can accurately predict residual disease[95]. A Korean retrospective study found that EUS had a similar area under the curve to simple visual detection (0.886 vs 0.870) for detecting remnant tumor[96]. A biopsy of the scar after close inspection may be useful, even if there is no visible evidence of remnant tumor on endoscopy, as residual disease may still be present.

Watch-and-wait strategy

Some patients may choose a watch-and-wait strategy, regardless of recommendations[1,97]. The significance of marginal status as a risk factor for recurrence remains uncertain[33,98]. One study of 436 patients with rNETs treated by ESD, with a median follow-up period of 61 months, found only one recurrence and one metastasis in 73 patients with R1 resection who opted for follow-up, compared to one recurrence and two metastases in 319 cases with R0 resection. There were no significant differences in survival between the two groups[94,99]. Similar results have been reported in multiple studies[22,94,100]. A meta-analysis comparing resection techniques for small rNETs found overall recurrence rates to be low (< 1%), with one local recurrence and one case of liver metastasis. However, the duration of follow-up was short[101]. Therefore, follow-up may be a feasible alternative to rigorous salvage therapy for patients who acknowledge the risk of possible recurrence and the need for long-term, potentially lifelong, follow-up[92,94]. When R1 resection occurs, the positive resection margins should be measured and discussed with pathologists to assess the risk of remnant tumor[102]. For grade 1 lesions ≤ 10 mm without LVI and no residual disease on scar biopsy, salvage procedure may not be necessary[1,8].

Endoscopic salvage therapy

Traditionally, rNETs with any risk factor for metastasis, including R1 resection, were considered for surgical resection with lymphadenectomy[17,94]. However, emerging evidence suggests that endoscopic procedures may be sufficient to achieve histological complete resection[103]. If there is evidence of residual disease in small tumors without lymph node involvement, local resection techniques such as TEM or ESD can be performed to achieve complete resection[8,92]. For patients who were not initially suspected of having an rNET and received nonrecommended endoscopic treatment (e.g., biopsy or cold snare), salvage treatment should be recommended due to high risk of residual lesions[92,94]. Some endoscopists recommended systematic scar resection, regardless of whether residual tumor is detected, to allow patients to avoid long-term follow-up if no other high risk factors are present[8,92].

Salvage resection techniques (ESD/EFR vs EMR) have been reported as the only significant factor associated with curative resection. mEMR can be performed if the scar is not visible and the lesion elevates satisfactorily after injection. Choosing mEMR may reduce the complicate rates, technical difficulty and medical costs. However, it is unlikely that no scar presents if the previous attempt was endoscopic resection rather than biopsy[92]. Therefore, ESD or EFR are the only options in this situation. Salvage endoscopic procedures using ESD or EFR have shown an R0 resection rate of near 100%[92]. When performing salvage ESD, care must be taken to ensure adequate resection depth (Figure 4). The inner circular muscular layer should be completely removed, and if scar blurs the border of muscle fibers, dissection under the scar is recommended to ensure negative margins[104,105]. A traction device can also be used to provide sustained countertraction during the procedure[106]. EFR has also been reported in salvage treatment for residual rNET lesions[107]. However, intentionally performing full-thickness resection may not be advisable when intermuscular dissection is likely to achieve adequate resection depth with a lower risk of adverse events.

Figure 4 Endoscopic salvage resection in a scar after endoscopic resection of rectal neuroendocrine tumor. A and B: Previous polypectomy showed a positive margin on histology. The scar exhibited poor lifting with submucosal injection; C and D: After circumferential incision, traction was applied with clips and a loop; E and F: During dissection at the scar site, the circular muscle layer was carefully removed to perform intramuscular dissection. Histological analysis showed no residue tumor.

Ju et al[108] proposed a type of polypectomy in salvage therapy of rNETs called wide hot snare polypectomy (WHSP). This technique involves placing a snare widely around the lesion to secure sufficient margins and then tightening it while pushing toward the lesion. They demonstrated a significantly higher R0 resection rate in the WHSP group (73%) compared to the ESD group (50%)[108]. However, this technique is more of a refinement of polypectomy rather than a novel procedure. The limitations of WHSP include the small size of the lesions (median 4 mm) and the fact that most previous treatments were biopsies (79%)[108].

Before considering salvage resection, endoscopic re-evaluation is essential to ensure complete removal of any residual rNET tumor[92]. Patients should undergo MRI evaluation of the scar area prior to second resection[8]. The primary benefit of salvage resection is to alleviate the burden of long-term follow-up[13,92]. However, it remains unclear whether re-resection of R1-resected small rNETs affects long-term outcomes.

PROGNOSIS AND FOLLOW-UP

Prognosis

rNETs have the best prognosis among all gastrointestinal NETs, with 5-year survival rate ranging 74% to 98%, according to various registry analysis[109,110]. Even in cases with metastasis, up to 77% to 89% of patients survived the first 5 years[36,38]. For rNETs ≤ 10 mm with complete resection and no other risk factors, the prognosis is usually curative. However, there are limited studies on long-term outcome following endoscopic treatment[31,111]. It should not be assumed that a T1 primary lesion will always have no lymph node or distant metastasis[3,112]. There are multiple published cases of small, well-differentiated rNETs recurring with lymph node involvement or distant metastases after apparent complete removal via endoscopic treatment[113,114].

Individualized survival prediction has garnered significant interest. Theoretically, these nomograms and predictive tools could provide more precise survival estimates by incorporating more factors such as age, tumor size, grade and LVI[115]. However, these tools have limited efficacy and poor generalization ability. Although they perform significantly better than the World Health Organization grade and American Joint Committee on Cancer tumor-node-metastasis classification, their C-index is only 0.65-0.70 for predicting overall survival, likely due to the limited case numbers in the training and validation datasets[49,116,117]. Recently, a GATIS score was developed to predict individualized survival for rNET patients based on a cohort of 1183 patients from 17 hospitals. In addition to common factors like tumor grade, T staging, tumor size, and age, the GATIS score further includes the prognostic nutritional index as a key component. Given the slow progression of rNETs, prognostic nutritional index may be a useful prognostic factor, as it reflects nutritional status[118,119]. The GATIS score demonstrated higher predictive power than tumor-node-metastasis stage and World Health Organization grade, with a C-index of 0.812 for overall survival and 0.865 for progression-free survival in the validation set[118]. However, the median follow-up duration of 34 months was short[120]. These nomograms and tools require further validation before they can be widely used to guide clinical decision-making in rNET management[115,121].

Follow-up

Most current guidelines recommend that patients with histologically complete resected rNET ≤ 10 mm and no risk factors for metastasis require no additional follow-up[13,40,56] (Table 1). However, localized recurrence or metachronous lesions may still occur, albeit at a low rate. Since these patients are typically in their 50s to 60s, minimal follow-up with colonoscopy as average-risk people is still encouraged. While it is not cost-effective to routinely perform imaging modalities such as MRI or PET/CT for surveillance, physicians should consider the possibility of recurrence based on the patient’s history, if there is unknown lesion are detected incidentally.

For patients with R1 resection without additional treatment or R0 resection with risk factors (e.g., tumor grade G2/G3, invasion of the MP, or LVI), the risk of metastatic spread is higher[8]. Immediate radical surgery is not absolutely necessary for these patients. However, there are reports of diffuse multiple rNETs remaining nonprogressive after 20 years of follow-up[112]. Metachronous metastatic recurrence may occur very late, and long-term follow-up is recommended for delayed recurrence (at least 20 years, potentially lifelong), starting from annual surveillance, although follow-up interval can be progressively lengthened[4,17,112]. This follow-up includes regular endoscopic examination and imaging. There have been limited studies on personalized follow-up stratified according to different risk factors; however, it could be a possible area to be explored.

Imaging, including abdominopelvic MRI with DWI, is recommended for rNETs with incomplete resection if a second endoscopic resection with R0 is not performed or if any of the aforementioned risk factors are present[13]. Most patients with NETs have prolonged survival, so radiation safety must be considered in their management. While imaging remains a cornerstone of follow-up, repeated CT-induced irradiation should be avoided, especially for patients with slowly or nonprogressive NETs. Nonionizing imaging techniques, such as MRI, should be considered as an alternative[17]. The role of functional imaging, particularly 68Ga-SSR-PET/CT, in follow-up remains unclear due to limited data[112]. According to European Neuroendocrine Tumor Society guideline, 68Ga-SSR-PET/CT is not recommended as a regularly performed modality during surveillance[13].

PROPOSED ALGORITHM FOR MANAGEMENT

As summarized in Figure 5, the initial evaluation of small rNET should thoroughly assess endoscopic features, including size, location, surface pit pattern and atypical changes including central depression. Imaging modalities, such as MRI-DWI, should be used to rule out lymph node involvement and distant metastasis. The optimal technique for treating localized small NETs, as well as the follow-up strategy, remains uncertain. However, efforts should be made to avoid cold biopsy or polypectomy. mEMR with submucosal stretching, including EMRL/EMRC, should be prioritized. For R1-resected small tumors without risk factors, salvage resection may be considered to reduce the risk of late recurrence. Long-term follow-up will be necessary if patients decline additional treatment. For patients with R0 resection and risk factors, extended follow-up should be considered[13].

Figure 5 Proposed algorithm for the management of rectal neuroendocrine tumors. SEL: Subepithelial lesion; NET: Neuroendocrine tumor; EUS: Endoscopic ultrasound; MP: Muscularis propria; mEMR: Modified endoscopic submucosal resection; ESD: Endoscopic submucosal dissection; EFR: Endoscopic full-thickness resection; EMRC: Cap-assisted endoscopic submucosal resection; EMRL: Endoscopic submucosal resection with a ligation device; MRI: Magnetic resonance imaging; 68Ga-SSR-PET/CT: Gallium-68-somatostatin receptor-positron emission tomography/computed tomography; R0 resection: Complete resection; R1 resection: Incomplete resection; TEM: Transanal endoscopic microsurgery.

CONCLUSION

rNETs are no longer considered rare due to increased detection through colonoscopy screening. Endoscopists should familiarized themselves with the endoscopic features of rNETs. Advanced endoscopic resection techniques should be preferred over biopsy or cold snaring. Specimen obtained should be thoroughly evaluated for margin status and risk factors related to metastasis. Follow-up strategies should be carefully selected based on histological findings to ensure optimal patient outcomes. Developing novel techniques with more adequate vertical margin distance may further improved the long-term outcome of rNETs.

ACKNOWLEDGEMENTS

Thanks for the generous help from Janet Brandsma of University of Massachusetts Medical Center on manuscript preparation and language revision.