Subepithelial lesions (SELs) of the gastrointestinal tract are protrusions covered by normal mucosa that originate between the mucosal layer and the serosal layer, such as the muscularis mucosa, submucosa, or muscularis propria. The detection rate of SELs has recently increased concomitantly with a rise in screening endoscopies, showing an incidence of approximately 1% to 2%.¹ SELs are most commonly found in the stomach, followed by the esophagus, duodenum, and colorectum. Although most SELs are asymptomatic and often benign, some SELs, such as gastrointestinal stromal tumors (GISTs), lymphomas, and neuroendocrine tumors (NETs), may have malignant potential.²

The endoscopic features of SELs, such as their color, consistency, size, mobility, and shape, allow presumptive diagnoses prior to definitive histologic diagnoses. Lesion location is also an important factor in the clinical presumptive diagnosis. For example, NETs can occur anywhere in the gastrointestinal tract, but are most commonly found in the rectum, with an incidence of approximately 1.04/100,000 persons, compared with an incidence of 0.3/100,000 persons for NETs in the stomach.³ The behavior of gastrointestinal SELs and the corresponding therapeutic strategies vary according to their clinical characteristics and the site affected. Generally, however, an accurate differential diagnosis based only on conventional endoscopic features and a histologic evaluation of subepithelial tissue acquired by endoscopic biopsy is difficult.

Endoscopic ultrasonography (EUS) is a noninvasive imaging method that visualizes the histological layers of the luminal walls of the gastrointestinal tract.⁴ EUS enables the determination of the layer of the gut wall from which SELs originate and accurate measurements of lesion size. In addition, an evaluation of the echogenicity, homogeneity, and margins of SELs may facilitate relatively accurate differential diagnoses. When tissue is required for histologic diagnosis, EUS-guided fine-needle biopsy shows high diagnostic accuracy (82%–89%).^5–8 Therefore, EUS has advantages in differentially diagnosing SELs of the gastrointestinal tract, as well as in determining treatment and predicting patients’ prognosis.

To date, however, the ability of EUS to evaluate rectal NETs has not been clearly determined, especially for evaluating small rectal NETs (< 10 mm).⁹ In addition, because rectal NETs usually affect both the mucosa and submucosa, the diagnostic yield of tissue sampled by conventional endoscopic biopsy is generally high, limiting the usefulness of tissue sampled by EUS. In this review, we describe the clinical applications of EUS imaging in diagnosing and treating rectal NET.

NETs, which are the most common type of SELs occurring in the rectum, arise from enterochromaffin cells of the intestine. On endoscopy, NETs are generally less than 10 mm in size, sessile, and single lesions. NETs usually show a hard consistency when pressed with biopsy forceps. These lesions are generally yellow in color, with smooth surfaces and clearly visualized blood vessels. The surface of large tumors may be accompanied by erosion, depression, ulceration, exudate, and hemorrhage.¹⁰

SELs of the rectum suspected of being NETs on conventional endoscopy can be evaluated in greater detail using EUS. The first and most important step is to identify the layer from which the lesion arises. The normal gut wall is composed of five layers: the mucosa, muscularis mucosa, submucosa, muscularis propria, and serosal layers. EUS at a high frequency of 12 to 20 MHz can reveal nine layers: the proper mucosa, consisting of a first layer with a border echo and a second layer; the muscularis mucosa, consisting of a third layer with border echo and a fourth layer; the submucosa or fifth layer; the muscularis propria, consisting of a sixth layer of inner muscle, a seventh layer with a border echo of connective tissue, an eighth layer of outer muscle; and the serosa or ninth layer.¹¹ The types of SELs differ based on their layer of origin, which is helpful for the differential diagnosis. Lesion echogenicity can be classified as anechoic, hypoechoic, isoechoic, and hyperechoic, relative to the degree of echo of the liver or spleen, which is defined as isoechoic.

Rectal NETs usually appear as well-circumscribed, smooth-margin, mildly hypoechoic, homogeneous, and oval or round lesions in the second and third layers of the gut wall (Fig. 1).¹² Hypoechoic lesions originating in the second layer of the rectum can be difficult to differentiate from leiomyomas and, more rarely, GISTs.

Figure 1. Endoscopic ultrasonography findings in a rectal neuroendocrine tumor, showing an 8-mm mildly hypoechoic lesion in the second and third layers (white arrow).

Several studies have shown that EUS was fairly accurate in assessing the size and depth of rectal NETs, even those < 10 mm in size. Park et al¹³ evaluated the accuracy of size and depth of invasion in rectal NETs ≤ 10 mm using EUS. They found no significant difference in size compared with histologic measurements (correlation coefficient = 0.727; P < 0.01), and the accuracy for depth was 92.5%. Another study of small rectal NETs < 10 mm located within the submucosal layer found that EUS had a 100% accuracy in measuring preoperative depth when compared with histology, with the mean sizes measured by these two methods not differing significantly (6.4 mm vs. 6.1 mm).¹⁴ EUS had a reported diagnostic yield for rectal NETs of 94.4% (34/36), a reported diagnostic yield for other non-NET SELs of 74.2% (23/31), a positive predictive value for rectal NETs of 80.9% (34/42), a negative predictive value for rectal NETs of 92.0% (23/25), and a diagnostic accuracy of 85.1%.¹⁵ Miniprobe EUS provided better results, but was not compared directly with conventional EUS.

The National Comprehensive Cancer Network (NCCN) and European Neuroendocrine Tumor Society guidelines recommend that all patients with rectal NETs undergo EUS or magnetic resonance imaging (MRI) to evaluate the lesion size and depth and regional lymph node metastases before treatment decisions are made.^16,17 Patients with incompletely resected incidental NETs should receive a similar evaluation.¹⁸ MRI can differentiate NETs from leiomyomas and GISTs, depending on the degree of enhancement of the tumors. Compared with muscle tissue, rectal NETs are usually isointense on T1-weighted image and isointense to hyperintense on T2-weighted image, and these lesions are homogeneously enhanced by contrast.¹⁹

An analysis of 247 patients with colorectal NET found that tumor size ≥ 11 mm and lymphatic invasion were risk factors for regional lymph node metastasis, whereas tumor size ≥ 21 mm and venous invasion were risk factors for distant metastasis.²⁰ The risk factors for regional or distant metastases of NETs include tumor size, involvement of the muscularis propria, differentiation index, and lymphovascular invasion,^17,21 along with multicentricity.²² The tumor size and depth of invasion can be determined by EUS prior to the final histologic evaluation. The T stage of rectal NETs is based on both the depth of invasion and tumor size. Tumors ≤ 20 mm that invade the lamina propria or submucosa are classified as T1 stage, while tumors measuring < 10 mm and 10 to 20 mm are classified as T1a and T1b, respectively. Tumors that invade the muscularis propria and those > 20 mm with invasion of the lamina propria or submucosa are classified as T2 stage.¹⁶ EUS is also helpful in assessing regional lymph node metastasis. An EUS evaluation of the size and depth of lesions and of the presence of regional lymph metastases can therefore help in planning appropriate treatment strategies for patients with rectal NET.

EUS evaluations of 120 patients with rectal NETs < 10 mm in diameter found that none showed invasion of the muscularis propria or regional lymph node metastases.¹³ In that study, all rectal NETs were classified as having grade 1 differentiation and L-cell phenotype. In addition, during a mean (± standard deviation) follow-up period of 408 ± 374 days, none of these patients had local or distant metastases. These findings suggested that EUS is not essential for the evaluation of small rectal NETs.

Other studies, however, have shown the presence of regional lymph node metastases in patients with rectal NETs < 10 mm in size. For example, lymph node metastases were detected in 3 of 43 patients (7.0%) with colorectal NETs < 10 mm who underwent surgery.²⁰ In addition, regional lymph node metastases were present in 1 of 97 patients (1.0%) with rectal NETs ≤ 5 mm in size and in 9 of 107 patients (8.4%) with rectal NETs 5 to 10 mm in size.²³ Another Japanese study showed regional lymph node metastases in 8 of 216 patients (3.7%) with rectal NETs < 5 mm and in 50 of 379 patients (13.2%) with rectal NETs measuring 5 to 10 mm.²⁴ A study analyzing the US National Cancer Center database found regional lymph node metastases in 25 of 169 patients (14.8%) with rectal NETs < 10 mm,²⁵ and a retrospective study in the US reported metastases in 2 of 66 patients (3.0%) with rectal NETs < 10 mm.²⁶ The results of these studies are summarized in Table 1.

Table 1 . Relationships between the Sizes of Rectal NETs and the Risk of Regional Lymph Node Metastasis.

Study	Country	Rectal NET size
		< 10 mm	≥ 10 mm and < 20 mm	≥ 20 mm
Soga24 (2005)	Japan	9.7% (58/595)	27.6% (42/152)	56.7% (17/30)
Konishi et al20 (2007)	Japan	6.4% (3/47)	40.0% (12/30)	57.6% (19/33)
Kasuga et al23 (2012)	Japan	4.9% (10/204)	54.5% (12/22)	66.7% (2/3)
Gleeson et al26 (2014)	US	3.0% (2/66)	66.7% (4/6)	73.3% (11/15)
Gamboa et al25 (2019)	US	14.8% (25/169)	54.1% (33/61)	68.6% (105/153)

Some rectal NETs < 10 mm in size may show invasion of the muscularis propria and may therefore be more likely to give rise to regional metastases. To our knowledge, however, no study has reported the incidence of involvement of the muscularis propria in patients with tumor size < 10 mm.²⁷ We encountered a patient with suspected muscularis propria invasion on EUS, despite the NET being 1 cm in size (Fig. 2). Histologic examination showed a high-grade (grade 2) NET, lymphovascular invasion, perineural invasion, and muscularis propria invasion, confirming the need to evaluate regional lymph node metastases in patients with lesions < 10 mm.

Figure 2. Muscularis propria invasion in a patient with a rectal neuroendocrine tumor (NET) < 10 mm in size. Endoscopic ultrasonography showed results suggesting muscularis propria invasion (white arrows). Endoscopic submucosal dissection was performed. The final histologic report revealed a grade 2 NET, 9 mm in size, with a mitotic count index of 3/10 high-power fields and a Ki-67 index of 5%. The patient was positive for lymphovascular and perineural invasion, and deep resection margins were observed.

Metastases have been detected in 25% to 65% of patients with well-differentiated rectal NETs 10 to 20 mm in size (T1b).^20,23–26 Low anterior resection or abdominoperineal resection should be considered in patients with rectal NETs > 20 mm in size, in those with invasion of the muscularis propria, and in those with lymph node metastases.²⁸

Patients with rectal NETs, even those with T1a lesions < 10 mm in size, are at risk of regional lymph node metastasis. Lymph node metastasis has been reported in patients with low-grade tumors (grade 1),^29,30 and distant metastases have been detected in some case reports.²² The risks of regional lymph node and distant metastases are very low, however, in patients with rectal NETs < 10 mm, muscularis propria invasion, and high-grade histology.⁹

Although regional lymph node metastases have been reported in patients with rectal NETs < 10 mm, none were reported in the absence of lymphatic invasion (0/28).²⁰ A retrospective analysis of 468 patients with rectal NETs reported that the rate of lymph node metastases in patients with tumors < 10 mm in size was 0.3%, indicating the extreme rarity of these metastases.³¹ In the present study, small rectal NETs in patients positive for lymph node metastases were accompanied by surface changes, erosion, and/or ulceration, allowing the risk of metastasis to be predicted in advance. An analysis of 347 patients with rectal NETs found that the risk factors for metastasis included tumor size ≥ 15 mm, increased mitosis rate, and lymphovascular invasion, and that none of the 237 patients with lesions < 10 mm were positive for lymph node metastasis.¹⁸ A study of 107 patients with small (< 10 mm) rectal NETs followed up for a median 31 months (range, 13–121 months) after endoscopic mucosal resection found no incidence of recurrence or metastasis, regardless of resection margin status, despite only four (3.7%) of these patients undergoing initial EUS.³² Evaluation of 95 patients with rectal NETs < 10 mm who underwent endoscopic mucosal resection or submucosal dissection found that only four were histologically positive for lymphovascular invasion, with only two having lymph node metastases.³³

Taken together, these findings suggested that, following incidental endoscopic resection of rectal NETs < 10 mm, patients may require additional examination using EUS or MRI, followed by surgery, depending on histologic results. Surveillance is sufficient in the absence of lymphovascular or muscularis propria invasion, or if the mitotic index is < 2/10 high power field.²¹ NCCN guidelines recommend surveillance in patients with low histologic grade (grade 1) tumors and small-to-intermediate resection margins but no surveillance in patients with completely resected small incidental NETs. Additional examination and/or treatment is recommended in patients with indeterminate margins following removal of intermediate histologic grade tumors.¹⁶

EUS determination of tumor size, depth of invasion, and regional lymph node involvement can help determine therapeutic strategies and predict the prognosis in patients suspected of having rectal NETs. However, rectal NETs < 5 mm in size have a very low possibility of recurrence or metastasis. In the absence of atypical features, such as surface changes, erosion, or ulcers, these tumors can be removed by endoscopic resection without further evaluation. Endoscopically resected incidental NETs < 10 mm have a high risk of recurrence or metastasis when the margins of high histologic grade tumors are positive, or when lymphovascular invasion is present. Regional lymph node metastases in these patients should be evaluated by EUS or MRI, whereas distant lymph node metastases should be evaluated by computed tomography (CT) or somatostatin receptor directed positron emission tomography-CT (SSR PET-CT). We suggest a diagnostic algorithm for rectal NETs using EUS (Fig. 3).

Figure 3. Suggestive diagnostic algorithm for rectal NETs. NET, neuroendocrine tumor; EMR, endoscopic mucosal resection; ESD, endoscopic submucosal dissection; LVI, lymphovascular invasion; EUS, endoscopic ultrasonography; MRI, magnetic resonance imaging; CT, computed tomography; SSR PET-CT, somatostatin receptor-directed positron emission computed tomography. †Atypical features: surface change, erosion, or ulcer.

SELs can be differentiated presumptively by their conventional endoscopic features, such as tumor location, size, color, consistency, and surface pattern. However, EUS or MRI may be required for a more accurate evaluation. EUS can help determine the characteristics of SELs, such as layer of origin, exact size, depth of invasion, echogenicity, and lesion borders. In addition, EUS can evaluate regional lymph node metastases. Endoscopic resection without EUS or MRI can be considered in patients with tumors measuring < 0.5 mm without atypical features. However, CT or SSR PET-CT should be considered when distant metastases are possible.

NETs, which are the most common type of SELs occurring in the rectum, arise from enterochromaffin cells of the intestine. On endoscopy, NETs are generally less than 10 mm in size, sessile, and single lesions. NETs usually show a hard consistency when pressed with biopsy forceps. These lesions are generally yellow in color, with smooth surfaces and clearly visualized blood vessels. The surface of large tumors may be accompanied by erosion, depression, ulceration, exudate, and hemorrhage.¹⁰

SELs of the rectum suspected of being NETs on conventional endoscopy can be evaluated in greater detail using EUS. The first and most important step is to identify the layer from which the lesion arises. The normal gut wall is composed of five layers: the mucosa, muscularis mucosa, submucosa, muscularis propria, and serosal layers. EUS at a high frequency of 12 to 20 MHz can reveal nine layers: the proper mucosa, consisting of a first layer with a border echo and a second layer; the muscularis mucosa, consisting of a third layer with border echo and a fourth layer; the submucosa or fifth layer; the muscularis propria, consisting of a sixth layer of inner muscle, a seventh layer with a border echo of connective tissue, an eighth layer of outer muscle; and the serosa or ninth layer.¹¹ The types of SELs differ based on their layer of origin, which is helpful for the differential diagnosis. Lesion echogenicity can be classified as anechoic, hypoechoic, isoechoic, and hyperechoic, relative to the degree of echo of the liver or spleen, which is defined as isoechoic.

Rectal NETs usually appear as well-circumscribed, smooth-margin, mildly hypoechoic, homogeneous, and oval or round lesions in the second and third layers of the gut wall (Fig. 1).¹² Hypoechoic lesions originating in the second layer of the rectum can be difficult to differentiate from leiomyomas and, more rarely, GISTs.

Figure 1. Endoscopic ultrasonography findings in a rectal neuroendocrine tumor, showing an 8-mm mildly hypoechoic lesion in the second and third layers (white arrow).

None.

No potential conflict of interest relevant to this article was reported.