Eur Radiol DOI 10.1007/s00330-015-3600-z

GASTROINTESTINAL

CT differentiation of poorly-differentiated gastric neuroendocrine tumours from well-differentiated neuroendocrine tumours and gastric adenocarcinomas Seong Ho Kim & Se Hyung Kim & Min-A Kim & Cheong-il Shin & Joon Koo Han & Byung Ihn Choi

Received: 2 July 2014 / Revised: 4 November 2014 / Accepted: 13 January 2015 # European Society of Radiology 2015

Abstract Purpose To evaluate the differential CT features of gastric poorly-differentiated neuroendocrine tumours (PDNETs) from well-differentiated NETs (WD-NETs) and gastric adenocarcinomas (ADCs) and to suggest differential features of hepatic metastases from gastric NETs and ADCs. Materials and methods Our study population was comprised of 36 patients with gastric NETs (18 WD-NETs, 18 PD-NETs) and 38 patients with gastric ADCs who served as our control group. Multiple CT features were assessed to identify significant differential CT findings of PD-NETs from WD-NETs and ADCs. In addition, CT features of hepatic metastases including the metastasis-to-liver ratio were analyzed to differentiate metastatic NETs from ADCs. Results The presence of metastatic lymph nodes was the sole differentiator of PD-NETs from WD-NETs (P=.001, odds ratio=56.67), while the presence of intact overlying mucosa with mucosal tenting was the sole significant CT feature differentiating PD-NETs from ADCs (P=.047, odds ratio=15.3) For hepatic metastases, metastases from NETs were more hyper-attenuated than those from ADCs. S. H. Kim : S. H. Kim (*) : C.
Conclusion The presence of metastatic LNs and intact overlying mucosa with mucosal tenting are useful CT discriminators of PD-NETs from WD-NETs and ADCs, respectively. In addition, a higher metastasis-to-liver ratio may help differentiate hepatic metastases of gastric NETs from those of gastric ADCs with high accuracy. Key Points • Presence of metastatic LNs is a useful differentiator of PDNETs from WD-NETs. • Intact overlying mucosa with mucosal tenting suggests PDNETs more than gastric ADCs. • Metastatic LNs are larger in size and greater in necrotic volume in PD-NETs. • Hepatic metastases from gastric NETs are more hyperattenuated than those from ADCs.

Keywords Stomach . Neuroendocrine tumour . Adenocarcinoma . Metastasis . Computed tomography

Abbreviations NETs Neuroendocrine tumours WHO World Health Organization WD-NETs Well-differentiated neuroendocrine tumours PD-NETs Poorly-differentiated neuroendocrine tumours ADCs Adenocarcinomas LNs Lymph nodes MDCT Multi-detector computed tomography HU Hounsfield units ROI Region of interest ROC Receiver operating characteristic AUC Area under the curve

Eur Radiol

Introduction

Materials and methods

Gastric NETs comprise less than 1 % of all gastric neoplasms and can be classified into well-differentiated NETs (WDNETs) or poorly-differentiated NETs (PD-NETs) with respect to its differentiation, and into low-grade (grade 1), intermediate-grade (grade 2), or high-grade (grade 3) NETs in terms of its grade (1, 2). Low- and intermediate-grade NETs are generally considered to be WD-NETs and highgrade NETs are considered to be PD-NETs. Thus, accurate discrimination of the degree of differentiation as well as its grade is of clinical importance as the management of gastric NETs is determined by these factors and whether the disease is localized or metastatic (3); typically, endoscopic or local resection along with simple endoscopic surveillance at 6 month intervals is recommended for WD-NETs, while radical gastrectomy with lymph node (LN) dissection is usually performed for localized PD-NETs (4, 5). It is also of clinical importance to differentiate gastric NETs from gastric adenocarcinomas (ADCs) as somatostatin and its analogue have been proposed for the new medical treatment of gastric NETs (6, 7). Furthermore, different treatment methods ranging from radiofrequency ablation (RFA) or transarterial chemoembolization (TACE) to liver transplantation for hepatic metastases can be attempted in patients with gastric NETs, often resulting in good symptomatic relief and up to 80 % 5-year survival in select cases (8–10), but not in patients with gastric ADCs, Thus, an attempt to discriminate the origin of hepatic metastases would be of great clinical significance. As multi-detector computed tomography (MDCT) with gastric distension has been a mainstay in the evaluation of primary gastric tumour extent and distant metastasis and is also known to be better than gastroscopy for the diagnosis of Bormann type IV ADCs, CT may be the most appropriate complementary modality to endoscopy (11, 12). However, until now the differential imaging features between WD-NETs, PD-NETs, and gastric ADCs have not been thoroughly investigated. As NETs are histologically located in the subepithelial area or deeply intramurally, appearing as submucosal nodules on endoscopy (5, 13–16) in addition to the fact that NETs are usually hypervascular tumors, we hypothesized that several CT features including the presence of an intact overlying epithelial layer and their enhancement degree may be helpful for the differentiation of gastric NETs from ADCs in both primary and metastatic tumours. Accordingly, the purpose of our study is to retrospectively evaluate the differential CT features of gastric PD-NETs from WD-NETs and gastric ADCs and to suggest differential imaging features between hepatic metastases from gastric NETs and gastric ADCs.

Patients This retrospective study was approved by our institutional review board and the requirement for informed consent was waived. Between May 2003 and January 2013, a total of 46 patients with gastric NETs proven either through surgery or endoscopic biopsy was found in our pathology database. Among them, we included patients who fulfilled the following criteria: (a) patients pathologically proven to have NETs of the stomach, (b) patients who had preoperative CT images with optimal gastric distension, and (c) patients whose primary gastric lesions were detectable on axial scans of CT images. We excluded 10 patients who had CT images with suboptimal gastric distension (n=5), whose primary gastric lesions were not detectable (n=3), and patients in whom the origin of the primary lesions were difficult to assess (n=2). Finally, 36 patients with histopathologically-proven gastric NETs (M:F=25:11; mean age, 60.6; age range, 29~82) comprised our study population. Eighteen of the 36 patients were proven to have WD-NETs (M:F=11:7; mean age, 55.6; age range, 29 ~75), while the remaining 18 patients were proven to have PD-NETs (M:F=14:4; mean age, 65.6; age range, 47~82). To serve as the control group, we separately selected 38 patients with gastric ADCs (M:F=28:10, mean age, 60.6; age range, 33~81) pathologically-proven through either surgery or endoscopic biopsy between January and December 2010. They were matched in terms of age, sex, and tumour stage with the patients proven to have NETs; 19 patients were matched with the WD-NETs group (M:F=13:6, mean age, 55.5; age range, 33~77), and the other 19 patients were matched with the PD-NETs group (M:F=15:4, mean age, 65.7; age range, 48~81). Only the patients in the latter group were assessed for comparison of CT features between PD-NETs and gastric ADCs (Fig. 1). Among the 36 cases of gastric NETs, 11 had hepatic metastases, while 5 out of the 38 cases of gastric ADCs had hepatic metastases. Hepatic metastases were confirmed by histology (n=4) or based on a combination of CT findings and consecutive studies (n=12) determined by a consensus of two experienced abdominal radiologists (S.H.K and C.I.S with more than 10 years of experience). These 16 cases were enrolled for the assessment of hepatic metastases (Fig. 1).

Clinical and histological features The clinical features of all patients were analyzed by one author (S.H.K) using the electronic medical records of our hospital. For each patient, the type of treatment for both primary and metastatic tumours, type of surgery, and type of neoadjuvant and adjuvant treatment were analyzed.

Eur Radiol Fig. 1 Flow diagram of enrolled patients. NETs neuroendocrine tumours, CT computed tomography, WD-NETs welldifferentiated neuroendocrine tumours, PD-NETs poorlydifferentiated neuroendocrine tumours, ADCs adenocarcinomas

For histological analysis, all specimens of gastric NETs and ADCs obtained through either surgical resection or biopsy were reviewed by an experienced gastrointestinal pathologist (M.A.K with 15 years of experience). The following findings were described on the pathologic reports: tumour size, location of the tumour, Bormann type, pTNM staging on histology, degree of differentiation, histological grade, and histological diagnosis. The degree of differentiation was reported in cases of gastric NETs as well-differentiated or as poorly-differentiated, and the histological grade of gastric NETs was assessed as low-grade (grade 1), intermediate-grade (grade 2), or high-grade (grade 3) NETs.

patients with gastric ADCs underwent contrast-enhanced CT after ingestion of 500 mL of water. Contrast-enhanced CT images were obtained after administration of an iodinated contrast agent (Ultravist 370, Bayer Schering Pharma, Berlin, Germany) at a dose of 1.5 mL/kg and a rate of 3~5 mL/sec using an automatic power injector. Dynamic-enhanced images were obtained in 31 patients with gastric NETs and in 37 patients with gastric ADCs by scanning the images 13~17 sec, 60~75 sec, and 2~3 min after the attenuation of the descending thoracic aorta reached 100 Hounsfield Units (HU) using the bolus tracking technique for arterial, portal, and delay phases, respectively. The remaining 5 patients with gastric NETs and 1 patient with gastric ADC underwent single phase CT using the portal phase only.

CT technique Image analysis A variety of CT systems were used in our study owing to its retrospective design. All 74 patients with gastric NETs and ADCs underwent MDCT with 4, 8, 16, 64, 128, or 320 detector-rows either at our institution (n=68) or at an outside hospital (n=6). The parameters used for these MDCTs were as follows: detector configuration (0.63~1 mm), pitch (0.89~ 1.35), rotation time (0.5~0.75 sec), tube voltage (120 kVp), tube current (150~250 mAs), slice thickness (2.5~5 mm), and reconstruction interval (2~5 mm). Among the 36 patients with gastric NETs, 31 patients (86.1%, 31/36) underwent stomach protocol CT with gastric distension achieved by ingestion of either an effervescent agent (n=27) or more than 1 L of water (n=4), while 31 patients (81.6%; 31/38) with gastric ADCs underwent stomach protocol CT with gastric distension achieved via ingestion of an effervescent agent in all 31 patients. The remaining 5 patients with gastric NETs and 7

Evaluation of the primary tumour in the stomach The morphologic features of the primary gastric neoplasms as well as their enhancement patterns on CT images were independently assessed in the WD-NETs, PD-NETs, and gastric ADCs groups by two radiologists (S.H.K and S.H.K with 3 and 16 years of experience, respectively) in consensus, on a picture archiving and communication system (PACS) workstation monitor (m-view, INFINITT, Seoul, Korea). Both radiologists were aware that the study population had either gastric NETs or gastric ADCs, but were blinded to their histological subtype, the degree of differentiation, and histological grade. The following CT findings of the primary gastric lesion were assessed: longitudinal and transverse location of the tumour, multiplicity, maximum tumour size, tumour

Eur Radiol

margin, TNM staging, homogeneity, degree of enhancement on each phase, dynamic pattern of enhancement, and the presence of intact overlying mucosa with mucosal tenting. Longitudinal location was divided into the upper half (cardia, fundus, and high body) and lower half (mid body, low body, and antrum), while the transverse location was divided into the anterior wall, posterior wall, lesser curvature, and greater curvature sides. Multiplicity was assessed into three categories: single, multiple, or diffuse. Maximum tumour size was measured on the axial CT scan and tumour margin was defined as either a well-defined margin or an ill-defined margin. TNM staging was assessed in the same manner as in prior studies (17, 18). The degree of enhancement was evaluated as high attenuation, iso-attenuation, or low attenuation on arterial, portal, and delay phases, referencing the enhancement of the normal gastric mucosa. Thereafter, the dynamic pattern of enhancement was assessed as follows: persistent low attenuation (low/low), progressive enhancement pattern (low/iso, low/high, or iso/high), wash-in/wash-out pattern (iso/low, high/iso, or high/low), and persistent high attenuation (high/high). Intact overlying mucosa with mucosal tenting was considered to be positive when the normal looking enhancing mucosal layer was entirely or partially preserved on the tumour as well as when symmetric mucosal elevation at both ends of the tumour was demonstrable (Figs. 2, 3, 4 and 5). Complete disruption of the enhancing thin mucosal layer and irregularly thickened mucosal layer covering the entire tumour were considered to be features negative for the presence of an intact overlying mucosa with mucosal tenting.

Fig. 2 A 72-year old male with a well-differentiated neuroendocrine tumours (Grade 1). a, b. Axial CT images with left posterior oblique position on arterial phase demonstrate two tiny well-defined enhancing nodules (arrow) at the greater curvature side of mid body. Normal mucosal layer seems to be well-maintained in both two tiny lesions. c, d. 3D surfacerendered (C) and virtual gastroscopy (D) images also well depict two polypoid lesions at the greater curvature side of gastric mid body (arrows)

Evaluation of lymph node (LN) metastases LN metastases were evaluated in all patients with gastric NETs and ADCs. We defined the number of nodal stations according to the Japanese classification of gastric cancer (JCGC) (12, 19). The presence of necrosis was also evaluated in cases of metastatic LNs. We considered LN necrosis to be present when the total volume of the necrotic portion was greater than 50% of the entire volume of the metastatic LNs. Finally, the shortest diameter of the biggest metastatic LN was measured for the size comparison. Evaluation of hepatic metastases and peritoneal seeding Both subjective and objective analyses were performed for the hepatic metastases by the same two radiologists (S.H.K and S.H.K). For the subjective analysis of hepatic metastases, two reviewers assessed the metastatic lesions either as hyperattenuated or hypo-attenuated in comparison to the adjacent normal parenchyma. The highest enhancing peripheral portion was included for subjective interpretation. For objective analysis, the radiologists measured the mean HU at the most strongly enhancing portion of the hepatic metastases as well as at the normal liver parenchyma by drawing regions of interest (ROIs) on the arterial phase using a PACS workstation. Thereafter, the metastasis-to-liver ratio was calculated in both gastric NET and ADC groups using the following equation: (HU at the most strongly enhancing portion of hepatic metastases) / (HU at the normal liver parenchyma). As all metastatic

Eur Radiol

the backward projection method were performed to determine the most significant differential CT features. For the comparison between hepatic metastases from gastric NETs and from gastric ADCs, the independent t-test was used to compare the average values of their metastasis-to-liver ratios and Fisher’s exact test was used for comparison of the subjective interpretation between the two groups. Receiver operating characteristic (ROC) curve analysis was performed to ascertain the optimal cut-off value of tumor size in differentiating WDNETs from PD-NETs and for the metastasis-to-liver ratio in differentiating hepatic metastases from gastric NETs and from gastric ADCs. For all statistical analyses methods, a p-value <0.05 was considered to indicate a statistical significance.

Results Clinical and histological findings

Fig. 3 A 55-year old man with a large cell poorly-differentiated neuroendocrine tumour (Grade 3). a, b. Axial CT images with left posterior oblique position on portal phase show a Borrmann type I-like tumour (arrows) at the posterior wall of gastric angle. Note an intact overlying mucosa with mucosal tenting appearance (arrowheads) and metastatic lymph node (*) with necrosis at celiac axis

lesions in the liver showed peripheral rim-like enhancement in both gastric NETs and ADCs, ROIs for measurement of the most strongly enhancing portion was drawn at the peripheral portion of the metastatic lesions. Finally, the presence or absence of peritoneal seeding was also assessed. Peritoneal seeding was confirmed by histology (n=0) or based on a combination of CT findings and follow-up studies (n=3) determined by a consensus of two experienced abdominal radiologists (**BLINDED** with more than 10 years of experience).

All clinical and histological findings are summarized in Table 1. Ten (55.6 %) out of 18 WD-NETs were assessed as grade 1, while the other 8 (44.4 %) were assessed as grade 2. All PD-NETs were assessed as grade 3. Six patients with WDNETs (33.3 %, 6/18) underwent localized treatment, such as endoscopic submucosal dissection (n=5) and wedge resection (n=1), but none of the patients with PD-NETs and gastric ADCs underwent localized treatment. Neoadjuvant chemotherapy was only performed in patients with PD-NETs, even though there were only 2 such cases. For adjuvant therapy, a fluoropyrimidine anticancer drug, TS-1, was administered in three patients with gastric ADCs. Treatments other than conventional chemotherapy including somatostatin analogs (n= 1), interferons (n=2), and TACE (n=5) were performed in patients with gastric NETs with hepatic metastases. In patients with gastric ADCs and liver metastases, however, only conventional chemotherapy (n=4) was administered. The mean survival time of the 5 patients with gastric NETs and hepatic metastases treated with TACE was 30.5 months, ranging from 12~42 months, while the mean survival of the 4 patients with gastric ADCs and hepatic metastases treated with conventional chemotherapy was 19 months, ranging from 3~29 months. Interpreting CT findings of the primary tumour in the stomach and LN metastases

Statistical analysis All statistical analyses were performed using commercially available software, SPSS 21.0 for Windows (SPSS Inc, Chicago, IL, USA). The prevalence of each CT finding between WD-NETs and PD-NETs as well as between PD-NETs and gastric ADCs were compared using Fisher’s exact test for categorical variables and the independent t-test for continuous variables. Thereafter, binary logistic regression analyses using

Univariate analyses of the CT findings between WD-NETs (Fig. 2) and PD-NETs (Figs. 3, 4 and 5) revealed that PDNETs were significantly larger in size (5.39±2.32 cm vs 1.72±0.93 cm, P<.0001), more frequently located in the lower-half of the stomach (77.8 %; 14/18 vs 33.3 %; 6/18, P=.009), more often solitary in number (100 %; 18/18 vs 72.2 %; 13/18, P=.023), and more easily metastasized to LNs (94.4 %; 17/18 vs 22.2 %; 4/18, P<.0001) (Table 2).

Eur Radiol Fig. 4 A 68-year old man with a large cell poorly-differentiated neuroendocrine tumour (Grade 3). a, b. Axial (a) and coronal multiplanar reconstruction (b) images obtained in a left posterior oblique position on portal phase demonstrate a well-defined tumour (arrow) with Borrmann type II morphology at the greater curvature side of gastric antrum. Note an intact overlying mucosa with mucosal tenting appearance (arrowheads). c, d. 3D surfacerendered (c) and virtual gastroscopy (d) images also well depict an ulcerofungating mass (arrow) with central ulceration (*)

Fig. 5 A 75-year old male with advanced gastric cancer (a, b) and a 77year old female with a large cell poorly-differentiated neuroendocrine tumour (c, d). a, b. Axial CT images with supine position on arterial phase shows ill-defined, poorly-enhancing wall thickening at the lesser curvature side of mid body with showing a Borrmann type III feature (arrows). Normal mucosal layer is completely disrupted. Multiple low attenuated lesions (arrowheads) with subtle rim enhancement were seen in the liver. Metastasis-to-liver ratio in this case was calculated as 0.76. Note small amount of ascites with fat infiltration at perisplenic area (*),

suggesting accompanied peritoneal seeding. C, Axial CT image with left posterior oblique position on arterial phase demonstrates well-defined Borrmann type II-like tumour (arrow) with strong enhancement. Intact overlying mucosa with mucosal tenting appearance (arrowheads) is evident. D, Axial CT image with left posterior oblique position on arterial phase demonstrates rim-enhancing nodules (arrows) in the liver suggesting hypervascular hepatic metastases. Metastasis-to-liver ratio was calculated as 3.3 in this case. Note a necrotic lymph node (arrowhead) at the perigastric area

Eur Radiol Table 1 Clinical and histological features of WD-NETs, PD-NETs, and gastric ADCs

Characteristics

Tumour differentiation WD-NETs

PD-NETs

ADCs

5 (27.8%) 8 (44.4%) 3 (16.7%) 2 (11.1%)

1 (5.6%) 2 (11.1%) 2 (11.1%) 13 (72.2%)

0 (0%) 2 (10.5%) 3 (15.8%) 14 (73.7%)

3 (23.1%) 9 (69.2%) 1 (7.7%) 0 (0%)

1 (9.1%) 3 (27.3%) 5 (45.5%) 2 (18.1%)

0 (0%) 1 (8.3%) 7 (58.3%) 4 (33.3%)

10 (55.6%) 8 (44.4%) 0 (0%)

0 (0%) 0 (0%) 18 (100%)

5 (27.8%) 13 (72.2%)

7 (38.9%) 11 (61.1%)

Clinical T stage (CT) T1 T2 T3 T4 Pathologic T stage (Pathologic) T1 T2 T3 T4 Histological grade Grade 1 Grade 2 Grade 3 Biopsy vs resection Biopsy Endoscopic or surgical resection Type of surgery Endoscopic submucosal dissection Wedge resection

WD-NETs well-differentiated neuroendocrine tumours, PDNETs poorly-differentiated neuroendocrine tumors, ADCs adenocarcinomas, CT computed tomography. * Combined treatment for liver metastasis was performed in patients with WDNETs and PD-NETs

Subtotal gastrectomy Total gastrectomy Extended total gastrectomy Palliative chemotherapy No Yes Neoadjuvant chemotherapy No Yes Adjuvant therapy None Chemotherapy Fluoropyrimidine anticancer drug Herceptin Treatment for liver metastasis* None Chemotherapy Somatostatin analogue (Octreotide) Interferon Transarterial chemoembolization

On ROC analysis, the optimal cut-off tumour size value for differentiating PD-NETs from WD-NETs was 3.05 cm with an area under the curve (AUC), sensitivity, and specificity of 0.970, 84.2, and 94.7%, respectively. On binary logistic regression analysis, the presence of metastatic LNs was the sole differentiator of

7 (36.8%) 12 (63.2%)

5 (38.5%) 1 (7.7%)

0 (0%) 0 (0%)

0 (0%) 0 (0%)

5 (38.5%) 2 (15.3%) 0 (%)

6 (54.5%) 4 (36.4%) 1 (9.1%)

8 (66.7%) 4 (33.3%) 0 (0%)

18 (100%) 0 (0%)

13 (72.2%) 5 (27.8%)

16 (84.2%) 3 (15.8%)

18 (100%) 0 (0%)

16 (88.9%) 2 (11.1%)

19 (100%) 0 (0%)

18 (100%) 0 (0%) 0 (0%) 0 (0%)

9 (50%) 9 (50%) 0 (0%) 0 (0%)

7 (36.8%) 8 (42.1%) 3 (15.8%) 1 (5.3%)

1 0 0

1 7 1

1 4 0

1 2

1 3

0 0

PD-NETs from WD-NETs (P=.001) with an odds ratio of 56.67 (Table 3). In terms of the differentiation of primary tumours between PD-NETs and gastric ADCs, PD-NETs more frequently showed well-defined margins (55.6%; 10/18 vs 0%; 0/19, P<.0001) and more often accompanied with intact overlying

Eur Radiol Table 2 Prevalence of independent CT features between WD-NETs and PD-NETs

Characteristics

Tumour differentiation

Tumour size (cm)

CT computed tomography, WDNETs well-differentiated neuroendocrine tumours, PDNETs poorly-differentiated neuroendocrine tumours, LN lymph node, ns not significant *A P value<.05 is considered to indicate statistical significance **Upper half of the stomach includes cardia, fundus, and high body, while lower half of the stomach includes mid body, low body, and antrum ***15 out of 18 patients with WD-NETs and 16 out of 18 patients with PD-NETs had undergone multiphasic dynamic CT

Longitudinal location** Lower half Upper half Multiplicity Single Multiple Diffuse Tumour margin Well-defined Ill-defined Homogeneity Homogenous Heterogeneous Intact overlying mucosa with mucosal tenting Present Absent Dynamic enhancement pattern*** Persistent low attenuation Progressive enhancement pattern Wash-in/wash-out pattern Persistent high attenuation LN metastasis Present Absent Mean diameter of the largest LN (cm) LN necrosis Present Absent Japanese lymph node classification N1 N2 N3 Peritoneal carcinomatosis Present Absent

mucosa with mucosal tenting on CT when compared with gastric ADCs (55.6 %; 10/18 vs 5.3 %; 1/19, P = .001) (Figs. 3, 4 and 5). In terms of LN metastases, although distant nodal metastasis (N3 nodal group) was more frequently found in ADCs (33.3%, 6/18) than in NETs (11.8 %, 2/17), the difference was not statistically significant. However, the mean diameter of the largest metastatic LN was significantly larger in PD-NETs (2.12±1.29 cm) than in ADCs (1.22±0.68 cm) (P=.014). In addition, metastatic LNs in PD-NETs tended to show more extensive necrosis of greater than 50 % of the entire volume of the LNs compared to ADCs (58.8 %; 10/17 vs 22.2 %; 4/14, P=.027) (Table 4). Binary regression analysis revealed the presence of intact overlying mucosa with

P value*

Well-differentiated

Poorly-differentiated

1.72±0.93

5.39±2.32

<.0001

6 (33.3%) 12 (66.7%)

14 (77.8%) 4 (22.2%)

.009

13 (72.2%) 5 (27.8%) 0 (0%)

18 (100%) 0 (0%) 0 (0%)

.023

15 (83.3%) 3 (16.7%)

10 (55.6%) 8 (44.4%)

ns

16 (88.9%) 2 (11.1%)

13 (72.2%) 5 (27.8%)

ns

9 (50%) 9 (50%)

10 (55.6%) 8 (44.4%)

ns

0 (0%) 1 (6.7%) 0 (0%) 14 (93.3%)

2 (12.5%) 1 (6.3%) 0 (0%) 13 (81.2%)

ns

4 (22.2%) 14 (77.8%)

17 (94.4%) 1 (5.6%)

<.0001

1.15±0.24

2.12±1.29

ns

1 (25%) 3 (75%)

10 (58.8%) 7 (41.2%)

ns

3 (75%) 1 (25%) 0 (0%)

12 (70.6%) 3 (17.6%) 2 (11.8%)

ns

0 (0%) 18 (100%)

0 (0%) 18 (100%)

ns

mucosal tenting as the sole significant differential CT feature of PD-NETs from ADCs (P=.047) with an odds ratio of 15.3 (Table 5). Hepatic metastases and peritoneal seeding from gastric NETs versus gastric ADCs Hepatic metastases from gastric NETs appeared to be significantly more hyper-attenuated than those from gastric ADCs on subjective analysis (81.8 %; 9/11 vs 20 %; 1/5, P=.036). The average HU value at the most strongly enhancing portion of hepatic metastases was higher in those from gastric NETs than in those from gastric ADCs on the arterial phase (Fig. 5),

Eur Radiol Table 3 Multivariate analysis for CT findings of PD-NETs compared with WD-NETs Characteristics LN metastasis Present Absent (Reference) Longitudinal location** Lower half Upper half (Reference)

Odds ratio

95% CI

P value*

56.67 1.00

5.17 - 621.02 -

.001

5.49 1.00

0.56 - 54.05 -

ns

CT computed tomography, PD-NETs poorly-differentiated, WD-NETs well-differentiated neuroendocrine tumours, CI confidence interval, LN lymph node, ns not significant *A P value<.05 is considered to indicate statistical significance **Upper half of the stomach includes cardia, fundus, and high body, while lower half of the stomach includes mid body, low body, and antrum

albeit with no statistical significance (140.31 ± 65.08 vs 104.28±53.47) (P=.299). However, the mean metastasis-toliver ratio in gastric NETs was significantly greater than that in gastric ADCs (2.15±0.74 vs 1.13±0.51, P=.015) (Table 6) (Fig 6). The optimized cut-off value for the metastasis-toliver ratio in differentiating hepatic metastases from gastric NETs and those from gastric ADCs was 1.42 with an AUC, sensitivity, and specificity of 0.873, 81.8 % (9/11), and 80.0 % (4/5), respectively (P=.002). In addition, peritoneal carcinomatosis was observed to be present solely in gastric ADCs (n=3), albeit without statistical significance (Table 4).

Discussion Although several radiologic reports investigating the imaging features of gastric NETs have been published, these studies were performed prior to 2010 in which newly standardized terminology was published by the World Health Organization (20, 21). In the past, the terms Bcarcinoid tumours,^ Bmalignant carcinoids,^ Bwell-differentiated neuroendocrine tumours,^ or Bwell-differentiated neuroendocrine carcinomas^ were confusingly and often interchangeably used. Such inconsistencies in the histological nomenclature of NETs among studies prior to 2010 have resulted in much confusion, making an exact comparison among the studies difficult. To the best of our knowledge, this is the first radiologic report describing the CT features of systematically classified gastric NETs based on 2010 WHO classification. Our study results showed that the presence of intact overlying mucosa with mucosal tenting was the sole CT discriminator of PD-NETs from ADCs. This finding is well in line with our hypothesis and previous pathologic findings (5, 13–16). We had initially hypothesized that an intact overlying mucosa or epithelium would be frequently found on CT in patients with gastric NETs whereas patients with gastric

ADCs would not have an intact mucosal layer, and that this would be an important differential CT clue for NETs. Our assumption was based on the fact that gastric NETs arise from enterochromaffin cells located at the lamina propria, not from the surface epithelium, sparing the overlying surface epithelium (5, 13–16), and small gastric NETs typically manifest as subepithelial nodules on endoscopy or sometimes on CT (20, 21). If the NETs become larger, although the surface epithelium may in fact become denuded, some intact epithelium would still exist at the border of the lesion resulting in the depiction of mucosal tenting with an intact overlying epithelium. In our study, 10 out of 18 patients with PD-NETs showed intact overlying mucosa with mucosal tenting regardless of tumour size compared to only 1 out of 19 gastric ADCs, thereby supporting our hypothesis. Our results also showed different features of metastatic LNs between the two groups, appearing larger in size and greater in necrotic component in PD-NETs than in ADCs. In clinical practice, the exact histological differentiation of PD-NETs from ADCs through endoscopic biopsy alone is often difficult even for expert pathologists owing to the poorly-differentiated nature of the tumours and due to the small obtainable biopsy samples. To the contrary, CT can show the entire intra-lesional morphology as well as characteristics of LN metastases, providing more comprehensive information to physicians. Therefore, considering the different treatment strategies and prognoses of these tumours, knowledge of differential CT features between PD-NETs and ADCs can be of great importance in clinical practice. In the present study, hepatic metastases from gastric NETs were observed to be significantly more hyper-attenuated than those from ADCs according to subjective analysis, which was also verified on objective analysis. In addition, the mean metastasis-to-liver ratio on the arterial phase of NETs was 2.15 which was significantly higher than the 1.13 of metastatic ADCs. This result well coincides with those of previous investigations (22) as well as our initial assumptions. Because NETs are commonly hypervascular tumours unlike ADCs, metastatic NETs may also show hypervascularity on arterial phase CT images. Compared to metastatic ADCs in which the 5-year survival is only 4% even with systemic chemotherapy (23), metastatic NETs show a quite optimistic 5-year survival rate of 50~80 % with aggressive surgical or non-surgical approaches such as somatostatin analogues, interferons, and TACE (24–26). Given that there is a definite survival benefit of surgical or liver-directed local treatments for metastatic NETs, our study results regarding the CT differentiation of hepatic metastasis between the two disease entities may be especially noteworthy. Indeed, in our study, 5 patients with metastatic NETs in the liver were treated through TACE and their mean survival (30.5 months) was longer than that (19 months) of the 4 patients with metastatic ADCs who were treated with systemic cytotoxic chemotherapy.

Eur Radiol Table 4 Prevalence of independent CT features between PD-NETs and ADCs

Characteristics

Tumour diagnosis

Tumour size (cm)

CT computed tomography, PDNETs poorly-differentiated neuroendocrine tumours, ADCs adenocarcinomas, LN lymph node, ns not significant *A P value<.05 is considered to indicate statistical significance **Upper half of the stomach includes cardia, fundus, and high body, while lower half of the stomach includes mid body, low body, and antrum ***16 out of 18 patients with PDNETs and 18 out of 19 patients with ADCs who were matched with PD-NETs had undergone multiphasic dynamic CT

Longitudinal location** Lower half Upper half Multiplicity Single Multiple Diffuse Tumour margin Well-defined Ill-defined Homogeneity Homogenous Heterogeneous Intact overlying mucosa with mucosal tenting Present Absent Dynamic enhancement pattern*** Persistent low attenuation Progressive enhancement pattern Wash-in/Wash-out pattern Persistent high attenuation LN metastasis Present Absent Tumor size (cm) Mean diameter of the largest LN (cm) LN necrosis Present Absent Japanese lymph node classification N1 N2 N3 Peritoneal carcinomatosis Present Absent

The distinction of PD-NETs from WD-NETs is one of the most important diagnostic steps in regard to these neoplasms as the biologic behaviour of the WD group is often rather indolent, whereas PD-NETs are highly aggressive. The therapeutic strategy also differs significantly between these two categories of tumours; patients with localized WD-NETs are treated through either endoscopic polypectomy (size<1 cm or number of lesions<5), antrectomy, or local excision (size> 1 cm or number of lesions>5) while patients with localized PD-NETs are treated with radical gastrectomy and extensive LN dissection (3). In our study, PD-NETs tended to be larger

P value*

PD-NETs

ADCs

5.39±2.32

5.83±2.52

ns

14 (77.8%) 4 (22.2%)

17 (89.5%) 2 (10.5%)

ns

18 (100%) 0 (0%) 0 (0%)

17 (89.5%) 0 (0%) 2 (10.5%)

ns

10 (55.6%) 8 (44.4%)

0 (0%) 19 (100%)

<.0001

13 (72.2%) 5 (27.8%)

11 (57.9%) 8 (42.1%)

ns

10 (55.6%) 8 (44.4%)

1 (5.3%) 18 (94.7%)

.001

2 (12.5%) 1 (6.3%) 0 (0%) 13 (81.2%)

0 (0%) 2 (11.1%) 0 (0%) 16 (88.9%)

ns

17 (94.4%) 1 (5.6%) 5.27±2.32 2.12±1.29

18 (94.7%) 1 (5.3%) 5.83±2.52 1.22±0.68

ns

10 (58.8%) 7 (41.2%)

4 (22.2%) 14 (77.8%)

12 (70.6%) 3 (17.6%) 2 (11.8%)

7 (38.9%) 5 (27.8%) 6 (33.3%)

ns

0 (0%) 18 (100%)

3 (15.8%) 16 (84.2%)

ns

ns .014 .027

in size, lower half located, more often solitary in number and more easily metastasized to LNs when compared with WDNETs on CT. Among them, metastasis to LNs was the most significant differentiator of PD-NETs from WD-NETs according to binary regression analysis. Thus, we believe that CT findings may provide additional confidence on the determination of the tumour differentiation, helping clinicians to select the most appropriate management strategy. Our study has several limitations. First, as our study was retrospectively designed, various MDCT scanners were used, resulting in differences in the parameters, degrees of gastric

Eur Radiol Table 5 Multivariate analysis for CT findings of PD-NETs compared with ADCs Characteristics Intact overlying mucosa with mucosal tenting Present Absent (Reference) Tumour margin Well-defined Ill-defined (Reference) LN necrosis Present Absent (Reference)

Odds ratio

95% CI

15.30 1.00

1.04 - 225.39 -

P value*

.047

1.00

-

ns

5.33 1.00

0.69 – 41.43 -

ns

CT computed tomography, PD-NETs poorly-differentiated neuroendocrine tumours; ADCs adenocarcinomas, CI confidence interval, ns not significant, LN lymph node * A P value<.05 is considered to indicate statistical significance

distension, patients’ positions, and doses of contrast media used. However, all patients in our study except for the two patients who had undergone MDCT had slice thicknesses of less than 5 mm, which may be acceptable for imaging interpretation. Second, there is a possibility of selection bias in our study, as we excluded cases with small gastric lesions which were not visualized on CT, but visualized on endoscopy only. Such a selection bias may be unavoidable, however, and the assessment of tiny gastric lesions remains a limitation of CT. Third, the number of patients in our study for hepatic metastases assessment was relatively small. Further studies with a larger number of patients are warranted to better evaluate the differences in CT findings between hepatic metastases from gastric NETs and from gastric ADCs. In conclusion, the presence of metastatic LNs and intact overlying mucosa with mucosal tenting can be useful CT Table 6 ADCs

Characteristics of hepatic metastases between NETs and

Characteristics

Tumour diagnosis NETs

Fig. 6 Dot graphs of metastasis-to-liver ratio between hepatic metastases from gastric neuroendocrine tumours (NETs) and those from gastric adenocarcinomas (ADCs). Mean metastasis-to-liver ratio of NETs (2.15) was significantly greater than that (1.13) of gastric ADCs

discriminators of PD-NETs from WD-NETs and gastric ADCs, respectively. In addition, a larger size and greater necrotic volume of metastatic LNs can favor LN metastases from PD-NETs over gastric ADCs. Finally, hepatic metastases from gastric NETs appear more hyper-attenuated than those from ADCs. Use of these CT findings may allow physicians to select the most appropriate management option. Acknowledgments The authors thank Chris Woo, B.A. for his English editorial assistance in preparing the manuscript. The scientific guarantor of this publication is Se Hyung Kim, Associate Professor, Department of Radiology, Seoul National University Hospital. The authors of this manuscript declare no relationships with any companies, whose products or services may be related to the subject matter of the article. This study has received funding by the Basic Science Research Program through the National Research Foundation of Korea [NRF] funded by the Ministry of Science, ICT & Future Planning [2013R1A1A3005937]. No complex statistical methods were necessary for this paper. Institutional Review Board approval was obtained. Written informed consent was waived by the Institutional Review Board. No study subjects or cohorts have been previously reported. Methodology: retrospective, observational, performed at one institution.

P value*

ADCs

References Subjective analysis Hypo-attenuated Hyper-attenuated Objective analysis

2 (18.2%) 9 (81.8%)

4 (80%) 1 (20%)

.036

Mean HU of peripheral 140.31±65.08 104.28±53.47 ns enhancing portion (arterial phase) Mean metastasis-to-liver 2.15±0.74 1.13±0.51 .015 ratio (arterial phase) NETs neuroendocrine tumours, ADCs adenocarcinomas, HU Hounsfield unit, ns not significant *A P value<.05 is considered to indicate statistical significance

1. Bosman FT, Carneiro F, Hruban RH, Theise ND (eds) (2010) WHO classification of tumours of the digestive system. IARC France, Lyon 2. Rindi G, Kloppel G, Alhman H et al (2006) TNM staging of foregut (neuro)endocrine tumors: a consensus proposal including a grading system. Virchows Archiv Int J Pathol 449:395–401 3. Christopoulos E. (2007) Gastric neuroendocrine tumors: biology and management. Ann Gastroenterol. 18 4. Scherübl H, Jensen RT, Cadiot G, Stölzel U, Klöppel G (2011) Management of early gastrointestinal neuroendocrine neoplasms. World J Gastrointest Endosc 3:133–139 5. Crosby DA, Donohoe CL, Fitzgerald L et al (2012) Gastric neuroendocrinetumours. Dig Surg 29:331–348

Eur Radiol 6. Caplin M, Hodgson H, Dhillon A et al (1998) Multimodality treatment for gastric carcinoid tumor with liver metastases. Am J Gastroenterol 93:1945–1948 7. Tomassetti P, Migliori M, Caletti GC, Fusaroli P, Corinaldesi R, Gullo L (2000) Treatment of type II gastric carcinoid tumors with somatostatin analogues. N Engl J Med 343:551–554 8. Ahlman H, Friman S, Cahlin C et al (2004) Liver transplantation for treatment of metastatic neuroendocrine tumors. Ann N Y Acad Sci 1014:265–269 9. Ruszniewski P, Rougier P, Roche A et al (1993) Hepatic arterial chemoembolization in patients with liver metastases of endocrine tumors. a prospective phase II study in 24 patients. Cancer 71: 2624–2630 10. Wessels FJ, Schell SR (2001) Radiofrequency ablation treatment of refractory carcinoid hepatic metastases. J Surg Res 95:8–12 11. Kim SH, Lee JM, Han JK et al (2005) Effect of adjusted positioning on gastric distention and fluid distribution during CT gastrography. AJR Am J Roentgenol 185:1180–1184 12. Kim JI, Kim YH, Lee KH et al (2013) Type-specific diagnosis and evaluation of longitudinal tumor extent of borrmann type iv gastric cancer: CT versus gastroscopy. Korean J Radiol 14:597–606 13. Rindi G, Bordi C, Rappel S, La Rosa S, Stolte M, Solcia E (1996) Gastric carcinoids and neuroendocrine carcinomas: pathogenesis, pathology, and behavior. World J Surg 20:168–172 1 4 . N e u r o e n d o c r i n e t u m o r. h t t p : / / e n . w i k i p e d i a . o rg / w i k i/ Neuroendocrine_tumor. Last modified on 8 May 2014, Visited on 1 June 2014 15. Bordi C, D'Adda T, Azzoni C, Ferraro G (1998) Pathogenesis of ECL cell tumors in humans. Yale J Biol Med 71:273 16. Modlin IM, Tang LH (1996) The gastric enterochromaffin-like cell: an enigmatic cellular link. Gastroenterology 111:783–810

17. Ahn HS, Kim SH, Kodera Y, Yang H-K (2013) Gastric cancer staging with radiologic imaging modalities and UICC staging system. Dig Surg 30:142–149 18. Park HS, Lee JM, Kim SH et al (2010) Three-dimensional MDCT for preoperative local staging of gastric cancer using gas and water distention methods: a retrospective cohort study. Am J Roentgenol 195: 1316–1323 19. Japanese Gastric Cancer Association (1998) Japanese classification of gastric carcinoma – 2nd English edition. Gastric Cancer 1:10–24 20. Binstock AJ, Johnson CD, Stephens DH, Lloyd RV, Fletcher JG (2001) Carcinoid tumors of the stomach: a clinical and radiographic study. AJR Am J Roentgenol 176:947–951 21. Chang S, Choi D, Lee SJ et al (2007) Neuroendocrine neoplasms of the gastrointestinal tract: classification, pathologic basis, and imaging features1. Radiograp Rev Public Radiol Soc North Am Inc 27:1667– 1679 22. Dromain C, de Baere T, Baudin E et al (2003) MR imaging of hepatic metastases caused by neuroendocrine tumors: comparing four techniques. Am J Roentgenol 180:121–128 23. Jemal A, Siegel R, Xu J, Ward E (2010) Cancer statistics, 2010. CA Cancer J Clin 60:277–300 24. Elias D, Lasser P, Ducreux M et al (2003) Liver resection (and associated extrahepatic resections) for metastatic well-differentiated endocrine tumors: a 15-year single center prospective study. Surgery 133:375–382 25. Harring TR, Nguyen NTN, Goss JA, O'Mahony CA (2011) Treatment of liver metastases in patients with neuroendocrine tumors: a comprehensive review. Int J Hepatol 2011:154541 26. Vogl TJ, Naguib NN, Zangos S, Eichler K, Hedayati A, Nour-Eldin N-EA (2009) Liver metastases of neuroendocrine carcinomas: interventional treatment via transarterial embolization, chemoembolization and thermal ablation. Eur J Radiol 72:517–528