- Email: [email protected]
Novel narrow-band imaging magnifying endoscopic classification for early gastric cancer
Digestive and Liver Disease 42 (2010) 704–708
Contents lists available at ScienceDirect
Digestive and Liver Disease journal homepage: www.elsevier.com/locate/dld
Digestive Endoscopy
Novel narrow-band imaging magnifying endoscopic classification for early gastric cancer Akira Yokoyama a,1 , Haruhiro Inoue a,∗ , Hitomi Minami a , Yoshiki Wada a , Yoshitaka Sato a , Hitoshi Satodate a , Shigeharu Hamatani b , Shin-ei Kudo a a b
Digestive Disease Center, Showa University Northern Yokohama Hospital, Yokohama, Japan Department of Pathology, Showa University Northern Yokohama Hospital, Yokohama, Japan
a r t i c l e
i n f o
Article history: Received 1 December 2009 Accepted 14 March 2010 Available online 11 May 2010 Keywords: Early gastric cancer Magnifying endoscopy Narrow-band imaging
a b s t r a c t Background: Narrow-band imaging magnifying endoscopy is widely used in Japan, but still there is no set of consistent guidelines for gastric lesions. Aims: To introduce a new narrow-band imaging magnifying endoscopic classification and report the accuracy of diagnosis in comparison to underlying histopathology of gastric lesions. Methods: Two hundred and fifty-seven consecutive patients with early gastric cancer lesions were enrolled into this study. Narrow-band imaging magnifying images were classified into four categories based on abnormal microvascular patterns and irregularities in the superficial glandular structure: finenetwork pattern, corkscrew pattern, intra-lobular loop pattern-1 and intra-lobular loop pattern-2. The narrow-band imaging magnifying endoscopic classification was compared with the histopathological findings. Results: Amongst the differentiated-type adenocarcinoma lesions, fine-network pattern, intra-lobular loop pattern-1, intra-lobular loop pattern-2 and corkscrew pattern were observed in 15.7%, 59.6%, 24.2% and 0.5%, respectively. Differentiated-type adenocarcinomas mainly exhibited fine-network pattern or intra-lobular loop pattern. In undifferentiated-type adenocarcinoma lesions, intra-lobular loop pattern2 and corkscrew pattern were observed in 41.2% and 58.8%, respectively. Therefore, undifferentiatedtype adenocarcinomas were all classified as intra-lobular loop pattern-2 and corkscrew pattern. The histopathological types were not equivalent with the narrow-band imaging magnifying classification categories (P < 0.001). Conclusions: The new narrow-band imaging magnifying classification that incorporates the intra-lobular loop pattern may be able to predict the histological subtype of most gastric carcinomas. © 2010 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved.
1. Introduction Yao and co-workers first reported unique magnifying endoscopic findings based mainly on the subepithelial microvascular (MV) architecture characteristics of differentiated-type early gastric cancer [1,2]. These findings differed depending upon the histological type. Differentiated-type adenocarcinomas were characterised by three main findings: disappearance of regular subepithelial capillary network (SECN), a demarcation line and irregular-microvascular pattern (IMVP). Undifferentiated-type adenocarcinomas were characterised by a reduced MV pattern
∗ Corresponding author at: Digestive Disease Center, Showa University Northern Yokohama Hospital, 35-1, Chigasaki-chuo, Tsuzuki, Yokohama, Kanagawa 2248503, Japan. Tel.: +81 45 949 7000; fax: +81 45 949 7263/7927. E-mail address: [email protected] (H. Inoue). 1 Tel.: +81 45 949 7000; fax: +81 45 949 7263.
[3]. In this study, we focus further on morphological changes in the surface glandular structure of cancerous tissue. Nakayoshi et al. first reported magnifying endoscopic classification of patterns based on the MV patterns of depressed-type early gastric cancer [4]. In their study, they demonstrated that fine-network pattern (FNP) was typical for differentiated-type early gastric cancer and corkscrew pattern (CSP) was typical for undifferentiated-type early gastric cancer. However, many lesions in their study showed an unclassified pattern. Our experience with magnification endoscopy suggests that neither FNP nor CSP are applicable to early gastric cancer lesions, especially not so in more than 70% of differentiated-type early gastric cancers. Therefore, we focussed our examination on unclear pattern, defining a new category corresponding to unclear pattern, called “ILL (intra-lobular loop pattern)” [5]. ILL has a villous glandular structure containing loop-like microvessels. We also introduced a new narrow-band imaging (NBI) magnifying endoscopic classification for not only depressed-type lesions but also for flat- and elevated-types. NBI magnifying images were clas-
1590-8658/$36.00 © 2010 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.dld.2010.03.013
A. Yokoyama et al. / Digestive and Liver Disease 42 (2010) 704–708
705
sified into four categories based on abnormal MV pattern and irregularity of superficial glandular structure. Conventional “FNP” and “CSP” were refined based on abnormal MV patterns and irregularity of superficial glandular structure. The purpose of this study was to compare the accuracy of our new NBI classification of gastric lesions with their underlying definitive histopathology.
2. Materials and methods The protocol was approved by the local medical ethics committee, and informed consent was obtained from all patients before the procedure. In this prospective study, we examined 257 lesions in 240 cases of early gastric cancer (223 differentiated-type and 34 undifferentiated-type) from October 2004 to September 2008. The mean age of the patients was 65 years (35–89 years). The male:female ratio was 2.6:1. All patients had endoscopic or surgical treatment at the Digestive Disease Center of Showa University Northern Yokohama Hospital and were classified into four groups before treatment. We observed the length and breadth of the lesions using magnifying endoscopy with NBI, paying special attention to MV pattern and surface glandular structure of each lesion. The examinations were conducted by four experienced endoscopists who were blinded to the histological findings. Endoscopic images were stored electronically and randomly allocated to two readers for evaluation. Sensitivity, specificity and diagnostic accuracy of each endoscopic modality were assessed with reference to histopathology. The Olympus-EVIS LUCERA SPECTRUM system and a highresolution magnifying upper gastrointestinal endoscope (GIFH260Z, Olympus, Tokyo, Japan) were used in all cases. The structure–enhancement function of the video processor was set to 6 or 8. A hood (MB-162, Olympus, Tokyo, Japan. ELASTIC TOUCH, Top, Tokyo, Japan) mounted on the tip of the scope enabled the optimal distance of 3 mm to be maintained between the mucosal surface and the tip of the scope under high-power magnification [1]. Initially the lesion was observed carefully without magnification. The entire lesion was then examined sequentially with both moderate and higher power magnification. This was to ensure that higher power magnification endoscopy was undertaken in the area precisely where the gastric cancer was located. This is an essential part of the assessment as under high-power magnification, the range of focus is narrow with small visual field that makes assessment of the relationship between the MV pattern and surface glandular structure difficult if high magnification is used without first utilising moderate magnification. The boundary between the area of both the MV pattern and the surface glandular structure using NBI magnifying endoscopy was determined, focussing on the “demarcation line” as recommended by Yao et al. [3]. Next, part of the lesion was observed under higher-power magnification by focussing on the “white zone” as recommended by Yagi et al. In this way, the surface glandular structure pattern of many lesions could be observed clearly without crystal violet staining [6]. If the surface glandular structure pattern could not be observed clearly, NBI with 0.05% crystal violet staining was used to visualise the surface glandular structure. Four types could be distinguished with respect to the MV patterns and surface glandular structures. If more than one pattern was observed in the lesion, the lesion was classified according to the predominant pattern visible. The endoscopic images were then compared with the histopathological findings. The histopathological findings were classified according to the Japanese classification of early gastric cancer [7]. Cases with an ulcer covered with corpus albicans were excluded as the MV pattern and irregularities in the surface glandular structure could not be detected.
Fig. 1. Narrow band imaging magnifying classification. (A) Fine network pattern (FNP) and (B) Corkscrew pattern (CSP).
3. Novel NBI magnifying endoscopic classification (Figs. 1 and 2) The magnified NBI images were classified into four categories based on the type of abnormal MV pattern and irregularities of the superficial glandular structure. FNP has a small glandular opening to the mucosal surface with an abnormal mesh-like MV structure that encircles each glandular opening. ILL-1 has a villous glandular structure containing loop-like microvessels. In ILL-2, the villous has begun to break apart. In CSP, almost all glandular structures have disappeared and there are numerous abnormal corkscrew-like vessels. 4. Statistics The SPSS for Windows Version 11.0 statistical software package (SPSS Inc, Chicago, III) was used for data analysis. The Yates 4 × 2 Chi-square test was used to compare our results with histopathological types based on NBI magnifying classification. P values less than 0.05 were considered to be statistically significant. 5. Results Two hundred and twenty-three differentiated-type adenocarcinoma lesions were examined. FNP was observed in 35 lesions (15.7%), ILL-1 in 133 lesions (59.6%), ILL-2 in 54 lesions (24.2%) and CSP in 1 lesion (0.5%). Differentiated-type adenocarcinomas mainly showed FNP or ILL with the exception of one lesion. More than 80% of differentiated-type adenocarcinomas were classified as ILL-1 or ILL-2. All 34 undifferentiated-type adenocarcinomas were classified as ILL-2 or CSP. ILL-2 was observed in 14 lesions (41.2%) and CSP in 20 lesions (58.8%). If the gastric cancer lesions are divided into histopathological types (differentiated-type adenocarcinomas and undifferentiated-type adenocarcinomas) according to our NBI magnifying classification, the NBI magnifying classification categories are not equivalent (P < 0.001; Table 1).
706
A. Yokoyama et al. / Digestive and Liver Disease 42 (2010) 704–708
Fig. 3. A case of elevated-type early gastric cancer (0-IIa) shows FNP. Conventional endoscopy (A) shows a mild red elevated lesion on the lesser curvature of the upper body in the stomach. NBI magnifying endoscopy on magnification (B, C) shows an abnormal mesh-like microvascular structure (FNP) on the elevated aspect. The pathological picture (D, HE staining) demonstrates well differentiated-type adenocarcinoma located in the lamina propria of the stomach.
Fig. 2. Narrow Band Imaging magnifying classification. (A) Intralobular Loop Pattern 1 (ILL-1) and (B) Intralobular Loop Pattern 2 (ILL-2).
In addition, we examined the predominant part of the morphology of the lesion using chromoendoscopy with 0.2% indigo carmine. We separated the lesions into two groups according to morphology: elevated-type early gastric cancer that consists of 0-I type, IIa type, and IIa + IIc type (predominantly IIa); and flat- and depressed-type early gastric cancer which consists of IIb type and IIc type. FNP was observed in six elevated-type early gastric cancer lesions (17.1%) (Fig. 3) and in 29 flat- and depressed-type early gastric cancer lesions (82.9%). ILL-1 was observed in 74 elevated-type early gastric cancer lesions (55.6%) and in 59 flat- and depressedtype early gastric cancer lesions (44.4%) (Fig. 4). ILL-2 was observed in 16 elevated-type early gastric cancer lesions (23.5%) and in 52 flat- and depressed-type early gastric cancer lesions (76.5%). CSP was observed in 1 elevated-type early gastric cancer lesion (4.8%) and 20 flat- and depressed-type early gastric cancer lesions (95.2%). Therefore, more than 80% of FNP, 75% of ILL-2, and 95% of CSP were observed in flat- and depressed-type early gastric cancer lesions. On the other hand, ILL-1 was observed in both elevated-type early gastric cancer lesions, and flat- and depressed-type early gastric cancer lesions. If the gastric cancer lesions are divided according to morphology (elevated-type early gastric cancer lesions and flatand depressed-type early gastric cancer lesions) using our NBI mag-
Fig. 4. A case of depressed early gastric cancer (0-IIc) shows ILL-1 on the posterior wall of the middle body in the stomach (A). NBI magnifying endoscopy (B) and magnifying endoscopy with crystal violet staining (C) show numerous microvessels that can be seen inside the surface glandular structures (ILL-1). The pathological picture (D, HE staining) demonstrates well differentiated-type adenocarcinoma located in the lamina propria of the stomach.
Table 1 Narrow Band Imaging (NBI) magnifying classification and histopathological types. Fine network pattern (FNP)
Differentiated-type (n = 223) Undifferentiated-type (n = 34)
Intralobular Loop Pattern 1 (ILL-1)
n
%
n
35 0
15.7 0
133 0
Intralobular Loop Pattern 2 (ILL-2) % 59.6 0
Corkscrew pattern (CSP)
n
%
n
%
54 14
24.2 41.2
1 20
0.5 58.8
Differentiated-type adenocarcinomas mainly showed FNP or ILL. More than 80% of differentiated-type adenocarcinomas were classified as ILL-1 or ILL-2. Undifferentiatedtype adenocarcinomas were all classified as ILL-2 or CSP. If the gastric cancer lesions are divided into histopathological types according our NBI magnifying classification, the NBI magnifying classification categories are not equivalent (P < 0.001). P values were determined using the Yates 4 × 2 Chi-square test.
A. Yokoyama et al. / Digestive and Liver Disease 42 (2010) 704–708
707
Table 2 Narrow Band Imaging (NBI) magnifying classification and forms. Elevated-type
Flat- and depressed-type
0-I
FNP (n = 35) ILL-1 (n = 133) ILL-2 (n = 68) CSP (n = 21)
0-IIa
0-IIa + IIc
0-IIb
0-IIc
n
%
n
%
n
%
n
%
n
%
0 13 1 0
0 9.8 1.5 0
4 47 7 0
11.4 35.3 10.3 0
2 14 8 1
5.7 10.5 11.8 4.8
1 4 0 0
2.9 3 0 0
28 55 52 20
80 41.4 76.5 95.2
More than 80% of Fine network pattern (FNP), more than 75% of Intralobular Loop Pattern 2 (ILL-2), and more than 95% of Corkscrew pattern (CSP) were observed in flat- and depressed-type cancer lesions. Intralobular Loop Pattern 1 (ILL-1) was observed in not only elevated-type cancer lesions, but also flat- and depressed-type cancer lesions. If the gastric cancer lesions are divided into forms according our NBI magnifying classification, the NBI magnifying classification categories are not equivalent (P < 0.001). P values were determined using the Yates 4 × 2 Chi-square test.
nifying classification, the NBI magnifying classification categories are not equivalent (P < 0.001; Table 2). 6. Discussion Despite widespread use of endoscopic resection of early gastric cancer in Japan [8–13], still there is no set of consistent guidelines for magnification endoscopy with respect to gastric lesions. There is no comparable classification equivalent to “Kudo’s Pit Pattern Classification in the colon” [14] and “IPCL Pattern Classification in the esophagus” [15,16]. Accurate evaluation with respect to the horizontal margins and determination of the underlying depth of invasion and histopathology of early gastric cancer is needed prior to treatment [17,18]. Yao et al. reported that NBI magnification endoscopy is useful for the determination of the horizontal margin particularly in differentiated-type adenocarcinomas [3]. This “demarcation line” is useful to identify the lateral spreading margin of differentiatedtype adenocarcinomas. However, we also need to determine the differentiation of neoplasia before evaluation of superficial spread as this determines further management. If an undifferentiatedtype adenocarcinoma is diagnosed, the lesion should be treated surgically. The endoscopic diagnosis for differentiation of gastric cancer using magnification endoscopy has been reported previously by Nakayoshi et al. [4]. They observed FNP in 66.1% of the 109 cases of differentiated-type adenocarcinoma. On the other hand, amongst the 56 cases of undifferentiated-type adenocarcinoma, CSP was observed in 85.7%. Unclassified patterns were present in 30.3% of differentiated-type adenocarcinomas in addition to 10.7% of undifferentiated-type adenocarcinomas. We have utilised a new NBI magnification endoscopic classification to differentiate not only depressed-type lesions, but also flatand elevated-type lesions without an “unclassified pattern”. Two patterns (ILL-1 and ILL-2) were added to enable more precise assessment of the superficial glandular structure. We observed that differentiated-type adenocarcinomas mainly demonstrated FNP and ILL with the exception of one lesion, with greater than 80% of differentiated-type adenocarcinomas classified as either ILL1 or ILL-2. Amongst the 34 undifferentiated-type adenocarcinoma lesions, ILL-2 was observed in 41.2%, and CSP in 58.8%. Therefore, undifferentiated-type adenocarcinomas were all classified as ILL2 or CSP. Conversely, FNP and ILL-1 were all differentiated-type adenocarcinomas. All CSP were undifferentiated-type adenocarcinomas, with the exception of one lesion. ILL-2 were mainly differentiated-type adenocarcinomas, but also included some undifferentiated-type adenocarcinomas. Therefore, FNP and ILL-1 predicted for differentiated-type adenocarcinomas, and CSP predicted for undifferentiated-type adenocarcinomas. ILL-2 mainly predicted for differentiated-type adenocarcinomas, although it may also be present in underlying undifferentiated-type adenocarcinomas, highlighting the importance of careful evaluation of
ILL-2. This may also account for the variations in pathological differentiation of gastric cancer and the existence of borderline cases presenting with characteristics of both differentiated and undifferentiated-type adenocarcinoma. Moreover, when a lesion with mixed-type differentiation is observed, the criteria for pathological diagnosis are not the same in Japan as they are in Europe and North America. In Japan, a pathological diagnosis is based on the size of the lesion, whereas in Europe and America, it depends on the grade of differentiation. Therefore, it is not surprising that endoscopic diagnosis is sometimes difficult since there are many borderline gastric cancer cases for which pathological diagnosis is not clear. This underlies one of the limitations of the study, as ILL-2 was common to borderline cases. ILL-2 was mainly differentiated-type adenocarcinomas but it also included some undifferentiated-type adenocarcinomas. Nonetheless, in ILL-2, it may be possible to predict histopathological types by focussing on the microvessels in the background of a villous glandular structure that shows signs of breaking apart. If corkscrew (CS)-like microvessels – like those observed with CSP – are observed, it may be possible to predict undifferentiated-type adenocarcinoma (Fig. 5). In our study, CS-like microvessels were observed in 64% of the undifferentiated-type adenocarcinoma ILL-2 lesions. In contrast, they were not observed in any of the differentiated-type adenocarcinomas. It is our opinion that the focus of future studies should
Fig. 5. A case of depressed early gastric cancer (0-IIc) shows ILL-2 on the posterior wall of the middle body in the stomach (A). NBI magnifying endoscopy (B) and magnifying endoscopy with indigo carmine (C) on magnification show villous glandular structure with signs of breaking apart from each other (ILL-2), and the microvessels in the background of a villous glandular structure that shows signs of breaking apart is CS-like microvessels. The pathological picture (D, HE staining) demonstrates undifferentiated-type adenocarcinoma located in the lamina propria of the stomach.
708
A. Yokoyama et al. / Digestive and Liver Disease 42 (2010) 704–708
be on better definition of ILL-2. With regard to the morphology of lesions, FNP was mainly observed in flat- and depressed-type early gastric cancer lesions. FNP was originally proposed by Nakayoshi et al. as a classification for depressed-type adenocarcinoma. However, we observed FNP in six elevated-type early gastric cancer lesions. Both CSP and ILL-2 were mainly observed in flat- and depressedtype early gastric cancer lesions. On the other hand, ILL-1 was observed in all types of lesions including elevated-type, flat- and depressed-type early gastric cancer. In both the elevated-type adenocarcinomas and the flat- and depressed-type adenocarcinomas, cases of ILL exceeded cases of FNP or CSP. Overall, more than 70% of early gastric cancers were classified as ILL-1 or ILL-2. Traditional management of gastric lesions has involved endoscopic evaluation of their morphology and histological confirmation of diagnosis after biopsy. The decision regarding endoscopic or surgical treatment has largely been based on these results. There is overwhelming interest in better and more objective endoscopic diagnosis in Japan largely owing to the high incidence of gastric cancer. Only a few Japanese endoscopists can claim confidence in the diagnosis of the extent of differentiation of gastric cancer using conventional endoscopy without magnification. The combination of NBI magnification and conventional endoscopy may assist in the formulation of a more objective and precise diagnosis. Conventional endoscopic examination shows that differentiated-type adenocarcinomas demonstrate rubor whilst undifferentiated-type adenocarcinomas are discoloured. The utilisation of NBI magnification endoscopy may help in the differentiation of different types of gastric cancer. Differentiated-type adenocarcinomas show abundant vascular patterns, “FNP” or “ILL”; whereas undifferentiated-type adenocarcinomas typically demonstrate poor vascular patterns, “CSP” or “ILL-2”. Therefore, the routine use of our novel NBI magnification classification based on the type of abnormal MV pattern and irregularities of superficial glandular structure may prove useful in the diagnosis of different types of gastric cancer lesions. In our study, 10% of all gastric cancer lesions showed mixed differentiation with both differentiated and undifferentiated types within the same lesion. Therefore, undertaking routine biopsy at random may incorrectly diagnose one type of differentiation over the other depending on the location of the random biopsy. It is not unusual to discover that biopsy of a suspected differentiated-type adenocarcinoma may harbour foci of an undifferentiated-type adenocarcinoma in some areas. If this is the case, then the management decision regarding treatment is altered. Hanaoka et al. reported that undifferentiated-type-predominant mixed-type gastric cancer with submucosal invasion carries a high risk of lymph node metastasis [19]. Our novel NBI magnification classification may assist in performing target “optical” biopsies in suspicious areas to identify lesions with a poorer prognosis. There are some limitations to our study with respect to the ILL2 subtype. Firstly, ILL-2 was influenced by previously performed biopsies. Secondly, ILL-2 is affected by histopathological types and also by depth of invasion. Some of these may be overcome by comparing our classification to three-dimensional configuration images. Nonetheless, we have demonstrated that magnification endoscopy and NBI does suggest a number of histological subtypes for early gastric cancer. It does not, however, identify submucosal cancer with overlying normal non-neoplastic mucosa. Despite its limitations, our NBI magnification classification system needs to
be subjected to further prospective studies as it may prove to be highly effective in the routine assessment and diagnosis of early gastric cancer. 7. Conclusion Our novel NBI magnification classification, which includes the ILL pattern may be able to accurately predict the histological subtype of most gastric carcinomas. Conflict of interest statement None declared. Acknowledgements We would like to express our gratitude to Mr Amyn Haji MA MBBChir MSc FRCS, London, for his invaluable support in the correction of this manuscript. References [1] Yao K, Oishi T, Matsui T, et al. Novel magnified endoscopic findings of microvascular architecture in intramucosal gastric cancer. Gastrointest Endosc 2002;56:279–84. [2] Yao K, Oishi T. Microgastroscopic findings of mucosal microvascular architecture as visualized by magnifying endoscopy. Dig Endosc 2001;13:S27–33. [3] Yao K, Yao T, Iwashita A. Determining the horizontal extent of early gastric carcinoma: two modern techniques based on differences in the mucosal microvascular architecture and density between carcinoma and noncarcinomatous mucosa. Dig Endosc 2002;14:S83–7. [4] Nakayoshi T, Tajiri H, Matsuda K, et al. Narrow-band imaging system with magnifying endoscopy for early gastric cancer: correlation of vascular pattern with histopathology (including video). Endoscopy 2004;36:1080–4. [5] Inoue H, Kodama K, Minami H, et al. NBI magnifying endoscopic classification using crystal violet staining. Nippon Rinsho 2008;66:1023–7. [6] Yagi K, Nakamura A, Sekine A, et al. Magnifying endoscopy with narrow band imaging for early differentiated gastric adenocarcinoma. Dig Endosc 2008;20:115–22. [7] Japanese Gastric Cancer Association. Japanese classification of gastric carcinoma: 2nd English edition. Gastric Cancer 1998;1:10–24. [8] Karita M, Tada M, Okita K. The successive strip biopsy partial resection technique for large early gastric and colon cancers. Gastrointest Endosc 1992;38:174–8. [9] Inoue H, Takeshita K, Hori H, et al. Endoscopic mucosal resection with cap-fitted panendoscope for esophagus, stomach, and colon mucosal lesions. Gastrointest Endosc 1993;39:58–62. [10] Torii A, Sasaki M, Kajiyama T, et al. Endoscopic aspiration mucosectomy as curative endoscopic surgery: analysis of 24 cases of early gastric cancer. Gastrointest Endosc 1995;42:475–9. [11] Hanazaki K, Wakabayashi M, Sodeyama H, et al. Clinicopathologic features of submucosal carcinoma of the stomach. J Clin Gastroenterol 1997;24:150–5. [12] Rembacken BJ, Gotoda T, Fujii T, et al. Endoscopic mucosal resection. Endoscopy 2001;33:709–18. [13] Soetikno RM, Gotoda T, Nakanishi Y, et al. Endoscopic mucosal resection. Gastrointest Endosc 2003;57:2003. [14] Kudo S, Tamura S, Nakajima T, et al. Diagnosis of colorectal tumorous lesions by magnifying endoscopy. Gastrointest Endosc 1996;44:8–14. [15] Inoue H, Honda T, Yoshida T, et al. Ultra-high magnification endoscopy of normal esophageal mucosa. Dig Endosc 1996;8:134–8. [16] Inoue H, Honda T, Nagai K, et al. Ultra-high magnification endoscopic observation of carcinoma in situ. Dig Endosc 1997;9:16–8. [17] Gotoda T, Yanagisawa A, Sasako M, et al. Incidence of lymph node metastasis from early gastric cancer: estimation with a large number of cases at two large centers. Gastric Cancer 2000;3:219–25. [18] Gotoda T, Sasako M, Ono H, et al. Evaluation of necessity for gastrectomy with lympho node dissection for patients with submucosal invasive gastric cancer. Br J Surg 2001;88:444–9. [19] Hanaoka N, Tanabe S, Mikami T, et al. Mixed-histologic-type submucosal invasive gastric cancer as a risk factor for lymph node metastasis: feasibility of endoscopic submucosal dissection. Endoscopy 2009;41:427–32.
Contents lists available at ScienceDirect
Digestive and Liver Disease journal homepage: www.elsevier.com/locate/dld
Digestive Endoscopy
Novel narrow-band imaging magnifying endoscopic classification for early gastric cancer Akira Yokoyama a,1 , Haruhiro Inoue a,∗ , Hitomi Minami a , Yoshiki Wada a , Yoshitaka Sato a , Hitoshi Satodate a , Shigeharu Hamatani b , Shin-ei Kudo a a b
Digestive Disease Center, Showa University Northern Yokohama Hospital, Yokohama, Japan Department of Pathology, Showa University Northern Yokohama Hospital, Yokohama, Japan
a r t i c l e
i n f o
Article history: Received 1 December 2009 Accepted 14 March 2010 Available online 11 May 2010 Keywords: Early gastric cancer Magnifying endoscopy Narrow-band imaging
a b s t r a c t Background: Narrow-band imaging magnifying endoscopy is widely used in Japan, but still there is no set of consistent guidelines for gastric lesions. Aims: To introduce a new narrow-band imaging magnifying endoscopic classification and report the accuracy of diagnosis in comparison to underlying histopathology of gastric lesions. Methods: Two hundred and fifty-seven consecutive patients with early gastric cancer lesions were enrolled into this study. Narrow-band imaging magnifying images were classified into four categories based on abnormal microvascular patterns and irregularities in the superficial glandular structure: finenetwork pattern, corkscrew pattern, intra-lobular loop pattern-1 and intra-lobular loop pattern-2. The narrow-band imaging magnifying endoscopic classification was compared with the histopathological findings. Results: Amongst the differentiated-type adenocarcinoma lesions, fine-network pattern, intra-lobular loop pattern-1, intra-lobular loop pattern-2 and corkscrew pattern were observed in 15.7%, 59.6%, 24.2% and 0.5%, respectively. Differentiated-type adenocarcinomas mainly exhibited fine-network pattern or intra-lobular loop pattern. In undifferentiated-type adenocarcinoma lesions, intra-lobular loop pattern2 and corkscrew pattern were observed in 41.2% and 58.8%, respectively. Therefore, undifferentiatedtype adenocarcinomas were all classified as intra-lobular loop pattern-2 and corkscrew pattern. The histopathological types were not equivalent with the narrow-band imaging magnifying classification categories (P < 0.001). Conclusions: The new narrow-band imaging magnifying classification that incorporates the intra-lobular loop pattern may be able to predict the histological subtype of most gastric carcinomas. © 2010 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved.
1. Introduction Yao and co-workers first reported unique magnifying endoscopic findings based mainly on the subepithelial microvascular (MV) architecture characteristics of differentiated-type early gastric cancer [1,2]. These findings differed depending upon the histological type. Differentiated-type adenocarcinomas were characterised by three main findings: disappearance of regular subepithelial capillary network (SECN), a demarcation line and irregular-microvascular pattern (IMVP). Undifferentiated-type adenocarcinomas were characterised by a reduced MV pattern
∗ Corresponding author at: Digestive Disease Center, Showa University Northern Yokohama Hospital, 35-1, Chigasaki-chuo, Tsuzuki, Yokohama, Kanagawa 2248503, Japan. Tel.: +81 45 949 7000; fax: +81 45 949 7263/7927. E-mail address: [email protected] (H. Inoue). 1 Tel.: +81 45 949 7000; fax: +81 45 949 7263.
[3]. In this study, we focus further on morphological changes in the surface glandular structure of cancerous tissue. Nakayoshi et al. first reported magnifying endoscopic classification of patterns based on the MV patterns of depressed-type early gastric cancer [4]. In their study, they demonstrated that fine-network pattern (FNP) was typical for differentiated-type early gastric cancer and corkscrew pattern (CSP) was typical for undifferentiated-type early gastric cancer. However, many lesions in their study showed an unclassified pattern. Our experience with magnification endoscopy suggests that neither FNP nor CSP are applicable to early gastric cancer lesions, especially not so in more than 70% of differentiated-type early gastric cancers. Therefore, we focussed our examination on unclear pattern, defining a new category corresponding to unclear pattern, called “ILL (intra-lobular loop pattern)” [5]. ILL has a villous glandular structure containing loop-like microvessels. We also introduced a new narrow-band imaging (NBI) magnifying endoscopic classification for not only depressed-type lesions but also for flat- and elevated-types. NBI magnifying images were clas-
1590-8658/$36.00 © 2010 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.dld.2010.03.013
A. Yokoyama et al. / Digestive and Liver Disease 42 (2010) 704–708
705
sified into four categories based on abnormal MV pattern and irregularity of superficial glandular structure. Conventional “FNP” and “CSP” were refined based on abnormal MV patterns and irregularity of superficial glandular structure. The purpose of this study was to compare the accuracy of our new NBI classification of gastric lesions with their underlying definitive histopathology.
2. Materials and methods The protocol was approved by the local medical ethics committee, and informed consent was obtained from all patients before the procedure. In this prospective study, we examined 257 lesions in 240 cases of early gastric cancer (223 differentiated-type and 34 undifferentiated-type) from October 2004 to September 2008. The mean age of the patients was 65 years (35–89 years). The male:female ratio was 2.6:1. All patients had endoscopic or surgical treatment at the Digestive Disease Center of Showa University Northern Yokohama Hospital and were classified into four groups before treatment. We observed the length and breadth of the lesions using magnifying endoscopy with NBI, paying special attention to MV pattern and surface glandular structure of each lesion. The examinations were conducted by four experienced endoscopists who were blinded to the histological findings. Endoscopic images were stored electronically and randomly allocated to two readers for evaluation. Sensitivity, specificity and diagnostic accuracy of each endoscopic modality were assessed with reference to histopathology. The Olympus-EVIS LUCERA SPECTRUM system and a highresolution magnifying upper gastrointestinal endoscope (GIFH260Z, Olympus, Tokyo, Japan) were used in all cases. The structure–enhancement function of the video processor was set to 6 or 8. A hood (MB-162, Olympus, Tokyo, Japan. ELASTIC TOUCH, Top, Tokyo, Japan) mounted on the tip of the scope enabled the optimal distance of 3 mm to be maintained between the mucosal surface and the tip of the scope under high-power magnification [1]. Initially the lesion was observed carefully without magnification. The entire lesion was then examined sequentially with both moderate and higher power magnification. This was to ensure that higher power magnification endoscopy was undertaken in the area precisely where the gastric cancer was located. This is an essential part of the assessment as under high-power magnification, the range of focus is narrow with small visual field that makes assessment of the relationship between the MV pattern and surface glandular structure difficult if high magnification is used without first utilising moderate magnification. The boundary between the area of both the MV pattern and the surface glandular structure using NBI magnifying endoscopy was determined, focussing on the “demarcation line” as recommended by Yao et al. [3]. Next, part of the lesion was observed under higher-power magnification by focussing on the “white zone” as recommended by Yagi et al. In this way, the surface glandular structure pattern of many lesions could be observed clearly without crystal violet staining [6]. If the surface glandular structure pattern could not be observed clearly, NBI with 0.05% crystal violet staining was used to visualise the surface glandular structure. Four types could be distinguished with respect to the MV patterns and surface glandular structures. If more than one pattern was observed in the lesion, the lesion was classified according to the predominant pattern visible. The endoscopic images were then compared with the histopathological findings. The histopathological findings were classified according to the Japanese classification of early gastric cancer [7]. Cases with an ulcer covered with corpus albicans were excluded as the MV pattern and irregularities in the surface glandular structure could not be detected.
Fig. 1. Narrow band imaging magnifying classification. (A) Fine network pattern (FNP) and (B) Corkscrew pattern (CSP).
3. Novel NBI magnifying endoscopic classification (Figs. 1 and 2) The magnified NBI images were classified into four categories based on the type of abnormal MV pattern and irregularities of the superficial glandular structure. FNP has a small glandular opening to the mucosal surface with an abnormal mesh-like MV structure that encircles each glandular opening. ILL-1 has a villous glandular structure containing loop-like microvessels. In ILL-2, the villous has begun to break apart. In CSP, almost all glandular structures have disappeared and there are numerous abnormal corkscrew-like vessels. 4. Statistics The SPSS for Windows Version 11.0 statistical software package (SPSS Inc, Chicago, III) was used for data analysis. The Yates 4 × 2 Chi-square test was used to compare our results with histopathological types based on NBI magnifying classification. P values less than 0.05 were considered to be statistically significant. 5. Results Two hundred and twenty-three differentiated-type adenocarcinoma lesions were examined. FNP was observed in 35 lesions (15.7%), ILL-1 in 133 lesions (59.6%), ILL-2 in 54 lesions (24.2%) and CSP in 1 lesion (0.5%). Differentiated-type adenocarcinomas mainly showed FNP or ILL with the exception of one lesion. More than 80% of differentiated-type adenocarcinomas were classified as ILL-1 or ILL-2. All 34 undifferentiated-type adenocarcinomas were classified as ILL-2 or CSP. ILL-2 was observed in 14 lesions (41.2%) and CSP in 20 lesions (58.8%). If the gastric cancer lesions are divided into histopathological types (differentiated-type adenocarcinomas and undifferentiated-type adenocarcinomas) according to our NBI magnifying classification, the NBI magnifying classification categories are not equivalent (P < 0.001; Table 1).
706
A. Yokoyama et al. / Digestive and Liver Disease 42 (2010) 704–708
Fig. 3. A case of elevated-type early gastric cancer (0-IIa) shows FNP. Conventional endoscopy (A) shows a mild red elevated lesion on the lesser curvature of the upper body in the stomach. NBI magnifying endoscopy on magnification (B, C) shows an abnormal mesh-like microvascular structure (FNP) on the elevated aspect. The pathological picture (D, HE staining) demonstrates well differentiated-type adenocarcinoma located in the lamina propria of the stomach.
Fig. 2. Narrow Band Imaging magnifying classification. (A) Intralobular Loop Pattern 1 (ILL-1) and (B) Intralobular Loop Pattern 2 (ILL-2).
In addition, we examined the predominant part of the morphology of the lesion using chromoendoscopy with 0.2% indigo carmine. We separated the lesions into two groups according to morphology: elevated-type early gastric cancer that consists of 0-I type, IIa type, and IIa + IIc type (predominantly IIa); and flat- and depressed-type early gastric cancer which consists of IIb type and IIc type. FNP was observed in six elevated-type early gastric cancer lesions (17.1%) (Fig. 3) and in 29 flat- and depressed-type early gastric cancer lesions (82.9%). ILL-1 was observed in 74 elevated-type early gastric cancer lesions (55.6%) and in 59 flat- and depressedtype early gastric cancer lesions (44.4%) (Fig. 4). ILL-2 was observed in 16 elevated-type early gastric cancer lesions (23.5%) and in 52 flat- and depressed-type early gastric cancer lesions (76.5%). CSP was observed in 1 elevated-type early gastric cancer lesion (4.8%) and 20 flat- and depressed-type early gastric cancer lesions (95.2%). Therefore, more than 80% of FNP, 75% of ILL-2, and 95% of CSP were observed in flat- and depressed-type early gastric cancer lesions. On the other hand, ILL-1 was observed in both elevated-type early gastric cancer lesions, and flat- and depressed-type early gastric cancer lesions. If the gastric cancer lesions are divided according to morphology (elevated-type early gastric cancer lesions and flatand depressed-type early gastric cancer lesions) using our NBI mag-
Fig. 4. A case of depressed early gastric cancer (0-IIc) shows ILL-1 on the posterior wall of the middle body in the stomach (A). NBI magnifying endoscopy (B) and magnifying endoscopy with crystal violet staining (C) show numerous microvessels that can be seen inside the surface glandular structures (ILL-1). The pathological picture (D, HE staining) demonstrates well differentiated-type adenocarcinoma located in the lamina propria of the stomach.
Table 1 Narrow Band Imaging (NBI) magnifying classification and histopathological types. Fine network pattern (FNP)
Differentiated-type (n = 223) Undifferentiated-type (n = 34)
Intralobular Loop Pattern 1 (ILL-1)
n
%
n
35 0
15.7 0
133 0
Intralobular Loop Pattern 2 (ILL-2) % 59.6 0
Corkscrew pattern (CSP)
n
%
n
%
54 14
24.2 41.2
1 20
0.5 58.8
Differentiated-type adenocarcinomas mainly showed FNP or ILL. More than 80% of differentiated-type adenocarcinomas were classified as ILL-1 or ILL-2. Undifferentiatedtype adenocarcinomas were all classified as ILL-2 or CSP. If the gastric cancer lesions are divided into histopathological types according our NBI magnifying classification, the NBI magnifying classification categories are not equivalent (P < 0.001). P values were determined using the Yates 4 × 2 Chi-square test.
A. Yokoyama et al. / Digestive and Liver Disease 42 (2010) 704–708
707
Table 2 Narrow Band Imaging (NBI) magnifying classification and forms. Elevated-type
Flat- and depressed-type
0-I
FNP (n = 35) ILL-1 (n = 133) ILL-2 (n = 68) CSP (n = 21)
0-IIa
0-IIa + IIc
0-IIb
0-IIc
n
%
n
%
n
%
n
%
n
%
0 13 1 0
0 9.8 1.5 0
4 47 7 0
11.4 35.3 10.3 0
2 14 8 1
5.7 10.5 11.8 4.8
1 4 0 0
2.9 3 0 0
28 55 52 20
80 41.4 76.5 95.2
More than 80% of Fine network pattern (FNP), more than 75% of Intralobular Loop Pattern 2 (ILL-2), and more than 95% of Corkscrew pattern (CSP) were observed in flat- and depressed-type cancer lesions. Intralobular Loop Pattern 1 (ILL-1) was observed in not only elevated-type cancer lesions, but also flat- and depressed-type cancer lesions. If the gastric cancer lesions are divided into forms according our NBI magnifying classification, the NBI magnifying classification categories are not equivalent (P < 0.001). P values were determined using the Yates 4 × 2 Chi-square test.
nifying classification, the NBI magnifying classification categories are not equivalent (P < 0.001; Table 2). 6. Discussion Despite widespread use of endoscopic resection of early gastric cancer in Japan [8–13], still there is no set of consistent guidelines for magnification endoscopy with respect to gastric lesions. There is no comparable classification equivalent to “Kudo’s Pit Pattern Classification in the colon” [14] and “IPCL Pattern Classification in the esophagus” [15,16]. Accurate evaluation with respect to the horizontal margins and determination of the underlying depth of invasion and histopathology of early gastric cancer is needed prior to treatment [17,18]. Yao et al. reported that NBI magnification endoscopy is useful for the determination of the horizontal margin particularly in differentiated-type adenocarcinomas [3]. This “demarcation line” is useful to identify the lateral spreading margin of differentiatedtype adenocarcinomas. However, we also need to determine the differentiation of neoplasia before evaluation of superficial spread as this determines further management. If an undifferentiatedtype adenocarcinoma is diagnosed, the lesion should be treated surgically. The endoscopic diagnosis for differentiation of gastric cancer using magnification endoscopy has been reported previously by Nakayoshi et al. [4]. They observed FNP in 66.1% of the 109 cases of differentiated-type adenocarcinoma. On the other hand, amongst the 56 cases of undifferentiated-type adenocarcinoma, CSP was observed in 85.7%. Unclassified patterns were present in 30.3% of differentiated-type adenocarcinomas in addition to 10.7% of undifferentiated-type adenocarcinomas. We have utilised a new NBI magnification endoscopic classification to differentiate not only depressed-type lesions, but also flatand elevated-type lesions without an “unclassified pattern”. Two patterns (ILL-1 and ILL-2) were added to enable more precise assessment of the superficial glandular structure. We observed that differentiated-type adenocarcinomas mainly demonstrated FNP and ILL with the exception of one lesion, with greater than 80% of differentiated-type adenocarcinomas classified as either ILL1 or ILL-2. Amongst the 34 undifferentiated-type adenocarcinoma lesions, ILL-2 was observed in 41.2%, and CSP in 58.8%. Therefore, undifferentiated-type adenocarcinomas were all classified as ILL2 or CSP. Conversely, FNP and ILL-1 were all differentiated-type adenocarcinomas. All CSP were undifferentiated-type adenocarcinomas, with the exception of one lesion. ILL-2 were mainly differentiated-type adenocarcinomas, but also included some undifferentiated-type adenocarcinomas. Therefore, FNP and ILL-1 predicted for differentiated-type adenocarcinomas, and CSP predicted for undifferentiated-type adenocarcinomas. ILL-2 mainly predicted for differentiated-type adenocarcinomas, although it may also be present in underlying undifferentiated-type adenocarcinomas, highlighting the importance of careful evaluation of
ILL-2. This may also account for the variations in pathological differentiation of gastric cancer and the existence of borderline cases presenting with characteristics of both differentiated and undifferentiated-type adenocarcinoma. Moreover, when a lesion with mixed-type differentiation is observed, the criteria for pathological diagnosis are not the same in Japan as they are in Europe and North America. In Japan, a pathological diagnosis is based on the size of the lesion, whereas in Europe and America, it depends on the grade of differentiation. Therefore, it is not surprising that endoscopic diagnosis is sometimes difficult since there are many borderline gastric cancer cases for which pathological diagnosis is not clear. This underlies one of the limitations of the study, as ILL-2 was common to borderline cases. ILL-2 was mainly differentiated-type adenocarcinomas but it also included some undifferentiated-type adenocarcinomas. Nonetheless, in ILL-2, it may be possible to predict histopathological types by focussing on the microvessels in the background of a villous glandular structure that shows signs of breaking apart. If corkscrew (CS)-like microvessels – like those observed with CSP – are observed, it may be possible to predict undifferentiated-type adenocarcinoma (Fig. 5). In our study, CS-like microvessels were observed in 64% of the undifferentiated-type adenocarcinoma ILL-2 lesions. In contrast, they were not observed in any of the differentiated-type adenocarcinomas. It is our opinion that the focus of future studies should
Fig. 5. A case of depressed early gastric cancer (0-IIc) shows ILL-2 on the posterior wall of the middle body in the stomach (A). NBI magnifying endoscopy (B) and magnifying endoscopy with indigo carmine (C) on magnification show villous glandular structure with signs of breaking apart from each other (ILL-2), and the microvessels in the background of a villous glandular structure that shows signs of breaking apart is CS-like microvessels. The pathological picture (D, HE staining) demonstrates undifferentiated-type adenocarcinoma located in the lamina propria of the stomach.
708
A. Yokoyama et al. / Digestive and Liver Disease 42 (2010) 704–708
be on better definition of ILL-2. With regard to the morphology of lesions, FNP was mainly observed in flat- and depressed-type early gastric cancer lesions. FNP was originally proposed by Nakayoshi et al. as a classification for depressed-type adenocarcinoma. However, we observed FNP in six elevated-type early gastric cancer lesions. Both CSP and ILL-2 were mainly observed in flat- and depressedtype early gastric cancer lesions. On the other hand, ILL-1 was observed in all types of lesions including elevated-type, flat- and depressed-type early gastric cancer. In both the elevated-type adenocarcinomas and the flat- and depressed-type adenocarcinomas, cases of ILL exceeded cases of FNP or CSP. Overall, more than 70% of early gastric cancers were classified as ILL-1 or ILL-2. Traditional management of gastric lesions has involved endoscopic evaluation of their morphology and histological confirmation of diagnosis after biopsy. The decision regarding endoscopic or surgical treatment has largely been based on these results. There is overwhelming interest in better and more objective endoscopic diagnosis in Japan largely owing to the high incidence of gastric cancer. Only a few Japanese endoscopists can claim confidence in the diagnosis of the extent of differentiation of gastric cancer using conventional endoscopy without magnification. The combination of NBI magnification and conventional endoscopy may assist in the formulation of a more objective and precise diagnosis. Conventional endoscopic examination shows that differentiated-type adenocarcinomas demonstrate rubor whilst undifferentiated-type adenocarcinomas are discoloured. The utilisation of NBI magnification endoscopy may help in the differentiation of different types of gastric cancer. Differentiated-type adenocarcinomas show abundant vascular patterns, “FNP” or “ILL”; whereas undifferentiated-type adenocarcinomas typically demonstrate poor vascular patterns, “CSP” or “ILL-2”. Therefore, the routine use of our novel NBI magnification classification based on the type of abnormal MV pattern and irregularities of superficial glandular structure may prove useful in the diagnosis of different types of gastric cancer lesions. In our study, 10% of all gastric cancer lesions showed mixed differentiation with both differentiated and undifferentiated types within the same lesion. Therefore, undertaking routine biopsy at random may incorrectly diagnose one type of differentiation over the other depending on the location of the random biopsy. It is not unusual to discover that biopsy of a suspected differentiated-type adenocarcinoma may harbour foci of an undifferentiated-type adenocarcinoma in some areas. If this is the case, then the management decision regarding treatment is altered. Hanaoka et al. reported that undifferentiated-type-predominant mixed-type gastric cancer with submucosal invasion carries a high risk of lymph node metastasis [19]. Our novel NBI magnification classification may assist in performing target “optical” biopsies in suspicious areas to identify lesions with a poorer prognosis. There are some limitations to our study with respect to the ILL2 subtype. Firstly, ILL-2 was influenced by previously performed biopsies. Secondly, ILL-2 is affected by histopathological types and also by depth of invasion. Some of these may be overcome by comparing our classification to three-dimensional configuration images. Nonetheless, we have demonstrated that magnification endoscopy and NBI does suggest a number of histological subtypes for early gastric cancer. It does not, however, identify submucosal cancer with overlying normal non-neoplastic mucosa. Despite its limitations, our NBI magnification classification system needs to
be subjected to further prospective studies as it may prove to be highly effective in the routine assessment and diagnosis of early gastric cancer. 7. Conclusion Our novel NBI magnification classification, which includes the ILL pattern may be able to accurately predict the histological subtype of most gastric carcinomas. Conflict of interest statement None declared. Acknowledgements We would like to express our gratitude to Mr Amyn Haji MA MBBChir MSc FRCS, London, for his invaluable support in the correction of this manuscript. References [1] Yao K, Oishi T, Matsui T, et al. Novel magnified endoscopic findings of microvascular architecture in intramucosal gastric cancer. Gastrointest Endosc 2002;56:279–84. [2] Yao K, Oishi T. Microgastroscopic findings of mucosal microvascular architecture as visualized by magnifying endoscopy. Dig Endosc 2001;13:S27–33. [3] Yao K, Yao T, Iwashita A. Determining the horizontal extent of early gastric carcinoma: two modern techniques based on differences in the mucosal microvascular architecture and density between carcinoma and noncarcinomatous mucosa. Dig Endosc 2002;14:S83–7. [4] Nakayoshi T, Tajiri H, Matsuda K, et al. Narrow-band imaging system with magnifying endoscopy for early gastric cancer: correlation of vascular pattern with histopathology (including video). Endoscopy 2004;36:1080–4. [5] Inoue H, Kodama K, Minami H, et al. NBI magnifying endoscopic classification using crystal violet staining. Nippon Rinsho 2008;66:1023–7. [6] Yagi K, Nakamura A, Sekine A, et al. Magnifying endoscopy with narrow band imaging for early differentiated gastric adenocarcinoma. Dig Endosc 2008;20:115–22. [7] Japanese Gastric Cancer Association. Japanese classification of gastric carcinoma: 2nd English edition. Gastric Cancer 1998;1:10–24. [8] Karita M, Tada M, Okita K. The successive strip biopsy partial resection technique for large early gastric and colon cancers. Gastrointest Endosc 1992;38:174–8. [9] Inoue H, Takeshita K, Hori H, et al. Endoscopic mucosal resection with cap-fitted panendoscope for esophagus, stomach, and colon mucosal lesions. Gastrointest Endosc 1993;39:58–62. [10] Torii A, Sasaki M, Kajiyama T, et al. Endoscopic aspiration mucosectomy as curative endoscopic surgery: analysis of 24 cases of early gastric cancer. Gastrointest Endosc 1995;42:475–9. [11] Hanazaki K, Wakabayashi M, Sodeyama H, et al. Clinicopathologic features of submucosal carcinoma of the stomach. J Clin Gastroenterol 1997;24:150–5. [12] Rembacken BJ, Gotoda T, Fujii T, et al. Endoscopic mucosal resection. Endoscopy 2001;33:709–18. [13] Soetikno RM, Gotoda T, Nakanishi Y, et al. Endoscopic mucosal resection. Gastrointest Endosc 2003;57:2003. [14] Kudo S, Tamura S, Nakajima T, et al. Diagnosis of colorectal tumorous lesions by magnifying endoscopy. Gastrointest Endosc 1996;44:8–14. [15] Inoue H, Honda T, Yoshida T, et al. Ultra-high magnification endoscopy of normal esophageal mucosa. Dig Endosc 1996;8:134–8. [16] Inoue H, Honda T, Nagai K, et al. Ultra-high magnification endoscopic observation of carcinoma in situ. Dig Endosc 1997;9:16–8. [17] Gotoda T, Yanagisawa A, Sasako M, et al. Incidence of lymph node metastasis from early gastric cancer: estimation with a large number of cases at two large centers. Gastric Cancer 2000;3:219–25. [18] Gotoda T, Sasako M, Ono H, et al. Evaluation of necessity for gastrectomy with lympho node dissection for patients with submucosal invasive gastric cancer. Br J Surg 2001;88:444–9. [19] Hanaoka N, Tanabe S, Mikami T, et al. Mixed-histologic-type submucosal invasive gastric cancer as a risk factor for lymph node metastasis: feasibility of endoscopic submucosal dissection. Endoscopy 2009;41:427–32.