The esophageal submucosal glands are considered to be a continuation

248 Significance of Involvement by Squamous Cell Carcinoma of the Ducts of Esophageal Submucosal Glands Analysis of 201 Surgically Resected Superficial Squamous Cell Carcinomas Yusuke Tajima, M.D. 1,2 Yukihiro Nakanishi, M.D. 3 Yuji Tachimori, M.D. 4 Hoichi Kato, M.D. 4 Hiroshi Watanabe, M.D. 4 Hajime Yamaguchi, M.D. 5 Kimio Yoshimura, M.D. 6 Mitsuo Kusano, M.D. 2 Tadakazu Shimoda, M.D. 1 1 Clinical Laboratory Division, National Cancer Center Hospital and Research Institute, Tokyo, Japan. 2 Second Department of Surgery, Faculty of Medicine, Showa University, Tokyo, Japan. 3 Pathology Division, National Cancer Center Research Institute, Tokyo, Japan. 4 Department of Surgery, National Cancer Center Hospital, Tokyo, Japan. 5 Department of Internal Medicine, National Cancer Center Hospital, Tokyo, Japan. 6 Cancer Information and Epidemiology Division, National Cancer Center Research Institute, Tokyo, Japan. BACKGROUND. Ductal involvement (DI) is often observed in superficial squamous cell carcinoma of the esophagus (SSCCE), defined as carcinoma with invasion limited to the submucosa. The purpose of this study was to clarify the clinicopathologic significance of DI in SSCCE. METHODS. Two hundred one surgically resected lesions from 140 patients with SSCCE were examined histopathologically. Clinicopathologic factors, such as macroscopic type, tumor location, maximum tumor size, depth of invasion, lymphatic and blood vessel permeation, lymph node metastasis, and prognosis, were examined to investigate the association between these factors and the presence of DI. RESULTS. Of the 201 SSCCE lesions, 43 (21.3%) had 152 DIs (1 32 DIs per lesion). The DI always remained in situ, and there were no tumors showing submucosal invasion through the DI. As for the relation between clinicopathologic factors and the presence of DI, multivariate analysis showed only maximum tumor size to correlate with the presence of DI (P 0.0001). There were no significant differences between DI negative and DI positive lesions in tumor location, macroscopic type, lymphatic and blood vessel permeation, lymph node metastasis, or prognosis. In 83 mucosal carcinomas, including in situ carcinomas or carcinomas that invaded no deeper than the lamina muscularis mucosa, no lymph node metastasis was found, and the 5-year survival rate was 100% (unaffected by the presence of DI). Among these 83 lesions, DI was found in 11 (13.8%), of which 6 (7.2%) had DI extending to the submucosal layer. CONCLUSIONS. These results indicate that DI as a pathway of tumor spread to the deeper layer is of little significance in squamous cell carcinoma of the esophagus, and that mucosal carcinomas with DI that extends to the submucosa should not be classified as submucosal carcinoma. Cancer 2000;89:248 54. 2000 American Cancer Society. Supported in part by a Grant-in-Aid for cancer research from the Ministry of Health and Welfare of Japan. Address for reprints: Tadakazu Shimoda, M.D., Clinical Laboratory Division, National Cancer Center Hospital, 1-1, Tsukiji 5 chome, Chuoku, Tokyo 104-0045, Japan. Received August 27, 1999; revisions received December 27, 1999 and March 15, 2000; accepted March 15, 2000. KEYWORDS: squamous cell carcinoma, superficial esophageal carcinoma, ductal involvement, glandular involvement, multivariate analysis. The esophageal submucosal glands are considered to be a continuation of the minor salivary glands of the mouth and pharynx and scattered throughout the entire esophagus. The glands are drained by ducts, lined by a single layer of cuboidal epithelium, which penetrate the muscularis mucosae and epithelium to open into the esophageal lumen. 1 Intraductal spread is one of the forms of tumor spread in squamous cell carcinoma of the esophagus (SCCE). 2,3 Nishimaki et al. and the Japanese Society for Esophageal Disease reported that intraductal spread was observed in 28 (22.2%) of 126 superficial squamous 2000 American Cancer Society

Ductal Involvement in Esophageal Carcinoma/Tajima et al. 249 cell carcinoma of the esophagus (SSCCE) defined as carcinoma with invasion limited to the submucosa. 2,4 Takubo et al. reported that of 175 SCCE tumors including 28 SSCCEs, 33 (19%) were positive for ductal involvement (DI), and that glandular involvement (GI) of tumor cells in the duct extending to the end portions of the esophageal glands was found in 12 (36%) tumors. 3 They speculated that a high incidence of DI in SCCE reflected its possible importance as a route to deep tissue in the early stage. In addition to the esophagus, DI (GI) also is frequently observed in squamous cell carcinomas of the oral floor 5 and the uterine cervix. 6,7 The importance of GI as a pathway of tumor spread to deeper layers has been recognized in melanoma of the palm and sole. 8 However, there have been no previous reports describing the clinocopathologic significance of DI and GI in esophageal carcinoma. Recently, the incidence rate of superficial esophageal carcinoma (SEC) defined as above has increased due to the greater frequency of endoscopic examinations by using the Lugol dye method. 9,10 The treatment of esophageal carcinoma is mainly surgical resection with systematic lymph node dissection. Endoscopic mucosal resection (EMR), a radical procedure that aims to eliminate mucosal carcinoma but is minimally invasive, has been performed successfully with excellent results. 9,11 According to the Japanese Guidelines for Clinical and Pathological Studies on Carcinoma of the Esophagus, tumors with DI that extends to the submucosa but does not definitely invade the submucosal stroma should not be classified as submucosal carcinoma. 4 Ductal involvement, which is present in mucosal carcinoma, may not be resected in EMR unless the submucosa beneath the tumor is sufficiently removed, leaving tumor tissue in residual parts of the submucosal layer. 2 Accordingly, we examined the clinicopathologic characteristics of DI in SSCCE by using 201 surgically resected lesions. MATERIALS AND METHODS Patients We examined 201 SSCCEs from 140 patients who had consecutively undergone esophagectomies at the National Cancer Center Hospital,Tokyo, Japan, between October 1987 and December 1996 (Table 1). The patients included 129 men and 11 women, aged 40 86 years (mean, 61.5 years). Thoraco-abdominal (two fields) or cervico-thoracic-abdominal (three-fields) lymphadenectomy was performed in all patients, and a mean of 55.8 26.6 lymph nodes were dissected from each patient. No preoperative radio- and/or chemotherapy was performed. Thirty-eight patients (27.1%) had multicentric lesions in the esophagus. TABLE 1 Patient Characteristics Characteristic No. of patients (%) Gender Male 129 Female 11 Mean age SD (yrs) 61.5 9.4 Depth of invasion Mucosal carcinoma 83 (41.3) Submucosal carcinoma 118 (58.7) Multicentric esophageal lesions No. of lesions 1 102 (72.9) 2 26 (18.6) 3 6 (4.3) 4 2 (1.4) 5 3 (2.1) 6 1 (0.7) Multiple primary tumors in other organs Pharynx 18 (36.0) Stomach 14 (28.0) Thyroid 4 (8.0) Larynx 3 (6.0) Colon 2 (4.0) Oral cavity 2 (4.0) Tongue 2 (4.0) Others 5 (10.0) Total 50 tumors in 43 patients Mean no. of dissected lymph nodes SD 55.8 26.6 SD: standard deviation. Forty-three patients (30.7%) had synchronous or metachronous multiple primary carcinomas in other organs. Pathologic Review Serial 5-mm-thick tissue sections of the entire lesion were cut from specimens fixed in 10% buffered formalin and embedded in paraffin and stained with hematoxylin and eosin. In this study, diagnostic criteria for DI were defined by the presence of ductal cancerization accompanied by nonneoplastic ductal epithelium lined by a single layer of cuboidal cells or mucinous cells in the center of and/or external to the cancerous nest. However, lesions without these histopathologic findings, but with desmoplastic reaction around the cancerous nest, were defined as tumor direct invasion into the stroma. The clinicopathologic review included tumor location, maximum tumor size, depth of invasion, lymphatic and blood vessel permeation, lymph node metastasis, and prognosis. The relations between the presence of DI and these clinicopathologic features were analyzed using univariate and multivariate analysis. In addition, histopathologic characteristics of DI positive mucosal carcinoma in which

250 CANCER July 15, 2000 / Volume 89 / Number 2 FIGURE 1. (a) Ductal involvement in superficial squamous cell carcinoma of the esophagus is shown. Cancer cells do not extend to the lamina muscularis mucosae. (b) Ductal involvement in superficial squamous cell carcinoma of the esophagus is shown. Cancer cells extend to the lamina muscularis mucosa but do not reach the end portions of the esophageal glands. (c) Ductal involvement in superficial squamous cell carcinoma of the esophagus is shown. Cancer cells reach the end portions of the esophageal glands, a portion of which are replaced by cancer cells. EMR may be indicated as a radical procedure also were examined. The histopathologic review was performed by three experienced pathologists (Y.T., Y.N., and T.S.) without knowledge of the patients clinical details. Results were compared, and discrepancies were resolved by reaching consensus after further histopathologic review. Statistical Analysis Logistic regression analysis was performed to assess the univariate and multivariate effects of various clinicopathologic factors. 12 The correlation between maximum tumor size and number of DI per tumor was assessed using the Spearman rank correlation coefficient test. The correlation between maximum tumor size and extent of DI was assessed using the Kruskal Wallis test. Five-year survival rates after esophagectomy were examined using the Kaplan Meier method. 13 The equality of the survival rates was assessed using the generalized Wilcoxon test. 14 Patients who died of operative complications or other diseases were excluded. The level of significance was set at P values less than 0.05. RESULTS Number of DIs per Lesion and Extent of DI in 201 Superficial Squamous Cell Carcinomas of the Esophagus Of 201 SSCCE lesions, 43 (21.3%) had 152 DIs. The number of DIs per lesion was 1 in 24 (55.8%) lesions, 2 in 5 (11.6%), 3 in 4 (9.3%), 4 in 2 (4.7%), 5 in 2 (4.7%),

Ductal Involvement in Esophageal Carcinoma/Tajima et al. 251 FIGURE 1. (continued) 8 in 1 (2.3%), 10 in 1 (2.3%), 11 in 2 (4.7%), 16 in 1 (2.3%), and 32 in 1 (2.3%). Of 152 DIs, 68 (44.7%) were limited to the mucosal layer including the lamina muscularis mucosa (intramucosal) (Fig. 1a), and 60 (39.5%) extended to the submucosal layer through the lamina muscularis mucosae but did not reach the end portions of the esophageal glands (submucosal) (Fig. 1b), whereas 24 (15.8%) did reach the end portions of the esophageal glands (glandular involvement; GI) (Fig. 1c). There was no lesion in which carcinomatous epithelium in the duct distinctly invaded the submucosal stroma with destruction of the basement membrane and desmoplastic stromal reaction around the tumor. Clinicopathologic Characteristics of DI Positive Lesions Results of the correlation between histopathologic findings and the presence of DI by using univariate and multivariate analysis of 201 SSCCE lesions are shown in Table 2. Univariate analysis showed maximum tumor size and depth of invasion to have a significant influence on DI (P 0.0001and P 0.02, respectively). Multivariate analysis showed that only maximum tumor size correlated with the presence of DI (P 0.0001). The prevalences of DI positive lesions with a maximum tumor size less than 2 cm, greater than 2 cm but less than 4 cm, greater than 4 cm but less than 6 cm, and greater than 6 cm were 7.0% (5 of 72 lesions), 15.7% (11 of 70), 38.2% (23 of 34), and 56.0% (14 of 25), respectively. In 43 DI positive lesions, the numbers of DIs per lesion with a maximum tumor size less than 2 cm, greater than 2 cm but less than 4 cm, greater than 4 cm but less than 6 cm, and greater than 6 cm were 1.40 0.89, 2.18 2.68, 2.08 1.32, and 6.71 8.78 (mean standard deviation), respectively. A larger maximum tumor size correlated significantly with a higher prevalence of DI and a higher number of DIs per lesion (P 0.0001 and P 0.03; Spearman rank correlation, 0.40, respectively). The deepest extensions of DI with tumor size less than 2 cm, greater than 2 cm but less than 4 cm, greater than 4 cm but less than 6 cm, and greater than 6 cm were intramucosal in 1 (20.0%), 5 (45.5%), 5 (38.5%), and 6 (42.9%) lesions, submucosal in 2 (40.0%), 2 (18.2%), 7 (53.8%), and 4 (28.6%) lesions, and into the glandular acini in 2 (40.0%), 4 (36.4%), 1 (7.7%), and 4 (28.6%) lesions, respectively. There was no significant correlation between maximum tumor size and extent of DI (Table 3). Regarding the correlations of the presence of DI with lymph node metastasis and prognosis in 140 SSCCE patients, the tumor showing the deepest invasion was evaluated in multicentric tumors. Of 105 patients with DI positive lesions, lymph node metastasis was found in 37 (35.2%) (mucosal carcinoma, 0 of 27 patients [0%]; submucosal carcinoma, 37 of 78 patients [47.4%]). Conversely, of 35 patients with DI negative lesions, lymph node metastasis was found in 14 (40.0%) (mucosal carcinoma, 0 of 7 patients [0%]; submucosal carcinoma, 14 of 28 patients [50.0%]). These differences were not significant. When the data for the same patients were analyzed using the presence of GI instead of DI, similar results were obtained (data not shown). The 5-year survival rates of 105 patients with DI negative lesions and 35 patients with DI positive lesions were 82.1% (mucosal carcinoma, 100%; sub-

252 CANCER July 15, 2000 / Volume 89 / Number 2 TABLE 2 Variables Associated with the Presence of Ductal Involvement in Univariate and Multivariate Analysis of 201 Lesions with Superficial Squamous Cell Carcinoma of the Esophagus Variable DI negative (n 158) DI positive (n 43) Univariate analysis (P value) Multivariate analysis OR 95% CI P value Location Cervical 5 1 Thoracic upper 28 8 Thoracic middle 73 19 NS NS Thoracic lower 52 15 Macroscopic type I 21 4 IIa 9 1 IIc 89 20 NS NS III 3 0 Combined 36 18 Mean of maximum tumor size (cm) 2.6 1.9 5.4 4.1 0.0001 1.536 1.268 1.864 0.0001 Depth of invasion (cm) m 72 11 0.02 NS sm 86 32 Lymphatic permeation Absent 104 26 NS NS Present 54 17 Blood vessel permeation Absent 135 36 NS NS Present 23 7 DI: ductal involvement; OR: odds ratio; CI: confidence interval; NS: not significant; I: elevated type; IIa: slightly elevated type; IIc: slightly depressed type; III: depressed type; combined: combined type. mucosal carcinoma, 75%) and 85.0% (mucosal carcinoma, 100%; submucosal carcinoma, 80.8%), respectively. These differences were not significant. When the data for the same patients were analyzed using the presence of GI instead of DI, similar results were obtained (data not shown). Histopathologic Characteristics of 11 DI Positive Lesions with Mucosal Superficial Squamous Cell Carcinomas of the Esophagus Of 83 mucosal squamous cell carcinomas including 38 in situ carcinomas and 45 invasive carcinomas, 11 (13.8%) including 3 in situ carcinomas had 29 DIs. Maximum tumor size was 2.83 1.41 cm (mean standard deviation; range, 1.0 5.5 cm) in 11 DI positive lesions, compared with 1.67 1.35 cm (0.2 6.5 cm) in 72 DI negative lesions. The difference was statistically significant (P 0.01). In 11 DI positive lesions, the mean number of DI per tumor was 2.64 2.77 (mean standard deviation; range, 1 10), and the deepest extension of DI was intramucosal in 5 tumors, submucosal in 3 tumors, and into the glandular acini in 3 tumors (Table 4). DISCUSSION The respective prevalences of DI in mucosal and submucosal SSCCE were reported to be 11.9% and 33.9% by Nishimaki et al. 2 and 23.5% and 9.0% by Takubo et al. 3 In this study, the prevalences were 13.3% and 27.1%, respectively, which is similar to the results of previous studies. Our results showed that maximum tumor size and depth of invasion had a significant influence on DI by univariate analysis, but that only maximum tumor size correlated with the presence of DI by multivariate analysis. In 43 DI positive lesions, larger maximum tumor size was significantly correlated with a higher number of DI per lesion. These results indicate that DI develops in association with horizontal tumor growth. The larger tumor size might be accompanied by a higher number of esophageal glands and ducts beneath the tumor and accordingly by a higher incidence of DI. Takubo et al. showed by electron microscope examination that the cancer cells of intraepithelial spread entered the duct by raising the normal ductal epithelium in the ductal cavity despite the basal lamina being retained, and normal gland duct cells and cancer cells either were attached with desmosomes directly or were separated by degenerated epithelial cells. These findings were similar to those observed for mucosal intraepithelial spread of carcinomas. 3,15 Thus, higher prevalence of DI might correlate with a tendency for intraepithelial horizontal spread of the tumor. Conversely, there were no signif-

Ductal Involvement in Esophageal Carcinoma/Tajima et al. 253 TABLE 3 Relation between Maximum Tumor Size and Ductal Involvement in Superficial Squamous Cell Carcinoma of the Esophagus Deepest extension of DI Maximum tumor size (cm) (n) No. of DI positive lesions (%) Mean no. of DI per lesion Intramucosal (%) Submucosal (%) Into the glandular acini (%) 2 (72) 5 (7.0) 1.40 0.89 1 (20.0) 2 (40.0) 2 (40.0) 4 (70) 11 (15.7) 2.18 2.68 5 (45.5) 2 (18.2) 4 (36.4) 6 (34) 13 (38.2) 2.01 1.32 5 (38.5) 7 (53.8) 1 (7.7) 6 (25) 14 (56.0) 6.71 8.78 6 (42.9) 4 (28.6) 4 (28.6) DI: ductal involvement. TABLE 4 Histopathologic Characteristics of 11 Ductal Involvement Positive Lesions with Mucosal Squamous Cell Carcinoma of the Esophagus Lesion no. Tumor size (cm) No. of DI per lesion Deepest extension of DI 1 a 1.0 1 Intramucosal 2 1.5 1 Submucosal 3 1.9 3 Into the glandular acini 4 2.0 2 Into the glandular acini 5 a 2.3 1 Intramucosal 6 2.3 1 Intramucosal 7 a 2.8 10 Intramucosal 8 2.8 1 Into the glandular acini 9 4.5 5 Submucosal 10 4.5 3 Submucosal 11 5.5 1 Intramucosal DI: ductal involvement. a In situ carcinoma. icant differences between DI negative and DI positive lesions in tumor location, macroscopic type, lymphatic and blood vessel permeation, lymph node metastasis, or prognosis. Therefore, the presence of DI in surgically resected specimens is of no clinical significance. Furthermore, in 83 mucosal carcinomas, no lymph node metastasis was found, and the 5-year survival rate was 100%, i.e., unaffected by the presence of DI. Our results indicate that mucosal carcinomas with DI that extends to the submucosa should not be classified as submucosal carcinoma. Studies of patients with SSCCE revealed that 0 8.7% of mucosal carcinoma had lymph node metastasis, whereas the rate of lymph node metastasis in those with submucosal carcinomas ranged from 43.6 to 53.3%. 16,17 It has been proposed that EMR be indicated as a radical resection procedure for mucosal carcinomas. 9,11 Our results showed that no lymph node metastasis was found in mucosal carcinoma irrespective of the presence of DI. Therefore, additional surgical resection after EMR is not needed in mucosal carcinoma. However, even in mucosal carcinoma, there exists a possibility of incomplete clearance of the tumor tissue due to the presence of DI extending to the submucosal layer or reaching the end portions of esophageal glands. Daley et al. reported the problem of involvement of ducts of minor salivary glands extending to the deep margin of resection in superficial specimens from the mouth. They found a higher recurrence rate in the mouth in cases with ductal cancerization. 5 The authors think that radiotherapy is one of therapeutic approaches to such cases of incomplete clearance of the tumor tissue. In this study, among 83 mucosal carcinomas, DI with extension to the submucosal layer or reaching the end portions of esophageal glands was not common, but each was present in 3 tumors (3.6%). Therefore, careful observation is needed to detect the extension of DI to the submucosal layer in EMR specimens, especially in large mucosal carcinomas. In this study, diagnostic criteria for DI were defined by the presence of ductal cancerization accompanied by nonneoplastic ductal epithelium lined by a single layer of cuboidal cells or mucinous cells in the center of and/or external to the cancerous nest. Lesions without these histopathologic findings, but with desmoplastic reaction around the cancerous nest, were regarded as direct tumor invasion into the stroma. Endoscopic mucosal resection has been indicated for mucosal carcinoma as a radical procedure. 9 11 However, additional surgical resection with systematic lymph node dissection after EMR is needed in submucosal carcinoma due to the high incidence of lymph node metastasis. 16 18 Therefore, it is important to judge accurately whether a small cancerous nest in the submucosal layer is DI or direct tumor invasion in deciding the necessity for additional surgical resection based on the histopathologic findings in EMR specimens. Iwase reported that in 28 of 33 (84.8%) microinvasive carcinomas of the cervix uteri, inceptive foci of

254 CANCER July 15, 2000 / Volume 89 / Number 2 invasion from GI were detected histologically. 7 In this study, there were no cases in which the tumor showed submucosal invasion through DI. It seems likely that the difference in the prevalence of tumor invasion into the stroma through DI between squamous cell carcinoma of the cervix uteri and that of the esophagus may be related to the difference in the number of glands beneath the squamous epithelium. Our results indicate that DI as a pathway of tumor spread to the deeper layer is of little significance in squamous cell carcinoma of the esophagus. REFERENCES 1. Enterline H, Thompson J. Pathology of the esophagus. New York: Springer-Verlag, 1984:14. 2. Nishimaki T, Tanaka O, Suzuki T, Aizawa K, Watanabe H, Muto T. Tumor spread in superficial esophageal cancer: histopathologic basis for rational surgical treatment. World J Surg 1993;17:766 72. 3. Takubo K, Takai A, Takayama S, Sasajima K, Yamashita K, Fujita K. Intraductal spread of esophageal squamous cell carcinoma. Cancer 1987;59:1751 7. 4. Japanese Society for Esophageal Disease. Guidelines for clinical and pathologic studies on carcinoma of the esophagus [in Japanese]. 9th ed. Tokyo: Kanehara Shuppan, 1999. 5. Daley TD, Lovas JG, Peters E, Wysocki GP, McGaw TW. Salivary gland duct involvement in oral epithelial dysplasia and squamous cell carcinoma. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1996;81:186 92. 6. Gompel C, Silverberg SG. The cervix. In: Gompel C, Silverberg SG, editors. Pathology in gynecology and obstetrics. 3rd ed. Philadelphia: J.B. Lippincott Company, 1985:66 148. 7. Iwase H. Histologic analysis of microinvasive carcinoma of the cervix uteri with graphic reconstruction [in Japanese with English abstract]. Acta Obstet Gynaecol Jpn 1986;38: 393 401. 8. Seiji M, Takahashi M. Acral melanoma in Japan. Hum Pathol 1982;13:607 9. 9. Kodama M, Kakegawa T. Treatment of superficial cancer of the esophagus: a summary of responses to a questionnaire on superficial cancer of the esophagus in Japan. Surgery 1998;123:432 9. 10. Endo M, Kawano T. Detection and classification of early squamous cell esophageal cancer. Dis Esophagus 1997;10:155 8. 11. Makuuchi H, Machimura T, Mizutani K. Endoscopic mucosal resection for early carcinoma of esophagus. In: Takanishi T, editor. Recent advances in management of digestive cancer. Tokyo: Springer-Verlag, 1993:94 7. 12. Lemeshow S, Hosmer D. Estimating odds ratio with categorically scaled covariates in multiple logistic regression analysis. Am J Epidemiol 1988;119:147 51. 13. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457 81. 14. Gehan EA. A generalized Wilcoxon test for comparing arbitrarily single-censored samples. Biometrika 1965;52:203 24. 15. Takubo K, Nishimura H, Taniguchi Y, Sasajima K, Nakagawa H, Miyamoto H, et al. Junctions between intraepithelial carcinoma and non-neoplastic tissue of the esophagus: light and electron microscopic studies. Acta Pathol Jpn 1984;34:785 96. 16. Tachibana M, Yoshimura H, Kinugasa S, Hashimoto N, Dhar D, Abe S, et al. Clinicopathological features of superficial squamous cell carcinoma of the esophagus. Am J Surg 1997; 174:49 53. 17. Nabeya K, Nakata Y. Extent of resection and lymphadenectomy in early squamous cell esophageal cancer. Dis Esophagus 1997;10:159 61. 18. Kato H, Tachimori Y, Watanabe H, Yamaguchi H, Ishikawa T, Itabashi M. Superficial esophageal carcinoma: surgical treatment and the results. Cancer 1990;66:2319 23.