Annals of Surgical Oncology, 6(7):664 – 670 Published by Lippincott Williams & Wilkins © 1999 The Society of Surgical Oncology, Inc. Is Lymphadenectomy Necessary for Early Gastric Cancer? Steven N. Hochwald, MD, Murray F. Brennan, MD, David S. Klimstra, MD, Steven Kim, MD, and Martin S. Karpeh, MD Background: Although early gastric cancer (T1, NX) in Japan has been reported to have an excellent prognosis, the experience with this cancer in the United States is limited. The treatment of these tumors in Japan is becoming less aggressive as “good prognostic factors” are increasingly recognized. Our objective was to identify predictors of nodal disease and survival in a large cohort of Western patients with T1 tumors. Methods: A retrospective review of our prospective data base from July 1985 to March 1998 revealed 165 patients undergoing surgical resection for T1 gastric tumors. Clinicopathological factors analyzed and compared included presence of positive nodes, tumor size ($4.5 vs. ,4.5 cm), depth (mucosal vs. submucosal), grade (poor vs. moderate and well), and tumor site (proximal vs. distal), presence of venous or perineural invasion, and Lauren’s classification. Factors predicting lymph node involvement and disease-specific survival were evaluated by univariate and multivariate analysis. Results: Median follow-up time was 36 months. The actuarial 5-year survival was 88%. Thirteen patients (8%) died of disease. Lymph node involvement was present in 31 tumors (19%), with a 5-year survival of 91% with negative nodes vs. 78% with positive nodes. On univariate and multivariate analysis, the presence of tumor submucosal invasion (P , .05), venous invasion (P 5 .02), and size of 4.5 cm and larger (P 5 .02) was significantly associated with an increased risk for nodal positivity. On univariate analysis of survival, node-positive tumors (P 5 .02) and tumors 4.5 cm and larger (P 5 .008) were significantly associated with decreased survival. On multivariate analysis, only node-positive tumors were significantly (P 5 .01) associated with decreased survival. Those tumors that were limited to the mucosa and less than 4.5 cm in size (n 5 47) had a 4% rate of positive nodes. In contrast, those tumors that were 4.5 cm and larger and had penetrated into the submucosa (n 5 16) had a 56% chance of positive nodes. Conclusions: Early gastric carcinoma in North America has an excellent prognosis, similar to that in Japan. Tumors that are limited to the mucosa and smaller than 4.5 cm could be considered for limited resection without lymphadenectomy. Key Words: Early gastric cancer—Lymphadenectomy. Early gastric cancer was first defined in 1962 by the Japanese Society of Gastroenterological Endoscopy as adenocarcinoma confined to the mucosa or submucosa irrespective of lymph node involvement (T1NX).1 The need for such a definition was based on the observation that gastric cancer of this type had a favorable prognosis. Survival rates of more than 90% have been shown in several series of patients with early gastric cancer from Japan.2,3 However, because of differences in the prevalence of gastric cancer in certain regions of the world, there is limited experience with early gastric cancer in North America, and varied treatment outcome has been reported.4 –7 The most widely used treatment for early gastric cancer is gastrectomy with radical D2 lymphadenectomy. Although initially proposed in the United States, this was universally accepted as the treatment of choice in Japan for many years but has only recently regained acceptance as a satisfactory treatment in some centers in Europe and Received March 23, 1999; accepted July 20, 1999. From the Departments of Surgery (SNH, MFB, MSK) and Pathology (DSK), Memorial Sloan-Kettering Cancer Center, New York, New York; and the Department of Surgery (SK), University of Medicine and Dentistry, Newark, New Jersey. Presented at the 52nd Annual Meeting of the Society of Surgical Oncology, Orlando, FL, March 4 –7, 1999. Address correspondence to: Martin S. Karpeh, MD, Department of Surgery, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 10021; Fax: 212-794-5847. 664 LYMPHADENECTOMY FOR EARLY GASTRIC CANCER North America.8 Whereas operative mortality is approximately 1% in Japan, a range in mortality from 1% to 10% has been published in studies from Europe.8 –11 As experience with early gastric cancer has increased, despite their previous aggressive approach, Japanese investigators are now advocating more conservative treatment including endoscopic mucosal resection and limited gastric resection without lymphadenectomy, in selected patients with early gastric cancer.12 Whether this is suitable for patients in the West is unknown. It continues to be debated whether early gastric cancer behaves differently in Japan than in the West. Our objective was to evaluate, in a large series of patients treated in a single institution in North America, factors that predicted lymph node involvement and decreased survival in early gastric cancer. We analyzed our data to identify which patients had the lowest rate of lymph node metastases and potentially could have benefited from tumor resection without lymphadenectomy. PATIENTS AND METHODS Between July 31, 1985, and March 31, 1998, 1558 patients with primary gastric carcinoma were admitted to the surgical service at Memorial Sloan-Kettering Cancer Center and entered onto a prospective data base. Of these patients, 1509 underwent surgical exploration and 49 did not. One hundred sixty-five (11%) of those patients undergoing surgery had pT1 tumors at the time of pathological analysis, and they form the basis of this study. All patients underwent a complete resection of the tumor. Careful analysis of clinicopathological variables was undertaken. Clinical and patient factors analyzed were age, sex, and ethnic distribution, tumor site, type of resection, type of lymphadenectomy, and symptoms at the time of presentation. Standard processing of pathological material was performed throughout the study period and consisted of hematoxylin and eosin staining. Pathological variables analyzed included presence or absence of nodal disease, depth of tumor penetration into the gastric wall (mucosa or submucosa), tumor differentiation (well, moderate, or poor), presence of venous or neural invasion, Lauren’s histological type (intestinal, diffuse, or mixed), and tumor size. Univariate analysis of factors potentially associated with having metastatic lymph nodes was performed by using x2 analysis. Kaplan-Meier analysis was performed, with a log-rank test to compare survival between factors. Multivariate analysis of factors influencing nodal involvement and survival was performed by using the Cox proportional hazards model. Significance was defined as P , .05. 665 RESULTS Demographics The median age of 165 patients with T1 gastric carcinoma, undergoing surgical resection, was 67 years (range, 29 – 88 years). The male-to-female ratio was 1.5:1. The ethnic distribution and presenting symptoms are shown in Table 1. Most patients were white (n 5 129; 78%), with Asian patients (n 5 20; 12%) being the second largest group. Only 5 patients (3%) were completely asymptomatic at presentation. The most frequent complaints were epigastric pain (n 5 53; 32%) and dyspepsia (n 5 27; 16%). The percentage of cases undergoing surgical exploration for gastric carcinoma for Tis or T1 lesions was 10% or less of cases in the 1980s, but during the last several years it has increased to approximately 20% (Fig. 1). Clinical Factors Tumors were classified as being in five locations, based on preoperative evaluation and surgical findings (Table 2). Fifty-four patients (33%) had tumors located either in the gastroesophageal junction or in the upper one-third of the stomach and were defined as proximal tumors. Most tumors (n 5 107; 65%) were located in the body/middle one-third or the antrum/pyloric region of the stomach and were defined as distal tumors. Five patients had tumors that diffusely involved the whole stomach. The type of resection reflects the tumor location (Table 2). Forty-nine patients (30%) had either an esophago/ proximal gastrectomy or a proximal subtotal gastrectomy. However, most patients had a distal subtotal gastrectomy (n 5 89; 54%). Twenty-two patients underwent a total gastrectomy, and five patients underwent a wedge resection of the stomach. Most patients underwent a D2 lymphadenectomy (n 5 96; 58%). A D1 lymphadenectomy was performed in 51 (31%) and a D3 lymphadenectomy in 9 (6%) patients. TABLE 1. Demographics of 165 patients with T1 gastric carcinoma Ethnicity and symptoms Ethnic distribution Black Hispanic Asian White Symptoms Epigastric pain Dyspepsia Gastrointestinal bleed Nausea/vomiting Dysphagia n (%) 5 (3) 11 (7) 20 (12) 129 (78) 53 (32) 27 (16) 18 (11) 13 (8) 11 (7) Ann Surg Oncol, Vol. 6, No. 7, 1999 666 S. N. HOCHWALD ET AL. and no tumor was identified in the subsequent surgical resection specimen (Table 3). Tumors were categorized as poorly (n 5 58% and 36%), moderately (n 5 52% and 32%), or well (n 5 46% and 28%) differentiated. In 9 cases, tumor differentiation was not available (Table 3). Venous invasion was indicated in 12% (n 5 20) of 143 cases for which these were available. Perineural invasion was present in only 4% (n 5 7) of 126 cases for which these data were available. Most tumors were classified as being of Lauren’s intestinal (n 5 106% and 64%) histological type, with mixed (n 5 9% and 6%) and diffuse (n 5 39% and 24%) types being relatively less frequent (Table 3). FIG. 1. Percentage of explorations for gastric carcinoma performed between 1985 and 1997 that were for carcinoma in situ (Tis) or T1 tumors. The denominator includes all patients explored for gastric carcinoma during this time period. No lymphadenectomy was performed in 5 patients, and in 4 patients the type of lymphadenectomy could not be determined. The median number of lymph nodes retrieved in the surgical specimen was 20. Thirty-one (19%) of 165 patients had positive nodes. The median number of positive nodes was two (range, 1–20). Pathological Factors The mean tumor size for all patients was 2.63 6 1.95 cm (range, 0.2–9.0 cm). The mean tumor size was significantly larger for the first 50 patients (3.1 cm), treated from 1985 to 1990, than for the next 108 patients (2.4 cm), treated from 1991 to 1998 (P 5 .04). Most patients had tumor penetration into the submucosa (n 5 99; 60%); 62 (38%) patients had penetration confined to the mucosa. In 4 patients, the extent of tumor penetration could not be determined because either the original pathology report did not indicate it and slides were not available for review, or the tumor had been removed in fragments at the time of endoscopic biopsy Tumor site Gastroesophageal junction Proximal/upper one-third Body/middle one-third Antrum/pyloric Diffuse/whole Type of resection Esophago/proximal gastrectomy Proximal subtotal gastrectomy Distal subtotal gastrectomy Total gastrectomy Wedge/sleeve resection Ann Surg Oncol, Vol. 6, No. 7, 1999 Survival Analysis With a median follow-up time of 36 months for the entire group (n 5 165), actuarial disease-specific survival was 88% at 5 years (Fig. 2). For those patients undergoing resection between July 1, 1985, and DecemTABLE 3. Pathological factors Factor TABLE 2. Tumor site and type of resection Site and resection Predicting Lymph Node Metastases Factors associated with having metastatic lymph nodes were determined by x2 analysis. On univariate analysis, perineural invasion (yes vs. no; P 5 .33), tumor site (proximal vs. distal; P 5 .13), Lauren’s histological type (intestinal vs. mixed and diffuse; P 5 .08), and tumor differentiation (well and moderate vs. poor; P 5 .06) had no significant impact on nodal positivity. However, tumor size (,4.5 cm, n 5 131 vs. $4.5 cm, n 5 27; P 5 .006), presence of venous invasion (no, n 5 123 vs. yes, n 5 20; P 5 .001), and depth of tumor penetration (mucosa, n 5 62 vs. submucosa, n 5 99; P 5 .001) were significantly associated with having positive nodes (Table 4). On multivariate analysis, tumors that were 4.5 cm or larger or had venous invasion or submucosa penetration were independently significantly associated with nodal positivity (Table 5). n (%) 29 (18) 25 (15) 45 (27) 62 (38) 5 (3) 29 (18) 20 (12) 89 (54) 22 (13) 5 (3) Depth of tumor penetration Mucosa Submucosa Unknown Tumor differentiation Well Moderate Poor Unknown Lauren’s histological type Intestinal Mixed Diffuse Unknown n (%) 62 (38) 99 (60) 4 (2) 46 (28) 52 (32) 56 (36) 9 (5) 106 (64) 9 (5) 39 (24) 11 (7) LYMPHADENECTOMY FOR EARLY GASTRIC CANCER 667 TABLE 4. Univariate analysis of factors predicting positive nodes Factor n Perineural invasion Tumor site (proximal vs. distal) Lauren’s histological type (intestinal vs. mixed and diffuse) Differentiation (well and moderate vs. poor) Size ,4.5 cm $4.5 cm Venous invasion No Yes Depth Mucosa Submucosa P .33 .13 .08 .06 .006 131 27 .001 123 20 .001 62 99 ber 1, 1993 (n 5 91), the median follow-up time was 58 months and the 5-year survival was 87%. Factors associated with survival were determined by Kaplan-Meier and log-rank analysis (Table 6). On univariate analysis, there was no significant impact of perineural invasion (yes vs. no; P 5 .56), Lauren’s histological type (intestinal vs. mixed and diffuse; P 5 .68), venous invasion (yes vs. no; P 5 .91), depth of tumor penetration (mucosa vs. submucosa; P 5 .26), tumor site (proximal vs. distal; P 5 .06), and tumor differentiation (well and moderate vs. poor; P 5 .09) on disease-specific survival. On univariate analysis, only nodal status and tumor size significantly impacted on survival. Node-negative tumors (n 5 134; 5-year survival, 91%) had a significantly better survival compared with node-positive tumors (n 5 31; 5-year survival, 78%; P 5 .02) (Fig. 3). In addition, tumors 4.5 cm or larger (n 5 27; 5-year survival, 75%) had a significantly decreased survival compared with tumors smaller than 4.5 cm (n 5 131; 5-year survival, 90%; P 5 .008) (Fig. 4). On multivariate analysis, only nodal status significantly impacted independently on survival, with node-positive patients faring worse (P 5 .02; TABLE 5. Multivariate analysis of factors predicting positive nodes Factor P Relative risk CI (95%) Perineural invasion Tumor site Lauren’s histological type (intestinal vs. others) Tumor differentiation (poor vs. others) Depth (submucosa vs. mucosa) Venous invasion Size ($4.5 vs. ,4.5 cm) .19 .58 .37 2.8 1.4 2.0 2.0–3.5 0.8–2.1 1.2–2.8 .78 .05 .02 .02 1.2 4.8 3.4 3.3 0.5–1.9 4.0–5.5 2.9–4.0 2.8–3.8 CI, confidence interval. FIG. 2. Disease-specific survival of 165 patients with early gastric carcinoma undergoing treatment. Actuarial 5-year survival is 88%. Median follow-up is 36 months. Table 7). The relative risk of death was increased by 4.6 in the presence of lymph node metastases. The prevalence of important pathological factors in node-positive and node-negative tumors was determined. It was noted that in those tumors that were node positive, 87% had submucosal penetration (n 5 27) compared with 55% (n 5 71) in those that were node negative. In those tumors that were node positive, 53% were poorly differentiated (n 5 16) compared with 33% (n 5 41) in those that were node negative. In node-positive tumors, TABLE 6. Univariate analysis of factors associated with survival Factor Perineural invasion No Yes Lauren’s histological type Intestinal Mixed or diffuse Venous invasion No Yes Depth Mucosa Submucosa Tumor site Proximal Distal Tumor differentiation Well or moderate Poor Nodal status Negative Positive Size ,4.5 cm .4.5 cm n 5-y survival (%) 119 7 94 100 106 48 87 91 123 20 93 92 62 99 94 83 53 107 83 92 98 58 85 92 134 31 91 78 131 27 90 75 P .56 .68 .91 .26 .06 .09 .02 .008 Ann Surg Oncol, Vol. 6, No. 7, 1999 668 S. N. HOCHWALD ET AL. TABLE 7. Multivariate analysis of factors associated with survival Factor P Relative risk CI (95%) Perineural invasion Lauren’s histotype (intestinal vs. others) Depth (submucosa vs. mucosa) Size ($4.5 vs. ,4.5 cm) Tumor site Venous invasion Differentiation (poor vs. others) Nodal status .86 .43 1.0 1.1 0.9–1.1 1.0–1.2 .57 .99 .28 .10 .16 0.9 1.0 1.2 1.2 0.31 0.6–1.1 0.8–1.3 1.0–1.5 1.1–1.4 0.0–1.1 .02 4.6 4.0–5.3 CI, confidence interval. FIG. 3. Disease-specific survival of patients with node-negative (dashed line; n 5 134) and node-positive tumors (solid line; n 5 31). Actuarial 5-year survival was 91% for those node negative and 78% for those node positive (P 5 .02). 44% (n 5 11) had evidence of venous invasion compared with 8% (n 5 9) of node-negative tumors. Finally, in node-positive tumors, 35% (n 5 11) were 4.5 cm or larger compared with 13% (n 5 16) in node-negative tumors (Fig. 5). Pathological factors that would be available on a preoperative basis that could be used to determine appropriate surgical therapy were analyzed. Those tumors that were (1) limited to the mucosa and (2) smaller than 4.5 cm (n 5 47) had a 4% rate of positive nodes. Those tumors that were (1) limited to the mucosa, (2) smaller FIG. 4. Disease-specific survival of patients with tumors less than 4.5 cm in size (dashed line; n 5 131) and tumors 4.5 cm or larger (solid line; n 5 27) (P 5 .008). Ann Surg Oncol, Vol. 6, No. 7, 1999 than 4.5 cm, and (3) well or moderately differentiated (n 5 32) had a 0% rate of positive nodes. In contrast, those tumors that had submucosal penetration or were 4.5 cm or larger (n 5 109) had a 27% rate of positive nodes. Those tumors that both penetrated into the submucosa and were 4.5 cm or larger (n 5 16) had the greatest rate (56%) of positive nodes. DISCUSSION Our 5-year survival for the treatment of patients with early gastric cancer in a large series from North America is similar to that reported by the Japanese. The Japanese have demonstrated high relative survival at 10 and 15 years, comparable with the rest of their population.13 The estimated 10-year disease-specific survival is 77% in this series, which is consistent with other recent Western reports that indicate survival between 77% and 85%.14 –16 Although different survival rates should be compared cautiously, because they are dependent on the length of follow-up, whether relative survival or deaths from gastric carcinoma alone are considered and the percentage FIG. 5. Percent of important pathological factors in node-positive and node-negative tumors. LYMPHADENECTOMY FOR EARLY GASTRIC CANCER of tumors that have submucosa vs. mucosa involvement, it seems that early gastric cancer carries a marginally poorer prognosis in the Western population. We found that early gastric cancer comprised 11% of all cases of gastric carcinoma undergoing surgical exploration at our institution from 1985 to 1998. Tumor size at the time of presentation was significantly larger in the first 50 patients evaluated in the 1980s and decreased to an average of 2.4 cm for the next 108 patients. In addition, the percentage of cases of gastric carcinoma that were early gastric cancer may have been increasing recently (Fig. 1). Indeed, more than 20% of cases in 2 of the last 4 years were T1. The recurrence rate of early gastric carcinoma in Western series has been reported to be 10% to 20% of resected cancers. When all cases including palliative surgery and patients who are not surgical candidates are included, the recurrence rate decreases to approximately 10%.8 It is clear that early gastric cancer is detected more frequently in Japan than in the West. The Japanese indicate that the percentage of early cancers among resected cases was 40% in 1985 with a gradual increase in detection rates throughout the country over the preceding 20 years.17 The discrepancy between Japan and the West appears to be related to the mass screening of the asymptomatic population, which has been done in Japan for almost 40 years. Similar to our findings, it appears that the rate of early gastric cancer is increasing in Europe.18 This may reflect advances in healthcare awareness in the United States and Europe as well as the earlier use of endoscopy than in the past. With increased prevalence of early gastric lesions in the West, it becomes even more important that alternatives to radical gastrectomy be explored. Recent prospective, randomized trials in Europe, comparing D1 and D2 lymphadenectomy for gastric carcinoma, indicate that mortality for D1 lymphadenectomy is 4% to 6% and for D2 lymphadenectomy is up to 13%.10,11 At the same time, the Japanese have published data indicating the efficacy of tumor resection without lymphadenectomy and/or gastrectomy.19 Tumor resection can then be accomplished through the use of laparoscopic gastrectomy or endoscopic mucosal resection. If more patients with early gastric cancer who do not need a lymphadenectomy can be identified, then a full laparotomy may be avoided. The problem with a limited surgical procedure in patients with early gastric cancer is identifying lesions that do not have associated positive nodes. There are limited data concerning this issue in the Western literature because of the infrequent nature of this disease. On multivariate analysis, we found that tumors 4.5 cm or larger, those with venous invasion, and those with sub- 669 mucosal penetration were independently associated with nodal positivity. In addition, nodal positivity was significantly associated with decreased survival. However, those tumors that were limited to the mucosa and smaller than 4.5 cm (n 5 47) had a 4% recurrence rate of positive nodes. Those limited to the mucosa, smaller than 4.5 cm, and well or moderately differentiated (n 5 32) had a 0% recurrence rate of positive nodes. Monoclonal antibody staining of pathology specimens in gastric cancer may increase the rate of positive nodes; but, to date, there are few studies addressing the clinical relevance of a histologically negative but immunohistochemically positive lymph node. Because patients with T1 gastric cancers who have standard hematoxylin and eosin-positive lymph nodes have a good prognosis, it will be very difficult to determine the benefit of immunohistochemistry in T1 tumors that are negative on hematoxylin and eosin staining. However, if this becomes important, perhaps new criteria for defining high- and low-risk primary cancers will need to be established. The crucial factor in the decision for or against limited surgery is the preoperative differentiation between carcinomas of mucosal and those of submucosal extent. In Japan, this is done solely on the basis of the macroscopic appearance during endoscopy. Protruded-type carcinomas (type I and type IIA) with a diameter of less than 25 mm and excavated types (type IIC) with a diameter of less than 20 mm are almost always free of lymph node metastases.19 Another variable that is used in Japan to select patients for endoscopic mucosal resection is histological differentiation. It has been observed that undifferentiated carcinomas often show diffuse, ill-defined superficial growth and are not candidates for endoscopic mucosal resection. In addition, the existence of intramural tumor ulcer or ulcer scar suggests breakdown of the muscularis mucosa, suggesting that invasion into the submucosa had already occurred.20 In the West, experience with the macroscopic classification advocated by the Japanese, for selecting patients with early gastric cancer suitable for limited surgery, is limited. It requires an experienced endoscopist with significant experience with gastric cancer, and findings are subjective. A more objective decision, using endoluminal ultrasonography, may be preferable. It has been shown in multiple studies that the accuracy of endosonography for T stage is approximately 85%.21 One disadvantage of endosonography is that it is operator dependent and often not available, requiring highly trained examiners. Clearly, not all early gastric carcinomas behave in the same manner. There is a subset of tumors that do not require a gastrectomy or lymphadenectomy for effective Ann Surg Oncol, Vol. 6, No. 7, 1999 670 S. N. HOCHWALD ET AL. treatment. We believe that those patients who have tumors that are limited to the mucosa by ultrasonography and that are less than 4.5 cm in diameter should be initially considered for limited resection without lymphadenectomy. Further information, which should be available in the preoperative period and may aid in decision making, is histological differentiation of the tumor. Poorly differentiated tumors or those with evidence of venous invasion should not be considered for limited resection without lymphadenectomy. Acknowledgment: Supported by the Richard Gelb Foundation. REFERENCES 1. Murakami T. Pathomorphological diagnosis: definition and gross classification of early gastric cancer. Gann Monogr Cancer Res 1971;11:53–5. 2. Itoh H, Oohata Y, Nakamura K, Nagata T, Mibu R, Nakayama F. Complete ten-year postgastrectomy follow-up of early gastric cancer. Am J Surg 1989;158:14 – 6. 3. Shiozawa N, Kodama M, Chida T, Arakawa T, Tur GE, Koyama K. Recurrent death among early gastric cancer patients: 20 years’ experience. Hepatogastroenterology 1994;41:244 –7. 4. Farley DR, Donohue JH. Early gastric cancer. Surg Clin N Am 1992;72:401–21. 5. Friesen G, Dockerty MB, ReMine WH. Superficial carcinoma of the stomach. Surgery 1962;51:300 –12. 6. Lehnert T, Sternberg SS, Sprossmann M, DeCosse JJ. Early gastric cancer. Am J Surg 1989;157:202–7. 7. Lawrence M, Shiu MH. Early gastric cancer. Ann Surg 1991;213: 327–34. 8. Everett SM, Axon ATR. Early gastric cancer in Europe. Br J Surg 1997;142–50. Ann Surg Oncol, Vol. 6, No. 7, 1999 9. Moreaux J, Bougaran J. Early gastric cancer: a 25-year surgical experience. Ann Surg 1993;217:347–55. 10. Bonenkamp JJ, Songun I, Hermans J, et al. Randomised comparison of morbidity after D1 and D2 dissection for gastric cancer in 996 Dutch patients. Lancet 1995;345:745– 8. 11. Cuschieri A, Flayers P, Fielding J, et al. Postoperative morbidity and mortality after D1 and D2 resections for gastric cancer: preliminary results of the MRC randomised controlled surgical trial. Lancet 1996;347:995–9. 12. Nishi M, Ishihara S, Nakajima T, Ohta K, Ohyama S, Ohta H. Chronological changes of characteristics of early gastric cancer and therapy: experience in the Cancer Institute Hospital of Tokyo, 1950 –1994. J Cancer Res Clin Oncol 1995;121:535– 41. 13. Yamazaki H, Oshima A, Murakami R, Endoh S, Ubukata T. A long-term follow-up study of patients with gastric cancer detected by mass screening. Cancer 1989;63:613–7. 14. Houghton PWJ, Mortensen NJM, Allan A, Williamson RCN, Davies JD. Early gastric cancer: the case for long term surveillance. BMJ 1985;291:305– 8. 15. Percivale P, Bertoglio S, Muggianu M, et al. Long-term postoperative results in 54 cases of early gastric cancer: the choice of surgical procedure. Eur J Surg Oncol 1989;15:436 – 40. 16. Guadagni S, Reed PI, Johnston BJ, et al. Early gastric cancer: follow-up after gastrectomy in 159 patients. Br J Surg 1993;80: 325– 8. 17. Shimizu S, Tada M, Kawai. Early gastric cancer: its surveillance and natural course. Endoscopy 1995;27:27–31. 18. Sue-Ling HM, Martin I, Griffith J, et al. Early gastric cancer: 46 cases treated in one surgical department. Gut 1992;33:1318 –22. 19. Sano T, Kobori O, Muto T. Lymph node metastasis from early gastric cancer: endoscopic resection of tumour. Br J Surg 1992; 79:241– 4. 20. Yamao T, Shirao K, Ono H, et al. Risk factors for lymph node metastasis from intramucosal gastric carcinoma. Cancer 1996;77: 602– 6. 21. Miller FH, Kochman ML, Talamonti MS, Ghahremani GG, Gore RM. Gastric cancer: radiologic staging. Radiol Clin North Am 1997;35:331– 49.