Macroscopic, endoscopic, and histopathologic classification of colorectal carcinoma
The tumor size (mm) was measured along the major axis on the resected specimen. The macroscopic shape of the carcinoma was classified as elevated or depressed (10, 11). The elevated type was further subclassified as pedunculated (Ip), semipedunculated/sessile (Isp/Is), creeping (12), or superficially elevated (IIa) (10, 11). The lesions were also classified based on the histologic type and the level of invasion. The histologic types included well-differentiated, moderately-differentiated, poorly-differentiated, signet ring cell, and mucinous adenocarcinoma. The tumors were also classified according to the invasion level. The Japanese Classification of Colorectal Carcinoma proposed by the Japanese Society for Cancer of the Colon and Rectum (10) was used to distinguish between minimally and massively submucosa-invasive colorectal carcinoma. Carcinoma invading the submucosal layer to a depth less than 300¦Ìm was considered minimally invasive.
Lymph node metastasis was diagnosed histologically in the laparotomy cases, and was diagnosed using computer tomography in the cases treated with endoscopic resection. The lymph node size over 1 cm was decided to be metastasis.

Endoscopic resection technique
Standard snare electrocautery techniques were used to endoscopically resect the Ip lesions. A modified endoscopic mucosal resection technique (13) was used for the Isp/Is, creeping, IIa, and depressed lesions. Briefly, after the tumor was identified using a video fiber optic colonoscope (Olympus CF-200I/CF-230I or CF-20I,. Tokyo, Japan), saline was injected into the submucosal layer under the tumor. The superficial-half of the elevated mucosa, including the tumor, was then resected using the specially designed hexagonal
Snare (13). This snare could be firmly fixed near the tumor with the attached needle, and completely encircled the lesions. The superficial-half of the elevated mucosa was resected because of avoiding damage to the muscle layer. A biopsy was performed using a diathermy (hot biopsy) to resect the tumors less than 5 mm in size.

Statistical analysis
A logistic regression analysis with the stepwize forward selection method was performed using a computer software; Statistical Analysis System (SAS, SAS Institute Co. Ltd). A p -value less than .05 was considered significant.

Results
Correlation between the histologic type and the invasion level
Table 1 shows the correlation between the histologic type and the invasion level of the resected colorectal carcinoma tumors. The rate of well-differentiated adenocarcinoma in all the tumors was 92.4%, and 6.4, 0.3, 0.6, and 0.3 % for moderately-differentiated, poorly-differentiated, signet ring cell, and mucinous adenocarcinoma, respectively. According to the results in Table 1, intramucosal or minimally submucosa-invasive carcinoma was found in 96 % (293/305) of the well-differentiated adenocarcinoma lesions. Conversely, more than 50 % of the signet ring cell and moderately-differentiated adenocarcinoma lesions were massively submucosa-invasive. Further, all of the poorly-differentiated and mucinous adenocarcinoma lesions were also massively invasive. There was no evidence of metastasis to the lymph nodes or vessels in any of the intramucosal or minimally invasive submucosal lesions (n=302), however 14.3 % (4/28) of the massively submucosa-invasive lesions (2 moderately differentiated, 1 signet ring cell, and 1 mucinous adenocarcinoma) had associated lymph node metastasis.

Correlation between the macroscopic finding (size and shape) and the invasion level
No massively submucosa-invasive colorectal carcinoma was found in the IIa lesions, whereas 2.2 % (2/93) of the Ip and 4.3 % (1/23) of the creeping tumors were massively submucosa-invasive (Table 2). Two of the three massively invasive lesions (1 creeping and 1 Ip ) were greater than 21 mm in size, and the size of the other one (Ip) was 8 mm. Massively submucosa-invasive carcinoma was also identified in the Isp/Is (7.3 %) and depressed (38.9%) tumors. Interestingly, 4 of which were less than 10 mm in size. These data suggest that any size of early colorectal carcinoma belongs to the IIa type, and the Ip or creeping type lesions less than 20 mm in size can be endoscopically resected when they were well-differentiated adenocarcinoma. However indication of endoscopic resection for the Isp/Is and the depressed lesions seem to depend on the size. In addition, there were 4 cases of positive lymph node metastasis; the shape, size, and histological type were as follows; Is/34 mm (moderately-differentiated adenocarcinoma), Isp/21 mm (signet ring cell carcinoma), Isp/16 mm (mucinous adenocarcinoma), and depressed/25 mm (moderately-differentiated adenocarcinoma).

Correlation between the endoscopic finding and the invasion level
The 41 submucosa-invasive colorectal carcinoma lesions (16 minimally and 25 massively) which were clearly observed by a colonoscope were analyzed to investigate the correlation between the endoscopic finding and the invasion level. The creeping lesions were excluded from this analysis because of lack of the correlation between the size and the level of invasion in this type (12). Since the characteristic endoscopic findings of submucosa-invasive colorectal carcinoma have been reported to include impression of tenseness, easy bleeding, white coating (erosion), white spots, and abnormal micro vessels (14-23), a logistic regression analysis for the each finding was performed between the minimally and massively submucosa-invasive lesions (Table 3). Interestingly, impression of tenseness was only the statistically significant (p<0.05) endoscopic finding to predict massively submucosa-invasive carcinoma. The ŽÒ tenseness ŽÓ was used for the lesions which had less mobility (within 2 mm) when pushed by an forceps. These lesions also had less transformation of the shape during air insufflation. Other endoscopists have expressed ŽÒ tenseness ŽÓ as solid impression (14, 15), stiffness (18, 22), or expansiveness (14, 24).

Logistic regression analysis for Isp/Is and depressed colorectal carcinoma
To decide indications of endoscopic resection for the Isp/Is and the depressed colorectal carcinoma, the other logistic regression analysis of the 187 resected lesions (162 intramucosal or minimally submucosa-invasive and 25 massively submucosa-invasive; 151 Isp/Is and 36 depressed) was performed using the size and shape as variables. We found that the possibility (P) of minimally submucosa-invasive carcinoma can be predicted using the following logistic regression analysis:

1
P=-----------------------------------
1+exp (-¦Â0 -¦²¦Âj¡¦xj)

(¦Â0: constant, ¦Âj: partial regression coefficient of xj, xj: explanatory variable of No. j)

where the shape variables were defined using the following grid: Shape
variable : Isp/Is depressed
X1 : 0 1

1
P =------------------------------------------------------------------
¡¡¡¡¡¡¡¡¡¡¡¡ 1+exp[-4.09313- (-2.0926¡¦x1 - 0.105465¡¦size)]

The confidence intervals of the analysis were shown in Table 4.

For example, for a 20 mm depressed lesion,

1
P=-------------------------------------------------------------
1+exp[-4.09313- (-2.0926 ¡ß1 - 0.105465 ¡ß20)]

1 1
=----------------------- = --------------------- = 0.4728342
1+exp (0.10877) 1+1.1149059


therefore, the possibility that the lesion would be minimally submucosa-invasive is 47%.
Possibilities of minimal invasion to the submucosal layer were calculated using the present logistic regression analysis for different sizes of the Isp/Is and the depressed colorectal carcinoma lesions (Table 5). Because the prior probability of this analysis was 87% (162 intramucosal or minimally submucosa-invasive lesions / 187 all lesions), the lesion whose possibility is less than 87% should be considered massively submucosa-invasive.

Discussion
This study was designed to endoscopically distinguish the intramucosal and minimally submucosa-invasive colorectal carcinoma from the massively submucosa-invasive one in order to avoid the unwarranted endoscopic resection for the massively invasive lesions. For this purpose, the histologic types, the macroscopic findings (size and shape), the endoscopic findings, and the invasion levels of 330 early colorectal carcinoma resected endoscopically or by laparotomy in our and satellite hospitals were analyzed. We also performed two different logistic regression analysis to investigate the correlations of the invasion level to the endoscopic finding (for 41 submucosa-invasive lesions), and the size (for 187 Isp/Is and depressed lesions), respectively.
Histologically, we found that the proportion of massive invasion to the submucosal layer in all the well-differentiated adenocarcinoma lesions was only 3.9% (12/305), whereas that in the moderately-differentiated, poorly-differentiated, signet ring cell or mucinous adenocarcinoma was over 50% (Table 1). Further, while no lymph node metastasis was found in the well-differentiated adenocarcinoma lesions (0/305), 9.5% (2/21), 50% (1/2), and 100% (1/1) of the metastasis were detected in the moderately differentiated, signet ring cell, and mucinous adenocarcinoma lesions, respectively. These results indicate that only the well-differentiated colorectal adenocarcinoma lesions are suitable targets for endoscopic resection. Many recent reports have demonstrated that well-differentiated colorectal adenocarcinoma is less likely to invade to the submucosal layer and accompany with lymph node metastasis than other histologic ones (6, 25-28). For example, Egashira et al. (29) have histologically analyzed 44 surgically resected colorectal adenocarcinoma invading to the submucosal layer, and found that predominant moderately-differentiated adenocarcinoma showed significantly deeper submucosal invasion than predominant well-differentiated one (29). They also reported that the positive rate of lymph node metastasis in predominant moderately-differentiated adenocarcinoma tended to be higher than that in predominant well-differentiated one. These reports may confirm our data about the correlation between the histologic type and invasion level of early colorectal carcinoma.
Since our results showed that any size of the IIa colorectal carcinoma lesions as well as all of the creeping and the almost Ip ones less than 20 mm in size were not massively submucosa-invasive (Table 2), it is likely that these lesions can be treated with endoscopic resection when they are well-differentiated adenocarcinoma. Our data could be supported by Kudo et al. (11) and Terai et al. (30) who reported that no or few massively submucosa-invasive carcinama were found in the Ip and the IIa type colorectal carcinoma in any size. The creeping colorectal tumors are also reported to have laterally-spreading growth tendency without invasion to the submucosal layer (13, 31, 32). Conversely, our results demonstrated that the rate being massively submucosa-invasive in the Isp/Is and the depressed lesions increased in a size-dependent manner (Table 2), so that the logistic regression analysis in the present study may support the decision of the treatment way for these lesions (Table 5). Because the prior probability of this analysis was 87% (162 intramucosal or minimally submucosa-invasive lesions / 187 all lesions), the colorectal carcinoma lesion whose possibility is less than 87% is likely to be massively submucosa-invasive. According to the results in Table 5, all of the depressed lesions and the Isp/Is ones over 20 mm in size are likely to be massively submucosa-invasive, and they should be treated with laparotomy. This analysis was chosen in this study because the variable (size) was irregular distribution. When the variable is regular, a discriminant analysis can be also used.
To further support the endoscopic distinction between minimally and massively submucosa-invasive colorectal carcinoma, we investigated the correlation between the endoscopic finding and the invasion level of the carcinoma using the other logistic regression analysis, and found that only impression of tenseness, but not easy bleeding, white coating, white spots, or abnormal microvessels is the finding predicting massively submucosa-invasive one (Table 3). This finding may be useful when a treatment plan (endoscopic resection or laparotomy) for the equivocal cases, especially the Isp/Is colorectal carcinoma lesions, is decided. Our results could be supported by the reports from Oda et al. (24) and Tamura et al. (33) who also asserted that this endoscopic finding is a good sign for colorectal tumors with massively submucosal invasion.
In conclusion, our findings in this study suggest that the early colorectal well-differentiated adenocarcinoma lesions which belongs to the IIa type in any size, and the creeping or Ip type less than 20 mm in size, are suitable targets for endoscopic resection. Conversely, other histologic type lesions should be resected by laparotomy. The logistic regression analysis in this study suggests that all of the depressed and the Isp/Is lesions over 20 mm in size should be also treated with laparotomy. This analysis technique also suggests that the endoscopic finding, impression of tenseness, may be a helpful sign predicting massively submucosal invasion, especially in the equivocal cases.

Acknowledgement
We thank Dr. Koji Nagahara in Koseiren Bisai Hospital and Dr. Kazuo Goto in Gifu Prefecture Tajimi Hospital for their great help to collect materials. We also thank Mr. Norio Sugimoto (Sanwa Kagaku Co. Ltd., Nagoya, Japan) for his great help to analyze our data.

References
1. Witt TR, Winawer SJ. Cancer in a colonic polyp, or malignant colonic adenomas - is polypectomy sufficient? Gastroenterology 1981; 81: 625-630.
2. William IW, Hiromi S. Definitive treatment of "malignant" polyps of the colon. Ann. Surg. 1975; 182: 516-525.
3. George EB, Douglas AS, Charles JM, Maus S, Michael RD. Carcinoma arising in adenomas of colon and rectum. JAMA 1952; 148: 1465-1469.
4. Clayton HS, Paul HL, Victor AG, Henry S. The treatment of pedunculated adenomatous colorectal polyps with focal cancer. Surgery, Gynecology & Obstetrics 1974; 139: 845-850.
5. Hase K, Mochizuki H, Utsunomiya K, Yoshizumi T, Yoshimura K, Kuranaga K, et al. Management of submucosal invasive colorectal cancer in view of long-term follow-up outcome [in Japanese with English abstract]. The Japanese Journal of Gastroenterological Surgery 1996; 29: 1013-1021.
6. Tanaka S, Haruma K, Teixeira CR, Tatsuta S, Ohtsu N, Hiraga Y, et al. Endoscopic treatment of submucosal invasive colorectal carcinoma with special reference to risk factors for lymph node metastasis. Journal of Gastroenterology 1995; 30: 710-717.
7. Oohara N, Nagasako K, Baba R, Tanaka Y, Sato S, Yashiro K, et al. A study on the lymph-node metastasis and the prognosis in patients of colorectal sm cancer [in Japanese with English abstract]. The Journal of The Japan Society of Coloproctology 1991; 44: 952-956.
8. Ohta H, Ueno M, Seki M, Tsuchiya S, Kinoshita M, Yamada H, et al. Reoperation after endoscopical polypectomy of early (SM) carcinoma of the colorectum [in Japanese with English abstract]. Journal of Japan Surgical Association 1993; 54:1742-1746.
9. Tanaka S, Yokota T, Saito D, Okamoto S, Oguro Y, Yoshida S. Clinicopathologic features of early rectal carcinomo and indications for endoscopic treatment. Dis Colon Rectum 1995; 38: 959-963.
10. Yasutomi M, Baba S, Hojo K, Kato Y, Kodaira S, Koyama Y, et al., editors. Japanese classification of colorectal carcinoma First english edition Part III Endoscopic findings and nanagement. Tokyo: Kanehara &Co., Ltd., 1997. p. 67-74.
11. Kudo S. Endoscopic mucosal resection of flat and depressed types of early colorectal cancer. Endoscopy 1993; 25: 455-461.
12. Iwashita A, Yamada Y, Yao T, Arita M, Yao K, Uchida Y, et al. Clinicopathological study on colorectal flat elevation with conglomerated nodular surface [in Japanese with English abstract]. Stomach and Intestine 1992; 27: 409-419.
13. Kanamori T, Itoh M, Yokoyama Y, Tsuchida K. Injection- incision-assisted snare resection of large sessile colorectal polyps. Gastrointest Endosc 1996; 43: 189-195.
14. Masumitu H, Yoshida S, Tsubomizu Y, Okamoto H, Satake Y, Fujita R. Endoscopic diagnosis of early colorectal cancer befor polypectomy with special refference to the depth of cancer [in Japanese with English abstract]. Gastroenterol Endosc 1987; 29: 1755-1762.
15. Ooishi T, Egawa N, Ookawa H, Monma K, Tajima T, Yamanaka A. Endoscopic study on colorectal cancer with submucosal involvemenet [in Japanese with English abstract]. Gastroenterol Endosc 1991; 33: 2581-2587.
16. Ikegami M, Shimoda T, Komaki T, Kuroda H. Macroscopic features of submucosal invasive colorectal cancers [in Japanese with English abstract]. Stomach and Intestine 1994; 29: 1237-1247.
17. Yokota T, Matsui T, Fukuda H, Yokoyama T, Oka M, Hisai H, et al. Evaluation of the depth of early colorectal cancers by endoscopy [in Japanese with English abstract]. Stomach and Intestine 1994; 29: 1261-1269.
18. Koizumi K, Kazami A, Handa T, Kai S, Maruyama M, Yanagisawa A, et al. The endoscopic findings of the colorectal pedunculated submucosal invasive carcinoma [in Japanese with English abstract]. Early Colorectal Cancer 1998;2: 389-394.
19. Tamegai Y, Hamano K, Komatsubara N. Differential diagnosis of the degree of submucosal invasion with IIa+IIc type colorectal cancers [in Japanese with English abstract]. Early Colorectal Cancer 1998; 2: 401-410.
20. Okamura S, Segawa K, Ohashi S, Mitake M, Urano F, Shimodaira M, et al. Clinicopathological study and colonoscopic diagnosis of the NPG type colorectal cancer with submucosal invasion [in Japanese with English abstract]. Gastroenterol Endosc 1997; 39: 911-917.
21. Iwamoto Y, Nagasako K, Suzuki S, Hayashi N. The endoscopic diagnosis of invasion of colorectal sessile tumors [in Japanese with English abstract]. Gastroenterol Endosc 1997; 39: 1370-1375.
22. Okamoto H, Sakai Y, Tani T, Yamazaki T, Okada T, Saito H, et al. Characteristic endoscopic findings of colorectal submucosal invasive cancers [in Japanese with English abstract]. Gastroenterological Endoscopy 1996; 38: 2577-2582.
23.Muto T, Kamiya J, Sawada T, Sugihara K, Morioka Y. Clinicopathological study of white spots of the colonic mucosa around polyps, with special reference to the endoscopic diagnosis of invasive carcinoma. Gastrointest Endosc 1984;30: 231-233.
24. Oda Y, Fujii T, Koba I, Muro K, Boku N, Ohtus A, et al. Statistical analysis on the relationship between the degree of histological atypism and the endoscopic findings of colorectal sessile polyp [in Japanese with English abstract]. Gastroenterol Endosc 1995; 37: 760-766.
25. Okabe S, Murase N, Masahiro T, Endo M. Significance of components of moderately differentiated adenocarcinoma as an adverse prognostic factor in submucosal invasive colorectal carcinamoa [in Japanese with English abstract]. Endoscopia Digestiva 1996; 8: 937-944.
26. Yagyu T, Kikkawa N, Mishima H, Nishisyou I, Shin E, Hasuike Y, et al. Indication for resection following endoscopic polypectomy or local excision for invasive polyps of the lower rectum [in Japanese with English abstract]. The Japanese Journal of Gastroenterological Surgery 1997; 60: 920-924.
27. Inoue Y, Suzuki M, Yoshida K, Tezuka T, Takasaki K, Murata Y, et al. A Study of Submucosal Carcinoma of the Colon and Rectum with Lymph Node Metastasis [in Japanese with English abstract]. The journal of the Japan society of coloproctology 1998; 51: 159-167.
28. Ichinose M, Ikegami M, Matsushima M, Suzuki H, Ushigome S. Colorectal carcinomas with submucosal invasion: A clinicopathological study on risk factors for heatic and lymphnodal metastasis [in Japanese with English abstract]. Gastroenterol Endosc 1999; 41: 1163-1174.
29. Egashira Y, Hirata I, Katu K. Relationship between differentiation of carcinoma and grading of malignancy in colorectal carcinoma invading the submucosal layer [in Japanese with English abstract]. Endoscopia Digestiva 1996;8: 915-921.
30. Terai T, Sakamoto N, Nihei H, Ohno Y, Kobayashi O, Miwa H, et al. Endoscopic findings of superficial, elevated type submucosal invading cancers of the colon and rectum-special reference to Iia type tumors and Laterally spreading tumors [in Japanese with English abstract]. Early Colorectal Cancer 1998; 2: 411-419.
31. Watanabe K, Okawa K, Ishiguro S, Oba H, Moriyoshi Y, Nebiki H, et al. Investigation of colorectal creeping tumor [in Japanese with English abstract]. Gastroenterol Endosc 1995;37:754-759.
32. Mizuta Y, Nomoto T, Kajiyama H, Minamino A, Minamino Y, Yano Y,et al. Endoscopic diagnosis and management of colorectal nodule-aggregating lesions [in Japanese with English abstract]. Gastroenterol Endosc 1997; 39: 169-174.
33. Tamura S, Kudo S, Nakajima T, Yamano H, Asano M, Kusaka H, et al. Depth diagnosis of early colorectal cancer from the viewpoint of endoscopic findings [in Japanese with English abstract]. Gastroenterol Endosc 1997; 39: 1781-1792.