Endoscopically Removed Malignant Colorectal Polyps: Clinicopathologic Correlations

Transcription

1 GASTROENTEROLOGY 1995;108: Endoscopically Removed Malignant Colorectal Polyps: Clinicopathologic Correlations HARRY S. COOPER,* LUDWIG M. DEPPISCH,* WILLIAM K. GOURLEY, ~ ELLEN I. KAHN, [I ROBERT LEV, PAUL N. MANLEY, # ROBERT R. PASCAL,** ALl H. QIZILBASH, *t ROBERT R. RICKERT, ~ JAN F. SILVERMAN, IIII and JOHN A. WIRMAN *Department of Pathology, Fox Chase Cancer Center, Philadelphia, Pennsylvania; ~Department of Pathology, Northside Medical Center, Western Reserve Care System, Youngstown, Ohio; ~Department of Pathology, University of Texas Medical Branch, Galveston, Texas; fldepartment of Pathology, North Shore University Hospital, Manhasset, New York; ~Department of Pathology, Roger Williams General Hospital, Providence, Rhode Island; #Department of Pathology, Kingston General Hospital, Kingston, Ontario, Canada; **Department of Pathology, Emory University Hospital, Atlanta, Georgia; ttdepartment of Pathology, North York General Hospital, North York, Ontario, Canada; ~Department of Pathology, St. Barnabas Medical Center, Livingston, New Jersey; IIIIDepartment of Pathology, East Carolina University School of Medicine, Greenville, North Carolina; and ~Department of Pathology, University of Cincinnati Medical Center, Cincinnati, Ohio Background~Aims: Treatment options for patients with endoscopically removed malignant colorectal polyps are polypectomy alone vs. polypectomy followed by surgery. The aim of this study was to define histopathologic parameters that can be used for clinically relevant treatment decisions. Methods: Five pathologists evaluated 140 polyps for the presence or absence of unfavorable histology. Unfavorable histology was tumor at or near (-<1.0 mm) the margin and/or grade III and/or lymphatic and/or venous invasion. Adverse outcome was recurrent and/or local cancer and/or lymph node metastasis. Results: Adverse outcome was 19.7% (14 of 71), 8.6% (2 of 23), and 0% (0 of 46) when unfavorable histology was present, indefinite (lack of agreement), and absent, respectively (P < , present vs. absent). Four patients with cancer >1.0 mm from the margin had an adverse outcome (2 with lymphatic invasion and 2 indefinite for lymphatic invasion). Four patients with negative resections later developed distant metastases. Eight patients (6.3%) died of disease, and 2 of 69 without unfavorable histology (both indefinite for lymphatic invasion) had an adverse outcome. Interobserver strength of agreement was substantial to almost perfect for margin, grade, and venous invasion and fair to substantial for lymphatic invasion. Conclusions: This system is usable clinically. Patients with unfavorable histology are probably best managed by resection postpolypectomy, whereas in the absence of unfavorable histology, they probably can be treated by polypectomy only. T he malignant polyp of the colorectum represents a form of early colorectal cancer. These lesions are often removed via endoscopic polypectomy. After examination by a pathologist and a diagnosis of malignant polyp, the clinician is presented with a therapeutic decision. Is the polypectomy itself adequate therapy, or does the patient need a subsequent definitive surgical resection? There have been many reports addressing this question) -18 However, in some reports, the follow-up for patients treated by polypectomy only has been short, 1'7'9'l~'17'18 whereas in other series, not all polyps were initially removed endoscopically. 4al Some series reported only on cases with subsequent resection, s'h and some series seem to rely on review of pathology records rather than having a pathologist carefully review the slides as a coinvestigator. 7'9 In many series, it is difficult to ascertain the outcome as it relates to the interactions of the different parameters studied. Accordingly, we undertook a large multi-institutional study of endoscopically removed malignant colorectal polyps to determine which parameters provide information as to how best to manage a patient with an endoscopically removed malignant polyp. To date, this is the largest clinicopathologic study of endoscopically removed malignant colorectal polyps that has studied the outcome of polyps treated by both polypectomy only and polypectomy followed by subsequent resection. In this study, the minimum follow-up for patients treated by polypectomy only was 5 years. We also provide data as to the interaction of histological parameters regarding outcome. The data from multiple interobserver reviews provide information as to the usability of this system and its pitfalls and/or limitations. Materials and Methods Cases of endoscopically removed malignant colorectal polyps were solicited from the membership of the Gastrointestinal Pathology Society. Criteria for inclusion into the study 1995 by the American Gastroenterological Association /95/53.00

2 1658 COOPER ET AL. GASTROENTEROLOGY Vol. 108, No. 6 were the following. (1) The polyp was initially removed endoscopically. (2) The polyp had to be a malignant polyp, which is defined as a polyp in which cancer has invaded into the submucosa. (3) Those malignant polyps treated by polypectomy only had to have a minimum of a 5-year follow-up period. (4) Any malignant polyp that was initially obtained via endoscopic polypectomy and with subsequent resection was considered usable for the study, regardless of the length of follow-up (e.g., the resection specimen in and of itself provided immediate follow-up regarding the presence or absence of lymph node metastases and/or residual local cancer). Five Gastrointestinal Pathology Society members (H.S.C., L.M.D., E.I.K., R.R.P., and R.R.R.) reviewed all of the slides. The results of this study reflect the data obtained from these 5 reviewers. Each case was evaluated according to a form specially prepared for this study. Criteria for margins, grade, lymphatic invasion, and venous invasion were circulated to the study group before review. The cases were evaluated for the following. (1) Whether the polyp was a malignant polyp (submucosal invasion). (2) Grade of the cancer (I, II, or III). 19 Comments were also made whether the cancer had a colloid or mucinous component. (3) The status of the resection margin (e.g., whether the cancer was or was not at the resection margin). Our definition of the margin of resection was the actual free edge of the submucosal transsection point that contained diathermy change. (4) The distance of cancer from the resection margin as measured with a micrometer. Three reviewers used a micrometer placed into the ocular of the microscope, whereas 2 reviewers used a handheld micrometer. (5) The presence or absence of lymphatic invasion by cancer as defined by tumor ceils within a true endothelial-lined channel in the absence of red blood cells. 2 (6) The presence or absence of venous invasion as defined by cancer in an endothelial-lined channel surrounded by a smooth muscle wall. 2 (7) The presence of any background adenoma or polypoid carcinoma. The definitions for these histological parameters were identical to those used daily by pathologists in their clinical practice Histological parameters for each case (e.g., lymphatic invasion, grade of cancer, status of margins, and so on) were considered present or absent only when at least 4 of the 5 pathologists agreed. A parameter was considered indefinite when the split was 3 vs. 2 or 2 vs. 3. While each reviewer measured the actual distance of cancer from the resection margin, the data for margins were evaluated as tumor at the margin (positive) or -< 1 mm from the margin, >1 and -<2 mm from the margin, and >2 mm from the resection margin. For the evaluation of margin status as it related to outcome (adverse vs. no adverse), the margin > 1 mm and --<2 mm group and the margin >2 mm group were combined into one group called margin negative (tumor > 1 mm from the resection margin). This was based on the data showing that the margin >1 mm and --<2 mm group was statistically no different than the margin >2 mm group as related to outcome, and both groups were each statistically significantly different than the margin --<1 mm group as related to outcome. In 15 cases, 2 slides were reviewed. In 2 cases, 3 slides were reviewed; in 150 cases, only 1 slide was reviewed. The configuration of the polyp (e.g., pedunculated, sessile, pseudopedunculated, or unknown) was determined from the gross macroscopic configuration of the polyp on the slide, gross descriptions from the pathology reports, and endoscopic reports. One can confidently confirm the configuration of a polyp as pedunculated from the gross appearance of the slide. However, for a polyp to be considered sessile or pseudopedunculated (a short-stalked or pseudostalked polyp readily distinguishable from a sessile polyp per endoscopic reports and personal communication with experienced endoscopists), this information had to come from actual endoscopic findings. The category of unknown configuration was used when there was no gross macroscopic evidence of a pedunculated polyp or no conclusive endoscopic findings. Adverse outcome was defined as recurrence (local and/or distant) in those cases treated by polypectomy only and positive lymph nodes and/or residual local cancer in definitive surgical resection specimens. Adverse outcome was defined as above to simulate the actual clinical situation faced by the physician when an endoscopically removed polyp is diagnosed as a malignant polyp (e.g., are there histological parameters that can accurately predict that no residual local cancer and/or no lymph node metastases will be present and/or is there is a good chance that residual local cancer is present or lymph node metastases are present), and they are faced with the decision of polypectomy alone vs. subsequent definitive surgical resection. In this study, an unfavorable histological parameter is defined as cancer at or near (--< 1 ram) the resection margin and/ or grade III cancer and/or lymphatic invasion by cancer and/or venous invasion by cancer. A favorable histological parameter is the absence of all of the above. These definitions are based on previous studies from the literature. ~-3'5-8'11-14 The indefinite category is defined as above. Interobserver analysis was analyzed by 1( analysis 21 and analysis of variance. Fisher's Exact Test was used to compare groups. A P value of <0.05 was considered significant. Survival statistics were performed using the Kaplan-Meier product limit method. 22 Differences in survival between the polypectomy only vs. polypectomy and resection groups were calculated by Wilcoxon rank-sum test using "c statistics for ranks. Results One hundred sixty-seven polyps were submitted for the study. Of these, 140 met the criteria for inclusion into the study. Eleven were considered not to be malig- nant polyps, 4 were indefinite for the diagnosis of a malignant polyp, 4 were poorly oriented and could not be evaluated, and 8 polypectomy-only cases did not meet the minimum 5-year follow-up period. Of the 140 polyps studied, 36 were treated by polypectomy only and 104 were treated by polypectomy followed by subsequent surgical resection.

3 June 1995 MALIGNANT COLORECTAL POLYPS 1659 Table 1. Outcome Related to Histological Parameters Unfavorable Adverse No adverse histology a outcome (%) outcome (%) Total (%) Present 14/71 (19.7) 57/71 (80.3) e,~'g 71/140(50.7) Indefinite 2/23 (8.6) ~'~ 21/23 (91.3) ~ 23/140(16.4) Absent 0/46 (0.0) 46/46 (100.0) 46/140(32.8) Total 16/140 (11.4) d 124/140 (88.6) 140/140 (100.0) NOTE. Regarding adverse vs. no adverse outcome: present unfavorable vs. indefinite unfavorable, NS; present unfavorable vs. absent unfavorable, P < ; indefinite unfavorable vs. absent unfavorable, NS; polypectomy only vs. polypectomy followed by resection, NS. Abbreviations for Tables 1-6: M+, cancer at or near (-<1.0 mm from) margin; M-, cancer >1.0 mm from margin; LI+, LI~, LI, lymphatic invasion present, indefinite, and absent; Ill+, IIl~, II1-, grade III cancer present, indefinite, and absent; VI+, Vl~, VI-, venous invasion present, indefinite, and absent. "Unfavorable histology is tumor at or near margin (~1.0 mm) and/or lymphatic invasion present and/or venous invasion present and/or grade III cancer. See Materials and Methods for definition of present, indefinite, and absent. bone patient with Ll~. COne patient with LI~ and colloid cancer. dthir[een patients with lymph node metastases and 1 patient with local residual cancer in resection specimen (palypectomy followed by resection); 2 patients with distant and/or local recurrence (polypectomy only). Adverse outcome in polypectomy only cases, 5.5% (2 of 36). Adverse outcome in polypectomy followed by resection cases, 3.5% (14 of 104). "One patient with negative resection (grade Ill signet cell cancer and M+) and distant metastases 4 years later, tone patient with negative resection, LI+, and liver metastases 2 years later. gone patient with negative resection, LI+ and VI+, and liver metastases 7 years later. hone patient with negative resection, II1~ and Ll~, and liver metastases 2 years later. The median and mean follow-up of those treated by polypectomy only was 7.0 and 8.1 years, respectively. Follow-up of those that had polypectomy followed by a negative surgical resection was 6.0 (median) and 5.8 (mean) years. Follow-up of those that had polypectomy followed by resection with positive lymph nodes and/or local disease was 5.0 years (median) and 5.5 years (mean). Eleven patients with surgical resection postpolypectomy were lost to follow-up subsequent to their surgical resection, and 2 of these 11 had positive lymph nodes in their resection specimen. Of the 140 polyps studied, 16 (11.4%) had an adverse outcome. The incidence of adverse outcome in the polypectomy only group and polypectomy followed by resection group was 5.5% (2 of 36) and 13.5% (14 of 104), respectively. This difference was not statistically significant. Thirteen patients had lymph node metastases, 1 had residual local cancer in the resection specimen, and 2 treated by polypectomy only had distant and/or local recurrences. This data is presented in Table 1 as it relates to the presence, indefinite, or absence of unfavorable histology. Of the 69 that did not have unfavorable histological parameters present (indefinite and absent group), two (2.9%) had an adverse outcome; one was a colloid cancer indefinite for lymphatic invasion, and one was indefinite for lymphatic invasion. In the group with an adverse outcome, the polyp size was cm (mean, 1.7). Eleven were from the sigmold colon, 2 from the rectum, and 3 from unknown sites. There were 8 men and 8 women ranging in age from 47 to 74 years (mean, 59 years). In the group without an adverse outcome, the polyp size was 0.5-~.0 cm (mean, 1.4 cm). Seventy-two were from the sigmoid colon, 19 from the rectum, 14 from the descending colon, 2 from the transverse colon, and 17 from unknown sites. There were 74 men and 50 women ranging in age from 38 to 84 years (mean, 68 years). Margins Table 2 shows in detail the outcome regarding the status of resection margins. Twelve of 56 patients (21.4%) with tumor at or near (--<1 mm) the resection margin had an adverse outcome. The presence of tumor at or near the margin vs. the margin negative group was significant regarding outcome (P < 0.003). Four patients with a tumor > 1.0 mm from the resection margins had an adverse outcome. Two of those 4 patients had lymphatic invasion, and 2 had histological findings indefinite for lymphatic invasion. No patient with a tumor >1 mm from the resection margin and absent unfavorable histological parameters had an adverse outcome. Lymphatic and Venous Invasion Tables 3 and 4 show the outcome regarding lymphatic and venous invasion, respectively. Seventeen (12.1%) and 22 (15.7%) patients were diagnosed with lymphatic invasion and indefinite for lymphatic invasion, respectively. Three of 17 patients (17.6%) with lymphatic invasion and 4 of 22 patients (18.1%) indefinite for lymphatic invasion had an adverse outcome. There was a significant difference (P < 0.025) between the lymphatic invasion present group and the lymphatic invasion absent group regarding outcome. There was a significant difference (P < 0.04) between the indefinite for lymphatic invasion (and absent unfavorable histology) group vs. the absent for lymphatic invasion (and absent unfavorable histology) group regarding outcome. The incidence of venous invasion was 3.5% (5 of 140), and adverse outcome was 40% (2 of 5); however, all with an adverse outcome were associated with the presence of other unfavorable histological parameters. Grade III Cancer Table 5 shows data regarding grade. Of interest, all grade III cancers were associated with the presence of other unfavorable histology (P < 0.01).

4 1660 COOPER ET AL. GASTROENTEROLOGY Vol. 108, No. 6 Table 2. Outcome Regarding Status of Resection Margin Margin positive or close (--<1 mm) (n = 56) (40.5%) Margin negative (>1 mm) (n = 82) (59.4%) Unfavorable histology Adverse (%) No adverse (%) Total (%) Adverse (%) No adverse (%) Total (%) Present 4 (50.0) 4 (50.0) a 8 (100) 2 (13.3) b'c 13 (86.6) a'e 15 (100) Indefinite 1 (8.3) 11 (91.6) 12 (100) 2 (9.5) f'g 19 (90.4) h 21 (100) Absent 7 (19.4) 29 (80.5) 36 (100) 0 (0) 46 (100) 46 (100) Total 12 (21.4) 44 (78.6) 56 (100) 4 (4.9) 78 (95.1) 82 (100) NOTE. Regarding adverse vs. no adverse outcome: M+ vs. M-, P < M+ (unfavorable histology, present vs. indefinite, P = 0.05; present vs. absent, NS; indefinite vs. absent, NS). M- (unfavorable histology, present vs. indefinite, NS; present vs. absent, P = 0.057; indefinite vs. absent, NS). M+ (unfavorable histology present) vs. M- (unfavorable histology present), NS; M+ (unfavorable histology absent) vs. M- (unfavorable histology absent), P < aone patient with negative resection (111+ and M+), distant metastases 4 years later. ~One patient with Ll+. cone patient with LI+. aone patient with negative resection (LI+) and distant metastases 2 years later. eone patient with negative resection (LI and VI+), distant metastases 7 years later. lone patient with Ll~. gone patient with Ll~ and colloid cancer. hone patient with negative resection (111~ and Ll~), distant metastases 2 years later. Polypoid Carcinoma Twenty-five (17.8%) were polypoid cancer, and 23 (16.4%) were indefinite for polypoid cancer. The incidence of adverse outcome was 20%, 17.4%, and 7.6% in the polypoid cancer, indefinite for polypoid cancer, and absent for polypoid cancer groups, respectively. However, all polypoid cancers with an adverse outcome (100%) were associated with the presence of unfavorable histological parameters. Polyp Configuration Seventy-four (52.8%), 17 (12.1%), 36 (25.7%), and 13 (9.3%) were pedunculated, pseudopedunculated, un- known, and sessile, respectively. The incidence of adverse outcome was 2.7%, 29.4%, 19.4%, and 15.4% for pedunculared, pseudopedunculared, unknown, and sessile cases, respectively. The difference in outcome between pedunculated vs. the other configurations (pseudopedunculated, unknown, and sessile cases) was due to the statistically significant association of the sessile, pseudopedunculated, and unknown groups with the presence of unfavorable histology. Survival Data In this study, there were 8 deaths (6.3%). Of the 8 patients who died, 2 (5.5%) were treated by polypectomy only and 6 (6.6%) with surgical resections postpo- Table 3. Outcome Regarding Lymphatic Invasion Lymphatic invasion present (n = 17) (12.1%) Lymphatic invasion indefinite (n = 22) (15.7%) Lymphatic invasion absent (n = 101)(72.1%) Unfavorable Adverse No adverse Total Adverse No adverse Total Adverse No adverse Total histology (%) (%) (%) (%) (%) (%) (%) (%) (%) Present 1 (16.6) 5 (83.3) ~ 6 (100) 2 (28.5) 5 (71.4) 7 (100) 9 (19.1) 38 (80.8) 47 (100) Indefinite 0 (0) 1 (100) 1 (100) 0 (0) 3 (100) 3 (100) 0 (0) 8 (100) 8 (100) Absent 2 (20) 8 (80) 10 (100) 2 (16.6) 10 (83.3) 12 (100) 0 (0) 46 (100) 46 (100) Total 3 (17.6) 14 (82.3) 17 (100) 4 (18.1) 18 (81.8) 22 (100) 9 (8.9) 92 (91.1) 101 (100) NOTE. Regarding adverse vs. no adverse outcome. LI+ vs. LI- = P < 0.025; LI+ vs. LIB, NS; LI~ vs. LI-, NS. LI+ (unfavorable histology; present vs. indefinite, NS; present vs. absent, NS; indefinite vs. absent, NS), Ll~ (unfavorable histology; present vs. indefinite, NS, present vs. absent, NS, indefinite vs. absent, NS), LI- (unfavorable histology; present vs. indefinite, NS; present vs. absent, P < ; indefinite vs. absent, NS). Ll+ (unfavorable histology absent) vs. LI- (unfavorable histology absent), P < LI+ (unfavorable histology present) vs. LI- (unfavorable histology present), NS. LI~ (unfavorable histology absent) vs. LI- (unfavorable histology absent, P < LI~ (unfavorable histology present) vs. LI- (unfavorable histology present), NS. LI+ and LI~ (unfavorable histology absent) vs, LI- (unfavorable histology absent), P < 0,008. LI+ and LIL (unfavorable histology present) vs. LI- (unfavorable histology present), NS. aone patient with negative resection (LI+ and VI+), distant metastases 7 years later. bone patient with negative resection (LI+) and distant metastases 2 years later. COne patient with negative resection (111~ and LI~), distant metastases 2 years later. ~One patient with negative resection (111+ and M+) and distant metastases 4 years later.

5 June 1995 MALIGNANT COLORECTAL POLYPS 1661 Table 4. Outcome Regarding Venous Invasion Venous invasion present Venous invasion indefinite (n = 5) (3.5%) (n = 9) (6.4%) Venous invasion absent (n = 126) (90.0%) Unfavorable Adverse No adverse Total Adverse No adverse Total Adverse No adverse Total histology (%) (%) (%) (%) (%) (%) (%) (%) (%) Present 2 (50) 2 (50) a 4 (100) 0 (0) 4 (100) 4 (100) 12 (19.4) 50 (80.6) 62 (100) Indefinite 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 2 (11.1) 16 (88.8) 18 (100) Absent 0 (0) 1 (100) 1 (100) 0 (0) 5 (100) 5 (100) 0 (0) 46 (100) 46 (100) Total 2 (40) 3 (60) 5 (100) 0 (0) 9 (100) 9 (100) 14 (11.0) 112 (89.0) 126 (100) NOTE. Regarding adverse vs. no adverse outcome: VI+ vs. VI-, NS; Vl vs. VIb NS; Vl~ vs. Vl-, NS. VI+ (unfavorable histology: present vs. indefinite, NS; present vs. absent, NS; indefinite vs. absent, NS). VII (unfavorable histology: present vs. indefinite, NS; present vs. absent, NS; indefinite vs. absent, NS). VI- (unfavorable histology: present vs. indefinite, NS; present vs. absent, P < 0.007; indefinite vs. absent, NS). aone patient with negative resection (LI+), distant metastases 7 years later. lypectomy. There was no statistical difference between the polypectomy only vs. the polypectomy followed by resection groups. Of the 6 polypectomy and resection cases with death, 4 had negative resections but developed distant metastases 2, 2, 4, and 7 years later. This latter figure represents 4.4% of negative resections in our study (4 of 90). Three of these had unfavorable histological parameters, and 1 was indefinite for both grade III and lymphatic invasion. Seven of 8 patients who died had the presence of unfavorable histological parameters. The only patient who died in the absence of the presence of unfavorable histological parameters was indefinite for both grade III and lymphatic invasion. Colloid Cancer There were 6 cases (4.3%) of colloid cancer. One (16.6%) had lymph node metastases and was also indefinite for lymphatic invasion. One with a signet cell carcinoma and cancer close to the resection margin had a negative resection but later died of metastatic disease. Interobserver Results Table 6 presents the data for interobserver variation. 1~ analysis 2 indicated that the strength of agreement was almost perfect for margins, substantial for venous invasion, and grade III and fair, moderate, and substantial for lymphatic invasion. Analysis of variance indicated that pathologist D was responsible for the fair strength of agreement for lymphatic invasion. If pathologist D was removed from lymphatic invasion, then the 1~ analysis shows that the strength of agreement was substantial for AB, AC, BC, and CE and moderate for BE and AE. One of the 2 with adverse outcome and diagnosed as indefinite for lymphatic invasion was considered positive for lymphatic invasion by the original pathologist. Literature Comparison For comparison, Tables 7 and 8 present data from the literature regarding outcome as it relates to margins, grade, lymphatic, and venous invasion. Discussion The malignant polyp of the colorectum represents a form of early colorectal cancer. The diagnosis is often made after an endoscopic polypectomy. At this point, the clinician is faced with the therapeutic decision as to whether polypectomy alone is adequate therapy or Table 5, Outcome Regarding Grade III Cancer Grade III cancer present Grade Ill cancer indefinite Grade Ill cancer absent (n = 8) (5.7%) (n = 10) (7.1%) (n = 122) (87.1%) Unfavorable Adverse No adverse Total Adverse No adverse Total Adverse No adverse Total histology (%) (%) (%) (%) (%) (%) (%) (%) (%) Present 3 (37.5) 5 (62.5) a 8 (100) 0 (0) 5 (100) 5 (100) 11 (18.9) 47 (81.0) 58 (100) Indefinite 0 (0) 0 (0) 0 (0) 0 (0) 2 (100) 2 (100) 2 (11.1) 16 (88.8) 18 (100) Absent 0 (0) 0 (0) 0 (0) 0 (0) 3 (100) 3 (100) 0 (0) 46 (100) 46 (100) Total 3 (37.5) 5 (62.5) 8 (100) 0 (0) 10 (100) 10 (100) 13 (10.7) 109 (89.3) 122 (100) NOTE. Regarding adverse vs. no adverse outcome: II1+ vs. II1-, P < 0.05; II1+ vs. II1~, NS; 111~ vs. II1-, NS. III- (unfavorable histology, present vs. indefinite, NS; present vs. absent, P < 0.001; indefinite vs. absent, NS). Ill+ (unfavorable histology present) vs. III- (unfavorable histology present), NS. "One patient with negative resection (M+) and distant metastases 4 years later. ~One patient with negative resection (LI~) and distant metastases 2 years later.

6 1662 COOPER ET AL. GASTROENTEROLOGY Vo1.108, No. 6 Table 6. Interobserver Diagnosis: Number of Events Diagnosed by Each Pathologist Margins Grade Lymphatic Venous Pathologist Agree Disagree Ill invasion invasion A B C D E NOTE. n = 140 malignant polyps. In 6 cases, margins were indefinite. Eight cases were grade III and 10 cases were grade IIl~, 17 cases were LI, and 22 cases were LI~. Five cases were VI and 9 cases were VII. K analysis values: margins (range, ), grade III (range, ), lymphatic invasion (range, ), venous invasion (range, ). See text. whether the patient should undergo a subsequent definitive surgical resection. Our study in conjunction with that in the literature provides data as to how to best handle these patients. In our study, there was an overall adverse outcome in 16 of 140 cases (11.4%). Thirteen had lymph node metastases, 1 had residual cancer in a resection specimen, and 2 had local and/or distant recurrence (polypectomy only cases). The incidence of adverse outcome was 14 of 71 (19.7%) with unfavorable histology present, 2 of 23 (8.6%) with indefinite unfavorable histology, and 0 of 46 (0%) in cases with absent unfavorable histology. Fourteen of 16 patients (87.5%) with an adverse outcome had unfavorable histology, 2 of 16 (12.5 %) with adverse outcome had indefinite histology, and 0 of 16 (0%) with an adverse outcome had absence of unfavorable histology. The two cases of adverse outcome associated with indefi- nite histology were both diagnosed as indefinite for lymphatic invasion. In our study, 21.4% of cases with cancer at or near the resection margin had an adverse outcome compared with 33% reported in the literature. 1'3'5-* '*3-15'1va8 We found that the presence of cancer at or near the margin was significantly associated with an adverse outcome even in the absence of other unfavorable parameters (P < 0.002). Of the 4 cases with an adverse outcome and tumor > 1 mm from the margin, 2 had lymphatic invasion and 2 were indefinite for lymphatic invasion. Hence, the incidence of adverse outcome was 0% with cancer > 1 mm from the margin in the absence of unfavorable histology. Review of the literature indicates that in the absence of unfavorable histology and a negative margin, the incidence of adverse outcome was 1.7%. 1'3'5< ' 13-15,17-18 In our study, 17 (12.1%) and 22 (15.7%) were diagnosed as positive and indefinite for lymphatic invasion, respectively. The incidence of adverse outcome was 17,6% (3) and 18.1% (4) in the positive and indefinite for lymphatic invasion groups, respectively. Review of the literature 1'5's indicates that the incidence of lymphatic invasion is 15.6% (range, 9.7%-44%), and 38.8% with lymphatic invasion had an adverse outcome (range, 16.6%-62.5%). In the literature, 1'%8 all (100%) with lymphatic invasion and an adverse outcome were associated with the presence of other unfavorable histology. In our study in the absence of other unfavorable histology, lymphatic invasion (P < 0.035), indefinite for lymphatic invasion (P < 0.04), and the combined group (P < 0.008) were significantly associated with an adverse outcome when compared with the negative for lymphatic Table 7. Outcome of Endoscopically Removed Malignant Colorectal Polyps as Related to Status of Resection Margins Polypectomy followed Polypectomy only by resection All cases Adverse Adverse NED (%) (%) Total NED (%) Adverse (%) Total NED (%) (%) Total Tumor at or near resection margin Negative resection margin 28 (84.8) 5 (15.2) (62.0) 46 (38.0) (67.0) 51 (33.0) (98.4) 4 (1.6) a'b (88.2) 10 (11.8) ' (96.0) 14 (4.0) (0.8) e'f 4 (4.7) e'g 6 (1.7) e'f'g NOTE. Adverse outcome: polypectomy only = recurrence (local or distant). Polypectomy followed by resection = residual local disease, and/ or lymph node metastases in resection specimen. Data from literature review. ±'3,5-1,13-15,17,18 NED, No residual disease in resection specimen (polypectomy followed by resection) and no recurrence following polypectomy only. aone patient with venous invasion. ~One patient with vascular invasion. CFour patients with grade III cancer. Two patients with vascular invasion. eexcluding cases with grade III cancer and/or lymphatic invasion and/or venous invasion. ~wo elderly patients cer[ified as dying of carcinomatosis. gfour patients with tumor in stalk, not otherwise specified. "Closeness" to resection margin unknown.

7 June 1995 MALIGNANT COLORECTAL POLYPS 1663 Table 8. Outcome of Endoscopically Removed Malignant Colorectal Polyps: Regarding Lymphatic Invasion, Venous Invasion, and Grade III Cancer Adverse outcome + (%) (%) No evidence All cases of disease Incidence (%) (%) (%) Lymphatic invasion (n = 115) 1'5'8 7 (100) 0 (0) 7 (38.8) 11 (61) Venous invasion (n = 136) *'12 (0) 1 (100) 1 (3.1) 31 (96.9) Venous and/or lymphatic invasion (n = 185) ~1'~3 13 (81) 3 (19) 16 (39) 25 (61) Grade Ill cancer (n = 488) 1'4'5'6'8'~ '~1'1s 16 (100) 0 (0) 16 (36) 29 (64) 18 (15.6) 32 (20.5) 41 (22.1) 45 (9.2) NOTE. Adverse outcome, recurrence following polypectomy only, residual local disease, and/or lymph node metastases m resection specimen postpolypectomy. --, Adverse outcome when other associated unfavorable histological parameters (positive margin, lymphatic invasion, venous invasion, or grade III cancer) are also present; -, adverse outcome when other associated unfavorable histological parameters are absent. All cases, incidence of adverse outcome in all cases (those with and without other unfavorable histological parameters), n = total number of polyps studied. invasion group. These data indicate that, for therapeutic decisions, those diagnosed as indefinite for lymphatic invasion should really be considered positive for lymphatic invasion. We found venous invasion in 5 (3.5%) cases. Two of the 5 had an adverse outcome; both had other unfavorable histological parameters. The literature is conflicting regarding venous invasion. Geraghty et al. 12 found venous invasion in 31 of 80 patients (38.7%); however, only 1 patient (3.2%) was found to be dead of disease. Muller et a1.13 found venous invasion in 5 of 34 malignant polyps (14.7%), and 3 also had concomitant lymphatic/vascular invasion. Of the 2 patients with only venous invasion, both had lymph node metastasis. Nivatvongs et al. 11 reported a 22% incidence of venous invasion/lymphatic invasion, and 31% of these had lymph node metastases. However, all of their cases with lymph node metastases were Haggitt level 4 (carcinoma invading into the submucosa of the bowel wall below the level of the stalk of the polyp but above the muscularis propria). In our study, the incidence of grade III cancer was 5.7% and the incidence of adverse outcome was 37.5% as compared with 9.2% and 36%, respectively, as reported in the literature. 1'4-<8'1 '11'I3 In our study, 100% grade III cancers were associated with other unfavorable histology. This was significant when compared with the indefinite grade III group (P < 0.01) and the negative for grade III group (P < 0.01), probably reflecting the inherent aggressive biological nature of grade III cancers. In the literature, all of grade III cancers (100%) with an adverse outcome were associated with the presence of other unfavorable histological features. 1'4-<8'1 '1I'13 Regarding polypoid carcinomas, 5 of 25 (23%) were associated with an adverse outcome; however, all 5 (100%) with an adverse outcome were associated with the presence of unfavorable histological parameters. From our data and that of others, 1'5-7'1 '11 it would seem that polypoid cancers may be no more aggressive than adenomas with invasive cancer and can be managed using the same histological criteria. Regarding colloid (excluding signet ceil cancer) cancer, our data and the literature 4'11'12 indicate that those cases with an adverse outcome were associated with the presence of unfavorable histological parameters. Malignant polyps with colloid cancer can probably be managed with the same criteria of other malignant polyps. However, we make this statement cautiously and believe more cases of this type need to be studied before definite conclusions can be drawn. There is an extremely high incidence of adverse outcome associated with signet cell cancer, as noted in our study and the literature. 12'13 Regarding polyp configuration, our findings are similar to that reported in the literature l'4,<s'i2j7 in that there is no difference in adverse outcome when comparing polyp configurations in and of themselves, but outcome is more importantly related to status of margin, histological grade, and the presence or absence of lymphatic invasion or venous invasion rather than polyp configuration. In our study, 4 of 90 (4.4%) with a negative resection later developed distant metastases. This situation has been previously reported in the literature. 1'3'6 Three of these 4 had unfavorable histological parameters present, and one was indefinite for both grade III and lymphatic invasion. Therefore, in a small percentage of negative resections (4.4%), there will be subsequent death regardless of what form of therapy is undertaken. In these instances, the cancer had probably metastasized distantly before initial polypectomy. In our study, there were 8 deaths (6.3%). Of the 8 who died, 2 were treated by polypectomy only and 6 with resections postpolypectomy. Of these latter 6, 4 had negative resections but developed distant metastases 2, 2, 4, and 7 years later. Seven of the 8 deaths had unfavorable histology present, whereas 1 of 8 was indefinite for both

8 1664 COOPER ET AL. GASTROENTEROLOGY Vo1.108, No. 6 grade III and lymphatic invasion. No patient died of disease who was considered to have absence of unfavorable histology. In this study, the interobserver correlation (strength of agreement) was substantial to almost perfect for margins, grade, and venous invasion but was only fair, moderate, or substantial for lymphatic invasion. This strength of agreement for diagnosing lymphatic invasion is similar to the strength of agreement between pathologists diagnosing dysplasia in ulcerative colitis. 23'24 In our study, pathologists A-E diagnosed lymphatic invasion in 27, 16, 10, 56, and 31, respectively. The criteria used for diagnosing lymphatic invasion were identical to that used by pathologists in their everyday practice. 2 We know of no other specific criteria that can increase this accuracy. Studies using immunohistochemical markers have not offered any diagnostic advantage) 3'25 The difficulty of accurately diagnosing lymphatic invasion has been commented on by others. 1 'I2'26 Both Williams and Geraghty 1 and Geraghty et al. 12 believe that diagnosing lymphatic invasion is very subjective and, because of this, they do not even bother. This variability could explain cases with reported favorable histological parameters and adverse outcome. In our study, 2 cases diagnosed as indefinite for lymphatic invasion with adverse outcome with absence of unfavorable histological parameters were probably positive for lymphatic invasion. In any clinicopathologic system, it is important to have a low rate of false-negative cases (e.g., those cases without unfavorable histological parameters but who had an adverse outcome). In the literature, 1'3'5-1 '13-15'17-~8 the overall incidence of false-negative cases is 1.7%. According to our study design, our false-negative rate was 2.9% (2 of 69 patients without the presence of unfavorable histology). These 2 revolved around the ability to diagnose lymphatic invasion present (both considered indefinite for lymphatic invasion). Possible ways of lowering this false-negative rate would be review of multiple levels of slides and/or review of the case with one or more pathologists, as has been suggested for diagnosing dysplasia in ulcerative colitis. 23'24 Our false-negative rate could be reduced to 0% by considering the indefinite for lymphatic invasion group as positive for the presence of lymphatic invasion. Our data supports this combination of groups. In summary, our study indicates that the parameters that we studied can be readily used for patient care and, in conjunction with the literature, support the management of patients with endoscopically removed malignant colorectal polyps as follows; patients with (1) cancer at or near (--< 1 mm) the resection margin and/or (2) grade III cancer and/or (3) lymphatic invasion (including in- definite for lymphatic invasion per our criteria) and/or (4) venous invasion are probably best treated by definitive surgical resection (if medically feasible) postpolypectomy. Patients with cancer >1 mm from the resection margin and grade I or II and no lymphatic or venous invasion can be treated safely with polypectomy only. However, the treatment of the patient with an endoscopically removed malignant colorectal polyp must be individualized for each patient, taking all factors into consideration. References 1. Cooper HS. Surgical pathology and endoscopically removed malignant polyps of the colon and rectum. Am J Surg Pathol 1983;7: Cooper HS. The role of the pathologist in the management of patients with endoscopically removed malignant colorectal polyp. Pathol Annu 1988; 23: Morson BC, Whiteway JE, Jone EA, Macrae FA, Williams CB. Histopathology and prognosis of malignant colorecta] polyps treated by endoscopic polypectomy. Gut 1984;25: Haggitt RC, Goltzbach RE, Softer EE, Wirbule LD. Prognostic factors in colorectal carcinomas arising in adenomas: implications of lesions removed by endoscopic polypectomy. Gastroenterology 1985; 89: Cranley JP, Petras RE, Carey WD, Paradis K, Sivak MB. When is endoscopic polypectomy adequate therapy for colonic polyps containing invasive carcinoma? Gastroenterology 1986;91: Kyzer S, Begin LR, Gordon PH, Mitmaker B. The care of patients with colorectal polyps that contain invasive adenocarcinoma. Endoscopic polypectomy or colectomy? Cancer 1992;70: Christie JP. Polypectomy or colectomy? Management of 106 consecutive encountered colorectal polyps. Am Surg 1988;54: Coverlizza S, Risio M, Ferrari A, Fenoglio-Preiser CM, Rossini FP. Colorectal adenomatous containing invasive carcinoma. Pathologic assessment of lymph node metastatic potential. Cancer 1989; 64: Richards WO, Webb WA, Morris S J, Davis RC, McDaniel L, Jones L, Littauer S. Patient management after endoscopic removal of the cancerous colon adenoma. Ann Surg 1987; 205: Williams CB, Geraghty JM. The malignant polyp--when to operate: the St. Mark's experience. Can J Gastroenterol 1990;4: Nivatvongs S, Rojanasakul A, Rayman HM, Dozlis RR, Wolff BG, Pembert JH, Baert W, Jacques LF. The risk of lymph node metastasis in colorectal polyps with invasive adenocarcinoma. Dis Colon Rectum 1991;34: Geraghty JM, Williams CB, Talbot IC. Malignant colorectal polyp: venous invasion and successful treatment by endoscopic polypectomy. Gut 1991;32: Muller S, Chesner IM, Igan MA, Rowlands DC, Collard M J, Swatbrick ET, Newman J. Significance of venous and lymphatic invasion in malignant polyps of the colon and rectum. Gut 1990; 30: Wolff WI, Shinya H. Definite treatment of malignant polyps of the colon. Ann Surg 1975; 185: Shatney Ch, Lober PH, Gilbertson V, Sosin H. Management of focally malignant pedunculated adenomatous colorectal polyps. Dis Colon Rectum 1976; 19: Lipper S, Kahn LB, Ackerman LV. The significance of microscopic invasive cancer in endoscopically removed polyp of the large

9 June 1995 MALIGNANT COLORECTAL POLYPS 1665 bowel. The clinicopathologic study of 51 cases. Cancer 1983; 52: Nivatvongs S, Goldberg SM. Management of patients who have polyps containing invasive carcinoma removed via colonoscope. Dis Col Rectum 1978; 21: Langer JC, Cohen Z, Taylor BR, Stafford S, JeeJeeBhoy KN, Cullen JB. Management of patients with polyp containing malignancy removed by colonoscopic polypectomy. Dis Colon Rectum 1984; 27: Dukes CE, Bussey HJM. The spread of rectal cancer and its effect in prognosis. Br J Cancer 1958; 12: Minsky BD, Mies C, Recht A, Rich TA, Chaffey JT. Resectable adenocarcinoma of the rectosigmoid and rectum. II. influence of blood vessel invasion. Cancer 1988; 61: Landis R J, Koch GC. The measurement of observer agreement for categorical data. Biometrics 1977;33: Kaplan El, Meier P. Nonparametric estimation for incomplete observation. J Am Stat Assoc 1958;33: Dixon MF, Brown L JR, Gilmour HM, Price AB, Smeeton NC, Talbot IC, Williams GT. Observer variation in the assessment of dysplasia in ulcerative colitis. Histopathology 1988;13: Melville DM, Jass JR, Morson BC, Pollock D J, Richman PI, Shepherd NA, Ritchie JK, Love SB, Lennard-Jones JE. Observer study of the grading of dysplasia in ulcerative colitis. Comparison with clinical outcome. Hum Pathol 1989;20: Bettelheim R, Mitchell D, Gusterson BA. Immunohistochemistry in the identification of vascular invasion in breast cancer. J Clin Pathol 1984; 37: , Gilchrist KW, Gould VE, Hirschel S, Imbriglia JE, Patchefsky, AS, Penner DW, Pickren J, Schwartz IS, Wheeler JE, Barnes JM, Mansour EG. Interobserver variation in the identification of breast carcinoma in intramammary lymphatics. Hum Pathol 1982; 13: Received June 10, Accepted February 22, Address requests for reprints to: Harry S. Cooper, M.D., Department of Pathology, Fox Chase Cancer Center, 7701 Burholme Avenue, Philadelphia, Pennsylvania Fax (215) The authors thank the many members of the Gastrointestinal Pathology Society who kindly provided cases and follow-up, the Thomas Jefferson University Tumor Registry and the Hahnemann University Tumor Registry for thorough help with patient follow-up, Arlene Barto- Iome and Connie Terzyk for excellent secretarial assistance, and Dr. S. N. S. Murthy, Dr. Marcia Polansky, and Narasim Murthy for their comments and extensive expert help with the statistical analysis.