ORIGINAL CONTRIBUTION
Mural and Extramural Venous Invasion and Prognosis in Colorectal Cancer Katherine M. Gibson, M.Sc., F.R.A.C.S.1 • Charles Chan, Ph.D., F.R.C.P.A.2 Pierre H. Chapuis, D.S., F.R.A.C.S.3 • Owen F. Dent, Ph.D.3 • Les Bokey, M.S., F.R.A.C.S.1 1 Department of Surgery, Liverpool Hospital and School of Medicine, University of Western Sydney, New South Wales, Australia 2 Department of Anatomical Pathology, Concord Hospital and Discipline of Pathology, University of Sydney, New South Wales, Australia 3 Department of Colorectal Surgery, Concord Hospital and Discipline of Surgery, University of Sydney, and School of Medicine, University of Western Sydney, New South Wales, Australia
BACKGROUND: Extramural venous invasion is a known independent predictor of poor prognosis after resection of colorectal adenocarcinoma, but the prognostic value of mural venous invasion alone and the association between venous invasion and prognosis within tumor stages has received little research attention.
or extramural invasion (both p < 0.001) than in those without invasion, and this persisted after adjustment for other prognostic variables. Equivalent bivariate associations were found in stage D, but only the effect of extramural invasion persisted after adjustment.
OBJECTIVE: This study aimed to determine whether
single surgical group and may not be generalizable to other settings. Only hematoxylin and eosin staining was used.
associations between mural and extramural venous invasion and outcome differ among tumor stages after adjustment for other factors known to influence prognosis. DESIGN: This study is a retrospective analysis of prospectively collected data. SETTINGS: Data were drawn from a registry of 3040 consecutive patients undergoing resection between 1980 and 2005 under the care of specialist surgeons in a tertiary referral public hospital and an affiliated private hospital. A standardized protocol was used for the pathological assessment of specimens. MAIN OUTCOME MEASURES: The primary outcomes
measured were overall survival, cancer-specific survival, and recurrence. RESULTS: There was no significant association between
venous invasion and survival in stages A (n = 544) or B (n = 1078). In stage C (n = 899), overall survival time was significantly shorter in patients with mural invasion alone Financial Disclosures: None reported. Correspondence: Owen Dent, Ph.D., Department of Colorectal Surgery, Concord Hospital, NSW 2139, Australia. E-mail: Owen.Dent@ netspeed.com.au Dis Colon Rectum 2014; 57: 916–926 DOI: 10.1097/DCR.0000000000000162 © The ASCRS 2014
916
LIMITATIONS: Our findings arise from the experience of a
CONCLUSIONS: The association between venous invasion and prognosis was stage specific. Both mural venous invasion alone and extramural venous invasion independently predicted overall survival in patients with stage C tumors, but not in patients with stages A, B, or D tumors. Although mural invasion alone was rare, the separate reporting of both mural and extramural invasion in patients with stage C tumor is informative and desirable. KEY WORDS: Colorectal cancer; Venous invasion; Survival; Recurrence; Prognosis.
T
he presence of tumor cells invading veins outside the bowel wall (extramural venous invasion, EMVI) in resected colorectal cancer (CRC) specimens is a known predictor of poor prognosis1–4 and is one of the core components in the Royal College of Pathologists data set for reporting CRC.5 However, the invasion of veins within the bowel wall (mural venous invasion, MVI) has received relatively little attention, and the prognostic significance of MVI alone, independent of EMVI and other variables, particularly tumor stage, has not been fully described.1–3,6–8 We retrospectively analyzed a large prospectively recorded series of CRC resections to investigate MVI and EMVI in relation to long-term patient outcomes of overDiseases of the Colon & Rectum Volume 57: 8 (2014)
Diseases of the Colon & Rectum Volume 57: 8 (2014)
all survival, CRC-specific survival, and tumor recurrence. Our hypothesis was that the associations between mural and extramural venous invasion and outcome would differ among tumor stages after adjustment for other factors known to influence prognosis.
PATIENTS AND METHODS Data were drawn from a prospective registry of consecutive CRC resections that was initiated in 1971 at Concord Hospital, a public tertiary referral hospital in Sydney, and contains detailed clinical, operative, pathology, adjuvant therapy, and follow-up information.9,10 It has the approval of the Sydney South West Health Area Ethics Committee. All resections were performed by specialist colorectal surgeons using a standardized technique.11,12 Resections between 1980 and 2005 inclusive were selected for analysis, and all nondeceased patients were followed for a minimum of 5 years. Patients were excluded if they had previous CRC, a noninvasive tumor, IBD, familial adenomatous polyposis coli, distant metastases resected within 3 months of resection of the primary, or received preoperative radiotherapy. Pathological examination of the resected specimen followed a standard protocol.9,13 Only adenocarcinomas were included in the registry, and, in analyses, histological type was categorized as mucinous or signet-ring versus nonmucinous/signet-ring. Where multiple tumors were present, only the most advanced-stage lesion was included. Tumor size was measured as the greatest surface dimension and blocks were taken to demonstrate maximum direct tumor penetration of the bowel wall. Additional blocks were taken to demonstrate the relationship between tumor and any adherent structure or tissue,14 as well as lines of resection and the free serosal surface.15 Tumor grade was assessed, taking into account the degree of differentiation and anaplasia, the nature of the tumor margin (pushing or infiltrating), and the presence of small vessel invasion, and it was categorized as high grade versus other. Venous invasion, assessed by hematoxylin and eosin staining, was recorded as involvement of thick- or thin-walled veins, either within or beyond the bowel wall. A vascular channel was defined as an endothelium-lined space surrounded by a smooth muscle coat and/or containing red blood cells within the lumen. When doubt existed as to whether a structure involved was a vein, a negative finding was recorded. An apical lymph node was defined as the most proximal node found within 1 cm of the vessel ligation at the apex of a vascular pedicle.16 For advanced stage tumors, the proportion of involved lymph nodes was calculated as a percentage of the total number of nodes harvested. Before 2002, over 90% of specimens were reported on or reviewed by a single pathologist (R.C. Newland, Concord Hospital).
917
All pathology features were looked for in every specimen and their presence or absence recorded explicitly. There were no missing data on any variable. In analyses, all patient and tumor covariates that were not natural binary variables were dichotomized at conventional or appropriate cutting points to simplify comparisons of effect sizes among covariates in multivariable models. Tumors were staged according to the Australian Clinico-Pathological Staging system for CRC.13 The 4 main stages of this system (A, B, C, D) are directly equivalent to the main stages (I, II, III, IV) of the American Joint Committee on Cancer/International Union against Cancer pTNM system,17 but, importantly, Australian Clinico-Pathological Staging differs in that all lesions with macro- or microscopic tumor in any resection margin are coded as stage D and included in analyses as such. These patients, similar to patients with pTNM stage D (who have metastatic residual disease), experience markedly diminished survival.18 Substages were not used in analyses. Postoperative chemotherapy, either adjuvant or palliative and 5-fluorouracil-based for the most part, was introduced in 1992 and used on 3% of all patients with stage B tumors, 25% of all stage C, and 20% of all stage D. Postoperative radiotherapy was introduced in 1990 and applied to only 46 patients overall (1.5%); therefore, it was not included in analyses. Follow-up and Assessment of Survival and Recurrence
Patients were seen at least every 6 months for the first 2 years after resection and yearly thereafter until death or December 31, 2010. Surveillance included clinical examination, sigmoidoscopy, chest x-ray, and serial CEA measurements. For rectal cancer, a CT scan was performed annually as well as either sigmoidoscopy or colonoscopy. For colon cancer, colonoscopy was generally repeated between 3 and 5 years after resection. Recurrence was defined as clinically or radiologically suspected or biopsy-proven tumor in the pelvis, perineal scar, or peritoneal cavity (local recurrence), or newly diagnosed distant metastasis (systemic recurrence). The occurrence, date, and cause of death were ascertained principally from the patient’s surgeon or family physician or hospital records; from a close relative if necessary; or, in a small number of cases, from the national death registration system. Overall survival time was measured from the date of resection to the date of death due to any cause. Times were censored at last contact for patients who were lost to follow-up, who remained alive in December 2010, or who had been followed for at least 14 years. For competingrisk regression analysis of death due to CRC, the survival times of patients who died of causes other than CRC were measured until the date of death, and these patients were coded as having experienced a competing risk. Recurrence was not recorded for colon cancer before 1995 and therefore was analyzed only in patients with rectal can-
918
cer. Time to recurrence was measured until the date of diagnosis of recurrence and was censored at last contact for patients who were lost to follow-up or who remained alive and recurrence-free as above. Patients who died without recurrence were classified as having experienced a competing risk. Statistical Analysis
All analyses were conducted based on intention to treat. The χ2 test was used to assess the significance of differences in proportions and the Cochrane-Armitage test to assess trend. The Kaplan-Meier method and log-rank test were used to assess differences in overall survival between strata of venous invasion and covariates. Proportional hazards regression or competing-risk regression and the Wald test were used to assess the effects of venous invasion and covariates on survival time or time to recurrence. In regression analyses, venous invasion was coded as 2 dummy variables: MVI versus all other and EMVI versus all other. In multivariable modeling, all covariates having an association with survival with a Wald test p < 0.1 were entered into an initial regression model. Sequential removal of covariates with p > 0.05, beginning with the highest p value, resulted in a provisional final model containing only covariates with p ≤ 0.05. Excluded variables were then reintroduced singly, but none achieved significance. Postoperative chemotherapy was included among the covariates but not as a time-dependent variable. This was necessary because the starting time of chemotherapy had not been recorded systematically. However, it was reasoned that, because chemotherapy was commenced relatively soon after surgery in most cases, and because it was used relatively infrequently (even in stage C tumors), the effect on resulting models would be negligible. The year of resection was included in models to adjust for any systematic temporal change in covariates associated with survival. The assumption of proportional hazards was assessed graphically and was not materially violated in any variable included in a regression model. Possible interactions between venous invasion and other covariates were examined by introducing product terms singly into regression models, but no statistically significant interactions were found. The level for 2-tailed statistical significance was p ≤ 0.05 with CIs at the 95% level. Analyses were performed with SPSS version 21 (IBM Australia Limited, 2013) and Stata release 12 (Stata Corporation, College Station, TX, 2011).
RESULTS Of 3343 potentially eligible patients, 111 had previous CRC, 44 had carcinoma in situ, 34 had IBD or polyposis coli, 27 had distant metastases resected, and 87 received preoperative radiotherapy or chemoradiotherapy. These
GIBSON ET AL: VENOUS INVASION IN COLORECTAL CANCER
TABLE 1. Clinical and pathological characteristics of 3040 patients Characteristic Male sex Age ≥ 75 y Colonic tumor Tumor maximum surface dimension ≥ 5 cm Mucinous or signet-ring tumor Direct tumor spread beyond muscularis propria Number of lymph nodes identified (median) Nodal metastasis Apical node involved ≥ 4 nodes involved High grade Free serosal surface involved Adjacent organ or structure infiltrated Tumor stage A B C D Received postoperative chemotherapy
n (%) 1949 (64.1) 964 (31.7) 1771 (58.3) 1425 (46.9) 286 (9.4) 2364 (77.8) 12 1313 (43.2) 162 (5.3) 496 (10.1) 759 (25.0) 404 (13.3) 127 (4.2) 544 (17.9) 1078 (35.5) 899 (29.6) 519 (17.1) 361 (11.9)
patients were excluded, leaving 3040 whose clinical and pathological characteristics are shown in Table 1. The same 9 surgeons performed 93% of these operations. Postoperative chemotherapy was introduced gradually from 1992 and applied selectively, not routinely. The overall percentages of patients who received postoperative chemotherapy were 2.5% in stage B, 25.1% in stage C, and 20.4% in stage D. Postoperative radiotherapy was used in 1.5% of patients overall. Venous invasion was not found in 2352 specimens (77.4%); MVI alone was found in 101 (3.3%), EMVI alone in 459 (15.1%), and both forms in 128 (4.2%; Table 2). The frequency of EMVI increased markedly with increasing stage, but there was no significant equivalent trend for MVI alone (p for trend 0.066). Preliminary analyses showed no significant difference in survival between patients with EMVI alone and those with both MVI and EMVI, so these 2 patient groups were combined under the term EMVI. Follow-up and Survival Time
Overall survival time ranged from 64 to 349 months for 862 patients (16.7%) who had not died, with a median of 153 months. Survival time in 2030 (66.8%) deceased patients ranged from 0.4 to 320 months with a median of 40 months. Forty-seven patients (1.5%) were lost to followup with a survival time ranging from 3 to 178 months (median, 80 months), and 101(3.3%) died before discharge from the hospital after resection. Venous Invasion and Overall Survival by Tumor Stage
In 544 patients with stage A tumor, 16 (2.9%) had MVI alone and 3 had EMVI (0.6%); these numbers were insufficient for reliable survival analysis.
919
Diseases of the Colon & Rectum Volume 57: 8 (2014)
TABLE 2. Association between tumor stage and venous invasion Venous invasion
Stage A n = 544
Stage B n = 1078
Stage C n = 899
Stage D n = 519
All stages n = 3040
525 (96.5) 16 (2.9) 2 (0.4) 1 (0.2)
936 (86.8) 24 (2.2) 82 (7.6) 36 (3.3)
646 (71.9) 42 (4.7) 178 (19.8) 33 (3.7)
245 (47.2) 19 (3.7) 197 (38.0) 58 (11.2)
2352 (77.4) 101 (3.3) 459 (15.1) 128 (4.2)
Absent Mural alone Extramural Both
Values are stated as n (%). χ2 p < 0.001.
In 1078 patients with stage B tumor, 24 (2.2%) had MVI alone and 118 had EMVI (11.0%). Despite an indication of poorer survival in those with EMVI (Fig. 1), there were no pairwise differences in overall survival (none vs MVI alone, p = 0.90; none vs EMVI, p = 0.172; MVI alone vs EMVI, p = 0.39). The survival curves converged completely after 120 months (not shown), and the appearance of early poorer survival among the 118 patients with EMVI may be the result of confounding by another variable rather than indicative of a type II error. These results were effectively uninfluenced by adjuvant chemotherapy because only 2.5% of stage B patients were so treated. The possibility of a difference in cancer-specific survival was tested by competing risks bivariate regression, which showed no association for either MVI alone (Wald p = 0.74) or EMVI (Wald p = 0.62). In 899 patients with stage C tumor, 42 (4.7%) had MVI alone and 211 had EMVI (23.5%). There was no significant difference in survival between those with MVI
alone and EMVI (p = 0.69), but survival was significantly shorter in these groups than in patients without venous invasion (none vs MVI alone, p < 0.001; none vs EMVI, p < 0.001) (Fig. 2). Twenty-five percent of patients with stage C tumor received chemotherapy, which might have reduced an otherwise greater survival difference. In 519 patients with stage D tumor, 19 (3.7%) had MVI alone and 255 had EMVI (49.1%). The same survival pattern applied in stage D patients as in stage C patients (MVI alone vs EMVI, p = 0.93; none vs MVI alone, p = 0. 01; none vs EMVI, p < 0.001) (Fig. 3). One-fifth of patients with stage D tumor received adjuvant chemotherapy, which might have reduced an otherwise greater survival difference. In summary, a bivariate association between venous invasion and diminished overall survival was present in stage C and D tumors, but not in stage A and B tumors. Proportion surviving 1.0
Proportion surviving 1.0
0.9
0.9
0.8
0.8
0.7
0.7
0.6
None Mural
0.6
0.5
None
0.4
0.5 Extramural
0.4
Mural
0.3
0.3
0.2
0.2
0.1
Extramural
0.1
12 12
24
36
48
60 72 Months
84
96 108 120
None
936 873 822 766 715 662 614 553 483 416 361 21
20
19
18
17
14
13
12
11
10
Extramural 118 105
95
83
78
68
64
59
55
52
49
Mural
24
FIGURE 1. Overall survival of patients who had stage B colorectal cancer with mural, extramural, and no venous invasion. There were no pairwise differences in survival (none vs MVI alone, p = 0.90; none vs EMVI, p = 0.172; MVI alone vs EMVI, p = 0.39). EMVI = extramural venous invasion; MVI = mural venous invasion.
36
48
60 72 Months
84
96 108 120
Number at risk None
Number at risk
24
646 583 514 447 402 357 327 300 268 240 220 30
26
19
16
14
12
7
7
5
Extramural 211 174 132
98
78
64
53
47
42
40
35
Mural
42
36
FIGURE 2. Overall survival of patients who had stage C colorectal cancer with mural, extramural, and no venous invasion. There was no significant difference in survival between those with MVI alone and EMVI (p = 0.69), but survival was significantly shorter in these groups than in patients without venous invasion (none vs MVI alone, p < 0.001; none vs EMVI, p < 0.001). EMVI = extramural venous invasion; MVI = mural venous invasion.
920
GIBSON ET AL: VENOUS INVASION IN COLORECTAL CANCER
Proportion surviving 1.0
Venous Invasion, Other Features, and Overall Survival in Stage C
0.9
Associations between clinicopathological variables and venous invasion in 899 patients with stage C tumor are shown in Table 3. Whereas differences in the proportions of patients with MVI alone were very small, EMVI was significantly more common in those with rectal cancer, direct spread beyond the muscularis propria, ≥4 involved lymph nodes, an involved apical node, ≥30% of nodes involved, high-grade tumor, and involvement of a free serosal surface. There were no associations with sex, age, tumor size, histological type, infiltration of an adjacent structure, or receiving postoperative chemotherapy. Within the colon there was no significant difference between the right and left colon (χ2 p = 0.15). A multivariable regression model showed significant independent associations with survival for both MVI alone (HR, 1.5; 95% CI, 1.1-2.2; p = 0.016) and EMVI (HR, 1.5; 95% CI, 1.2-1.8; p < 0.001) after adjustment for other clinicopathological features (Table 4). Importantly, there was no association between tumor site and survival, indicating that the effect of venous invasion applied independently of site, and thus analyses of the effects of venous invasion separately for the colon and rectum were not necessary. The “year of resection” term adjusts for any covariate that changed systematically over the span of the study and was significantly related to overall survival (the more recent the resection, the longer the
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Mural 12 24
None
Extramural 36
48
60 72 Months
84
96 108 120
Number at risk 245 152
81
51
32
24
20
15
10
10
10
9
4
0
0
0
0
0
0
0
0
Extramural 255 107
40
11
7
5
3
2
1
1
1
None Mural
19
FIGURE 3. Overall survival of patients who had stage D colorectal cancer with mural, extramural, and no venous invasion. There was no significant difference in survival between those with MVI alone and EMVI (p = 0.93), but survival was significantly shorter in these groups than in patients without venous invasion (none vs MVI alone, p = 0.01; none vs EMVI, p < 0.001). EMVI = extramural venous invasion; MVI = mural venous invasion.
TABLE 3. Association between clinicopathological variables and venous invasion in 899 patients with stage C tumor
Male Female Age ≥ 75 y No Rectum Colon Maximum surface dimension ≥ 5 cm No Mucinous or signet-ring No Spread beyond muscularis propria No ≥ 4 nodes involved No Apical node involved No ≥ 30% of nodes involved No High grade No Free serosal surface involved No Adjacent structure infiltrated No Postoperative chemotherapy No Values are stated as n (%).
n (100%)
None
Mural alone
Extramural
χ2 p
566 333 268 631 404 495 412 487 99 800 775 124 278 621 75 824 299 600 321 578 171 728 38 861 226 673
410 (72) 236 (71) 200 (75) 446 (71) 267 (66) 379 (77) 294 (71) 352 (72) 80 (80) 566 (71) 534 (69) 112 (90) 168 (60) 478 (77) 44 (59) 602 (73) 173 (58) 473 (79) 189 (59) 457 (79) 105 (61) 541 (74) 24 (63) 622 (72) 174 (77) 472 (70)
29 (5) 13 (4) 9 (3) 33 (5) 20 (5) 22 (4) 15 (4) 27 (6) 2 (2) 40 (5) 36 (5) 6 (5) 14 (5) 28 (5) 4 (5) 38 (5) 19 (6) 23 (4) 17 (5) 25 (4) 10 (6) 32 (4) 0 42 (5) 9 (4) 33 (5)
127 (22) 84 (25) 59 (22) 152 (24) 117 (29) 94 (19) 103 (25) 108 (22) 17 (17) 194 (24) 205 (27) 6 (5) 96 (35) 115 (19) 27 (36) 184 (22) 107 (36) 104 (17) 115 (36) 96 (17) 56 (33) 115 (21) 14 (37) 197 (23) 43 (19) 168 (25)
0.48 0.33 0.002 0.28 0.09
Mural and Extramural Venous Invasion and Prognosis in Colorectal Cancer Katherine M. Gibson, M.Sc., F.R.A.C.S.1 • Charles Chan, Ph.D., F.R.C.P.A.2 Pierre H. Chapuis, D.S., F.R.A.C.S.3 • Owen F. Dent, Ph.D.3 • Les Bokey, M.S., F.R.A.C.S.1 1 Department of Surgery, Liverpool Hospital and School of Medicine, University of Western Sydney, New South Wales, Australia 2 Department of Anatomical Pathology, Concord Hospital and Discipline of Pathology, University of Sydney, New South Wales, Australia 3 Department of Colorectal Surgery, Concord Hospital and Discipline of Surgery, University of Sydney, and School of Medicine, University of Western Sydney, New South Wales, Australia
BACKGROUND: Extramural venous invasion is a known independent predictor of poor prognosis after resection of colorectal adenocarcinoma, but the prognostic value of mural venous invasion alone and the association between venous invasion and prognosis within tumor stages has received little research attention.
or extramural invasion (both p < 0.001) than in those without invasion, and this persisted after adjustment for other prognostic variables. Equivalent bivariate associations were found in stage D, but only the effect of extramural invasion persisted after adjustment.
OBJECTIVE: This study aimed to determine whether
single surgical group and may not be generalizable to other settings. Only hematoxylin and eosin staining was used.
associations between mural and extramural venous invasion and outcome differ among tumor stages after adjustment for other factors known to influence prognosis. DESIGN: This study is a retrospective analysis of prospectively collected data. SETTINGS: Data were drawn from a registry of 3040 consecutive patients undergoing resection between 1980 and 2005 under the care of specialist surgeons in a tertiary referral public hospital and an affiliated private hospital. A standardized protocol was used for the pathological assessment of specimens. MAIN OUTCOME MEASURES: The primary outcomes
measured were overall survival, cancer-specific survival, and recurrence. RESULTS: There was no significant association between
venous invasion and survival in stages A (n = 544) or B (n = 1078). In stage C (n = 899), overall survival time was significantly shorter in patients with mural invasion alone Financial Disclosures: None reported. Correspondence: Owen Dent, Ph.D., Department of Colorectal Surgery, Concord Hospital, NSW 2139, Australia. E-mail: Owen.Dent@ netspeed.com.au Dis Colon Rectum 2014; 57: 916–926 DOI: 10.1097/DCR.0000000000000162 © The ASCRS 2014
916
LIMITATIONS: Our findings arise from the experience of a
CONCLUSIONS: The association between venous invasion and prognosis was stage specific. Both mural venous invasion alone and extramural venous invasion independently predicted overall survival in patients with stage C tumors, but not in patients with stages A, B, or D tumors. Although mural invasion alone was rare, the separate reporting of both mural and extramural invasion in patients with stage C tumor is informative and desirable. KEY WORDS: Colorectal cancer; Venous invasion; Survival; Recurrence; Prognosis.
T
he presence of tumor cells invading veins outside the bowel wall (extramural venous invasion, EMVI) in resected colorectal cancer (CRC) specimens is a known predictor of poor prognosis1–4 and is one of the core components in the Royal College of Pathologists data set for reporting CRC.5 However, the invasion of veins within the bowel wall (mural venous invasion, MVI) has received relatively little attention, and the prognostic significance of MVI alone, independent of EMVI and other variables, particularly tumor stage, has not been fully described.1–3,6–8 We retrospectively analyzed a large prospectively recorded series of CRC resections to investigate MVI and EMVI in relation to long-term patient outcomes of overDiseases of the Colon & Rectum Volume 57: 8 (2014)
Diseases of the Colon & Rectum Volume 57: 8 (2014)
all survival, CRC-specific survival, and tumor recurrence. Our hypothesis was that the associations between mural and extramural venous invasion and outcome would differ among tumor stages after adjustment for other factors known to influence prognosis.
PATIENTS AND METHODS Data were drawn from a prospective registry of consecutive CRC resections that was initiated in 1971 at Concord Hospital, a public tertiary referral hospital in Sydney, and contains detailed clinical, operative, pathology, adjuvant therapy, and follow-up information.9,10 It has the approval of the Sydney South West Health Area Ethics Committee. All resections were performed by specialist colorectal surgeons using a standardized technique.11,12 Resections between 1980 and 2005 inclusive were selected for analysis, and all nondeceased patients were followed for a minimum of 5 years. Patients were excluded if they had previous CRC, a noninvasive tumor, IBD, familial adenomatous polyposis coli, distant metastases resected within 3 months of resection of the primary, or received preoperative radiotherapy. Pathological examination of the resected specimen followed a standard protocol.9,13 Only adenocarcinomas were included in the registry, and, in analyses, histological type was categorized as mucinous or signet-ring versus nonmucinous/signet-ring. Where multiple tumors were present, only the most advanced-stage lesion was included. Tumor size was measured as the greatest surface dimension and blocks were taken to demonstrate maximum direct tumor penetration of the bowel wall. Additional blocks were taken to demonstrate the relationship between tumor and any adherent structure or tissue,14 as well as lines of resection and the free serosal surface.15 Tumor grade was assessed, taking into account the degree of differentiation and anaplasia, the nature of the tumor margin (pushing or infiltrating), and the presence of small vessel invasion, and it was categorized as high grade versus other. Venous invasion, assessed by hematoxylin and eosin staining, was recorded as involvement of thick- or thin-walled veins, either within or beyond the bowel wall. A vascular channel was defined as an endothelium-lined space surrounded by a smooth muscle coat and/or containing red blood cells within the lumen. When doubt existed as to whether a structure involved was a vein, a negative finding was recorded. An apical lymph node was defined as the most proximal node found within 1 cm of the vessel ligation at the apex of a vascular pedicle.16 For advanced stage tumors, the proportion of involved lymph nodes was calculated as a percentage of the total number of nodes harvested. Before 2002, over 90% of specimens were reported on or reviewed by a single pathologist (R.C. Newland, Concord Hospital).
917
All pathology features were looked for in every specimen and their presence or absence recorded explicitly. There were no missing data on any variable. In analyses, all patient and tumor covariates that were not natural binary variables were dichotomized at conventional or appropriate cutting points to simplify comparisons of effect sizes among covariates in multivariable models. Tumors were staged according to the Australian Clinico-Pathological Staging system for CRC.13 The 4 main stages of this system (A, B, C, D) are directly equivalent to the main stages (I, II, III, IV) of the American Joint Committee on Cancer/International Union against Cancer pTNM system,17 but, importantly, Australian Clinico-Pathological Staging differs in that all lesions with macro- or microscopic tumor in any resection margin are coded as stage D and included in analyses as such. These patients, similar to patients with pTNM stage D (who have metastatic residual disease), experience markedly diminished survival.18 Substages were not used in analyses. Postoperative chemotherapy, either adjuvant or palliative and 5-fluorouracil-based for the most part, was introduced in 1992 and used on 3% of all patients with stage B tumors, 25% of all stage C, and 20% of all stage D. Postoperative radiotherapy was introduced in 1990 and applied to only 46 patients overall (1.5%); therefore, it was not included in analyses. Follow-up and Assessment of Survival and Recurrence
Patients were seen at least every 6 months for the first 2 years after resection and yearly thereafter until death or December 31, 2010. Surveillance included clinical examination, sigmoidoscopy, chest x-ray, and serial CEA measurements. For rectal cancer, a CT scan was performed annually as well as either sigmoidoscopy or colonoscopy. For colon cancer, colonoscopy was generally repeated between 3 and 5 years after resection. Recurrence was defined as clinically or radiologically suspected or biopsy-proven tumor in the pelvis, perineal scar, or peritoneal cavity (local recurrence), or newly diagnosed distant metastasis (systemic recurrence). The occurrence, date, and cause of death were ascertained principally from the patient’s surgeon or family physician or hospital records; from a close relative if necessary; or, in a small number of cases, from the national death registration system. Overall survival time was measured from the date of resection to the date of death due to any cause. Times were censored at last contact for patients who were lost to follow-up, who remained alive in December 2010, or who had been followed for at least 14 years. For competingrisk regression analysis of death due to CRC, the survival times of patients who died of causes other than CRC were measured until the date of death, and these patients were coded as having experienced a competing risk. Recurrence was not recorded for colon cancer before 1995 and therefore was analyzed only in patients with rectal can-
918
cer. Time to recurrence was measured until the date of diagnosis of recurrence and was censored at last contact for patients who were lost to follow-up or who remained alive and recurrence-free as above. Patients who died without recurrence were classified as having experienced a competing risk. Statistical Analysis
All analyses were conducted based on intention to treat. The χ2 test was used to assess the significance of differences in proportions and the Cochrane-Armitage test to assess trend. The Kaplan-Meier method and log-rank test were used to assess differences in overall survival between strata of venous invasion and covariates. Proportional hazards regression or competing-risk regression and the Wald test were used to assess the effects of venous invasion and covariates on survival time or time to recurrence. In regression analyses, venous invasion was coded as 2 dummy variables: MVI versus all other and EMVI versus all other. In multivariable modeling, all covariates having an association with survival with a Wald test p < 0.1 were entered into an initial regression model. Sequential removal of covariates with p > 0.05, beginning with the highest p value, resulted in a provisional final model containing only covariates with p ≤ 0.05. Excluded variables were then reintroduced singly, but none achieved significance. Postoperative chemotherapy was included among the covariates but not as a time-dependent variable. This was necessary because the starting time of chemotherapy had not been recorded systematically. However, it was reasoned that, because chemotherapy was commenced relatively soon after surgery in most cases, and because it was used relatively infrequently (even in stage C tumors), the effect on resulting models would be negligible. The year of resection was included in models to adjust for any systematic temporal change in covariates associated with survival. The assumption of proportional hazards was assessed graphically and was not materially violated in any variable included in a regression model. Possible interactions between venous invasion and other covariates were examined by introducing product terms singly into regression models, but no statistically significant interactions were found. The level for 2-tailed statistical significance was p ≤ 0.05 with CIs at the 95% level. Analyses were performed with SPSS version 21 (IBM Australia Limited, 2013) and Stata release 12 (Stata Corporation, College Station, TX, 2011).
RESULTS Of 3343 potentially eligible patients, 111 had previous CRC, 44 had carcinoma in situ, 34 had IBD or polyposis coli, 27 had distant metastases resected, and 87 received preoperative radiotherapy or chemoradiotherapy. These
GIBSON ET AL: VENOUS INVASION IN COLORECTAL CANCER
TABLE 1. Clinical and pathological characteristics of 3040 patients Characteristic Male sex Age ≥ 75 y Colonic tumor Tumor maximum surface dimension ≥ 5 cm Mucinous or signet-ring tumor Direct tumor spread beyond muscularis propria Number of lymph nodes identified (median) Nodal metastasis Apical node involved ≥ 4 nodes involved High grade Free serosal surface involved Adjacent organ or structure infiltrated Tumor stage A B C D Received postoperative chemotherapy
n (%) 1949 (64.1) 964 (31.7) 1771 (58.3) 1425 (46.9) 286 (9.4) 2364 (77.8) 12 1313 (43.2) 162 (5.3) 496 (10.1) 759 (25.0) 404 (13.3) 127 (4.2) 544 (17.9) 1078 (35.5) 899 (29.6) 519 (17.1) 361 (11.9)
patients were excluded, leaving 3040 whose clinical and pathological characteristics are shown in Table 1. The same 9 surgeons performed 93% of these operations. Postoperative chemotherapy was introduced gradually from 1992 and applied selectively, not routinely. The overall percentages of patients who received postoperative chemotherapy were 2.5% in stage B, 25.1% in stage C, and 20.4% in stage D. Postoperative radiotherapy was used in 1.5% of patients overall. Venous invasion was not found in 2352 specimens (77.4%); MVI alone was found in 101 (3.3%), EMVI alone in 459 (15.1%), and both forms in 128 (4.2%; Table 2). The frequency of EMVI increased markedly with increasing stage, but there was no significant equivalent trend for MVI alone (p for trend 0.066). Preliminary analyses showed no significant difference in survival between patients with EMVI alone and those with both MVI and EMVI, so these 2 patient groups were combined under the term EMVI. Follow-up and Survival Time
Overall survival time ranged from 64 to 349 months for 862 patients (16.7%) who had not died, with a median of 153 months. Survival time in 2030 (66.8%) deceased patients ranged from 0.4 to 320 months with a median of 40 months. Forty-seven patients (1.5%) were lost to followup with a survival time ranging from 3 to 178 months (median, 80 months), and 101(3.3%) died before discharge from the hospital after resection. Venous Invasion and Overall Survival by Tumor Stage
In 544 patients with stage A tumor, 16 (2.9%) had MVI alone and 3 had EMVI (0.6%); these numbers were insufficient for reliable survival analysis.
919
Diseases of the Colon & Rectum Volume 57: 8 (2014)
TABLE 2. Association between tumor stage and venous invasion Venous invasion
Stage A n = 544
Stage B n = 1078
Stage C n = 899
Stage D n = 519
All stages n = 3040
525 (96.5) 16 (2.9) 2 (0.4) 1 (0.2)
936 (86.8) 24 (2.2) 82 (7.6) 36 (3.3)
646 (71.9) 42 (4.7) 178 (19.8) 33 (3.7)
245 (47.2) 19 (3.7) 197 (38.0) 58 (11.2)
2352 (77.4) 101 (3.3) 459 (15.1) 128 (4.2)
Absent Mural alone Extramural Both
Values are stated as n (%). χ2 p < 0.001.
In 1078 patients with stage B tumor, 24 (2.2%) had MVI alone and 118 had EMVI (11.0%). Despite an indication of poorer survival in those with EMVI (Fig. 1), there were no pairwise differences in overall survival (none vs MVI alone, p = 0.90; none vs EMVI, p = 0.172; MVI alone vs EMVI, p = 0.39). The survival curves converged completely after 120 months (not shown), and the appearance of early poorer survival among the 118 patients with EMVI may be the result of confounding by another variable rather than indicative of a type II error. These results were effectively uninfluenced by adjuvant chemotherapy because only 2.5% of stage B patients were so treated. The possibility of a difference in cancer-specific survival was tested by competing risks bivariate regression, which showed no association for either MVI alone (Wald p = 0.74) or EMVI (Wald p = 0.62). In 899 patients with stage C tumor, 42 (4.7%) had MVI alone and 211 had EMVI (23.5%). There was no significant difference in survival between those with MVI
alone and EMVI (p = 0.69), but survival was significantly shorter in these groups than in patients without venous invasion (none vs MVI alone, p < 0.001; none vs EMVI, p < 0.001) (Fig. 2). Twenty-five percent of patients with stage C tumor received chemotherapy, which might have reduced an otherwise greater survival difference. In 519 patients with stage D tumor, 19 (3.7%) had MVI alone and 255 had EMVI (49.1%). The same survival pattern applied in stage D patients as in stage C patients (MVI alone vs EMVI, p = 0.93; none vs MVI alone, p = 0. 01; none vs EMVI, p < 0.001) (Fig. 3). One-fifth of patients with stage D tumor received adjuvant chemotherapy, which might have reduced an otherwise greater survival difference. In summary, a bivariate association between venous invasion and diminished overall survival was present in stage C and D tumors, but not in stage A and B tumors. Proportion surviving 1.0
Proportion surviving 1.0
0.9
0.9
0.8
0.8
0.7
0.7
0.6
None Mural
0.6
0.5
None
0.4
0.5 Extramural
0.4
Mural
0.3
0.3
0.2
0.2
0.1
Extramural
0.1
12 12
24
36
48
60 72 Months
84
96 108 120
None
936 873 822 766 715 662 614 553 483 416 361 21
20
19
18
17
14
13
12
11
10
Extramural 118 105
95
83
78
68
64
59
55
52
49
Mural
24
FIGURE 1. Overall survival of patients who had stage B colorectal cancer with mural, extramural, and no venous invasion. There were no pairwise differences in survival (none vs MVI alone, p = 0.90; none vs EMVI, p = 0.172; MVI alone vs EMVI, p = 0.39). EMVI = extramural venous invasion; MVI = mural venous invasion.
36
48
60 72 Months
84
96 108 120
Number at risk None
Number at risk
24
646 583 514 447 402 357 327 300 268 240 220 30
26
19
16
14
12
7
7
5
Extramural 211 174 132
98
78
64
53
47
42
40
35
Mural
42
36
FIGURE 2. Overall survival of patients who had stage C colorectal cancer with mural, extramural, and no venous invasion. There was no significant difference in survival between those with MVI alone and EMVI (p = 0.69), but survival was significantly shorter in these groups than in patients without venous invasion (none vs MVI alone, p < 0.001; none vs EMVI, p < 0.001). EMVI = extramural venous invasion; MVI = mural venous invasion.
920
GIBSON ET AL: VENOUS INVASION IN COLORECTAL CANCER
Proportion surviving 1.0
Venous Invasion, Other Features, and Overall Survival in Stage C
0.9
Associations between clinicopathological variables and venous invasion in 899 patients with stage C tumor are shown in Table 3. Whereas differences in the proportions of patients with MVI alone were very small, EMVI was significantly more common in those with rectal cancer, direct spread beyond the muscularis propria, ≥4 involved lymph nodes, an involved apical node, ≥30% of nodes involved, high-grade tumor, and involvement of a free serosal surface. There were no associations with sex, age, tumor size, histological type, infiltration of an adjacent structure, or receiving postoperative chemotherapy. Within the colon there was no significant difference between the right and left colon (χ2 p = 0.15). A multivariable regression model showed significant independent associations with survival for both MVI alone (HR, 1.5; 95% CI, 1.1-2.2; p = 0.016) and EMVI (HR, 1.5; 95% CI, 1.2-1.8; p < 0.001) after adjustment for other clinicopathological features (Table 4). Importantly, there was no association between tumor site and survival, indicating that the effect of venous invasion applied independently of site, and thus analyses of the effects of venous invasion separately for the colon and rectum were not necessary. The “year of resection” term adjusts for any covariate that changed systematically over the span of the study and was significantly related to overall survival (the more recent the resection, the longer the
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Mural 12 24
None
Extramural 36
48
60 72 Months
84
96 108 120
Number at risk 245 152
81
51
32
24
20
15
10
10
10
9
4
0
0
0
0
0
0
0
0
Extramural 255 107
40
11
7
5
3
2
1
1
1
None Mural
19
FIGURE 3. Overall survival of patients who had stage D colorectal cancer with mural, extramural, and no venous invasion. There was no significant difference in survival between those with MVI alone and EMVI (p = 0.93), but survival was significantly shorter in these groups than in patients without venous invasion (none vs MVI alone, p = 0.01; none vs EMVI, p < 0.001). EMVI = extramural venous invasion; MVI = mural venous invasion.
TABLE 3. Association between clinicopathological variables and venous invasion in 899 patients with stage C tumor
Male Female Age ≥ 75 y No Rectum Colon Maximum surface dimension ≥ 5 cm No Mucinous or signet-ring No Spread beyond muscularis propria No ≥ 4 nodes involved No Apical node involved No ≥ 30% of nodes involved No High grade No Free serosal surface involved No Adjacent structure infiltrated No Postoperative chemotherapy No Values are stated as n (%).
n (100%)
None
Mural alone
Extramural
χ2 p
566 333 268 631 404 495 412 487 99 800 775 124 278 621 75 824 299 600 321 578 171 728 38 861 226 673
410 (72) 236 (71) 200 (75) 446 (71) 267 (66) 379 (77) 294 (71) 352 (72) 80 (80) 566 (71) 534 (69) 112 (90) 168 (60) 478 (77) 44 (59) 602 (73) 173 (58) 473 (79) 189 (59) 457 (79) 105 (61) 541 (74) 24 (63) 622 (72) 174 (77) 472 (70)
29 (5) 13 (4) 9 (3) 33 (5) 20 (5) 22 (4) 15 (4) 27 (6) 2 (2) 40 (5) 36 (5) 6 (5) 14 (5) 28 (5) 4 (5) 38 (5) 19 (6) 23 (4) 17 (5) 25 (4) 10 (6) 32 (4) 0 42 (5) 9 (4) 33 (5)
127 (22) 84 (25) 59 (22) 152 (24) 117 (29) 94 (19) 103 (25) 108 (22) 17 (17) 194 (24) 205 (27) 6 (5) 96 (35) 115 (19) 27 (36) 184 (22) 107 (36) 104 (17) 115 (36) 96 (17) 56 (33) 115 (21) 14 (37) 197 (23) 43 (19) 168 (25)
0.48 0.33 0.002 0.28 0.09
