ORIGINAL CONTRIBUTION
Lymphovascular Invasion: A Comprehensive Appraisal in Colon and Rectal Adenocarcinoma John Hogan, M.B.1,2 • Kah Hoong Chang1 • Gerald Duff, M.B.1 • Georges Samaha, M.B.2,3 Niall Kelly, M.B.1,2 • Michael Burton, M.B.1 • Emily Burton, M.B.1 John Calvin Coffey, Ph.D.1,2,4 1 Department of Surgery, University Hospital Limerick, Limerick, Ireland 2 University of Limerick, Limerick, Ireland 3 Department of Surgery, Johns Hopkins Hospital, Baltimore, Maryland 4 Centre for Interventions in Infection, Inflammation and Immunity (4i), Graduate Entry Medical School, University of Limerick, Ireland
BACKGROUND: Colon and rectal adenocarcinomas differ
RESULTS: Five hundred twenty-seven patients were
at a multitude of levels. The association between outcome and predictor in 1 group may obscure the relationship between outcome and predictor in the other.
included in the analysis (379 with colon cancer and 148 with rectal cancer). On univariate analysis, lymphovascular invasion positivity was associated with adverse locoregional recurrence in colon (p = 0.002) but not rectal adenocarcinoma (p = 0.13). Conversely, lymphovascular invasion positivity was associated with adverse systemic recurrence in rectal (p = 0.002) but not colon adenocarcinoma (p = 0.35). On multivariate analysis, lymphovascular invasion positivity was an independent predictor of adverse disease-free survival in colon (p = 0.02) and rectal adenocarcinoma (p < 0.001). Regarding overall survival, lymphovascular invasion positivity was a poor prognostic indicator in rectal adenocarcinoma only (p = 0.04).
OBJECTIVE: The current study aims to evaluate the prognostic properties of lymphovascular invasion in colon and rectal adenocarcinoma separately. MATERIALS AND METHODS (DESIGN, SETTING AND PATIENTS): A comparative retrospective analysis
was undertaken to determine the prognostic properties of lymphovascular invasion in colon and rectal adenocarcinomas. Patients were classified as lymphovascular invasion positive and lymphovascular invasion negative in separate colon and rectal cancer cohorts. Within cohorts, a univariate analysis was undertaken to determine the association between lymphovascular invasion positivity and local/systemic recurrence and overall/disease-free survival. Findings were evaluated by using Kaplan-Meier estimates, log-rank analysis, and a Cox proportional hazards multivariate model. MAIN OUTCOME MEASURE: The primary outcomes measured were overall and disease-free survival. Financial Disclosures: None reported. Podium presentation at the meeting of The American Society of Colon and Rectal Surgeons, Hollywood, FL, May 17-21, 2014. Correspondence: J. Calvin Coffey, Ph.D., F.R.C.S.I., Department of Colorectal Surgery, University Hospital Limerick, Limerick, Ireland. E-mail: [email protected] Dis Colon Rectum 2015; 58: 547–555 DOI: 10.1097/DCR.0000000000000361 © The ASCRS 2015 Diseases of the Colon & Rectum Volume 58: 6 (2015)
LIMITATIONS AND CONCLUSIONS: In this retrospective analysis, lymphovascular invasion positivity was associated with different patterns of disease recurrence in colon and rectal cancer. Lymphovascular invasion positivity was associated with adverse overall survival in rectal cancer only. KEY WORDS: Colon; Rectum; Cancer; Lymphovascular invasion; Outcome.
C
olorectal cancer (CRC) is the third most common cancer worldwide with 102,000 new cases of colon adenocarcinoma and 40,000 new cases of rectal adenocarcinoma diagnosed in the United States in 2013. TNM staging is the standard in outcome prediction.1–5 However, 20% of node-negative patients develop disease recurrence, hence, the importance of additional prognostic parameters in distinguishing high- and low-risk patients.6–9 Lymphovascular invasion (LVI) (characterized by the extension of tumor cells into lymphatic and/or blood 547
Copyright © The American Society of Colon & Rectal Surgeons, Inc. Unauthorized reproduction of this article is prohibited.
548
vessels) is used as an adjunct to TNM staging in stage II CRC. Lymphovascular invasion positivity (LVI+) is a critical prognostic indicator and, as such, is documented in over 90% of histopathology reports. Substantial evidence exists indicating an association between LVI+ and adverse overall/disease-free survival, validating its use in clinical oncology. Systemic inflammation, perineural invasion, emergency presentation, and other pathological features are also established prognostic indicators.10–13 However, limited reporting of perineural invasion (0.5% of reports) and the relative rarity of emergency presentations (10%–15%) precludes their use as adjuncts in population-based staging.14–18 Colon (CC) and rectal cancer (RC) differ at a multitude of levels (epidemiology, etiology, genomic, proteomic, treatment modalities, survival, etc).19–22 The prognostic properties of LVI+ have traditionally been assessed in combined data sets (CC and RC together). However, the association between predictor and outcome in 1 disease category may obscure the true relationship between predictor and outcome in the other. This group has recently demonstrated this potential caveat.6 A retrospective comparative analysis comparing mucinous and nonmucinous adenocarcinoma of the colon alone demonstrated an independent association between mucinous and improved overall survival contrary to the findings of studies combining CC and RC. Consequently, the role of LVI+ in oncological outcomes is yet to be separately established in CC and RC,23,24
OBJECTIVE This study aims to separately evaluate the role of LVI+ in CC and RC with respect to overall survival (OS), disease-free survival (DFS), and locoregional and systemic recurrence.
MATERIALS AND METHODS A retrospective comparative study was undertaken to evaluate the prognostic properties of LVI in colon and rectal adenocarcinoma. Following institutional and ethical approval, a database was generated including all patients with a histopathological diagnosis of colon and rectal adenocarcinoma undergoing curative intent surgery during the interval 2000 to 2010. Data were obtained from pathology/endoscopy/radiology reports, the Hospital Inpatient Enquiry System, and direct contact with primary care physicians. Collated data included age, sex, TNM stage (American Joint Committee on Cancer 7th edition), level of differentiation (well, moderate, and poorly differentiated), lymphovascular invasion, lymph node yield, number of metastatic lymph nodes, tumor size, anatomical location, mode of presentation (emergency vs elective), and margin status. Postoperatively, patients entered a standard surveillance program comprising annual CT for 3 years and colo-
HOGAN ET AL: COMPARING LVI IN COLON/RECTAL CANCER
noscopy at years 1 and 3. Mean surveillance duration was 42.4 months. Surveillance data were recorded to 2013 and included in the final database. The incidence of and time to recurrence/death were recorded providing for a comprehensive survival analysis. The current study aimed to evaluate and compare the prognostic properties of LVI+ in the context of colon and rectal adenocarcinoma separately. Primary end points included OS (all-cause mortality) and DFS (locoregional and systemic recurrence). Patient survival status was determined through direct contact with each primary care physician and precise dates of death recorded. Final histological diagnosis and the presence of LVI+ were determined at a multidisciplinary team meeting where slides were reviewed by an independent panel of pathologists. Magnetic resonance imaging and endoscopy findings were reviewed, and the anatomical location of tumors (colon vs rectum) was determined. Patients were included if they were undergoing curative intent surgery and if they had adequate follow-up. Patients were excluded if they had documented metastatic disease at the time of diagnosis (stage IV), were undergoing neoadjuvant therapy, and had died within 30 days of surgery. Three cohorts were generated for comparison: CC, RC, and CRC (combined colon and rectal adenocarcinoma). Within cohorts, cases were categorized as lymphovascular invasion positive (LVI+) or lymphovascular invasion negative (LVI–). The following analysis was performed in the context of CC, RC, and CRC separately. A χ2 test was used to evaluate the differences in distribution of clinicopathological factors between LVI+ and LVI– groups (in CC, RC, and CRC). A t test/Mann-Whitney U test (depending on the distribution of data) was used to compare continuous variables between groups. Kaplan-Meier estimates were generated comparing LVI+ and LVI– groups in CC, RC, and CRC. Differences in distribution of survival curves (LVI+ vs LVI–) were further evaluated by using a log-rank test. A multivariate Cox proportional hazards model was used to determine the independent association between LVI+ and OS/DFS in CC, RC, then CRC. Results were compared across 3 cohorts to evaluate the differences in prognostic value of LVI+ in CC, RC, and CRC. All calculations were performed in SPSS (Chicago Inc. version 15). p < 0.05 was considered statistically significant.
RESULTS Demographics (CC vs RC)
Five hundred twenty-seven patients underwent surgery for CRC during the study period. Of these patients, 379 were categorized as CC and 148 as RC (Table 1). There was no difference in the incidence of LVI+ and LVI– between CC and RC (101/379, 26.6% vs 44/148, 29.7%, p = 0.78). The incidence of locoregional recurrence (LR) (27/379,
Copyright © The American Society of Colon & Rectal Surgeons, Inc. Unauthorized reproduction of this article is prohibited.
549
Diseases of the Colon & Rectum Volume 58: 6 (2015)
Table 1. Differences in distribution of clinicopathological characteristics in colon and rectal cancer Parameter
Colon (n = 379)
Rectum (n = 148)
Colorectal (n = 527)
Sex, male, n (%) Age, mean, y N stage, n (%) 0 1 2 AJCC stage, n (%) I II III Total lymph node yield, mean Positive lymph node yield, mean Lymph node ratio, mean Lymphovascular invasion present, n (%) Local recurrence, incidence, n (%) Systemic recurrence, incidence, n (%)
212 (55) 67.1 ± 11.4
106 (71) 65.5 ± 11.1
318 (60.3) 66.8 ± 11.3
0.03 0.39
210 (55.4) 117 (30.8) 52 (13.7)
94 (60.1) 35 (23.6) 19 (12.8)
304 (57.7) 152 (28.8) 71 (13.5)
0.53 0.26 0.99
79 225 223 9.93 ± 6.56 1.52 ± 2.91 0.155 ± 0.249 145 (27.5) 30 (5.69) 119 (22.6)
0.02 0.35 0.08 0.63 0.85 0.45 0.78 0.91 0.33
43 (11.3) 167 (44.1) 169 (44.5) 10.1 ± 6.6 1.57 ± 2.92 0.1625 ± 0.254 101 (26.6) 27 (7.12) 92 (24.2)
36 (24.32) 58 (39.19) 54 (42.19) 9.49 ± 6.28 1.41 ± 2.89 0.138 ± 0.237 44 (29.7) 3 (2.02) 27 (18.2)
p
A χ2 test was used to assess differences in the distribution of clinicopathological characteristics in colon and rectal cancer. There were more stage I tumors in RC (p = 0.02) than in CC. AJCC = American Joint Committee on Cancer.
7.12% vs 3/148, 2.02%, p = 0.91) and systemic recurrence (SR) (92/379, 24.2% vs 27/148, 18.2%, p = 0.33) was similar in CC and RC. The proportion of stage I tumors was greater in RC (36/148, 24.32%) than in CC (43/379, 11.3%, p = 0.02). (Table 1). Nonetheless, the incidence of LR (p = 0.91) and SR (p = 0.33) was similar between CC and RC (reflecting the limitation of χ2 test in determining oncological outcomes). Local Recurrence (LVI+ vs LVI– Groups)
Twenty-seven patients developed LR in CC during the follow-up period. Local recurrence was defined as tumor recurrence at the anastomotic site. Twenty of 27 (74%) patients were classified as LVI+. Twenty-four of 27 (88.89%) patients were classified as stage III, and 3 (11.11%) were classified as stage II. The majority of tumors developing LR were high risk (stage III, LVI+). A χ2 test was used to compare the incidence of LR between LVI+ and LVI– groups with respect to CC, RC, and CRC. Lymphovascular invasion positivity was associated with increased incidence of LR in CC (p = 0.002) and CRC (p < 0.001) but not RC (p = 0.13) (Table 2). Regarding CC, 19 (18.8%) of LVI+ (n = 101) patients developed LR compared with 8 (2.89%) in the LVI– group (n = 278) (p = 0.002). Regard-
ing CRC, 21 (14.4%) of LVI+ (n = 145) patients developed LR compared with 9 (2.36%) patients in the LVI– group (n = 382) (p < 0.001). In RC, there was no difference in the incidence of LR between LVI+ (n = 44) and LVI– (n = 104) groups (2, 4.54% vs 1, 0.96%, p = 0.13) (Table 2). Survival distributions of LVI+/LVI– groups were plotted in Kaplan-Meier estimates. The differences in the distribution of survival curves (LR) were compared by using a logrank test. Three graphs were generated representing CC, RC, and CRC. Lymphovascular invasion positivity was associated with adverse LR in CC (p < 0.001) (Fig. 1A) and CRC (p < 0.001) (Fig. 1C) but not RC (p = 0.15) (Fig. 1B). Systemic Recurrence (LVI+ vs LVI– Groups)
A χ2 test was used to compare the incidence of SR between groups (LVI+ vs LVI–) with respect to CC, RC, and CRC. Lymphovascular invasion positivity was associated with an increased incidence of SR in RC (p = 0.002) and CRC (p = 0.02) but not CC (p = 0.35) (Table 3). Regarding RC, 15 (34.1%) of LVI+ patients (n = 44) developed SR compared with 12 (11.5%) patients in the LVI– group (n = 104) (p = 0.002). Similarly, in CRC, 43 (29.3%) LVI+ patients (n = 145) developed SR compared with 76 (19.8%) patients in the LVI– group (n = 382) (p = 0.02). There was no association
Table 2. Patients with established locoregional recurrence Colon cancer
Total LR, n (%)
LVI+
LVI–
101 19 (18.8)
278 8 (2.89)
Rectal cancer p 0.002
Colorectal cancer
LVI+
LVI–
p
LVI+
LVI–
p
44 2 (4.54)
104 1 (0.96)
0.13
145 21 (14.4)
382 9 (2.36)
< 0.001
Patients with established LR were categorized as LVI+ and LVI–. The incidence of LR was compared between LVI+ and LVI– groups by using a χ test. LR was greater in LVI+ groups (in comparison with LVI– groups) in CC and CRC but not RC. CC = colon cancer; CRC = colorectal cancer; LR = locoregional recurrence; LVI+ = lymphovascular invasion positive; LVI– = lymphovascular invasion negative; RC = rectal cancer. 2
Copyright © The American Society of Colon & Rectal Surgeons, Inc. Unauthorized reproduction of this article is prohibited.
550
HOGAN ET AL: COMPARING LVI IN COLON/RECTAL CANCER
A 1.0
Colon cancer
B 1.0
Rectal cancer
C
Colorectal cancer
1.0
0.8
0.8
0.8
0.6
0.6
0.6
0.4
0.4
0.4
0.2
0.2
0.2
p < 0.001
p = 0.15
p < 0.001
5
5
5
D
10 15 20 25 30 Local recurrence Colon cancer
E
10 15 20 25 30 Local recurrence
1.0
1.0
Rectal cancer
F 1.0
0.8
0.8
0.8
0.6
0.6
0.6
0.4
0.4
0.4
0.2
0.2
0.2 p < 0.001
p = 0.28 5 10 15 20 25 Systemic recurrence
30
10 15 20 25 30 Local recurrence Colorectal cancer
p < 0.001
5 10 15 20 25 Systemic recurrence
30
5 10 15 20 25 30 Systemic recurrence
Figure 1. X axis = survival (months), y axis = cumulative survival. A–C, Survival distributions (LR) of lymphovascular invasion positive (LVI+) and negative (LVI–) cases were plotted and compared in Kaplan-Meier estimates (log-rank analysis). Lymphovascular invasion positivity was associated with an increased rate of LR in CC (p < 0.001) and CRC (p < 0.001) but not RC (p = 0.15). D–F, Survival distributions (SR) of LVI+ and LVI– cases were plotted and compared in Kaplan-Meier estimates (log-rank analysis). Lymphovascular invasion positivity was associated with increased rate of SR in RC (p < 0.001) and CRC (p < 0.001) but not CC (p =0.28). CC = colon cancer; CRC = colorectal cancer; LR = locoregional recurrence; RC = rectal cancer; SR = systemic recurrence.
between LVI+ and SR in CC (p = 0.35) (Table 3). The survival distributions of LVI+/LVI– groups were plotted and compared in Kaplan-Meier estimates. Differences in the distribution of survival curves (SR) were further evaluated by using a log-rank test. Three graphs were generated representing RC, CRC, and CC. Lymphovascular invasion positivity was associated with adverse SR in RC (p < 0.001) (Fig. 1E) and CRC (p < 0.001) (Fig. 1F) but not CC (p = 0.28) (Fig. 1D). Multivariate Analysis: Association Between LVI+ and LR
Lymphovascular invasion status and established prognostic factors were incorporated into a multivariate Cox proportional hazards model to evaluate the independent association between LVI+ and recurrence. As expected,
stage III tumors were independent predictors of adverse LR with respect to CC (HR, 2.17; 95% CI, 1.64–3.62; p = 0.02; Table 4), RC (HR, 2.57; 95% CI, 1.01–6.93; p = 0.05; Table 5) and CRC (HR, 2.37; 95% CI, 1.32–3.51; p < 0.001; Table 6). Lymphovascular invasion positivity was an independent predictor of adverse LR in CC (HR, 1.90; 95% CI, 1.10–3.27; p = 0.02) and CRC (HR, 1.79; 95% CI, 1.11–2.90; p = 0.02) but not RC (HR, 1.57; 95% CI, 0.55–4.49; p = 0.40). Multivariate Analysis: Association Between LVI+ and SR
The independent association between LVI+ and SR was evaluated in a multivariate Cox proportional hazards
Table 3. Patients with established systemic recurrence Colon cancer
Total SR, n (%)
LVI+
LVI–
101 28 (27.7)
278 64 (23)
Rectal cancer p 0.35
LVI+
LVI–
44 15 (34.1)
104 12 (11.5)
Colorectal cancer p 0.002
LVI+
LVI–
145 43 (29.3)
382 76 (19.8)
p 0.02
Patients with established SR were categorized LVI+ and LVI–. The incidence of SR was compared between LVI+ and LVI– groups by using a χ test. SR was greater in LVI+ groups (in comparison with LVI– groups) in CRC and RC but not CC. CC = colon cancer; CRC = colorectal cancer; LVI+ = lymphovascular invasion positive; LVI– = lymphovascular invasion negative; RC = rectal cancer; SR = systemic recurrence. 2
Copyright © The American Society of Colon & Rectal Surgeons, Inc. Unauthorized reproduction of this article is prohibited.
551
Diseases of the Colon & Rectum Volume 58: 6 (2015)
Table 4. Analysis of independent association between LVI+ and LR/SR in CC Locoregional recurrence Parameter Age Male Stage III Total LN LVI
Systemic recurrence
p
HR
95% CI
p
HR
95% CI
0.94 0.77 0.02 0.81 0.02
1.00 1.08 2.17 1.00 1.90
0.98–1.02 0.65–1.79 1.64–3.62 0.96–1.03 1.10–3.27
0.53 0.60 0.02 0.03 0.88
0.99 1.12 1.84 0.96 1.04
0.98–1.01 0.74–1.71 1.16–3.09 0.93–1.00 0.64–1.67
Established prognostic factors and LVI were incorporated into a Cox proportional hazards multivariate model to evaluate the independent association between LVI+ and LR/ SR. LVI+ was an independent predictor of adverse LR in CC (p = 0.02) and CRC (p = 0.02) but not RC (p = 0.40). LVI+ was an independent predictor of adverse SR in RC (p = 0.04) but not CC (p = 0.88) and CRC (p = 0.31). CC = colon cancer; CRC = colorectal cancer; LR = locoregional; LN = lymph node; LVI = lymphovascular invasion; LVI+ = lymphovascular invasion positive; LVI– = lymphovascular invasion negative; RC = rectal cancer; SR = systemic recurrence.
model. As expected, stage III tumors were independent predictors of adverse SR in CC (HR, 1.84; 95% CI, 1.16– 3.09; p = 0.02; Table 4), RC (HR, 1.66; 95% CI, 0.69–4.32; p = 0.05; Table 5) and CRC (HR, 1.79; 95% CI, 1.18–2.84; p = 0.02; Table 6). Lymphovascular invasion positivity was not associated with adverse SR in CC (HR, 1.04; 95% CI, 0.64–1.67; p = 0.88) and CRC (HR, 1.24; 95% CI, 0.82– 1.86; p = 0.31) but was an independent predictor of adverse SR in RC (HR, 2.57; 95% CI, 1.04–6.39; p = 0.04) (Table 4). Multivariate Analysis: Association Between LVI+ and DFS/OS
The association between LVI+ and DFS was evaluated in a multivariate Cox proportional hazards model (Table 5). Lymphovascular invasion positivity was an independent predictor of adverse DFS in CC (HR, 1.49; 95% CI, 1.07–2.07; p = 0.02; Table 7), RC (HR, 2.69; 95% CI, 1.36–5.32; p < 0.001; Table 8) and CRC (HR, 1.61; 95% CI, 1.20–2.16; p < 0.001; Table 9). Kaplan-Meier estimates comparing LVI+ and LVI– groups demonstrated adverse DFS in LVI+ patients in CC (p < 0.001), RC (p < 0.001), and CRC (p < 0.001) (Fig. 2A–C). In the context of OS, LVI+ was not an independent predictor of adverse outcomes (multivariate analysis) in CC (p = 0.11, Table 7) and CRC (p = 0.20, Table 9). Lymphovascular invasion positivity was an independent predictor of adverse
OS in RC (HR, 2.32; 95% CI, 0.75–7.13; p = 0.04; Table 8). Kaplan-Meier estimates comparing survival distributions (OS) of LVI+ and LVI– groups demonstrated adverse survival in CC (p < 0.001), CRC (p < 0.001), and RC (p = 0.05) (Fig. 2D–2F).
DISCUSSION Colon and rectal adenocarcinomas differ at a multitude of levels and are increasingly regarded as distinct biological entities. In particular, the differences in surgical technique; adjuvant, neoadjuvant therapies; and survival rates may obscure the true relationship between outcome and prognostic indicator when colon and rectal adenocarcinomas are combined as a single cohort in data set analysis. This study evaluated differences in relationships between LVI+ and oncological end points in colon (CC) and rectal adenocarcinomas (RC) separately. Results regarding CC and RC are presented adjacent to the CRC group to highlight the discrepancies between combined data sets and CC/RC cohorts (which may have significant clinical consequences). Univariate and multivariate analyses demonstrated that the relationship between LVI+ and outcome in CC differed from LVI+ and outcome in RC. Previous studies evaluating outcomes in CRC data sets may be flawed with 1 group skewing the results of the other.
Table 5. Analysis of independent association between LVI+ and LR/SR in RC Locoregional recurrence Parameter Age Male Stage III Total LN LVI
Systemic recurrence
p
HR
95% CI
p
HR
95% CI
0.37 0.88 0.05 0.09 0.40
1.02 1.09 2.57 0.92 1.57
0.98–1.07 0.37–3.21 1.01–6.93 0.83–1.01 0.55–4.49
0.49 0.42 0.05 0.76 0.04
0.99 0.68 1.66 0.99 2.57
0.95–1.03 0.24–1.80 0.69–4.32 0.91–1.07 1.04–6.39
Established prognostic factors and LVI were incorporated into a Cox proportional hazards multivariate model to evaluate the independent association between LVI+ and LR/ SR. LVI+ was an independent predictor of adverse LR in CC (p = 0.02) and CRC (p = 0.02) but not RC (p = 0.40). LVI+ was an independent predictor of adverse SR in RC (p = 0.04) but not CC (p = 0.88) and CRC (p = 0.31). CC = colon cancer; CRC = colorectal cancer; LR = locoregional; LN = lymph node; LVI = lymphovascular invasion; LVI+ = lymphovascular invasion positive; LVI– = lymphovascular invasion negative; RC = rectal cancer; SR = systemic recurrence.
Copyright © The American Society of Colon & Rectal Surgeons, Inc. Unauthorized reproduction of this article is prohibited.
552
HOGAN ET AL: COMPARING LVI IN COLON/RECTAL CANCER
Table 6. Analysis of independent association between LVI+ and LR/ SR in CRC Locoregional recurrence Parameter Age Male Stage III Total LN LVI
Systemic recurrence
p
HR
95% CI
p
HR
0.74 0.97 < 0.001 0.33 0.02
1.00 1.01 2.37 0.98 1.79
0.98–1.02 0.65–1.57 1.32–3.51 0.95–1.02 1.11–2.90
0.32 0.61 0.02 0.03 0.31
0.99 1.10 1.79 0.96 1.24
95% CI 0.98–1.01 0.76–1.59 1.18–2.84 0.93–1.00 0.82–1.86
Established prognostic factors and LVI were incorporated into a Cox proportional hazards multivariate model to evaluate the independent association between LVI+ and LR/SR. LVI+ was an independent predictor of adverse LR in CC (p = 0.02) and CRC (p = 0.02) but not RC (p = 0.40). LVI+ was an independent predictor of adverse SR in RC (p = 0.04) but not CC (p = 0.88) and CRC (p = 0.31). CC = colon cancer; CRC = colorectal cancer; LR = locoregional; LN = lymph node; LVI = lymphovascular invasion; LVI+ = lymphovascular invasion positive; LVI– = lymphovascular invasion negative; RC = rectal cancer; SR = systemic recurrence.
Kaplan-Meier estimates demonstrated an adverse association between LVI+ and both OS and DFS in all 3 cohorts (CC, RC, and CRC). On multivariate analysis however, LVI+ was associated with adverse OS in RC only (the association was not significant in CC). Conversely, LVI+ was associated with adverse DFS in CC, RC, and CRC. The pattern of disease recurrence differed between CC and RC. Lymphovascular invasion positivity increased LR in CC and SR in RC. These findings persisted in Kaplan-Meier estimates (compared by using log-rank analysis) and univariate and multivariate analysis. Surprisingly, the incidence of LR was greater in CC than in RC. This did not approach statistical significance and may reflect the greater proportion of stage I tumors in RC (statistically significant). The analysis of LVI in patients receiving adjuvant and neoadjuvant therapies were not undertaken in the current study. However, preliminary work suggests that LVI retains its prognostic properties in CC and RC (with respect to OS and DFS). Substantial evidence exists validating the association between LVI+ and adverse survival (disease-free and overall) in combined CC and RC data sets. Disparity exists between the results of the current study and published literature for a number of reasons, such as 1) differences between outcome variables evaluated (OS, DFS, LR, and/or SR), 2) depth of analysis (majority of
studies do not analyze SR and LR separately), and 3) the combination of CC and RC, which differ at a multitude of levels (proteomic, genomic, epidemiological, and oncological outcome). Few studies overcome all these potential limitations. Huh and colleagues27 performed a univariate/multivariate survival analysis evaluating LVI and oncological outcomes in CC and RC. Lymphovascular invasion was associated with adverse DFS in both CC and RC (similar to the findings of the current study). However, methods differed substantially with respect to the depth of analysis. Huh did not examine the role of LVI in LR and SR separately, and, as such, the concordance between Huh and the current results is unclear.27 Patel et al28 reported no association between LVI and LR in early-stage RC validating the results discussed in this article. Systemic recurrence was not evaluated, nor did the authors compare outcomes between CC and RC. Bhangu et al29 evaluated LVI in SR and LR in RC alone. On multivariate analysis of T4 rectal tumors, vascular invasion was associated with adverse OS and DFS. When LR and SR were analyzed separately, the authors reported no association between LVI and LR but an adverse association between LVI and SR (in RC). This is one of few studies evaluating both LR and SR in the context of CC or RC alone with the use of patients from a single institute.29 Similar to the results of this article, LVI has
Table 7. Evaluation of the independent association between LVI+ and DFS/OS in CC Disease-free-survival Parameter Age Male Stage III Total LN LVI
Overall survival
P value
HR
95% CI
P value
HR
0.30 0.79 0.05 0.15 0.02
1.01 1.04 1.49 0.98 1.49
0.98–1.02 0.77–1.40 1.12–2.08 0.96–1.01 1.07–2.07
0.40 0.92 0.91 0.01 0.11
1.01 1.02 1.04 0.96 1.35
95% CI 0.99–1.02 0.73–1.42 0.63–1.61 0.93–0.99 0.93–1.95
Established prognostic factors and LVI were incorporated into a Cox proportional hazards multivariate model to evaluate the independent association between LVI+ and DFS/OS. LVI+ was an independent predictor of adverse DFS in CC (p = 0.02), RC (p < 0.001) (Table 8), and CRC (p < 0.001) (Table 9). LVI+ was associated with adverse OS in RC (p = 0.04) but not CC (p = 0.11) and CRC (p = 0.20). CC = colon cancer; CRC = colorectal cancer; DFS = disease-free survival; LN = lymph node; LVI = lymphovascular invasion; LVI+ = lymphovascular invasion positive; OS = overall survival; RC = rectal cancer.
Copyright © The American Society of Colon & Rectal Surgeons, Inc. Unauthorized reproduction of this article is prohibited.
553
Diseases of the Colon & Rectum Volume 58: 6 (2015)
Table 8. Evaluation of the independent association between LVI+ and DFS/OS in RC Disease-free-survival Parameter Age Male Stage III Total LN LVI
Overall survival
p
HR
95% CI
p
HR
0.21 0.24 0.17 0.67 < 0.001
1.02 0.63 1.64 0.99 2.69
0.99–1.05 0.29–1.36 0.69–3.32 0.93–1.05 1.36–5.32
0.15 0.11 0.77 0.70 0.04
1.04 2.82 1.25 0.98 2.32
95% CI 0.99–1.10 0.78–10.11 0.45–4.72 0.89–1.09 0.75–7.13
Established prognostic factors and LVI were incorporated into a Cox proportional hazards multivariate model to evaluate the independent association between LVI+ and DFS/OS. LVI+ was an independent predictor of adverse DFS in CC (p = 0.02) (Table 7), RC (p < 0.001), and CRC (p < 0.001) (Table 9). LVI+ was associated with adverse OS in RC (p = 0.04) but not CC (p = 0.11) and CRC (p = 0.20). CC = colon cancer; CRC = colorectal cancer; DFS = disease-free survival; LN = lymph node; LVI = lymphovascular invasion; LVI+ = lymphovascular invasion positive; OS = overall survival; RC = rectal cancer.
unique prognostic properties in the context of RC (adverse SR but not LR). In summary, LVI+ was associated with increased LR in CC and increased SR in RC. This prompts the question as to why LVI+ is associated with LR in CC but not RC. Total mesorectal excision is composed of the removal of tumors and the entire unit of regional spread and is associated with decreased LR. No such standard of "mesocolic excision" exists with respect to CC, and potentially metastatic lymphatics remain in situ adversely affecting LR in CC.30 Conversely, LVI+ was associated with increased SR in RC but not CC. This discrepancy may be explained by the differences in venous drainage of the colon and rectum. The superior, middle, and inferior rectal veins drain the rectum. Middle and inferior rectal veins drain to the inferior vena cava via the iliac vessels. The colon drains via the superior mesenteric vein to the portal vein. In this manner, parts of the rectum drain directly to the systemic circulation, which may explain the increased incidence of SR in LVI+ RC. This may also account for the finding that LVI+ was associated with adverse OS in RC only (on multivariate analysis). Lymphovascular invasion positivity increased LR in CC, and LR is frequently amenable to secondary surgery. Therefore mortality rates and OS remain unchanged. In RC, LVI+ was associated with increased SR, which is less amenable to curative interventions, and so OS is adversely affected.31,32
LIMITATIONS This study has some limitations. 1) It is a retrospective analysis and the RC cohort is relatively small. This increases the potential for type II statistical error. The incidence of LR in RC is low, further increasing the potential for statistical error. A multi-institutional collaborative study or highvolume single-institute study may overcome this issue in future studies. 2) The role of LVI in patients undergoing adjuvant/neoadjuvant therapy is unclear and was not evaluated in the current article. 3) Median lymph node retrieval was less than 12 in CC and RC, raising issues regarding appropriate staging. However, significant disparity exists within the literature regarding the minimum lymph node yield associated with improved survival. 4) Data regarding grade of LVI and the differentiation between extramural/ intramural invasion was not available for analysis. Sample size was sufficient to detect statistical significance, and this is one of a few studies to evaluate the prognostic properties of LVI+ in colon and rectal adenocarcinomas separately.
CONCLUSIONS In conclusion, the pattern of disease recurrence differs between LVI+ CC and LVI+ RC, resulting in discrepancies in the prognostic properties of LVI+ in CC and RC. Future studies assessing the prognostic indicators in CRC should evaluate CC and RC separately.
Table 9. Evaluation of the independent association between LVI+ and DFS/OS in CRC Disease-free-survival Parameter Age Male Stage III Total LN LVI
Overall survival
p
HR
95% CI
p
HR
95% CI
0.20 0.69 0.01 0.10
Lymphovascular Invasion: A Comprehensive Appraisal in Colon and Rectal Adenocarcinoma John Hogan, M.B.1,2 • Kah Hoong Chang1 • Gerald Duff, M.B.1 • Georges Samaha, M.B.2,3 Niall Kelly, M.B.1,2 • Michael Burton, M.B.1 • Emily Burton, M.B.1 John Calvin Coffey, Ph.D.1,2,4 1 Department of Surgery, University Hospital Limerick, Limerick, Ireland 2 University of Limerick, Limerick, Ireland 3 Department of Surgery, Johns Hopkins Hospital, Baltimore, Maryland 4 Centre for Interventions in Infection, Inflammation and Immunity (4i), Graduate Entry Medical School, University of Limerick, Ireland
BACKGROUND: Colon and rectal adenocarcinomas differ
RESULTS: Five hundred twenty-seven patients were
at a multitude of levels. The association between outcome and predictor in 1 group may obscure the relationship between outcome and predictor in the other.
included in the analysis (379 with colon cancer and 148 with rectal cancer). On univariate analysis, lymphovascular invasion positivity was associated with adverse locoregional recurrence in colon (p = 0.002) but not rectal adenocarcinoma (p = 0.13). Conversely, lymphovascular invasion positivity was associated with adverse systemic recurrence in rectal (p = 0.002) but not colon adenocarcinoma (p = 0.35). On multivariate analysis, lymphovascular invasion positivity was an independent predictor of adverse disease-free survival in colon (p = 0.02) and rectal adenocarcinoma (p < 0.001). Regarding overall survival, lymphovascular invasion positivity was a poor prognostic indicator in rectal adenocarcinoma only (p = 0.04).
OBJECTIVE: The current study aims to evaluate the prognostic properties of lymphovascular invasion in colon and rectal adenocarcinoma separately. MATERIALS AND METHODS (DESIGN, SETTING AND PATIENTS): A comparative retrospective analysis
was undertaken to determine the prognostic properties of lymphovascular invasion in colon and rectal adenocarcinomas. Patients were classified as lymphovascular invasion positive and lymphovascular invasion negative in separate colon and rectal cancer cohorts. Within cohorts, a univariate analysis was undertaken to determine the association between lymphovascular invasion positivity and local/systemic recurrence and overall/disease-free survival. Findings were evaluated by using Kaplan-Meier estimates, log-rank analysis, and a Cox proportional hazards multivariate model. MAIN OUTCOME MEASURE: The primary outcomes measured were overall and disease-free survival. Financial Disclosures: None reported. Podium presentation at the meeting of The American Society of Colon and Rectal Surgeons, Hollywood, FL, May 17-21, 2014. Correspondence: J. Calvin Coffey, Ph.D., F.R.C.S.I., Department of Colorectal Surgery, University Hospital Limerick, Limerick, Ireland. E-mail: [email protected] Dis Colon Rectum 2015; 58: 547–555 DOI: 10.1097/DCR.0000000000000361 © The ASCRS 2015 Diseases of the Colon & Rectum Volume 58: 6 (2015)
LIMITATIONS AND CONCLUSIONS: In this retrospective analysis, lymphovascular invasion positivity was associated with different patterns of disease recurrence in colon and rectal cancer. Lymphovascular invasion positivity was associated with adverse overall survival in rectal cancer only. KEY WORDS: Colon; Rectum; Cancer; Lymphovascular invasion; Outcome.
C
olorectal cancer (CRC) is the third most common cancer worldwide with 102,000 new cases of colon adenocarcinoma and 40,000 new cases of rectal adenocarcinoma diagnosed in the United States in 2013. TNM staging is the standard in outcome prediction.1–5 However, 20% of node-negative patients develop disease recurrence, hence, the importance of additional prognostic parameters in distinguishing high- and low-risk patients.6–9 Lymphovascular invasion (LVI) (characterized by the extension of tumor cells into lymphatic and/or blood 547
Copyright © The American Society of Colon & Rectal Surgeons, Inc. Unauthorized reproduction of this article is prohibited.
548
vessels) is used as an adjunct to TNM staging in stage II CRC. Lymphovascular invasion positivity (LVI+) is a critical prognostic indicator and, as such, is documented in over 90% of histopathology reports. Substantial evidence exists indicating an association between LVI+ and adverse overall/disease-free survival, validating its use in clinical oncology. Systemic inflammation, perineural invasion, emergency presentation, and other pathological features are also established prognostic indicators.10–13 However, limited reporting of perineural invasion (0.5% of reports) and the relative rarity of emergency presentations (10%–15%) precludes their use as adjuncts in population-based staging.14–18 Colon (CC) and rectal cancer (RC) differ at a multitude of levels (epidemiology, etiology, genomic, proteomic, treatment modalities, survival, etc).19–22 The prognostic properties of LVI+ have traditionally been assessed in combined data sets (CC and RC together). However, the association between predictor and outcome in 1 disease category may obscure the true relationship between predictor and outcome in the other. This group has recently demonstrated this potential caveat.6 A retrospective comparative analysis comparing mucinous and nonmucinous adenocarcinoma of the colon alone demonstrated an independent association between mucinous and improved overall survival contrary to the findings of studies combining CC and RC. Consequently, the role of LVI+ in oncological outcomes is yet to be separately established in CC and RC,23,24
OBJECTIVE This study aims to separately evaluate the role of LVI+ in CC and RC with respect to overall survival (OS), disease-free survival (DFS), and locoregional and systemic recurrence.
MATERIALS AND METHODS A retrospective comparative study was undertaken to evaluate the prognostic properties of LVI in colon and rectal adenocarcinoma. Following institutional and ethical approval, a database was generated including all patients with a histopathological diagnosis of colon and rectal adenocarcinoma undergoing curative intent surgery during the interval 2000 to 2010. Data were obtained from pathology/endoscopy/radiology reports, the Hospital Inpatient Enquiry System, and direct contact with primary care physicians. Collated data included age, sex, TNM stage (American Joint Committee on Cancer 7th edition), level of differentiation (well, moderate, and poorly differentiated), lymphovascular invasion, lymph node yield, number of metastatic lymph nodes, tumor size, anatomical location, mode of presentation (emergency vs elective), and margin status. Postoperatively, patients entered a standard surveillance program comprising annual CT for 3 years and colo-
HOGAN ET AL: COMPARING LVI IN COLON/RECTAL CANCER
noscopy at years 1 and 3. Mean surveillance duration was 42.4 months. Surveillance data were recorded to 2013 and included in the final database. The incidence of and time to recurrence/death were recorded providing for a comprehensive survival analysis. The current study aimed to evaluate and compare the prognostic properties of LVI+ in the context of colon and rectal adenocarcinoma separately. Primary end points included OS (all-cause mortality) and DFS (locoregional and systemic recurrence). Patient survival status was determined through direct contact with each primary care physician and precise dates of death recorded. Final histological diagnosis and the presence of LVI+ were determined at a multidisciplinary team meeting where slides were reviewed by an independent panel of pathologists. Magnetic resonance imaging and endoscopy findings were reviewed, and the anatomical location of tumors (colon vs rectum) was determined. Patients were included if they were undergoing curative intent surgery and if they had adequate follow-up. Patients were excluded if they had documented metastatic disease at the time of diagnosis (stage IV), were undergoing neoadjuvant therapy, and had died within 30 days of surgery. Three cohorts were generated for comparison: CC, RC, and CRC (combined colon and rectal adenocarcinoma). Within cohorts, cases were categorized as lymphovascular invasion positive (LVI+) or lymphovascular invasion negative (LVI–). The following analysis was performed in the context of CC, RC, and CRC separately. A χ2 test was used to evaluate the differences in distribution of clinicopathological factors between LVI+ and LVI– groups (in CC, RC, and CRC). A t test/Mann-Whitney U test (depending on the distribution of data) was used to compare continuous variables between groups. Kaplan-Meier estimates were generated comparing LVI+ and LVI– groups in CC, RC, and CRC. Differences in distribution of survival curves (LVI+ vs LVI–) were further evaluated by using a log-rank test. A multivariate Cox proportional hazards model was used to determine the independent association between LVI+ and OS/DFS in CC, RC, then CRC. Results were compared across 3 cohorts to evaluate the differences in prognostic value of LVI+ in CC, RC, and CRC. All calculations were performed in SPSS (Chicago Inc. version 15). p < 0.05 was considered statistically significant.
RESULTS Demographics (CC vs RC)
Five hundred twenty-seven patients underwent surgery for CRC during the study period. Of these patients, 379 were categorized as CC and 148 as RC (Table 1). There was no difference in the incidence of LVI+ and LVI– between CC and RC (101/379, 26.6% vs 44/148, 29.7%, p = 0.78). The incidence of locoregional recurrence (LR) (27/379,
Copyright © The American Society of Colon & Rectal Surgeons, Inc. Unauthorized reproduction of this article is prohibited.
549
Diseases of the Colon & Rectum Volume 58: 6 (2015)
Table 1. Differences in distribution of clinicopathological characteristics in colon and rectal cancer Parameter
Colon (n = 379)
Rectum (n = 148)
Colorectal (n = 527)
Sex, male, n (%) Age, mean, y N stage, n (%) 0 1 2 AJCC stage, n (%) I II III Total lymph node yield, mean Positive lymph node yield, mean Lymph node ratio, mean Lymphovascular invasion present, n (%) Local recurrence, incidence, n (%) Systemic recurrence, incidence, n (%)
212 (55) 67.1 ± 11.4
106 (71) 65.5 ± 11.1
318 (60.3) 66.8 ± 11.3
0.03 0.39
210 (55.4) 117 (30.8) 52 (13.7)
94 (60.1) 35 (23.6) 19 (12.8)
304 (57.7) 152 (28.8) 71 (13.5)
0.53 0.26 0.99
79 225 223 9.93 ± 6.56 1.52 ± 2.91 0.155 ± 0.249 145 (27.5) 30 (5.69) 119 (22.6)
0.02 0.35 0.08 0.63 0.85 0.45 0.78 0.91 0.33
43 (11.3) 167 (44.1) 169 (44.5) 10.1 ± 6.6 1.57 ± 2.92 0.1625 ± 0.254 101 (26.6) 27 (7.12) 92 (24.2)
36 (24.32) 58 (39.19) 54 (42.19) 9.49 ± 6.28 1.41 ± 2.89 0.138 ± 0.237 44 (29.7) 3 (2.02) 27 (18.2)
p
A χ2 test was used to assess differences in the distribution of clinicopathological characteristics in colon and rectal cancer. There were more stage I tumors in RC (p = 0.02) than in CC. AJCC = American Joint Committee on Cancer.
7.12% vs 3/148, 2.02%, p = 0.91) and systemic recurrence (SR) (92/379, 24.2% vs 27/148, 18.2%, p = 0.33) was similar in CC and RC. The proportion of stage I tumors was greater in RC (36/148, 24.32%) than in CC (43/379, 11.3%, p = 0.02). (Table 1). Nonetheless, the incidence of LR (p = 0.91) and SR (p = 0.33) was similar between CC and RC (reflecting the limitation of χ2 test in determining oncological outcomes). Local Recurrence (LVI+ vs LVI– Groups)
Twenty-seven patients developed LR in CC during the follow-up period. Local recurrence was defined as tumor recurrence at the anastomotic site. Twenty of 27 (74%) patients were classified as LVI+. Twenty-four of 27 (88.89%) patients were classified as stage III, and 3 (11.11%) were classified as stage II. The majority of tumors developing LR were high risk (stage III, LVI+). A χ2 test was used to compare the incidence of LR between LVI+ and LVI– groups with respect to CC, RC, and CRC. Lymphovascular invasion positivity was associated with increased incidence of LR in CC (p = 0.002) and CRC (p < 0.001) but not RC (p = 0.13) (Table 2). Regarding CC, 19 (18.8%) of LVI+ (n = 101) patients developed LR compared with 8 (2.89%) in the LVI– group (n = 278) (p = 0.002). Regard-
ing CRC, 21 (14.4%) of LVI+ (n = 145) patients developed LR compared with 9 (2.36%) patients in the LVI– group (n = 382) (p < 0.001). In RC, there was no difference in the incidence of LR between LVI+ (n = 44) and LVI– (n = 104) groups (2, 4.54% vs 1, 0.96%, p = 0.13) (Table 2). Survival distributions of LVI+/LVI– groups were plotted in Kaplan-Meier estimates. The differences in the distribution of survival curves (LR) were compared by using a logrank test. Three graphs were generated representing CC, RC, and CRC. Lymphovascular invasion positivity was associated with adverse LR in CC (p < 0.001) (Fig. 1A) and CRC (p < 0.001) (Fig. 1C) but not RC (p = 0.15) (Fig. 1B). Systemic Recurrence (LVI+ vs LVI– Groups)
A χ2 test was used to compare the incidence of SR between groups (LVI+ vs LVI–) with respect to CC, RC, and CRC. Lymphovascular invasion positivity was associated with an increased incidence of SR in RC (p = 0.002) and CRC (p = 0.02) but not CC (p = 0.35) (Table 3). Regarding RC, 15 (34.1%) of LVI+ patients (n = 44) developed SR compared with 12 (11.5%) patients in the LVI– group (n = 104) (p = 0.002). Similarly, in CRC, 43 (29.3%) LVI+ patients (n = 145) developed SR compared with 76 (19.8%) patients in the LVI– group (n = 382) (p = 0.02). There was no association
Table 2. Patients with established locoregional recurrence Colon cancer
Total LR, n (%)
LVI+
LVI–
101 19 (18.8)
278 8 (2.89)
Rectal cancer p 0.002
Colorectal cancer
LVI+
LVI–
p
LVI+
LVI–
p
44 2 (4.54)
104 1 (0.96)
0.13
145 21 (14.4)
382 9 (2.36)
< 0.001
Patients with established LR were categorized as LVI+ and LVI–. The incidence of LR was compared between LVI+ and LVI– groups by using a χ test. LR was greater in LVI+ groups (in comparison with LVI– groups) in CC and CRC but not RC. CC = colon cancer; CRC = colorectal cancer; LR = locoregional recurrence; LVI+ = lymphovascular invasion positive; LVI– = lymphovascular invasion negative; RC = rectal cancer. 2
Copyright © The American Society of Colon & Rectal Surgeons, Inc. Unauthorized reproduction of this article is prohibited.
550
HOGAN ET AL: COMPARING LVI IN COLON/RECTAL CANCER
A 1.0
Colon cancer
B 1.0
Rectal cancer
C
Colorectal cancer
1.0
0.8
0.8
0.8
0.6
0.6
0.6
0.4
0.4
0.4
0.2
0.2
0.2
p < 0.001
p = 0.15
p < 0.001
5
5
5
D
10 15 20 25 30 Local recurrence Colon cancer
E
10 15 20 25 30 Local recurrence
1.0
1.0
Rectal cancer
F 1.0
0.8
0.8
0.8
0.6
0.6
0.6
0.4
0.4
0.4
0.2
0.2
0.2 p < 0.001
p = 0.28 5 10 15 20 25 Systemic recurrence
30
10 15 20 25 30 Local recurrence Colorectal cancer
p < 0.001
5 10 15 20 25 Systemic recurrence
30
5 10 15 20 25 30 Systemic recurrence
Figure 1. X axis = survival (months), y axis = cumulative survival. A–C, Survival distributions (LR) of lymphovascular invasion positive (LVI+) and negative (LVI–) cases were plotted and compared in Kaplan-Meier estimates (log-rank analysis). Lymphovascular invasion positivity was associated with an increased rate of LR in CC (p < 0.001) and CRC (p < 0.001) but not RC (p = 0.15). D–F, Survival distributions (SR) of LVI+ and LVI– cases were plotted and compared in Kaplan-Meier estimates (log-rank analysis). Lymphovascular invasion positivity was associated with increased rate of SR in RC (p < 0.001) and CRC (p < 0.001) but not CC (p =0.28). CC = colon cancer; CRC = colorectal cancer; LR = locoregional recurrence; RC = rectal cancer; SR = systemic recurrence.
between LVI+ and SR in CC (p = 0.35) (Table 3). The survival distributions of LVI+/LVI– groups were plotted and compared in Kaplan-Meier estimates. Differences in the distribution of survival curves (SR) were further evaluated by using a log-rank test. Three graphs were generated representing RC, CRC, and CC. Lymphovascular invasion positivity was associated with adverse SR in RC (p < 0.001) (Fig. 1E) and CRC (p < 0.001) (Fig. 1F) but not CC (p = 0.28) (Fig. 1D). Multivariate Analysis: Association Between LVI+ and LR
Lymphovascular invasion status and established prognostic factors were incorporated into a multivariate Cox proportional hazards model to evaluate the independent association between LVI+ and recurrence. As expected,
stage III tumors were independent predictors of adverse LR with respect to CC (HR, 2.17; 95% CI, 1.64–3.62; p = 0.02; Table 4), RC (HR, 2.57; 95% CI, 1.01–6.93; p = 0.05; Table 5) and CRC (HR, 2.37; 95% CI, 1.32–3.51; p < 0.001; Table 6). Lymphovascular invasion positivity was an independent predictor of adverse LR in CC (HR, 1.90; 95% CI, 1.10–3.27; p = 0.02) and CRC (HR, 1.79; 95% CI, 1.11–2.90; p = 0.02) but not RC (HR, 1.57; 95% CI, 0.55–4.49; p = 0.40). Multivariate Analysis: Association Between LVI+ and SR
The independent association between LVI+ and SR was evaluated in a multivariate Cox proportional hazards
Table 3. Patients with established systemic recurrence Colon cancer
Total SR, n (%)
LVI+
LVI–
101 28 (27.7)
278 64 (23)
Rectal cancer p 0.35
LVI+
LVI–
44 15 (34.1)
104 12 (11.5)
Colorectal cancer p 0.002
LVI+
LVI–
145 43 (29.3)
382 76 (19.8)
p 0.02
Patients with established SR were categorized LVI+ and LVI–. The incidence of SR was compared between LVI+ and LVI– groups by using a χ test. SR was greater in LVI+ groups (in comparison with LVI– groups) in CRC and RC but not CC. CC = colon cancer; CRC = colorectal cancer; LVI+ = lymphovascular invasion positive; LVI– = lymphovascular invasion negative; RC = rectal cancer; SR = systemic recurrence. 2
Copyright © The American Society of Colon & Rectal Surgeons, Inc. Unauthorized reproduction of this article is prohibited.
551
Diseases of the Colon & Rectum Volume 58: 6 (2015)
Table 4. Analysis of independent association between LVI+ and LR/SR in CC Locoregional recurrence Parameter Age Male Stage III Total LN LVI
Systemic recurrence
p
HR
95% CI
p
HR
95% CI
0.94 0.77 0.02 0.81 0.02
1.00 1.08 2.17 1.00 1.90
0.98–1.02 0.65–1.79 1.64–3.62 0.96–1.03 1.10–3.27
0.53 0.60 0.02 0.03 0.88
0.99 1.12 1.84 0.96 1.04
0.98–1.01 0.74–1.71 1.16–3.09 0.93–1.00 0.64–1.67
Established prognostic factors and LVI were incorporated into a Cox proportional hazards multivariate model to evaluate the independent association between LVI+ and LR/ SR. LVI+ was an independent predictor of adverse LR in CC (p = 0.02) and CRC (p = 0.02) but not RC (p = 0.40). LVI+ was an independent predictor of adverse SR in RC (p = 0.04) but not CC (p = 0.88) and CRC (p = 0.31). CC = colon cancer; CRC = colorectal cancer; LR = locoregional; LN = lymph node; LVI = lymphovascular invasion; LVI+ = lymphovascular invasion positive; LVI– = lymphovascular invasion negative; RC = rectal cancer; SR = systemic recurrence.
model. As expected, stage III tumors were independent predictors of adverse SR in CC (HR, 1.84; 95% CI, 1.16– 3.09; p = 0.02; Table 4), RC (HR, 1.66; 95% CI, 0.69–4.32; p = 0.05; Table 5) and CRC (HR, 1.79; 95% CI, 1.18–2.84; p = 0.02; Table 6). Lymphovascular invasion positivity was not associated with adverse SR in CC (HR, 1.04; 95% CI, 0.64–1.67; p = 0.88) and CRC (HR, 1.24; 95% CI, 0.82– 1.86; p = 0.31) but was an independent predictor of adverse SR in RC (HR, 2.57; 95% CI, 1.04–6.39; p = 0.04) (Table 4). Multivariate Analysis: Association Between LVI+ and DFS/OS
The association between LVI+ and DFS was evaluated in a multivariate Cox proportional hazards model (Table 5). Lymphovascular invasion positivity was an independent predictor of adverse DFS in CC (HR, 1.49; 95% CI, 1.07–2.07; p = 0.02; Table 7), RC (HR, 2.69; 95% CI, 1.36–5.32; p < 0.001; Table 8) and CRC (HR, 1.61; 95% CI, 1.20–2.16; p < 0.001; Table 9). Kaplan-Meier estimates comparing LVI+ and LVI– groups demonstrated adverse DFS in LVI+ patients in CC (p < 0.001), RC (p < 0.001), and CRC (p < 0.001) (Fig. 2A–C). In the context of OS, LVI+ was not an independent predictor of adverse outcomes (multivariate analysis) in CC (p = 0.11, Table 7) and CRC (p = 0.20, Table 9). Lymphovascular invasion positivity was an independent predictor of adverse
OS in RC (HR, 2.32; 95% CI, 0.75–7.13; p = 0.04; Table 8). Kaplan-Meier estimates comparing survival distributions (OS) of LVI+ and LVI– groups demonstrated adverse survival in CC (p < 0.001), CRC (p < 0.001), and RC (p = 0.05) (Fig. 2D–2F).
DISCUSSION Colon and rectal adenocarcinomas differ at a multitude of levels and are increasingly regarded as distinct biological entities. In particular, the differences in surgical technique; adjuvant, neoadjuvant therapies; and survival rates may obscure the true relationship between outcome and prognostic indicator when colon and rectal adenocarcinomas are combined as a single cohort in data set analysis. This study evaluated differences in relationships between LVI+ and oncological end points in colon (CC) and rectal adenocarcinomas (RC) separately. Results regarding CC and RC are presented adjacent to the CRC group to highlight the discrepancies between combined data sets and CC/RC cohorts (which may have significant clinical consequences). Univariate and multivariate analyses demonstrated that the relationship between LVI+ and outcome in CC differed from LVI+ and outcome in RC. Previous studies evaluating outcomes in CRC data sets may be flawed with 1 group skewing the results of the other.
Table 5. Analysis of independent association between LVI+ and LR/SR in RC Locoregional recurrence Parameter Age Male Stage III Total LN LVI
Systemic recurrence
p
HR
95% CI
p
HR
95% CI
0.37 0.88 0.05 0.09 0.40
1.02 1.09 2.57 0.92 1.57
0.98–1.07 0.37–3.21 1.01–6.93 0.83–1.01 0.55–4.49
0.49 0.42 0.05 0.76 0.04
0.99 0.68 1.66 0.99 2.57
0.95–1.03 0.24–1.80 0.69–4.32 0.91–1.07 1.04–6.39
Established prognostic factors and LVI were incorporated into a Cox proportional hazards multivariate model to evaluate the independent association between LVI+ and LR/ SR. LVI+ was an independent predictor of adverse LR in CC (p = 0.02) and CRC (p = 0.02) but not RC (p = 0.40). LVI+ was an independent predictor of adverse SR in RC (p = 0.04) but not CC (p = 0.88) and CRC (p = 0.31). CC = colon cancer; CRC = colorectal cancer; LR = locoregional; LN = lymph node; LVI = lymphovascular invasion; LVI+ = lymphovascular invasion positive; LVI– = lymphovascular invasion negative; RC = rectal cancer; SR = systemic recurrence.
Copyright © The American Society of Colon & Rectal Surgeons, Inc. Unauthorized reproduction of this article is prohibited.
552
HOGAN ET AL: COMPARING LVI IN COLON/RECTAL CANCER
Table 6. Analysis of independent association between LVI+ and LR/ SR in CRC Locoregional recurrence Parameter Age Male Stage III Total LN LVI
Systemic recurrence
p
HR
95% CI
p
HR
0.74 0.97 < 0.001 0.33 0.02
1.00 1.01 2.37 0.98 1.79
0.98–1.02 0.65–1.57 1.32–3.51 0.95–1.02 1.11–2.90
0.32 0.61 0.02 0.03 0.31
0.99 1.10 1.79 0.96 1.24
95% CI 0.98–1.01 0.76–1.59 1.18–2.84 0.93–1.00 0.82–1.86
Established prognostic factors and LVI were incorporated into a Cox proportional hazards multivariate model to evaluate the independent association between LVI+ and LR/SR. LVI+ was an independent predictor of adverse LR in CC (p = 0.02) and CRC (p = 0.02) but not RC (p = 0.40). LVI+ was an independent predictor of adverse SR in RC (p = 0.04) but not CC (p = 0.88) and CRC (p = 0.31). CC = colon cancer; CRC = colorectal cancer; LR = locoregional; LN = lymph node; LVI = lymphovascular invasion; LVI+ = lymphovascular invasion positive; LVI– = lymphovascular invasion negative; RC = rectal cancer; SR = systemic recurrence.
Kaplan-Meier estimates demonstrated an adverse association between LVI+ and both OS and DFS in all 3 cohorts (CC, RC, and CRC). On multivariate analysis however, LVI+ was associated with adverse OS in RC only (the association was not significant in CC). Conversely, LVI+ was associated with adverse DFS in CC, RC, and CRC. The pattern of disease recurrence differed between CC and RC. Lymphovascular invasion positivity increased LR in CC and SR in RC. These findings persisted in Kaplan-Meier estimates (compared by using log-rank analysis) and univariate and multivariate analysis. Surprisingly, the incidence of LR was greater in CC than in RC. This did not approach statistical significance and may reflect the greater proportion of stage I tumors in RC (statistically significant). The analysis of LVI in patients receiving adjuvant and neoadjuvant therapies were not undertaken in the current study. However, preliminary work suggests that LVI retains its prognostic properties in CC and RC (with respect to OS and DFS). Substantial evidence exists validating the association between LVI+ and adverse survival (disease-free and overall) in combined CC and RC data sets. Disparity exists between the results of the current study and published literature for a number of reasons, such as 1) differences between outcome variables evaluated (OS, DFS, LR, and/or SR), 2) depth of analysis (majority of
studies do not analyze SR and LR separately), and 3) the combination of CC and RC, which differ at a multitude of levels (proteomic, genomic, epidemiological, and oncological outcome). Few studies overcome all these potential limitations. Huh and colleagues27 performed a univariate/multivariate survival analysis evaluating LVI and oncological outcomes in CC and RC. Lymphovascular invasion was associated with adverse DFS in both CC and RC (similar to the findings of the current study). However, methods differed substantially with respect to the depth of analysis. Huh did not examine the role of LVI in LR and SR separately, and, as such, the concordance between Huh and the current results is unclear.27 Patel et al28 reported no association between LVI and LR in early-stage RC validating the results discussed in this article. Systemic recurrence was not evaluated, nor did the authors compare outcomes between CC and RC. Bhangu et al29 evaluated LVI in SR and LR in RC alone. On multivariate analysis of T4 rectal tumors, vascular invasion was associated with adverse OS and DFS. When LR and SR were analyzed separately, the authors reported no association between LVI and LR but an adverse association between LVI and SR (in RC). This is one of few studies evaluating both LR and SR in the context of CC or RC alone with the use of patients from a single institute.29 Similar to the results of this article, LVI has
Table 7. Evaluation of the independent association between LVI+ and DFS/OS in CC Disease-free-survival Parameter Age Male Stage III Total LN LVI
Overall survival
P value
HR
95% CI
P value
HR
0.30 0.79 0.05 0.15 0.02
1.01 1.04 1.49 0.98 1.49
0.98–1.02 0.77–1.40 1.12–2.08 0.96–1.01 1.07–2.07
0.40 0.92 0.91 0.01 0.11
1.01 1.02 1.04 0.96 1.35
95% CI 0.99–1.02 0.73–1.42 0.63–1.61 0.93–0.99 0.93–1.95
Established prognostic factors and LVI were incorporated into a Cox proportional hazards multivariate model to evaluate the independent association between LVI+ and DFS/OS. LVI+ was an independent predictor of adverse DFS in CC (p = 0.02), RC (p < 0.001) (Table 8), and CRC (p < 0.001) (Table 9). LVI+ was associated with adverse OS in RC (p = 0.04) but not CC (p = 0.11) and CRC (p = 0.20). CC = colon cancer; CRC = colorectal cancer; DFS = disease-free survival; LN = lymph node; LVI = lymphovascular invasion; LVI+ = lymphovascular invasion positive; OS = overall survival; RC = rectal cancer.
Copyright © The American Society of Colon & Rectal Surgeons, Inc. Unauthorized reproduction of this article is prohibited.
553
Diseases of the Colon & Rectum Volume 58: 6 (2015)
Table 8. Evaluation of the independent association between LVI+ and DFS/OS in RC Disease-free-survival Parameter Age Male Stage III Total LN LVI
Overall survival
p
HR
95% CI
p
HR
0.21 0.24 0.17 0.67 < 0.001
1.02 0.63 1.64 0.99 2.69
0.99–1.05 0.29–1.36 0.69–3.32 0.93–1.05 1.36–5.32
0.15 0.11 0.77 0.70 0.04
1.04 2.82 1.25 0.98 2.32
95% CI 0.99–1.10 0.78–10.11 0.45–4.72 0.89–1.09 0.75–7.13
Established prognostic factors and LVI were incorporated into a Cox proportional hazards multivariate model to evaluate the independent association between LVI+ and DFS/OS. LVI+ was an independent predictor of adverse DFS in CC (p = 0.02) (Table 7), RC (p < 0.001), and CRC (p < 0.001) (Table 9). LVI+ was associated with adverse OS in RC (p = 0.04) but not CC (p = 0.11) and CRC (p = 0.20). CC = colon cancer; CRC = colorectal cancer; DFS = disease-free survival; LN = lymph node; LVI = lymphovascular invasion; LVI+ = lymphovascular invasion positive; OS = overall survival; RC = rectal cancer.
unique prognostic properties in the context of RC (adverse SR but not LR). In summary, LVI+ was associated with increased LR in CC and increased SR in RC. This prompts the question as to why LVI+ is associated with LR in CC but not RC. Total mesorectal excision is composed of the removal of tumors and the entire unit of regional spread and is associated with decreased LR. No such standard of "mesocolic excision" exists with respect to CC, and potentially metastatic lymphatics remain in situ adversely affecting LR in CC.30 Conversely, LVI+ was associated with increased SR in RC but not CC. This discrepancy may be explained by the differences in venous drainage of the colon and rectum. The superior, middle, and inferior rectal veins drain the rectum. Middle and inferior rectal veins drain to the inferior vena cava via the iliac vessels. The colon drains via the superior mesenteric vein to the portal vein. In this manner, parts of the rectum drain directly to the systemic circulation, which may explain the increased incidence of SR in LVI+ RC. This may also account for the finding that LVI+ was associated with adverse OS in RC only (on multivariate analysis). Lymphovascular invasion positivity increased LR in CC, and LR is frequently amenable to secondary surgery. Therefore mortality rates and OS remain unchanged. In RC, LVI+ was associated with increased SR, which is less amenable to curative interventions, and so OS is adversely affected.31,32
LIMITATIONS This study has some limitations. 1) It is a retrospective analysis and the RC cohort is relatively small. This increases the potential for type II statistical error. The incidence of LR in RC is low, further increasing the potential for statistical error. A multi-institutional collaborative study or highvolume single-institute study may overcome this issue in future studies. 2) The role of LVI in patients undergoing adjuvant/neoadjuvant therapy is unclear and was not evaluated in the current article. 3) Median lymph node retrieval was less than 12 in CC and RC, raising issues regarding appropriate staging. However, significant disparity exists within the literature regarding the minimum lymph node yield associated with improved survival. 4) Data regarding grade of LVI and the differentiation between extramural/ intramural invasion was not available for analysis. Sample size was sufficient to detect statistical significance, and this is one of a few studies to evaluate the prognostic properties of LVI+ in colon and rectal adenocarcinomas separately.
CONCLUSIONS In conclusion, the pattern of disease recurrence differs between LVI+ CC and LVI+ RC, resulting in discrepancies in the prognostic properties of LVI+ in CC and RC. Future studies assessing the prognostic indicators in CRC should evaluate CC and RC separately.
Table 9. Evaluation of the independent association between LVI+ and DFS/OS in CRC Disease-free-survival Parameter Age Male Stage III Total LN LVI
Overall survival
p
HR
95% CI
p
HR
95% CI
0.20 0.69 0.01 0.10
