A Comprehensive Study of Extramural Venous Invasion in Colorectal Cancer

Colorectal cancer is a common malignancy and a leading cause of cancer related death. Cancer staging following resection is key to determining any adjuvant therapy in those patients with high risk disease. In colorectal cancer, tumour stage and lymph node stage are the main pathological factors which have been considered to influence outcome. Increasing emphasis is now being placed on other factors, especially the presence of extramural venous invasion (EMVI). It is important to understand the relationship of EMVI with other pathological factors and to confirm that in an individual centre that EMVI is being detected at an appropriate rate and is of prognostic significance. This comprehensive study assesses the reporting and prognostic significance of EMVI in a single centre, using prospectively collected data from histopathology reports of a cohort of 2405 patients who underwent surgery for colorectal cancer over a nine year period. Overall, EMVI was reported in 27.9% of colorectal cancer excision specimens. In tumours (n = 1928) that had not received neoadjuvant therapy, the presence of EMVI varied significantly depending on tumour site (χ2 = 12.03, p<0.005), tumour stage (χ2 = 268.188, p<0.001), lymph node stage (χ2 = 294.368, p<0.001) and Dukes’ stage (χ2 = 253.753, p<0.001). Multivariate analysis confirmed EMVI as a significant independent prognostic indicator (p<0.001). In conclusion, the presence of EMVI as an independent prognostic indicator is shown and is related to other pathological and prognostic factors. This study emphasises the requirement for the accurate identification of EMVI in colorectal cancer excision specimens and also understanding the relationship of EMVI with other prognostic factors.


Introduction
Colorectal carcinoma is the most common malignancy of the gastrointestinal tract and is a leading cause of cancer related deaths [1]. Prognosis is predicted by the extent of tumour spread locally, the presence of lymphatic system involvement and/or metastasis to other organs which are encompassed in TNM or Dukes' staging. Outwith these classical staging criteria, the presence of tumour cells within veins outside the bowel wall (extramural venous invasion) is an important predictor of tumour recurrence or metastasis. Consequently extramural venous invasion (EMVI) is an independent indicator of poor prognosis in colorectal carcinoma [2][3][4][5][6][7][8][9][10][11]. It should be acknowledged that in addition to that of extramural venous channels, there is an emerging role for intramural venous channels located in the submucosa or muscularis propria (intramural venous invasion, MVI) [5,7]. However, the prognostic significance of MVI alone is still unclear [4,12].
Whilst not directly affecting overall tumour stage, the presence of EMVI does confer high risk disease status [13,14]. Indeed, about 25-30% of patients with lymph node negative disease die from recurrent or metastatic disease, emphasising the role of vascular spread in colorectal cancer dissemination [15][16][17].
Therefore, the United Kingdom Royal College of Pathologists (RCPath) has proposed a minimum rate of detection of EMVI in colorectal cancer excision specimens. This study reports on data collected from 2005 to 2013, during which time the RCPath dataset for colorectal cancer (2 nd edition) published in 2007 recommended a frequency of detection of EMVI of 20% [18].
EMVI was defined by Talbot and colleagues as the presence of tumour within an endothelium-lined space that is either surrounded by a rim of muscle or contains red blood cells [19]. Additionally, the so called orphan artery sign is suggestive; where a rounded or elongated tumour profile that is not in direct continuity with the advancing tumour margin is identified adjacent to an artery, especially when no accompanying vein can be seen [3]. Whilst it is possible to detect EMVI on a haematoxylin and eosin stained slide, the application of ancillary techniques such as elastin stains or immunohistochemistry have been shown to significantly increase detection rates [2,7,8,16,[20][21][22][23][24].
This comprehensive study assesses the trend in the evaluation of EMVI in a large cohort of patients with colorectal cancer over a nine year period in a single centre, focusing on its prognostic significance with respect to other colorectal cancer staging criteria and potential confounding factors.

Study population
This study included 2405 patients who had their colorectal cancer pathology resection specimens reported by the Department of Pathology, Aberdeen Royal Infirmary from 2005-2013. The Department of Pathology at Aberdeen Royal Infirmary is a regional pathology centre receiving pathology specimens from four acute hospitals over three health authorities. Aberdeen Royal Infirmary (an academic teaching hospital centre) and Dr Gray's Hospital, Elgin (a district general hospital) are both in NHS Grampian, while Balfour Hospital, Kirkwall (NHS Orkney) and Gilbert Bain Hospital, Lerwick (NHS Shetland) are both remote and rural hospitals. Approximately 75% of the colorectal cancer surgery is performed at Aberdeen Royal Infirmary. Relevant pathological data was extracted at the time of the weekly colorectal cancer multidisciplinary team meeting from the pathology reports of the resected colorectal cancer excision specimens and a database constructed.
The database was compiled using prospectively collected histopathological data from colorectal cancers resected between 2005 and 2013. The information recorded in this database includes age, gender, year of operation, administration of neoadjuvant therapy, whether the tumour was screen detected, tumour site, tumour differentiation, tumour (T) stage, the presence of EMVI, total number of lymph nodes examined, number of lymph nodes involved by metastatic tumour, lymph node (N) stage and Dukes' stage. Information for each parameter was available for every patient. Additionally, survival data (all-cause mortality) was available for a subset of 1004 patients (number of deaths = 243). The histopathology of all the cases in this database were reported according to the criteria set out in the RCPath dataset for the reporting of colorectal cancer excision specimens (2 nd edition 2007-14) which incorporates guidance from TNM5 and all cases were also subject to multi-disciplinary review [18]. Throughout this period of time NHS Grampian has been a centre for the NHS Scotland bowel screening programme (2000-2006, pilot centre for evaluation of programme, 2007-part of national programme following its implementation throughout Scotland).

Assessment of EMVI
The presence of tumour within venous structures beyond the bowel wall was assessed on haematoxylin and eosin stained sections of tumour, with the additional use of elastic haematoxylin and eosin staining in cases where EMVI was suspected but not clearly identified on microscopic examination of the haematoxylin and eosin stained sections (Fig 1).

Statistics
The database was compiled in Excel 2007 and then imported into IBM SPSS version 21 for Windows 7 TM (IBM, Portsmouth, UK) to perform data analysis. Statistical comparisons between groups were performed using chi-square (χ 2 ) test. Survival curves were prepared using the Kaplan-Meier method with log-rank (Mantel-Cox) analysis. Multivariate analysis with Cox regression (proportional hazard analysis) was also performed.

Ethics
The project was carried out with ethics approval (ref. no. 08/S0801/81) from the North of Scotland research ethics committee. The research ethics committee did not require written participant consent and data was anonymised prior to analysis.

Results
The clinico-pathological parameters collected for the dataset are summarised in Table 1. In this study, EMVI was reported in 27.9% of all colorectal cancer cases (n = 2405). EMVI was reported in 31.4% of cases where no neoadjuvant therapy had been received (n = 1928) compared with 13.4% in cases receiving neoadjuvant therapy (n = 477). Cases were subsequently grouped for analysis into those that had received neoadjuvant therapy or those that had not and cases that were screen detected or were not. Further subdivision into stage (primary tumour stage, lymph node stage or Dukes' stage) and site of primary tumour was carried out for analysis. The relationship of EMVI with tumour site Tumours were grouped into the categories of proximal (appendix, caecum, ascending colon, hepatic flexure and transverse colon tumours), distal (splenic flexure, descending colon and sigmoid colon tumours) and rectum depending on their anatomical location. When all cases of colorectal cancer were considered (n = 2405), the frequency of EMVI differed significantly depending on the primary tumour site. EMVI was reported in 35.0% of proximal, 29.0% of distal and 17.0% of rectum cases (χ 2 = 66.83, p<0.001) (Fig 2).
In cases that had not received neoadjuvant therapy (n = 1928), the frequency of EMVI differed significantly depending on site of primary tumour. EMVI was reported in 35.0% of proximal, 28.0% of distal and 26.5% of rectum cases (χ 2 = 12.03, p<0.005). For cases that had received neoadjuvant therapy (n = 477), the frequency of EMVI differed significantly depending on site of primary tumour. EMVI was reported in 25.0% of proximal, 50.0% of distal and 11.8% of rectum cases (χ 2 = 20.82, p<0.001).
When cases that had not been screen detected were analysed (n = 2077) the frequency of EMVI differed significantly depending on site of primary tumour. EMVI was reported in 36.6% of proximal, 29.3% of distal and 15.2% of rectum cases (χ 2 = 77.97, p<0.001). Whereas, in screen detected cases (n = 328) EMVI reporting did not differ significantly between sites of primary tumour. EMVI was reported in 21.6% of proximal, 24.4% of distal and 24.7% of rectum cases (χ 2 = 0.373, p = 0.830).

The relationship of EMVI with tumour stage
The rate of EMVI also varied significantly depending on tumour (T) stage of primary tumour when all cases were considered. EMVI was reported in 1.8% of T0 (complete pathological response of primary tumour to neoadjuvant therapy), 0.7% of T1, 3.4% of T2, 27.6% of T3 and 56.9% of T4 tumours (χ 2 = 390.212, p<0.001) (Fig 3).
When cases that had not been screen detected were analysed, the presence of EMVI differed significantly depending on T stage of primary tumour. EMVI was reported in 2.1% of T0, 0% of T1, 2.6% of T2, 27.1% of T3 and 57.3% of T4 cases (χ 2 = 344.376, p<0.001). There was also a significant difference in the frequency of EMVI depending on T stage of primary tumour in screen detected cases. EMVI was reported in 0% of T0, 2.6% of T1, 6.2% of T2, 31.1% of T3 and 51.6% of T4 cases (χ 2 = 43.866, p<0.001).

The relationship of EMVI with lymph node stage
When all cases were considered, the presence of EMVI differed significantly depending on lymph node (N) stage of primary tumour. EMVI was reported in 14.9% of N0, 35.5% of N1 and 65.3% of N2, cases (χ 2 = 392.878, p<0.001) (Fig 4).
In cases that had not received neoadjuvant therapy, there was a significant difference in EMVI rate depending on N stage, with EMVI present in 17.5% of N0, 37.2% of N1 and 66.6% of N2 tumours respectively (χ 2 = 294.368, p<0.001). When cases that had received neoadjuvant therapy are considered, there is also a significant difference in EMVI rate depending on N stage. EMVI was present in 7.2% of N0, 24.4% of N1 and 52.9% of N2 tumours respectively (χ 2 = 66.222, p<0.001).
Cases that were not screen detected showed a significant difference in EMVI rate depending on N stage, with EMVI present in 15.1% of N0, 35.6% of N1 and 66.4% of N2 tumours respectively (χ 2 = 351.707, p<0.001). A significant difference in the frequency of EMVI depending on N stage was also seen in cases that had been screen detected, with EMVI present in 13.4% of N0, 34.9% of N1 and 55.6% of N2 cases respectively (χ 2 = 38.513, p<0.001).

The relationship of EMVI with Dukes' stage
The frequency of EMVI reporting in different Dukes' stage tumours was also considered. When all cases were analysed, EMVI was reported in 1.1% of Dukes' A cases, 21.6% of Dukes' B cases and 46.8% of Dukes' C cases. (χ 2 = 365.373, p<0.001) (Fig 5). For

Multivariate analysis
Multi-variate analysis showed that EMVI was a highly significant independent prognostic factor (p<0.001) when other tumour dependent factors including tumour stage, lymph node stage, Dukes stage, tumour site and degree of tumour differentiation as well as patient specific factors of age and gender are considered (Tables 2 and 3). EMVI was not a significant prognostic cofactor in screen detected cases but was in nonscreen detected cases (Tables 4 and 5).

Discussion
It is important for individual centres reporting colorectal excision specimen to demonstrate an appropriate frequency of detection of pathological prognostic factors taking account of confounding factors e.g. proportion of patients receiving neoadjuvant therapy and proportion of cases identified through a bowel cancer screening programme. Individual centres should also demonstrate that the pathological factors that are assumed to be prognostically significant are indeed prognostically significant in their population. This large study from a single centre  utilises prospectively collected data to comprehensively assess the evaluation of EMVI in colorectal cancer resection specimens over a nine year period and to consider potential confounding factors including neoadjuvant therapy and the influence of bowel cancer screening. It uses a prospectively collected dataset to examine in detail the significance of the reporting of EMVI on mortality in this population. All of the data derives from a single histopathology department, thereby reducing data collection and reporting bias.
There is well documented variation between centres in the demonstration of EMVI [1,5,7,12,16,[19][20][21][22][23][24][25][26][27]. Despite which, EMVI is recognised as an important prognostic feature [12,16,22,[28][29][30]. As such, reporting of the presence or absence of EMVI in colorectal cancer excision specimens is recommended by professional organisations including the Royal College of Pathologists (UK) and the College of American Pathologists [3,9]. The consequence of accurate detection of EMVI is that patients with lymph node negative colorectal cancer (stage II/ Dukes' B disease) but with EMVI and/or other adverse prognostic features (including tumour perforation, serosal involvement, incomplete tumour resection) may benefit from and should be considered for adjuvant chemotherapy [14,31,32]. The presence of EMVI has become prognostically more relevant following the introduction in the UK of bowel cancer screening programmes, as this has resulted in the more frequent resection of lymph node negative tumours.
This study shows that when all cases are analysed, detection of EMVI exceeds the standard set by the Royal College of Pathologists over the same time period [18]. In addition, exclusion of cases in which preoperative neoadjuvant therapy had been given and which were potentially down staged, resulted in both the 2 nd edition RCPath standard (20%) as well as the revised 3 rd edition (2014) RCPath standard (30%) being surpassed [3,18]. Albeit the data included in this study is prior to the publication of the third edition of the RCPath colorectal cancer dataset [3].  Table 2. The relationship between clinico-pathological characteristics and overall survival, when tumour stage and lymph node status are covariates and primary tumour site is grouped as proximal, distal or rectum. As might be expected, EMVI is more common in higher stage (T-, N-and Dukes' stage) tumours and this study has demonstrated a stage dependent significant difference irrespective of whether the cases were screen-detected or not or if there had been neoadjuvant therapy or not. It has also shown the frequency of the identification of EMVI varied significantly depending on site of primary tumour when all cases, non-screen-detected cases, cases that had received neoadjuvant therapy and also those cases that had not. There was no significant tumour site dependent difference in the frequency of EMVI reporting in colorectal cancer cases that were detected by the bowel screening programme. It is not clear if that represent the smaller number of bowel cancer screening detected cases or reflects a difference in the biology between symptomatic colorectal cancer and screen detected cancers. Whilst the frequency of EMVI appears closely related to tumour stage, multivariate analysis has shown that EMVI is an independent predictor of poor prognosis, as are patient age and tumour stage. This study allowed assessment of all-cause mortality in a subgroup of patients. Overall, EMVI was associated with statistically significantly reduced survival. Only in screendetected cases (n = 125) and in Dukes' B cases following neoadjuvant therapy (n = 71) was EMVI not associated with a significant reduction in survival. However, this lack of significance may reflect the limited amount of survival data available in those groups.
EMVI is clearly an important parameter with regards to patient prognosis and the postoperative management of colorectal cancer. Adequate reporting is therefore paramount and may be optimised with the use of ancillary techniques such as elastic haematoxylin and eosin staining [2,7,8,16,[20][21][22][23]. The data presented supports the need for accurate assessment and reporting of EMVI along with other prognostic factors such as lymph node yield, lymph node ratio and relevant biomarkers to facilitate the decision making process of colorectal cancer treatment [33][34][35][36][37].