Assessment of a Staging System for Sigmoid Colon Cancer Based on Tumor Deposits and Extramural Venous Invasion on Computed Tomography

This prognostic study assesses whether preoperative computed tomographic findings can be used for sigmoid colon cancer prognosis.


Introduction
There is increasing awareness that preoperative identification of high-risk tumors on imaging has improved rectal cancer outcomes. [1][2][3] To date, such a strategy does not exist for colon cancer and may partly explain why rectal cancer outcomes have improved and now have overtaken colon cancer outcomes. [4][5][6][7] Novel treatment options for advanced, nonmetastatic colon cancer, such as neoadjuvant chemotherapy, 8 complete mesocolic excision, or hyperthermic intraperitoneal therapy, 9 may improve outcomes among patients with an otherwise poor prognosis but require accurate preoperative imaging to identify locally advanced tumors. At present, preoperative clinical staging of colon cancer by computed tomography (CT) is based on the TNM system. 10 Nodal disease, which is the traditional histopathologic indication for adjuvant chemotherapy, is no longer recommended for clinical staging 11 because it cannot be reliably identified by CT, with specificity of 55% to 67% on meta-analysis. [12][13][14] More recently, DNA analysis of primary tumor, lymph nodes, and metastases in patients with colorectal cancer has found that most distant metastases had an origin different from lymph nodes. 15 Tumor stage has traditionally not been seen to be relevant because it was not prognostic. In 2007, advanced T3 (>5-mm depth of spread beyond the muscularis propria) and T4 categories were identified as criteria for colon cancer with poor prognosis on CT. 16 Patients with advanced T3 and T4 tumors identified on CT had worse recurrence-free survival at 3 years than patients with T1 and T2 tumors (53%) and those with early T3 tumors (87%). The use of preoperative treatment for all patients with T3 tumors was recommended in a recent colon cancer trial, 8,17 but the preoperative treatment failed to reduce recurrence; this finding may have been associated with inclusion of a group at low risk of recurrence (early T3 tumors with <5-mm spread).
With the emergence of novel and varied treatment strategies, an accurate preoperative staging system permits an opportunity to stratify treatment as well as surveillance. Risk factors for recurrence that were not evaluated by TNM are now standardly reported on CT report proformas for colon cancer, 18 including advanced T3 substage, 16,19 localized peritoneal infiltration, 20,21 extramural venous invasion (EMVI), 19,22,23 and discontinuous tumor deposits. 24,25 In this study, we aimed to investigate whether these known disease features on CT are associated with oncologic outcome of sigmoid colon cancer and whether they could be used to optimize preoperative staging systems.

Data Sources and Baseline Variables
This prognostic study used retrospective data from patients identified in a prospectively maintained institutional database at the Royal Marsden Hospital, London, England, that included patient data from national and international tertiary referral centers in addition to a local referral network of 6 Patients were eligible for inclusion if they had nonmetastatic adenocarcinoma arising in the sigmoid colon on CT scan and underwent colonic resection. The mesorectal-mesocolic junction as recognized by the sigmoid take-off 28 was used as the distal landmark to define sigmoid tumors.
Exclusion criteria included synchronous tumors, metastatic disease, early tumors removed with endoscopic treatment only, or disease that was later found not to be colorectal cancer. Patients who received palliative treatment without resection of their primary tumor and patients who died within 30 days of undergoing a surgical procedure were also excluded. If no follow-up data were available for international or national patients who returned to their referral hospital after undergoing a surgical procedure, they were excluded. Data were extracted on basic demographic characteristics, such as age, sex, emergency presentation, and preoperative or postoperative treatment (ie, radiotherapy or chemotherapy).

Radiologic Analysis
Computed tomography was performed at National Health Service and international institutions on multidetector CT scanners from a range of manufacturers using standard CT abdomen-pelvis protocols with 1-mm section intervals in more than 90% of cases. Images were viewed with multiplanar reconstruction when possible; this was a limitation in fewer than 10% of cases. Data on CT reports from the local hospital were extracted when reported by gastrointestinal radiologists on standardized CT reporting proformas for colon cancer developed by the Royal College of Radiologists. 18 Results were otherwise rereported by 2 gastrointestinal radiologists who had more than 5 years of experience and who were blinded to the clinical information. The CT data were recorded on T stage (including T3 substaging by depth of spread), nodal disease, discontinuous tumor deposits, EMVI, and peritoneal infiltration localized to the tumor site (ie, not disseminated).
Two CT-staging classifications for poor and good prognosis groups based on T stage have previously been described. 8,16,19,20,29 The CT-TNM system uses conventional TNM staging and classifies all T3 and T4 tumors to a poor prognosis group and all T1 and T2 tumors to a good prognosis group. The CT-TNM system was used to select patients with T3 or T4 colon cancer for preoperative therapy by a prospective multicenter trial. 8 The CT-T3 substage system combines high-risk T3 substage tumors (T3c-T3d) and T4 tumors to create a poor prognosis group, whereas good prognosis T3 substage tumors (T3a-T3b) are grouped with T1 and T2 tumors. The CT-T3 substage system has been described in previous reports, 16,19,20,29 with high sensitivity and specificity for accurate substaging on meta-analysis. 12

Pathologic Analysis and Clinical End Points
Pathologic variables of T stage, nodal disease, tumor deposits, differentiation grade, EMVI, peritoneal disease, and circumferential resection margin status were recorded on the standard data set issued by the Royal College of Pathologists and extracted. 30 Data were extracted on disease recurrence and death from clinical medical notes. Recurrence site was confirmed on imaging and was reclassified as local recurrence or distant recurrence.

Statistical Analysis
The primary end point was to investigate the association between known preoperative prognostic factors on CT and disease recurrence: local recurrence, distant recurrence, and overall (local or distant recurrence). The secondary end point was to compare the prognostic accuracy of different CT staging systems based on TNM, T3 substage, and a CT staging system based on significant risk factors for disease-free survival on multivariate analysis. All included patients had complete data for radiologic staging. Local institutions were contacted to seek missing data particularly on patient follow-up. Missing data for pathologic variables were quantified. Patients who were lost to follow-up had the duration of follow-up quantified as the period between the date of surgical procedure and date of the last clinic appointment or radiologic scan. Survival analysis was defined for disease-free survival as time to event in months from the date of surgical procedure to the date of local  (Figure 1). However, the CT-tumor deposits and EMVI classification system based on CT-detected tumor deposits and EMVI best identified a poor prognosis group (HR, 2.81; 95% CI, 1.95-4.05) (Figure 1).

Discussion
The TNM stage as defined in pathologic studies has been applied for CT staging of colon cancer. Our data suggest that the TNM category does not translate to CT staging and does not provide prognostically accurate information before surgery. In contrast to the TNM system, CT-T3 substage enabled a poor prognosis group to be identified. On multivariate analysis, CT evaluation of lymph for patients at risk of cancer recurrence, identification of a poor prognosis group through a better CT staging system is now necessary to match the improvements seen in the treatment of rectal   cancer. 1-3 Advanced T3 stage of colonic cancer on histopathologic analysis has been shown to be an adverse prognostic factor, 31 but tumor stage-for-stage matching on CT imaging and pathologic specimens correlated poorly on meta-analysis. 12,13 However, early vs advanced T3 tumor stage can be distinguished on CT with better sensitivity (87%-90%) and specificity (69%-78%) 16,19 and acceptable interreporter reliability. 20,21 As with pathologic results, 31 T3 substaging on CT is prognostic for recurrence. 16,19,22 Inclusion of all T3 sigmoid colon tumors into a poor prognosis group failed as a staging system because it did not stratify patients at risk of disease recurrence. This was explained by the equivalent recurrence rates of CT-staged T3a-T3b and T1-T2 tumors, which has not been shown previously to our knowledge. This result may explain the recent results of the FOXTROT study 17 ; preoperative treatment for patients with cT3a-cT3b-staged disease would expose these patients to toxic effects for negligible benefit because their risk of recurrence is already low.

JAMA Network Open | Imaging
For colon cancer, T3 substage of more than 5-mm extramural spread should therefore be included on all CT reports. Given its high risk for recurrence, T3 substage poor prognosis disease should be investigated as an indication for neoadjuvant therapy.  19,22,23 In this study, tumor deposits could be differentiated from lymph nodes on CT ( Figure 2B) and were associated with a significantly poorer prognosis on multivariate analysis.

CT-T3 Substage and CT-Tumor Deposits and EMVI: Step-by-Step Guide
Identification of features of advanced colonic disease on CT requires multiplanar reconstruction on CT multidetector scanners. As shown in this study, diagnostic accuracy can be maintained across a range of different CT manufacturers using local abdomen-pelvis protocols.

Local Tumor Stage
Tumor spread beyond the muscularis propria can be measured on multiplanar reconstruction of the tumor. 16,19,20,23 The CT diagnosis of a stage T3 lesion is based on the presence of tumor soft tissue extending into the pericolonic fat with a broad-based bulging or nodular configuration in continuity ( Figure 3A).

Extramural Venous Invasion
Tumor that invades veins beyond the muscularis propria can be identified on imaging. 19,23 On CT, the draining colonic veins can be identified as tubular structures in continuity on adjacent sections on multiplanar reconstruction. Linear or serpiginous extension of the primary tumor into the pericolonic veins can be seen because the tumor infiltrates and expands the vessel, creating an irregular contour ( Figure 3B). Because this can be a subtle finding in smaller veins, it is likely that only larger-vessel EMVI is detected on CT.

Discontinuous Tumor Deposits
Tumor deposits are deposits of cancer cells that are discontinuous with the primary tumor and not associated with a lymph node. They can be distinguished from smoothly enlarged round lymph nodes on CT because they are more likely to have an irregular contour and mixed signal density.
Although both tumor deposits and lymph nodes may appear to have an irregular contour, tumor deposits interrupt the course of a vein (best seen on multiplanar reconstruction), whereas lymph nodes lie alongside a vein ( Figure 3C).

Limitations
This study has limitations. It is possible that other centers may not be able to reproduce the level of accuracy in detection of adverse features such as advanced T3c substage, EMVI, or tumor deposits.
Gastrointestinal radiologists can identify CT features that are not routinely reported. [19][20][21][22] The prognostic importance of these features may make the case for greater specialist reporting of colon cancer. Although workshop training can improve identification of high-risk colonic tumors on CT, 8 regular interface between specialist radiologists with the colorectal multidisciplinary team may be more important to improve an understanding of the disease and its outcomes.
Other limitations of this retrospective study were the lack of follow-up data for patients returning to their homes abroad or elsewhere in the United Kingdom and who could not be followed up. Omission of relevant covariates (eg, comorbidity) that were unavailable may have led to bias.
There may also be undiscovered predictor data variables (eg, tumor biomarkers) that were confounders.  Further research should be aimed at establishing the interreader reliability of the radiologist for adverse CT prognostic features. One option is to test our T3 substage and tumor deposits and EMVI classifications against TNM in other test cohorts (eg, using CT for patients with colon cancer in the FOXTROT study, 17 which is currently underway). This would also establish whether the adverse CT features for sigmoid colon cancer in this study carry the same prognostic weights for other sites of colon cancer.

Conclusions
In this study, advanced T3 substage disease in colon cancer was identified on CT and was associated with a significantly adverse prognosis for disease-free survival. Tumor deposits and EMVI were independent adverse prognostic features visible on CT that were most strongly associated with worse disease-free survival. Lymph node assessment of colon cancer on CT should not be relied on since it had no prognostic value.