High-risk features in potentially resectable colon cancer: a prospective MDCT-pathology agreement study.

Purpose: Neoadjuvant chemotherapy in potentially resectable high-risk Stage II and Stage III colon cancer has demonstrated promising results in the PRODIGE 22-ECKINOXE Phase II trial. Identification of adverse morphologic features, namely T3 with >5 mm extramural extension/T4 stages and/or N2, is fundamental and requires accurate noninvasive imaging. Our aim was to assess the value of optimized preoperative MDCT to stratify potentially resectable colon cancer patients for neoadjuvant therapy.

Methods: this is an observational prospective cross-sectional radiologic-pathologic agreement study. All patients with colon cancer referred to our Institution's Radiology department for preoperative MDCT staging between 01-10-2013 and 11-02-2015 underwent independent reading based on axial and multiplanar reconstruction images by 3 radiologists with 3, 6, and 20 years of experience in gastrointestinal radiology. T stage, extramural extension if T3 (≤5 mm or >5 mm), and N stage were recorded. Surgical specimens subsequently obtained underwent micro-pathologic analysis by a gastrointestinal pathologist with 9 years of experience in gastrointestinal pathology. Main outcome measures were sensitivity, specificity, PPV, NPV, AUROC, diagnostic accuracy, and interobserver agreement of optimized MDCT, and pathologic analysis of the surgical specimen considered the reference standard.

Results: 74 patients [43 males; median age 73 (45-89)] were eligible. MDCT sensitivity, specificity, PPV, NPV, AUROC, and diagnostic accuracy ranged between 42.9-76.2, 75.5-90.6, 55.2-76.2, 80.0-90.6, 0.67-0.83 and 0.76-0.86%, respectively, for the identification of T3 > 5 mm/T4 disease, with moderate interobserver agreement (0.49); and 8.3-33.3, 93.5-98.4, 20-66.7, 84.1-88.2, 0.51-0.65 and 0.80-0.86%, respectively, for the identification of N2 disease, with absent interobserver agreement (0.10).

Conclusions: Specificity of MDCT in the stratification of patients for neoadjuvant therapy may be high enough to prevent overtreatment. However, it may lead to undertreatment in a meaningful proportion of patients. Observer performance may benefit from targeted training programs, given the variability and observer dependence of the results. Limitations include 4-slice MDCT equipment, time to surgery and lack of long-term outcome information based on imaging parameters per se.