570 Background: The significance of vascular emboli (VE) in stage III colorectal cancer (CRC), the mechanism of their formation and their therapy strategy remain obscure demanding enhanced research. Methods: Data from 323 consecutive patients (192 non-VE, 131 VE) receiving radical surgery and adjuvant chemotherapy in our institution between Jan. 2009 and Nov. 2014 were retrospectively collected. The follow-up deadline was Feb. 2016. Potential prognostic risk factors were tested using univariate and multivariate survival analyses. mRNAs differentially expressed between VE and non-VE stage III CRC were analyzed using Agilent Gene Expression oligo-microarrays (Version 6.5). Patient-derived xenograft (PDX) nude mouse models were constructed to evaluate the efficacy of adjuvant chemotherapy plus targeted drugs (cetuximab or bevacizumab) on VE and non-VE stage III CRC. Results: VE was significantly associated with gross tumor morphology (p = 0.001), histologic type (p < 0.001), lymph node status (p < 0.001), sub-class of stage III (p =0.001), and serum CA199 level (p = 0.022). VE and lymph node status were independent risk factors for overall survival (OS) (p < 0.001, p = 0.008) and disease-free survival (DFS) (p < 0.001, p = 0.007). The median OS and DFS in VE stage III CRC patients were 38 months and 24 months, respectively. Compared with pericarcinous tissue, 809 genes (396 up-, 413 down-regulated) were differently expressed in no-VE tissue, and 1513 genes (898 up-, 615 down-regulated) in VE tissue. The top ten up-regulated protein-coding genes in VE stage III CRC were ITGBL1 (p<0.001), PROCR (p<0.001), CENPW (p<0.001), ZNF485 (p<0.001), VEGFA (p<0.001), DSG3 (p<0.001), CYP24A1 (p=0.001), COL10A1 (p=0.001), HIST3H2A (p=0.001)and PSAT1 (p=0.002). Conclusions: VE appears to be an independent risk factor for the prognosis of stage III CRC patients treated with radical surgery and adjuvant chemotherapy. Differently regulated genes seem to be involved in the formation and progress of VE. The therapy scheme should be more individualized taking into account the VE status.
Deep learning can predict lymph node status directly from histology in colorectal cancer
