Construction of a prediction model for mismatch repair defects in colorectal cancer based on preoperative contrast-enhanced CT and clinicopathologic features
YUAN Shuo
ZHANG Bin
AN Zhiqiang
ZHANG Kai
ZHANG Chengyang
WEI Yanze
Abstract：Objective To investigate the correlation between preoperative enhanced CT and clinicopathologic features of colorectal cancer according to microsatellite instability(MSI-H)/mismatch repair defects(dMMR).Methods A total of 203 patients who attended Puyang People's Hospital affiliated with Xinxiang Medical College from August 2022 to July 2024 were retrospectively collected and included,and they were divided into the dMMR group(n=39)and the mismatch repair(pMMR)group(n=164).Preoperative enhanced CT features and clinicopathological features were analyzed,one-way and multifactorial Logistic analysis was performed,and the predictive model was constructed by screening relevant factors based on multifactorial Logistic analysis,creating column-line diagrams,evaluating the calibration chart and clinical applicability of the model and performing internal validation.Results In multifactorial analysis,lesion location(OR=0.37,P=0.046),degree of differentiation(moderately differentiated OR=0.24,P=0.009),percentage of intratumoral hypoluminescence(1/3-2/3:OR=1.233,P=0.010;>2/3:OR=6.90,P=0.008),short diameter of the largest lymph node≥8 mm(OR=2.67,P=0.039),and tumor short diameter(OR=1.59,P=0.002)were independent influences on the occurrence of dMMR(P<0.05).The mean AUC of internal validation of the model was 0.878(0.817-0.934)and 0.824(0.737-0.877),respectively,with good model differentiation.Conclusion The combination of CT imaging features and clinicopathologic features of colorectal cancer correlates with MMR status,which provides noninvasive MMR prediction,and the construction of a column-line graph prediction model shows good diagnostic performance.
Keywords：colorectal cancermicrosatellite instabilitymismatch repair proteinsimmunohistochemistry
Publication Date：2025-03-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：10( 330-339 )