CDX2 expression loss is commonly associated with mismatch repair deficiency (dMMR) in colorectal cancer (CRC). However, there are only a few studies that have attempted to correlate CDX2 expression loss with specific MMR genes (MLH1, MSH2, MSH6, PMS2). This is a retrospective study of 327 patients who underwent surgery due to CRC. Nine patients (2.9%) had two synchronous CRCs, making the total sample 336 CRC. Histopathological data such as tumor type, tumor grade, perineural, lymphatic, and vascular invasion, pT stage, pN stage, peritumoral and intratumoral lymphocytic infiltration were collected and recorded in the database. After immunohistochemical analysis, CDX2 expression, MLH1, MSH2, MSH6, and PMS2 deficiency were also recorded. CDX2 expression loss was detected in 19 out of 336 CRCs (5.9%) and was associated with ascending colon CRC, partially mucinous adenocarcinoma, poorly differentiated carcinoma, and dMMR. Forty-four (13.1%) of CRCs were dMMR. We found a statistically significant association between CDX2 expression loss and MLH1 and PMS2 deficiency. Considering that most expression phenotypes include pairs of MMR genes, we analyzed MLH1/PMS2 and MSH2/MSH6 as heterodimers. Analysis of heterodimers showed a similar result, namely, that MLH1/PMS2 heterodimer deficiency was significantly associated with CDX2 expression loss. We also constructed a regression model for CDX2 expression loss and for dMMR. Poor tumor differentiation and MLH1/PMS2 heterodimer deficiency have been identified as potential predictors for CDX2 expression loss. CRC in the ascending colon and CDX2 expression loss have been identified as positive potential predictors of dMMR with rectal cancer as negative potential predictor of dMMR. Our study showed a significant association between CDX2 expression loss and MLH1 and PMS2 deficiency in CRC. We also managed to produce a regression model for CDX2 expression and showed that poor tumor differentiation and MLH1/PMS2 heterodimer deficiency are independent factors for CDX2 expression loss. We were the first to include CDX2 expression in a regression model for dMMR and showed that CDX2 expression loss can be used as a predictive factor for dMMR, which should be confirmed by further studies.