Based on a multi-centered and a large sample size, this study aims to analyze the relationship between preoperative and postoperative serum CEA and recurrence of rectal cancer without preoperative therapy. This retrospective cohort study enrolled stage I to III rectal cancer patients without preoperative therapy (N = 1,022) who received radical resection of rectal cancer from 2 hospitals in China. Based on the preoperative and postoperative serum carcinoembryonic antigen, the patients were subdivided into 3 groups ie, normal preoperative CEA (≤5.0 ng/mL, N = 627), elevated preoperative (>5.0 ng/mL) but normalized postoperative CEA (normalized postoperative CEA, N = 255), as well as elevated preoperative and postoperative CEA (elevated postoperative CEA, N = 67). The generalized additive model was used to assess the relationship between carcinoembryonic antigen and the risk of recurrence. Further, the Cox regression model was used to evaluate the relationship between carcinoembryonic antigen and 3-year recurrence-free survival (RFS) after adjusting for potential confounders. The 3-year RFS of patients with elevated postoperative CEA was 45.8% (95% CI, 35.2% -59.5%), which was significantly lower compared to the other two groups of patients (normalized postoperative CEA: 75.9%, 95% CI, 70.8%-81.4%; and normal preoperative CEA: 84.9%, 95% CI, 82.2%-87.8%) (P <0.001). Based on multivariable Cox model analysis, the elevated postoperative CEA was a prognostic factor for 3 years RFS (hazard ratio [HR], 3.08; 95% CI, 2.05-4.66; P<0.001). At the same time, normalized postoperative CEA was insignificantly correlated with 3-year RFS (HR, 1.38; 95% CI, 1.00-1.92; P = 0.05) and was not an independent risk factor. We found that preoperative and postoperative serum CEA of rectal cancer patients were related to the 3-year recurrence-free survival rate. Moreover, the risk of recurrence in the normalized postoperative CEA group of patients was insignificantly different from that of the normalized preoperative CEA patients. Therefore, it is necessary to combine preoperative and postoperative CEA to predict the prognosis of patients with rectal cancer, rather than using it alone. © 2021 Pu et al.