This study aimed to explore the relationship between tumor size (Ts) and prognosis in endometrial cancer (EC). A total of 52,208 patients with EC who underwent total hysterectomy were selected from the Surveillance, Epidemiology, and End Results Program database. Overall survival (OS) and endometrial cancer-specific survival (ESS) were chosen as survival outcomes. The Cox proportional hazards model was used to explore the effect of Ts on prognosis. The restricted cubic splines based on the Cox regression model were used to determine the nonlinear relationship between Ts and survival. When Ts was analyzed as a categorical variable, the risk of death increased with Ts, with the highest risk in patients with Ts > 9 cm with regard to all-cause death (ACD) (hazard ratio [HR] 1.317; 95% confidence interval [CI], 1.196-1.450; P < 0.001) and endometrial cancer-specific death (ESD) (HR, 1.378; 95% CI, 1.226-1.549; P < 0.001). As a continuous variable, Ts showed a nonlinear relationship with ACD (HR, 1.061; 95% CI, 1.053-1.069; P < 0.001) and ESD (HR, 1.062; 95% CI, 1.052-1.073; P < 0.001). The risk of mortality increased quickly with Ts when Ts was less than 7.5 cm and then leveled off when Ts was larger than 7.5 cm in all patients. Among patients with lymph node metastasis, the risk of poor prognosis decreased rapidly with Ts when Ts was less than 3.5 cm, and subsequently increased sharply with Ts when Ts ranged from 3.5 cm to 7.5 cm, and then increased slowly when Ts was larger than 7.5 cm (P < 0.001 for nonlinearity). There was a nonlinear relationship between Ts and prognosis in patients with EC. Clinicians should not ignore the impact of small tumors on prognosis in EC patients with lymph node metastasis.