Background: Clinical characteristics combined with new biomarkers help discriminate between atypical uterine smooth muscle tumors (AUSMT) and leiomyosarcomas (LMS). Patients and Methods: We retrospectively collected a series of leiomyomas (LM), AUSMT, and LMS. Estrogen receptors (ER), progesterone receptors (PR), p16, Ki-67, and p53 expression were assessed by immunohistochemistry. For AUSMT patients, immunohistochemistry evaluations were performed at the time of diagnosis and at recurrences. Results: A total of 27 cases of AUSMT, 22 LM, and 31 LMS were identified. The expression of ER and PR decreased from LM to LMS (ER+: LM 95.5%, AUSMT 88.9%, LMS 41.9%, p < 0.001; PR+: LM 100%, AUSMT 88.9%, LMS 38.2%, p = 0.002). By contrast, p16 and p53 expression increased (p16+: LM 4.5%, AUSMT 40.7%, LMS 45.2%, p = 0.004; p53: LM 9.1%, AUSMT 33.3%, LMS 58.1%, p = 0.001). At a median follow-up of 33.47 months, 40.7% of patients with AUSMT experienced recurrent disease, 6 patients relapsed as AUSMT and 5 as LMS. In univariate analysis was observed that ER status (p = 0.027) and p53 expression (p = 0.015) predicted risk of relapse. Conclusions: Treatment of AUSMT should be centralized in dedicated centers. International collaborations are needed to optimize research strategy, which may lead to the identification of new useful biomarkers and to improvement in the clinical management of this rare disease.