Aims To evaluate an immunohistochemical panel differentiating endometrial stromal sarcoma (ESS) from uterine leiomyosarcoma (ULMS) and leiomyoma (LM).

Methods 94 cases (28 ESS, 41 ULMS, 25 LM) were retrieved and arrayed. 10 immunomarkers (estrogen receptor (ER), progesterone receptor (PR), CD10, smooth muscle actin, desmin, h-caldesmon, transgelin, GEM, ASC1, stathmin1) were used. A predictive model was constructed and examined by receiver operating characteristics curve analysis to determine area under the curve (AUC).

Results The combination of ER⁺/PR⁺/CD10⁺/GEM⁻/h-caldesmon⁻/transgelin⁻ can predict ESS versus ULMS with AUC predictive value of 0.872 (95% CI 0.784 to 0.961, p<0.0001). The combination of ER⁺/PR⁺/CD10⁺/h-caldesmon⁻/transgelin⁻ can predict low grade (LG) ESS from ‘LG’ ULMS with AUC predictive value of 0.914 (95% CI 0.832 to 0.995, p<0.0001). Finally, ULMS and ESS, including the LGs, were more likely to be stathmin1⁺ than LM.

Conclusions Due to the different clinical course and management, adding novel antibodies (GEM, transgelin) to the well established immunohistochemistry panel seemed to be useful in distinguishing ESS from ULMS and LG ESS from ‘LG’ ULMS. Finally, stathmin1 expression could be of value in differentiating LM from uterine sarcomas.