Myxoid liposarcoma (MLPS) is a malignant adipocytic neoplasm with predilection for the extremities. MLPS is genetically defined by a t(12;16) translocation leading to FUS-DDIT3 (95%) or more rarely t(12;22) leading to EWSR1-DDIT3. Low-grade MLPS is characterized by bland spindle cells within a myxoid matrix containing delicate "chicken-wire" vasculature, whereas high-grade ("round cell") MLPS may be indistinguishable from other round cell sarcomas. In many cases, cytogenetic or molecular genetic techniques are applied to confirm the diagnosis. A recent study documented the utility of DDIT3 immunohistochemistry (IHC) in the differential diagnosis of adipocytic and myxoid soft tissue tumors. The purpose of this study was to evaluate DDIT3 IHC as a surrogate for molecular testing in high-grade MLPS. IHC was performed using a mouse monoclonal antibody directed against the N-terminus of DDIT3 on whole tissue sections from 50 high-grade MLPS cases and 319 histologic mimics used as controls (170 on whole tissue sections and 149 on a tissue microarray). Histologic mimics included Ewing sarcoma, CIC-rearranged sarcoma, sarcomas with BCOR genetic alterations, poorly differentiated synovial sarcoma, alveolar and embryonal rhabdomyosarcomas, mesenchymal chondrosarcoma, desmoplastic small round cell tumor, and neuroblastoma. Nuclear staining in >5% of cells was considered positive. By IHC, 48 (96%) high-grade MLPS showed strong diffuse nuclear staining for DDIT3. Of the controls, 2% of cases were positive, with no more than 25% nuclear staining. An additional 19% of control cases displayed less than 5% nuclear staining. Overall, DDIT3 IHC showed 96% sensitivity and 98% specificity for high-grade MLPS; strong, diffuse staining is also 96% sensitive but is 100% specific. IHC using an antibody directed against the N-terminus of DDIT3 is highly sensitive and specific for high-grade MLPS among histologic mimics and could replace molecular genetic testing in many cases, although limited labeling may be seen in a range of other tumor types.