Logistic regression analysis of myxoid sarcomas: a cytologic study.

Abstract

The objective of this study was to identify key diagnostic cytologic criteria for the most common myxoid sarcomas studied by fine-needle aspiration cytology. We reviewed 27 myxoid malignant fibrous histiocytomas, 8 chordomas, 16 chondrosarcomas, and 12 myxoid liposarcomas in which both cytologic specimens and final histopathologic diagnoses were available. All specimens were coded as to the presence or absence of the following variables: high cellularity, low cellularity, tissue fragments, epithelial fragments, pale/loose ground substance, dense ground substance, chondroid fragments, large amount of myxoid material, small amount of myxoid material, capillary vessel networks, pleomorphism, binucleate cells, multinucleate cells, physaliphorous cells, cells in lacunae, signet ring cells, lipoblasts, fibroblast-like cells, histiocyte-like cells, stellate cells, long filamentous cells, short spindle cells, osteoclastic giant cells, nuclei with pointed ends, nuclei with cigar-shaped ends, fish-hook nuclei, round/ovoid nuclei, naked nuclei, large nucleoli, small nucleoli, mitotic figures, abnormal mitotic figures, intracytoplasmic hemosiderin deposits, background cells, fat, cytoplasmic vacuoles, and pleomorphic giant cells. A logistic regression analysis was performed to identify the variables predictive of myxoid malignant fibrous histiocytoma, chordoma, myxoid chondrosarcoma, and myxoid liposarcoma. The statistical analysis selected pleomorphic giant cells and the presence of fibroblast-like cells as most predictive of malignant fibrous histiocytoma, physaliphorous cells as most closely associated with chordoma, chondroid fragments as most predictive of chondrosarcoma, and lipoblasts as most predictive of liposarcoma. While myxoid lesions have many overlapping cytologic features, key criteria including the presence of lipoblasts, physaliphorous cells, chondroid fragments, and pleomorphic giant cells are useful in subclassifying these neoplasms.