Logistic regression analysis of small round cell neoplasms: a cytologic study

Abstract

The diagnosis of small round cell neoplasms (SRCN) in children is difficult by both histologic and cytologic methods. These neoplasms are unified morphologically by the scanty cytoplasm surrounding relatively round nuclei containing a primitive chromatin pattern. Further categorization is achieved histologically by the recognition of architectural differentiation and cytoplasmic features. Numerous series and case reports documenting the cytologic features of SRCN have been published in the English-language literature, but relatively few studies have statistically analyzed the diagnostic utility of these features. Our logistic regression analysis of 59 cases of SRCN indicates that the presence of an extremely scanty cytoplasm favors the diagnosis of Ewing's sarcoma (P = 0.004), as does the positive expression of cytoplasmic vacuoles (P = 0.02). Strap or tadpole cells closely correlate with the diagnosis of rhabdomyosarcoma (P < 0.001). The positive expression of rosettes strongly supports the presence of a neuroblastoma (P < 0.001). The diagnosis of Wilms' tumor is confirmed by the expression of tubules (P < 0.001). The presence of lymphoglandular bodies strongly favors the diagnosis of lymphoma (P < 0.001). While some cytologic features are highly correlated with specific SRCN, these features are not invariably present, and definitive diagnosis may require immunohistochemical or ultrastructural analysis performed on cell block material.