Cytologic evaluation of pancreatobiliary brushings is specific but poorly sensitive for malignancy. Detection of polysomic cells by fluorescence in situ hybridization (FISH) is significantly more sensitive than routine cytology with similar specificity. The purpose of this study was to identify cytologic criteria most associated with malignancy in specimens unaffected by sample failure. Endoscopic brushings were split equally for routine cytologic and FISH analyses per clinical practice. We retrospectively evaluated 16 cytologic criteria on Papanicolaou-stained slides. We assumed that the presence of polysomic cells by FISH indicated successful tumor sampling in specimens from patients with pathologic evidence of malignancy on follow-up. We compared cytologic criteria of malignant brushings with corresponding positive FISH results (positive control, n = 39) with those without evidence of malignancy and corresponding negative FISH results (negative control, n = 30). The presence of single abnormal cells, irregular nuclear membranes, and enlarged nuclei were independent predictors of malignancy by logistic regression (P < .05).