BACKGROUND: Ki-67, cytokeratin 13, and/or cytokeratin 17 detection by immunohistochemistry has been reported to be useful for the diagnosis of oral precancerous lesions. However, the use of these markers remains controversial because of the lack of appropriately designed statistical studies. We assessed the hypothesis that Ki-67, cytokeratin 13, or cytokeratin 17 immunohistochemistry could facilitate the diagnosis of oral precancerous lesions and/or predict prognosis. METHODS: Epithelial dysplasia was classified as low grade (none or mild dysplasia) or high grade (moderate dysplasia, severe dysplasia, or carcinoma in situ). This study included 58 low-grade and 36 high-grade dysplasia cases. We used logistic regression to assess the diagnostic values of Ki-67, cytokeratin 13, and cytokeratin 17 for high-grade dysplasia. Correlations between these markers and the prognosis of oral atypical epithelium were assessed using the Cox proportional hazards model. RESULTS: Ki-67 overexpression and cytokeratin 13 loss were independent diagnostic markers for high-grade dysplasia (odds ratios, 1.92 and 2.53; 95% confidence intervals, 1.03-3.58, and 1.19-5.38, respectively). The area under the curve of Ki-67 was 0.73 and that of cytokeratin 13 was 0.72. However, the combination of Ki-67 and cytokeratin 13 yielded the area under the curve of 0.78. Ki-67 overexpression was significantly associated with recurrence and/or malignant transformation of oral atypical epithelium (hazard ratio, 7.25; 95% confidence interval, 1.07-48.92).CONCLUSIONS: Ki-67 overexpression and cytokeratin 13 loss may be useful for distinguishing oral precancerous lesions from reactive atypical epithelium. Moreover, Ki-67 overexpression may be a risk factor for recurrence and/or malignant transformation of oral atypical epithelium.