BACKGROUND Abnormalities in tumor suppressors p14, p16 and p53 are reported in several human cancers. In ovarian epithelial carcinogenesis, p16 and p53 show higher immunohistochemical staining frequencies in malignant tumors and are associated with poor prognoses. p14 was only analyzed in carcinomas, with conflicting results. There are no reports on its expression in benign and borderline tumors. This study aims to determine p14, p16 and p53 expression frequencies in ovarian benign, borderline and malignant tumors and their associations with clinical parameters. METHODS A cross-sectional study utilizing immunohistochemistry was performed on paraffin-embedded ovarian epithelial tumor samples. Clinical data were collected from medical records. Fisher's exact test and the Bonferroni correction were performed for frequency associations. Survival comparisons utilized Kaplan-Meier and log rank testing. Associations were considered significant when p < 0.05. RESULTS p14 absent expression was associated with malignant tumors (60 % positive) (p = 0.000), while 93 % and 94 % of benign and borderline tumors, respectively, were positive. p16 was positive in 94.6 % of carcinomas, 75 % of borderline and 45.7 % of benign tumors (p = 0.000). p53 negative staining was associated with benign tumors (2.9 % positive) (p = 0.016) but no difference was observed between borderline (16.7 %) and malignant tumors (29.7 %) (p = 0.560). No associations were found between expression rates, disease-free survival times or clinical variables. Carcinoma subtypes showed no difference in expression. CONCLUSIONS This is the first description of p14 expression in benign and borderline tumors. It remains stable in benign and borderline tumors, while carcinomas show a significant absence of staining. This may indicate that p14 abnormalities occur later in carcinogenesis. p16 and p53 frequencies increase from benign to borderline and malignant tumors, similarly to previous reports, possibly reflecting the accumulation of inactive mutant protein. The small sample size may have prevented statistically significant survival analyses and clinical correlations. Future studies should investigate genetic abnormalities in p14 coding sequences and include all types of ovarian epithelial tumors. Bigger sample sizes may be needed for significant associations.