Effectiveness of Toki's criteria and determination of variables for identification of HPV L1 protein in oral lesions.

ABSTRACT

OBJECTIVE: To evaluate the effectiveness of Toki's criteria in identifying the HPV L1 protein in oral lesions with the use of immunohistochemistry (IHC) and to determine which criteria optimize such identification. STUDY DESIGN: Retrospective study of 277 cases diagnosed as HPV lesions at 22 years. Tests of sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), kappa coefficients, and chi2 values, as well as two logistic regression analyses (p≤0.05), were conducted. RESULTS: Of the lesions studied, 96.4% (267 of 277) were positive for HPV using Toki's criteria and 28.5% (79 of 277) were positive for L1 by IHC. Toki's criteria showed sensitivity=93.67%, specificity=2.53%, PPV=6.99%, and NPV=46.55%. Neither concordance nor statistically significant associations were observed between both tests. The logistic regression of Toki's criteria was useful in the diagnosis of L1, correctly classified 71.8% of the lesions positive for L1, and showed a Hosmer-Lemeshow adjustment of p=0.614 and a Nagelkerke's coefficient of determination of 6.8%. The explanatory variables statistically significant at p≤0.05 were dyskeratosis (p=0.01) and papillomatosis (p=0.04). Forty-nine independent variables (clinical and histopathologic) were involved in the second regression analysis. The model correctly classified 85.2% of the lesions and showed a Hosmer-Lemeshow adjustment of p=0.696 and a Nagelkerke's coefficient of determination of 60.2%. The explanatory variables statistically significant at p≤0.05 were age younger than 35 years (p=0.001), multiple lesions (p=0.031), hyperorthokeratosis (p=0.019), focal intracellular edema (p=0.002), and the presence of 1 to more than 5 cells with degenerative changes in their nucleus (p=0.048). CONCLUSIONS: Toki's criteria are not adequate to make a diagnosis of lesions by HPV in the mouth, but the logistic regression analysis showed clinical and histopathologic variables which optimize the identification of lesions through the L1 protein. However, a PCR study is advisable when the presence of high-risk HPV is suspected.