A Prediction Model for the Differential Diagnosis of Bowen's Disease and Bowenoid Actinic Keratosis

ABSTRACT

Bowen's disease and bowenoid actinic keratosis share overlapping clinical and histopathological features, and there is no clear consensus regarding their differential diagnosis. This study aimed to identify clinicopathological differences between the two diseases and to develop a diagnostic prediction model in a Korean population. We retrospectively analyzed 253 patients pathologically diagnosed with Bowen's disease or bowenoid actinic keratosis at a single institution in South Korea. Clinical and histopathological variables were compared between the two groups, and significant variables were incorporated into a multivariable logistic regression model. The model was externally validated using an independent cohort of 207 patients from another institution in South Korea. In the development cohort, variables significantly associated with diagnosis included lesion location, adjacent pigmented lesions, solar elastosis, pathologically identical lesions in the same region, and the number of atypical cells within a 3 mm width of the epidermis. The prediction model showed good discriminative performance, with an area under the receiver operating characteristic curve of 0.897. In the external validation cohort, the model demonstrated excellent performance, with an area under the curve of 0.994 and an accuracy of 95.17%. These findings suggest that integration of clinical and histopathological features may help differentiate Bowen's disease from bowenoid actinic keratosis in Korean clinical settings. Further validation in ethnically and geographically diverse populations is needed before broader application of the proposed model.