Predicting keratinocyte carcinoma in patients with actinic keratosis: development and internal validation of a multivariable risk-prediction model

Background Patients with actinic keratosis (AK) are at increased risk for developing keratinocyte carcinoma (KC) but predictive factors and their risk rates are unknown. Objectives To develop and internally validate a prediction model to calculate the absolute risk of a firstKCin patients withAK. Methods The risk-prediction model was based on the prospective population-based Rotterdam Study cohort. We hereto analysed the data of participants with at least oneAKlesion at cohort baseline using a multivariable Cox proportional hazards model and included 13 a priori defined candidate predictor variables considering phenotypic, genetic and lifestyle risk factors.KCs were identified by linkage of the data with the Dutch Pathology Registry. Results Of the 1169AKparticipants at baseline, 176 (15 center dot 1%) developed aKCafter a median follow-up of 1 center dot 8 years. The final model with significant predictors was obtained after backward stepwise selection and comprised the presence of four to nineAKs [hazard ratio (HR) 1 center dot 68, 95% confidence interval (CI) 1 center dot 17-2 center dot 42], 10 or moreAKs (HR2 center dot 44, 95%CI1 center dot 65-3 center dot 61),AKlocalization on the upper extremities (HR0 center dot 75, 95%CI0 center dot 52-1 center dot 08) or elsewhere except the head (HR1 center dot 40, 95%CI0 center dot 98-2 center dot 01) and coffee consumption (HR0 center dot 92, 95%CI0 center dot 84-1 center dot 01). Evaluation of the discriminative ability of the model showed a bootstrap validated concordance index (c-index) of 0 center dot 60. Conclusions We showed that the risk ofKCin patients withAKcan be calculated with the use of four easily assessable predictor variables. Given the c-index, extension of the model with additional, currently unknown predictor variables is desirable. Linked Comment: Kimet al. Br J Dermatol2020; 183:415-416.