Background. Psoriatic arthritis risk prediction and early detection in patients with psoriasis may help prevent irreversible musculoskeletal changes and improve patients outcomes.
 Aims. To develop and validate predictive model for psoriatic arthritis risk assessment and classification for patients with moderate-to-severe psoriasis based on demographic and clinical characteristics.
 Materials and methods. Data of psoriasis patient registry of Russian Society of Dermatovenereologists and Cosmetologists was analyzed. Significant differences between independent variables of interest among patients with and without psoriatic arthritis were determined by means of 2-test or MannWitney test. Predictive models were developed stepwise by means of logistic regression analysis. Independent variables of low significance were excluded from the model. Regression coefficients were considered significant if p 0.05. The optimal cut-off value was derived from ROC-analysis. The model performance was evaluated by calculation of AUC, sensitivity and specificity on training and test datasets. Finally, adjusted regression coefficients, AUC, sensitivity and specificity were derived for pooled data.
 Results. Training sample included 3245 patients with psoriasis, 920 of them had diagnosis of psoriatic arthritis. The final predictive model included five significant predictors: psoriasis duration, medical history of psoriatic erythroderma, family history of psoriatic arthritis, arterial hypertension, and fatty liver. All regression coefficients were highly significant (p 0.001). The AUC of prediction model adjusted on pooled data was 0,7473, sensitivity 70%, specificity 66% for cut-off value 0.212.
 Conclusions. Developed predictive model for risk assessment of psoriatic arthritis may contribute to its earlier detection in patients with psoriasis taking into account the degree of influence of significant predictors. The proposed classification may be used to discriminate patients at higher risk of psoriatic arthritis.
Read full abstract