APPLICATION OF A PREDICTION MODEL TO IDENTIFY PSORIATIC ARTHRITIS PATIENTS WITH A HIGHER DIFFICULTNESS-TO-TREAT – ANALYSIS FROM A SINGLE-CENTRE COHORT

Background: Over the last two decades, treatment strategies for PsA have dramatically improved thanks to the application of treat-to-target and personalized medicine concepts and the introduction of new drug classes such as biologic and targeted-synthetic disease-modifying anti-rheumatic drugs (b/tsDMARDs). The clinical heterogeneity, the multiorgan involvement, and the comorbidities associated to psoriatic arthritis (PsA) are elements that contribute to the complexity of its management, despite the recent expansion in therapeutic options. To date, a true official definition of a D2T PsA has not yet been elaborated by the international scientific societies and the only attempt so far has been to translate the criteria established for D2T RA also to PsA, to produce a preliminary definition [1]. Objectives: The aim of our work is to identify the prevalence of difficult-to-treat (D2T) PsA patients from a monocentric cohort according to the definition proposed and to further characterize patients’ features associated with an increased difficultness of treatment. Methods: 267 consecutive, adult PsA patients treated with b/tsDMARDs were enrolled in our study. We collected demographic, clinical and clinimetrical data for each patient. According to the proposed definition , we derived the prevalence of D2T PsA patients from our cohort. Then, we calculated the number of failed b/tsDMARDs normalized by the expected number of failed b/tsDMARDs considering the follow up time for each patient. Patients with a higher difficultness-to-treat were defined by a ratio of observed and expected failed drugs (OBS/EXP) > 1. Using the OBS/EXP ratio as the response variable and clinical and demographic characteristics as independent variables, a generalized linear model with Poisson distribution was applied to explore potential features associated with difficultness-to-treat. A univariate analysis with Wald statistics followed by a multivariate analysis highlightened the difficultness-to-treat prediction model according to patients’ features. To pursue the robustness of the selection of the covariates, a bagging procedure was applied. First, a backward variable selection based on the Akaike Information Criterion (AIC) was performed on a bootstrap of 1000 samples. Then, the regression coefficients’ mean from each sample was calculated to perform a sufficiently robust prediction of the OBS/EXP ratio for each patient. Results: Among the 267 enrolled patients, only 8 (2.9%) satisfied the proposed D2T PsA criteria (Table 1). Following the application of the predictive model to 177 patients, 46 of them (26%) showed a higher difficultness-to-treat (OBS/EXP>1). At the univariate analysis, female sex ( p =0.037), psoriasis pattern ( p =0.05), presence of fibromyalgia ( p =0.01), and steroid therapy ( p <0.001) were associated with higher difficultness. The association between fibromyalgia, nail and pustular psoriasis, and steroid use was confirmed at the multivariate analysis (Figure 1). Conclusion: The proposed definition of D2T PsA patients [1] seems to capture only a small proportion of the prevalent patients with a higher difficultness-to-treat, as defined by a statistical predictive model. These data may provide a clue for possible criteria in the definition of D2T PsA. REFERENCES: [1] Perrotta FM, Scriffignano S, Ciccia F, Lubrano E. Clinical Characteristics of Potential “Difficult-to-treat” Patients with Psoriatic Arthritis: A Retrospective Analysis of a Longitudinal Cohort. Rheumatol Ther. 2022;9(4):1193-1201. doi:10.1007/s40744-022-00461-w Acknowledgements: NIL. Disclosure of Interests: None declared. DOI: 10.1136/annrheumdis-2024-eular.5952 Keywords: Outcome measures, Disease-modifying Drugs (DMARDs) Citation: , volume 83, supplement 1, year 2024, page 1465Session: Psoriatic arthritis (Publication Only)

Outcome measures, Disease-modifying Drugs (DMARDs)