Journal Article

AB0897 PREDICTIVE MODELS TO FORECAST DIFFICULT-TO-MANAGE AXIAL SPONDYLOARTHRITIS

Villalba, A., Plasencia-Rodríguez, C., Juárez, M., Díaz-Almirón, M., Monjo-Henry, I., Benavent, D., Balsa, A., Peiteado, D., Navarro-Compán, V., Novella-Navarro, M., Nuño, L.

2024

Overview

Background:Difficult-to-manage (D2M) axial spondyloarthritis (axSpA) is an emerging concept, whose definition is yet to be established. The characterisation of these subgroup of patients is relevant, as it may contribute to a broader understanding of the reasons behind treatment failure and to the development of new therapeutic strategies [1].Objectives:To develop a predictive model to forecast patients from the early stages of treatment with b/tsDMARDs.Methods:We analysed data from an observational prospective cohort from La Paz Hospital between 2004-2019 which included patients diagnosed of axSpA initiating a b/tsDMARD, and who fulfilled one of these two definitions: a) D2M: failure to at least 2 b/tsDMARDs, b) good responders (GR): patients remaining their first bDMARD for at least 3 years or withdrawing it because of sustained disease control. Clinical, laboratory, therapy-related information and disease activity measures prior to starting the first b/tsDMARD (baseline), as well as disease activity measures 6-months after initiating it, were collected. Delta-ASDAS was estimated as the difference between baseline and 6-month ASDAS. After excluding the variables with the greater number of missing values, all the variables associated with D2M-axSpA in the univariable regression analyses were selected in order to create Classification And Regression Tree (CART) models. The cohort was randomly split into two independent groups: a training set (80%) and a validation set (20%). Later on, the CART model inputted the most associated factors with D2M-axSpA selecting an optimal cut-off point for classification. Subsequent splits were made, using the Gini index, to divide the population into two branches, until the terminal node. Finally, the model’s performance was assessed using the validation set.Results:Among the 101 patients included in the cohort, 41 (41.6%) were classified as D2M, 59 (58.4%) were male with a mean age of 43 years old. D2M patients were less frequently HLA-B27 positive, had more peripheral manifestations (enthesitis), extra-musculoskeletal manifestations (IBD), comorbidities, and scored higher in composite disease activity indices 6 months after starting a first bDMARD (but did not in baseline indices). Two different CART models were obtained, with a lesser number of patients included in the second model because of the missing values. These 2 models with their cut-off points and the probability of D2M axSpA after each step are shown in Figure 1. The first model (Figure 1, model a) had a pre-test probability of D2M of 44%, and identified BASDAI (cut-off point < 4) after 6 months of therapy with the first bDMARD and age at the beginning of the first bDMARD (cut-off point ≥ 44 years old) as the most relevant factors. The second model (Figure 1, model b) had a pre-test probability of D2M of 47%. It used delta-ASDAS (cut-off point ≥ 1.1) as the variable for the first step, and tender joint count (cut-off point < 1) after 6 months of therapy with the first bDMARD and baseline age (cut-off point ≥ 53) for the second step. After validation, the CART models achieved an AUC of 0.94 (95% CI 0.87-1) and 0.83 (95% CI 0.67-1), respectively, and both of them classified properly 83% of the patients.Conclusion:This study identified two predictive models, which may be applied using everyday information, and could identify D2M-axSpA patients only after the first 6 months of bDMARD therapy. Next step will involve further validation in an external cohort.REFERENCES:[1] Wendling D, Verhoeven F, Prati C. Is the Difficult-to-Treat (D2T) concept applicable to axial spondyloarthritis? Joint Bone Spine. 2023;90(3):105512.Figure 1.CART models predicting D2M-axSpA. The value at each node represents the most frequently expected outcome (D2M in red or GR in green).GR: good responders, D2M: difficult-to-manage, BASDAI: Bath Ankylosing Spondylitis Disease Activity, ASDAS: Ankylosing Spondylitis Disease Activity Score, TJC: tender joint count.Acknowledgements:NIL.Disclosure of Interests:Manuel Juárez: None declared, Diego Benavent Eli Lilly, Janssen and UCB Pharma, Victoria Navarro-Compán Eli Lilly, Janssen, MSD, Novartis, Pfizer and UCB Pharma, AbbVie, Eli Lilly, Galapagos, Moonlake, MSD, Novartis, Pfizer and UCB Pharma, AbbVie and Novartis, Mariana Díaz-Almirón: None declared, Marta Novella-Navarro UCB, Lilly, Galapagos and Janssen, Diana Peiteado: None declared, Alejandro Villalba Janssen, Irene Monjo-Henry Roche, Novartis, UCB and Gedeon Richter, Laura Nuño: None declared, Alejandro Balsa AbbVie, Amgen, Pfizer, Galapagos, Novartis, Gilead, BMS, Nordic, Sanofi, Sandoz, Lilly, UCB and Roche, Chamaida Plasencia-Rodríguez AbbVie, Pfizer, Novartis, Lilly and Roche.

• Publisher: BMJ Publishing Group Ltd and European League Against Rheumatism,Elsevier B.V,Elsevier Limited
• Subject: Ankylosing spondylitis / Arthritis / Axial skeleton / Branches / Classification / Comorbidity / Disease control / Disease-modifying Drugs (DMARDs) / Inflammatory diseases / Patients / Prediction models / Prognostic factors / Regression analysis / Rheumatic diseases / Scientific Abstracts