Objectives: To identify phenotype clusters and their trajectories in psoriatic arthritis (PsA) and examine the association of the clusters with treatment response in a real-world setting.
Methods: In the multicentre PsA Research Consortium (PARC) study, we applied factor analysis of mixed data to reduce dimensionality and collinearity, followed by hierarchical clustering on principal components. We then evaluated the transition of PsA clusters and their response to new immunomodulatory therapy and tumour necrosis factor inhibitor (TNFi).
Results: Among 627 patients with PsA, three clusters were identified: mild PsA and psoriasis only (PsO) (Cluster 1, 47.4%), severe PsA and mild PsO (Cluster 2, 34.3%) and severe PsA and severe PsO (Cluster 3, 18.3%). Among 339 patients starting or changing, significant differences in response were observed (mean follow-up of 0.7 years, SD 0.8), with Cluster 3 showing the largest improvements in cDAPSA and PsAID. No differences were found among those starting TNFi (n=218). cDAPSA remission and PsAID patient acceptable symptom state were achieved in 10% and 54%, respectively. Clusters remained stable over time despite treatment changes, though some transitions occurred, notably from Cluster 3 to milder clusters.
Conclusion: Data-driven clusters with distinct therapy responses identified in this real-world study highlight the extensive heterogeneity in PsA and the central role of psoriasis and musculoskeletal severity in treatment outcomes. Concurrently, these findings underscore the need for better outcome measures, particularly for individuals with lower disease activity.
Keywords: disease activity; machine learning; psoriatic arthritis; treatment.
© Author(s) (or their employer(s)) 2024. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ on behalf of EULAR.