Background/Purpose: The course of axial spondyloarthritis (SpA) is heterogeneous and remains to be better defined. DESIR is a longitudinal French cohort of early undifferentiated axial SpA, allowing to define different disease courses. We recently performed a cluster analysis, based on baseline characteristics and identified 2 SpA clusters: one, characterized by an isolated axial disorder (A for axial), the other by additional high prevalence of peripheral manifestations (B for both) (Costantino et al. Arthritis Rheumatol, 2016). These two clusters may be linked to different levels of disease severity. The aim of the present study was to assess whether cluster-based SpA phenotypes identified at baseline are stable over time and might predict different severity disease outcomes at follow-up.

Methods: We analysed longitudinal data from the 535 patients fulfilling the ASAS classification criteria (either axial or peripheral criteria) and who completed all follow-up visits until year 5 (months 6, 12, 18, 24, 36, 48 and 60) out of the 679 patients in the DESIR cohort used to define clusters at baseline. We performed a linear mixed-effect analysis for quantitative variables and a generalized linear mixed-effect analysis for qualitative variables with cluster and follow-up visits as fixed effects and subjects as random effect. P-values were obtained by likelihood ratio tests of the full model with cluster against the model without cluster as fixed effects. To determine a way to predict a definite cluster in a given patient at baseline, a classification and regression tree (CART) analysis was performed. A 10-fold cross validation was used to estimate misclassification rates.

Results: Over the time, both clusters continued to show significant differences with respect to prevalence of peripheral involvement, higher disease activity, worse patient-reported outcome, higher frequency of conventional DMARDs and TNF inhibitor use in cluster B,  and higher prevalence of radiographic and MRI sacroiliitis at 2 and 5 years in cluster A (Table 1). CART analysis (Figure 1) shows that probability to maintain the same cluster over time is very high if a given patient belongs to a definite cluster-phenotype (A or B), with sensitivity of 98%, specificity of 87% and positive likelihood ratio of 7.62 at baseline (Table 2).

Conclusion: Cluster-based SpA phenotypes identify different severity outcomes at 5 years. We develop a decision tree for cluster determination, which shows accurate classification performance. This algorithm is easy to use (only three clinical parameters) and could help to identify patients at risk of poor outcome in early axial SpA.

« Back to 2019 ACR/ARP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/cluster-based-spondyloarthritis-phenotypes-defined-at-baseline-are-predictive-of-different-severity-outcomes-at-5-year-in-the-desir-cohort/