SAT0033 Time-dependent relationships between biological parameters and disease activity in systemic lupus erythematosus. (12th June 2018)

• Record Type: Journal Article
• Authors: Connelly, K.; Nim, H.; Vincent, F.; Petitjean, F.; Hoi, A.; Kohlmeyer, R.; Boyd, S.; Morand, E.
• Is Part Of: Annals of the rheumatic diseases. Volume 77(2018)Supplement 2
• Page Start: 881
• Page End: 882
• Publication Date: 2018-06-12
• Subjects: Rheumatism -- Periodicals; 616.723005
• DOI: 10.1136/annrheumdis-2018-eular.2034
• Languages: English
• ISSNs: 0003-4967

Abstract

Abstract : Background: Systemic lupus erythematosus (SLE) is a multisystem autoimmune disease characterised by high inter-patient variability of clinical features, pathology, and disease time-course. Relationships between biomarkers and disease remission/relapse cycles are especially complex and poorly understood. Objectives: To investigate the relationship between disease activity and biomarker expression in a longitudinally-followed SLE cohort. Methods: We measured 4 candidate protein biomarkers implicated in SLE (MIF, CCL2, CCL19 and CXCL10) and 13 routinely collected serum and urine biological parameters, and assessed disease activity (SLEDAI-2k) on each clinic visit. We analysed these data by first focusing on the magnitude of expression levels of the 17 biological markers and then on the temporal dimension, to untangle their relationship to disease activity. Results: Data from 843 visits in 110 SLE patients (median age 47, 83% female, 49% Asian ethnicity) were analysed. We demonstrated highly heterogeneous time-dependent relationships between disease activity and the measured biological markers. Using unbiased magnitude-based hierarchical clustering of biomarker expression levels, we isolated a patient subset (n=9) with distinctively heterogeneous patterns of expression of the 17 biological parameters, compared to the other (n=101) patients who were more homogeneous. The smaller subgroup had significantly higher levels of MIF, CCL2, CCL19 and CXCL10, but the larger subgroup had stronger associations between biological parameters and SLEDAI-2k, based on leave-one-out cross-validated regression analysis. In this subgroup, when we constructed a time-dependent regression model, compared to the equivalent time-agnostic regression model, the biological parameters had significantly stronger predictive power for disease activity, suggesting a time-dependent relationship. To disentangle the effect of magnitude versus temporal correlation, we used dynamic time-warping analysis to align longitudinal clinical and laboratory profiles. This revealed a further subset (n=69) in whom a time-dependent regression model showed significantly stronger associations between biological parameters and disease activity, despite no significant difference in simple magnitude. This subgroup was characterised by lower rates of flare, lower disease activity and lower damage scores, suggesting that this patient cluster is highly clinically meaningful. Conclusions: Using aggregated longitudinal clinical data and samples, we demonstrated significant subgroups of time-dependent relationships between disease activity and biological markers among patients with SLE. These results imply the association between biological parameters and disease activity may exist in a multi-dimensional time-dependent pattern. Longitudinal SLE data presents potential opportunities to identify patient-stratifying biomarker patterns that are concealed when time is not considered. This finding has significant implications for the design of SLE biomarker studies. Disclosure of Interest: None declared