POS0459 UNSUPERVISED CLUSTERING IDENTIFIES UNIQUE SUBSETS OF PATIENTS IN A RACIALLY AND ETHNICALLY DIVERSE RHEUMATOID ARTHRITIS COHORT

Background: Single biomarkers have limited utility to date in guiding RA clinical care. Machine learning algorithms may better identify and stratify patients with differential outcomes. Objectives: To determine if unsupervised machine methods can be employed a racially ethnically diverse cohort clusters of different disease activity trajectories, as measured by DAS28ESR. Methods: Data are derived from the longitudinal, observational University California, San Francisco Cohort. Along routine labs, medications assessments, multiple biomarker (MBDA) panel was obtained at each visit. The MBDA measures 12 serum biomarkers. Four patient were identified K-prototype clustering after collapsing all observations into cross sectional dataset. Plots created display longitudinal trajectories for cluster. Lasso regression applied associated Results: We 4 distinct our (Table 1) visually (Fig. 1). Cluster 1 (n=116) older (63.6±9.7), had highest proportion Asian participants (n=73, 63%) most study visits longest duration. 2 (n=70) mean DAS28ESR (5.5±0.7), dose prednisone (8.6±4.9 mg/day). 3 lowest number (n=14), biologic use (28.6%). largest cluster (n=173) shortest duration (4.9±3.8 years) (61.3%). In regressions, leptin found significant positive associations whole group well Clusters 4. EGF negative group, CRP 1. YKL40 VCAM1 3, respectively. Conclusion: unique activity. Although additional work is needed explore outcomes cluster, application combinations characteristics influencing Table Demographics cohort. Biomarkers significantly determined Values listed per standard deviation biomarker. Overall (N=373) (N=116) (N=70) (N=14) (N=173) Demographics: Age 54.8 ± 3.6 63.6 9.7 50.8 14.9 58.2 15.8 50.3 12.1 Female Sex 318 (85%) 101 (87%) 57 (81%) 11 (79%) 149 (86%) Race: Hispanic/Latino 181 (49%) 22 (19%) 47 (67%) 5 (36%) 107 (62%) 123 (33%) 73 (63%) 8 (11%) 6 (43%) 36 (21%) Black 35 (9%) (10%) (14%) 13 (8%) White & Other 34 9 (7%) 7 17 Clinical Characteristics: Rheumatoid Factor 315 104 (90%) 56 (80%) (93%) 142 (82%) ACPA 297 98 54 (77%) 133 Disease Duration 7.8 7.6 13.7 5.4 4.6 6.7 5.7 4.9 3.8 csDMARD 344 (92%) 108 63 160 Biologic DMARD 185 (50%) 45 (39%) 30 (29%) 106 (61%) Prednisone Dose 6.0 4.0 8.6 5.8 1.4 6.3 2.8 Body Mass Index 28.2 4.5 26.6 28.7 28.0 6.2 29.1 4.7 4.2 1.1 1.0 5.5 0.8 3.9 3.7 0.9 -0.16* -0.41*** -- -0.20** Leptin 0.15** 0.21* 0.21** c-reactive protein 0.34** 0.51*** -0.73* 0.26* -EGF: epidermal growth factor; VCAM1: Vascular cell adhesion 1; YKL40: Chitinase-3-like -*p<0.05, **p<0.01, ***p<0.001 Figure trajectory plots 95%CIs Acknowledgements: This supported Rheumatology Research Foundation Scientist Development Award. Disclosure Interests: None declared