RT Journal Article SR Electronic T1 OP0114 MACHINE LEARNING TOOLS IDENTIFY PATIENT CLUSTERS AND SWOLLEN AND TENDER JOINT CORRELATION PATTERNS IN A LARGE DATABASE FROM THE SECUKINUMAB PSORIATIC ARTHRITIS CLINICAL DEVELOPMENT PROGRAM JF Annals of the Rheumatic Diseases JO Ann Rheum Dis FD BMJ Publishing Group Ltd and European League Against Rheumatism SP 131 OP 131 DO 10.1136/annrheumdis-2019-eular.1910 VO 78 IS Suppl 2 A1 Kormaksson, Matthias A1 Pournara, Effie A1 Ligozio, Gregory A1 Pricop, Luminita A1 Abrams, Ken A1 Kirkham, Bruce A1 Reich, Kristian A1 McInnes, Iain YR 2019 UL http://ard.bmj.com/content/78/Suppl_2/131.1.abstract AB Background Identifying patient phenotypes using machine-learning (ML) techniques amidst the variability and heterogeneity of the clinical manifestations of psoriatic arthritis (PsA) could be the first critical step towards better understanding of the disease eventually leading to individualized medicine.1 Objectives To identify distinct clusters of patients with PsA based on patients’ tender joint (TJ) and swollen joint (SJ) counts and correlation patterns among TJ and SJ counts at baseline as captured in the secukinumab FUTURE trials program.Methods Pairwise correlations were explored among 76 SJ and 78 TJ measurements of >2,700 patients with PsA across 5 phase III studies with ≈425,000 data entries at baseline and were visualized using heatmaps. Due to high correlations between SJs and corresponding TJs, a composite variable “swollen/tender joint count” was constructed for each joint. Hierarchical clustering was then performed on the composite using “1-correlation” as the dissimilarity metric and Ward’s agglomeration method for pairwise grouping of joints. A dendrogram was used to visualize and assess the resulting joint groupings.Results The hierarchical clustering algorithm grouped the 78 individual joints into distinct and natural clusters (Figure 1A). At higher level of the dendrogram, the algorithm grouped separately all foot, larger (jaws, clavicles, ankles, hips, wrists, knees, shoulders, elbows), and hand joints. Cutting the dendrogram at 15 clusters separated all the joints into distinct groups; hand joints (distal and proximal phalanges, metacarpals and thumbs), and foot joints (distal and proximal phalanges, metatarsals and big toes). Similar clustering algorithms were explored to identify patient clusters at baseline with distinct swelling and tenderness patterns across the identified joint groups. High correlation between swelling/tenderness of the left and swelling/tenderness of the corresponding right joint was observed across all individual joints (Figure 1B); a high correlation was also observed between swelling and tenderness at all individual joints. More localized patterns showed that there is a gradual decrease in correlation (from highest to lowest) among TJs and SJs in adjacent vs non-adjacent fingers, which is evident from grey-scale patterns (Figure 1C). Specifically, a gradual decrease in correlation between the swelling of 2nd distal interphalangeal joint and the tenderness of the 2nd–5th distal phalanges was noted.Conclusion Machine learning methodology confirmed a natural grouping of joints in patients with psoriatic arthritis based on baseline swelling and tenderness and revealed complex correlation patterns. Additional cluster analyses have demonstrated distinct patient clusters across the identified joint groups. Further investigating potential associations of other disease manifestations such as skin and nail involvement to define additional phenotypes may explain differences in disease pathogenesis and treatment outcomes.Reference [1] Grys BT, et al., J Cell Biol. 2017; 216(1): 65-71.Disclosure of Interests Matthias Kormaksson Shareholder of: Novartis, Employee of: Novartis, Effie Pournara Shareholder of: Novartis, Employee of: Novartis, Gregory Ligozio Shareholder of: Novartis, Employee of: Novartis, Luminita Pricop Shareholder of: Novartis, Employee of: Novartis, Ken Abrams Shareholder of: Novartis, Employee of: Novartis, Bruce Kirkham Grant/research support from: Abbvie, Janssen, Lilly, Novartis, Roche, UCB, Consultant for: Abbvie, Janssen, Lilly, Novartis, Roche, UCB, Speakers bureau: Abbvie, Janssen, Lilly, Novartis, Roche, UCB, Kristian Reich Consultant for: Abbvie, Affibody, Amgen, Biogen, Boehringer Ingelheim Pharma, Celgene, Centocor, Covagen, Forward Pharma, Fresenius Medical Care, GlaxoSmithKline, Janssen-Cilag, Kyowa Kirin, Leo, Lilly, Medac, Merck Sharp & Dohme Corp., Novartis, Miltenyi Biotec, Ocean Pharma, Pfizer, Regeneron, Samsung Bioepis, Sanofi, Takeda, UCB Pharma, Valeant, Xenoport; Speakers bureau: Abbvie, Affibody, Amgen, Biogen, Boehringer Ingelheim Pharma, Celgene, Centocor, Covagen, Forward Pharma, Fresenius Medical Care, GlaxoSmithKline, Janssen-Cilag, Kyowa Kirin, Leo, Lilly, Medac, Merck Sharp & Dohme Corp., Novartis, Miltenyi Biotec, Ocean Pharma, Pfizer, Regeneron, Samsung Bioepis, Sanofi, Takeda, UCB Pharma, Valeant, Xenoport, Iain McInnes Grant/research support from: AstraZeneca, Celgene, Compugen, Novartis, Roche, UCB Pharma, Consultant for: AbbVie, Celgene, Galvani, Lilly, Novartis, Pfizer, UCB Pharma