Exploring Heterogeneity in Rheumatoid Arthritis: Patient Profiling Through Principal Component and Cluster Analysis of the BRASS Registry

Jeffrey Curtis¹, Michael Weinblatt ², Kenneth Saag ¹, Vivian Bykerk ³, Christina Charles-Schoeman ⁴, Stefano Fiore ⁵, Gregory St John ⁶, Toshio Kimura ⁷, Shen Zheng ⁵, Clifton Bingham ⁸, Grace Wright ⁹, Martin Bergman ¹⁰, Kamala Nola ¹¹, Daniel Furst ⁴ and Nancy Shadick ², ¹University of Alabama at Birmingham, Birmingham, AL, ²Brigham and Women's Hospital, Boston, MA, ³Hospital for Special Surgery, New York City, NY, ⁴University of California, Los Angeles, CA, ⁵Sanofi Genzyme, Bridgewater, NJ, ⁶Regeneron Pharmaceuticals, Inc., Tarrytown, NY, ⁷Regeneron Pharmaceuticals, Inc, Tarrytown, NY, ⁸Johns Hopkins University, Baltimore, MD, ⁹Private Practice, New York City, NY, ¹⁰Drexel University College of Medicine, Stockholm, Sweden, ¹¹Lipscomb University College of Pharmacy & Health Sciences, Nashville, TN

Meeting: 2019 ACR/ARP Annual Meeting

Keywords: Epidemiologic methods, population studies and phenotypes, registry, Rheumatoid arthritis (RA)

Session Information

Date: Sunday, November 10, 2019

Title: Epidemiology & Public Health Poster I: RA

Session Type: Poster Session (Sunday)

Session Time: 9:00AM-11:00AM

Background/Purpose: Data-driven principal component (PC) and cluster analysis has the potential to identify previously unknown patient subgroups within a rheumatoid arthritis (RA) registry to establish prognosis, predict disease trajectory, and help inform treatment. The Brigham and Women’s Rheumatoid Arthritis Sequential Study (BRASS), established in 2003, is a single-center, prospective observational registry cohort providing a comprehensive set of clinical disease activity measures in >1400 patients with RA. Our objective was to use PC and cluster analysis of baseline demographic, socio-economic, health and disease characteristics in BRASS to identify and characterize distinct patient clusters in RA.

Methods: Patient variables recorded at entry into BRASS were refined and combined using PC analysis to reduce dimensionality and collinearity. The number of PCs was established by eigenvalue >1, cumulative variance, and interpretability. Patients were clustered using a k-means approach with non-hierarchical, exclusive, and complete clustering, with minimum cluster size 5% of population, and maximum 19 clusters. The final number of clusters was determined according to the cubic clustering criterion and pseudo F.

Results: Analysis of baseline data from 1443 patients identified 41 PCs that capture the fundamental characteristics involved in RA. Cluster analysis distinguished 5 patient clusters. Each cluster reflected a different profile of PCs, and can be described based on overall health, RA disease activity and duration (Table). Key differentiators between clusters include comorbidity PCs (metabolic comorbidities predominate in cluster 1, neurologic in cluster 4, and orthopedic in cluster 5) and patient characteristics/social PCs (greatest number of doctor visits and family history of MI in cluster 3, greatest BMI in cluster 1, highest income in cluster 2, lowest income in cluster 5, and least emotional support in cluster 1).

Conclusion: Data-driven cluster analysis of RA patient characteristics at entry into the BRASS registry identified five distinct patient phenotypes, providing a convenient method to potentially derive novel insights into the multifactorial drivers, commonly co-occurring health conditions, and manifestations of RA. Investigation of longitudinal outcomes in these different clusters in the BRASS registry and validation in an independent dataset is ongoing.

Disclosure: J. Curtis, AbbVie, 2, 5, Abbvie, 2, 5, AbbVie, Amgen, Bristol-Myers Squibb, Corrona, Lilly, Janssen, Myriad, Pfizer, Regeneron, Roche, and UCB, 2, 5, Amgen, 2, 5, Amgen Inc., 2, 5, BMS, 2, 5, Bristol-Myers Squibb, 2, 5, Corrona, 2, 5, Crescendo, 2, 5, Eli Lilly, 2, 5, Eli Lilly and Company, 2, 5, Genentech, 2, 5, Janseen, 5, Janssen, 2, 5, Janssen Research & Development, LLC, 2, Lilly, 2, 5, Myriad, 2, 5, Patient Centered Outcomes Research Insitute (PCORI), 2, Pfizer, 2, 5, Radius Health, Inc., 9, Regeneron, 2, 5, Roche, 2, 3, 5, Roche/Genentech, 5, UCB, 2, 5; M. Weinblatt, Abbvie, 5, AbbVie, 5, Amgen, 5, BMS, 2, 5, Bristol Myers Squibb, 2, 5, Bristol-Myers Squibb, 2, 5, Canfite, 1, 4, Corrona, 5, Crescendo Bioscience, 2, 5, Eli Lilly and Company, 5, Gilead, 5, Glaxo-Smith Kline, 5, GlaxoSmithKline, 5, GSK, 5, Horizon, 5, Lilly, 5, Lily, 5, Lycera, 1, 4, 5, Merck, 5, Novartis, 5, Pfizer, 5, Roche, 5, Samsung, 5, Samsung Bioepis Co., Ltd., 5, Sanofi Regeneron, 2, Sanofi/Regeneron, 2, Sanofi-Regeneron, 2, Scipher, 1, 4, 5, Set Point, 5, SetPoint, 5, Squibb, 5, Vorso, 1; K. Saag, Abbvie, 5, AbbVie, 5, Amgen, 2, 5, Ampel, 2, Bayer, 5, Gilead, 5, Horizon, 2, 5, Ironwood/AstraZeneca, 2, 5, Kowa, 5, kowa, 5, Mereo, 2, Radius, 5, Radius Health, 2, 5, Roche/Genentech, 5, SOBI, 2, 5, Sobi, 2, 5, Takeda, 2, 5, Teijin, 5, Tejin, 5; V. Bykerk, AbbVie, 5, Amgen, 1, 2, 3, 5, 8, Brainstorm Therapeutics, 1, 2, 3, 5, 8, Bristol-Myers Squibb, 5, Genentech, 5, Gilead, 5, NIH, 2, Pfizer, 1, 2, 3, 5, 8, Regeneron, 5, Regeneron Pharmaceuticals, Inc, 5, Sanofi, 5, Sanofi/Genzyme-Regeneron, 5, Sanofi-Genzyme/Regeneron, 1, 2, 3, 5, 8, Scipher, 1, 2, 3, 5, 8, The Cedar Hill Foundation, 9, UCB, 1, 2, 3, 5, 8, UCB Pharma, 5; C. Charles-Schoeman, Abbvie, 2, AbbVie, 2, Amgen, 5, BMS, 2, Bristol Myers Squibb, 2, Gilead, 5, Octapharma, 2, 5, Pfizer, 2, 5, Regeneron, 5, Regeneron/Sanofi, 5, Sanofi, 5; S. Fiore, Sanofi, 1, 3; G. St John, Regeneron, 1, 3, 4, Regeneron Pharmaceuticals, Inc, 1, 3; T. Kimura, Regeneron, 1, 3, Regeneron Pharmaceuticals, Inc, 1, 3; S. Zheng, Sanofi, 3, 5; C. Bingham, Abbvie, 5, AbbVie, 5, BMS, 2, 5, Bristol Meyer Squibb, 2, 5, Bristol Myers-Squibb, 2, 5, Bristol-Myers Squibb, 2, 5, Eli Lilly, 5, Eli/Lilly, 5, Genentech/Roche, 5, Janssen, 5, Janssen Research & Development, LLC, 2, Pfizer Inc, 5, Regeneron/Sanofi, 5, Sanofi/Regeneron, 5; G. Wright, AbbVie, 5, 8, Abbvie, 5, 8, Amgen, 5, 8, Autoimmune, 5, 8, BMS, 5, 8, Exagen, 5, 8, Lilly, 5, 8, Myriad, 5, 8, Myriad Autoimmune, 5, 8, Novartis, 5, 8, Pfizer, 5, 8, Regeneron, 5, 8, Sanofi Genzyme, 5, 8, UCB, 5, 8; M. Bergman, Abbvie, 5, 8, AbbVie, 5, 8, AbbVie, BMS, Celgene Corporation, Genentech, Janssen, Merck, Novartis, Pfizer, Sanofi, 5, AbbVie, Celgene Corporation, Novartis, Pfizer, Sanofi, 8, Amgen, 5, 8, BMS, 5, 8, Celgene, 5, 8, Genentech, 5, Genentech/Roche, 5, 8, Genentech-Roche, 5, Gilead, 5, GlaxoSmithKline, 8, GSK, 8, Horizon, 5, Janssen, 5, 8, JNJ (parent of Janssen), 1, JNJ stock, 1, Johnson & Johnson, 1, 4, Johnson and Johnson, 1, Merck, 5, 8, Novartis, 5, 8, Pfizer, 5, 8, Sandoz, 5, Sanofi, 5, 8, Sanofi/Regeneron, 5, 8, Sanofi-Regeneron, 5, 8; K. Nola, Coherus, 8, Gilead, 1, 5, Johnson & Johnson, 1, Proctor & Gamble, 1, Regeneron, 5, Sanofi Genzyme, 5; D. Furst, Actelion, 2, 5, Actelion Pharmaceuticals, 2, 5, Amgen, 2, 5, BMS, 2, 5, CME, 5, 8, Corbus, 2, 5, Galapagos, 2, 5, Galapogos Novartis, 5, GlaxoSmithKline, 2, GSK, 2, 5, NIH, 2, Novartis, 2, 5, Pfizer, 2, 5, Roche/Genentech, 2, 5, Sanofi, 2, 5; N. Shadick, BMS, 2, Crescendo Biosciences, 2, Mallinckrodt, 2, Sanofi Regeneron, 2, Sanofi/Regeneron, 2.

To cite this abstract in AMA style:

Curtis J, Weinblatt M, Saag K, Bykerk V, Charles-Schoeman C, Fiore S, St John G, Kimura T, Zheng S, Bingham C, Wright G, Bergman M, Nola K, Furst D, Shadick N. Exploring Heterogeneity in Rheumatoid Arthritis: Patient Profiling Through Principal Component and Cluster Analysis of the BRASS Registry [abstract]. Arthritis Rheumatol. 2019; 71 (suppl 10). https://acrabstracts.org/abstract/exploring-heterogeneity-in-rheumatoid-arthritis-patient-profiling-through-principal-component-and-cluster-analysis-of-the-brass-registry/. Accessed .

« Back to 2019 ACR/ARP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/exploring-heterogeneity-in-rheumatoid-arthritis-patient-profiling-through-principal-component-and-cluster-analysis-of-the-brass-registry/