Predicting Trends of Radiographic Outcomes Using Baseline Data

• Activity Number: 62
• Type: Topic Contributed
• Date/Time: Sunday, July 29, 2007 : 4:00 PM to 5:50 PM
• Sponsor: WNAR
• Author(s): Grace Park*+ and Weng-Kee Wong and Abdelmonem A. Afifi and David A. Elashoff and Dinesh Khanna and John T. Sharp and Richard H. Gold and Harold E. Paulus
• Companies: University of California, Los Angeles and University of California, Los Angeles and University of California, Los Angeles and University of California, Los Angeles and University of Cincinnati and University of Washington and University of California, Los Angeles and University of California, Los Angeles
• Keywords: cluster analysis ; K-medians ; rheumatoid arthritis ; multinomial outcomes

Abstract

Observational studies may lead to diverse visit times and methods of evaluating radiographic progression of joint damage in rheumatoid arthritis (RA). Instead of traditional progression rates for radiographic outcomes, we examined the usefulness of radiographic profile patterns through clustering algorithms to assess progression rates at set time intervals. Hands and feet radiographic scores were analyzed from a prospective RA cohort. Progression rates were determined by interpolating between set time intervals past the first radiographic observation. Patients were then grouped on their sets of progression rates by clustering algorithms (K-medians) based on Euclidean distances, and cluster membership was examined as a multinomial outcome measure regressed on baseline covariates. We conclude that patterns may correspond better with clinical status and treatment versus traditional rates.