Patterns of radiographic outcomes in early, seropositive rheumatoid arthritis: a baseline analysis

We examine radiographic profile patterns using clustering algorithms to assess progression rates at set time intervals in a rheumatoid arthritis (RA) observational study. Hands/feet radiographic scores were analyzed for 190 early, seropositive RA patients with ≥ 3 radiographic observations from a prospective cohort. Assessments at 6 months, 1 year, and yearly thereafter were requested for demographic, therapeutic, functional, laboratory, radiographic, and clinical data. Progression rates for the total sharp scores [erosion (E)+joint space narrowing (JSN)] were interpolated for intervals of 0 to 6 months, 6 month-1 year, 1-2 years, and 2-3 years past first radiographic observation. Patients were grouped on their sets of rates by K-median clustering algorithms, and categorical group membership was regressed onto baseline characteristics using multinomial models. The number of clusters was determined using one-way MANOVA, and baseline differences across clusters by Kruskal-Wallis tests. The median RA duration was 6.1 months, mean age 52 years, median disease activity score (DAS) 4.6, mean radiographic observations 4.6 (range 3-8) for this mostly female (77%), Caucasian (78%) sample. 3 patterns were determined: increasing (n = 41; 22%), increasing then decreasing (n = 41; 22%), and flat (n = 108; 57%). High baseline C-reactive protein was associated with a worsening radiographic progression (p < 0.005), as were HAQ-DI (p = 0.07), JSN (p < 0.01), and E (p = 0.03). Our conclusions are that radiographic progression patterns graphically supplement traditional linear rates, and are flexible to use in both clinical and observational studies. The identified clusters and rates may correspond better with clinical status and treatment over the disease course than linear progression rates alone.