Disease progression in patients can be highly variable following treatment initiation. One possible explanation is the existence of distinct disease activity/response trajectories influenced by individual patient variables at baseline, such as demographics, disease characteristics, socioeconomic status and health status. To identify distinct disease activity trajectories and distinguishing baseline factors over 12 months using data from controlled clinical trials, a group-based trajectory modelling strategy (Nagin, 2009) was applied to a disease-activity measure, to find unique longitudinal groups of patients with similar disease activity. Trajectory models are latent-variable models, fit as polynomials. The number of groups and polynomial degree of each group were specified and fit for all combinations of up to 6 groups and up to a 4th degree polynomial; a best-fit model was chosen using Bayesian information criteria. Data mining techniques are then applied to patient variables (baseline characteristics) to determine what can inform treating physicians in assessing in early treatment, that is, the likelihood of which trajectory a newly-treated patient will follow.