Objective. To identify and describe different trajectories of physical activity over 2 years and to identify baseline predictors for each trajectory within a large cohort of people with rheumatoid arthritis (RA). Methods. The sample included 2,752 people ages 18-75 years and independent in daily living. Data on sociodemographic, disease-related, and psychosocial variables and physical activity (total weekly hours of vigorously/moderately intense activity and walking) were collected from registers and by questionnaires at baseline, and at 14 and 26 months. K-means cluster analysis was used to identify different physical activity trajectories. Multinomial logistic regression was used to identify predictors of trajectory membership. Results. Three trajectories were identified: 1 stable high (n 5 272; with an average of 25 hours of physical activity/week), 1 decreasing (n 5 564; changing from 22 to 8 hours), and 1 stable low (n 5 1,916; with an average of 3 hours). Predictors of the stable high trajectory versus the other 2 were male sex and already established physical activity at baseline. Predictors of the stable high trajectory versus decreasing trajectory were lower age and less social support for exercise, while predictors of stable high versus the stable low trajectory were less activity limitation and higher exercise self-efficacy. Conclusion. The results indicate that distinct trajectories of physical activity over 2 years in people with RA exist, and that stable, high physical activity is mainly predicted by male sex and already established physical activity at baseline. Additional predictors, amenable to change, include activity limitation, social support, and self-efficacy for exercise, suggesting a wide perspective on maintained physical activity in RA.