INTRODUCTION: Although mean physical activity in COPD patients declines by 400-500steps/day annually, it is unknown whether the natural progression is the same for all patients. We aimed to identify distinct physical activity progression patterns using a hypothesis-free approach and to assess their determinants. METHODS: We pooled data from two cohorts (usual care arm of Urban Training [NCT01897298] and PROactive initial validation [NCT01388218] studies) measuring physical activity at baseline and 12 months (Dynaport MoveMonitor). We identified clusters (patterns) of physical activity progression (based on levels and changes of steps/day) using k-means, and compared baseline sociodemographic, interpersonal, environmental, clinical and psychological characteristics across patterns. RESULTS: In 291 COPD patients (mean±SD 68±8 years, 81% male, FEV1 59±19%pred) we identified three distinct physical activity progression patterns: Inactive (n=173 [59%], baseline: 4621±1757 steps/day, 12-month change (Δ): -487±1201 steps/day), ActiveImprovers (n=49 [17%], baseline: 7727±3275 steps/day, Δ:+3378±2203 steps/day) and ActiveDecliners (n=69 [24%], baseline: 11 267±3009 steps/day, Δ: -2217±2085 steps/day). After adjustment in a mixed multinomial logistic regression model using Active Decliners as reference pattern, a lower 6-min walking distance (RRR [95% CI] 0.94 [0.90-0.98] per 10m, P=.001) and a higher mMRC dyspnea score (1.71 [1.12-2.60] per 1 point, P=.012) were independently related with being Inactive. No baseline variable was independently associated with being an Active Improver. CONCLUSIONS: The natural progression in physical activity over time in COPD patients is heterogeneous. While Inactive patients relate to worse scores for clinical COPD characteristics, Active Improvers and Decliners cannot be predicted at baseline.