Background: Depressive symptoms are common in multiple sclerosis (MS), yet there is little information about the pattern and predictors of changes in depressive symptoms over time. Objective: We examined changes in depressive symptoms over a 30-month period and the demographic, clinical and behavioral predictors of such changes in relapsing–remitting MS (RRMS). Methods: 269 persons with RRMS completed the Hospital Anxiety and Depression Scale (HADS) and a demographic/clinical scale, Godin Leisure-Time Exercise Questionnaire (GLTEQ) and Patient Determined Disease Steps (PDDS) scale every 6 months over a 30-month period. Data were analyzed using latent class growth modeling (LCGM). Results: LCGM identified a two-class model for changes in HADS depression scores over time. Class 1 involved lower initial status (i.e. fewer depressive symptoms) and linear decreases in depressive symptoms over time (i.e. improving HADS scores), whereas Class 2 involved higher initial status (i.e. more depressive symptoms) and linear increases in depressive symptoms over time (i.e. worsening HADS scores). LCGM further indicated that being older (OR = 2.46; p < .05), married (OR = 2.62; p < .05), employed (OR = 4.29; p < .005) and physically active (OR = 2.71; p < .05) predicted a greater likelihood of belonging to C1 than C2. Conclusions: Depressive symptoms change over time in persons with RRMS, and the pattern of change can be predicted by modifiable and non-modifiable factors.