Background: The heterogeneity of late-life depression hampers diagnosis and treatment. Data-driven methods have identified several subtypes of depression in older persons, but the longitudinal stability of these subtypes remains unknown. Methods: In total 111 older persons with a major depressive disorder both at baseline and 2-year follow-up from the Netherlands Study of Depression in Older persons (NESDO) were included. Latent class analysis was performed to identify subtypes of depression at baseline and at 2-year follow-up, and latent transition analysis was used to examine the stability of these subtypes over time. Transition rates between subtypes and characteristics of groups were examined. Results: Two subtypes were identified in both baseline (T0) and follow-up data (T1), including a ‘melancholic’ subtype (prevalence 80.2% (T0) and 62.2% (T1)), and an ‘atypical’ subtype (prevalence 19.8% (T0) and 37.8% (T1)). The melancholic subtype was characterized by decreased appetite and weight and had a stability of 0.86. The atypical subtype was characterized by increased appetite and weight and had a stability of 0.93, although the discriminating power of different symptoms had decreased at T1. Mean age and education differed significantly between stable and transitioning subgroups, other characteristics did not differ between subgroups. Limitations: Limited sample size might have hampered the analyses. Conclusions: Subtypes of late-life depression are relatively stable, but symptoms of depression (like weight loss) seem to blur with symptoms of (patho)physiological aging. This underlines the clinical relevance of depression subtyping, but also the importance of further research into subtypes and the influence of aging.