Multiple system atrophy (MSA) is a rapidly progressive neurodegenerative disorder with different metabolic patterns in parkinsonian (MSA-P) and cerebellar (MSA-C) subtypes, but their longitudinal changes are not well understood. To characterize longitudinal changes of glucose metabolism and network connectivity in MSA using 18F-FDG-PET for early diagnosis and disease progression tracking, we retrospectively analyzed 29 MSA patients (17 MSA-P, 12 MSA-C) and 28 healthy controls. Regional glucose uptake was assessed as standardized uptake value ratios (SUVR). We examined inter-regional connectivity via correlation analysis and modeled metabolic decline using nonlinear mixed-effects models. Clinical progression was measured using the Unified Multiple System Atrophy Rating Scale (UMSARS). MSA-P displayed progressive hypometabolism in the putamen, cerebellum, and frontal cortex, while MSA-C showed declines primarily in the cerebellum and frontal regions. Longitudinal modeling indicated a faster putaminal decline in MSA-P (β = -0.015 ± 0.006) than in MSA-C (β = 0 ± 0.011), whereas cerebellar metabolism declined over time in both groups with overlapping slopes in MSA-P (β = -0.022 ± 0.006) and MSA-C (β = -0.022 ± 0.011). Regional metabolic reductions correlated with UMSARS progression (putamen in MSA-P: r = -0.59, p < 0.001; cerebellum in MSA-C: r = -0.62, p < 0.001). Significant connectivity disruptions were noted in frontal, basal ganglia, and cerebellar-parietal circuits. Longitudinal FDG-PET reveals distinct metabolic decline patterns in MSA subtypes-putamen in MSA-P and cerebellum in MSA-C-linked to clinical severity. These findings may inform clinical practice and trial design, supporting the use of FDG-PET for biological staging, monitoring disease progression, and potentially evaluating treatment responses.