BACKGROUND: Altered autonomic modulation, measured by heart rate variability (HRV), has been found to be associated with dementia risk in the elderly. However, long-term follow-up study evaluating the association between autonomic modulation from middle-age and the incidence of dementia has been limited. METHODS: This retrospective cohort analyzed data from Taiwan\'s National Health Insurance Database covering the period from 2001 to 2017, with a linkage to citywide health examinations conducted by Tainan Metropolitan City, Taiwan. We included subjects aged 45-64 years. The mean follow-up period was 15.75 +/- 3.40 years. The measurements of HRV included resting heart rate, high frequency (HF), low frequency (LF), standard deviation of normal-to-normal R-R intervals (SDNN), ratio between the 30th and 15th R-R interval after standing up from the supine position (30/15 ratio), ratio between the R-R intervals during expiration and inspiration, and the ratio between the high- and low-frequency components (LF/HF). The main study outcome was the incidence of dementia. We performed multivariable Cox proportional hazard regression models to compare the risk of dementia among different HRV subgroups. RESULTS: We included 565 participants with a mean age of 53 (SD: 6) years, of whom 44% were male. The risk of dementia was significantly increased in association with lower parasympathetic HRV modulation, including SDNN (HR: 3.23, 95% CI: 1.55-6.73) and 30/15 ratio (HR: 3.52, 95%CI: 1.67-7.42). Moreover, the risk of dementia was increased in subjects with higher LF/HF ratios (HR: 2.05, 95% CI: 1.12-3.72). CONCLUSIONS: Lower parasympathetic activity and higher sympathetic-vagal imbalance in middle-age were associated with dementia risk.