To explore the relationship between plasma amyloid-β (Aβ) and cognitive decline during 2 year follow-up in a population-based cohort in Xi'an rural areas. The study was conducted in Qubao village in Xi'an suburbs cognitively normal residents over 40 years old were recruited from October 2014 to March 2015 and given a face-to-face standardized interview. Mini-mental state examination (MMSE) was employed to evaluate the global cognitive function, and quantification of plasma Aβ was measured by sandwich enzyme-linked immunosorbent assay (ELISA) at baseline. Two years later, MMSE was tested at the end of study. Then logistic regression was performed to analyze the relationship between baseline Aβ and cognitive change during 2 year follow-up. A total of 1 020 participants completed the study, among whom 223 subjects (21.9%) presented MMSE scores decline (defined as MMSE scores decreased ≥2 points). Compared with those without decline, participants in the MMSE decline group were older (<0.001) and had lower education level (<0.001), while gender, hypertension, hyperlipemia, diabetes mellitus and APOE genotype were not significantly different between two groups. One-way analysis of variance (ANOVA) showed that the MMSE score decline was slighter in the lower tertile of baseline Aβ(1)-40 compared with middle tertile (=0.012), while MMSE decline were similar between different Aβ(1)-42 level groups and Aβ(1-42)/Aβ(1-40) ratio groups (=0.758, =0.671, respectively). Multivariable logistic regression analysis showed that MMSE scores in the lower baseline plasma Aβ(1-40) level declined more slowly (=0.565, 95% 0.379-0.845, =0.005). However, the MMSE decline were also similar among different baseline plasma Aβ(1-42) levels groups and Aβ(1-42)/Aβ(1-40) ratio groups. Population with lower level of baseline plasma Aβ(1-40) manifests lower cognitive decline during 2 years, however further investigation on dynamics of plasma Aβ and long term follow up are needed.