Importance: Preclinical studies suggest that amylin has a U-shaped dose-response association with risk of Alzheimer disease (AD). The association of plasma amylin with AD in humans is unknown. Objectives: To measure amylin concentration in plasma by using enzyme-linked immunosorbent assay and to study the association between plasma amylin, incidence of AD, and brain structure in humans. Design, Setting, and Participants: This cohort study used data from the Framingham Heart Study offspring cohort from 1998 to 2015. Using a Monte Carlo approach, participants were divided into 3 plasma amylin concentration groups: (1) low (<75 pmol/L), (2) high (75-2800 pmol/L), and (3) extremely high (≥2800 pmol/L). Data analyses were conducted October 5, 2017, to December 18, 2018. Exposures: Baseline plasma amylin concentrations at examination 7. Main Outcomes and Measures: Incidence of dementia or AD and brain volumetric measures from structural magnetic resonance imaging data. Results: From the Framingham Heart Study offspring cohort, 3061 participants (mean [SD] age at baseline, 61.0 [9.5] years; 1653 [54.0%] women) who had plasma amylin measurements, dementia incidence, and brain volume measurements on record were included in this study. The distribution of plasma amylin concentrations was highly skewed (median [interquartile range], 7.5 [4.6-18.9] pmol/L; mean [SD], 302.3 [1941.0] pmol/L; range, 0.03-44 623.7 pmol/L). Compared with the low plasma amylin concentration group, the high plasma amylin concentration group had a lower rate of AD incidence (2.3% vs 5.6%; P = .04), but the extremely high plasma amylin concentration group had a higher rate of AD incidence (14.3%; P < .001). After adjusting for age, sex, education, body mass index, diabetes, cardiovascular disease, high-density lipoprotein level, and APOE4, high plasma amylin was not associated with decreased AD risk (hazard ratio, 0.42 [95% CI, 0.16-1.14]; P = .09) but was positively associated with volume of gray matter in the temporal lobe (β = 0.17 [SE, 0.05]; P < .001). In contrast, extremely high plasma amylin concentration was associated with a higher AD risk (hazard ratio, 2.51 [95% CI, 1.38-4.57]; P = .003) but not associated with temporal lobe volume (β = 0.02 [SE, 0.07]; P = .82). Conclusions and Relevance: This study found that plasma amylin concentration was associated with AD incidence and brain structure with a U-shaped pattern. These findings are consistent with preclinical findings that suggest amylin is a neuropeptide that is physiological; however, at extremely high concentrations, it may lead to amylin aggregation and therefore may be a risk factor for AD.