Sedentary behavior is defined as waking behaviors involving energy expenditure of 1.5 metabolic equivalent units or less while in a sitting or reclining posture and was determined using a machine learning–based analysis of 1 week of wrist-worn accelerometer data. A, Histogram shows participant counts across the range of mean daily sedentary behavior and vertical lines represent dementia cases. B, Model is fully adjusted (see Methods). The reference value (hazard ratio [HR] = 1; dotted horizontal line) was set by the median exposure variable (9.27 hours/d for sedentary behavior time) and the HRs are plotted on a log scale. The shaded areas reflect the 95% CIs for the HRs. The model depicted in part B was adjusted for age, sex, education, Townsend Deprivation Index, presence of APOE ε4 allele, ethnicity, chronic conditions (heart or vascular disease, diabetes, or cancer), self-reported health, smoking status, alcohol consumption, diet, body mass index, self-reported depression, and time spent engaged in moderate to vigorous physical activity.