A key uncertainty in the obesity-mortality association continues to be how this association changes over the life course. Prior studies tend to rely on cross-sectional design with static weight status taken at the time of the survey. This study tracks a cohort of individuals and employs lifelong body mass index information from the Framingham Heart Study original cohort (1948-2010). We focus on respondents who were younger than age 45 at time of their first survey (n = 2,176) and evaluate how the mortality risk associated with obesity changes over three age groups (below 45, 45-59, and 60 and above) and how population compositions may contribute to this pattern. We find the hazard ratio associated with obesity compared to normal weight decreases over three age groups, but this pattern is influenced by different ages of onset of obesity, inconsistency in the reference group (normal weight) over ages, and mortality selection effects. These factors explain away the decreasing effect of obesity (with onset before age 45) on mortality up to age 60; after age 60, the detrimental effect still declines, but to a much less degree. Later onset of obesity, however, is not significantly associated with excess mortality risks after age 60.