In Reply We thank Dr Schiappacasse and colleagues for their thoughtful comments on our article, and for raising the potential role of social engagement as a confounder or mediator of the observed associations. In our original analyses, we did not adjust for social engagement because comprehensive data on this construct were collected after each study’s baseline, specifically, beginning in 1992 in the NHS and in 1988 in the HPFS compared with baseline in 1980 and 1986, respectively. Including these variables would have substantially reduced follow-up duration and statistical power; therefore, we adjusted for marital status and living arrangements, which were available in the original baseline data.