To identify studies for inclusion in the meta-analysis, we searched the PubMed and Scopus databases up to February 2012 and screened the reference lists of relevant articles for additional studies. We focused on prospective cohort studies, with five-factor model personality traits assessed at baseline in cognitively healthy participants who were evaluated at follow-up for incident AD. In cases with multiple publications from the same sample, we considered one effect for each trait from each sample. To reduce variability across studies, we generally chose the risk estimates from the main model, with age, sex, education, and ethnicity as covariates. The logHR and SE were scaled in each study to correspond to the effect associated with 1 SD difference on the trait. We performed random-effect model meta-analyses. Heterogeneity was evaluated using the Q statistic, and publication bias was evaluated statistically with the Kendall’s tau and Egger test. The meta-analyses were conducted using the “Comprehensive Meta-Analysis” software package.
