Background. A previous companion paper to this report (Valenzuela & Sachdev, Psychological Medicine 2006, 36, 441–454) suggests a link between behavioural brain reserve and incident dementia; however, the issues of covariate control and ascertainment bias were not directly addressed. Our aim was to quantitatively review an independent set of longitudinal studies of cognitive change in order to clarify these factors.

Method. Cohort studies of the effects of education, occupation, and mental activities on cognitive decline were of interest. Abstracts were identified in MEDLINE (1966–September 2004), CURRENT CONTENTS (to September 2004), PsychINFO (1984–September 2004), Cochrane Library Databases and reference lists from relevant articles. Eighteen studies met inclusion criteria. Key information was extracted by both reviewers onto a standard template with a high level of agreement. Cognitive decline studies were integrated using a non-parametric method after converting outcome data onto a common effect size metric.

Results. Higher behavioural brain reserve was related to decreased longitudinal cognitive decline after control for covariates in source studies (ϕ=1·70, p<0·001). This effect was robust to correction for both multiple predictors and multiple outcome measures and was the result of integrating data derived from more than 47000 individuals.

Conclusions. This study affirms that the link between behavioural brain reserve and incident dementia is most likely due to fundamentally different cognitive trajectories rather than confound factors.