Au et al. Respond to “Body Mass Index and Risk of Dementia”

We appreciate the opportunity to respond to Brenowitz’s astute commentary (1) on our recent paper (2), and would like to further highlight some key points.

FROM PHENOMENON TO MECHANISMS

As noted by Brenowitz, few studies have investigated the impact of body mass index (BMI) on dementia risk across the mid- to late life course in the same subjects. Over 38 years of follow-up of Framingham Offspring participants, we were able to observe changing trends in the association between BMI and dementia, from a positive association in midlife to an inverse trend in late life (2). But as Brenowitz pointed out, inherent to any observational study is the limitation of determining the actual mechanisms underlying these associations. We concur that further clinical or laboratory studies are warranted to identify causal pathways and elucidate the directionality of the relationship along the age continuum. In our own discussion, we used findings from previous studies to suggest that one potential trigger of the neurodegenerative processes at an earlier age is mediated through vascular and dysmetabolic pathways and/or through cell-signaling proteins secreted by the adipose tissue. Brenowitz’s suggestion of considering Mendelian randomization or mediation analyses to further help triangulate evidence around BMI, obesity, and dementia is insightful and offers an opportunity to further capitalize on ways observational studies can be further leveraged beyond simply decades of prospective study.

FROM DIFFERENT ASSOCIATIONS BETWEEN BMI AT DIFFERENT AGE GROUPS AND DEMENTIA TO THE EFFECT OF BMI TRAJECTORY ON DEMENTIA

Our core finding that the relationship between BMI and dementia risk is heterogeneous across the adult age range, along with Brenowitz’s commentary, reinforces the need for future studies to confirm at what stages in the adult life course obesity increases risk for dementia and at what age weight loss due to dementia pathologies begins. We contend that perhaps the starting point should not be restricted to adult age. The Lancet Commission (3) provides a model that maps dementia risk factors across the entire life span but is limited in its observations based on research that has been reported. This raises the issue that few studies exist that span the entire adult age span in which weight history has been recorded. Further, the Framingham Heart Study stands virtually alone as the only cohort study that includes participants who have been followed beginning at young to middle adult age for dementia as a research area of interest. Thus, there is the pragmatic reality of how to conduct the research that is recognized and needed. One suggestion of how to do so is to consider how to tap the electronic health records that include weight/height history data as a means of retroactively constructing a prospective study that can span far more years of follow-up than current dementia study designs allow. Another is to tap longitudinal studies that started with young children and add dementia-related ancillary studies, similar to what was done with the Framingham Heart Study. The Bogalusa Heart Study (4, 5) is one such example in which investigators have started to add dementia-related endophenotypes to expand their initial focus beyond cardiovascular disease (6, 7). As to our own work, we seek to apply more novel machine-learning analytical methods to provide a different approach to examine fluctuating trajectories of BMI. In doing so, we hope to further elucidate which specific patterns of BMI changes are associated with increased dementia risk and which are not. Clearly, the larger take-home message from this study highlights the importance of the availability of life-course data in shedding light on the pathogenesis of risk factors in the primary prevention of dementia.