Chuang YF, An Y, Bilgel M, Wong DF, Troncoso JC, O'Brien RJ, Breitner JC, Ferruci L, Resnick SM, Thambisetty M. Midlife adiposity predicts earlier onset of Alzheimer's dementia, neuropathology and presymptomatic cerebral amyloid accumulation. Mol Psychiatry. 2015 Sep 1; PubMed. Correction.

Recommends

Please login to recommend the paper.

Comments

  1. This study addresses an extremely important topic and highlights how much uncertainty there still is about the impact of BMI across the life course on AD risk.

    The most relevant questions for most people interested in this research are: Would losing weight reduce my risk of developing AD or age of onset of AD? Would gaining weight increase my risk? In other words, would changing my behavior to lose or gain weight affect my AD risk? These questions are very difficult to answer because some people lose weight intentionally, while others lose weight because of underlying diseases that may influence AD risk. This paper did not find any association between change in BMI and age of onset of AD. 

    There is a puzzle here: Why is BMI in midlife associated with AD risk but changes in weight do not predict better outcomes (i.e., later onset)? We don't know the answer to that yet, but this study is a good step forward to addressing that critical next question. 

    The methods here are innovative. For example, the authors did not directly measure BMI at age 50 for most people in the study (the average age at baseline was 59 years). The authors used a clever statistical technique to "look back" and infer people's BMI at age 50 based on later measurements, though it should be noted that there is little evidence on whether this statistical technique performs as well as direct measurements.

    The autopsy and imaging data provide further novelty here, although these data are from a small sample of all the people in the study.

    View all comments by Maria Glymour

  2. We thank Dr. Glymour for her interest in our work and for the insightful comments above. Our main aim in the recently reported study was to relate midlife (defined here as 50 years of age) adiposity with the age at onset (AAO) of Alzheimer’s disease (AD). The AAO is a phenotype that has received relatively less attention compared to conventional measures of disease risk. For example, we recently reported on the relationship(s) between several novel genetic risk variants for AD and the AAO phenotype (Thambisetty et al., 2013).

    We believe that this is a measure that deserves more attention because understanding how risk factors for AD may accelerate disease onset is important if we are to test interventions that may delay symptom onset. This is, in fact, a stated goal of the National Plan to address Alzheimer’s disease.

    We agree with Dr. Glymour that testing associations between changes over time in BMI with AD onset is of great interest. It is also more challenging for some of the reasons she mentions. As individuals approach the average age of AD symptom onset, they are subject to numerous confounding influences on BMI. These include co-morbid medical conditions of old age which may influence either weight gain or weight loss and thereby affect AD onset. We were also mindful of the effect of accumulating AD and/or other neuropathology as individuals approach the average age of symptom onset, which could plausibly affect eating behavior, appetite, and body weight regulation. For these reasons, we chose to focus on a single measure of BMI at age 50 years in this analysis.

    Equally importantly, Dr. Glymour’s comments illustrate the need to define the optimal BMI threshold(s) in relation to AD risk and AAO. A change in BMI over time in relation to disease risk is perhaps more meaningful only if such changes are interpretable in terms of being “toward” or “away from” optimal BMI value(s). In our study, we were unable to establish such a threshold value of BMI because the relationship between midlife BMI and AAO across the available BMI values was linear in our sample. This is the reason we believe it is important to test these relationships in larger samples representing a wider range of BMI values.

    The definitive answer to whether or not such changes in BMI may alter the trajectory of AD may ultimately have to come from well-designed clinical trials with a large sample size over a sufficiently long follow-up time. This would be both expensive and time-consuming. Whether or not investing in such definitive clinical trials targeting midlife risk factors is worthwhile is perhaps a topic for another debate in itself.

    Lastly, Dr. Glymour raises an important point about estimated and “directly” measured BMI at age 50 years. In our sample of 1,394 participants in the Baltimore Longitudinal Study of Aging (BLSA), we had 8,061 actual BMI measurements (recorded every year or every two years) during the course of follow-up. Thus, each participant had, on average, almost six measurements of BMI throughout their participation in the study. The value of BMI at age 50 years is then derived from all the available, repeated BMI measurements, applying a linear mixed-effects model to the entire longitudinal sample to derive BLUE (Best Linear Unbiased Estimates) of BMI at 50 years of age. This method also allows us to assess the reliability of BMI estimates for each participant at age 50. Thus, we further accounted for the reliability of our BMI estimates in subsequent analyses by assigning greater weights to BMI values with higher reliability.

    View all comments by Yi-Fang Chuang

Make a Comment

To make a comment you must login or register.