Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
Observational Study
, 102 (7), 2310-2320

Body Mass Index and Risk of Alzheimer's Disease: A Mendelian Randomization Study of 399,536 Individuals

Affiliations
Observational Study

Body Mass Index and Risk of Alzheimer's Disease: A Mendelian Randomization Study of 399,536 Individuals

Liv Tybjærg Nordestgaard et al. J Clin Endocrinol Metab.

Abstract

Context: Recently, data on 2,000,000 people established that low body mass index (BMI) is associated with increased risk of dementia. Whether this observational association reflects a causal effect remains to be clarified.

Objective: We tested the hypothesis that there is a causal association between low BMI and high risk of Alzheimer's disease.

Design, setting, and participants: Using a Mendelian randomization approach, we studied 95,578 individuals from the Copenhagen General Population Study (CGPS) with up to 36 years of follow-up and consortia data on 303,958 individuals from the Genetic Investigation of Anthropometric Traits (GIANT) and the International Genomics of Alzheimer's Project (IGAP).

Main outcome measure: Risk of Alzheimer's disease.

Results: The causal odds ratio for a 1-kg/m2 genetically determined lower BMI was 0.98 [95% confidence interval (CI), 0.77 to 1.23] for a weighted allele score in the CGPS. Using 32 BMI-decreasing variants from GIANT and IGAP the causal odds ratio for Alzheimer's disease for a 1-standard deviation (SD) lower genetically determined BMI was 1.02 (95% CI, 0.86 to 1.22). Corresponding observational hazard ratios from the CGPS were 1.07 (95% CI, 1.05 to 1.09) and 1.32 (95% CI, 1.20 to 1.46) for a 1-kg/m2 and a 1-SD lower BMI, respectively.

Conclusions: Genetic and hence lifelong low BMI is not associated with increased risk of Alzheimer's disease in the general population. These data suggest that low BMI is not a causal risk factor for Alzheimer's disease and that the corresponding observational association likely is explained by reverse causation or confounding.

Figures

Figure 1.
Figure 1.
Risk of Alzheimer’s disease as a function of BMI in quartiles in the CGPS. Individuals with Alzheimer’s disease before blood sampling were excluded, leaving 95,526 individuals for this analysis. Multifactorial adjustment was for age (as time scale), sex, hypertension, diabetes, smoking, alcohol intake, physical inactivity, postmenopausal status and hormonal replacement therapy in women, lipid-lowering therapy and educational level. P for trend from Cox regression trend test.
Figure 2.
Figure 2.
BMI (left) and corresponding theoretically predicted (middle) and observed (right) hazard ratios for Alzheimer’s disease as a function of BMI weighted allele score/simple allele score group. Theoretically predicted hazard ratios were calculated from delta BMI and the known association of BMI with risk of Alzheimer’s disease in the observational study (see Fig. 1). Observed hazard ratios were multifactorially adjusted for age (as time scale), sex, hypertension, diabetes, smoking, alcohol intake, physical inactivity, postmenopausal status and hormonal replacement therapy in women, lipid-lowering therapy, and educational level. As a positive control, estimates were also presented for type 2 diabetes (lower part of Fig. 2; multifactorial adjustment omitted diabetes as a covariate). BMI values are mean (±standard error of the mean). P values are test for trend.
Figure 3.
Figure 3.
Association of potential confounders with BMI, Alzheimer’s disease, and BMI weighted allele score/simple allele score group. Potential confounders were dichotomized: age (≥70 vs <70); sex (male vs female), hypertension (hypertension vs no hypertension), physical activity (high vs low physical activity), education (short, <8 years, vs long education), and menopausal status (postmenopausal vs premenopausal). Potential confounders were defined as described in the Table 1 footnote. For each potential confounder, logistic regression analysis was used to calculate sex- and age-adjusted odds ratios and P values for, respectively, a 1-kg/m2 lower observational BMI, Alzheimer’s disease vs no event, and a 1-kg/m2 genetically lower BMI for weighted/simple allele score group.
Figure 4.
Figure 4.
Risk of Alzheimer’s disease for a 1-kg/m2 lower observational and causal genetically determined BMI. The hazard ratio for a 1-kg/m2 lower observational BMI was calculated using Cox regression, whereas the odds ratio for genetically low BMI was derived from an instrumental variable analysis. P value: significance of hazard ratio or odds ratio. F, strength of the genetic instrument (>10 indicates sufficient strength); R2, percent contribution of genetic instrument to the variation in BMI.
Figure 5.
Figure 5.
Risk of Alzheimer’s disease for a 1-SD lower observational and causal genetically determined BMI. The hazard ratio for a 1-SD lower observational BMI was calculated using Cox regression, whereas the odds ratio for genetically low BMI was derived from the inverse-variance weighted method (upper panel). Estimated effects on Alzheimer’s disease risk are plotted against estimated effects on BMI (percent of 1 SD on the log scale) for 32 genetic variants (lower panel). Vertical gray lines show 95% CIs for each individual genetic variant. We graphically illustrated the overall causal estimate shown as a red solid line with 95% CIs as red dashed lines (lower panel). The regression coefficient of –0.021 represents the overall causal estimate. P value: significance of hazard ratio or odds ratio.

Similar articles

See all similar articles

Cited by 7 PubMed Central articles

See all "Cited by" articles

Publication types

MeSH terms

Feedback