Investigating Causal Relations Between Risk Tolerance,Risky Behaviors,and Alzheimer’s Disease: A Bidirectional Two-Sample Mendelian Randomization Study

Abstract

Background:

Several studies have shown risky behaviors and risk tolerance are associated with Alzheimer’s disease. However, the underlying causality remains unclear. Risky behavior and risk tolerance may induce the onset of Alzheimer’s disease, and/or vulnerability to Alzheimer’s disease may result in more risky behaviors.

Objective:

To examine bidirectional relationships between risky behavior, risk tolerance, and Alzheimer’s disease using Mendelian randomization method for assessing potential causal inference.

Methods:

This bidirectional two-sample Mendelian randomization study used independent genetic variants associated with risky behaviors and risk tolerance (n = 370, 771– 939, 908), and Alzheimer’s disease (n = 71, 880 – 37, 613) as genetic instruments from large meta-analyses of genome-wide association studies.

Results:

Our results support a strong protective casual effect of risk-taking tendency on AD (odds ratio [OR] = 0.79; 95% CI, 0.67– 0.94, p = 0.007). There was weak statistically significant relationship between number of sexual partners and AD (OR = 0.50, 95% CI, 0.27– 0.93, p = 0.04), and between family history of AD and automobile speeding propensity (OR = 1.018, 95% CI, 1.005 to 1.031; p = 0.007). Contrary to expectations, there was no statistically significant causal effect of AD on risk-taking tendency (β= 0.015, 95% CI, – 0.005 to 0.04; p = 0.14).

Conclusion:

Under Mendelian randomization assumptions, our results suggest a protective relationship between risk-taking tendency and the risk of AD. This finding may provide valuable insights into Alzheimer’s disease pathogenesis and the development of preventive strategies.

Keywords

Alzheimer’s disease causality mendelian randomization analysis risk behavior

INTRODUCTION

Alzheimer’s disease (AD) is the most common form of dementia in the elderly, and it is pathologically characterized by the accumulation of intracellular tangles containing tau, and extracellular plaques containing amyloid-β (Aβ) [1]. Although some progress has been made in the studies focus on the biological mechanisms and symptoms improvements, there are no curative or protective treatments available to date [2].

Lifestyle exposure factors play a pivotal role in AD [3 –5]. Large scale case-control and cohort studies have shown healthy behaviors, such as frequent participation in cognitively stimulating activities and regular physical activity, can produce protective effects on AD [6]. Meanwhile, few studies have demonstrated that risky behaviors and risk tolerance, such as smoking and drinking, are positively associated with the development of AD [7 –9]. However, whether causality exists remains to be determined due to the nature of study design and confounding bias.

Mendelian randomization (MR) is an approach that can investigate the causal effects between potentially modifiable risk factors and the outcomes. Because this method relies on the use of genetic variants that are randomly allocated before birth, it can reduce the susceptible to measurement errors and minimize the effects of residual confounding and reverse causation compared with conventional multivariable regression approaches [10]. In the field of AD, recent MR studies have provided support that healthy lifestyles have a direct causal relationship with AD [11 –13]. Yet, these studies are almost entirely focused on the modified environment factors and general lifestyles, while rarely looking into the role of specific behavioral phenotypes. In an attempt to fill the evidence gap, we aim to investigate the potential causal relationship between risky behaviors and the risk for AD using two-sample bidirectional MR analysis.

MATERIALS AND METHODS

This study utilized publicly available de-identified genome-wide association study (GWAS) summary statistical data; the ethical approval had been obtained in all original studies. A total of 3, 976, 961 participants were available for the summarized data; the corresponding sample size of GWAS in each study is shown below. Data were analyzed from January 29, 2019 through March 31, 2020.

Data sources and instruments

We conducted MR using a recently published meta-analysis of GWAS on risky behaviors and risk tolerance [14] in the UK Biobank and 23andMe Study participants. This study examined the following 6 highly correlated risky behavior phenotypes and 1 principal component (PC) of four of risky behavior phenotypes: 1) self-reported general risk tolerance (n = 939, 908); 2) adventurousness (defined as the self-reported tendency to be adventurous versus cautious, n = 557, 923); 3) automobile speeding propensity (the tendency to drive faster than the speed limit, n = 404, 291); 4) drinks per week (the average number of alcoholic drinks consumed per week, n = 414, 343); 5) ever smoker (whether one has ever been a smoker, 518, 633); 6) number of sexual partners (the lifetime number of sexual partners, n = 370, 711); and 7) risk-taking tendency (the first PC of four risky behaviors, including automobile speeding propensity, drinks per week, ever smoker, number of sexual partners, n = 315, 894). The GWAS of this study identified statistically significant associations of 611 independent loci, which is the mainly single-nucleotide polymorphisms (SNPs) that we used for subsequent MR analysis.

We manually checked all the identified SNPs that are directly associated with AD by a literature retrieval and using PhenoScanner GWAS database (http://www.phenoscanner.medschl.cam.ac.uk/) [15, 16], excluding criteria included: 1) SNP that is significantly associated with AD (p < 5×10^–8), and 2) loci that has a probable role in the pathogenesis of AD. We also used LDlink (https://ldlink.nci.nih.gov/) API, including LDpair and LDproxy [17, 18], to exclude one of the variants for the linkage disequilibrium (LD, R² > 0.2) and find the proxy SNPs (R² >0.9) that are not available in the AD GWAS dataset. The strength of the genetic instrument was judged by F statistics [19]. We calculated the effect size and sample size of each individual SNPs [20] and excluded the SNPs that have F statistics less than 10 according to standard practice (Fig. 1). The quality control of genetic instruments also applies to AD. The list of instrument SNPs for risk-taking tendency is given in Supplementary Tables 1 to 2.

Fig. 1

A Flow Diagram for the Process for Identifying Genetic Variants to be included in MR Analysis. AD, Alzheimer’s disease; MR, mendelian randomization; F-stat, F statistics.

Alzheimer’s disease

We utilized the GWAS summary statistics from the largest and most recent dataset from Psychiatric Genomics Consortium (PGC) for AD. In this dataset, the phenotype is defined by AD or AD-by-proxy, which means subject with at least one parent who is diagnosed as AD is qualified. The value of by-proxy phenotypes for GWAS was recently demonstrated by Liu et al. for 12 common diseases, including substantial gains in statistical power for AD [21, 22]. This discovery cohort of GWAS identified totally 31 independent loci, 18 of them are significant for AD (n = 79, 145), while left 13 are significant for AD-by-proxy (n = 376, 113). We also drew on another two-large famous GWAS summary statistics datasets from International Genomics of Alzheimer’s Project (IGAP) and The Centre for Cognitive Ageing and Cognitive Epidemiology at the University of Edinburgh (CCACE) to verify our results [23, 24]. The IGAP is a large two-stage study based upon GWAS on individuals of European ancestry. The CCACE is a center of excellence to advance research into how ageing affects cognition (cognitive aging), and how mental ability in youth affects health and longevity (cognitive epidemiology). As instruments, we applied the quality control as mentioned before. The list of instrument SNPs for phenotype is given in Supplementary Tables 3 and 4.

Statistical analysis

We adopted the two sample MR as our main statistical methods. This method using two different study samples to estimate associations between the instrument-exposure factor and instrument-outcome factor, in order to estimate a causal effect of the exposure on the outcome [25]. This can be useful when the exposure and outcome are not available in one independent sample. Our research conformed to this requirement. Another similar approach is single sample MR, which uses one dataset in the instrumental variable analysis to yield the causal estimate of the exposure on the outcome.

We used the latest functions and packages (TwoSampleMR) in R to conduct MR analysis. First, we extracted the SNPs being used to instrument the exposure from the outcome GWAS. We then performed harmonization of the direction of SNPs between exposure and outcome associations, and we excluded the strand-ambiguous SNPs with intermediate effect allele frequencies (>0.3). The sensitivity analysis was also applied using the data above.

We chose the inverse variance-weighted (IVW) fixed-effect method as our primary method for each direction MR analysis of causal effects. This method essentially assumes the intercept is zero and associated a weighted regression of SNP-exposure effects with SNP-outcome effects. It will return an unbiased estimate in the absence of horizontal pleiotropy, or when horizontal pleiotropy is balanced [26]. Since the functional pathways and proportion of pleiotropic genetic instruments were not clear, we compared IVW results with other MR methods, including MR-Egger, weighted median, simple mode-based, and weighted mode-based methods in order to make different assumptions at the cost of reduced statistical power [27]. The MR-Egger method assumes no intercept term in the model and provides less biased results when there are directional pleiotropy and considerable heterogeneity [27]. The weighted mean approach selects the median MR estimate as a casual estimate [28]. We reported β values when the outcome is continuous (i.e., the first principal component (PC) of four risky behaviors) and converted to ORs when the outcome was dichotomous (i.e., AD status).

We used the Cochran Q statistic, leave-one-out sensitivity analysis, and MR Egger intercept test of deviation from the null as main methods to identify pleiotropic variants [29]. These tests aimed to capture the SNPs that may influence the outcome through an unaccounted causal pathway.

RESULTS

Causal effects of risk tolerance and risky behaviors on AD

We found evidence of a protective causal effect of risk-taking tendency on the risk of AD (IVW OR, 0.79 for family history of AD per 1-SD increase in first principal component; 95% CI, 0.67– 0.94, p =0.007) (Table 1). We verified the results in another large dataset and yield similar patterns of effects (Supplementary Table 5). The number of sexual partners also showed causal effects on reduced risk of AD (MR-Egger OR, 0.50 for AD per 1-SD increase in lifetime number of sexual partners; 95% CI, 0.27– 0.93, p = 0.038). However, this causal relationship could not be observed in another dataset and other MR methods. The Cochran Q statistic for risk-taking tendency (Q = 19.82; p = 0.76) and number of sexual partners (Q = 22.72; p = 0.70) indicated no notable heterogeneity across instrument SNP effects. The MR Egger intercept test for risk-taking tendency (intercept, – 0.006; standard error, 0.005; p = 0.23) suggested no horizontal pleiotropy, while the number of sexual partners (intercept, 0.013; standard error, 0.0056 p = 0.03) showed that the IVW estimate for the number of sexual partners might be biased.

Table 1

MR Results for the Casual Relationship Between Risk-Taking Tendency and AD

Method	OR (97% CI)^a	p	No. of SNPs
IVW	0.79 (0.67– 0.94)	0.007	26
Weighted median	0.80 (0.63– 1.02)	0.07	26
MR Egger	1.02 (0.66– 1.58)	0.93	26

IVW, inverse variance– weighted; AD, Alzheimer’s disease; MR, mendelian randomization; OR, odds ratio; SNP, single-nucleotide polymorphism. ^aIndicates odds for Family history of Alzheimer’s disease per 1-SD increase in first principal component.

By using literature retrieval and API provided by PhenoScanner GWAS, we found none of the SNPs, including 26 significant SNPs of risk-taking tendency (Fig. 2) and 29 significant SNPs of number of sexual partners, were associated with AD or had any probable role in the pathogenesis of AD.

Fig. 2

MR Plots for Casual Effects of Risk-Taking Tendency on AD. A) Scatterplot of single-nucleotide polymorphism (SNP) potential effects on Risk-Taking Tendency versus AD, with the slope of each line corresponding to estimated MR effect per method. B) Forest plot of individual and combined SNP MR-estimated effects sizes. Data are expressed as raw β values with 95% CI. p < 5×10^–8 for top SNPs.

Other risky behaviors, including general risk tolerance (IVW OR, 1.16 for family history of AD per 1-SD increase in general risk tolerance; 95% CI, 0.87– 1.52, p = 0.28), automobile speeding propensity (IVW OR, 0.82 for family history of AD per 1-SD increase in automobile speeding propensity; 95% CI, 0.58– 1.15, p = 0.25), drinks per week (IVW OR, 0.99 for family history of AD per 1-SD increase in drinks per week; 95% CI, 0.78– 1.26, p = 0.94) and ever smoker (IVW OR, 0.90 for family history of AD per ever versus never smokers; 95% CI, 0.76– 1.06, p = 0.20), were analyzed with AD or AD-by-proxy, and we also found no evidence of causal relationships

Causal effects of AD on risk tolerance and risky behaviors

In this direction, we found some characteristics but weak causal relationships between AD and risky behaviors. We did not find any causal relationship of AD with risk-taking tendency (β= 0.015 in first principal component per AD versus control status; 95% CI– 0.005 to 0.04; p = 0.14). However, AD had a causal effect on speeding (IVW OR, 1.018 in automobile speeding propensity increase per family history of AD versus control status; 95% CI, 1.005 to 1.031; p = 0.007) at very weak level (Table 2, Fig. 3). Weighted median and MR Egger analysis yielded similar pattern of effects. There were some week casual relationships between AD and other risky behaviors and risk tolerance, including number of sexual partners (β= 0.032 in number of sexual partners per AD (late onset) versus control status; 95% CI, 0.0016 to 0.063; p = 0.04), drink per week (β= – 0.0002 in drink per week per family history of AD versus control status; 95% CI, – 0.0003 to – 4.26×10^–5; p = 0.01) and ever smoker (β= – 0.002 in ever smoker per AD (late onset) versus control status; 95% CI, -0.004 to – 0.0006; p = 0.04).

Table 2

MR Results for the Causal Relationship Between AD and Automobile Speeding Propensity

Method	OR (97% CI)^a	p	No. of SNPs
IVW	1.016 (1.005– 1.029)	0.007	4
Weighted median	1.017 (1.004– 1.031)	0.012	4
MR Egger	1.01 (0.99– 1.03)	0.34	4

IVW, inverse variance– weighted; AD, Alzheimer’s disease; MR, mendelian randomization; OR, odds ratio; SNP, single-nucleotide polymorphism. ^aIndicates odds for automobile speeding propensity increase per Family history of Alzheimer’s disease versus control status.

Fig. 3

MR Plots for Casual Effects of AD on Automobile Speeding Propensity. A) Scatterplot of single-nucleotide polymorphism (SNP) potential effects on AD versus automobile speeding propensity, with the slope of each line corresponding to estimated MR effect per method. B) Forest plot of individual and combined SNP MR-estimated effects sizes. Data are expressed as raw β values with 95% CI. p < 5×10^–8 for top SNPs.

DISCUSSION

Recent studies have suggested that a healthier lifestyle may protect against the risk for AD [30]. However, insufficient data is available regarding the causal relationship between risky behaviors and AD. By applying MR analysis method with the genetic instruments selected from a large-scale GWAS database, our study demonstrate that risk-taking tendency have a protective role on AD. The sensitivity analysis has enhanced the robustness of the causal relationship in the absence of any detectable pleiotropic effect. Another MR analysis in an independent dataset has also validated our results. On the other hand, several observational studies and mendelian analyses have also reached the same conclusions, in which drinks per week and ever smoker have no causal effects on AD [31] [11]. The R-squared value for each exposure also located at the reasonable section, which is from around 0.2 to 0.8 (Supplementary Table 6). It can increase the power of the statistical analysis.

As previously described, the risk-taking tendency is the first PC of four risky behaviors. It is originally interpreted as the willingness to take action without the consideration of negative consequences. Here, we differed risk-taking tendency from concrete risky behaviors, such as drinking and smoking, and expanded its meaning to relate with irrational decision-making and low self-control ability. In the aspect of AD, Seo et al. have reported that poor demanding inhibition and goal-directed behaviors could be the characteristics of the very early cognitive sign in the course of AD[32]. By the use of Iowa gambling task, Sinz et al. have also found the decision-making under ambiguity and decision-making under risk were impaired in patients with mild AD, characterized by a lack of advantageous strategies [33]. Although such behaviors may be attributed to the neuropathological changes in the ventromedial prefrontal cortex and amygdala [34], the relationship, however, has not proved to be straight forward. It is possible that that risky behaviors may be the result of hormesis-based adaptive neuronal response for maintaining functional connectivity and neuroplasticity [35, 36]. Nevertheless, further research is needed to gain more clarity.

Our results provide several contributions to current literature. First, in the bidirectional MR analysis between risking-taking tendency and AD, we observed only 1 direction of this relationship, where risk-taking tendency demonstrated a potential casual effect on AD, which cuts down the possible inverse association between them and offers more precise information on the protective effect of risking-taking tendency on AD. Second, the causal relationship between AD and risky behaviors, risk tolerance, provided some weak but no least evidence that AD may cause all risky behaviors that we included. Since the β value for each risky behavior is relatively small, it meant that AD played a small casual effect on these behaviors. Generally speaking, people with these risky behaviors are often at a young age [37]. To explain these, we inferred that the vulnerability to AD might have causal effects on these risky behaviors for the young. To the best of our knowledge, there is no epidemiologic population-based study exploring the relationship between risky behaviors and vulnerability to AD. Therefore, our MR result interpretation is more conjectural and less evidence-based and needed more independent line of validation of this association to be done to confirm it.

Limitations

This study has several limitations. First, since we lack the individual-level data and the definite information of potential confounders, we cannot evaluate the direct association of individual genetic variants with potential confounders between risky behaviors and AD. Second, our analysis was based on several GWAS datasets. There is a potential that underlying overlap between samples might bias the results in several MR methods. In our datasets, the samples of risky behaviors are completely from UK Biobank, while the samples of AD are from several different cohort, including UK biobank and IGAP. There is around 80% overlap. However, IVW method adopts the second-order weights which can address this bias and the limitation can have little impact on our results. Third, we only selected the significant SNPs as the exposure genetic instruments, which made these SNPs difficult to explain the total variation for complex exposure traits. Moreover, even we used the up-to-date literature retrieval, biological effects of these SNPs were not available. As a result, we were not able to fully filer out pleiotropic mechanisms of these loci. Forth, although genetic variants could exhibit remarkable heterogeneity about functional activity in different ethnic and racial backgrounds, the GWAS dataset we used was mainly derived from European ancestry [38]. Thus, the present findings may not be applicable to other ethnic-racial groups. Fifth, some of our results were different between two datasets and in different methods. Except the above reasons, the difference between standard of AD diagnosis in the two can also explain it. Finally, our study showed the causal relationships between risky behaviors and AD, but not for AD symptoms. This restricts the clinical application of our results. On the other hand, it will be the future direction to explore the causal effect of risk-taking tendency and AD symptoms, such as memory complaints.

Despite these limitations, to our knowledge, this is the first and most comprehensive MR analysis for risky behavior phenotypes, risk tolerance, and AD. By adopting a strict sensitivity analysis [25], including literature searches for potential pleiotropic variants, filter out SNPs in direct action loci, we were able to provide strong evidence of causal protective role on risk-taking tendency on AD. It is of great importance to find this because few modifiable factors are known to be protective for AD. Our study should not be interpreted as an encouragement for risky behaviors, but rather as new insights for further studies.

Conclusions

In the study, we utilized MR to find potential positive factors for AD. Our findings provide strong evidence for a causal effect of risk-taking dependency on AD. Meanwhile, we also found that the vulnerability to AD may result in some risky behavior. Nonetheless, these findings have potential implications as an effective prevention strategy for AD and reveal some underlying mechanisms of the characteristics in AD patients.

Authors’ disclosures available online (https://www.j-alz.com/manuscript-disclosures/20-0773r3).

Footnotes

ACKNOWLEDGMENTS

This study was supported by grants from the Shanghai Technology and Science Key Project in Healthcare (No. 17441902100), National Natural Science Foundation of China (91849126), Shanghai Municipal Science and Technology Major Project (No.2018SHZDZX01) and Zhangjiang Lab, Tianqiao and Chrissy Chen Institute, and the State Key Laboratory of Neurobiology and Frontiers Center for Brain Science of Ministry of Education, Fudan University.

This work was made possible by the generous sharing of GWAS summary statistics. We thank PGC for providing summary results data for these analyses. The investigators within PGC contributed to the design and implementation of PGC and/or provided data but did not participate in the analysis or writing of this report. PGC was made possible by the generous participation of the control subjects, the patients, and their families. We also thank the participants, researchers, and staff associated with the many other studies from which we used data for this report. We thank the International Genomics of Alzheimer’s Project (IGAP) for providing summary results data for these analyses. The investigators within IGAP contributed to the design and implementation of IGAP and/or provided data but did not participate in analysis or writing of this report. IGAP was made possible by the generous participation of the control subjects, the patients, and their families. The i– Select chips were funded by the French National Foundation on Alzheimer’s disease and related disorders. EADI was supported by the LABEX (laboratory of excellence program investment for the future) DISTALZ grant, Inserm, Institut Pasteur de Lille, Université de Lille 2, and the Lille University Hospital. GERAD was supported by the Medical Research Council (Grant no 503480), Alzheimer’s Research UK (Grant no 503176), the Wellcome Trust (Grant no 082604/2/07/Z) and German Federal Ministry of Education and Research (BMBF): Competence Network Dementia (CND) grant no 01GI0102, 01GI0711, 01GI0420. CHARGE was partly supported by the NIH/NIA grant R01 AG033193 and the NIA AG081220 and AGES contract N01– AG– 12100, the NHLBI grant R01 HL105756, the Icelandic Heart Association, and the Erasmus Medical Center and Erasmus University. ADGC was supported by the NIH/NIA grants: U01 AG032984, U24 AG021886, U01 AG016976, and the Alzheimer’s Association grant ADGC– 10– 196728.

The supplementary material is available in the electronic version of this article:

References

Rauch

, Huen

, Miller

, Chaudry

, Lau

, Sanes

, Johanson

, Stopa

, Burgess

(2011) Changes in brain beta-amyloid deposition and aquaporin 4 levels in response to altered agrin expression in mice. J Neuropathol Exp Neurol 70, 1124–1137.

Ang

, Tai

, Lo

, Seet

, Soong

(2010) Neurodegenerative diseases: Exercising toward neurogenesis and neuroregeneration. Front Aging Neurosci 2, 25.

Martinez

, Martin

, Alcelay

, Flores

, Valiente

, Juanbeltz

, Beldarrain

, Lopez

, Gonzalez-Fernandez

, Salazar

, Gandarias

, Borda

, Marques

, Amillano

, Zabaleta

, de Pancorbo

(2009) The COMT Val158 Met polymorphism as an associated risk factor for Alzheimer disease and mild cognitive impairment in APOE 4 carriers. BMC Neurosci 10, 125.

Lemche

(2018) Early life stress and epigenetics in late-onset Alzheimer’s dementia: A systematic review. Curr Genomics 19, 522–602.

Kunkle

, Grenier-Boley

, Sims

, Bis

, Damotte

, Naj

, Boland

, Vronskaya

, van der Lee

, Amlie-Wolf

, et al. (2019) Genetic meta-analysis of diagnosed Alzheimer’s disease identifies new risk loci and implicates Abeta, tau, immunity and lipid processing. Nat Genet 51, 414–430.

Pope

, Shue

, Beck

(2003) Will a healthy lifestyle help prevent Alzheimer’s disease. Annu Rev Public Health 24, 111–132.

Ott

, Slooter

, Hofman

, van Harskamp

, Witteman

, Van Broeckhoven

, van Duijn

, Breteler

(1998) Smoking and risk of dementia and Alzheimer’s disease in a population-based cohort study: The Rotterdam Study. Lancet 351, 1840–1843.

Graves

, van Duijn

, Chandra

, Fratiglioni

, Heyman

, Jorm

, Kokmen

, Kondo

, Mortimer

, Rocca

, Shalat

, Soininen

, Hofman

, the Eurodem Risk Factors Research Group (1991) Alcohol and tobacco consumption as risk factors for Alzheimer’s disease: A collaborative re-analysis of case-control studies. EURODEM Risk Factors Research Group. Int J Epidemiol 20 Suppl 2, S48–57.

Tyas

(2001) Alcohol use and the risk of developing Alzheimer’s disease. Alcohol Res Health 25, 299–306.

10.

Richmond

, Anderson

, Dashti

, Jones

, Lane

, Strand

, Brumpton

, Rutter

, Wood

, Straif

, Relton

, Munafo

, Frayling

, Martin

, Saxena

, Weedon

, Lawlor

, Smith

(2019) Investigating causal relations between sleep traits and risk of breast cancer in women: Mendelian randomisation study. BMJ 365, 12327.

11.

Larsson

, Traylor

, Malik

, Dichgans

, Burgess

, Markus

, CoSTREAM Consortium, on behalf of the International Genomics of Alzheimer’s Project (2017) Modifiable pathways in Alzheimer’s disease: Mendelian randomisation analysis. BMJ 359, j5375.

12.

Østergaard

, Mukherjee

, Sharp

, Proitsi

, Lotta

, Day

, Perry

, Boehme

, Walter

, Kauwe

, Gibbons

; Alzheimer’s Disease Genetics Consortium; GERAD1 Consortium; EPIC-InterAct Consortium, Larson

, Powell

, Langenberg

, Crane

, Wareham

, Scott

(2015) Associations between potentially modifiable risk factors and Alzheimer disease: A mendelian randomization study. PLoS Med 12, e1001841; discussion e1001841.

13.

Mukherjee

, Walter

, Kauwe

JSK

, Saykin

, Bennett

, Larson

, Crane

, Glymour

, Adult Changes in Thought Study Investigators; Religious Orders Study/Memory and Aging Project Investigators; Alzheimer’s Disease Genetics Consortium (2015) Genetically predicted body mass index and Alzheimer’s disease-related phenotypes in three large samples: Mendelian randomization analyses. Alzheimers Dement 11, 1439–1451.

14.

Karlsson Linner

, Biroli

, Kong

, Meddens

SFW

, Wedow

, Fontana

, Lebreton

, Tino

, Abdellaoui

, Hammerschlag

, et al. (2019) Genome-wide association analyses of risk tolerance and risky behaviors in over 1 million individuals identify hundreds of loci and shared genetic influences. Nat Genet 51, 245–257.

15.

Staley

, Blackshaw

, Kamat

, Ellis

, Surendran

, Sun

, Paul

, Freitag

, Burgess

, Danesh

, Young

, Butterworth

(2016) PhenoScanner: A database of human genotype-phenotype associations. Bioinformatics 32, 3207–3209.

16.

Kamat

, Blackshaw

, Young

, Surendran

, Burgess

, Danesh

, Butterworth

, Staley

(2019) PhenoScanner V2: An expanded tool for searching human genotype-phenotype associations. Bioinformatics 35, 4851–4853.

17.

Myers

, Chanock

, Machiela

(2020) LDlinkR: An R Package for rapidly calculating linkage disequilibrium statistics in diverse populations. Front Genet 11, 157.

18.

Machiela

, Chanock

(2015) LDlink: A web-based application for exploring population-specific haplotype structure and linking correlated alleles of possible functional variants. Bioinformatics 31, 3555–3557.

19.

Pichler

, Del Greco

, Gogele

, Lill

, Bertram

, Do

, Eriksson

, Foroud

, Myers

, PD Consortium

GWAS

, Nalls

, Keller

, International Parkinson’s Disease Genomics Consortium; Wellcome Trust Case Control Consortium, Benyamin

, Whitfield

, Genetics of Iron Status Consortium, Pramstaller

, Hicks

, Thompson

, Minelli

(2013) Serum iron levels and the risk of Parkinson disease: A Mendelian randomization study. PLoS Med 10, e1001462.

20.

Brion

, Shakhbazov

, Visscher

(2013) Calculating statistical power in Mendelian randomization studies. Int J Epidemiol 42, 1497–1501.

21.

Jansen

, Savage

, Watanabe

, Bryois

, Williams

, Steinberg

, Sealock

, Karlsson

, Hagg

, Athanasiu

, Voyle

, Proitsi

, Witoelar

, Stringer

, Aarsland

, Almdahl

, Andersen

, Bergh

, Bettella

, Bjornsson

, Braekhus

, Brathen

, de Leeuw

, Desikan

, Djurovic

, Dumitrescu

, Fladby

, Hohman

, Jonsson

, Kiddle

, Rongve

, Saltvedt

, Sando

, Selbaek

, Shoai

, Skene

, Snaedal

, Stordal

, Ulstein

, Wang

, White

, Hardy

, Hjerling-Leffler

, Sullivan

, van der Flier

, Dobson

, Davis

, Stefansson

, Pedersen

, Ripke

, Andreassen

, Posthuma

(2019) Genome-wide meta-analysis identifies new loci and functional pathways influencing Alzheimer’s disease risk. Nat Genet 51, 404–413.

22.

Liu

, Erlich

, Pickrell

(2017) Case-control association mapping by proxy using family history of disease. Nat Genet 49, 325–331.

23.

Marioni

, Harris

, Zhang

, McRae

, Hagenaars

, Hill

, Davies

, Ritchie

, Gale

, Starr

, Goate

, Porteous

, Yang

, Evans

, Deary

, Wray

, Visscher

(2018) GWAS on family history of Alzheimer’s disease. Transl Psychiatry 8, 99.

24.

Lambert

, Ibrahim-Verbaas

, Harold

, Naj

, Sims

, Bellenguez

, DeStafano

, Bis

, Beecham

, Grenier-Boley

, et al. (2013) Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for Alzheimer’s disease. Nat Genet 45, 1452–1458.

25.

Davies

, Holmes

, Davey Smith

(2018) Reading Mendelian randomisation studies: A guide, glossary, and checklist for clinicians, BMJ 362, k601.

26.

Bowden

, Del Greco

, Minelli

, Davey Smith

, Sheehan

, Thompson

(2017) A framework for the investigation of pleiotropy in two-sample summary data Mendelian randomization. Stat Med 36, 1783–1802.

27.

Bowden

, Davey Smith

, Burgess

(2015) Mendelian randomization with invalid instruments: Effect estimation and bias detection through Egger regression. Int J Epidemiol 44, 512–525.

28.

Bowden

, Davey Smith

, Haycock

, Burgess

(2016) Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol 40, 304–314.

29.

Burgess

, Thompson

(2017) Interpreting findings from Mendelian randomization using the MR-Egger method. Eur J Epidemiol 32, 377–389.

30.

Barnard

, Bush

, Ceccarelli

, Cooper

, de Jager

, Erickson

, Fraser

, Kesler

, Levin

, Lucey

, Morris

, Squitti

(2014) Dietary and lifestyle guidelines for the prevention of Alzheimer’s disease. Neurobiol Aging 35 Suppl 2, S74–78.

31.

Tyas

, Manfreda

, Strain

, Montgomery

(2001) Risk factors for Alzheimer’s disease: A population-based, longitudinal study in Manitoba, Canada. Int J Epidemiol 30, 590–597.

32.

Seo

, Kim

, Lee

, Choo

(2016) Altered executive function in pre-mild cognitive impairment. J Alzheimers Dis 54, 933–940.

33.

Sinz

, Zamarian

, Benke

, Wenning

, Delazer

(2008) Impact of ambiguity and risk on decision making in mild Alzheimer’s disease. Neuropsychologia 46, 2043–2055.

34.

Hare

, Camerer

, Rangel

(2009) Self-control in decision-making involves modulation of the vmPFC valuation system. Science 324, 646–648.

35.

Raefsky

, Mattson

(2017) Adaptive responses of neuronal mitochondria to bioenergetic challenges: Roles in neuroplasticity and disease resistance. Free Radic Biol Med 102, 203–216.

36.

Ahmed

, Ke

, Vucic

, Ittner

, Seeley

, Hodges

, Piguet

, Halliday

, Kiernan

(2018) Physiological changes in neurodegeneration - mechanistic insights and clinical utility. Nat Rev Neurol 14, 259–271.

37.

Hessler

, Katz

(2010) Brief report: Associations between emotional competence and adolescent risky behavior. J Adolesc 33, 241–246.

38.

Morales

, Welter

, Bowler

, Cerezo

, Harris

, McMahon

, Hall

, Junkins

, Milano

, Hastings

, Malangone

, Buniello

, Burdett

, Flicek

, Parkinson

, Cunningham

, Hindorff

, MacArthur

JAL

(2018) A standardized framework for representation of ancestry data in genomics studies, with application to the NHGRI-EBI GWAS Catalog. Genome Biol 19, 21.