Causal Relationships Between Relative Intake from the Macronutrients and Alzheimer’s Disease: A Two-Sample Mendelian Randomization Study

Abstract

Background:

Some observational studies indicated the associations of relative carbohydrate, sugar, fat, and protein intake and Alzheimer’s disease (AD). But it remains unclear whether the associations are causal.

Objective:

This study aimed to identify the effects of relative carbohydrate, sugar, fat, and protein intake in the diet on AD.

Methods:

A two-sample Mendelian randomization was employed. Finally, 14 independent lead SNPs remained in the Social Science Genetic Association Consortium. These SNPs of relative carbohydrate, sugar, fat, and protein intake at the level of genome-wide significance (p < 5×10^–8) were used as instrumental variables. The summary data for AD were acquired from the International Genomics of Alzheimer’s Project with a total of 54,162 individuals (17,008 AD patients and 37,154 control participants).

Results:

This two-sample Mendelian randomization indicated that increased relative protein intake (per 1 standard deviation) causally decreased the AD risk (OR = 0.48, 95% CI: 0.24–0.95, p = 0.036), and increased relative fat intake may decrease the risk of AD (OR = 0.22, 95% CI: 0.06–0.86, p = 0.029). No statistical significance with AD risk was seen for relative carbohydrate or relative sugar intake.

Conclusion:

A higher relative intake of protein can causally reduce the risk of AD in the elderly. Additionally, a higher relative intake of fat may be protective against AD. No evidence showed that AD was associated with relative carbohydrate and sugar intake.

Keywords

Alzheimer’s disease dietary carbohydrates dietary fats dietary proteins Mendelian randomization analysis

INTRODUCTION

According to the Global Burden of Disease 2017 study, Alzheimer’s disease (AD) has become the tenth leading cause of death globally, affecting people’s ability to function [1]. However, the pathogenesis of AD is still unclear. The existing clinical therapies of AD are symptomatic treatment [2]. Therefore, it is of great significance to prevent the development of AD in the early stage or even before the onset of AD.

Dietary factors may perform an important role in age-related cognitive decline and AD [3]. Most studies had focused the main point on the effects of dietary patterns like the Mediterranean diet [4] and dietary components such as fruits and vegetables on AD [5]. The increased total energy intake has an impact on the risk of AD [6]. Carbohydrate, fat, and protein are the primary energy sources. However, it was still inconsistent with the effects of energy source nutrients on AD. As for fat, it has been found that total dietary fat had an aggravating effect on AD [7]. But a study showed that a high total fat intake provided a protective effect for AD [8]. Another study proved that medium-fat intake was beneficial to AD [6]. At the same time, there was an article showing that there was no correlation between the intake level of total fat and AD [9]. As for protein, it has been described that a high protein diet may confer some protection against AD, reducing the risk of developing AD [8, 10]. But another research found no statistically significant difference in protein intake between patients with AD and mild cognitive impairment (MCI) and the control group [11].

Regarding carbohydrate, the risk of MCI or dementia increased nearly two times among the highest carbohydrate intake quartile versus the lowest in a prospective cohort [8]. However, in a cross-sectional study, the investigator discovered no difference in carbohydrate intake between AD patients and non-AD patients [11]. And as for sugar, people who consume sugar or sugar-sweetened beverages every day have a higher risk of AD than those who never or rarely eat sugar [12, 13]. Existing conflicting research is challenging to make clear causal inferences on the effects of macronutrients on AD because of confounding factors and the lack of use of relative intakes from macronutrients between studies. Thus, the causal effects of relative intakes from macronutrients on AD are warranted.

Mendelian randomization (MR) provides a new method to ensure the rationality of the time sequence of causal inference by using genetic variations as instrumental variables [14]. The existing research focused on assessing the effects of blood vitamins and minerals on AD and the role of food on AD [15, 16]. In this study, a two-sample MR analysis was performed to provide evidence for causal effects between the relative intake of fat, carbohydrate, and proteins and AD.

METHODS

The method of two-sample MR analysis has been widely used before [17]. This study used a two-sample MR approach to explore the potential causal relationship between dietary macronutrients and AD and to estimate the size of its effect. In MR analysis, instrumental variables used as genetic variants need to satisfy three assumptions: 1) genetic variants must be strongly associated with the exposure of interest; 2) genetic variants must be not associated with potential confounders of the association between the exposure and the outcome; and 3) there are no effects of the genetic variants on the outcome, independent of the exposure [18] (Fig. 1).

Fig. 1

Schematic diagram of two-sample Mendelian randomization.

Exposure data on diet composition

The genome-wide association studies (GWAS) of relative intake from the macronutrients, fat, protein, carbohydrate, and sugar had over 235,000 populations with European ancestry [19]. All cohorts of this GWAS had an average total energy expenditure of 2,064 kcal (SD: 609.9 kcal), of which 49% were carbohydrates, 35% were fats, and 16% proteins. This summary data of dietary macronutrients is based on a correction (percentage of total energy intake) between macronutrients and total energy intake [20].

To the best of our knowledge, this is the largest GWAS of the discovery sample that encompasses people of European countries (white British, Scottish, Irish, and others). Our discovery cohort is the UK Biobank (UKB). The UKB was established in 2006 and represented a large group of the entire UK population aged 45–69 when recruiting [21]. Anyone registered with the Health Commission and living near one of the 22 assessment centers was qualified to participate in this cohort. Around 40% of UKB samples completed at least one 24-h dietary recall questionnaire. Finally, 211,063 individuals passed internal quality control, and 175,253 questionnaires were kept. And the replication samples consist of 14 cohorts from western and northern European countries like Italy, Spain, France, Sweden, and Germany. Food-frequency questionnaires were used by all the replication cohorts.

Outcome data on AD

Genetic associations with AD were obtained from the meta-analysis of GWAS on individuals of European ancestry (N_case = 17,008 and N_control =37,154) contributed by the International Genomics of Alzheimer’s Project (IGAP) [22]. In short, the meta-analysis consists of 4 GWAS: the Genetic and Environmental Risk in Alzheimer’s Disease Consortium, the European Alzheimer’s Disease Initiative consortium, the Cohort for Heart and Ageing Research in Genomic Epidemiology consortium, and the Alzheimer’s Disease Genetics Consortium. In all cohorts included in this meta-analysis, the average age of onset of AD cases was between 68.5 to 82.3 years, and the percentage of women in AD controls and patients are 57.3% and 61.3%. Details of the IGAP were published before [22].

Extracting instrumental variables

Single nucleotide polymorphisms (SNPs) extrac-ted from summary-level data were robustly associated with relative intake from the macronutrients with genome-wide significance (p < 5×10^–8). All SNPs were not in a state of linkage disequilibrium to ensure that the SNPs were independent of each other. We computed F statistics for the individual SNPs to assess the strength of our instruments. All the above F statistics greater than 10 were widely used to indicate a strong instrument. We used two comprehensive genetic cross-reference websites to check pleiotropy to ensure that significant SNPs were singly associated with AD via relative intake from fat, protein, carbohydrate, and sugar. And they are the GWAS catalog (https://www.ebi.ac.uk/gwas/) [23], and PhenoScanner (https://www.phenoscanner.medschl.cam.ac.uk) [24], which provide well-acknowledged associations of known SNPs with their phenotypes.

We use non-UKB samples data to avoid overleap and overestimation for the outcome. Analyses were conducted without the apolipoprotein E variant (rs429358) included, as apolipoprotein E was previously shown to be pleiotropy [25]. Only biallelic SNPs were used as instrumental variables [26]. SNPs for exposure were checked in the outcome GWAS dataset. The minor allele frequency threshold of palindromic SNPs is 0.3. Proxies were identified for any SNPs which were not found in the outcome database (R² > 0.8 based on LD information in 1000 Genomes Project). Next, we harmonized the SNPs on exposure and outcome to the same allele. The 14 SNPs remained and F > 10 (Supplementary Table 1), after excluding 20 SNPs with potential pleiotropy (Supplementary Table 2) and two palindromic SNPs (rs9927317 and rs1104608).

Statistical analysis

First of all, inverse variance weighting (IVW) was used as the main analysis method for causal inference. The premise of adopting the IVW method is that all SNPs are valid instrumental variables, so this method can get accurate estimation results [27].

Sensitivity analysis

As a sensitivity analysis, MR-Egger regression with bootstrapped SE was used to evaluate the bias caused by invalid instrumental variables and pleiotropic instrumental variables. Its intercept value can evaluate the size of gene pleiotropy. When the regression intercept item is 0 or the intercepted item is not statistically significant, there is no gene pleiotropy, and its impact can be ignored [28]. We also took a weighted median (WM) method [29]. When at least half of the SNPs are valid instrumental variables, WM can be used to obtain an estimate consistent with the final effect [29]. The penalized weighted median (PWM) method uses the penalized weights to construct a distribution function and estimate the causal effect, weakening the influence of the estimated causal effect due to the pleiotropy of genetic variation on the estimate of the overall causal effect [29]. Finally, when significant heterogeneity was estimated in Cochran’s Q statistics [30]. Additional analysis was carried out with Mendelian randomization pleiotropy residual sum and outlier (MR-PRESSO), which can identify and rectify the effects of possible confounders [31]. Here is the flowchart of this article depicted (Fig. 2).

Fig. 2

Flowchart of two-sample Mendelian randomization. SSGAC consortium, the Social Science Genetic Association Consortium; IGAP, the International Genomics of Alzheimer’s Project; GWAS, genome-wide association study; AD, Alzheimer’s disease; IVW, inverse-variance weighted; MR, Mendelian randomization; MR PRESSO, MR Pleiotropy RESidual Sum and Outlier; SNP, single-nucleotide polymorphism.

In order to investigate the causal relationship between relative intake of the macronutrients and AD, we conducted a bidirectional analysis to explore whether relative intake of the macronutrients can affect the risk of AD.

All statistical analysis was based on R software (version: 4.0.3). We used the R package “MendelianRandomization” (version 0.4.3) [32], “TwoSampleMR” (version 0.5.6) [26], and “MRPRESSO” (version 1.0) [31] for MR analysis. A two-sided p value < 0.05 was considered statistically significant.

RESULTS

Four genome-wide significant SNPs for relative carbohydrate intake, four genome-wide significant SNPs for sugar, two uncorrelated significant SNPs for fat, and four uncorrelated significant SNPs for protein remained after excluding palindromic SNPs and confounders (Table 1). None of these SNPs were given in the GWAS catalog or PhenoScanner as strongly associated with AD confounders.

Table 1

Main characteristics of genetic variants in the macronutrients and AD GWAS datasets

	SNP	Chr	Gene	EAF	EA	NEA	β1	SE	β2	se
Carbohydrate	rs10206338	2	AC007100.1	0.578	A	G	–0.016	0.003	0.0034	0.0161
	rs9987289	8	AC022784.6	0.090	A	G	–0.026	0.0048	0.014	0.0279
	rs2472297	15	CYP1A1	0.260	T	C	–0.018	0.003	0.0370	0.0196
	rs838144	19	IZUMO1	0.509	T	C	–0.023	0.003	0.0231	0.0161
Sugar	rs12713415	2	AC007100.1	0.7219	C	G	–0.01911	0.00327	–0.0052	0.0170
	rs13202107	6	SNORD66	0.1948	A	G	–0.02016	0.00358	0.0297	0.0196
	rs341228	18	L3MBTL4	0.297	T	C	0.019	0.00313	–0.0177	0.0172
	rs62132802	19	RARB	0.361	T	C	–0.02017	0.00353	–0.0062	0.0163
Protein	rs445551	2	CTNNA2	0.317	A	G	0.01919	0.00339	–0.0128	0.0172
	rs13146907	4	KLB	0.625	A	G	–0.0219	0.00284	0.0299	0.0165
	rs1461729	8	AC022784.6	0.098	A	G	0.03183	0.00459	0.0071	0.0274
	rs838133	19	FGF21	0.425	A	G	–0.03182	0.00301	0.0261	0.1188
Fat	rs57193069	7	MAD1L1	0.553	A	G	–0.016	0.0028	0.0445	0.0162
	rs33988101	19	MAMSTR	0.499	T	G	–0.029	0.00276	0.0329	0.0162

SNP, single nucleotide polymorphism identifier; Chr, chromosome; Gene, nearest gene; EAF, effect allele frequency; EA, effect allele; NEA, non-effect allele; β1, regression coefficient of exposure; SE, standard error of regression coefficient of expose; β2, regression coefficient of outcome; se, standard error of regression coefficient of outcome.

Relationship between relative carbohydrate intake and AD

Genetically instrumented relative carbohydrate intake was not associated with AD based on these five genome-wide significant SNPs using IVW (with fixed effects), WM, PWM, MR-Egger, or MR-PRESSO (Table 2).

Table 2

MR estimates of the associations of log-transformed standardized residuals of relative intake from carbohydrate, sugar, fat, and protein with AD

Expose	No. of SNPs	MR method	OR	95% CI	p	Q statistic (p)	Intercept p
Carbohydrate	4	Inverse variance weighted¹	0.93	(0.26,3.27)	0.906	5.89 (0.117)
		Weighted median	0.90	(0.27,2.96)	0.859
		Penalized Weighted median	0.90	(0.26,3.11)	0.864
		MR-Egger (bootstrap)	9.27	(0.48,178.99)	0.052		0.559
		MR-PRESSO	0.93	(0.26,3.27)	0.913
Sugar	4	Inverse variance weighted¹	0.71	(0.17,2.89)	0.633	5.98 (0.113)
		Weighted median	0.52	(0.15,1.77)	0.293
		Penalized Weighted median	0.48	(0.13,1.73)	0.262
		MR-Egger (bootstrap)	0.24	(0.02,2.36)	0.102		0.986
		MR-PRESSO	0.71	(0.17,2.89)	0.666
Fat	2	Inverse variance weighted²	0.22	(0.06,0.86)	0.029
Protein	4	Inverse variance weighted¹	0.48	(0.24,0.95)	0.036	1.63 (0.44)
		Weighted median	0.45	(0.20,1.03)	0.058
		Penalized Weighted median	0.45	(0.20,1.02)	0.057
		MR-Egger (bootstrap)	0.25	(0.03,1.92)	0.088		0.562
		MR-PRESSO	0.48	(0.27,0.87)	0.015

Carbohydrate: 4 independent SNPs, p value < 5×10^–8; Sugar: 4 independent SNPs, p value < 5×10^–8; Fat: 2 independent SNPs, p value < 5×10^–8; Protein: 4 independent SNPs, p value < 5×10^–8; AD, Alzheimer’s disease; MR, Mendelian randomization; SNPs: single nucleotide polymorphisms; OR: odds ratio; 95% CI: 95% confidence interval; MR-PRESSO: Mendelian randomization pleiotropy residual sum and outlier (MR-PRESSO). ¹fixed effects model; ²multiplicative random effects model.

Relationship between relative sugar intake and AD

Genetically instrumented relative sugar intake was not associated with AD based on these five genome-wide significant SNPs using the IVW method (with fixed effects) (OR = 0.71, 95% CI: 0.17–2.89, p = 0.633). The results of WM, PWM, and MR-Egger were not significant (Table 2).

Relationship between relative fat intake and AD

Genetically instrumented relative fat intake was associated with a lower risk of AD. These inverse relationships were generally obvious to different MR analysis methods. IVW method (with multiplicative random effects) was OR = 0.22 (95% CI: 0.06–0.86, p = 0.029) (Table 2).

Relationship between relative protein intake and AD

Genetically instrumented relative protein intake was associated with AD based on four single genome-wide significant SNP using IVW. IVW method (OR = 0.48, 95% CI: 0.24–0.95, p = 0.036) showed the reverse relationship between dietary protein intake and AD. MR-PRESSO method (OR = 0.48, 95% CI: 0.27–0.87, p = 0.015) was in agreement with IVW method. MR Egger intercept for protein gave no statistical pleiotropy for AD. The heterogeneity test showed no significant heterogeneity with Cochran’s Q = 1.97, p = 0.579 (Table 2).

Genetically predicted AD on relative intake of the macronutrients

For the bidirectional MR analysis, we selected 19 significant, independent SNPs as IVs from IGAP and extracted these SNPs from the relative intake of the macronutrients individually.

The results suggested no association between AD and relative carbohydrate, sugar, fat, and protein intake (Supplementary Table 3). And MR Egger intercept indicated there was no pleiotropy among those 19 IVs. The heterogeneity test showed no significant heterogeneity.

DISCUSSION

Dietary fat intake and AD

This study indicated the potential protective effect of dietary fat on AD. A recent study showed that the overall cognitive scores of people on a high-fat diet are 9.0% higher than those on a low-fat diet [33]. In a Chinese cohort, the cooking oil intake of MCI patients was lower than the normal people intake (29.76 mL VS 35.20 mL cooking oil per day) [34]. The ketogenic diet (usually a very high-fat, low-carbohydrate diet) may also alter the process of AD. A modified Mediterranean ketogenic diet that was < 10% carbohydrate, 60% ∼65% fat, and 30% –35% protein in macronutrient composition (% of total calories) can improve cerebral perfusion, peripheral metabolic profiles, and cerebral ketone body uptake [35]. A ketogenic diet can significantly increase ketone body levels, improve ketone utilization, effectively reduce brain glucose utilization, lower insulin levels, increase long-chain and medium-chain fatty acids, affect lipid processing, and reduce inflammation [36]. So these changes may make the ketogenic diet improve cognition in AD.

Furthermore, some studies have confirmed that omega-3 fatty acids had a protective effect on AD [9, 37]. Docosahexaenoic acid (DHA) and eicosapentaenoic acid can play a key role on brain cells. DHA, an important structural component of membrane phospholipids in the brain, maintains the integrity, fluidity, and function of neuronal membranes. Eicosapentaenoic acid is an inhibitor of prostaglandins, thromboxanes, and leukotrienes and has been proven to regulate metabolism and immune processes. DHA can reduce the secretion of cell death pro-apoptotic protein Bcl-2 antagonist and amyloid-β (Aβ), increase PI3-kinase neuroprotective pathway and improve cognitive function [38].

However, some observational studies reported that total dietary fat had no effect on AD [9]. A disadvantage of this epidemiological study was that the investigator did not evaluate the subjects’ diet at baseline. If people with cognitive impairment change their diets, this may lead to incorrect inferences of reverse causality. Another possible explanation for this was that this observational study used the specific daily fat intake, not the macronutrient ratio of fat, to verify the association between fat intake and AD, which is not adjusted by total energy intake.

Another study proved that moderate total fat intake has a protective effect on AD. Moderate intake of unsaturated fats (second quartile) in middle age has a protective effect, and moderate intake of saturated fats (second quartile) may increase the risk of AD [6]. The limitation of this study is the reliability of dietary data. Researchers only evaluated the fat intake of spreads, milk, and yogurt, which represented only half of the fat intake in the diet. In addition, a dietary assessment during middle age did not represent the long-term diet after the baseline survey.

Dietary protein intake and AD

Consistent with our study, a cohort study reported that a higher dietary protein intake was associated with a 21% reduced risk of MCI or dementia, independent of potential confounders [8]. Greater protein consumption was associated with a lower likelihood of high brain Aβ burden, which may be related to reducing blood pressure with a high dietary protein intake [10]. Aβ was considered as the leading cause of AD in the amyloid hypothesis. When Aβ clumped together to form deposits in the brain, it triggered neurodegenerative processes that led to the loss of memory and cognitive ability that is observed in AD [39]. If Aβ accumulation is the cause of neurodegeneration, then changing this accumulation will prevent AD.

Maintaining sufficient protein intake can help the elderly maintain muscle mass and strength [40]. Decreased muscle strength and low lean body mass have been shown to be important risk factors for cognitive decline. A recent study showed that the decrease in total brain volume and cognitive ability in the brain is significantly associated with the decrease in lean body mass [41]. Another study showed that for every unit increase in muscle strength, the risk of AD was reduced by 43%. This association still existed after adjusting for covariates such as body mass index, physical activity, lung function, vascular disease, and apolipoprotein E4 status. In addition, the growth of muscle strength was associated with a slower decline in overall cognitive function and a lower risk of MCI [42]. So dietary protein may protect the brain’s cognitive function by preserving muscle mass and muscular strength.

There was some reasonable evidence that protein and amino acid intake can improve sleep quality and sleep duration. Sleep is closely related to short-term cognitive problems and has an important impact on the long-term development of AD [43]. Recent evidence showed that amino acids in the human body also played an irreplaceable role in the regulation of sleep. L-tryptophan is the upstream precursor of melatonin, which is a hormone that regulates human sleep by affecting the sleep-wake cycle [43]. Thereby increasing melatonin by regulating L-tryptophan may benefit cognitive health.

Dietary carbohydrate, sugar intake, and AD

The previous study showed the relatively high intake of carbohydrate may increase the risk of MCI or dementia in older people [8]. Another research indicated that daily carbohydrate intake of AD and MCI patients did not differ from that of the control group (42.0±1.4 and 42.6±1.5 versus 40.5±1.2 g/day, p = 0.540) [11]. As for sugar, research reported higher intake of sugary beverages was associated with markers of pre-clinical AD [12]. Our bidirectional MR study provided no evidence for a causal relationship between relative intake of the macronutrients and AD.

However, the MR method gave a different conclusion: no statistical significance with AD risk was seen for relative carbohydrate intake or relative sugar intake. A possible explanation for this result was that the exposure data used in this study was the macronutrient distribution of carbohydrate and sugar, not the specific daily intake, whereas the daily intake data was used in most observational studies. Another possible explanation for this result was that only a very few SNPs were retained to make an MR analysis, which led to an unstable result. So this result should be treated with caution. The larger amount of dietary data from genome-wide association studies in the future would solve this issue promisingly.

Overall, much of the current evidence is contradictory. The MR method provides a new approach to exploring the association of dietary macronutrients with AD and can largely avoid confounding factors and reverse causality [14].

There were several limitations to this study. First, two-sample MR assumes that there is a linear relationship between exposure and outcome. If there is a non-linear relationship between the two, this method is not applicable. Second, more detailed dietary data such as refined carbohydrates, omega-3 fatty acids, and vegetable protein or animal protein are not available. They are more representative of the quality of the diet. Third, our analysis was limited to the existing summary data of macronutrient intake that explained only a small part of the phenotypic variation. And we had to exclude the SNPs that were closely related to confounding factors (Supplementary Table 2). The following were the changes that can be explained after eliminating confounding factors: protein changed from 0.153% to 0.107%; fat changed from 0.170% to 0.055%; carbohydrate changed from 0.199% to 0.062%; sugar changed from 0.207% to 0.061%. Moreover, the average age of populations in exposure samples was far less than that of populations in outcome samples, causing partly wrong estimation of causal relationships.

In conclusion, there was evidence of a causal relation between relative fat intake and protein intake on AD. A Diet with a higher proportion of fat intake and protein intake might benefit AD prevention. Clarifying the underlying pathways would be meaningful to dietary recommendations, AD prevention, and public health.

Footnotes

ACKNOWLEDGMENTS

This work was supported by the National Natural Science Foundation of China (81872616). We also want to acknowledge the participants and investigators of the studies. Data on the relative intake of the macronutrients have been contributed by the SSGAC consortium and available at https://www.thessgac.org/ for the Social Science Genetic Association Consortium. Data on Alzheimer’s disease have been contributed by the IGAP consortium and available at .

Authors’ disclosures available online ().

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

References

(2018) Global, regional, and national age-sex-specific mortality for 282 causes of death in 195 countries and territories, 1980-2017: A systematic analysis for the Global Burden of Disease Study 2017. Lancet 392, 1736–1788.

Panza

, Seripa

, Solfrizzi

, Imbimbo

, Lozupone

, Leo

, Sardone

, Gagliardi

, Lofano

, Creanza

, Bisceglia

, Daniele

, Bellomo

, Greco

, Logroscino

(2016) Emerging drugs to reduce abnormal β-amyloid protein in Alzheimer’s disease patients. Expert Opin Emerg Drugs 21, 377–391.

Angeloni

, Businaro

, Vauzour

(2020) The role of diet in preventing and reducing cognitive decline. Curr Opin Psychiatry 33, 432–438.

Sofi

, Macchi

, Abbate

, Gensini

, Casini

(2010) Effectiveness of the Mediterranean diet: Can it help delay or prevent Alzheimer’s disease? J Alzheimers Dis 20, 795–801.

Loef

, Walach

(2012) Fruit, vegetables and prevention of cognitive decline or dementia: A systematic review of cohort studies. J Nutr Health Aging 16, 626–630.

Laitinen

, Ngandu

, Rovio

, Helkala

, Uusitalo

, Viitanen

, Nissinen

, Tuomilehto

, Soininen

, Kivipelto

(2006) Fat intake at midlife and risk of dementia and Alzheimer’s disease: A population-based study. Dement Geriatr Cogn Disord 22, 99–107.

Kalmijn

, Launer

, Ott

, Witteman

, Hofman

, Breteler

(1997) Dietary fat intake and the risk of incident dementia in the Rotterdam Study. Ann Neurol 42, 776–782.

Roberts

, Roberts

, Geda

, Cha

, Pankratz

, O’Connor

, Knopman

, Petersen

(2012) Relative intake of macronutrients impacts risk of mild cognitive impairment or dementia. J Alzheimers Dis 32, 329–339.

Morris

, Evans

, Bienias

, Tangney

, Bennett

, Aggarwal

, Schneider

, Wilson

(2003) Dietary fats and the risk of incident Alzheimer disease. Arch Neurol 60, 194–200.

10.

Fernando

WMADB

, Rainey-Smith

, Gardener

, Villemagne

, Burnham

, Macaulay

, Brown

, Gupta

, Sohrabi

, Weinborn

, Taddei

, Laws

, Goozee

, Ames

, Fowler

, Maruff

, Masters

, Salvado

, Rowe

, Martins

(2018) Associations of dietary protein and fiber intake with brain and blood amyloid-β. J Alzheimers Dis 61, 1589–1598.

11.

Doorduijn

, de van der Schueren

MAE

, van de Rest

, de Leeuw

, Hendriksen

HMA

, Teunissen

, Scheltens

, van der Flier

, Visser

(2020) Energy intake and expenditure in patients with Alzheimer’s disease and mild cognitive impairment: The NUDAD project. Alzheimers Res Ther 12, 116.

12.

Pase

, Himali

, Jacques

, DeCarli

, Satizabal

, Aparicio

, Vasan

, Beiser

, Seshadri

(2017) Sugary beverage intake and preclinical Alzheimer’s disease in the community. Alzheimers Dement 13, 955–964.

13.

, Liu

, Khan

, Yan

, Wang

(2019) Dietary habits and cognitive impairment risk among oldest-old Chinese. J Gerontol B Psychol Sci Soc Sci 74, 474–483.

14.

Burgess

, Foley

, Zuber

(2018) Inferring causal relationships between risk factors and outcomes from genome-wide association study data. Ann Rev Genomics Hum Genet 19, 303–327.

15.

Carnegie

, Zheng

, Sallis

, Jones

, Wade

, Evans

, Zammit

, Munafò

, Martin

(2020) Mendelian randomisation for nutritional psychiatry. Lancet Psychiatry 7, 208–216.

16.

Zhou

, Taylor

, Karhunen

, Zhan

, Rovio

, Lahti

, Sjögren

, Byberg

, Lyall

, Auvinen

, Lehtimäki

, Kähönen

, Hutri-Kähönen

, Perälä

, Michaëlsson

, Mahajan

, Lind

, Power

, Eriksson

, Raitakari

, Hägg

, Pedersen

, Veijola

, Järvelin

, Munafò

, Ingelsson

, Llewellyn

, Hyppönen

(2018) Habitual coffee consumption and cognitive function: A Mendelian randomization meta-analysis in up to 415,530 participants. Sci Rep 8, 7526.

17.

Lawlor

(2016) Commentary: Two-sample Mendelian randomization: Opportunities and challenges. Int J Epidemiol 45, 908–915.

18.

Davies

, Holmes

, Davey Smith

(2018) Reading Mendelian randomisation studies: A guide, glossary, and checklist for clinicians.k. BMJ 362, 601.

19.

Meddens

SFW

, de Vlaming

, Bowers

, Burik

CAP

, Linnér

, Lee

, Okbay

, Turley

, Rietveld

, Fontana

, Ghanbari

, Imamura

, McMahon

, van der Most

, Voortman

, Wade

, Anderson

, Braun

KVE

, Emmett

, Esko

, Gonzalez

, Kiefte-de Jong

, Langenberg

, Luan

, Muka

, Ring

, Rivadeneira

, Snieder

, van Rooij

FJA

, Wolffenbuttel

BHR

; 23and Me Research Team; EPIC- InterAct Consortium; Lifelines Cohort Study, Smith

, Franco

, Forouhi

, Ikram

, Uitterlinden

, van Vliet-Ostaptchouk

, Wareham

, Cesarini

, Harden

, Lee

, Benjamin

, Chow

, Koellinger

(2021) Genomic analysis of diet composition finds novel loci and associations with health and lifestyle. Mol Psychiatry 26, 2056–2069.

20.

Raubenheimer

, Simpson

, Le Couteur

, Solon-Biet

, Coogan

(2016) Nutritional ecology and the evolution of aging. Exp Gerontol 86, 50–61.

21.

Hewitt

, Walters

, Padmanabhan

, Dawson

(2016) Cohort profile of the UK Biobank: Diagnosis and characteristics of cerebrovascular disease. BMJ Open 6, e009161.

22.

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.

23.

Welter

, MacArthur

, Morales

, Burdett

, Hall

, Junkins

, Klemm

, Flicek

, Manolio

, Hindorff

, Parkinson

(2014) The NHGRI GWAS Catalog, a curated resource of SNP-trait associations. Nucleic Acids Res 42, D1001–D1006.

24.

Staley

, Blackshaw

, Kamat

, Ellis

, Surendran

, Sun

, Paul

, Freitag

, Burgess

, Danesh

, Young

, Butterworth

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

25.

Zetterberg

, Alexander

, Spandidos

, Blennow

(2009) Additional evidence for antagonistic pleiotropic effects of APOE. Alzheimers Dement 5, 75.

26.

Hemani

, Zheng

, Elsworth

, Wade

, Haberland

, Baird

, Laurin

, Burgess

, Bowden

, Langdon

, Tan

, Yarmolinsky

, Shihab

, Timpson

, Evans

, Relton

, Martin

, Davey Smith

, Gaunt

, Haycock

(2018) The MR-Base platform supports systematic causal inference across the human phenome. Elife 7, e34408.

27.

Greco

M FD

, Minelli

, Sheehan

, Thompson

(2015) Detecting pleiotropy in Mendelian randomisation studies with summary data and a continuous outcome. Stat Med 34, 2926–2940.

28.

Bowden

, Davey Smith

, Burgess

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

29.

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.

30.

Huedo-Medina

, Sánchez-Meca

, Marín-Martínez

, Botella

(2006) Assessing heterogeneity in meta-analysis: Q statistic or I2 index? Psychol Methods 11, 193–206.

31.

Verbanck

, Chen

C-Y

, Neale

, Do

(2018) Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 50, 693–698.

32.

Yavorska

, Burgess

(2017) MendelianRandomization: An R package for performing Mendelian randomization analyses using summarized data. Int J Epidemiol 46, 1734–1739.

33.

Jackson

, Bay

, Barter

, Ni

, Caudle

, Serra

, Wharton

, Hackney

(2020) The role of nutrition and inflammation on cognition in a high-risk group for Alzheimer’s disease. J Alzheimers Dis Rep 4, 345–352.

34.

Dong

, Xiao

, Cai

, Xu

, Wang

, Pan

, Yuan

(2016) Diet, lifestyle and cognitive function in old Chinese adults. Arch Gerontol Geriatr 63, 36–42.

35.

Neth

, Mintz

, Whitlow

, Jung

, Solingapuram Sai

, Register

, Kellar

, Lockhart

, Hoscheidt

, Maldjian

, Heslegrave

, Blennow

, Cunnane

, Castellano

C-A

, Zetterberg

, Craft

(2020) Modified ketogenic diet is associated with improved cerebrospinal fluid biomarker profile, cerebral perfusion, and cerebral ketone body uptake in older adults at risk for Alzheimer’s disease: A pilot study. Neurobiol Aging 86, 54–63.

36.

Taylor

, Sullivan

, Mahnken

, Burns

, Swerdlow

(2018) Feasibility and efficacy data from a ketogenic diet intervention in Alzheimer’s disease. Alzheimers Dementia (N Y) 4, 28–36.

37.

Gustafson

, Bäckman

, Scarmeas

, Stern

, Manly

, Mayeux

, Gu

(2020) Dietary fatty acids and risk of Alzheimer’s disease and related dementias: Observations from the Washington Heights-Hamilton Heights-Inwood Columbia Aging Project (WHICAP). Alzheimers Dement 16, 1638–1649.

38.

Boudrault

, Bazinet

, Ma

DWL

(2009) Experimental models and mechanisms underlying the protective effects of n-3 polyunsaturated fatty acids in Alzheimer’s disease. J Nutr Biochem 20, 1–10.

39.

Selkoe

, Hardy

(2016) The amyloid hypothesis of Alzheimer’s disease at 25 years. EMBO Mol Med 8, 595–608.

40.

Sahni

, Mangano

, Hannan

, Kiel

, McLean

(2015) Higher protein intake is associated with higher lean mass and quadriceps muscle strength in adult men and women. J Nutr 145, 1569–1575.

41.

Burns

, Johnson

, Watts

, Swerdlow

, Brooks

(2010) Reduced lean mass in early Alzheimer disease and its association with brain atrophy. Arch Neurol 67, 428–433.

42.

Boyle

, Buchman

, Wilson

, Leurgans

, Bennett

(2009) Association of muscle strength with the risk of Alzheimer disease and the rate of cognitive decline in community-dwelling older persons. Arch Neurol 66, 1339–1344.

43.

Glenn

, Madero

, Bott

(2019) Dietary protein and amino acid intake: Links to the maintenance of cognitive health. Nutrients 11, 1315.