Dietary Patterns and Cognitive Decline in an Australian Study of Ageing

ORIGINAL ARTICLE Dietary patterns and cognitive decline in an Australian study of ageing

SL Gardener1,2, SR Rainey-Smith1,2, MB Barnes3, HR Sohrabi1,2, M Weinborn1,2,4, YY Lim5, K Harrington5, K Taddei1,2,YGu6,7, A Rembach5, C Szoeke5, KA Ellis5,8,9, CL Masters5, SL Macaulay10, CC Rowe11, D Ames8,9, JB Keogh12, N Scarmeas6,7,13,14 and RN Martins1,2 for the AIBL Research Group15

The aim of this paper was to investigate the association of three well-recognised dietary patterns with cognitive change over a 3-year period. Five hundred and twenty-seven healthy participants from the Australian Imaging, Biomarkers and Lifestyle study of ageing completed the Cancer Council of Victoria food frequency questionnaire at baseline and underwent a comprehensive neuropsychological assessment at baseline, 18 and 36 months follow-up. Individual neuropsychological test scores were used to construct composite scores for six cognitive domains and a global cognitive score. Based on self-reported consumption, scores for three dietary patterns, (1) Australian-style Mediterranean diet (AusMeDi), (2) western diet and (3) prudent diet were generated for each individual. Linear mixed model analyses were conducted to examine the relationship between diet scores and cognitive change in each cognitive domain and for the global score. Higher baseline adherence to the AusMeDi was associated with better performance in the executive function cognitive domain after 36 months in apolipoprotein E (APOE) ε4 allele carriers (Po0.01). Higher baseline western diet adherence was associated with greater cognitive decline after 36 months in the visuospatial cognitive domain in APOE ε4 allele non-carriers (Po0.01). All other results were not significant. Our findings in this well-characterised Australian cohort indicate that adherence to a healthy diet is important to reduce risk for cognitive decline, with the converse being true for the western diet. Executive function and visuospatial functioning appear to be particularly susceptible to the influence of diet.

Molecular Psychiatry (2015) 20, 860–866; doi:10.1038/mp.2014.79; published online 29 July 2014

INTRODUCTION a healthy diet and slower cognitive decline.3–9 By contrast, studies Improved healthcare and nutrition are contributing signiﬁcantly to have also concluded that higher consumption of a healthy dietary – increased life expectancy, subsequently increasing the prevalence pattern is not protective against cognitive decline.10 13 These of age-related onset diseases, particularly Alzheimer’s disease previous studies have limitations, most notably, lack of compre- (AD). Cognitive decline is the progressive loss of cognitive funct- hensive cognitive testing whereby multiple cognitive domains are – – ions, including memory, and may lead to dementia, of which AD is assessed.3,7 9,11 13 There is a need for further longitudinal investi- the most common type.1 While there is no cure or effective treat- gation in highly characterised ageing cohorts. ment for AD, early intervention prevention programs hold consi- The aim of our study was to investigate the individual derable promise. Following particular dietary patterns represents association of three dietary patterns; (1) Australian-style Mediter- 14 one potential intervention strategy that is accessible to all. ranean diet (AusMeDi; ), (2) western diet and (3) prudent diet Individual diets contain both nutrient and non-nutrient sub- (a ‘healthy’ dietary pattern), with cognitive change over 3 years, stances rather than single foods. It may therefore be more useful assessed using a comprehensive neuropsychological battery. to examine indices of food and nutrient intake that express several This was investigated in a well-characterised, Australian elderly related aspects of diet concurrently rather than focus on con- cohort drawn from the larger Australian Imaging, Biomarkers sumption of single nutrients.2 and Lifestyle study of ageing (AIBL;15). The AusMeDi was Results from previous studies on the association between dietary constructed using an ‘a priori’ method and the western and patterns and future cognitive decline are inconsistent. Several prudent diet patterns constructed by factor analysis; an ‘a studies have shown a positive relationship between adherence to posteriori’ method.

1Centre of Excellence for Alzheimer's disease Research and Care, School of Medical Sciences, Edith Cowan University, Joondalup, WA, Australia; 2Sir James McCusker Alzheimer's Disease Research Unit (Hollywood Private Hospital), Perth, WA, Australia; 3CSIRO Computational Informatics, Glen Osmond, SA, Australia; 4School of Psychology, University of Western Australia, Crawley, WA, Australia; 5The Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Parkville, VIC, Australia; 6Taub Institute for Research of Alzheimer’s Disease and the Ageing Brain, Columbia University, New York, NY, USA; 7Gertrude H. Sergievsky Centre, Columbia University, New York, NY, USA; 8National Ageing Research Institute, Parkville, VIC, Australia; 9Academic Unit for Psychiatry of Old Age, St Vincent’s Health, Department of Psychiatry, The University of Melbourne, Kew, VIC, Australia; 10CSIRO Preventative Health Flagship, CMSE Parkville, VIC, Australia; 11Department of Nuclear Medicine and Centre for PET, Austin Health, Heidelberg, VIC, Australia; 12School of Pharmacy and Medical Sciences and Sansom Institute for Health Research, Division of Health Sciences, University of South Australia, Adelaide, SA, Australia; 13Department of Neurology, Columbia University College of Physicians and Surgeons, New York, NY, USA and 14Department of Social Medicine, Psychiatry, and Neurology, National and Kapodistrian University of Athens, Athens, Greece. Correspondence: Professor RN Martins, Centre of Excellence for Alzheimer's disease Research and Care, School of Medical Sciences, Edith Cowan University, 270 Joondalup Drive, Joondalup, 6027 WA, Australia. E-mail: [email protected] 15A full list of the AIBL investigators is available at www.aibl.csiro.au Received 31 October 2013; revised 17 June 2014; accepted 18 June 2014; published online 29 July 2014 Dietary patterns and cognitive decline SL Gardener et al 861 MATERIALS AND METHODS group was summed to derive the 33 groups for each individual. Factor Participants analysis (principle components) was conducted to derive dietary patterns.29 Food groups were entered as weight in grams per day. The This report describes data from 527 healthy control (cognitively ‘normal’) participants taken from the AIBL study15 who completed the Cancer factors were rotated by a varimax procedure resulting in non-correlated Council of Victoria food frequency questionnaire (CCVFFQ;16) at baseline. factors, to facilitate factor interpretability. In determining number of factors to retain, we considered components with an eigenvalue 41.25, Scree The AIBL study is a longitudinal study of 1112 volunteers including healthy 30 controls, those with mild cognitive impairment (MCI), and AD who are tests and interpretability. A cut-off of 0.20 was used to determine factor being assessed for prospective research into ageing and AD. MCI and AD loadings included in each pattern, and two major patterns were extracted, participants were excluded from this analysis as we are using a food labelled western pattern and prudent pattern; Table 1 shows factor frequency questionnaire that requires estimations of food intake over the loadings for these patterns. The dietary pattern score was constructed by previous year, and there is a potential for mis-classification due to limited summing intakes of the food groups weighted by factor loadings. accuracy in estimations. All AIBL volunteers were aged 60 years and above at baseline, and excluded if they had a history of non-AD dementia, Statistical analysis schizophrenia, bipolar disorder, significant current depression, Parkinson’s disease, cancer (other than basal cell skin carcinoma) within the last 2 All statistical analyses were performed using R version 3.0.1 (R Foundation years, symptomatic stroke, insulin-dependent diabetes, uncontrolled for Statistical Computing, Vienna, Austria). A P-value of 0.01 or smaller fi diabetes mellitus or current regular alcohol use exceeding two standard determined a signi cant result to balance the risk of type I and type II drinks per day for women or four per day for men. Further details errors, due to the large number of statistical tests performed. regarding recruitment, assessment, inclusion and exclusion criteria are Means, standard deviations and percentages are provided for the entire fi ε described in Ellis et al.15 The AIBL study is approved by the institutional cohort as well as following strati cation by APOE 4 allele carrier status. 2 ethics committees of Austin Health, St Vincent’s Health, Hollywood Private Independent samples t-tests and χ -analyses were conducted to evaluate Hospital and Edith Cowan University.15 group differences as appropriate. A series of repeated measures linear mixed model analyses (using maximum likelihood estimation and an unstructured covariance matrix) Cognitive assessments were conducted to examine the relationship between baseline diet score A comprehensive neuropsychological battery of well-validated measures and time (baseline, 18 and 36 month follow-up) with respect to cognitive was administered according to standard protocols (described else- change. In the first model (model 1), baseline diet score, time, APOE ε4 15,17 where; ). The battery assessed six cognitive domains (verbal allele carrier status (presence or absence of ε4 allele; the most common memory,18,19 visual memory,20 executive function,21,22 language,22,23 21,24 20,25 genetic risk factor for AD), body mass index (BMI), country of birth attention and visuospatial functioning ) at baseline, 18 and 36 (Australia or other), years of education (⩽12 years or 412 years), past month follow-up. Composite scores were calculated for each of these smoking status, energy intake, history of angina, stroke, hypertension, fi cognitive domains by rst converting raw scores for individual measures to heart attack and diabetes were entered as fixed factors; participant as a overall sample-based Z scores, then averaging Z scores for the relevant random factor; age as a covariate; diet score x APOE ε4 allele carrier status measures to compute a single composite score for that domain. Neuro- as an interaction term; and cognitive composite score as the dependent psychological tests were assigned to one of the six cognitive domains on variable. Model 2 included the same covariates as model 1, but without the the basis of a consensus among neuropsychologists, psychologists and neurologists involved in the AIBL study (Supplementary Table 1 lists cardiovascular disease (CVD) risk factors of past smoking status, history of neuropsychological tests used to calculate composite scores;26). We made hypertension, angina, stroke, heart attack and diabetes. We planned to fi a global cognitive score for each individual by summing the six cognitive repeat analyses following APOE genotype strati cation if there was ε domain scores and dividing by six. evidence of a diet score x APOE 4 allele carrier status interaction. The portion of variance explained was estimated by comparing the fixed effect variance with and without that particular term in the linear mixed model. Genotyping Fasting blood samples obtained using standard venipuncture of the antecubital vein, were collected into EDTA tubes containing Prostaglandin E1 (Sapphire Biosciences, Waterloo, NSW, Australia; 33·3 ng ml− 1)to Table 1. Factor loadings for the two major factors (diet patterns) prevent platelet activation. Whole blood was centrifuged to separate identified from principle components analysis leucocytes and DNA isolated from them, using Qiagen Midiprep kits, and apolipoprotein E (APOE) genotype was determined using PCR reaction Western diet pattern Prudent diet pattern amplification and restriction enzyme digest techniques.27 Factor 1 Factor Factor 2 Factor AusMeDi loadings loadings AusMeDi score was constructed following the most commonly described Red meats 0.50 Dark-yellow vegetables 0.72 method;14 however, cohort sex-specific medians rather than traditional fi Processed meats 0.49 Other vegetables 0.61 sex-speci c medians were used. A value of 1 was assigned for; (i) each Chips 0.49 Green leafy vegetables 0.47 beneficial component (fruits, vegetables, legumes, cereals and fish) if fi fi Re ned grains 0.48 Fruit 0.41 caloric-adjusted consumption was at or above the cohort sex-speci c Poultry 0.43 Cruciferous vegetables 0.39 median; (ii) each detrimental component (meat and dairy) if caloric- Condiments 0.42 Nuts 0.30 fi adjusted consumption was below the sex-speci c median; and (iii) a ratio Potatoes 0.40 Whole grains 0.29 of monounsaturated to saturated fats at or above the median. Individuals Sweets 0.38 Tomatoes 0.29 4 were assigned a value of 1 for mild-to-moderate alcohol consumption ( 5 Other breakfast 0.35 Fish 0.29 o 4 o to 25 g per day for females and 10 to 50 g per day for males). cereals AusMeDi score was generated for each participant by summing scores for Meat pies 0.32 Low-fat dairy 0.26 each category; AusMeDi score ranged from 0–9, with higher score indicat- 28 Margarine 0.31 Potatoes 0.25 ing greater adherence. Supplementary Table 2 provides the cohort sex- High-fat dairy 0.30 High-fat dairy − 0.25 specific medians used to construct the AusMeDi and compares them with Dark-yellow 0.30 Poultry 0.22 the traditional MeDi sex-specific medians. vegetables Fruit juice 0.30 Garlic 0.21 Western and prudent dietary patterns Snacks 0.29 Snacks 0.20 Beer 0.29 We first classified 101 food and drink items into 33 pre-defined food Hamburger 0.24 groups, minimising within person variations in intakes of individual foods. Pizza 0.23 Output from the CCVFFQ is provided as intake in grams per day of food Low-fat dairy − 0.21 and drink items. The grams per day intake of food or drink in each food

Table 2. Descriptive statistics for the healthy control cohort who completed the Cancer Council of Victoria food frequency questionnaire, and subgroups of the cohort based on stratiﬁcation by APOE ε4 allele carrier status

Total sample APOE ε4 carrier APOE ε4 non - carrier P-values for APOE (n = 527) (n = 148) (n = 379) genotype differences

Age, years 69.3 ± 6.4 68.4 ± 6.1 69.6 ± 6.5 0.054 Sex, men; n (%) 210 (39.8) 61 (41.2) 149 (39.3) 0.689 Country of birth, Australian; n (%) 417 (79.1) 120 (81.1) 297 (78.4) 0.490 Presence of APOE ε4 allele; n (%) 148 (28.1) Baseline body mass indexa,kgm2 26.3 ± 4.2 26.0 ± 4.2 26.4 ± 4.1 0.340 Baseline energy intake, kCal 1708.1 ± 557.2 1726.1 ± 577.7 1701.1 ± 549.6 0.644 Diabetes; n (%)b 38 (7.2) 13 (8.8) 25 (6.6) 0.383 Hypertension; n (%)b 201 (38.1) 51 (34.5) 150 (39.6) 0.277 Angina; n (%)b 30 (5.7) 10 (6.8) 20 (5.3) 0.510 Heart attack; n (%)b 26 (4.9) 9 (6.1) 17 (4.5) 0.447 Stroke; n (%)b 10 (1.9) 3 (2.0) 7 (1.8) 0.892 Past smoker; n (%) 228 (43.3) 66 (44.6) 162 (42.7) 0.700 Education ⩽ 12 years; n (%) 241 (45.7) 77 (52.0) 164 (43.3) 0.070 Baseline western diet score 144.3 ± 148.9 160.8 ± 152.2 137.9 ± 147.3 0.113 Baseline prudent diet score 307.5 ± 130.0 285.9 ± 116.1 316.0 ± 134.3 0.017 Baseline AusMeDi score 4.57 ± 1.6 4.7 ± 1.5 4.5 ± 1.7 0.164 Abbreviations: AusMeDi, Australian-style Mediterranean diet; APOE, apolipoprotein E. If not otherwise described, data are presented as mean ± s.d. of the mean. Bold indicates statistical signiﬁcance (Po0.05); characteristics compared using independent sample t-test for continuous variables and χ2 for categorical variables. aBody mass index is calculated as weight in kilograms divided by height in meters squared. bHistory of diabetes, hypertension, angina, heart attack or stroke; medical history obtained by participant self-report.

RESULTS In both models, there was a negative trend toward western diet This cohort comprised 527 participants (39.8% male) with average adherence and change in the visuospatial functioning composite age of 69.3 years. Participants had an average BMI of 26.3 kg m2, score (Po0.05; Table 3). Evidence of a western diet x APOE over 79% were born in Australia, 28% were carriers of the APOE ε4 genotype interaction (P = 0.025) warranted further analysis follow- fi ε allele (genotypes ε2/4, ε3/4 or ε4/4) and over 45% had 12 or less ing strati cation of the data by APOE 4 allele carrier status. The ε o years of education (Table 2). Following stratification by APOE association was driven by APOE 4 allele non-carriers (P 0.01; genotype, the only difference between APOE ε4 allele carriers and Table 4 and Figure 1b); that is, a higher western diet score was non-carriers was a higher mean prudent diet score in non-carriers associated with more decline in visuospatial functioning over 3 ε of the APOE ε4 allele (P = 0.017; Table 2). years in APOE 4 allele non-carriers. Western diet explained 3.6% All mean cognitive composite domain scores were higher at of the variance, and decreased to 3.4% without CVD risk factors. fi baseline than at 36 month follow-up; decline in cognitive For comparison, the other statistically signi cant variables (age, performance over 36 months is expected among the demo- BMI and years of education) explained 14.9%, 1.8% and 9.6%, ’ – respectively of the fixed effect variance in visuospatial functioning graphic assessed. Pearson s product moment correlations were fi conducted to determine the relationships between the dietary composite score. No signi cant relationships between prudent diet adherence and cognition were found, and the global patterns assessed. As expected, prudent and western diet scores cognitive score was not associated with adherence to any of the were negatively correlated (r = − 0.31, Po0.001). AusMeDi and three dietary patterns. prudent diet scores were positively correlated (r = 0.27, Po0.001). Western and AusMeDi diet scores were not significantly correlated. DISCUSSION Linear mixed models were used to analyse the relationship The objective of this study was to investigate the association of between AusMeDi, prudent and western diet, the six cognitive three dietary patterns; AusMeDi, prudent and western diet, with domains assessed and the global cognitive score. Two models cognitive change over a 3-year period. The main findings from this were analysed, one with a range of covariates included, and the study are that higher baseline adherence to the ‘healthy’ AusMeDi second model with the same covariates but excluding CVD risk is associated with less decline in the executive function composite factors. In both models, there was a positive trend toward score 36 months later in APOE ε4 allele carriers. By contrast, higher AusMeDi adherence and change in executive function composite baseline adherence to the ‘unhealthy’ western diet score is o score in the cohort as a whole (P 0.05; Table 3). Marginally associated with increased cognitive decline in the visuospatial significant AusMeDi x APOE genotype interaction (P = 0.011) functioning composite score 36 months later in APOE ε4 allele warranted further analysis following stratification of the data by non-carriers. APOE ε4 allele carrier status. The positive trend was only We saw no association between adherence to any of the three significant among APOE ε4 allele carriers (Po0.01; Table 4 and dietary patterns and global cognitive score over 36 months, Figure 1a); that is, a higher AusMeDi score was associated with less suggesting that the relationship between high and low dietary decline in executive function over 3 years in APOE ε4 allele pattern adherences is specific to decline only in the executive carriers. Diet explained 8.1% of the variance, increasing to 8.6% function and visuospatial functioning domains. without CVD risk factors. Even though 8.6% is a small portion of A potential mechanism through which diet may exert its effect the fixed effect variance in executive function composite score, on cognition is via the vascular system.31–33 Scarmeas et al.5 the other statistically significant variables (age, years of education investigated whether there was attenuation of the association and history of angina) contributed less, explaining only 5.6%, 5.6% between MeDi and AD when vascular variables were included in and 6.4%, respectively. their logistic regression models: they reported that their

Table 3. Results of linear mixed model analyses examining the association between baseline diet scores and change in cognitive performance over 36 months

AusMeDi score Prudent diet score Western diet score

Model Fully Adjusted without Fully Adjusted without Fully Adjusted without Change in composite cognitive domain adjusteda CVD risk factorsb adjusteda CVD risk factorsb adjusteda CVD risk factorsb

Verbal memory 0.0157 0.0154 0.0001 0.0001 − 0.0002 − 0.0002 Visual memory − 0.0085 − 0.0091 − 0.0000 − 0.0002 − 0.0002 − 0.0001 Executive function 0.0261 0.0265 0.0000 0.0002 − 0.0000 0.0000 Language 0.0013 0.0015 0.0002 0.0002 − 0.0002 − 0.0002 Attention 0.0089 0.0097 -0.0000 0.0007 − 0.0002 − 0.0002 Visuospatial functioning − 0.0267 − 0.0269 0.0003 0.0004 − 0.0005 − 0.0004 Global cognitive score 0.0010 0.0009 0.0001 0.0002 − 0.0002 − 0.0002 Abbreviations: AusMeDi, Australian-style Mediterranean diet; CVD, cardiovascular disease. P40.01. Standardised β-values shown. aFully adjusted model includes age, sex, years of education, APOE ε4 allele carrier status, country of birth (Australia vs other), baseline body mass index, baseline energy intake kCal, past smoking status, and history of hypertension, angina, stroke, heart attack and diabetes as covariates. bAdjusted model without CVD risk factors includes the same covariates as the fully adjusted model, without past smoking status, history of hypertension, angina, stroke, heart attack and diabetes. association between MeDi and AD was not mediated by vascular and protectins.35 Neuroinflammation is a process involved in AD comorbidities. To investigate the influence of vascular variables in which is thought to be triggered by amyloid beta (Aβ), a hallmark our population, we analysed two adjusted models, one controlling of AD pathology.36 It could be that the protective effect of for vascular comorbidities and one without inclusion of these AusMeDi adherence compared with the prudent diet is due to the covariates. Addition of CVD risk factors did not make any consumption of fish oils such as EPA and DHA. Conversely, the associations previously significant non-significant, or vice versa. negative association seen with the western diet and aspects of We saw a change in the significance level of the AusMeDi and cognition may be due to low consumption of DHA and EPA for executive function association in carriers of the APOE ε4 allele example, rather than high consumption of a detrimental food when CVD risk factors were excluded (from Po0.01 to Po0.001). group. Previous studies have reported associations between However, the effect size was increased in the model without CVD dietary patterns and biomarkers of inflammation, for example, risk factors, indicating that, in agreement with Scarmeas et al.,5 the MeDi has been associated with lower levels of inflammatory CVD risk factors do not appear to significantly contribute to the markers;28,37–40 in other dietary pattern studies, a healthy dietary observed relationship between AusMeDi and cognition, and the pattern (like our prudent diet) has been shown to lower inflam- improved significance may be due to lowering the degrees of mation marker levels and an unhealthy dietary pattern (like our freedom. western diet) has been shown to increase levels of inflammatory The deleterious effect of the western diet on visuospatial markers.41–46 Although the mechanism mediating the dietary functioning was smaller in size and this association is likely to be effects cannot be discerned from the present results, inflammation driven by variation on the Rey Complex Figure Test copy as our is a prime candidate that warrants future investigation. clock test data were negatively skewed and kurtosed. Within the Our results suggest that there are differential effects of diet on Rey Complex Figure Test copy, there is an executive function cognition that are likely to be contingent on APOE ε4 allele domain, namely planning and use of strategy.20,34 It could be that carriage; a finding that warrants further investigation. Compared instead of a negative association between the western diet and with the APOE ε2 and ε3 isoforms, APOE ε4 is a less functional visuospatial functioning domain, we are actually observing a cholesterol transporter.47 Management of cholesterol by fish oils negative association with the western diet in another area of in specific diets, like the AusMeDi, has been proposed to prevent executive function (planning), and therefore the only domain diet and treat the negative effects of the APOE ε4 genotype.48 Mice is affecting is executive function. Of note, the measures studies have shown a positive influence of fish oil containing diets comprising the executive function domain in the present study on behavioural and cognitive performances on APOE ε4 allele were heavily loaded with measures requiring cognitive flexibility, target replacement mice.49 A possible explanation for the switching and generativity, but did not include executive tasks AusMeDi effects observed only in APOE ε4 allele carriers in our relying heavily on planning. A recommendation for future studies study is that the non-carriers are already functioning highly in is to more comprehensively measure the executive function their cholesterol transport due to their ε2/ε3 genotype so components included in the visuospatial functioning domain to AusMeDi components are not able to exert a significant effect. confirm our reasoning. Ours is not the first study to report differential effects of lifestyle We saw no association with the healthy prudent dietary pattern factors on aspects of cognitive decline and AD-related pathology and cognitive decline or the global cognitive score. The AusMeDi that are contingent on APOE ε4 allele carriage. Notably, Brown and prudent diet patterns are similar but have some significant et al.50 showed that brain and blood Aβ levels are modulated by differences, most notably characterisation of fish intake which physical activity by APOE genotype-dependent mechanisms. contributes heavily to the AusMeDi but only represents a small Specifically, the authors reported that high physical activity is component of the prudent diet score. Eicosapentaenoic acid (EPA) associated with lower plasma Aβ1–42/1–40 in ε4 allele non-carriers, and docosahexaenoic acid (DHA) are n-3 fatty acids found in oily while higher exercising ε4 allele carriers had lower levels of brain fish. These fatty acids are able to affect a number of aspects of amyloid. Brown et al.50 concluded that physical activity appears to inflammation and mechanisms underlying their actions include attenuate the high levels of brain amyloid associated with ε4 altered cell membrane phospholipid fatty acid composition, carriers, and the apparent lack of effect on plasma Aβ levels may disruption of lipid rafts and inhibition of activation of the pro- reflect poor clearance of Aβ in the periphery in individuals with inflammatory transcription factor nuclear factor kappa B thereby the ε4 allele. Taken together, the results of both studies suggest reducing expression of inflammatory genes. EPA and DHA that in future the assessment of lifestyle in the context of APOE increase anti-inflammatory and inflammation resolving resolvins genotype may help the development of strategies tailored to the

© 2015 Macmillan Publishers Limited Molecular Psychiatry (2015), 860 – 866 Dietary patterns and cognitive decline SL Gardener et al 864 cance ﬁ 4 allele 0.0003 0.0000 ε − − ++ 0.0003 − ed by APOE ﬁ ) 0.0001 0.0001 a adjusted model without CVD n strati — 0.0001 0.0003 0.0003 0.0003 0.0005 0.0005 0.0002 0.0000 0.0001 0.0003 0.0004 0.0003 0.0003 0.0002 0.0002 − − − − − − 0.0006* (3.4% − ) a −− ttack and diabetes. Bold indicates statistical signi 0.0001 0.0003 0.0002 0.0001 0.0003 0.0003 − − − − − − diabetes as covariates. Model 2 0.0006* (3.6% − 0.0001 0.0001 − − ++ fully adjusted model includes age, sex, years of education, country of birth (Australia 0.0001 0.0001 0.0000 0.0000 — − − − 0.0000 0.0000 − − 1 0.0001 0.0004 0.0003 1 0.0001 0.0001 0.0001 1212 1 2 12 −− 0.0000 0.0000 − − ) 0.0000 0.0001 a 0.0690 0.0004 0.0004 0.0001 0.0001 0.0297 0.0000 0.0001 0.0003 0.0003 − 0.0815** (8.6% ) a ++ 0.0584 − AusMeDi score Prudent diet score Western diet score 0.0772* (8.1%

Portion of variance explained by diet score. a

a Figure 1. ( ) Higher baseline Australian-style Mediterranean diet (AusMeDi) adherence is associated with longitudinally increased performance in the executive function composite score in apolipo- 0.0041 0.0272 0.0276 0.000 0.0142 0.0213 0.0119 0.0002 0.0252 0.0200 0.000 0.0090 ε b) − − − − protein E (APOE) 4 allele carriers. ( Higher baseline western diet adherence is associated with longitudinally poorer performance in the visuospatial functioning composite score in APOE ε4 allele non- -values shown. −−

β carriers. Blue, 0 months; Black, 18 months; Red, 36 months. 0.0069 0.0164 0.0023 0.0110 − − − −

needs of the APOE ε4 allele carriers in the Alzheimer population which have the poorest prognosis with current treatment strategies. The positive relationship between AusMeDi adherence and

0.001). Standardised better performance on the executive function composite score

o described in this paper is consistent with previous ﬁndings P

** in other populations. In several American population studies, Results of linear mixed model analyses examining the association between baseline diet scores and change in cognitive performance over 36 months whe

4 allele carrier status 5

ε Scarmeas et al. have reported higher MeDi adherence associated 4–6 0.01, with lower AD risk and slower cognitive decline. A cross- Language Verbal memoryVisual memory Executive function 0.0149 0.0077 0.0170 0.0101 0.0368 Attention 0.0014 0.0026 0.0427 0.0289 Global cognitive score Visuospatial functioning o ‘ ’ P sectional Australian study reported that cognitively normal ModelAPOE 1 2 1 2 Change in composite cognitive domain * risk factors includes the same covariates as the fully adjusted model, without past smoking status, history of hypertension, angina, stroke, heart a Table 4. Abbreviations: AusMeDi, Australian-style Mediterranean diet;vs APOE, other), apolipoprotein baseline E; body CVD, mass cardiovascular index, disease. baseline Model energy 1 intake kCal,( past smoking status, and history of hypertension, angina, stroke, heart attack and carrier status individuals had higher adherence to the MeDi than MCI and AD

Molecular Psychiatry (2015), 860 – 866 © 2015 Macmillan Publishers Limited Dietary patterns and cognitive decline SL Gardener et al 865 participants.51 In a French study, it was observed that each as cognitively healthy at baseline. We recognise that ideally an additional MeDi score unit was associated with fewer Mini-Mental objective measure such as a nutrient biomarker panel that reflects State Examination (MMSE) errors at follow-up.3 A study conducted dietary pattern adherence would be employed as an additional in Chicago found that higher MeDi scores were associated with approach to circumvent the limitations of participant self-reported slower rates of cognitive decline.7 By contrast, the Personality and dietary intake. Although this approach has its own limitations, that Total Health through life study observed that greater MeDi is, a biomarker panel cannot be that exhaustive to include every adherence was not protective against cognitive decline. The possible nutrient in food. We report changes in cognition over apparent lack of protection in this study could be explained by the 3 years associated with dietary pattern adherence amongst heterogeneous nature of the study population. Furthermore, only individuals classified as ‘cognitively normal’ at baseline. Additional 66 participants from the original 1528 demonstrated any cognitive data collected at continuing follow-ups will further inform us of decline in the 4-year follow-up; a limitation with respect to the relationship between diet, cognitive decline and the transition generating sufficient statistical power to detect effects of the from healthy control to MCI or AD. However, we acknowledge that MeDi.13 Samieri et al.52 found that long-term MeDi adherence was the potential benefit of a healthy diet on cognition suggested by related to moderately better cognition, but not with cognitive longitudinal, observational studies such as AIBL will in time be change in a large cohort of older women from the Nurses’ Health demonstrated by randomised controlled trials. Study. This study used a limited telephone interview to assess Many aspects of our study provide confidence in our findings. cognitive status, with only global cognition and verbal memory We have utilised a well-characterised cohort with the vast majority assessed. Another study by Samieri et al.,53 this time using the being Caucasian, increasing the internal validity of our results. We Women’s Health Study, found no association with MeDi and have taken a very conservative approach by controlling for a wide cognitive decline. Again this study used the same telephone range of demographic variables and using a P-value of 0.01 to interview with only global and verbal memory assessed. determine statistical significance. The dietary data were collected Construction of the MeDi scores used in the Nurses’ Health Study using an instrument previously validated in earlier epidemiological and Women’s Health Study studies was modified from the studies.56 Furthermore, our findings are consistent with previous ‘traditional’ MeDi to increase relevance to American populations, reported studies which describe a ‘healthy’ or ‘prudent’ pattern in for example, refined and whole grains were delineated in the opposition to a ‘western’ or ‘processed’ dietary pattern.8,9,57–59 To modified MeDi. our knowledge, this is the first study extensively comparing the There are limited studies relating ‘a posteriori’ dietary patterns AusMeDi, western and prudent diet scores to cognitive decline in such as the western and prudent patterns investigated here, to an elderly, Australian cohort. Our results suggest a detrimental cognitive decline in the elderly. This could be due to identified nature of a western diet and propose the importance of adhering patterns depending on the study cohort, thereby limiting inter- to a healthy dietary pattern with respect to reducing risk for cogni- study comparison.54 Samieri et al.8 identified five dietary patterns, tive decline, with executive function and visuospatial functioning the ‘healthy diet’ pattern was associated with significantly lower seemingly most susceptible to the influence of diet. MMSE errors. Akbaraly et al.9 found that higher consumption of a ‘whole-food’ pattern (similar to our prudent diet pattern) was associated with lower odds of cognitive deficit, and higher CONFLICT OF INTEREST consumption of a ‘processed food’ pattern (similar to our western The authors declare no conflict of interest. diet pattern) was associated with higher odds of cognitive deficit; however, adjustment for education significantly attenuated most ACKNOWLEDGMENTS of the associations seen. ‘A priori’ patterns such as the AusMeDi are relatively easy to We acknowledge the contribution of Rosalind Miller (CSIRO) who developed the compute and reflect adherence to specific dietary patterns or method of estimating the portion of fixed effect variance in linear mixed models. guidelines. However, this also represents a disadvantage, as the Samantha Gardener is supported by a University Postgraduate Award from the results can only be as good as these underlying guidelines. Dementia Collaborative Research Centres program. Funding for the study is provided fi by the CSIRO Flagship Collaboration Fund and the Science and Industry Endowment Availability of dietary guidelines is required to de ne dietary Fund (SIEF) in partnership with Edith Cowan University (ECU), The Florey Institute of indices, and generally the guidelines are not disease specific, Neuroscience and Mental Health, Alzheimer's Australia (AA), National Ageing hence adherence may reduce the risk of some diseases but not Research Institute (NARI), Austin Health, CogState Ltd., Hollywood Private Hospital, others. Dietary indices need to be updated as dietary recommen- and Sir Charles Gairdner Hospital. The study also receives funding from the National dations for the population being analysed are modified, a process Health and Medical Research Council (NHMRC), the Dementia Collaborative Research of continual revision. Of further note, our study was conducted in Centres program (DCRC2) and the McCusker Alzheimer's Research Foundation and an Australian cohort, whose range of dietary intake is likely to be Operational Infrastructure Support from the Government of Victoria. different to the diet typical of Mediterranean countries. Therefore, fl we have named our MeDi score the AusMeDi and it re ects a REFERENCES different dietary intake when compared with Mediterranean populations. Subjects with high AusMeDi score may be potentially 1 Carter CL, Resnick EM, Mallampalli M, Kalbarczyk A. Sex and gender differences in categorized as having a low MeDi score relative to subjects from Alzheimer's disease: recommendations for future research. J Womens Health (Larchmt) 2012; 21: 1018–1023. Mediterranean populations. However, our results support the 2 Kant AK. Indexes of overall diet quality: a review. J Am Diet Assoc 1996; 96: notion that the relevance of the MeDi score is transferable to 785–791. different populations. Limitations of ‘a posteriori’ methods include 3 Feart C, Samieri C, Rondeau V, Amieva H, Portet F, Dartigues JF et al. Adherence to 55 the fact that they are exploratory in nature and are based on a Mediterranean diet, cognitive decline, and risk of dementia. JAMA 2009; 302: complex statistical analyses that require investigator-led selection 638–648. of a limited number of components to summarise the food 4 Scarmeas N, Stern Y, Tang MX, Mayeux R, Luchsinger JA. Mediterranean diet and patterns.54 risk for Alzheimer's disease. Ann Neurol 2006; 59: 912–921. ’ 5 Scarmeas N, Stern Y, Mayeux R, Luchsinger JA. Mediterranean diet, Alzheimer The CCVFFQ utilised in this study relied on participants estima- 63 – tions of food intake over the previous year; this is a common disease, and vascular mediation. Arch Neurol 2006; :1709 1717. fi 6 Scarmeas N, Luchsinger JA, Mayeux R, Stern Y. Mediterranean diet and Alzheimer limitation of studies of diet and can lead to mis-classi cation of disease mortality. Neurology 2007; 69: 1084–1093. dietary pattern adherence due to limited accuracy. This is parti- 7 Tangney CC, Kwasny MJ, Li H, Wilson RS, Evans DA, Morris MC. Adherence to a cularly important with cognitively impaired participants; to Mediterranean-type dietary pattern and cognitive decline in a community address this limitation we have assessed participants classified population. Am J Clin Nutr 2011; 93:601–607.

© 2015 Macmillan Publishers Limited Molecular Psychiatry (2015), 860 – 866 Dietary patterns and cognitive decline SL Gardener et al 866 8 Samieri C, Jutand MA, Feart C, Capuron L, Letenneur L, Barberger-Gateau P. 36 Agostinho P, Cunha RA, Oliveira C. Neuroinflammation, oxidative stress and the Dietary patterns derived by hybrid clustering method in older people: association pathogenesis of Alzheimer's disease. Curr Pharm Des 2010; 16: 2766–2778. with cognition, mood, and self-rated health. J Am Diet Assoc 2008; 108: 37 Chrysohoou C, Panagiotakos DB, Pitsavos C, Das UN, Stefanadis C. Adherence to 1461–1471. the Mediterranean diet attenuates inflammation and coagulation process in 9 Akbaraly TN, Singh-Manoux A, Marmot MG, Brunner EJ. Education attenuates the healthy adults: The ATTICA Study. J Am Coll Cardiol 2004; 44:152–158. association between dietary patterns and cognition. Dement Geriatr Cogn Disord 38 Fung TT, McCullough ML, Newby PK, Manson JE, Meigs JB, Rifai N et al. 2009; 27: 147–154. Diet-quality scores and plasma concentrations of markers of inflammation and 10 Kesse-Guyot E, Andreeva VA, Lassale C, Ferry M, Jeandel C, Hercberg S et al. endothelial dysfunction. Am J Clin Nutr 2005; 82: 163–173. Mediterranean diet and cognitive function: a French study. Am J Clin Nutr 2013; 39 Panagiotakos DB, Dimakopoulou K, Katsouyanni K, Bellander T, Grau M, Koenig W 97: 369–376. et al. Mediterranean diet and inflammatory response in myocardial infarction 11 Psaltopoulou T, Kyrozis A, Stathopoulos P, Trichopoulos D, Vassilopoulos D, survivors. Int J Epidemiol 2009; 38: 856–866. Trichopoulou A. Diet, physical activity and cognitive impairment among elders: 40 Esposito K, Marfella R, Ciotola M, Di Palo C, Giugliano F, Giugliano G et al. Effect of the EPIC-Greece cohort (European Prospective Investigation into Cancer and a mediterranean-style diet on endothelial dysfunction and markers of vascular Nutrition). Public Health Nutr 2008; 11: 1054–1062. inflammation in the metabolic syndrome: a randomized trial. JAMA 2004; 292: 12 Vercambre MN, Grodstein F, Berr C, Kang JH. Mediterranean diet and cognitive 1440–1446. decline in women with cardiovascular disease or risk factors. J Acad Nutr Diet 41 Hickling S, Hung J, Knuiman M, Divitini M, Beilby J. Are the associations between 112 – 2012; : 816 823. diet and C-reactive protein independent of obesity? Prev Med 2008; 47:71–76. 13 Cherbuin N, Anstey KJ. The Mediterranean Diet is Not Related to Cognitive 42 Fung TT, Rimm EB, Spiegelman D, Rifai N, Tofler GH, Willett WC et al. Association Change in a Large Prospective Investigation: The PATH Through Life Study. Am J between dietary patterns and plasma biomarkers of obesity and cardiovascular 20 – Geriatr Psychiatry 2011; :635 639. disease risk. Am J Clin Nutr 2001; 73:61–67. 14 Trichopoulou A, Costacou T, Bamia C, Trichopoulos D. Adherence to a Medi- 43 Nettleton JA, Steffen LM, Mayer-Davis EJ, Jenny NS, Jiang R, Herrington DM et al. 348 terranean diet and survival in a Greek population. N Engl J Med 2003; : Dietary patterns are associated with biochemical markers of inflammation and – 2599 2608. endothelial activation in the Multi-Ethnic Study of Atherosclerosis (MESA). Am J 15 Ellis KA, Bush AI, Darby D, De Fazio D, Foster J, Hudson P et al. The Australian Clin Nutr 2006; 83: 1369–1379. Imaging, Biomarkers and Lifestyle (AIBL) study of aging: methodology and 44 Esmaillzadeh A, Kimiagar M, Mehrabi Y, Azadbakht L, Hu FB, Willett WC. Dietary baseline characteristics of 1112 individuals recruited for a longitudinal study of patterns and markers of systemic inflammation among Iranian women. J Nutr 21 – Alzheimer's disease. Int Psychogeriatr 2009; :672 687. 2007; 137:992–998. 16 Keogh JB, Lange K, Syrette J. Comparative analysis of two FFQ. Public Health Nutr 45 Lopez-Garcia E, Schulze MB, Fung TT, Meigs JB, Rifai N, Manson JE et al. Major 2010; 13: 1553–1558. dietary patterns are related to plasma concentrations of markers of inflammation 17 Pike KE, Ellis KA, Villemagne VL, Good N, Chetelat G, Ames D et al. Cognition and and endothelial dysfunction. Am J Clin Nutr 2004; 80: 1029–1035. beta-amyloid in preclinical Alzheimer's disease: data from the AIBL study. Neuro- 46 Nettleton JA, Matijevic N, Follis JL, Folsom AR, Boerwinkle E. Associations between psychologia 2011; 49:2384–2390. dietary patterns and flow cytometry-measured biomarkers of inflammation and 18 Wechsler D. A standardised memory scale for clinical use. J Psychol 1945; 19: cellular activation in the Atherosclerosis Risk in Communities (ARIC) Carotid Artery 87–95. MRI Study. Atherosclerosis 2010; 212: 260–267. 19 Delis DC, Kramer JH, Keplan E, Ober B. California Verbal Learning Test—Second 47 Mahley RW. Apolipoprotein E cholesterol transport protein with expanding role in Edition. Psychological Corporation: San Antonion, TX, 2000. cell biology. Science 1988; 240: 622–630. 20 Meyers JE, Meyers KR. Rey Complex Figure Test and Recognition Trial. Professional 48 Schipper HM. Apolipoprotein E implications for AD neurobiology, epidemiology Manual: Psychological Assessment Resource, Inc.; 1995. and risk assessment. Neurobiol Aging 2011; 32: 778–790. 21 Strauss, E Sherman, EMS Spreen, O. A Compendium of Neuropsychological Tests: 49 Kariv-Inbal Z, Yacobson S, Berkecz R, Peter M, Janaky T, Lutjohann D et al. Administration Norms, and Commentary. 3rd edn. Oxford University Press: The isoform-specific pathological effects of apoE4 in vivo are prevented by a New York, 2006. fish oil (DHA) diet and are modified by cholesterol. J Alzheimers Dis 2012; 28: 22 Delis DC, Kaplan E, Kramer JH. Delis-Kaplan Executive Funtion System (D-KEFS). 667–683. Psychological Corporation: San Antonio, TX, 2001. 50 Brown BM, Peiffer JJ, Taddei K, Lui JK, Laws SM, Gupta VB et al. Physical activity 23 Saxton J, Ratcliff G, Munro CA, Coffey EC, Becker JT, Fried L et al. Normative data on the Boston Naming Test and two equivalent 30-item short forms. Clin Neuro- and amyloid-beta plasma and brain levels: results from the Australian Imaging, 18 – psychol 2000; 14: 526–534. Biomarkers and Lifestyle Study of Ageing. Mol Psychiatry 2012; :875 881. 24 Wechsler D.. Wechsler Adult Intelligence Scale3rd edition(WAIS-III). Psychological 51 Gardener S, Gu Y, Rainey-Smith SR, Keogh JB, Clifton PM, Mathieson SL et al. Corporation: San Antonio, TX, 1997. Adherence to a Mediterranean diet and Alzheimer's disease risk in an Australian 2 25 Libon DJ, Swenson RA, Barnoski EJ, Sands LP. Clock drawing as an assessment tool population. Transl Psychiatry 2012; : e164. for dementia. Arch Clin Neuropsychol 1993; 8:405–415. 52 Samieri C, Okereke OI, ED E, Grodstein F. Long-term adherence to the Medi- 26 Harrington KD, Lim YY, Ellis KA, Copolov C, Darby D, Weinborn M et al. The terranean diet is associated with overall cognitive status, but not cognitive 143 – association of Abeta amyloid and composite cognitive measures in healthy older decline, in women. J Nutr 2013; : 493 499. adults and MCI. Int Psychogeriatr 2013; 25: 1667–1677. 53 Samieri C, Grodstein F, Rosner BA, Kang JH, Cook NR, Manson JE et al. Medi- 24 27 Hixson JE, Vernier DT. Restriction isotyping of human apolipoprotein E by gene terranean diet and cognitive function in older age. Epidemiology 2013; : – amplification and cleavage with HhaI. J Lipid Res 1990; 31:545–548. 490 499. 28 Gu Y, Luchsinger JA, Stern Y, Scarmeas N. Mediterranean diet, inflammatory and 54 Alles B, Samieri C, Feart C, Jutand MA, Laurin D, Barberger-Gateau P. Dietary metabolic biomarkers, and risk of Alzheimer's disease. J Alzheimers Dis 2010; 22: patterns: a novel approach to examine the link between nutrition and cognitive 483–492. function in older individuals. Nutr Res Rev 2012; 25:207–222. 29 Kleinbaum DG, Kupper LL, Muller KE. Variable Reduction and Factor Analysis. 55 Hu FB. Dietary pattern analysis: a new direction in nutritional epidemiology. Curr Applied Regression Analysis and other Multivariate Methods. PWS-KENT Publishing Opin Lipidol 2002; 13:3–9. Company: Boston, 1988. 56 Ireland PJ, Jolley D, Giles G, O'Dea K, Powles J, Rutishauser I, Wahlqvist ML, 30 Kim JO, Mueller CW. Fcator Analysis: Statistical Methods and Practical Issues. Sage Williams J. Development of the Melbourne FFQ: a food frequency questionnaire Publications, Thousand Oaks, CA, 1978. for use in an Australian prospective study involving an ethnically diverse cohort. 31 Breteler MM. Vascular risk factors for Alzheimer's disease: an epidemiologic per- Asia Pac J Clin Nutr 1994; 3:19–31. spective. Neurobiol Aging 2000; 21:153–160. 57 Gu Y, Nieves JW, Stern Y, Luchsinger JA, Scarmeas N. Food combination and 32 Luchsinger JA, Mayeux R. Cardiovascular risk factors and Alzheimer's disease. Curr Alzheimer disease risk: a protective diet. Arch Neurol 2010; 67:699–706. Atheroscler Rep 2004; 6:261–266. 58 Pryer JA, Nichols R, Elliott P, Thakrar B, Brunner E, Marmot M. Dietary patterns 33 Cummings JL. Alzheimer's disease. N Engl J Med 2004; 351:56–67. among a national random sample of British adults. J Epidemiol Community Health 34 Waber DP, Holmes JM. Assessing children's memory productions of the 2001; 55:29–37. Rey-Osterrieth Complex Figure. J Clin Exp Neuropsychol 1986; 8:563–580. 59 Bamia C, Orfanos P, Ferrari P, Overvad K, Hundborg HH, Tjonneland A et al. Dietary 35 Calder P. Omega-3 polyunsaturated fatty acids and inflammatory processes: patterns among older Europeans: the EPIC-Elderly study. Br J Nutr 2005; 94: nutrition of pharmacology. Br J Clin Pharmacol 2012; 75:645–662. 100–113.

Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp)