Mediterranean diet is associated with lower white matter lesion volume in Mediterranean cities and lower cerebrospinal fluid Aβ42 in non-Mediterranean cities in the EPAD LCS cohort

The Mediterranean diet (MedDiet) has been associated with better brain health and reduced incidence of dementia. Few studies have compared the effects of the MedDiet in early Alzheimer's disease or compared the effects of the diet within and outside of the Mediterranean region. The Mediterranean diet adherence screener (MEDAS) and MEDAS continuous scores were calculated at the baseline visit of the European Prevention of Alzheimer's Dementia Longitudinal Cohort Study (n = 1625). The scores were included in linear regression models to test for associations with hippocampal volume, log-transformed white matter lesion volume, cerebrospinal fluid pTau18, and Aβ42. Higher MEDAS scores were associated with lower log-transformed white matter lesion volume (β: -0.07, standard error [SE]: 0.02, p < 0.001). This association was only seen in the Mediterranean region (β: -0.12, SE: 0.03, p < 0.001). In the non-Mediterranean region, higher MEDAS continuous scores were associated with lower cerebrospinal fluid Aβ42 (β: -68.30, SE: 14.32, p < 0.001). More research is needed to understand the differences in the associations seen with the MedDiet and Alzheimer's disease biomarkers in different European regions.


Introduction
High adherence to a Mediterranean diet (MedDiet) has consistently been associated with a 10%-40% lower incidence of dementia (Scarmeas et al., 2018;Yusufov et al., 2017).The MedDiet is an eating pattern rich in fruit, vegetables, legumes, nuts, olive oil, and fish, with a moderate consumption of red wine and a low consumption of red meat and processed foods (Trichopoulou et al., 2015).
This association with better brain health has been shown in studies conducted within the Mediterranean basin.For example, a study using data from the InCHIANTI study in Italy that recruited cognitively healthy older adults a higher risk of cognitive decline in those with the lowest adherence to the MedDiet, compared to medium and high adherence (Tanaka et al., 2018).Similarly, data from the Bordeaux cohort of the Three Cities study, representing cognitively healthy older adults, found that higher adherence to a MedDiet was associated with slower global cognitive decline (Féart et al., 2009).In the Spanish arm of the European Prospective Investigation into Cancer cohort study, involving cognitively healthy participants at baseline, high adherence to the MedDiet was associated with a lower risk of dementia compared to low adherence (Andreu-Reinón et al., 2021).An analysis of data from the Navarra center of the PREDIMED randomized controlled trial of the control (low fat) diet versus MedDiet with supplemental extra virgin olive oil versus MedDiet with supplemental nuts suggested a benefit of being in the MedDiet plus nuts group for BDNF levels (Sánchez-Villegas et al., 2011).However, these results are not consistent with other studies conducted in the Mediterranean finding no associations between the MedDiet and either cognition or risk for dementia (Féart et al., 2009;Munoz-Garcia et al., 2019;Nicoli et al., 2021).
Outside of the Mediterranean, there is also a mixed evidence base for associations between MedDiet and brain health.A populationbased study in The Netherlands found that the highest adherence to the MedDiet was associated with better cognition compared to the lowest adherence (Nooyens et al., 2021).However, other studies found no associations, such as an Australian study of communitydwelling older adults that reported no significant associations between MedDiet scores and either overall cognition or cognitive decline (Chen et al., 2022).Interestingly, despite beneficial associations of adherence to the MIND diet (a combination of the Mediterranean diet and the dietary approach to stop hypertension [DASH]) with cognitive performance proximal to death, an autopsy study found no association between diet and brain pathology (Dhana et al., 2021).Again, this study investigated the effects of the MedDiet in a non-Mediterranean population living in the USA.This lack of consistency both inside and outside of the Mediterranean region warrants further investigation to understand more about whether there are groups that may be particularly responsive to the MedDiet.
To better understand the mechanisms that may explain associations, or a lack of, between the MedDiet and both cognition and dementia risk, research should focus on the underlying pathology of neurodegeneration.A recent systematic review from our group (Gregory et al., 2023) identified a lack of evidence for associations between MedDiet adherence and brain health as measured by hippocampal volume and white matter hyperintensity volumes (indicators of Alzheimer's disease [AD] and cerebrovascular disease [CVD], respectively).In a small study of cognitively healthy participants (N = 52), living in Manhattan, USA, individuals with lower adherence to the MedDiet had more cortical thinning in brain regions of interest for AD (Mosconi et al., 2014) than individuals with higher adherence to the MedDiet.Compared to studies investigating cognition and dementia incidence, there are few studies reporting on neuroimaging measures, and more research is needed.A systematic review reported associations between the MedDiet and the AD biomarkers of tau and amyloid beta as measured on Positron Emission Tomography (PET), with all 4 studies included in the evidence synthesis conducted outside of the Mediterranean (Hill et al., 2019).More research is needed to explore associations between MedDiet and cerebrospinal fluid (CSF) AD biomarkers, both within and outside of the Mediterranean.
Recent work by our groups reports an analysis of the European Prevention of Alzheimer's Dementia Longitudinal Cohort Study (EPAD LCS), which identified significant associations between MedDiet adherence and allocentric processing in the EPAD LCS, with the strongest longitudinal effect seen for those participants living in the Mediterranean region (Gregory et al., 2022).Allocentric processing is the ability to mentally manipulate objects from a stationary point of view and is thought to represent changes to the hippocampus that occur in the earliest stages of AD (Fidalgo and Martin, 2016).This suggests that it is important to test for associations between MedDiet adherence and hippocampal volume in this cohort.Exploratory analyses in this group found the associations were seen in female but not male participants, suggesting that a sex-stratified approach may be important when considering diet and brain health.

Objective
Our aim in this study was to test the hypothesis that adherence to a MedDiet would be associated with higher hippocampal volume and lower white matter lesion volume (WMLV) in a European cohort of participants living at risk for AD dementia.In addition, we also wanted to explore associations between MedDiet adherence with CSF AD biomarkers to replicate previously reported findings from PET studies.To understand if there are differential effects of the MedDiet depending on region, we stratified the sample to individually analyze the associations between the MedDiet and AD outcomes within and outside of the Mediterranean.Finally, we aimed to conduct an exploratory analysis to test for any differences in dietary associations with AD outcomes between male and female participants.

Methods
The data used in this study were from the EPAD LCS.The EPAD LCS is a pan-European cohort of participants representing a spectrum of AD risk states from cognitively healthy volunteers to participants with preclinical AD and Mild Cognitive Impairment (MCI) (Ritchie et al., 2016, Solomon et al., 2018).Participants attended sites for baseline assessments, with follow-up visits at 6, 12, 24, 36, and 48 months for the duration of the project (2015-2020).For the purposes of this study, we looked at the baseline data for crosssectional analysis.All data collected for the baseline visit were collected within a 1-month period (dietary data, sociodemographics, Magnetic Resonance Imaging (MRI) scans, and CSF sample collection).After excluding participants with missing data on the dietary questionnaire, neuroimaging variables, or CSF biomarkers (n = 472), 1625 participants were included in the analysis.

Ethics and consent
Each country participating in the EPAD LCS obtained relevant favorable ethical opinions and complied with local governance procedures for each study site.Participants provided written informed consent prior to engaging with any protocol activities.Participants were required to have the capacity to consent at the time of study entry, as assessed during the informed consent conversation by a trained and delegated member of the research team, and all participants were free of dementia at the time of their baseline visit.

Data
This analysis used the EPAD LCS v.IMI baseline data set, following the approval of a data access request (ep-ad.org/open-access-data/overview).Data used in the preparation of this article were obtained from the EPAD LCS data set V.IMI, doi:10.34688/epadlcs_v.imi_20.10.30.EPAD LCS is registered at www.clinicaltrials.govIdentifier: NCT02804789.

Sociodemographic data
Participants self-reported their age (in years), education (in years), history of dementia in first-degree biological relative, and sex.Participants were categorized by Mediterranean or non-Mediterranean site using the European Commission biogeographical region based on the site they attended (European Commission, 2022).A list of site allocation is provided in the supplementary materials (Supplementary Table S1).Both Toulouse and Perugia were situated near the border and for the main analysis are included in the Mediterranean region.The baseline clinical dementia rating (CDR) score was used to describe whether the participant was cognitively healthy or considered to have MCI.

Diet questionnaire and dietary score calculations
Participants completed the Healthy Ageing Through Internet Counseling in the Elderly dietary questionnaire, which focuses on MedDiet food components (Richard et al., 2019).The Healthy Ageing Through Internet Counseling in the Elderly dietary questionnaire was designed to serve as a MedDiet adherence screener and asks participants to report their average consumption on a daily or weekly of relevant food groups, including meats, fish and seafood, vegetables, fruits, nuts, legumes, olive oil, and alcohol.The questionnaire was self-completed using pen and paper at the study sites and is available on request (Richard et al., 2019).Using these data, we calculated 2 MedDiet scores for each participant: the Mediterranean diet adherence screener (MEDAS) score and the MEDAS continuous score using previously published scoring methods.Full details of scoring methodologies are available in the supplementary materials (Supplementary Table S2).Briefly, the MEDAS score was calculated by allocating participants 0 or 1 point for each of 14 food groups depending on whether they met consumption criteria (Estruch et al., 2018).The MEDAS continuous was developed by Shannon et al. with points allocated for the same targets as MEDAS but in a continuous basis from 0 to 1 (Shannon et al., 2019).Continuous MedDiet scores have been demonstrated as more sensitive to differences in diet quality and may be particularly important to use outside of the Mediterranean region (Shannon et al., 2019).

Imaging
Participants completed an MRI at baseline on either a 1.5 T or 3 T scanner, depending on availability at the site.Volumetric analysis was completed by IXICO using an automated process supplemented with a visual endpoint quality check from trained image analysts.Freesurfer analysis was used to derive variables of interest (Lorenzini et al., 2022), with hippocampal volume and WMLV extracted from the data set for this analysis.The volumes of the left and right hippocampal volumes were averaged for each participant and normalized by intracranial volume.WMLVs were log-transformed prior to data analysis.

APOEε4
At baseline, all participants provided a DNA sample, which was used for APOEe4 genotyping.This analysis was carried using Taqman Genotyping in a single laboratory on QuantStudio12K Flex, with further details available in the EPAD LCS v500.0 baseline paper (Ritchie et al., 2020).

Physical health and medical history
Data were extracted on smoking history (self-reported as current, former, or nonsmoker), hypertension, hypercholesterolemia, hyperglycemia, diabetes, stroke, antihypertensive medication use, and diabetic medication use (all self-reported during an interview with a study clinician, included as binary variables indicating absence or presence).Body mass index (BMI) was calculated from height and weight measured during the baseline site visit and was included as a continuous variable.Finally, the Cardiovascular Risk Factors, Aging and Dementia (CAIDE) score was calculated, as per the original publication, using age, sex, education, systolic blood pressure, BMI, hypercholesterolemia, physical activity (Kivipelto et al., 2006).The CAIDE score was calculated without APOEε4, as this was included separately as a covariate.

Statistical analysis
All statistical analyses were completed in R (Version 4.1.0).Participants with missing data in the variables of interest were excluded from the analysis.We initially tested the cohort as a whole, and we then fitted a series of linear regression models.First, we fitted univariate regression models to test for associations between MEDAS and the primary outcomes (hippocampal volume and WMLV), followed by secondary outcomes (CSF Aβ 42 , CSF pTau 181 ) (Model 1).Next, we fitted partially adjusted models (Model 2 adjusted for: sex, age, education, family history, APOEε4; Model 3 additionally adjusted for CAIDE) and fully adjusted models (Model 4 additionally adjusted for: smoking, BMI, hypertension, hypercholesterolemia, hyperglycemia, diabetes, stroke, antihypertensive medication use, diabetic medication use, living in the Mediterranean).For models where WMLV was the outcome variable, intracranial volume was additionally included as a covariate in all models.All covariates were included in the models simultaneously.Our preplanned subgroup analysis splits the data set into participants living in Mediterranean and non-Mediterranean regions.All analyses were repeated after removing the variable indicating whether participants lived in the Mediterranean from the fully adjusted models.We then repeated the analyses with the MEDAS continuous score.We treated the analysis of hippocampal volume and WMLV as co-primary outcomes, and all other analyses were secondary outcomes.As we ran 2 concurrent analyses for each of the co-primary outcomes (MEDAS, MEDAS continuous), we corrected analyses using the Benjamini-Hochberg false discovery rate procedure to adjust for multiple comparisons.The false discovery rate adjustments resulted in any p values of 0.01 and below being accepted as statistically significant.Following the initial approach outlined above, we additionally ran a sensitivity analysis, excluding Toulouse and Perugia from the Mediterranean region.Finally, we undertook 2 exploratory analyses.First, we split the data set into male and female, and reran the above-described analysis with sex removed as a covariate, for the MEDAS scores only.Second, we split the data set by CDR score (0 or 0.5) to explore whether any findings were driven or explained by cognitive status in the cognitively healthy group (CDR 0).

Descriptive statistics
We included 1625 participants in this analysis.Compared to participants who were excluded due to missing data, the participants included were more likely to be APOEε4 carriers (39.0% vs. 28.4%,p < 0.001), were more likely to have a family history of dementia (66.3% vs. 58.1%,p < 0.001), were younger (65.48 ± 7.40 years vs. 67.34± 7.73 years, p < 0.001), were more likely to be male (44.1% vs. 43.2%,p = 0.03), and were more likely to live in the Mediterranean region (33.2% vs. 28.2%,p = 0.02).Full descriptive details of the cohort are presented in Table 1 and Supplementary Fig. S1 for MedDiet score distributions.The majority of the participants were considered to be cognitively healthy, with 73% of the sample having a CDR of 0 (n = 1188), while 27% of the sample had a CDR of 0.5 indicating MCI (n = 434), with 3 participants excluded due to missing CDR scores at the baseline visit.There was a larger proportion of CDR 0.5 participants in the Mediterranean region.This is likely due to the recruitment methods used, where sites in the Mediterranean may have been more likely to recruit clinical cohorts compared to other centers.

Whole cohort analysis: imaging
Analysis of the full cohort identified significant associations between higher MEDAS scores and lower log-transformed WMLV in unadjusted, partially, and fully adjusted models.(fully corrected β: −0.06, standard error [SE]: 0.02, p < 0.001).There was a significant association between the MEDAS continuous score and lower logtransformed WMLV models 1, 2, and 4 although not in model 3 (see Table 2).There were no significant associations between either of the MedDiet scores and hippocampal volume.

Whole cohort analysis: CSF biomarkers
Higher adherence to the MedDiet as measured by the MEDAS continuous score was associated with lower CSF Aβ 42 in partially and fully adjusted models (fully adjusted β: −38.96,SE: 11.44, p: 0.0007).There was no significant association between the MEDAS score and Aβ 42 .There were no significant associations between either of the MedDiet scores and CSF pTau 181 .Full details of all models are presented in Table 2.

Secondary analysis by biogeographical region: Mediterranean region
For participants living in the Mediterranean region, higher MEDAS and MEDAS continuous scores were associated with lower log WMLV in unadjusted, partially, and fully adjusted models (fully adjusted MEDAS β: −0.14, SE: 0.03, p < 0.001; MEDAS continuous β: −0.14, SE: 0.03, p: 0.003) (see Fig. 2).There were no significant associations between either of the MedDiet scores and hippocampal volume, CSF Aβ 42 , or CSF pTau 181 .Full details are available in Table 4.

Secondary analysis by biogeographical region: non-Mediterranean region
Participants living in non-Mediterranean regions with higher MEDAS continuous had lower CSF Aβ 42 in unadjusted, partially, and fully adjusted models (MEDAS continuous β: −67.09,SE: 14.30, p < 0.001) (see Fig. 3).There were no significant associations between either MedDiet score and CSF pTau 181 , hippocampal volume, or log-transformed WMLV.Full details of all models are presented in Table 5.

Sensitivity analysis
We conducted a sensitivity analysis with the MEDAS scores only to remove Toulouse and Perugia from the Mediterranean region as these sit on or close to the border of the biogeographically defined  region.When these regions were excluded, there remains a significant association between higher MEDAS score and lower logtransformed WMLV in all models (Supplementary Table S3).

Exploratory analysis: sex
An exploratory analysis stratifying the data set into male and female participants found that the association between MEDAS score and log-transformed WMLV was seen in both male and female participants, with no other significant associations, although only the unadjusted and fully adjusted models were significant in female participants.Overall, this suggests that there are no clear sex differences in associations between MedDiet adherence and AD biomarkers in the EPAD LCS (Supplementary Table S4).

Discussion
When analyzed as a full cohort, we found that there were significant associations between higher MedDiet scores with both  lower log-transformed WMLV and lower CSF Aβ 42 .Within the Mediterranean region, higher adherence to the MedDiet was associated with lower log WMLV.In contrast, there was a significant association between greater MedDiet adherence and lower CSF Aβ 42 in those living in non-Mediterranean regions.
Our findings from the Mediterranean region support previous evidence suggesting a positive benefit from adhering to a MedDiet pattern (Ballarini et al., 2021;Hill et al., 2019;Hoscheidt et al., 2022;Mosconi et al., 2014;Scarmeas et al., 2018;Titova et al., 2013;Yusufov et al., 2017).However, our findings from the non-Mediterranean region warrant further exploration as these are contrary to previous findings in the literature that have reported significant associations between MedDiet and lower Aβ 42 on PET scans in non-Mediterranean countries (Ballarini et al., 2021;Hill et al., 2019;Rainey-Smith et al., 2018).At the surface level, we could conclude that adhering to a MedDiet is harmful for brain health to those living in non-Mediterranean regions.However, that seems to lack biological plausibility given what we know from multiple previous studies about the nutritional benefits offered by eating a MedDiet.
We posit that the likeliest explanation for these seemingly contradictory findings is that participants living in non-Mediterranean countries eating a MedDiet have almost certainly only adopted this dietary pattern later in life.Although there is a statistically significant difference between the MEDAS scores for those living in and out of the Mediterranean region, the very small difference between the means is  unlikely to represent clinically meaningful differences in the quality of diet between the 2 regions.While in Mediterranean countries, the MedDiet has been the naturally accepted pattern of foods, adoption outside of the Mediterranean region coincided with the spread of knowledge about the potential benefits.Therefore, it is very unlikely that the EPAD LCS participants living outside of the Mediterranean have had a lifelong adoption of this diet.Although there is a recent trend to more Westernization of diets in the Mediterranean, looking at the decades in which our participants have lived through, there is a clear bias toward the availability of MedDiet food in the Mediterranean region (Vareiro et al., 2009).The participants who enrolled into the EPAD LCS were also highly motivated to engage in brain health-positive behaviors, as seen in other similar European cohorts (Cattaneo et al., 2020) and may have started engaging in healthy eating patterns as part of this.One possibility is that, without being aware of any underlying pathology, small changes in their ability to function were apparent, and they sought out options to tackle these insipid deteriorations in their brain health, as was seen in a Canadian study where participants, unaware of their amyloid status, demonstrated changes in their driving behaviors (Bayat et al., 2021).All of these would suggest that it is important for us to look at younger populations, living outside of the Mediterranean, to understand if longer adherence to the MedDiet is associated with a lower risk for AD or amyloidosis.It is also important to note that, by splitting the data set by region, we have limited the sample size, particularly in the Mediterranean region, which means that results may need to be interpreted with caution.However, as the results were seen in the whole cohort (lower WMLV and lower Aβ 42 ) when we used data from 1625 participants, we think it is unlikely that the results in the stratified analyses are spurious findings.
It is also possible that there are important social determinants influencing these results that we were not able to measure within this analysis.This could include household income, access to health care or health insurance, and employment history.There remain health inequities in Europe driven by social determinants, including low incomes, with limited access to healthy diet identified as an important area to tackle with fiscal policies (Marmot et al., 2012).Factors such as living alone compared to living with a family are also thought to influence the likelihood of consuming a MedDiet, with compliance thought to be easier in family units where food preparation efforts are shared (Navarro-Martínez et al., 2022).Similarly, social determinants of health, particularly low income, and education are known to be important risk factors for AD and related dementias (Majoka and Schimming, 2021).Consideration of key social determinants of health should be included in future studies to model their impact on associations between diet and health.
The above-described outcomes remained significant when the data set categorized Toulouse and Perugia as non-Mediterranean sites, with no sex differences identified.The sensitivity analysis results suggest that the association between higher MedDiet scores and lower WMLV is driven by data from participants in the true biogeographical Mediterranean region.It is interesting that there were no sex differences seen in this analysis.Previous analysis by our group identified significant associations between MedDiet adherence and allocentric processing in the EPAD LCS, with the results seen in female but not male participants (Gregory et al., 2022).In this analysis, excluding Toulouse and Perugia had a significant effect on the results with a broader effect on cognition seen.The difference in these results suggests that the outcomes used in this current analysis may not correlate directly to cognitive performance as associated with MedDiet.Analysis of the Rush Memory and Aging Project found that, despite a positive association between the MIND diet and cognitive performance proximal to death, there was no association between diet and brain pathology (Dhana et al., 2021).Although the Four Mountains Test is commonly conceived of as a test of hippocampal volume (Chen et al., 2010;Khan et al., 2014;Nestor et al., 2003;Ritchie et al., 2017;Rowe et al., 2007), it may be that variables such as functional connectivity are more important to consider to explain the sex differences in the Four Mountains Test analysis (Derbie et al., 2022;Grande et al., 2022;Rinaldi et al., 2020).There were differences in the associations seen between the diet and outcomes when the data set was split by CDR score.As there was a significant difference in the proportion of participants with each CDR score by region, it is possible that this is reflecting the geographical differences we have described earlier.
One significant limitation of our study is that we were not able to energy adjust the analyses.The dietary data questionnaire was exclusively focused on components of the MedDiet and, as such, is missing several food groups required to calculate overall energy intake.This limitation may, in turn, explain some of the unexpected non-Mediterranean region findings.Participants living in non-Mediterranean regions may have been consuming higher quantities of foods associated with a Western diet, which we know can impact negatively on brain health (Hoscheidt et al., 2022), and may even attenuate any benefits of the MedDiet (Agarwal et al., 2021).We are also limited by the self-report nature of the questionnaire, introducing the possibility of bias.Participants may have wanted to present a better picture of their diet to the research team but also may have forgotten elements of their diet as recall over time is known to be fallible.This is a well-recognized problem in the nutritional field, and the method of data collection has to be balanced between reduction of bias (such as using food diaries) and time considerations (brief recall questionnaires) (Ravelli and Schoeller, 2020).This is a cross-sectional analysis and as such there are significant limitations in being able to estimate longterm dietary patterns.Further work should investigate long-term dietary patterns and biomarkers of AD.
Future work should investigate biogeographical differences in response to the MedDiet, particularly in cohorts that have more thorough dietary data to support appropriate model adjustments and in cohorts that better represent the populations from which they are drawn.

Conclusion
Our study found that adherence to the MedDiet was associated with lower log WMLV for participants living in the Mediterranean region and lower CSF Aβ 42 for participants living outside of the Mediterranean region.We hypothesize this difference may be due to lifelong compared to later-life adoption of the MedDiet; however, further research is needed to explore and test this hypothesis.Understanding more about the associations between the MedDiet, brain health, and AD, particularly when and for how long adhering to the dietary pattern is important, will be critical both for designing public health strategies and providing individual patient advice.

Verification
The manuscript is not being considered for publication elsewhere and has been read and approved by all authors.Please note this work was presented at AAIC 2022 by the first author (Sarah Gregory) and the abstract for this may be available online.All authors meet the criteria for authorship stated in the ICMJE guidelines.We have stated funding and any conflicts of interest in the submission process.
Collaboration Award (MR/T001852/1).The funders had no involvement in the protocol design, data collection, analysis, or manuscript preparation.For the purpose of open access, the author has applied a creative commons attribution (CC BY) license to any author-accepted manuscript version arising.

CRediT authorship contribution statement
Fig.1

Fig. 1 .
Fig. 1.Site location by biogeographical regions in Europe alongside summary of results within and outside of the biogeographical European Mediterranean region.

Fig. 2 .
Fig. 2. Scatterplot of association between MEDAS score and log-transformed white matter lesion volume in the Mediterranean and non-Mediterranean regions.

Table 2
Linear regression models of MedDiet scores (MEDAS, MEDAS continuous) with neuroimaging and CSF outcomes b Model 1: MedDiet score; Model 2: MedDiet score, sex, age, education, family history, APOEε4; Model 3: MedDiet score, sex, age, education, family history, APOEε4, CAIDE; Model 4: MedDiet score, sex, age, education, family history, APOEε4, CAIDE, smoking, body mass index, hypertension, hypercholesterolemia, hyperglycemia, diabetes, stroke, antihypertensive medication use, diabetic medication use, living in the Mediterranean.Intracranial volume included as a covariate in all models with white matter lesion volume as outcome measure.Key: CSF, cerebrospinal fluid; MedDiet, Mediterranean diet; MEDAS, Mediterranean diet adherence screener; SE, standard error.Bold and italicised values indicate statistical significance.a Note white matter lesion volume is log transformed.b Indicates p value remains significant after FDR correction.

Table 3
Comparison of descriptive statistics between participants living in Mediterranean and non-Mediterranean countries

Table 4
Linear regression models of MedDiet scores (MEDAS, MEDAS continuous) with neuroimaging and CSF outcomes in Mediterranean countries (n = 540) MedDiet score, sex, age, education, family history, APOEε4, CAIDE, smoking, body mass index, hypertension, hypercholesterolemia, hyperglycemia, diabetes, stroke, antihypertensive medication use, diabetic medication use.Intracranial volume included as a covariate in all models with white matter lesion volume as outcome measure.Key: CSF, cerebrospinal fluid; MedDiet, Mediterranean diet; MEDAS, Mediterranean diet adherence screener; SE, standard error.Bold and italicised values indicate statistical significance.