Larger dlPFC and vmPFC grey matter volumes are associated with high adherence to the Mediterranean diet: A cross-sectional study in older adults

Abstract

Dietary self-control is associated with inter-individual differences in neuroanatomy. Yet, whether such inter-individual differences are also associated with healthier dietary patterns is yet to be determined. In this cross-sectional study, a total of 100 northern Portuguese older community-dwellers were assessed with regards to i) the adherence to a healthy dietary eating pattern – the Mediterranean diet (MedDiet), and ii) grey matter density (GMD) of brain regions associated with valuation and dietary self-regulation, the ventromedial (vmPFC) and dorsolateral prefrontal cortex (dlPFC), through voxel-based morphometry. Healthy food choices were ascertained through the Mediterranean Diet Adherence Screener (MEDAS) where higher scores indicated greater adherence to the MedDiet. Voxel-based morphometry showed that greater grey matter density in the dlPFC and vmPFC associated with a higher adherence to the MedDiet. These results replicate previous links between dietary decision-making measured under laboratory conditions and the neuroanatomy of the brain's valuation and self-control system. Importantly, they shed new light on the potential relevance of inter-individual differences in the neuroanatomy of these two brain regions for adhering to healthier dietary patterns in everyday life.

Introduction

Healthier dietary habits influence weight [35], health status [44], incidence of non-communicable diseases [55], risk of all-cause mortality, attributable disability-adjusted life-years [[55], [59]], and cognitive function [[29], [51]]. Additionally, both past and current high diet quality associate with better physical performance in older age [45], and even changes in diet later in life associate with less functional deterioration [37]. Healthy dietary patterns comprise a consistent choice of food items that are less appealing over other more appetitive ones. Such dietary regulation has been hypothesized to depend on cognitive processes mediated by the prefrontal cortex [62].

Cognitive processes such as delay discounting, attention, and response inhibition, have been associated with unhealthy dietary behaviours [[63], [28]]. For instance, delay discounting increases both in obese and overweight subjects [[23], [62], [64]] and in those who smoke, are physically inactive, drink excessive amounts of alcohol or engage in substance misuse [61]. Attention to foods, more specifically time required to shift attention towards other items, was also associated with obese binge eaters [9], and a poor response inhibition to food cues predicted overeating regardless of the weight and weight gain [5]. Interestingly, besides being correlated with unhealthy dietary choices, these cognitive processes were associated with altered brain patterns activation of two specific areas: the ventromedial prefrontal cortex (vmPFC) and the dorsolateral prefrontal cortex (dlPFC) [28].

Several functional MRI studies have reported that the activation patterns of the vmPFC and dlPFC influence the results of food decision paradigms, being the dlPFC more activated when delayed options are chosen [[20], [21], [40], [28]]. Interestingly, structural differences, such as larger grey matter volumes of these areas, were associated with greater regulatory success during dietary decision-making [53]. These findings suggest that food choices might be influenced by inter-individual differences in neuroanatomy of the vmPFC and dlPFC. Yet, it is still unknown whether grey matter structural differences of these brain areas are associated with the adherence to healthier dietary patterns, i.e. a larger number of healthy dietary-related choices. Thus, we hypothesized that larger GMD is positively correlated with high adherence to the Mediterranean diet, which is a dietary pattern previously found to be beneficial for health. This hypothesis was tested by correlating vmPFC and dlPFC grey matter volumes of older community-dwellers with the adherence to Mediterranean diet using voxel-based morphometry (VBM).

Material and methods

Participants

A prospective cohort was recruited by randomly selecting participants of a previous larger cross-sectional study (SWITCHBOX Consortium project, https://www.switchbox-online.eu/). Briefly, this cross-sectional study included a representative sample of the Portuguese older population (n = 1051) concerning age, gender, and education, from Guimarães and Vizela (two cities in Northern Portugal). Adults aged 50 years and over, free of dementia or another neuropsychiatric disorder, as per medical record, and able to provide informed consent were included. Each participant underwent a neurocognitive assessment whose scores, through a principal component analysis, allowed the identification of four dimensions of cognitive performance and mood: memory (based on the selective reminding test variables; CLTR, LTS, and delayed recall), executive function (Stroop parameters; words, colors, and words/colors; FAS parameter, admissible; and forward/backward digit spans), global cognitive status (the mini-mental state examination – MMSE), and mood (geriatric depression scale – GDS). Participants with cognitive impairment according to the MMSE were excluded, i.e. a total MMSE score <17 for individual with ≤4 years of formal school education and/or ≥72 years of age, and MMSE score <23 otherwise. Based on the four cognitive dimensions, four performance clusters were identified: “very good”, “good”, “poor” and “very poor” [49]. A total of 246 individuals from the “very good” and “very poor” cognitive clusters were randomly selected and contacted. Of these, 120 were integrated into the prospective cohort whose main aim was to study the healthy cognitive ageing [49]. Out of the 120, six were excluded due to missing data (two without the MEDAS score and four without total energy) and 14 due to invalid MRI acquisition (excess of movement during the acquisition) or no MRI acquisition. In total, 100 individuals were included in this analysis.

The study was approved by the national and local ethics review boards and conducted in accordance with the Declaration of Helsinki. All participants provided informed consent to participate in the present study. Clinical assessments took place at the Clinical Academic Center-Braga (Braga, Portugal).

MRI data acquisition

The imaging sessions took place at the Hospital of Braga (Portugal) on a Siemens Magnetom Avanto 1.5 T MRI scanner (Siemens Medical Solutions, Erlangen, Germany), using a 12-channel receive-only head-coil. The analysis was based on a structural acquisition. A 3D T1-weighted magnetization prepared rapid gradient echo (MPRAGE) sequence was used with the following parameters: 176 sagittal slices, TR/TE = 2,730/3.48 ms, flip angle = 7°, slice thickness = 1.0 mm, in-plane resolution = 1.0 × 1.0 mm2, FoV = 256 × 256 mm.

MRI data processing

Prior to data processing, all acquisitions were visually inspected to confirm that none of the participants had brain lesions nor critical head motion or artefacts that could compromise data quality.

DICOM files were converted to NIfTI format, and each participant’s anatomical image was centred and then segmented into grey, white matter and cerebrospinal fluid using the SPM12 segmentation tool, run in MATLAB Version 2016 (Mathworks, USA). Individual images were then co-registered between participants using Diffeomorphic Anatomical Registration through Exponentiated Lie Algebra (DARTEL). Following the alignment of the grey and white matter, the registered images were normalized to the Montreal Neurological Institute (MNI) stereotactic space using the DARTEL template and spatially smoothed using a Gaussian kernel with full width at half maximum of 8 mm.

Dietary intake

To assess adherence to a healthy eating pattern, the Mediterranean Diet Adherence Screener (MEDAS) was filled by a trained dietitian. MEDAS is a 14-item questionnaire measuring adherence to Mediterranean diet [54]. Each question is scored as 0 or 1 according to the cut off points that better represented the observed dose–response relationships between Mediterranean diet and a significant protection against myocardial infarction in a case-control study [30]. Thus, one point is assigned when the conditions are met: olive oil is used as principal source of fat for cooking (item 1); consuming 4 or more tablespoons of olive oil per day (item 2); 2 or more servings of vegetables per day (item 3); 3 or more pieces of fruit per day (item 4); <1 serving of red meat, hamburger or sausages (item 5); <1 serving of animal fat (item 6); <1 cup of sugar-sweetened beverages per day (item 7); 7 or more servings of wine per week (item 8); 3 or more servings of pulses per week (item 9); 3 or more servings of fish or sea food per week (item 10); <2 commercial pastries per week (item 11); 3 or more servings of nuts per week (item 12); preferring white meat instead of red meat (item 13) and eating a dish with sauce of tomato, garlic, onion or leeks sautéed in olive oil (item 14). In case the conditions were not met, 0 points were recorded [54]. MEDAS reliably ranks the MedDiet adherence [[54], [39]] with a moderate correlation with dietary intake [54], corroborated with objective blood biomarkers [22]. Additionally, a fair agreement between the Portuguese version of MEDAS and the food frequency questionnaire-derived Mediterranean diet adherence score was reported [17].

To ascertain the total energy intake, a 24 h dietary recall was applied. A manual adaptation of the Automated Multiple-Pass Method [41] was applied. After obtaining a list of all food items eaten on the previous day, the interviewer probed for frequently forgotten foods such as savoury snacks or candies. Then, the time and name of the eating occasion (e.g. breakfast) and a detailed description of the type of food item, cooking method and other relevant aspects were collected. Lastly, a final revision of the dietary intake took place [41]. Once the data was collected, energy intake (in Kcal) was determined by Nutrilog SAS software (version 3.1).

Covariates

Upon enrolment, sociodemographic data such as gender, age and number of years of formal education was collected. Reported physical activity levels and smoking habits were also ascertained. The body mass index (BMI) was obtained according to the standardized manual based on international recommendations [60] by measuring weight (Tanita® BF 350 Body Composition Analyzer; Tanita Corporation, Tokyo, Japan), height (stadiometer Seca® 217; Seca GmBH & Co Kg, Hamburg, Germany) and computing as Kg/m². Global cognitive status was assessed with the mini-mental state examination (MMSE) [13] and depressive mood status was ascertained by the Geriatric Depression Scale (GDS) [67]. Variables that could explain structural differences such as age [[19], [50]], school attainment [2], BMI [14], cognitive performance and mood [58] were included in the models. Energy intake was added to control for confounding and reduce extraneous variation [65].

Statistical analysis

Data on descriptive analysis were obtained with SPSS® software (version 24.0) and are reported as mean and standard deviation (SD) for the continuous variables and number of cases and percentages for categorical variables. XjView 9.7 toolbox was used to display brain regions (https://www.alivelearn.net/xjview).

Voxel-based morphometric analyses were conducted using SPM12 run in MATLAB Version 2016 (Mathworks USA). To determine whether adherence to Mediterranean diet was associated with GMD in brain regions of interest (ROIs: vmPFC and dlPFC), a voxel wise linear multiple regression model was conducted following equation i:

$$\mathit{G}\mathit{M}\mathit{D}={\beta }_0+{\beta }_{\mathit{MEDAS}}+{\beta }_{\mathit{Age}}+{\beta }_{\mathit{Gender}}+{\beta }_{\mathit{Education}}+{\beta }_{\mathit{Energy}}+{\beta }_{\mathit{BMI}}+{\beta }_{\mathit{GDS}}+{\beta }_{\mathit{MMSE}}+{\beta }_{\mathit{TGV}}+∊$$

Model I predicted the grey matter density (GMD) through the following regressors: number of healthy dietary decisions/adherence to the Mediterranean diet (MEDAS), participants’ age (Age), gender (Gender), education attainment (Education), energy intake (Energy), BMI (BMI), depressive mood (GDS), general cognition (MMSE) and global GM volume (TGV).

To examine whether the GMD of the brain regions of interest was associated with other healthy lifestyle behaviours, another linear regression analysis was conducted following equation ii:

$$\mathit{G}\mathit{M}\mathit{D}={\beta }_0+{\beta }_{\mathit{MEDAS}}+{\beta }_{\mathit{PA}}+{\beta }_{\mathit{AC}}+{\beta }_{\mathit{Age}}+{\beta }_{\mathit{Gender}}+{\beta }_{\mathit{Education}}+{\beta }_{\mathit{Energy}}+{\beta }_{\mathit{BMI}}+{\beta }_{\mathit{GDS}}+{\beta }_{\mathit{MMSE}}+{\beta }_{\mathit{TGV}}+∊$$

Model II therefore included two additional independent variables: level of reported physical activity (PA) and alcohol intake (AC).

We asked whether a larger number of healthy behaviours, i.e. the combined number of healthy lifestyle behaviours (HB), was associated with larger GMD following equation iii:

$$\mathit{G}\mathit{M}\mathit{D}={\beta }_0+{\beta }_{\mathit{HB}}+{\beta }_{\mathit{Age}}+{\beta }_{\mathit{Gender}}+{\beta }_{\mathit{Education}}+{\beta }_{\mathit{Energy}}+{\beta }_{\mathit{BMI}}+{\beta }_{\mathit{GDS}}+{\beta }_{\mathit{MMSE}}+{\beta }_{\mathit{TGV}}+∊$$

Four types of behaviour were considered: adherence to Mediterranean diet, alcohol consumption, smoking habits, and physical activity levels. For each behaviour, one point was assigned if the participants reported i) high adherence to the MedDiet, i.e. MEDAS ≥ 10 [31], ii) alcohol consumption between 1 and 35 g/day [25], iii) physical activity at least once per week and iv) currently not smoking. If the condition was not met, a zero would be recorded. In total, one could vary from zero to four points, where a participant would score zero if none of the behaviours met the condition and four if one met all conditions.

According to our a priori hypothesis, the analysis was restricted to the following ROIs: left dlPFC [−40,40,20] and left vmPFC [−9,46,−15] and right dlPFC [40,40,20] and right vmPFC [9,46,−15] [53]. For that purpose, a small volume correction was applied with a lenient initial whole-brain threshold of p <.005 uncorrected. Small volume corrected voxels were regarded as significant when falling below the p_FWE < 0.05 family wise error (FWE) small volume corrected on the cluster level.

Note, all linear regression models were controlled for variance related to age, gender, formal years of education, BMI, total energy, depressive mood, global cognition, and global GM volume (following ANCOVA normalization).

Results

Characterization of the sample

The sample was composed by 49 women and 51 men with a mean age of 65.6 (±8.21) and 5.31 (±3.82) years of formal education. On average, participants had a BMI of 29.5 (±3.70), where 12 % were normal weight, 49 % overweight and 39 % obese. Adherence to the MedDiet was on average 8.22 (±2.01), which is below the cut-off defined for high adherence, i.e. equal or greater than 10. As for alcohol consumption, 48 % had less than 25 g/day, 22 % between 25 and 50 g/day and 29 % greater than 50 g/day. More than two thirds, i.e. 70 %, never engaged in physical activity and 65 % were non-smokers (Table 1).

Table 1.

Demographic profile of the full cohort (n = 100).

Variable	All subjects
Female (n,%)	49 (49.0)
Age (M ± SD)	65.6 (8.21)
School years (M ± SD)	5.31 (3.82)
MEDAS score (M ± SD)	8.22 (2.01)
BMI in kg/m2 (M ± SD)	29.5 (3.70)
Total of Energy in Kcal (M ± SD)	2070 (808)
Alcohol g/day (M ± SD)	36.5 (39.8)
MMSE score (M ± SD)	26.6 (3.20)
GDS score (M ± SD)	11.0 (6.64)
Physical Activity (n,%)
Never	70 (70.0)
<3 times/week	13 (13.0)
≥3 times/week	9 (9.0)
Daily	8 (8.0)
Smoking habits (n,%)
Non-smoker	65 (65.0)
Former smoker	26 (26.0)
Smoker	9 (9.0)
No healthy lifestyle behaviours a	1.72 (0.09)

a – Combined number of healthy lifestyle behaviours as per the criteria: 1) high adherence to the MedDiet, i.e. MEDAS ≥10, 2) alcohol consumption between 1 and 35 g/day, 3) physical activity at least once per week and 4) currently not smoking. BMI, body mass index; GDS, geriatric depression scale; M, mean; MEDAS, Mediterranean diet adherence screener; MMSE, mini-mental state examination; SD, standard deviation.

VBM analysis

Small volume-corrected analyses showed that higher adherence to the MedDiet was associated with larger GMD of the studied ROIs: left dlPFC (t = 2.92, d.f. 90, r = 0.29), right dlPFC (t = 3.11, d.f. 90, r = 0.31) and the left vmPFC (t = 3.51, d.f. 90, r = 0.36) (Fig. 1 and Fig. S1, Table 2). These results were controlled for global GM volume, age, gender, BMI, education, energy intake, mood, and global cognitive performance (equation i). When further controlling for reported level of physical activity and alcohol intake, these results remained significant (equation ii, Table 3). Additionally, neither alcohol consumption nor reported levels of physical activity were associated with the studied brain areas (equation ii). Solely, BMI was associated with GMD of the dlPFC. Specifically, participants with higher BMIs also had lower left dlPFC GMD (t = 3.46, d.f. 90, r = -0.34) (Table 4). When studied whether an increase in the combined number of healthy lifestyle behaviours (equation iii) was associated with larger GMD of the vmPFC and dlPFC, none of the regions of interest showed significant association.

Fig. 1.

Sagittal view of the statistical parametric maps of the grey matter volume correlation with adherence to the MedDiet (threshold at p < 0.005, uncorrected for display) controlled for age, gender, formal years of education, BMI, total of energy, depressive mood, global cognition and global GM volume. The regions where the adherence to the MedDiet, as measured by MEDAS, was significantly correlated with larger GMD are shown in yellow. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 2.

Peak coordinates and Z-scores values of the VBM analyses for adherence to the MedDiet (equation i).

Region	BA	Cluster size (KE)	X	Y	Z	Peak z-score
L dlPFC	10	25	−34	41	23	2.85
R dlPFC	10	27	36	42	18	3.02
L vmPFC	11	11	−14	47	−17	3.39

Clusters at p <.05 corrected for familywise error corrected for small volumes. BA, Brodmann area; dlPFC, dorsolateral prefrontal cortex; L, left; MedDiet, Mediterranean diet; R, right; VBM, voxel-based morphometry; vmPFC, ventromedial prefrontal cortex.

Table 3.

Peak coordinates and Z-scores value of the VBM analyses for adherence to the MedDiet further controlled for reported physical activity levels and alcohol consumption (equation ii).

Region	BA	Cluster size (KE)	X	Y	Z	Peak z-score
L dlPFC	10	26	−34	41	23	2.84
R dlPFC	10	55	38	44	20	3.23
L vmPFC	11	7	−14	47	−18	3.20

Clusters at p < 0.05 corrected for familywise error corrected for small volumes. BA, Brodmann area; dlPFC, dorsolateral prefrontal cortex; L, left; MedDiet, Mediterranean diet; R, right; VBM, voxel-based morphometry; vmPFC, ventromedial prefrontal cortex.

Table 4.

Peak coordinates and Z-scores value of the VBM analyses for BMI.

Region	BA	Cluster size (KE)	X	Y	Z	Peak z-score
L dlPFC
Eq i	46	68	−47	35	15	3.38
Eq ii	46	12	−45	39	14	3.07
Eq iii	46	64	−45	33	14	3.36

Clusters at p < 0.05 corrected for familywise error corrected for small volumes. BMI, body mass index; BA, Brodmann area; dlPFC, dorsolateral prefrontal cortex; L, left; MedDiet, Mediterranean diet; R, right; VBM, voxel-based morphometry.

Discussion

This study shows that older community-dwellers with higher adherence to MedDiet have larger GMD of bilateral dlPFC and left vmPFC. We found that larger GMD of the dlPFC and vmPFC were associated with MEDAS after controlling for age, gender, intracranial volume, BMI, total of energy, mood, global cognition, and school years. The results remained unchanged when further controlling for additional health-related behaviours adherence. Likewise, higher BMI was associated with lower GMD of the left dlPFC.

Here, higher MEDAS scores, i.e. larger number of healthy dietary behaviours, was associated with larger GMD of the vmPFC and dlPFC. These results expand on the work developed by Schmidt et al. [53] where larger GMD of vmPFC and dlPFC was associated with higher dietary regulatory success in two food choice fMRI paradigms [53]. Likewise, Yao et al. [66] found that GM density of the dlPFC correlated with higher conscious restriction of food intake as assessed by the three Factor Eating Questionnaire Scale [66]. Both studies imply that individual structural differences of these dietary self-regulatory brain areas relate to food choices. Yet, these studies did not ascertain whether the subjects’ dietary patterns were associated with the structural differences. Previous studies have reported larger total brain volume associated with higher adherence to healthier dietary patterns [43]. However, to our knowledge, none reported associations between the Mediterranean diet and the vmPFC and the dlPFC. In the present study, which controlled for several factors that could impact on brain grey matter, our findings suggest that healthier dietary patterns are associated with larger GMD of self-control brain regions even after controlling for intracranial volume and other lifestyle behaviours. These differences in volume seem to be specific for dietary self-control as the study of other healthy behaviour was not correlated with GMD of these brain regions. Note that the current study design precludes interpretation of causal relationships. Few studies have focused on changes over time of structural brain changes with differences in BMI and/or dietary habits. In a RCT testing whether the Mediterranean-DASH Intervention for Neurodegenerative Delay (MIND) diet could improve cognitive performance in middle-aged obese individuals, two groups of women with obesity underwent a 3-months hypocaloric intervention with a similar caloric restriction and balance of mixed foods, but the MIND group recorded a significant increase in the intake of green leafy vegetables, berries, olive oil, fish, and whole grains, along with a significant decrease in butter, cheese, red meats, fast foods, and sweets. When compared to control group, the MIND diet intervention group displayed an increase in the surface area of the inferior frontal gyrus, a brain region also involved in decision-making [3]. Whereas these finding suggest that changes in dietary habits associate with brain structural changes, it is unclear whether these two groups differed in terms of density of the grey matter of the two regions of interest at baseline. In fact, baseline differences have been shown to predict future change in BMI. For instance, Yokum et al. [68] reported that low grey matter volume of regions involved in the inhibitory control, such as the superior frontal gyrus and middle frontal gyrus, were associated with increased BMI at 12-month follow-up. Additionally, Cerit et al. [8] found that a greater baseline functional connectivity of the hippocampus, frontoparietal network and default mode network nodes predicted improvement in cognitive control and weight loss at 12-months follow-up in patients who underwent bariatric surgery. Thus, further research is required to understand the direction of the effect.

Both dlPFC and vmPFC activation have been associated with dietary self-control. For instance, when compared to lean subjects, obese participants had a less activated left dlPFC upon food cues [26]. Similarly, after a standardised meal, the left dlPFC remained less activated regardless of the size of the meal in obese subjects as opposed to lean subjects [38]. Moreover, participants who chose delayed rewards had a more active left dlPFC and the connectivity from the left dlPFC to vmPFC was stronger at the time of choice [20] when food items were regarded as healthy [10]. In addition, weight loss positively correlated with increased connectivity from vmPFC to the left dlPFC [36]. These findings suggest that lower activation of the left dlPFC and vmPFC upon food cues and after meals affects satiety and satiation, which then either leads to food overconsumption or as an acquired feature of obesity, rends difficult the weight loss [26]. Our results suggest that these differences in activation might be associated with structural differences. A likely explanation for the structural differences of volume is a lower activation of these brain regions with lower scores of MEDAS, i.e. fewer healthy dietary choices. As previous studies suggested that decreases on the regional cerebral blood flow precede volumetric changes of the respective area [4] and that repeated activation of a given brain region resulted in an increased grey matter density [[24], [33]], it is possible that subjects who activate more frequently the dlPFC and vmPFC to make more healthier dietary choices have larger GMD of the activated region. Support for this interpretation comes from findings on other domains. For instance, both enhanced activation and larger grey volumes of the insula were associated with increased strategic reasoning in experienced gamers [15]. To test this tentative mechanism, studies including mediation analyses while applying a neuroimaging approach and/or nutrient biomarkers could provide important information on the mechanisms [69]. In fact, this approach has been applied in the field of nutritional cognitive neuroscience providing interesting findings. For example, whereas white matter microstructure of the fornix mediated the relationship between the ratio of plasma polyunsaturated fatty acids and memory [71], the grey matter volume of the left rostral anterior cingulate cortex partially mediated the relationship between omega-3 polyunsaturated fatty acids biomarkers and cognitive flexibility [70]. Similarly, MedDiet associated with cognitive scores and larger grey matter volume of the dentate gyrus, but brain volume losses did not mediate the relationship between the MedDiet and cognitive scores [25]. Conversely, the association of the inflammation-related nutrient pattern with the visuospatial cognitive domain was mediated by the total grey matter volume [18].

It is widely known that as one ages, brain atrophy follows [12]. Yet, unlike the degeneration recorded in both white and grey matter within the prefrontal region, vmPFC volume remains unmodified with ageing as part of it locates in the orbital surface, which in turn seems to be one of the few regional PFC areas that remained unchanged [47]. In fact, our data also suggests that the GMD of the ROIs were not associated with age. Unlike vmPFC, GMD of the dlPFC seems to decrease with age [42]. However, our analysis did not find an effect of age on the GMD of the dlPFC. Besides volumetric alterations, age-related brain changes also raise the issue of whether these grey matter reductions influence cognitive skills. For instance, as working memory, attention, and executive control decline with ageing, decision making is likely to be affected [48] and therefore food-related decision making could be impacted. Although early work indicated that older adults had longer delay discounting [[16], [27]], subsequent research has reported the opposite for health rewards and social rewards [56]. Yet, food-related decision making comprises other factors such as food preferences. For instance, as older adults experience changes in taste and smell, these alterations may shift their preferences towards stronger flavours [52], or changes in cooking processes or chosen/added ingredients [57], which can affect health outcomes. Lastly, with age-related brain atrophy, a misreporting of the dietary intake could occur [72]. In the present study sensitivity analysis based on age was not possible due to the sample size. However, age-related cognition impairment is unlikely to affect our results as none of the participants had a diagnosis of dementia nor mild cognitive impairment. In addition, both age and global cognitive performance were included in the model presented in this study.

Several factors contribute to the robustness of this study. The large sample size, the broad set of confounding variables controlled for including global cognition and the use of an instrument to measure adherence to the Mediterranean diet with regards to the last year. The present study also has its limitations. The use of the MMSE has limitations, such as being sensitive to age and education [[1], [32]], and it exhibits floor and ceiling effects [34]. The latter might have resulted in immeasurable cognitive functioning variance. MEDAS does not capture certain eating habits such as ready-to-eat meals and savoury snacks, which might yield different scores [39]. Yet, in this sample the intake of these food items was regarded as unlikely as ascertained by the 24 h dietary recall. As for the volumes, a volumetric measure might not provide information on the functionality of the tissue. A larger volume due to hypertrophy [47] resulting from a compensation mechanism could explain the volumetric differences [7]. However, previous research has shown a link between the GM volumes of these areas and a greater regulatory success during dietary decision-making paradigms [53]. Note that, apart from the dlPFC and the vmPFC, other brain systems and related regions have been implicated in food intake. Briefly, the homeostatic system along with the reward circuit, attention and emotion and memory systems are interlinked with each other and with the cognitive control systems [[6], [11], [46]]. Lastly, as a cross-sectional observational study, neither can causality be established nor residual confounding ruled out. Larger samples and both prospective studies and intervention studies implicating these areas are required to understand the relationship between changes in the GMD of these areas and eating behaviours.

Conclusion

The present study has shown that healthier eating habits are associated with larger GMD of the dietary self-regulatory regions, going beyond the reported food choices or lab-related food choices and regardless of the BMI and age. Our findings suggest that food decisions and dietary self-control are associated with structural traits such as the density of the grey matter of vmPFC and dlPFC. In turn, this offers new insights for future research on dietary behaviour change.

Funding

The work was developed under the scope of the projects SwitchBox (European Commission, FP7; contract HEALTH-F2-2010-259772) and TEMPO-Better mental health during aging based on temporal prediction of individual brain aging trajectories (FCT - Fundação Calouste Gulbenkian; Contract Grant No P-139977). BR was funded by Physical Activity and Nutritional INfluences In ageing (PANINI) Innovative Training Network (European Commission, Horizon 2020, Grant number: 675003). TC, RM, PSM, and CP-N were supported by FCT PhD scholarships; SFRH/BD/90078/2012 to TC; PDE/BDE/113604/2015 from the PhD-iHES Programme to RM; PDE/BDE/113601/2015 to PSM; PD/BD/106050/2015 from the Inter-University Doctoral Programme in Aging and Chronic Disease (PhDOC) to CP-N. JMS was supported by a fellowship of the project SwitchBox. NCS was a recipient of a Research Assistantship through the FCT Investigator Programme 200∞Ciência. Financial support was provided by National funds, through the Foundation for Science and Technology (FCT) - project UIDB/50026/2020 and UIDP/50026/2020, and by the project NORTE-01-0145-FEDER-000039, supported by Norte Portugal Regional Operational Programme (NORTE 2020), under the PORTUGAL 2020 Partnership Agreement, through the European Regional Development Fund (ERDF).

CRediT authorship contribution statement

Belina Rodrigues: Data curation, Formal analysis, Methodology, Visualization, Writing – original draft, Software, Writing – review & editing. Carlos Portugal-Nunes: Investigation, Writing – review & editing. Ricardo Magalhães: Software, Formal analysis, Writing – review & editing. Liane Schmidt: Software, Formal analysis, Writing – review & editing. Pedro Silva Moreira: Investigation, Writing – review & editing. José Miguel Soares: Investigation, Writing – review & editing. Teresa Costa Castanho: Investigation, Writing – review & editing. Paulo Marques: Investigation, Writing – review & editing. Nuno Sousa: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing – review & editing. Nadine Correia Santos: Conceptualization, Methodology, Project administration, Supervision, Writing – review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: