- AD
-
Alzheimer disease
- 3C
-
Three-city
- CDR
-
clinical dementia rating
- HR
-
hazard ratio
- MCI
-
mild cognitive impairment
- MeDi
-
Mediterranean diet
- WHICAP
-
Washington Heights-Inwood Columbia Aging Project
The Mediterranean diet (MeDi) has been first described in the Seven-Country study by Keys et al., who reported low mortality rates, in particular from CHD, in populations of the Mediterranean basin with olive oil as the main fat source( Reference Keys, Menotti and Karvonen 1 ). The traditional diet of these populations was also characterised by a high consumption of plant foods such as fruits, vegetables and legumes, and a low consumption of meat. Since then, a growing body of evidence has emerged for a beneficial role of the MeDi in several chronic diseases, and the concept of MeDi as a healthy dietary model has been increasingly recognised( Reference Roman, Carta and Martinez-Gonzalez 2 ). In a recent meta-analysis involving more than two millions subjects, greater adherence to a Mediterranean-style diet, as evaluated by the ‘MeDi score’, was associated with longer survival, reduced risk for cardiovascular mortality and cancer incidence and mortality( Reference Sofi, Cesari and Abbate 3 , Reference Sofi, Abbate and Gensini 4 ). In addition, a limited number of studies have addressed the relation between the MeDi and cognitive function and neurodegenerative diseases( Reference Sofi, Abbate and Gensini 4 , Reference Alles, Samieri and Feart 5 ). In the present paper, we update available knowledge on the relationship between adherence to the MeDi and cognitive decline, risk of dementia or Alzheimer's disease (AD). We also discuss the reasons for some inconsistent results across studies.
Dementia: a worldwide public health challenge
Dementia and AD, its most frequent form, are responsible for a considerable public health challenge. The prevalence of dementia increases exponentially with age from approximately 1% in the age group 65–69 years to 30% at age 90 years and older( Reference Lobo, Launer and Fratiglioni 6 ). Given the longer life expectancy, the prevalence of dementia is expected to increase considerably according to most projections( Reference Brayne 7 ). In 2005, more than twenty-four million people suffered from dementia, and it was estimated that 4·6 million new cases of dementia were arising every year( Reference Reitz, Brayne and Mayeux 8 ). There is no aetiologic treatment for AD available yet, and the strongest risk factors for dementia and AD are age, and possession of variants of several genes, which are not modifiable( Reference Lambert, Heath and Even 9 – Reference Sleegers, Lambert and Bertram 11 ). Hence, finding effective preventive strategies for cognitive decline or dementia is of utmost importance. These preventive strategies should aim at delaying the onset of mild cognitive impairment (MCI)( Reference Petersen 12 ), an unstable but still potentially reversible stage, in order to avoid or delay the conversion to dementia.
Although the aetiology of dementia and AD is still partly unknown, a growing body of evidence suggests that complex interactions between genetic factors and environmental conditions, involving multiple pathophysiological mechanisms, probably occur to initiate and exacerbate the neurodegenerative process( Reference Fotuhi, Hachinski and Whitehouse 13 – Reference Jack, Knopman and Jagust 16 ). Among environmental factors, cardiovascular risk factors may lead to various lesions in the brain which modify the risk of late-onset dementia( Reference Fotuhi, Hachinski and Whitehouse 13 ). In addition, regular physical exercise or other lifestyle risk factors have been associated with dementia risk. However, overall, no definite conclusion on the impact of these factors could be established to date, since results from randomised controlled trials did not corroborate most of the observational findings, highlighting the complexity of this disease( Reference Plassman, Williams and Burke 17 – Reference Devanand, Lee and Luchsinger 19 ).
Among lifestyle factors, several epidemiological data underscored a possible protective role of nutrition( Reference Filley and Anderson 18 , Reference Luchsinger, Noble and Scarmeas 20 – Reference Barberger-Gateau, Lambert and Feart 25 ). Most nutrients that have been individually associated with cognitive decline, dementia or AD in preclinical or epidemiological studies are found in the MeDi: MUFA, found in large amounts in olive oil (the hallmark of the MeDi), long-chain n-3 PUFA, mainly provided by fish and seafood, vitamins B12, folate, and antioxidants (vitamins C and E, carotenoids, flavonoids and selenium) provided by plant foods( Reference Gomez-Pinilla 22 , Reference Reynolds 26 – Reference Barberger Gateau, Feart, Samieri and Yaffe 40 ). However, reports about consumption of single nutrients or foods have been inconsistent and an integrative approach of the diet considering the additive and/or synergistic effects of each food component should be privileged( Reference Jacobs, Gross and Tapsell 41 – Reference Tucker 44 ). Therefore, beyond the study of individual dietary components, there has been an increasing interest in the influence of dietary patterns on cognitive health( Reference Alles, Samieri and Feart 5 , Reference Barberger-Gateau, Raffaitin and Letenneur 45 , Reference Gu and Scarmeas 46 ). This review focused on available epidemiological knowledge on the relationship between adherence to the MeDi and cognitive function.
The Mediterranean diet: basic concepts
The Seven-Country study was the pioneer observational study that first described the Mediterranean-style diet in the early 1960s( Reference Keys, Menotti and Karvonen 1 ). This study evidenced that countries from South Europe (Italy, Yugoslavia and Greece) had a higher life expectancy and lower rates of CHD, cancers and some other chronic diseases( Reference Keys, Menotti and Karvonen 1 ). Hence, the authors hypothesised that the exceptional health of these populations from the Mediterranean basin may be attributed to their traditional diet( Reference Trichopoulou, Kouris-Blazos and Vassilakou 47 ). The MeDi is characterised by abundant consumption of plant foods such as fresh fruits, vegetables, breads, other forms of cereals, potatoes, beans, nuts and seeds; olive oil as the main source of fats, providing notably monounsaturated lipids; a low-to-moderate intake of dairy products in the form of cheese and yoghurt; a low-to-moderate consumption of fish depending of the proximity of the sea; a low-to-moderate consumption of poultry; fewer than four eggs consumed per week; low amount of red meat and wine consumed in low-to-moderate amounts, normally during meals( Reference Willett, Sacks and Trichopoulou 48 ). There is no single MeDi, but several definitions, because dietary habits vary considerably across countries bordering the Mediterranean sea( Reference Sofi 49 , Reference Bere and Brug 50 ). Nevertheless, a scientific consensus has been reached on what constitutes a MeDi today( Reference Bach-Faig, Berry and Lairon 51 ). At the same time, several indices have been developed to evaluate adherence to a Mediterranean-style diet( Reference Kourlaba and Panagiotakos 52 – Reference Mila-Villarroel, Bach-Faig and Puig 54 ).
Assessment of adherence to a Mediterranean diet
The most commonly used definition to assess adherence to the MeDi is the ‘MeDi score’, first proposed by Trichopoulou et al. ( Reference Trichopoulou, Kouris-Blazos and Wahlqvist 55 ). In its original form, the MeDi score included eight food groups–dietary components (vegetables, fruits, legumes, cereals, meat, dairy products, MUFA:SFA ratio and alcohol), and ‘fish and seafood’ intake was added as a ninth group, based on growing evidence on the beneficial role of long-chain n-3 PUFA in health( Reference Trichopoulou, Costacou and Bamia 56 ). The MeDi score is a sum-score of nine individual binary components (corresponding to the nine aforementioned food groups), and ranges therefore from zero (lower adherence) to nine points (higher adherence). Individual components are calculated as follows: a value of zero or one is assigned to each component, using cut-offs based on sex-specific medians of consumption in the population. For components presumed to be beneficial to health (e.g. vegetables, fruits, legumes, cereals and fish, and MUFA:SFA ratio), individuals whose consumption is below the median are assigned a value of zero v. one for the others. For components presumed to be detrimental to health (e.g. meat and dairy products), individuals whose consumption is below the median are assigned one point v. zero point for the others. For alcohol intake, where moderate consumption is supposed to be beneficial, more heterogeneity in the scoring system has been observed. Initially, a value of one is assigned to men who consume 10–50 g alcohol (from any source) per d, and to women who consume 5–25 g alcohol per d( Reference Trichopoulou, Costacou and Bamia 56 ). In studies based on populations less adherent to the traditional MeDi, these thresholds have been sometimes modified to better identify moderate drinkers( Reference Feart, Samieri and Barberger-Gateau 57 ). Finally, among the sources of alcohol, red wine also provides polyphenols which may be beneficial for various health outcomes; it has been sometimes separated from other sources of alcohol in the scoring system( Reference Tangney, Kwasny and Li 58 ).
An alternate index, the MedDiet score, has been used in relation to cognitive function( Reference Tangney, Kwasny and Li 58 ). In this index, first developed by Panagiotakos et al. ( Reference Panagiotakos, Pitsavos and Stefanadis 59 ), food intake is not translated into a binary component according to the median as in the MedDiet score, but is expressed in number of servings (per month, week or d, according to the food group considered). The MeDi score is a sum-score of eleven individual five points components: non-refined cereals, vegetables, fruits, olive oil, alcohol and full-fat dairy products (servings per d); legumes, fish, poultry and potatoes (servings per week), and red meat and meat products (servings per month). For components positively associated with the MeDi (i.e. non-refined cereals, potatoes, fruits, vegetables, legumes fish and olive oil), a score from zero, for rare or no consumption, up to five for almost daily consumption was assigned. On the other hand, for components inversely associated with the MeDi (i.e. red meat and products, and poultry and full fat dairy products), opposite scores were assigned (i.e. zero for almost daily consumption to five for rare or no consumption of meat and meat products, poultry and full-fat dairy products). For alcohol, in the original MedDiet score form, five points were assigned for a consumption of less than 300 ml alcohol per d, and zero point for no consumption or for consumption over than 700 ml per d. Thus, the range of the MedDiet score was between zero and fifty-five; higher score indicating greater adherence to a Mediterranean-style diet( Reference Panagiotakos, Pitsavos and Stefanadis 59 ). Advantages of this second index are the weighting of the selected food groups, depending on the frequency of consumption (thresholds are chosen based on a priori hypothesis) and regardless of the consumptions of the sample studied.
Altogether, these indices have been considered as efficient tools to assess adherence to the MeDi, whereas a lack of high correlation between both indices has also been reported( Reference Mila-Villarroel, Bach-Faig and Puig 54 ).
Potential biological mechanisms
Foods or nutrients from the MeDi might delay age-related cognitive decline by several underlying biological mechanisms, including vascular, antioxidant and anti-inflammatory pathways( Reference Dauncey 23 , Reference Barberger Gateau, Feart, Samieri and Yaffe 40 , Reference Serra-Majem, Roman and Estruch 60 , Reference Steele, Stuchbury and Munch 61 ). A comprehensive review recently published by Frisardi et al. provides an updated state-of-the-art on this issue( Reference Frisardi, Panza and Seripa 62 ).
First, dementia and cognitive decline have been related to various vascular risk factors( Reference Viswanathan, Rocca and Tzourio 63 ) and the role of nutrition, and especially of the MeDi, on vascular risk factors and CVD is well documented( Reference Sofi, Abbate and Gensini 4 ). A greater adherence to the MeDi was associated with a 10% reduced risk of fatal and non-fatal cardiovascular events( Reference Sofi, Abbate and Gensini 4 ). Dementia and cognitive decline have also been related to the metabolic syndrome, defined as a cluster of cardiovascular risk factors (i.e. having at least three over the five following cardio-metabolic abnormalities: hypertension, high waist circumference, hypertriglyceridemia, low-HDL-cholesterol and hyperglycaemia)( Reference Raffaitin, Feart and Le Goff 64 – Reference Profenno, Porsteinsson and Faraone 66 ). A recent meta-analysis of fifty epidemiological studies and trials involving more than 500 000 participants showed that MeDi adherence was inversely associated with the risk of the metabolic syndrome overall, and with each of its individual component( Reference Kastorini, Milionis and Esposito 67 ). For instance, compared with a low-fat diet, the MeDi improved plasma glucose, systolic blood pressure and cholesterol levels in individuals with high risk of CVD( Reference Estruch, Martinez-Gonzalez and Corella 68 ).
Secondly, oxidative damage have been implicated in the pathogenesis of AD, and the MeDi, rich in antioxidant compounds found in olive oil and wine (sources of polyphenols), fruits and vegetables (sources of vitamins C, E and carotenoids) may help lower oxidative stress in brain ageing. Among individuals with high cardiovascular risk enrolled in the PREDIMED study, a dietary-intervention trial, some components of the MeDi (i.e. total olive oil, walnuts and wine), with antioxidant properties or rich in polyphenols, have been independently associated with better cognitive function( Reference Valls-Pedret, Lamuela-Raventos and Medina-Remon 69 ). Moreover, the MeDi has been inversely associated with markers of oxidative stress( Reference Dai, Jones and Goldberg 70 ) and lipid peroxidation( Reference Gaskins, Rovner and Mumford 71 ). In a recent trial, a MeDi enriched in olive oil was found to lower expression of genes related to oxidative stress and inflammatory processes, and to decrease markers of lipid oxidation and systemic inflammation in plasma( Reference Konstantinidou, Covas and Munoz-Aguayo 72 ).
Inflammation is another key mechanism in the pathogenesis of AD, and a higher adherence to the MeDi has been associated with lower inflammatory markers( Reference Giugliano and Esposito 73 – Reference Gu, Luchsinger and Stern 75 ). A meta-analysis comparing Mediterranean to low-fat diets concluded that individuals assigned to a MeDi had more favourable changes in plasma high-sensitivity C-reactive protein and fasting glucose, in total cholesterol and in blood pressure than those assigned to a low-fat diet( Reference Nordmann, Suter-Zimmermann and Bucher 76 ). The anti-inflammatory properties of n-3 fatty acids may also be involved in the relationship between MeDi, inflammation and cognitive function, since higher MeDi adherence has been associated with higher levels of plasma n-3 fatty acids( Reference Panagiotakos, Kastorini and Pitsavos 77 ), in particular in the Three-City (3C) study( Reference Feart, Torres and Samieri 78 ).
Altogether, these results suggested that the association between MeDi adherence and cognitive functions might be mediated by vascular comorbidity, and underscored that non-vascular biological mechanisms, such as oxidative stress, inflammation and metabolic disorders, should also be considered as possible mediators( Reference Frisardi, Panza and Seripa 62 ).
Adherence to a Mediterranean diet and cognitive functions
The association between adherence to a Mediterranean-type diet and cognitive function or dementia has been only recently explored. To date, this relationship has been investigated in seven different cohorts, and overall, results mostly converge towards a beneficial effect of the MeDi on cognitive function, despite many differences between the populations studied (Table 1). We provide here a summary of existing prospective published data in this field, to the best of our knowledge to date.
Adherence to a Mediterranean diet (MeDi) and cognitive functions: summary of main longitudinal studies
WHICAP, Washington Heights-Inwood Columbia Aging Project; AD, Alzheimer disease; HR, hazard ratio; MCD, mild cognitive disorder; MCI, mild cognitive impairment; MMSE, Mini Mental State Examination; IST, Isaacs Set Test; BVRT, Benton Visual Retention Test; FCSRT, Free and Cued Selective Reminding Test; CDR, clinical dementia rating; CHAP, Chicago Health and Aging Project.
Results from the Washington Heights-Inwood Columbia Aging Project
The first association between greater adherence to a Mediterranean-style diet and lower incidence of AD was reported by Scarmeas et al. ( Reference Scarmeas, Stern and Tang 79 ). Beginning in 1992, 2258 US individuals older than 65 years free from dementia were enrolled in the Washington Heights-Inwood Columbia Aging Project (WHICAP) and followed for 4 years on average (range 0·2–13·9). During this follow-up, 262 incident cases of AD were identified. Adherence to a Mediterranean-style diet was assessed at baseline using the MeDi score (range 0–9), as described earlier. A higher adherence to the MeDi was associated with a reduced risk of AD (hazard ratio (HR) = 0·91, 95% CI 0·83, 0·98 for each additional point of MeDi score, P = 0·015), taking into account major potential confounders. Compared with individuals in the lowest tertile of MeDi score (scores 0–3, indicating a low adherence to the MeDi), those in the middle tertile of score (scores 4 or 5) had a 21% lower risk of AD and those in the highest tertile (scores 6–9) had a 40% lower risk of AD, with a significant trend for a dose–response effect (P for trend = 0·007), in fully adjusted models.
This key result was extended to MCI( Reference Scarmeas, Stern and Mayeux 80 ). In a sub-sample of 1393 individuals of the WHICAP followed for 4·5 years on average, 275 individuals developed MCI. A borderline significant association between greater adherence to the MeDi and lower risk of MCI was observed (HR = 0·72, 95% CI 0·52, 1·00, for one additional point of MeDi score, P = 0·05). Moreover, among 482 individuals with MCI at baseline, 106 developed AD during the follow-up. MCI patients with moderate (scores 4 or 5) or high (scores 6–9) MeDi adherence at baseline had, respectively, 45 and 48% lower risks of AD than those with lower MeDi adherence (scores 0–3; P for trend = 0·02). A potential bias could yet not be dismissed in these analyses in which assessment of nutritional habits was performed among individuals with memory deficits.
A more global healthy lifestyle was then studied in the WHICAP, combining dietary habits and physical exercise in relation to the risk of AD( Reference Scarmeas, Luchsinger and Schupf 81 ). Among 1880 individuals with information for diet and physical activity and non-demented at baseline, 282 individuals were diagnosed with AD during follow-up (5·4 years on average). Independent inverse associations between physical exercise, MeDi adherence and the risk of AD were found (HR for much v. no physical activity = 0·67, 95% CI 0·47, 0·95, P = 0·02; HR for high (scores 6–9) v. low MeDi adherence (scores 0–3) = 0·60, 95% CI 0·42, 0·87, P = 0·007).
Finally, the same authors investigated in the WHICAP the relation between adherence to the MeDi and mortality in AD patients( Reference Scarmeas, Luchsinger and Mayeux 82 ). Among 192 subjects diagnosed with AD at baseline in the WHICAP and followed for 4·4 years on average, eighty-five died. Compared with AD patients in the lowest tertile of MeDi score at baseline (scores 0–3), those in the highest tertile of the MeDi score (scores 6–9) had a significant lower mortality risk (HR = 0·27, 95% CI 0·10, 0·69, P for trend = 0·003), with a longer survival of 3·9 years on average. This result suggests that adherence to the MeDi may affect not only risk for AD but also subsequent disease course.
In order to investigate the hypothesis of a vascular mediation, Scarmeas et al. tested whether the association between MeDi and the risk of AD was attenuated when vascular risk factors (i.e. history of stroke, diabetes, hypertension, heart disease and lipids levels) were added in their statistical models( Reference Scarmeas, Stern and Mayeux 86 ). Surprisingly, the magnitude of the association between the MeDi score and AD risk was virtually unchanged when adding vascular factors in the models, suggesting that vascular factors/disease may not act as strong mediators in this relationship. Nevertheless, more recently, the same authors reported among 707 WHICAP participants (aged 80 years on average) that higher MeDi adherence was associated with lower cerebrovascular disease burden in the brain, as assessed by high-resolution structural MRI( Reference Scarmeas, Luchsinger and Stern 87 ). Interestingly, MeDi adherence was associated with less MRI infarcts, markers of large vessel disease, but not with less white matter hyperintensities volumes, markers of small vessel disease. This result thus suggested that large vessel disease could be in the biological pathway between MeDi and AD( Reference Scarmeas, Luchsinger and Stern 87 ). To discuss the inconsistency between this finding and their previous mediation study( Reference Scarmeas, Stern and Mayeux 86 ), the authors argued that self-reported history of stroke might be less accurate than markers of cerebrovascular disease provided by MRI( Reference Scarmeas, Luchsinger and Stern 87 ). Finally, the association between MeDi adherence and AD was also not attributable to inflammatory or metabolic markers since the introduction of high-sensitive C-reactive protein, fasting insulin or adiponectin as adjustment variables, did not modify the inverse association between MeDi adherence and incident AD observed in the WHICAP( Reference Gu, Luchsinger and Stern 75 ).
Altogether, these results suggest that vascular and inflammatory pathways explored in the WHICAP may be, at best, only partial mediators in the MeDi–AD association, and that other pathways may also be involved. Most of the aforementioned results from the WHICAP performed at Columbia University have been recently reviewed( Reference Devanand, Lee and Luchsinger 19 ).
Results from the Three-City study
To our knowledge, there is only one prospective study to date which examined the association of a MeDi to cognitive change and risk of dementia in Europe( Reference Feart, Samieri and Rondeau 83 ). Using data from the 3C Study, a French prospective cohort of older individuals (aged 65 years or over at baseline)( 88 ), the authors investigated the relationship between adherence to the MeDi at baseline, using the MeDi score (range 0–9), and change in cognitive performances assessed every 2–3 years using four neuropsychological tests. The 3C study was the first to analyse the relationship between MeDi adherence and cognitive decline assessed longitudinally. A total of 1410 individuals free from dementia at baseline were followed-up for 5 years. Global cognitive function was assessed using the Mini Mental State Examination, and the Isaacs Set Test, the Benton Visual Retention Test and the Free and Cued Selective Reminding Test assessed semantic verbal fluency, visual memory and verbal episodic memory, respectively. During the follow-up, ninety-nine incident cases of dementia (including sixty-six AD) were identified. After multivariable adjustment, higher adherence to the MeDi was significantly associated with better trajectories of global cognition and episodic memory (Mini Mental State Examination and Free and Cued Selective Reminding Test, respectively), especially in individuals who remained free from dementia over 5 years. However, no association was found between greater MeDi adherence and the risk of dementia (HR = 1·06, 95% CI 0·92, 1·21, P = 0·43) or AD (HR = 1·00, 95% CI 0·85, 1·19, P = 0·96). These results may suggest that the MeDi prevents brain ageing early in the prodromal phase of dementia( Reference Amieva, Le Goff and Millet 89 ), rather than in the very last few years preceding dementia diagnosis. However, the null association found between MeDi adherence and dementia risk in this study could also be explained by a lack of power due to a relatively short length of follow-up and a limited number of incident cases of dementia. As already observed in the WHICAP, additional adjustments for cardiovascular risk factors and reported history of stroke did not attenuate the negative relationship between MeDi adherence and cognitive decline in the 3C study( Reference Scarmeas, Luchsinger and Mayeux 82 ).
Combined results from the WHICAP and the 3C study were summarised in a meta-analysis which concluded that a two-point increase of the MeDi score was associated with a 13% reduced risk of neurodegenerative disease( Reference Sofi, Abbate and Gensini 4 ).
Investigators from the 3C study further examined the association between MeDi adherence and the onset of disability, which is a necessary condition for the diagnosis of dementia( 90 ). Among 1410 individuals from the 3C study, basic and instrumental activities of daily living were assessed by Lawton–Brody and Katz scales( Reference Feart, Peres and Samieri 91 ). MeDi adherence, assessed by the MeDi score, was inversely associated with the risk of incident disability in basic and instrumental activities of daily living in women (n 185 incident cases), but not in men (n 90 incident cases). Women with the highest MeDi adherence (scores 6–9) had a 50% reduction of incident disability in basic and instrumental activities of daily living than women with the lowest MeDi adherence (scores 0–3; P = 0·003). These findings suggested that adherence to a Mediterranean-style diet could contribute to slow down of the disablement process, at least in women.
Results from The Mayo Clinic Study of Aging
The relationship between adherence to the MeDi and the risk of MCI and dementia has also been investigated among 1141 individuals enrolled in The Mayo Clinic Study of Aging in the USA( Reference Roberts, Geda and Cerhan 84 ). After 2 years of follow-up, a non-significant 25% reduced risk of MCI (n 93 incident cases) or dementia (n 23 incident cases) was observed in individuals with greater MeDi adherence v. those with lower adherence (P = 0·24). However, with only 116 events recorded during a short follow-up, this study may have been underpowered to detect associations.
Results from the Chicago Health and Aging Project
In the Chicago Health and Aging Project, adherence to the MeDi was assessed with a novel tool, the MedDiet score (range 0–55)( Reference Tangney, Kwasny and Li 58 ) (see description aforementioned), which takes into account daily, weekly or monthly intakes of eleven food groups. This index was computed according to the traditional MeDi and the definition of the thresholds used was therefore not specific of the consumption of the population studied( Reference Panagiotakos, Pitsavos and Stefanadis 59 ). In this study, Tangney et al. investigated whether MeDi adherence was associated with cognitive change in older adults (n 3790) aged 65+ years at baseline. Four cognitive tests were administered to the participants, up to five times, for 7·6 years on average. The East Boston Memory test (immediate and delayed recalls), the Mini Mental State Examination and the Symbol Digit Modalities test were used to define a global measure of cognition. Interestingly, after multivariable adjustment, the MedDiet score was associated with slower rates of cognitive decline; the results were mostly unchanged in sensitivity analyses excluding persons with heart disease or stroke. Regarding the MedDiet score expressed in tertiles, participants in the highest tertile of MeDi score (range 30–45) had a significant slower rate of cognitive change (P = 0·0002). This result was also observed when only wine was considered in the computation of the MeDi score, instead of all alcoholic beverages. By contrast, the Healthy Eating Index-2005, developed to measure the dietary quality of an individual compared with recommendations of the 2005 Dietary guidelines( Reference Kennedy, Ohls and Carlson 92 ), was not associated with rate of cognitive decline in this study.
Results from the PATH Through Life study
An Australian study investigated whether MeDi adherence was associated with cognitive change among 1528 individuals aged 60–64 years participating in the PATH Through Life study( Reference Cherbuin and Anstey 85 ). These individuals were followed-up for 4 years and a dietary survey allowed us to compute the MeDi score (range 0–9) at baseline. Clinical assessments allowed us to identify sixty-six individuals (ten MCI, nineteen with impairment on the Clinical Dementia Rating, thirty-seven with any mild cognitive disorder) who transitioned from a normal cognitive stage at baseline to a cognitive impairment at the end of follow-up. In this study, a greater adherence to a MeDi was not associated with cognitive impairment, each disorder being separately studied. Here again, the small number incident cases due to a short length of follow-up could in part explain the lack of significant association. Although the corresponding data were not presented in this paper, we can assume that the pooled analyses, considering all individuals with incident cognitive impairment, whatever the sub-type, added no relevant information.
Results from the Women's Antioxidant Cardiovascular Study
A recent report examined the association between MeDi adherence and cognitive decline among women enrolled in the Women's Antioxidant Cardiovascular Study, a trial of secondary prevention of CVD( Reference Vercambre, Grodstein and Berr 93 ). These women (n 2504, aged 65+ years at baseline) had a history of CVD or risk factors and were therefore at higher risk of cognitive decline. The initial cognitive assessment, performed on average 3·5 years after the dietary survey, consisted of five cognitive tests administered by telephone. Repeated cognitive evaluation was performed every 2 years for 5 years on average. Adherence to the MeDi at baseline (i.e. 3·5 years before initial cognitive interview) was evaluated with the MeDi score (range 0–9) and the MedDiet score (range 0–55). In spite of a large sample and a prospective design with a similar length of follow-up than previous studies, no association was observed between adherence to the MeDi and 5-year cognitive decline. These results suggest that the prevention of cognitive decline might be more challenging in individuals with prevalent vascular disease or risk factors, and overall, strengthen the hypothesis that the MeDi may exert beneficial properties at early disease stages.
Results from the European Prospective Investigation into Cancer and Nutrition Greek cohort
The association of MeDi adherence to cognitive function was assessed in the Greek sample of the European Prospective Investigation into Cancer and Nutrition cohort( Reference Psaltopoulou, Kyrozis and Stathopoulos 94 ). Among 732 individuals, aged 60+ years at baseline, adherence to a MeDi was evaluated by the MeDi score (range 0–9). Six to 13 years after the dietary survey, the Mini Mental State Examination was administered to assess global cognitive function; since there were no cognitive evaluations at the time of dietary assessment, this study cannot be considered as a truly prospective study. In spite of the Mediterranean origin of the population, likely highly adherent to the traditional MeDi, only a weak non-significant association was observed between MeDi adherence and global cognition. However, the lack of repeated assessment of cognition to evaluate cognitive decline limited the scope of these results.
Possible reasons for some inconsistent results across studies
There is a strong biological rationale for a protective role of the MeDi in brain ageing. Indeed, this dietary pattern combines antioxidants, B vitamins, n-3 fatty acids and other compounds that have been inversely related to cognitive decline and to the risk of dementia( Reference Feart, Alles and Merle 95 ). It also likely captures additive or synergistic effects of several nutrients consumed together( Reference Jacobs, Gross and Tapsell 41 ). Yet, some discrepancies remain between studies on MeDi adherence and cognitive health worldwide.
Assessment of Mediterranean diet adherence
One of the reasons for these inconsistent results may rely on the method used to compute the MeDi score, which limits the generalisation of the results and prevents definite conclusions.
A major limitation of the original MeDi score proposed by Trichopoulou et al. ( Reference Trichopoulou, Kouris-Blazos and Wahlqvist 55 ) is the use of thresholds based on medians of intake of each MeDi component, which are, per se, population-specific. Therefore, a MeDi score is, by definition, population-specific and cannot be compared with a MeDi score computed in a different sample. This may have led to misclassifications, since low consumers from one cohort could be considered as high consumers in another cohort for a particular food group and vice versa (Fig. 1). Therefore, the use of sex-specific cut-off points to develop the MeDi score does not measure adherence to a universal traditional Mediterranean dietary pattern, but rather to a specific pattern( Reference Bach, Serra-Majem and Carrasco 96 ). The MedDiet score, proposed by Panagiotakos et al., aimed at addressing this limitation, since the frequency of consumption of each food group part of the index was used as cut-off with respect to the traditional MeDi( Reference Panagiotakos, Pitsavos and Stefanadis 59 ) and whatever the population studied.
Computation of the Mediterranean diet score( Reference Trichopoulou, Costacou and Bamia 56 ) among two countries with specific median of consumption of food groups.
Another limitation common to both indices is that a same score could be a reflection of several dietary patterns. For instance, a MeDi score of 4 could be assigned to an individual with a high intake of vegetables, fruits, fish and a high MUFA:SFA ratio (but a low consumption of legumes, cereals, a high consumption of meat and dairy products and a non-moderate alcohol intake). However, a MeDi score of 4 could also be assigned to an individual with a high consumption of cereals and fish and a low consumption of meat and dairy products (but a low consumption of fruits, vegetables, legumes, a low MUFA:SFA ratio and non-moderate alcohol intake). Obviously, these dietary patterns, simply summarised by a value of 4 for the MeDi score, are significantly different. Hence, several dietary patterns coexist in a single category of MeDi adherents (identified as low, moderate or high).
Finally, the scoring system used in both indices does not consider other food groups that could be relevant to the traditional MeDi or a more modern definition; for example, the use of supplements is not considered in these indices. Nevertheless, the indices already available showed satisfactory performances in assessing adherence to the MeDi, while the need to reach a consensus on the components to be included is still required( Reference Mila-Villarroel, Bach-Faig and Puig 54 ).
Design of studies on diet and cognitive health
Studies on the relationship between MeDi adherence and cognitive health should be interpreted with caution. In a slowly evolving dementia syndrome, successive emergence of cognitive deficits appear more than 10 years before the diagnosis of dementia( Reference Amieva, Le Goff and Millet 89 ). Therefore, imposing a reasonable delay (i.e. several years) between dietary assessment and cognitive evaluation is of utmost importance to avoid reverse causation, that is, cognitive impairment lead to a change in dietary habits, and not the reverse. In the 3C study, the benefit of a higher MeDi adherence on verbal episodic memory was observed at least 5 years before the clinical diagnosis of dementia, suggesting that after a window of opportunity, in the very last few years preceding dementia, neurodegenerative processes may be too advanced to be reversed by diet( Reference Feart, Samieri and Rondeau 83 ).
Moreover, in most studies, dietary habits are assessed at a single occasion (i.e. at the beginning of cognitive evaluation, or before) and are therefore assumed a good marker of long-term habits. In the WHICAP, adherence to a MeDi was remarkably stable over 8 years( Reference Scarmeas, Stern and Tang 79 , Reference Scarmeas, Stern and Mayeux 86 ); similar results were observed in the Chicago Health and Aging Project ( Reference Tangney, Kwasny and Li 58 ). However, there is still a possibility that diet assessed in late-life is a poor marker of midlife dietary habits, and midlife is likely the most relevant period to study risk factors for cognitive decline which evolve for years, if not decades.
It is also possible that the MeDi may not be relevant to cognitive health overall, but that the association between the MeDi and cognitive decline and dementia is actually driven by a limited number of specific foods. This issue was addressed in the WHICAP and in the PATH Through Life study. Among WHICAP participants, a mild-to-moderate alcohol consumption and higher vegetable consumption were independently associated with a reduced risk of AD after adjustment for all other individual components used to calculate the MeDi score( Reference Scarmeas, Stern and Tang 79 ). Despite a lack of association between MeDi adherence and incident cognitive impairment in the PATH Through Life study, significant associations were found between higher MUFA, dairies and alcohol consumptions and increased risk of MCI( Reference Cherbuin and Anstey 85 ); unexpectedly, a higher fish consumption was associated with a higher risk of cognitive disorder. However, these analyses were not controlled for all the food groups' part of the MeDi. Altogether, these results underscored the limits of the single food approach and the difficulty to understand the interactions of all components of the food matrix.
Finally, MeDi adherence is also part of a healthier lifestyle in general, especially in countries far from the Mediterranean basin, which can favourably affect cognition. By insufficiently controlling for lifestyle confounders, it is possible that some residual confounding persists in studies on MeDi and cognition. Besides, controlling for late-life risk factors cannot be sufficient to stand for lifelong exposure, which remains an issue( Reference Knopman 97 ).
Conclusion
Overall, although a growing body of scientific evidence suggests that the MeDi may promote healthy brain ageing, there are still some controversies among epidemiological studies. The replication of analyses presented here is therefore needed to encompass some limitations and to allow their generalisation( Reference Barnes 98 ). Specifically, large-scale studies in various populations with common methodology are required before considering the MeDi as an optimal dietary strategy to prevent cognitive decline or dementia.
Dietary habits reflect individual food preferences which are closely related not only to culture, education and socio-economic background but also to health status. The whole being more than the sum of its part, the promotion of the healthy Mediterranean-type dietary pattern, rather than individual food or nutrients, should be extended, in particular because MeDi would be the best known model to fulfil nutrient requirements( Reference Maillot, Issa and Vieux 99 , Reference Martinez-Gonzalez and Gea 100 ). This seems to be of particular importance since traditional food choices are changing with the progressive globalisation of food supply in young generations( Reference Sofi 49 ).
Acknowledgements
C. F. received fees for conferences from Danone. P. B.-G. served on a scientific advisory board for Caisse Nationale pour la Solidarite et l'Autonomie; had received funding for travel and speaker honoraria from Lesieur, Bausch & Lomb, Aprifel, Danone Institute, Canadian Association of Gerontology and the Jean Mayer Human Nutrition Research Center on Aging, Tufts University; served on the editorial boards of Disability and Rehabilitation; had received consultancy fees from Vifor Pharma; and received research support from Lesieur, Danone, Agence Nationale de la Recherche, Conseil Régional d'Aquitaine, Institut Carnot LISA and Groupe Lipides et Nutrition. C. S. and B. A. declare no conflict of interest. C. F. conducted the research; the paper was written by C. F., C. S. and P. B.-G., C. F. having responsibility for final content; P. B.-G., C. S. and B. A. provided significant advice and all authors read the draft critically.
