Epidemiological evidence

Epidemiological studies in this area that have focused on fruit and vegetables as a combined category have shown some positive findings. In the Whitehall II study, consuming <2 portions of fruit and vegetables daily was associated with an increased risk of poor executive function and, to a lesser extent, poorer episodic memory^{(Reference Sabia, Nabi and Kivimaki34)}. This association was apparent for fruit and vegetable intake at midlife but was strongest when measured cross-sectionally at 15-year follow-up. The supplementation with antioxidant vitamins and minerals 2 study also showed that higher fruit and vegetable intake was associated with better verbal memory at 13-year follow-up, with specific effects for fruit and for vitamin C-rich fruit and vegetables^{(Reference Péneau, Galan and Jeandel35)}. However, in this study the association with executive functioning was negative. One possible explanation for the discrepancy in findings relating to executive function is the classification of verbal fluency as memory tasks in the supplementation with antioxidant vitamins and minerals 2 study^{(Reference Péneau, Galan and Jeandel35)}, whereas this was categorised as executive function in the Whitehall II study^{(Reference Babateen, Shannon and O'Brien36)}. It is also important to note that in both studies no measure of cognition was taken at baseline, so assessment of cognitive decline was not possible. Support for a relationship between fruit and vegetable intake in late midlife and cognition in later life also comes from the health professionals follow-up study, which showed that higher consumption of both fruit and vegetables was associated with lower odds of poor subjective cognitive function at 18–22-year follow-up^{(Reference Yuan, Fondell and Bhushan37)}. Whilst no objective measure was included in this study, the experience of cognitive function subjectively may be particularly important in terms of quality of life. Performance on cognitive tests assessing memory and executive function improved with an increasing intake of fruit and vegetables in the Hordaland health study, with a marked dose-dependent relationship up to about 500 g/d. When analysed separately, fruit consumption was positively associated with performance on three out of five cognitive tests, whereas vegetable consumption was positively associated with performance on four tests^{(Reference Nurk, Refsum and Drevon38)}. Interestingly, data from the Boston Puerto Rican health study indicated that greater variety of fruit and vegetable intake but not quantity was associated with better cognition, and this association was strongest for vegetables^{(Reference Ye, Bhupathiraju and Tucker39)}.

Studies that have examined the role of vegetables separately from fruit have shown higher intake to be associated with smaller declines in processing speed and global cognitive function. This is despite vegetable intake being associated with lower information processing speed and worse cognitive flexibility at baseline, which may be driven by negative effects of allium (garlic, onion and leek) consumption^{(Reference Nooyens, Bueno-de-Mesquita and van Boxtel40)}. Total vegetable intake was also inversely associated with cognitive decline in the nurses' health study, despite no such association for fruit and vegetables combined. The highest intake of vegetable was equivalent to 1⋅5 years less decline at the 2-year follow-up^{(Reference Kang, Ascherio and Grodstein41)}. Similarly, those in the highest two quintiles of vegetable intake in the Chicago Health and Aging Project showed a 40 % reduction in decline on a global cognition measure at 6-year follow-up when compared to the lowest quartile intake^{(Reference Morris, Evans and Tangney42)}. Moreover, a specific association between cognitive decline and green leafy vegetables was observed^{(Reference Morris, Evans and Tangney42,Reference Morris, Wang and Barnes43)} . This is supported by the nurses' health study showing the greatest benefit for green leafy vegetables, but also a specific benefit for cruciferous vegetable, particularly for episodic memory^{(Reference Kang, Ascherio and Grodstein41)}. In the Doetinchem cohort study, higher intake of cabbage was associated with better memory and better global cognitive function at baseline, and with smaller decline in information processing speed at 5-year follow-up, whilst root vegetables were associated with smaller declines in cognitive flexibility and global cognition^{(Reference Nooyens, Bueno-de-Mesquita and van Boxtel40)}. Support for the roles of root and cruciferous vegetables also comes from the Hordaland health study showing strongest associations for carrots, and cruciferous vegetables, but also for citrus fruits^{(Reference Nurk, Refsum and Drevon38)}.

Effects of specific fruits have been explored with the nurses' health study showing that greater intake of blueberries and strawberries was associated with better episodic memory scores and slower rates of cognitive decline for global cognition, equivalent to up to 2⋅5 years of ageing^{(Reference Devore, Kang and Breteler44)}. Consumers of avocado were also shown to have better episodic memory and global cognition than non-consumers^{(Reference Cheng, Ford and Taylor45)}. Whilst not always categorised as fruit, a specific benefit has also been shown for tree nuts. The Doetinchem cohort study observed that higher nut intake was related to better global cognition, flexibility, memory and processing speed^{(Reference Nooyens, Bueno-de-Mesquita and van Boxtel40)}. Long-term (6-year) nut intake was linked to better global cognition in women aged ≥70 years but not with cognitive decline^{(Reference O'Brien, Okereke and Devore46)} and an association to 5-year cognitive decline in the Doetinchem cohort study was lost following adjustment for cardiovascular risk factors^{(Reference Nooyens, Bueno-de-Mesquita and van Boxtel40)}. This indicates a role for cardiovascular effects, which is supported by cognitive improvements shown following Mediterranean diet plus mixed tree nuts in those at high cardiovascular risk^{(Reference Valls-Pedret, Sala-Vila and Serra-Mir31)}.

Epidemiological data suggest a relationship between fruit and vegetable intake and cognition. The evidence appears to be stronger for an association between vegetables and cognition than for fruit, which may relate to the common inclusion of fruit juice in the fruit category, despite differences in their fibre content. Since fruit and fruit juice modulate different aspects of immune function, with juice potentially linked to proinflammatory pathway activation^{(Reference Nicodemus-Johnson and Sinnott47)}, the inclusion of fruit juice in the fruit category may be concealing effects of whole fruit. Specific benefits shown for berries, citrus fruits, avocado and nuts suggest fruit is worthy of further investigation in relation to cognition; nuts in particular represent a nutrient-dense food with less sugar than other fruits. In terms of vegetables, a specific benefit may be seen for cruciferous or green leafy vegetables. However, this is not consistently shown with one study showing benefits for both^{(Reference Kang, Ascherio and Grodstein41)}, whilst others show benefits for one or the other; a specific benefit for cabbage^{(Reference Nooyens, Bueno-de-Mesquita and van Boxtel40)} may indicate that green and leafy vegetables such as kale and savoy cabbage may confer the most favourable effects, potentially due to the high levels of glucosinolates, carotenoids, vitamins, minerals and flavonoids. In an analysis focused on flavonoids, the Lothian birth cohort 1936 study failed to show effects of fruit and/or vegetables^{(Reference Butchart, Kyle and McNeill48)}. This potentially highlights the importance of covariates in epidemiological research since relationships between vegetable intake and cognition were lost after controlling for childhood IQ, whilst those for total fruit intake were lost with the additional inclusion of smoking, socio-economic status, education and the apoE 14 allele.

Randomised controlled trials

RCT in this area (see Table 2) have tended to focus on fruit, specifically flavonoid-rich. The effects of orange juice consumption were assessed in a crossover trial of 60–81-year-olds. Cognition was assessed via five episodic memory tasks and six executive function tasks, pre- and post- a 500 ml serving of 305 mg flavanones or placebo for 8 weeks. The findings revealed improvement in global cognitive function, with no effects on episodic memory or executive function when assessed separately^{(Reference Kean, Lamport and Dodd49)}. Several studies have also assessed the effects of flavonoid-rich berries. Effects of 12 weeks' tart cherry juice were assessed in 65–80-year-olds on measures of episodic memory, sustained attention, information processing, executive function and subjective memory^{(Reference Chai, Jerusik and Davis50)}. Compared to placebo, 480 ml tart cherry juice daily led to fewer errors on an episodic memory task, and shorter movement reaction time, as well as increasing contentment with memory. Two studies have assessed effects of blueberry in older adults^{(Reference Miller, Hamilton and Joseph51,Reference Whyte, Cheng and Fromentin52)} . Twenty-four g/d freeze-dried blueberry extract (equivalent to one cup of fresh blueberries) led to improved episodic memory and executive function in 60–75-year-olds, with no effects on information processing or attention tasks^{(Reference Miller, Hamilton and Joseph51)}. Different doses and forms of blueberry have also been assessed in 65–80-year-olds^{(Reference Whyte, Cheng and Fromentin52)}. Effects of 500 and 1000 mg wild blueberry powder and 100 mg of a purified extract were assessed on measures of episodic memory, executive function and selective attention. The purified extract led to improved episodic memory and visual memory span at 12 but not 24 weeks, compared to placebo^{(Reference Whyte, Cheng and Fromentin52)}. Interestingly, the purified extract contained negligible fat, fibre, vitamin C and sugars, which were present in the other blueberry interventions. This indicates that the observed effects are attributable to the flavonoid content alone rather than a synergy with these other components. The lack of effects at 24 weeks is difficult to explain, particularly as systolic blood pressure was reduced at 12 and 24 weeks, but a tentative explanation is that this is reflective of a practice effect whereby improved strategy in performing the tasks has reduced task sensitivity. Twelve weeks' consumption of 4⋅9 g/d freeze-dried cranberry extract (equivalent to one cup of fresh cranberries) was also shown to improve episodic memory in 50–80-year-olds^{(Reference Flanagan, Cameron and Sobhan53)}. Reduced LDL-cholesterol was also shown with some evidence for sex differences in response, as well as increased perfusion in specific regions of the right brain. This is contrary to an earlier study showing no effects of cranberry juice on a range of cognitive measures in healthy older adults^{(Reference Crews, Harrison and Griffin54)}. These null findings may be due to the form or dose of cranberry administered, with a 909 ml beverage containing 27 % cranberry juice being delivered, or they could indicate that a 6-week intervention was not sufficient to elicit an effect on cognitive performance in this population.

Table 2.

Randomised controlled trials of fruit and vegetable juices and extracts

The effects of cocoa should also be considered here. Whilst not typically thought of as a fruit, cocoa represents the seed of the fruit of the cacao tree and has been studied for potential benefits due to its flavan-3-ol content. In one such study, high flavan-3-ol (993 mg), intermediate flavan-3-ol (520 mg) or low flavan-3-ol control (48 mg) were given for 8 weeks. Improvements to information processing and a measure of executive function were shown following both flavan-3-ol doses and the highest dose also improved verbal fluency^{(Reference Mastroiacovo, Kwik-Uribe and Grassi55)}. Importantly, no effects were observed on the mini-mental state examination, highlighting the lack of sensitivity of this measure in short intervention trials of healthy adults. This study also attempted to ascertain the determinants of cognitive change, observing that insulin resistance explained 17 % of cognitive change. Systolic blood pressure and lipid peroxidation were not predictors despite blood pressure and several metabolic markers being improved by the intervention. Another interesting study of cocoa showed that high flavan-3-ol (900 mg) improved novel object recognition reaction time when compared to control. Moreover, the authors reported that the improvement observed following intake of high flavan-3-ol cocoa for 12 weeks was equivalent to three decades of ageing, shifting the slope in expected age-related decline. Imaging data also showed an increase in cerebral blood volume in the dentate gyrus region of the hippocampus and this increase was inversely correlated with reaction time. This indicates a potential mechanism of action for improvements linked to increased blood flow, which may be specific to the hippocampus given the localised effect and the role of the hippocampus in the task assessed^{(Reference Brickman, Khan and Provenzano56)}. Significant improvements to global cognition were also observed in healthy elderly following supplementation with 494 mg of cocoa flavan-3-ol for 28 d. Interestingly, increases in serum brain-derived neurotrophic factor levels were also observed^{(Reference Neshatdoust, Saunders and Castle57)}.

Studies of nuts have produced mixed findings. In the dual-centre walnuts and healthy ageing study there were no effects of a 2-year walnut intervention^{(Reference Sala-Vila, Valls-Pedret and Rajaram58)}. However, a site-specific improvement to global cognition and functional MRI differences indicative of improved neural efficiency in the walnut group were observed following 30–60 g/d walnuts over 2 years. A study assessing effects of 12-week almond consumption failed to find effects on cognition when compared to an isoenergetic carbohydrate-rich snack in overweight and obese older adults, despite improvements to aspects of cardiometabolic health^{(Reference Coates, Morgillo and Yandell59)}. However, 12-week peanut consumption was observed to enhance memory, verbal fluency and processing speed, as well as increasing cerebrovascular reactivity and small artery elasticity in healthy overweight adults^{(Reference Barbour, Howe and Buckley60)}. Changes in cerebrovascular reactivity in the left middle cerebral artery were also positively correlated with changes in delayed memory and recognition. Although peanuts are legumes and therefore classified as vegetables, they are usually considered in the nut category due to their composition. These nut studies illustrate the issue of implementing a control in studies of whole foods. In two of the studies the control arm consisted of habitual diet minus nuts^{(Reference Sala-Vila, Valls-Pedret and Rajaram58,Reference Barbour, Howe and Buckley60)} , whereas in the other study nuts were compared to an isoenergetic snack^{(Reference Coates, Morgillo and Yandell59)}. The first option presents issues in that no intervention is present in the control arm, whereas the snack comparator makes interpreting findings difficult due to potential effects of the composition of the control employed.

RCT of vegetables are less prevalent than fruit, possibly due to the logistics of developing a control for the intervention. Fruit is typically studied in juice form, but vegetable juice is less commonly consumed and is potentially less likely to be accepted by participants where daily intake is required. For instance, there are a number of studies of beetroot juice, but only one of these has included repeated consumption in healthy older adults^{(Reference Babateen, Shannon and O'Brien36)}. A 13-week trial in overweight and obese older adults assessed effects of high (800 mg/d), moderate (400 mg/d) and low (200 mg/d) nitrate administered in beetroot juice and compared effects to a nitrate-depleted beetroot juice. Despite a wide array of cognitive tasks, there was no evidence of any effects of beetroot juice on cognition, and no effects on cerebral blood volume as assessed with near IR spectroscopy. This study was designed as a feasibility study and not intended to provide a definitive investigation of the effects of nitrate on cognition and cerebral blood flow, but the null findings are unexpected based upon the available evidence from short-term studies.

Evidence for cognitive benefits of fruit and vegetables from RCT is weaker than epidemiological data. There are generally only one or two RCT per food type, with the strongest evidence for cocoa where three studies exist. Although the available data are limited, the findings are generally positive despite potential issues with short duration of intervention and limited knowledge of optimal dosage. Not all studies show positive effects, the dual-centre walnuts and healthy ageing study^{(Reference Sala-Vila, Valls-Pedret and Rajaram58)} failed to detect effects of a 2-year walnut intervention when analysed in the population as whole. However, subgroup analysis revealed a site-specific improvement. Compared with other sites, participants at the site where improvement was observed, reported lower education and lower intake of the plant n-3, α-linolenic acid^{(Reference Sala-Vila, Valls-Pedret and Rajaram58)}. These findings potentially suggest that those at risk, either due to poorer nutrition or poorer education, may be more susceptible to cognitive improvements following a fruit or vegetable intervention. A number of studies^{(Reference Flanagan, Cameron and Sobhan53,Reference Mastroiacovo, Kwik-Uribe and Grassi55,Reference Brickman, Khan and Provenzano56,Reference Barbour, Howe and Buckley60)} indicate increased cerebral blood volume following fruit/legume intervention, as well as improved cardiometabolic markers, and in some cases a correlation between changes in cognition and cerebrovascular/metabolic markers. These findings may suggest that cognitive improvement is more likely in those who are at risk for cardiometabolic dysfunction.

However, RCT assessing effects of diet on cognition often impose strict exclusion criteria such as: no clinically significant coexisting medical conditions, including CVD, cerebrovascular events, neurological disorders, inflammatory diseases, metabolic disorders, gastrointestinal abnormalities, mental illness including depression; no prescription medication; no supplements; no current smokers; no individuals with BMI ≥30 kg/m². Whilst these criteria are sensible in terms of study design and minimising ‘noise’ from uncontrolled variables, they present a problem in terms of representativeness. For instance, 63 % of adults in England aged 65–74 reported a longstanding health condition in 2018^{(Reference Conolly and Craig61)}. In adults aged ≥65, at least 80 % were shown to have taken a prescribed medication in the past week, and more than one in ten took at least eight different prescribed medications per week^{(Reference Moody, Mindell and Faulding62)}. Whilst 75 % of 65–74-year-olds were reported to be overweight or obese^{(Reference Conolly and Craig61)}. In addition to problems relating to generalisability of findings, these studies create somewhat of a paradox in that they are exploring the importance of nutrition to health yet recruiting unusually healthy people whilst assuming suboptimal nutrition that can be enhanced by intervention.

Nutrition status and self-selection

Inclusion of nutrition status measures, such as dietary intake measures and nutritional biomarkers, in RCT of nutrition interventions on cognition is increasing, but still rare. Failure to measure nutritional status in cognitive studies presents obvious problems in not being able to ascertain if any null findings are due to already optimal status in study participants^{(Reference Morris and Tangney63)}. Indeed, post-hoc analyses indicate that beneficial effects of vitamins A, C and certain B-vitamins on cognitive decline may only be shown in those with low baseline intake (for a review see^{(Reference Morris64)}). However, due to self-selection bias, those who volunteer for RCT of nutrition are likely to be health conscious and potentially less likely to display suboptimal nutrition. This is nicely illustrated by the memory and attention supplementation trial where a validated (against blood markers) questionnaire was used to assess diet quality prior to enrolment, aiming for 50 % ‘optimal’ diet (a high intake of fruit, vegetables, legumes, olive oil and nuts, and a lower intake of processed foods) and 50 % suboptimal. Out of 501 volunteers, 461 (74⋅4 %) met the criterion for an ‘optimal’ diet indicative of elevated nutrient status^{(Reference Young, Gauci and Scholey65)}. When considered against the statistic that only 5⋅4 % of Australian adults meet the current recommendations for both fruit and vegetable intake, this again highlights the lack of representativeness of cognition-nutrition research volunteers. Since multivitamin use is more common in females and those with healthier lifestyle habits, higher educational attainment, higher socioeconomic status and of lower BMI^{(Reference Comerford66)}, it is potentially more likely that those who participate in nutrition research will share similar characteristics. This is supported by evidence from the Monongahela Valley Independent Elders Survey, where those who volunteered to participate were more likely to be women, and had more formal years of education, higher cognitive test scores, higher instrumental activities of daily living ability and lower mortality rates than randomly selected participants^{(Reference Ganguli, Lytle and Reynolds67)}. This indicates a further issue with cognitive reserve since those with more years of formal education and who engage in more physical activity are at reduced risk of cognitive decline^{(Reference Baumgart, Snyder and Carrillo68)}. Higher cognitive test scores in those who volunteer may also indicate the presence of ceiling effects in study participants, again increasing the likelihood of null findings.