Alzheimer's disease

AD is a common condition in older age. The overall prevalence among individuals aged 65 years and older has been estimated to be 10–13%⁽ Reference Evans, Funkenstein and Albert ^{2 ,} Reference Hebert, Scherr and Bienias ^{5 )} and the rate increases exponentially with older age: from 3% among 65–74-year-olds to 19% among 75–84-year-olds and nearly half (47%) of those aged 85 years and older⁽ Reference Evans, Funkenstein and Albert ^{2 )}. Because the oldest age groups are also the fastest growing in the USA, the number of individuals with AD is expected to almost triple from six million in 2010 to sixteen million by the year 2050⁽ Reference Hebert, Beckett and Scherr ^{6 )}. Thus, by mid-century approximately 25% of the USA population will be aged 65 years and older, and more than one-third will likely develop dementia⁽ Reference Hebert, Beckett and Scherr ^{6 )}.

The clinical diagnosis of AD requires impairment in memory and at least one other cognitive domain⁽ Reference McKhann, Drachman and Folstein ^{7 )}. The disease is neuropathologically characterised by extracellular deposition of amyloid β into senile plaques, intra-neuronal formation of neurofibrillary tangles composed of hyperphosphorylated τ protein, and loss of neurons and synapses in the neocortex, hippocampus and other subcortical areas of the brain^{( 8 )}. The primary risk factors for late-onset AD include older age, the apoE-ε4 genotype, head injury, family history, low education and low participation in cognitively stimulating activities⁽ Reference Reitz, Brayne and Mayeux ^{9 )}. CVD conditions and risk factors are the primary focus of the current research in AD⁽ Reference Reitz, Brayne and Mayeux ^{9 ,} Reference Davidglus, Bell and Berrettini ^{10 )}. Among these factors are physical activity⁽ Reference Larson, Wang and Bowen ^{11 )}, diet⁽ Reference Luchsinger, Noble and Scarmeas ^{12 )}, hypertension⁽ Reference Skoog, Lernfelt and Landahl ^{13 ,} Reference Kivipelto, Helkala and Laakso ^{14 )}, obesity⁽ Reference Whitmer, Gunderson and Barrett-Connor ^{15 )}, hypercholesterolaemia⁽ Reference Kivipelto, Helkala and Laakso ^{14 )}, diabetes⁽ Reference Luchsinger, Tang and Stern ^{16 )}, heart disease⁽ Reference Reitz, Brickman and Luchsinger ^{17 )}, stroke⁽ Reference Schneider, Wilson and Bienias ^{18 )}, metabolic syndrome⁽ Reference Kalmijn, Foley and White ^{19 ,} Reference Yaffe, Weston and Blackwell ^{20 )} and statin⁽ Reference Solomon, Kareholt and Ngandu ^{21 ,} Reference Beydoun, Beason-Held and Kitner-Triolo ^{22 )} and non-steroidal anti-inflammatory⁽ Reference t'Veld, Ruitenberg and Hofman ^{23 )} medications. The strong ties between CVD and diet and between CVD and dementia support the idea that diet also plays a role in dementia at the very least through its effect on cardiovascular-related conditions. Genetic causes of AD are restricted to early-onset cases in persons less than 60 years of age and account for only 1% of all cases⁽ Reference Bekris, Yu and Bird ^{24 )}.

Cognitive decline

Central to all degenerative dementias is decline in cognitive abilities. Thus, an important and complementary area of study is individual change over time in neuropsychological test performance. These types of studies do not require clinical neurological evaluation by expert neurologists and thus are less costly than are studies of dementia diagnosis. Also, larger populations can be studied to examine risk factor associations with cognitive decline earlier in the disease process resulting in greater statistical power and less bias than studies using clinical diagnosis of dementia⁽ Reference Morris, Evans and Hebert ^{25 )}. In this type of study design in which exposures of interest are related to within-person rates of cognitive decline, to the extent that potential confounding factors for individuals remain constant over the cognitive assessment periods, confounding bias is minimised⁽ Reference Morris, Evans and Hebert ^{25 )}. The outcome (change in neuropsychological test performance) is not subjected to biases in clinician diagnosis arising from cultural, educational or other reasons. The greater the number of cognitive tests administered at each assessment period and the greater the number of cognitive assessments, the more accurate the measure of cognitive decline.

Cognitive decline is not disease specific. Therefore, a finding of association between a risk factor and cognitive decline does not necessarily mean that one should observe an association between the risk factor and any one of the dementia diseases in that population. However, one might be concerned to observe a risk factor association with a specific type of dementia in the study population but not with cognitive decline. This could be an indicator of a spurious disease association or perhaps early disease effects on the risk factor.

Underlying biologic mechanisms of antioxidant nutrients

The brain is a site of high metabolic activity and is especially prone to oxidative stress and damage to neural tissue⁽ Reference Mecocci, MacGarvey and Kaufman ^{34 –} Reference Haigis and Yankner ^{36 )}. A prevailing theory is that oxidative damage and neural inflammation are the underlying biological mechanisms of neurodegenerative disorders like AD and Parkinson's disease⁽ Reference Christen ^{37 –} Reference Beal ^{40 )}. There are two types of antioxidants in the body: antioxidant enzymes that prevent the generation of toxic substances and antioxidant nutrients that neutralise free radicals and singlet molecules of O⁽ Reference Sardesai ^{41 )}. The antioxidant enzymes require the availability of selected nutrients, including Mn, Cu, Se and Zn⁽ Reference Sardesai ^{41 )}. Antioxidant nutrients include vitamin E (tocopherols), vitamin C, carotenoids and flavonoids. The antioxidant nutrients may play more important roles in the aging brain than in other organs of the body due to the fact that there are relatively fewer antioxidant enzymes for neuronal protection⁽ Reference Olanow ^{42 )}. There is also evidence that vitamin E may play a role in the developing brain. For example, infants with cystic fibrosis who had prolonged deficiency levels of α-tocopherol (<300 μ/dl plasma) had statistically significantly lower scores (P<0·05) on cognitive tests in childhood⁽ Reference Koscik, Farrell and Kosorok ^{31 ,} Reference Koscik, Lai and Laxova ^{43 )}.

Epidemiological evidence from prospective studies for vitamin E in food

The prospective epidemiological studies of dietary vitamin E (i.e. from food sources) consistently show statistically significant inverse associations with incident dementia and AD, and with cognitive decline (Table 1). The one negative study was the Washington Heights-Inwood Columbia Ageing Project⁽ Reference Luchsinger, Tang and Shea ^{44 )}. One possible explanation for the negative finding may be that the Washington Heights-Inwood Columbia Ageing Project study population had a narrow range of low intake levels (4–7 mg/d for the quartile medians). The top quartile median intake level of 7 mg/d was in the lowest tertile of intake of 10 mg/d in the Rotterdam study that reported positive findings for vitamin E⁽ Reference Engelhart, Geerlings and Ruitenberg ^{45 )}.

Table 1. Prospective studies of dietary antioxidant nutrients and dementia outcomesFootnote *

AD, Alzheimer's disease; BLSA, Baltimore Longitudinal Study of Aging; Dem, dementia; Cog, cognitive; CHAP, Chicago Health and Ageing Project; EVA, Etude du Vieillissement Arteriél; MCI, mild cognitive impairment; MMSE, mini-mental status examination; 3MS, three mini-mental staus examinations; NHS, Nurses' Health Study; PAQUID, Persones Agèe QUID; WHICAP, Washington Heights-Inwood Columbia Ageing Project.

* Upward arrows indicate statistically significant increased risk; downward arrows, statistically significant decreased risk; –, no association; blank, not examined.

† Marginally statistically significant (P=0·07) in low powered study.

‡ Statistically significant interaction; inverse association in current smokers. No association overall.

Findings from studies of biochemical levels of vitamin E are less consistent than that for the self-reported diet studies (Table 1). A comparison of the ranges of biochemical vitamin E levels revealed that differences in the ranges could account for null findings in some studies. For example, population levels in the negative study by Ravaglia et al. ⁽ Reference Ravaglia, Forti and Lucicesare ^{46 )} are likely too low to provide protective benefit; the highest tertile level of α-tocopherol of 6·5 μmol/mmol cholesterol was in the lowest tertile of that reported by Mangialasche et al., a positive study⁽ Reference Mangialasche, Kivipelto and Mecocci ^{47 )}. Similarly, the entire biochemical range of vitamin E levels in the Nurses’ Health Study⁽ Reference Kang and Grodstein ^{48 )}, a null study, may be within the optimum range for cognitive function; the lowest tertile of α-tocopherol (median 24·85 μmol/l) fell within the highest tertile of the Persones Agèe QUID study (<25 μmol/l)⁽ Reference Helmer, Peuchant and Letenneur ^{49 )}. In the Persones Agèe QUID study, which had a wide range of α-tocopherol levels, the risk of dementia was three times higher (P=0·03) in the lowest tertile (11–21 μmol/l) compared with the highest tertile (25–55 μmol/l) level⁽ Reference Helmer, Peuchant and Letenneur ^{49 )}. Another explanation for the seemingly inconsistent findings may be that total tocopherol intake is more important for neuroprotection than the individual tocopherols alone. In the Kungsholmen study⁽ Reference Mangialasche, Kivipelto and Mecocci ^{47 )} plasma levels of total tocopherols, total tocotrienols and total vitamin E were each inversely associated with incident AD, but none of the individual tocopherols in isolation was associated. In fact, none of the biochemical studies found association with the individual tocopherols (Table 1).

Evidence for other antioxidant nutrients

At this stage in the research there is weak evidence to support protection against dementia by dietary intake of other antioxidant nutrients (e.g. vitamin C, β-carotene and flavonoids). Only the Cache County Study⁽ Reference Zandi, Anthony and Khachaturian ^{50 )} found protective benefit with high dietary intakes of vitamin C and β-carotene against the development of dementia (Table 1). Investigators from the Rotterdam study initially reported statistically significant inverse associations after 6 years of follow-up for vitamin C, and for smokers only, for β-carotene and flavonoids⁽ Reference Engelhart, Geerlings and Ruitenberg ^{45 )}. However, these protective associations were not maintained after 9·6 years of follow-up; only the inverse association for vitamin E remained statistically significant⁽ Reference Devore, Grodstein and van Rooij ^{51 )} (Table 1).

Absence of evidence for antioxidant vitamin supplements

A number of prospective epidemiological studies⁽ Reference Luchsinger, Tang and Shea ^{44 ,} Reference Zandi, Anthony and Khachaturian ^{50 ,} Reference Fillenbaum, Kuchibhatla and Hanlon ^{52 –} Reference Maxwell, Hicks and Hogan ^{57 )} and four randomised clinical trials⁽ Reference Kang, Cook and Manson ^{58 –} Reference Petersen, Thomas and Grundman ^{61 )} examined the relation of vitamin E supplement use to AD or cognitive decline. With few exceptions⁽ Reference Zandi, Anthony and Khachaturian ^{50 )} both types of studies found null associations of vitamin E supplement use with the cognitive outcomes. Dose level may play a role in the negative findings of studies that investigated α-tocopherol supplements. Vitamin supplement users and clinical trial participants tend to be healthy behaviour-seeking individuals and their diets likely already provide sufficient levels of nutrients for optimum functioning. In the Chicago Health and Ageing Project, vitamin E supplement use was inversely associated with cognitive decline but only among the participants who had low levels of intake from food⁽ Reference Morris, Evans and Bienias ^{55 )}. In this study, vitamin E supplementation was not associated with further reductions in the rate of cognitive decline among participants who had high food-intake levels of vitamin E. Post-hoc analyses of the Women's Health Study randomised trial revealed a statistically significant effect of vitamin E supplementation on the rate of cognitive decline among women who had low dietary intake at the baseline (below the median of 6·1 mg/d)⁽ Reference Kang, Cook and Manson ^{60 )}. These investigators performed similar post-hoc analyses for the Women's Antioxidant and Cardiovascular Study but using a much higher cut-off (15 mg/d) than in the Women's Health Study and found no effect of vitamin E supplementation⁽ Reference Kang, Cook and Manson ^{58 )}. This raises the possibility that the level of nutrient insufficiency for cognitive function is closer to 6 mg/d than to 15 mg/d. Populations vary considerably in vitamin E intake levels. For example, nearly the entire New York Washington Heights-Inwood Columbia Ageing Project study cohort was below 6 mg/d⁽ Reference Luchsinger, Tang and Shea ^{44 )}, whereas less than 20% of the Chicago Health and Ageing Project study cohort had intakes this low⁽ Reference Morris, Evans and Bienias ^{54 )}.

Although the epidemiological evidence for a protective relation of dietary β-carotene on dementia is weak, there is evidence from the randomised trials that higher levels consumed through vitamin supplementation over the long term may have protective benefits. In the Physicians’ Health Study, β-carotene supplementation (50 mg every other day) for 18 years significantly reduced the rate of cognitive decline, but there was no effect of the β-carotene supplement after 1 year of supplementation⁽ Reference Grodstein, Kang and Glynn ^{59 )}. Also, in post-hoc analyses of the Women's Antioxidant and Cardiovascular Health Study, women in the lowest quintile of total carotenoid intake at baseline (<3·09 mg/d) who were randomised to the β-carotene arm had statistically significantly slower rate of cognitive decline compared with the women in the placebo arm (P for interaction 0·02)⁽ Reference Kang, Cook and Manson ^{58 )}.

Plausible explanations for inconsistent findings for food and supplements

There are a number of differences between dietary and supplement sources of nutrient intake that could account for the positive findings for dietary intake and negative findings for supplemental intake. One difference is that the vitamin E form used in the supplement studies was high-dose α-tocopherol, usually twenty-five to fifty times the levels consumed through diet. The vitamin E measured in the dietary studies was not a single form but a combination of four tocopherols (α-, β-, δ- and γ- tocopherols) and their sterioisomers (α-, β-, δ- and γ- tocotrienols). α-Tocopherol has its own transfer protein that preferentially incorporates α-tocopherol into the plasma over other tocopherol forms. Consumption of high doses of α-tocopherol (e.g. through high-dose vitamin supplements) may not be functionally optimal. Studies have shown that consumption of high-dose α-tocopherol decreases the absorption of γ-tocopherol, a potent anti-inflammatory⁽ Reference Yang, Lu and Ju ^{62 )} and that the highest level of antioxidant and anti-inflammatory action occurs with a combination of tocopherols⁽ Reference Devaraj, Leonard and Traber ^{63 )}. These studies highlight an important feature of diet which was not replicated in the vitamin supplement trials. Foods provide multiple dietary constituents at complementary dose levels to enhance absorption and metabolism for optimum physiological effect.

Biological mechanisms of lipids in dementia

Lipids represent the primary dry weight structure of the brain and the lipid, cholesterol, plays a central role in AD⁽ Reference Puglielli, Tanzi and Kovacs ^{68 )}. ApoE, a gene encoding cholesterol transport, is an established risk factor for AD in which the ApoE allele is associated with at least a doubling in risk⁽ Reference Evans, Beckett and Field ^{69 ,} Reference Bennett, Wilson and Schneider ^{70 )}. Depletion of brain cholesterol reduces the generation of amyloid β⁽ Reference Simons, Keller and De Strooper ^{71 )}. In animal models, a high-fat, hypercholesterolaemic diet has been shown to result in multiple deleterious effects on the brain including: significant amyloid β burden⁽ Reference Sparks, Liu and Gross ^{72 –} Reference Howland, Trusko and Savage ^{74 )}, phosphorylated τ⁽ Reference Ullrich, Pirchl and Humpel ^{75 )}, impaired cognition⁽ Reference Ullrich, Pirchl and Humpel ^{75 –} Reference Sparks, Friedland and Petanceska ^{78 )}, increased immunoreactivity⁽ Reference Sparks, Liu and Gross ^{72 ,} Reference Ullrich, Pirchl and Humpel ^{75 )} and microglia activation⁽ Reference Streit and Sparks ^{79 )} and decreased brain membrane fluidity and brain-derived neurotrophic factor⁽ Reference Pistell, Morrison and Gupta ^{77 )}. In contrast, animals fed diets enriched with corn oil (polyunsaturated fat) or olive oil (monounsaturated fat) had increased memory performance and activation of protein F₁ ⁽ Reference Wong, Murakami and Routtenberg ^{80 )}.

Dietary fat composition and dementia

The epidemiological evidence for the association of dietary fat composition and dementia comes from several lines of evidence including dietary intake and biochemical levels of fatty acids, and blood lipid levels. The prospective epidemiological studies of dietary fat composition fairly consistently show increased risk of dementia and faster rate of cognitive decline with higher intakes of saturated fat⁽ Reference Devore, Stampfer and Breteler ^{81 –} Reference Luchsinger, Min-Xing and Shea ^{87 )}, although there are some negative studies⁽ Reference Engelhart, Geerlings and Ruitenberg ^{88 ,} Reference Samieri, Feart and Letenneur ^{89 )}. The limited number of prospective studies that investigated the association of trans fat intake found increased dementia risk⁽ Reference Devore, Stampfer and Breteler ^{81 ,} Reference Morris, Evans and Bienias ^{84 ,} Reference Morris, Evans and Bienias ^{85 )}. Studies of polyunsaturated and monounsaturated fat intakes are very inconsistent with some finding inverse associations⁽ Reference Devore, Stampfer and Breteler ^{81 ,} Reference Beydoun, Kaufman and Satia ^{83 ,} Reference Morris, Evans and Bienias ^{84 ,} Reference Solfrizzi, Colacicco and D'Introno ^{90 )} and others no association⁽ Reference Eskelinen, Ngandu and Helkala ^{82 ,} Reference Heude, Ducimetiere and Berr ^{86 –} Reference Samieri, Feart and Letenneur ^{89 )} with dementia outcomes. One complicating factor in the interpretation of these studies is that intakes of ‘healthy fats’ (polyunsaturated and monounsaturated) are positively correlated with intakes of the unhealthy fats (saturated and trans). This can result in considerable negative confounding (i.e. confounding in the opposite direction of the true relation) that likely obscured real associations. Very few of these studies analysed the different dietary fats conjointly in the statistical analyses. Those that did control for intakes of other fats found inverse associations of polyunsaturated and monounsaturated fats with incident AD⁽ Reference Morris, Evans and Bienias ^{84 )} and cognitive decline⁽ Reference Devore, Stampfer and Breteler ^{81 ,} Reference Morris, Evans and Bienias ^{85 )}. Even so, the evidence for associations of polyunsaturated and monounsaturated fats in and of themselves is weak for dementia outcomes. It could be that the composition of the diet towards a higher ratio of unsaturated to saturated fats is what is most important for protection against dementia; of four studies that examined this relation⁽ Reference Devore, Stampfer and Breteler ^{81 ,} Reference Eskelinen, Ngandu and Helkala ^{82 ,} Reference Morris, Evans and Bienias ^{84 ,} Reference Morris, Evans and Bienias ^{85 )}, three found inverse associations⁽ Reference Devore, Stampfer and Breteler ^{81 ,} Reference Morris, Evans and Bienias ^{84 ,} Reference Morris, Evans and Bienias ^{85 )}.

Examination of the prospective blood lipid studies reveal an interesting pattern from middle-to-late adulthood in which, overall, blood cholesterol levels decrease with older age but the highest rate of decrease occurs among those who had the highest cholesterol levels in mid-life⁽ Reference Ferrara, Barrett-Connor and Shan ^{91 ,} Reference Solomon, Kareholt and Ngandu ^{92 )}. Studies relating mid-life cholesterol to late-life dementia have shown that compared with non-demented individuals, those who become demented exhibit a greater decrease in blood cholesterol level with older age⁽ Reference Solomon, Kareholt and Ngandu ^{92 )}. The same phenomenon has been demonstrated in longitudinal studies of blood pressure⁽ Reference Skoog, Lernfelt and Landahl ^{13 )} and weight⁽ Reference Whitmer, Gunderson and Barrett-Connor ^{15 )}. This may explain why most studies that examined blood cholesterol in mid-life have observed positive associations with the risk of dementia in late life⁽ Reference Kivipelto, Helkala and Laakso ^{14 ,} Reference Beydoun, Beason-Held and Kitner-Triolo ^{22 ,} Reference Solomon, Kareholt and Ngandu ^{92 –} Reference Notkola, Sulkava and Pekkanen ^{97 )}, but findings of studies that examined cholesterol levels in late life are either null⁽ Reference Yaffe, Weston and Blackwell ^{20 ,} Reference Rastas, Pirttila and Mattila ^{98 ,} Reference Li, Shofer, Kukull and Peskind ^{99 )} or inconsistent with some showing positive associations with dementia risk⁽ Reference Yaffe, Barrett-Connor and Lin ^{100 ,} Reference Moroney, Tang and Berglund ^{101 )} and others showing inverse associations⁽ Reference Reitz, Tang and Manly ^{102 ,} Reference Romas, Tang and Berglund ^{103 )}. One plausible explanation may be that the disease process of dementia accelerates the decreases with older age in levels of cholesterol, blood pressure and weight.

n-3 Fatty acids and the brain

The n-3 fatty acids are a class of polyunsaturated fat that includes three nutritionally important species: α-linolenic acid (18:3 n-3), EPA (20:5 n-3) and DHA (22:6 n-3). Mammals cannot synthesise α-linolenic acid and so it must be consumed through oils (e.g. soyabean, flaxseed, black currant and rapeseed) wheat germ, soyabeans, or nuts (e.g. walnuts). In human subjects, α-linolenic acid is elongated and desaturated to form EPA and DHA, but this conversion is not more than 0·5%⁽ Reference Harris, Mozaffarian and Lefevre ^{104 )}. Fish and seafood are a direct natural dietary source of DHA and EPA, although eggs and meats can also contain these n-3 fatty acids in animals with supplemented feed or who graze on certain grasses that contain α-linolenic acid.

DHA primary lipid in metabolically active brain regions

DHA is the primary n-3 fatty acid in the brain; other species of n-3 fatty acids are uncommonly found⁽ Reference Connor ^{105 )}. DHA is the predominant lipid in the most metabolically active areas of the brain including the cerebral cortex, synaptosomes and mitochondria⁽ Reference Carrie, Clement and de Javel ^{106 ,} Reference Connor, Neuringer and Lin ^{107 )}. In laboratory studies, animals fed diets enriched with n-3 PUFA had superior learning acquisition and memory performance⁽ Reference Carrie, Clement and de Javel ^{106 ,} Reference Gamoh, Hashimoto and Sugioka ^{108 –} Reference Yamamoto, Saitoh and Moriuchi ^{116 )} including in aged mice⁽ Reference Suzuki, Park and Tamura ^{109 ,} Reference Gamoh, Hashimoto and Hossain ^{117 –} Reference Sugimoto, Taga and Nishiga ^{119 )}, better regulation of neuronal membrane excitability⁽ Reference Itokazu, Ikegaya and Nishikawa ^{120 –} Reference Young, Gean and Chiou ^{125 )}, increased levels of neurotransmitters and receptors⁽ Reference Chalon, Delion-Vancassel and Belzung ^{126 –} Reference Innis ^{129 )}, increased hippocampal nerve growth⁽ Reference Ikemoto, Nitta and Furukawa ^{130 )}, greater fluidity of synaptic membranes⁽ Reference Suzuki, Park and Tamura ^{109 )} and reduced ischaemic damage to neurons⁽ Reference Okada, Amamoto and Tomonaga ^{131 )}.

Cardiovascular mechanisms of fish protection

Fish and n-3 fatty acids have favourable effects on the cardiovascular system including lower blood pressure, lower TAG and reduced cardiac death (one fish meal per week)⁽ Reference Harris, Mozaffarian and Lefevre ^{104 )}. In a meta-analysis of nine prospective studies of incident stroke there was a dose–response inverse association with increased fish consumption⁽ Reference He, Song and Daviglus ^{132 )}. Thus, in addition to the earlier-described mechanisms of DHA on brain health, dietary fish consumption may be protective particularly of vascular dementia through their beneficial effects on these cardiovascular risk factors.

Epidemiological evidence for fish and n-3 fatty acids and dementia

In prospective studies, consumption of fish was associated with lower risk of incident dementia⁽ Reference Barberger-Gateau, Letenneur and Deschamps ^{133 –} Reference Larrieu, Letenneur and Helmer ^{138 )}, stroke⁽ Reference Gillum, Mussolino and Madans ^{139 ,} Reference Iso, Rexrode and Stampfer ^{140 ,} Reference Keli, Feskens and Kromhout ^{141 ,} Reference Orencia, Daviglus and Dyer ^{142 )}, and cognitive decline⁽ Reference Heude, Ducimetiere and Berr ^{86 ,} Reference Morris, Evans and Tangney ^{143 )}. A number of studies investigated the relation of fish intake to risk of developing AD. In all cases, the estimated OR were in the protective direction and most were statistically significant at a low level of intake of just one fish meal per week⁽ Reference Barberger-Gateau, Letenneur and Deschamps ^{133 –} Reference Larrieu, Letenneur and Helmer ¹³⁸ ,Reference Devore, Grodstein and van Rooij ^{144 )}. The Rotterdam Study reported significant reduction in risk after 2·1 years of follow-up⁽ Reference Kalmijn, Launer and Ott ^{135 )}, but the association after 9 years was not statistically significant⁽ Reference Devore, Grodstein and van Rooij ^{144 )}. One plausible explanation for these discrepant findings may be misclassification error in the dietary exposure as diet was not reassessed to capture potential changes in fish consumption over the 9-year period. Two of the studies observed the inverse association with dementia only among persons who were apoE-ε4 negative⁽ Reference Huang, Zandi and Tucker ^{134 ,} Reference Barberger-Gateau, Raffaitin and Letenneur ^{145 )}, but this was not observed in other studies⁽ Reference Kalmijn, Launer and Ott ^{135 –} Reference Larrieu, Letenneur and Helmer ¹³⁸ ,Reference Devore, Grodstein and van Rooij ^{144 )}. Of interest, the relation of fish consumption to AD risk in the Cardiovascular Health Study was specific to fatty fish and not to fried fish⁽ Reference Huang, Zandi and Tucker ^{134 )}. This raises the question of whether fish preparation might alter the neuroprotective benefit given that frying reduces the n-3 fatty acid levels and increases saturated fat intake. In the Women's Health Initiative Observational Study, women who consumed five or more servings per week of baked or broiled fish had a 30% reduction in the risk of incident heart failure, whereas women who consumed one or more weekly servings of fried fish had an increase in risk (hazard ratio 1·5, 95% CI 7·2, 1·8)⁽ Reference Belin, Greenland and Martin ^{146 )}.

A few randomised trials on the effects of DHA supplements on cognitive decline reported negative findings. A major flaw in the designs of these trials⁽ Reference Quinn, Raman and Thomas ^{147 ,} Reference van de Rest, Geleijnse and Kok ^{148 )} is that they allowed fish consumption up to three fish meals per week among all trial participants. This is far above the level (one fish meal per week) that was observed to be inversely associated with dementia in the majority of epidemiological studies. As a result, the difference between treatment and placebo groups in intake levels at the level of benefit may have been minimised thus resulting in null findings of the supplement.

Several studies examined the relation between fish consumption and decline in cognitive test scores with two studies showing statistically significant reductions at just one fish meal per week⁽ Reference Morris, Evans and Tangney ^{143 ,} Reference van Gelder, Tijhuis and Kalmijn ^{149 )}, whereas the third study was null⁽ Reference van de Rest, Geleijnse and Kok ^{148 )}. Neither of the two positive studies observed statistically significant inverse associations with dietary intake levels of the marine n-3 fatty acids⁽ Reference Morris, Evans and Tangney ^{143 ,} Reference van Gelder, Tijhuis and Kalmijn ^{149 )}. However, of the six studies that measured biochemical levels of n-3 fatty acids, three observed statistically significant reductions in the rate of cognitive decline⁽ Reference Beydoun, Kaufman and Satia ^{83 ,} Reference Heude, Ducimetiere and Berr ^{86 ,} Reference Whalley, Deary and Starr ^{150 )} and two others observed inverse associations with dementia⁽ Reference Samieri, Feart and Letenneur ^{89 ,} Reference Schaefer, Bongard and Beiser ^{137 )}. The apparent inconsistency may be due to the fact that food frequency assessment of marine n-3 fatty acid intake is relatively imprecise. This is because most questionnaires employ at most three general questions on fish consumption and thus do not capture well the different levels of EPA and DHA contained in different types of fish. Correlations between dietary levels of the n-3 fatty acids and biochemical levels are typically about 0·35⁽ Reference Hunter and Willett ^{151 )}.

Vitamin B₁₂ and folate: cofactors in metabolic pathways

Vitamin B₁₂ and folate have essential and intertwined roles in human health as they are co-factors in the metabolism of DNA and erythrocytes synthesis and with methionine and S-adenosylmethionine synthesis. Both vitamin B₁₂ and folate (5-methyl-tetrahydrofolate (THF)) are required for the remethylation of homocysteine to methionine, while simultaneously demethylating 5-methyl-THF to THF⁽ Reference Herbert, Das, Shils, Olson and Shike ^{155 )}. Thus, with vitamin B₁₂ deficiency, there is an accumulation of homocysteine and 5-methyl-THF. Vitamin B₁₂ deficiency syndrome is characterised by peripheral neuropathy, megaloblastic anaemia, fatigue, depression and cognitive impairment⁽ Reference Savage and Lindenbaum ^{156 )}. The neuropathy characteristic of vitamin B₁₂ deficiency is probably the result of the block in this pathway because myelin basic protein is ‘hypomethylated’. The typical haematological manifestations of vitamin B₁₂ and folate deficiency (megaloblastic anaemia and granulocyte hypersegmentation) are thought to be the result of a reduction in DNA synthesis when 5,10-methylene-THF is lacking. Synthetic folic acid may correct this problem if the tissue contains an enzyme that can reduce folic acid to THF.

B-vitamins, cognitive decline and Alzheimer's disease

Epidemiologic studies and randomised clinical trials on the international scene for B-vitamins and cognition have produced seemingly inconsistent findings that may reflect the complex interrelation of vitamin B₁₂ and folate in the metabolic pathways as well as differences among study populations in vitamin deficiencies, vitamin supplementation and food fortification. An important public health issue is whether or under what conditions, B-vitamin supplementation and food fortification are of benefit or harm to older persons. Vitamin B₁₂ deficiency is common in older adults⁽ Reference Carmel ^{157 ,} Reference Pfeiffer, Caudill and Gunter ^{158 )}. Folate deficiency, on the other hand, is now a rare occurrence in the USA since the folic acid fortification⁽ Reference Kalmbach, Choumenkovitch and Troen ^{154 )}.

In recent years, there has been a lot of interest in vitamin B₁₂ and folate as risk factors for dementia. This interest is largely based on their relations as co-factors in the metabolism of homocysteine. Homocysteine has been related to the risk of developing AD in some⁽ Reference Seshadri, Beiser and Selhub ^{159 ,} Reference Haan, Miller and Aiello ^{160 )} but not all studies⁽ Reference Luchsinger, Tang and Shea ^{161 )} and remains a topic of interest. The mechanism for association is not known, although homocysteine and folate deficiency have been shown to be neurotoxic in mouse models of AD⁽ Reference Ho, Collins and Dhitavat ^{162 ,} Reference Tchantchou and Shea ^{163 )}.

Commonly, observational studies and randomised trials that examine nutrient effects on disease processes do not consider dose level when interpreting study results and trial design and, as noted earlier, this consideration is particularly important for folate because of folic acid fortification in some countries. Populations will also differ in vitamin B₁₂ nutriture depending on the prevalence of atrophic gastritis and the use of medications that affect gastric acidity in addition to dietary and supplemental intake. Thus, some of the inconsistent findings across studies of B-vitamin effects on cognitive decline and dementia may be due to the range of nutrient status in the study population. Of ten cohort studies that examined the relations of these B-vitamins and cognitive decline, four reported protective associations with folate⁽ Reference Mooijaart, Gussekloo and Frolich ^{164 –} Reference Kim, Kim and Shin ^{167 )} and three reported protective associations with vitamin B₁₂ ⁽ Reference Tucker, Qiao and Scott ^{165 ,} Reference Morris, Evans and Bienias ^{168 –} Reference Clarke, Birks and Nexo ^{170 )} but many found no association with one or the other B-vitamin⁽ Reference Tucker, Qiao and Scott ^{165 ,} Reference Clarke, Birks and Nexo ^{170 –} Reference McCaddon, Hudson and Abrahamsson ^{174 )}. The Chicago Health and Ageing Project study found a deleterious effect of faster decline among persons with food and/or supplement intakes exceeding 400 μg/d⁽ Reference Hebert, Beckett and Scherr ^{6 )}. This USA study took place primarily post-fortification. The deleterious findings with high levels of folic acid were also reported in a cross-sectional study of National Health and Nutrition Examination Survey data⁽ Reference Morris, Jacques and Rosenberg ^{175 )} in which persons with low vitamin B₁₂ status and elevated serum folate concentrations were more likely to manifest impaired cognitive performance than those with normal serum folate concentrations.

A number of prospective studies that examined the relation between folate and dementia found inverse associations⁽ Reference Luchsinger, Tang and Miller ^{176 –} Reference Kim, Stewart and Kim ^{180 )}, although several studies did not⁽ Reference Seshadri, Beiser and Selhub ^{159 ,} Reference Morris, Evans and Bienias ^{181 ,} Reference Nelson, Wengreen and Munger ^{182 )}. In contrast, the studies that examined the association of vitamin B₁₂ with AD were null⁽ Reference Seshadri, Beiser and Selhub ^{159 ,} Reference Luchsinger, Tang and Miller ^{176 –} Reference Corrada, Kawas and Hallfrisch ^{178 ,} Reference Kim, Stewart and Kim ^{180 –} Reference Nelson, Wengreen and Munger ^{182 )}. There have been a number of randomised trials of the effects of supplementation with folic acid and vitamin B₁₂ on cognitive decline. Most of the trials included a small number of subjects and supplemented over short periods of weeks or months, thus making it difficult to interpret the null results. Of several large randomised trials that tested B-vitamin supplementation for longer periods of time (i.e. 2–10 years)⁽ Reference Durga, van Boxtel and Schouten ^{183 –} Reference Kang, Cook and Manson ^{186 )}, only one⁽ Reference Durga, van Boxtel and Schouten ^{183 )} targeted individuals who had low folate status. In this trial, after 3 years of folic acid supplementation, the treated group had a statistically significant slower rate of cognitive decline compared with the placebo group⁽ Reference Durga, van Boxtel and Schouten ^{183 )}. In another randomised trial of folic acid, vitamin B₁₂ and vitamin B₆ supplementation in older persons with high homocysteine (VITACOG and ISRCTN 94410159), 2 years of treatment reduced decline in global cognition and in memory⁽ Reference de Jager, Oulhaj and Jacoby ^{184 )}. Treated subjects in this trial also had evidence of reduced brain atrophy compared with participants taking placebo⁽ Reference Smith, Smith and de Jager ^{187 )}. Two other trials tested a combination therapy of folate, vitamin B₁₂ and vitamin B₆ on cognitive decline and found no effect overall⁽ Reference McMahon, Green and Skeaff ^{185 ,} Reference Kang, Cook and Manson ^{186 )}. It is noteworthy that post-hoc analyses of one of these negative trials⁽ Reference Kang, Cook and Manson ^{186 )} found protective effects of the B-vitamin supplement among women who had low B-vitamin intake at baseline.

In summary, based on the current available evidence, insufficient levels of vitamin B₁₂ and folate may be associated with faster cognitive decline. While there is some evidence to support that insufficient folate may also increase the risk of developing AD this is not the case for vitamin B₁₂. In addition, there is a possibility that exposure to high levels of folic acid in vitamin supplements or fortified food may be also associated with cognitive decline. In view of the folate fortification policy in some countries, this needs to be investigated further. Mandatory food fortification and vitamin supplementation have resulted in unprecedented numbers of persons with high circulating folic acid concentrations in the USA⁽ Reference Pfeiffer, Caudill and Gunter ^{158 )}. This substantially raises the potential for misdiagnosis of subtle vitamin B₁₂ deficiency and increases prevalence of its neurologic syndrome. Masking of vitamin B₁₂ deficiency can occur when there is correction of the macrocytic anaemia but not of the neurologic consequences. In case reports nearly five decades ago, folic acid treatment of vitamin B₁₂-deficient individuals was reported to aggravate neurological complications⁽ Reference Savage and Lindenbaum ^{156 ,} Reference Chodos and Ross ^{188 )}. These older reports gain greater importance in light of the recent Chicago Health and Ageing Project and National Health and Nutrition Examination Survey studies⁽ Reference Morris, Evans and Bienias ^{168 ,} Reference Morris, Jacques and Rosenberg ^{175 )}. The possible adverse outcomes cannot be ignored in the face of the prevalence of subtle vitamin B₁₂ deficiency and the dramatic increase in folate intake that has occurred in the USA and other countries that have mandatory fortification.