Smoking is a risk factor for several chronic diseases, but its long-term relationship with dementia of various sub-types is still controversial. In fact, smoking is well known to increase the risk of stroke [269] and thus subsequent vascular types of dementia (VaD). However, many studies have concluded that smoking status influenced risk of VaD independently of stroke status and thus may have an effect beyond cerebrovascular disease. In addition, studies that have shown a direct impact of smoking on AD, suggest that smoking might in fact influence neurodegeneration. A vast amount of literature points to a role of smoking in oxidative stress and inflammation, both mechanisms believed to play a key role in AD [270].

However, it is also biologically plausible that smoking might protect against cognitive decline and AD, given that nicotine, a key active component of tobacco, may enhance the release of acetylcholine, increase the density of nicotnic receptors, therefore improving attention and information processing [271]. It is now known that AD is characterized by cholinergic system deficits which may be delayed through tobacco consumption [271,272].

Population-based evidence of an effect of smoking on cognitive outcomes was inconclusive, with most longitudinal studies reporting weak or null associations [63,72,74,76,77,82,84,90]. However, a number of other cohort studies have found a positive association between smoking and risk of incident dementia and AD [78,86,88,93,95,98,101,103,104] as well as incident cognitive impairment [75,81,90] and age-related cognitive decline [83,85,89,96,99,105].

For instance, the British 1946 birth cohort study pointed to the difficulty of finding an association between smoking and cognitive impairment given the differential high mortality of smokers especially among the elderly population [85]. After controlling for a range of socioeconomic and health status indicators (both physical and mental), they found that smokers who survive into later life may be at risk of clinically significant cognitive decline. However, these effects were accounted for largely by heavy smokers, i.e. those who smoked 20 cigarettes per day or more. Earlier research on middle aged adults suggested that current smoking and number of pack-years of smoking were related to reduced performance on tests of psychomotor speed and cognitive flexibility assessed approximately five years later [83]. Similar results were shown for cognitive decline in a large cohort study (Rotterdam study) conducted in multiple European countries [89] and in another more recent study conducted in the United States [92].

Among studies that examined incidence of AD in relation to smoking status, two of the largest European cohort studies reported conflicting results. While one found no relationship between smoking status and incident AD among a large sample of 34,439 older UK men (mean age: 81) [82], the recent 2011 study found that heavy smoking increased the risk of dementia and AD in a younger sample of 21,123 older Finnish adults (Mean age:60.1) that comprised both men and women [104].

In sum, 16 (55.2%) out of the 29 selected cohort studies linking smoking to the various cognitive outcomes found the relationship to be in the hypothesized direction in the entire population that was studied and for most outcomes of interest [75,81,83,85,86,88,89,92-96,98,99,101,104], while 2 found this relationship for some outcomes but not others [97,103] and 2 detected it for a sub-group of the total population [78,105], while the remaining 9 studies did not find an association [63,72,74,76,77,82,84,90] or found an association in the opposite direction [102]. (See Table 2 and Figure 2A).

Only 2 (28.6%) of the 7 cross-sectional studies found an association in the hypothesized direction [36,87], while 2 detected it for a sub-group of the total population [79,100], and the remaining 3 did not detect a significant association [73,80,91], (See Table 2 and Figure 2B).

Alcohol consumption in moderation was hypothesized to be protective against cognitive decline and impairment in old age. Several mechanisms may be involved in explaining the potential protective effect of moderate alcohol consumption on various cognitive outcomes. First, this effect might be mediated through cardiovascular risk factor reduction, partly through a dampening effect of ethanol on platelet aggregation, or through a modification of the serum lipid profile. Second, another potential mechanism in which alcohol can have a direct effect on cognitive function is through acetylcholine release in the hippocampus, which in turn enhances learning and memory [273].

The Rotterdam study [83] also examined the effect of alcohol use on cognition. They found that past alcohol consumption’s effect on speed and flexibility appeared to be slightly U-shaped, with the best performance observed among those who drank 1–4 glasses of alcohol per day, although this association was stronger among women than among men. Other studies also detected sex differences [106,107,111]. Light to moderate alcohol consumption was also found beneficial based on findings of other cohort and cross-sectional studies with a U- or J-shaped pattern observed [108-110,113,114,116,117,119-121,124-127]. However, in other studies, a linear dose–response relationship between alcohol use and improved cognition was noted, though the authors cautioned that these should not encourage increased alcohol consumption without an upper bound to this consumption [114,115,122,131].

In one cross-sectional study, a linear relationship between alcohol consumption and cognitive function was found in women but a U-shaped pattern was found in men [109]. One cohort study found that overall, moderate consumption was protective against poor cognitive function, but had an opposite relationship with cognitive function among ApoE4⁺ individuals [113], while another found that alcohol use in general was related to better cognition without effect modification by ApoE4 status [123]. Slower memory decline with increased alcohol consumption in men was found in one study, though the opposite relationship was found in the case of psychomotor speed among women [118]. The positive association between alcohol intake and memory was also noted in at least one other cross-sectional study for both men and women combined [128]. Moreover, heavy alcohol use was linked to poorer cognitive outcomes in a few studies [87,127,129,130]. Finally, only a few studies among those that were selected found no associations between alcohol consumption and cognitive outcomes [76,77,112].

In fact, 8 out of the 18 selected cohort studies (44%) linking alcohol consumption to the various cognitive outcomes, found the relationship to be in the hypothesized direction (but were U-shaped, J-shaped or linear) in the entire population that was studied and for most outcomes of interest [108,114,116,117,122,124,125,131], while 2 found this relationship for some outcomes but not others [115,120] and 4 detected it for a sub-group of the total population [83,111,113,118]. Moreover, 1 cohort studies have indicated that alcohol use was generally linked to poor cognitive outcomes for the total population [87]. Finally, 3 did not find any significant associations between alcohol consumption and the various cognitive outcomes that were under study [76,77,129]. (See Table 2 and Figure 2A).

9 of the 12 cross-sectional studies (75%) found an association in the hypothesized direction for the entire study population and for most outcomes of interest [106,109,110,119,121,123,126-128]. The remaining 3 studies either found this U-shaped or J-shaped association in a sub-group [107], and either failed to detect any association [112] or detected one that was not in line with the hypothesis, whereby alcohol use was generally found to result in poor cognitive outcomes [130]. (See Table 2 and Figure 2B).

Physical activity has many well-known benefits for preventing a number of chronic disorders, including coronary heart disease, stroke, diabetes mellitus and osteoporosis. However, its impact on cognitive functioning has not been studied extensively. Several mechanisms may underlie the potentially protective effects of physical activity on cognitive function, including sustained cerebral blood flow [274], improved aerobic capacity and cerebral nutrient supply [275,276] as well as growth factors, specifically the brain-derived neurotropic factor, which is a molecule that increases neuronal survival, enhances learning, and protects against cognitive decline [277,278].

Currently, 24 cohort and 4 cross-sectional studies have examined the hypothesized relationship. For instance, a recent cohort study of 716 dementia-free older adults from the Rush Memory and Aging Project who were followed-up for an average of 4 years found an inverse relationship between total daily physical activity and incident AD after controlling for age, sex, education, self-report physical, social, and cognitive activities, as well as current level of motor function, depressive symptoms, chronic health conditions, and ApoE4 allele status [156]. Furthermore, a recent cross-sectional study of 9344 women, 65 years and older, found a lower prevalence of cognitive impairment among those who reported being physically active versus those who reported being physically inactive at different stages of their lives [155].

These findings suggested that physical activity could represent an important and potent protective factor for cognitive decline and dementia in elderly persons. Significant findings were obtained by other recent cohort [132-145,147,148,150-154,156-158] and cross-sectional studies [100,146,149]. Only one cohort study resulted in non-significant findings [143].

In sum, of the 24 selected cohort studies linking physical activity to the various cognitive outcomes, 21(87.5%) found the relationship in the hypothesized direction in the entire population that was studied and for most outcomes of interest [132,133,135-140,142,144,145,147,148,150-154,156-158,279], while 1 found this relationship in ApoE4 carriers [134] and 1 in non-carriers [141]. In one cohort study, the association was against its hypothesized direction [143]. (See Table 2 and Figure 2A).

In addition, all 4 of the selected cross-sectional studies (100%) found an association in the hypothesized direction for the entire study population and for most outcomes of interest (See Table 2 and Figure 2B).

Caffeine is known to be the most widely used psychoactive drug worldwide. Its main source is coffee particularly in Western diets. Acting as a stimulant of the central nervous system, caffeine causes heightened alertness and arousal [280]. Previous literature yielded inconsistent findings about the effects of caffeine consumption on cognitive processes. In fact, caffeine improved perceptual speed and vigilance, as well as more complex functions such as memory [281]. Caffeine is one type of compound known as methylxanthines whose effects are mainly to block adenosine receptors in the brain, resulting in cholinergic stimulation. It was hypothesized that such stimulation would lead to improved memory [282]. The earliest large cross-sectional study conducted by Jarvis and colleagues found that caffeine improved cognitive performance [159]. Later on, other cross-sectional studies focusing on tea consumption found similar results [100,162,167,170]. Others, however, did not show evidence of a significant protective effect [160,168]. In sum, 4 of 7 selected cross-sectional studies linking caffeine consumption to various cognitive outcomes found the relationship to be in the hypothesized direction in the study population and for most outcomes of interest (57.1%), one found this association in men [100] and two failed to find a significant association [160,168]. (See Table 2 and Figure 2B).

Of 11 cohort studies, positive findings pertained to 3 (27.3%) [163,164,175], though this was found only for coffee intake in two studies [165,169], while 5 recent studies detected this association only among women or for specific exposures [165,169,171-173]. The remaining cohort studies (3 of 11, 27%) did not find an association between caffeine intake and cognitive change [161] or incident dementia [166,174]. Given the paucity of large cohort studies, more research is needed to establish causality (See Table 2 and Figure 2A).

Several findings suggest that oxidative stress may play an important role in the pathogenesis of AD. First, the brains of AD patients have lesions that are associated with exposure to free radicals. Moreover, oxidative stress among these patients is also marked by an increased level of antioxidants in the brain that act as free radical scavengers. Finally, in vitro studies suggest that exogenous antioxidants may reduce the toxicity of β-amyloids in the brains of AD patients [283-285]. Based on these findings, it may be hypothesized that dietary antioxidants may help reduce the risk of AD.

Those epidemiologic studies examined the longitudinal relationship between supplemental antioxidants and risk of AD and other dementias found conflicting results: While vitamin C supplement use was related to lower AD risk in one cohort study [178], combined supplementation of vitamin E and vitamin C was associated with reduced prevalence and incidence of AD and cognitive decline in three other cohort studies [189,193,198], whereas another study found this effect to be specific to Vitamin E supplements [186]. These findings of a protective effect of supplemental antioxidant use against cognitive impairment and decline was replicated in a large cohort study [185]. However, there were only borderline or little evidence of a cognitive benefit from use of antioxidant supplements, particularly vitamins C and E, according to at least five independent cohort studies [177,180,187,192,199].

There are several prospective cohort studies on the effect of dietary antioxidants on the risk of dementia. One study found that high dietary intake of vitamins C and E may reduce the risk of AD [182] with the relationship most pronounced among smokers. Morris and colleagues [183] found that dietary intake of vitamin E, but not other antioxidants, was associated with a reduced risk of incident AD, although this association was restricted to individuals without the Apolipoprotein E ϵ4 genotype. Similar findings were reported with cognitive decline as an outcome [184]. In a later study when both outcomes were considered it was concluded that certain forms of tocopherols not found in dietary supplements but found only in foods may be at play [194]. This observation was corroborated by at least one recent study [197]. Another study, however, suggested that dietary antioxidants were not able to reduce AD risk [187]. Similarly, Laurin and colleagues [188] found no association between midlife dietary intake of vitamins E and C and dementia incidence. At least five other cohort studies came to a similar conclusion [176,181,201,202]. In addition to examining associations of cognition with vitamins C and E, other studies found that carotenoids, particularly β-carotene intake, may be have beneficial effects of various cognitive outcomes [176], though others were not able to detect such an association [184,201,202].

Irrespective of the source of antioxidants, plasma concentration may be a good biomarker for oxidative stress status. In particular, an inverse association between plasma vitamin E among others and poor cognitive outcomes was found in at least two cross-sectional studies [179,190] and two cohort studies [196,200]. Another cross-sectional study, however, did not find evidence of an association between plasma antioxidants, including vitamin E and prevalent AD [191]. In addition, among studies that examined the influence of plasma carotenoids [179,195], only one detected a significant potential protective effect against cognitive impairment [195]. While these results are mixed, they suggest that at least one antioxidant has a protective effect against adverse cognitive outcomes.

In sum, of the 21 selected cohort studies linking antioxidants, with focus on vitamin E, to the various cognitive outcomes, 9 (42.9%) found the relationship to be in the hypothesized direction in the entire population that was studied and for most outcomes of interest [176,178,182,184,189,193,194,197,198], while 5 found this relationship for specific antioxidants or some outcomes but not others [180,183,186,196,200,201] and 1 detected it for a sub-group of the total population [183]. The remaining selected cohort studies (n = 6) did not find a significant association [177,185,187,188,192,199]. (See Table 2 and Figure 2A).

Similarly, of the 6 cross-sectional studies that were selected, 2 (33.3%) found an association in the hypothesized direction for the entire study population and for most outcomes of interest [190,195], 1 found the association to hold only for vitamin E [179], whereas 3 found no significant association [181,191,202]. (See Table 2 and Figure 2B).

An elevated level of plasma concentration of the sulfur amino acid Hcy (hyperhomocysteinemia) is recognized as an independent risk factor for cardiovascular, peripheral vascular, and cerebrovascular disease [286]. Accordingly, a potential influence of hyperhomocysteinemia on cognitive functioning among older adults was postulated and several studies were able to associate high levels of Hcy with increased risk of incident AD or all-cause dementia [206,217,224,228,233,234]. Studies have pointed to selective effect of Hcy on specific domains of cognition [214,287,288]. One explanation could be that Hcy might be affecting certain parts of the brain to a greater extent than others, and studies have linked Hcy to higher degree of white matter hyperintensities and with brain atrophy [289-293].

Even though blood Hcy levels increase with age and diminished renal function, it is largely determined by dietary intake of B-vitamins (mainly B-6 and B-12) and folate which are needed to convert Hcy into methionine and cysteine, through the methylation reactions [294]. Thus, Hcy status in plasma can be modified by dietary interventions. Moreover, vitamin B-12 plasma level has been shown to be inversely related to that of Hcy [295] and studies looking at Hcy levels and cognitive functioning also examined the effect of B-vitamins. In particular, vitamin B-12 was found to be protective against decline in at least three recent studies [204,213,219]. At least five other studies [204,213,214,216,217,219,221] concluded that fotate was protective against cognitive impairment or decline. For Vitamin B-6, two other studies suggested a protective effect [213,219]. An antagonistic interaction of folate and Vitamin B-12 with Hcy’s effect on cognition was noted in other studies [224,296,297]. Aside from its link to cardiovascular disease, Hcy was shown to have neurotoxic and excitotoxic properties in vitro[298,299], suggesting a direct influence on cognition.

Overall, of the 19 selected cohort studies linking Hcy to the various cognitive outcomes, 12(63.2%) found the relationship in the hypothesized direction in the entire population that was studied and for most outcomes of interest [206,215,217,219,222-224,228,231,233-235], while 2 found this relationship for some outcomes but not others or a sub-group [213,214] and 5 were not able to detect a significant association [203,209,210,229,230]. (See Table 2 and Figure 2A).

Similarly, of the 14 cross-sectional studies that were selected, 11(78%) found an association in the hypothesized direction for the entire study population and for most outcomes of interest [204,205,207,208,216,218,220,221,225,227,232], 1 found an association only among older adults above age 60y [212], 1 detected it among ApoE4⁺ individuals [226], and 1 found no significant relationship [211]. (See Table 2 and Figure 2B).

Another nutritional factor hypothesized to be protective against cognitive decline is higher intake of n-3 fatty acids and/or a better balance of n-3/n-6 fatty acids. Linoleic(LA ~ 18:2n-6) and α-linolenic (LNA ~ 18:3n-3) are two types of fatty acids that are essential for all members of the animal kingdom. These fatty acids and their respective derivatives are also commonly referred to as n-6 and n-3 fatty acids. Their essentiality lies in the fact that they cannot be synthesized de novo within the human or animal organism [300].

In the past, n-3 fatty acids were classified only as essential because of their ability to alleviate deficiency symptoms that include dermatitis, growth retardation and reproductive failure. However, n-3 fatty acids have other important neurological functions, which explain their high concentrations in neural and retinal tissues [301-303]. Some of the longer chain fatty acids that are synthesized from α-linolenic acid include Eicosapentanoic acid (EPA ~ 20:5 n-3), which through further elongation, desaturation and β-oxidation produces Docosahexaenoic acid (DHA ~ 22:6 n-3). On the other hand, products of linoleic acid which are also termed long-chain n-6 fatty acids include gamma-linoleic (GLA ~ 18:3 n-6), dihomogammalinolenic acid (DGLA ~ 20:3 n-6) and Arachidonic acid (AA ~ 20:4 n-6) [304]. Of all organs in the human body (excluding adipose tissue), the nervous system has the highest lipid content. The dry weight of an adult brain is 50% to 60% lipid, and 35% of the lipid content is accounted for by polyunsaturated fatty acids (PUFAs) [305].

A review of scientific articles and biochemistry textbooks [306] suggested that the fatty acid composition of neuronal cell membrane phospholipids reflects their intake in the diet. Fish oils, which contain high levels of C20 and C22 PUFA, exert the most profound influence on brain PUFA concentrations [306]. The ratio between n-3 and n-6 PUFA may influence various aspects of serotoninergic and catecholaminergic neurotransmission, and it has been shown that by increasing the density of neurotransmitter receptors for acetylcholine and dopamine, dietary n-3 PUFA can improve learning and memory processes [307].

Previous observational studies suggested that the biochemical composition of blood components in terms of fatty acids differs significantly between subjects with normal cognitive functioning and patients with some form of cognitive impairment. While the majority of these studies showed an inverse association of plasma and erythrocyte n-3 fatty acids with cognition among older adults [243-245,248,255,258], at least one found no association between biochemical markers of n-3 fatty acids and cognition [251].

Epidemiological studies involving self-reported dietary data of n-3 fatty acids had suggestive but slightly controversial results. One study by Morris and colleagues used cohort data on 815 subjects who were initially unaffected by AD (age range: 65-94y, mean follow-up period = 2.3y). Using standardized criteria, AD incidence was compared across n-3 fatty acid consumption groups, with those eating fish once per week compared to those who rarely or never eat fish having considerably lower incidence (RR = 0.4; 95% CI: 0.2, 0.9). Total n-3 fatty acid consumption was also associated with a reduced AD risk even after controlling for intake of other dietary fats, vitamin E and for cardiovascular conditions [239]. A similar finding was reported later on for a larger but comparable cohort when looking at fish consumption and cognitive decline over time [241].

In the Zutphen Elderly Study, cognitive functioning and decline over three years were assessed in a cohort of 476 men aged 69-89y using the Mini-Mental State Examination (MMSE). Findings indicated that high linoleic acid intake (the main n-6 fatty acid in the diet) was associated with cognitive impairment, even after controlling for age, education, cigarette smoking, alcohol consumption and energy intake (OR = 1.76, 95% CI: 1.04-3.01, comparing highest to lowest tertile). However, there was no distinctive association for n-3 fatty acids. Nevertheless, total fish consumption was suggestive of a protective effect, even though it did not reach significance [236].

Another larger cohort study–The Rotterdam Study– recruited 5,386 non-institutionalized participants, aged 55 + y at baseline, who had normal cognition and assessed their complete dietary intake with a semi-quantitative food-frequency questionnaire. After an average 2.1y of follow-up, lower risk of incident dementia and AD was found among fish consumers and therefore among those with higher intake of n-3 fatty acids (RR = 0.3; 95% CI: 0.1-0.9) [237]. However, when the study was conducted later with a longer follow-up (mean follow-up period of 6.0 years), it was concluded that high intake of total, saturated, trans fat, cholesterol and low intake of monounsatured fatty acids (MUFA), total PUFA, n-6 PUFA and n-3 PUFA were not associated with increased risk of dementia or its subtypes [238].

A cross-sectional study of 1,613 subjects aged 45–70 years that examined the association between fatty acid and fish intake with cognitive function, found that the risk of cognitive impairment was reduced with increased consumption of fatty fish and marine n-3 PUFA. per Standard Deviation (SD) increased intake, the ORs were 0.81 (95% CI: 0.66, 1.00) and 0.72 (95% CI: 0.57, 0.90), respectively [240]. Another recent study using the Athersclerosis Risk in Communities (ARIC) cohort data suggested that dietary intake of n-3 fatty acids (mainly DHA + EPA) reduced the risk of cognitive decline in verbal fluency but not other cognitive domains (i.e. delayed word recall and psychomotor speed). This protective effect was particularly strong among hypertensive subjects [246]. The potentially protective effect of dietary n-3 fatty acid was also reported in several other large epidemiological studies [242,247,249,253,254,256-258], but not in others [250-252].

In sum, 7 out of the 18 (39%) selected cohort studies linking n-3 fatty acids to the various cognitive outcomes found the relationship in the hypothesized direction in the entire population that was studied and for most outcomes of interest [237,239,247,248,250,253,257], while 1 found this relationship for some outcomes but not others [245], 4 detected it for a sub-group of the total population [242,243,246,258], and 6 found no association [236,238,241,250-252] (See Table 2 and Figure 2A). In addition, all of the 5 (100%) cross-sectional studies that were selected found an association in the hypothesized direction for the entire study population and for most outcomes of interest [240,244,249,254,255] (See Table 2 and Figure 2B).