Chapter  134:  B Vitamins and Berries and Age-Related Neurodegenerative Disorders

Overall Effects

Thiamine (B1). The overall findings of the effects of thiamine deficiency are summarized in Figure 1. Of the four studies (in five publications) using thiamine-deficient models, two examined the rats' clinical status and/or histopathology after 44 days or 35 weeks of thiamine-depleted diets, and three examined rats' performance in neurocognitive tests (Table 2).³⁴–^{36, 38, 39} The two studies that examined the rats' clinical status and/or histopathology found that a diet without thiamine significantly damaged the rats' brains and/or caused serious neurological pathology, including death. In the other two studies that examined the effects of thiamine deficiency on rats' performance in neurocognitive tests, the deficient regimens varied widely. Specifically, Terasawa 1999 fed the rats a low-thiamine diet (30 mg per 100 g diet) for approximately 42 days, and Ciccia 2000 fed the rats a vitamin-fortified chow in combination with a thrice-weekly thiamine injection (1 mg per kg body weight) for three episodes of thiamine deficiency. Each episode lasted about 4.5 weeks. Terasawa 1999 found rats fed thiamine-deficient diets had significantly impaired performances in the neurocognitive tests when compared to the control rats. Ciccia 2000 found that rats with three separate episodes of thiamine deficiency had significantly impaired performance in two of eight neurocognitive tests.

Pyridoxine (B6). Two studies used a vitamin B6-supplementation model in mice or rats and both studies examined different doses of B6 supplementation on performances in neurocognitive tests (Table 3).^{41, 42} No significant effects of B6 supplementation were found for rat learning or cognitive function. Dietary B6 supplementations showed some positive effect on motor function or behavior, although the effects were not consistent across the two studies. The studies did not find a linear dose-response relationship for the effects; however, the results suggested worse motor function was associated with a lower dose of dietary B6 supplementation.

Cyanocobalamin (B12). One study examined the effects of low-dose vitamin B12 supplementation on spontaneous movements and performance in the Morris Water Maze in rats with nucleus basalis magnocellularis (NBM) lesions (an animal model that mimics some of the cholinergic hypofunction and memory loss associated with AD; Table 4).³⁷ The study found that low dose (1 mg per kg diet) vitamin B12 supplementation alone had no significant effect on spontaneous movements and did not improve memory in rats with NBM lesions.

Folate. Three studies examined the effects of folate deficiency on animal performance in neurocognitive tests and on brain neurotransmitters or histopathology; two used a normal animal model, one used a PD model and one used an AD model (Table 5).^{26, 27, 40} Kim 2002 conducted a electron microscopic experiment comparing the cerebrocortical microvascular wall in brain cross sections with rats fed a diet without folate to that of rats fed a diet with folate (4 mg per kg diet). After 8 weeks of dietary treatments, a degenerative appearance of the cerebrocortical microvascular wall occurred in rats fed folate-deficient diet, while rats fed a diet with folate had a normal cerebral capillary wall. Duan 2002 tested the effects of folate deficiency using a rat model of PD. After 3 months of dietary treatments, both experiment and control animals received two intraperitoneal injections of 20 mg 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) per kilogram body weight. The MPTP toxin induced PD-like pathology and behavioral symptoms. In contrast to mice on the control diet that were resistant to the sub-toxic dose of MPTP, mice maintained on the folate-deficient diet exhibited profound motor dysfunction as indicated by a decrease in the period of time in which they were able to maintain themselves on the rotarod and by an increased number of falls. However, folate deficiency alone did not impair rats' performance on rotarod tests. Kruman 2002 examined the effects of folate deficiency on brain histology findings in amyloid precursor protein (APP) mutant mice. These mice developed age-dependent deposition of amyloid β-peptide in their brains. There was no significant difference in the levels of amyloid β deposition between APP mutant mice on the folate-deficient diet and those on the control diet. However, the analyses in the CA3 region of the hippocampus revealed a highly significant 20 percent loss of neurons in APP mutant mice on the folate-deficient diet compared with those on the control diets. These results suggested that folate deficiency renders hippocampal CA3 neurons in APP mutant mice vulnerable to death by a mechanism that does not involve increased amyloid β production or deposition.

Summary. Of the four studies that used thiamine-deficient models, two examined the rats' clinical status and/or histopathology after thiamine-depletion diets and all found that thiamine-depletion diets significantly damaged brain and/or cause serious neurological pathology, including death. The remaining two studies that examined rat performance in neurocognitive tests found that thiamine deficiency had significantly impaired performance in some neurocognitive tests.

Two studies used a B6-supplementation model in mice or rats to examine the effects of B6 treatments on performance in neurocognitive tests. Both studies that used the B6-supplementation model examined various doses of B6 supplementation on animal performance in neurocognitive tests. No significant effects of B6 supplementation were found on rats learning or cognitive function. Dietary B6 supplementations showed some positive effect on animal motor function or behavior, although the effects were not consistent across the two studies and did not show a linear dose-response relationship.

One study showed that low dose (1 mg per kg diet) vitamin B12 supplementation alone had no significant effect on spontaneous movements and did not improve memory in rats with NBM lesions.

Of the three studies that examined the effects of folate deficiency on animal performance in neurocognitive tests and brain neurotransmitters or histopathology, two used a normal animal model, one each used a PD model and an AD model. Results from the normal animal model showed a degenerative appearance of the cerebrocortical microvascular wall was shown in rats fed a folate-deficient diet for 8 weeks, but 3-month folate-deficient diet did not impair rats' performance on rotarod tests. The one study that tested the effects of folate deficiency using a rat model of PD found that mice which had been maintained on the folate-deficient diet exhibited profound motor dysfunction induced by MPTP, in contrast to mice on the control diet that were resistant to the sub-toxic dose of MPTP. The results from the one study that used an AD model to examine the effects of folate deficiency on brain histology findings in APP mutant mice suggested that folate deficiency renders hippocampal CA3 neurons in APP mutant mice vulnerable to death by a mechanism that dose not involve increased amyloid β production or deposition.

We found no study that examined the effects of riboflavin (B2) or mixed B vitamin treatments on outcomes of interest.

Overall, research has shown that there were negative effects of thiamine and folate deficiency or deprivation on animal's clinical status and/or histopathology, although not all deficient animals had worse performance in neurocognitive tests. Studies have found some positive effects of the supplementations of B6, B12, and folate on animal's performance in neurocognitive tests, but studies did not show a dose-response relationship. Only folate deficiency was examined in animal models of AD and PD; the results showed a synergistic effect with both PD and AD pathology.

Summary Findings From Studies Using Immature Animals. Eight studies in nine publications were not included in detail in this review because immature rodents were used.⁴³–⁵¹ In these studies, young rodents were treated for a short period time and tested before their growth ceased. Though these studies are not applicable to questions concerning “age-related neurodegenerative” changes, it is of interest how these studies compare to those with more appropriate models. Of the eight studies, three used thiamine-deficient models, two used B6-deficient models, two used a B12-supplementation model, and one used folate-deficient model. The results from these studies were similar to those summarized in this report. Detailed data can be found in the evidence tables.^*

Overall Effects

Thiamine (B1) deficiency. Warnock 1968 examined pyruvate metabolism and the differences in pyruvate transport across BBB, comparing normal rats to thiamine-deficient rats.58 The results showed that labeled pyruvate entered the brain directly in adult thiamine-deficient animals, while it did not directly enter the brain of normal adult animals. This indicated that selective transport across BBB was not functioning in a normal fashion in thiamine-deficient rats. Robertson 1971 evaluated the presence or absence of extravascular fluorescence bound to intraperitoneally administered bovine albumin in relation to the severity of BBB lesions.⁶⁰ They found that extravascular fluorescence was present in one of 24 rat brains from animals with early stage of thiamine deficiency (as indicated histologically by slight edema, or more marked spongy reticulation, frequently accompanied by vascular congestion). Extravascular fluorescence was present in 12 of 22 rat brains from animals with late stage of thiamine deficiency (as indicated histologically by hemorrhages, tissue degradation and neuronal fallout). There was no extravascular fluorescence seen in rat brains from control animals. These results suggested that BBB is intact with respect to albumin in the early lesions of thiamine deficiency. Thus intracellular edema associated with early deficiency results from a defect in cell membrane transport rather than a vascular leak of the inflammatory type across BBB. A follow-up study by Manz 1972 employed a similar protocol to further define the nature and sequence of permeability changes of the BBB, using horseradish peroxidase (a low molecular weight protein marker).⁵⁹ It was found that control rats and rats with early stages of thiamine deficiency had “qualitatively and quantitatively” the same pattern of peroxidase granules in phagocytes. None of the rats with early stage of thiamine deficiency had parenchymatous infiltration, while 21 of 30 rats with late stage thiamine deficiency did. Furthermore, control rats and rats with early stage of thiamine deficiency were devoid of peroxidase in the vascular basement membrane and the neural parenchyma. Although the interendothelial junctional complexes were morphologically intact in rats with late stage of thiamine deficiency, reaction product was deposited in the contraluminal side basement membrane zone of intercellular gaps. These results suggested that BBB damage seen in later stages corresponds to damage seen from cold-injury edema and other models of cerebral edema. The leakage of BBB appears to be predominantly through the mechanism of pinocytosis (introduction of fluids into a cell by invagination of the cell membrane, followed by formation of vesicles within the cells), not disruption of interendothelial junctions.

Folate supplementation. Lee et al. conducted a series of studies to examine the effects of dietary folate supplementation on cerebral endothelial function and cerebral vascular damage induced by hyperhomocysteinemia in vivo. Before allocation to dietary intervention, all rats were fed a diet with added homocysteine (3.0 g per kg diet) for 2 weeks to induce hyperhomocysteinemia along with the hyperhomocysteinemia-induced cerebrovascular endothelial dysfunction. The first study found that 2 weeks of dietary folate supplementation significantly ameliorated the hyperhomocysteinemia-induced cerebrovascular endothelial dysfunction, characterized by reduced endothelial nitric oxide synthase (eNOS) activity and glucose transporter protein (GLUT-1) activity. Specifically, the level of brain eNOS protein expression increased by 44 percent (P=0.04) and the GLUT-1 level increased by 27 percent (P=0.04), in the comparison of rats fed an additional 8 mg per kg diet of folate to those on homocysteine diet. However, an unexpected result was observed for the cerebral content of the vascular cell adhesion molecule (VCAM-1). Rats fed an additional 8 mg per kg diet of folate had a 43 percent (P=0.04) decrease in VCAM-1 level. A subsequent study evaluated the effects of 8 weeks of dietary folate supplementation on cerebral vascular damage induced by hyperhomocysteinemia in vivo, in particular, investigating the structural features of the cerebral vasculature by electron microscopy. Consistent with the results from the first study, folate supplementation significantly increased the cerebral GLUT-1 protein, which had been decreased by a homocysteine diet. In the folate supplemented group, damaged vessels, annihilation of cell organelles, degeneration of mitochondrial bilayer, and perivascular detachment were also observed, although the damage of the cerebral vasculature was described as “more serious” in rats fed homocysteine-supplemented diet. In addition, dietary supplementation with folate for 8 weeks significantly reduced the percentage of damaged vessels.

Summary. Three studies examined BBB transport and permeability showed an abnormal selective transport of pyruvate across BBB in adult thiamine-deficient animals and intracellular edema associated with early deficiency results from a defect in cell membrane transport rather than a vascular leak of the inflammatory type across BBB. The leakage of BBB appears to be predominantly through the mechanism of pinocytosis, not disruption of interendothelial junctions. Two studies examined the effects of dietary folate supplementation on cerebral endothelial function and cerebral vascular damage induced by hyperhomocysteinemia in vivo. The results suggest that folate supplementation may ameliorate the hyperhomocysteinemia-induced cerebrovascular endothelial dysfunction and reduce cerebrovascular damage induced by hyperhomocysteinemia.

Thiamine (Vitamin B1) Intervention Studies

Study Descriptions. Three randomized controlled trials (RCTs),⁶¹–⁶³ one non-randomized comparative trial (N-RCT), ⁶² and one uncontrolled cohort study ⁶⁴ reported data on thiamine supplementation among patients with probable or possible AD in four articles. All studies included subjects whose mean ages were greater than 70 years, and who met standard criteria for diagnosis of probable or possible dementia. Their average mini-mental status examination (MMSE) scores ranged from 14 to 18, indicating moderate to mild dementia. All RCTs tested three divided doses of 3 g per day thiamine, the N-RCT tested a maximum of 8 g per day thiamine and the single arm intervention trial tested a dose of 750 mg of a thiamine derivative. Thus, most studies used high-dose interventions compared to the US Recommended Daily Allowance (RDA) of 1.1 to 1.2 mg daily. The primary outcome for all studies was cognitive function (measured with MMSE or Alzheimer's Disease Assessment Scale [ADAS]). In addition, two of the studies evaluated outcomes on behavior and/or emotion.61,64 All studies had small sample sizes (fewer than 25 subjects). All studies were assessed to poor quality (C). Two studies have moderate applicability and three narrow applicability.

Overall Effect. Three of the five studies reported statistically significant effects in cognitive function after intervention with thiamine for short durations (1 to 3 months).^{61, 62, 64} Meador 1993 reported statistically significant improvement of cognitive score (ADAS) in the initial months with reduced deterioration at 11 to 13 months, suggesting some beneficial effect of high dose of thiamine in decreasing the rate of progression of AD. This study also reported clinically significant effect of thiamine supplementation in the majority of subjects at dosages above 4 g per day.

Among these, Blass 1988 was a randomized crossover trial. The study did not document a wash out phase and compared thiamine to the intervention of niacinamide 750 mg as an active placebo. The follow-up study to Blass 1988, by Nolan 1991, was an RCT with parallel design that compared thiamine treatment to lactose placebo. It found no significant effects on cognitive function after treating to the same dose of thiamine for one year.

Meador 1993 reported two different experiments, and tested higher doses of thiamine (greater than 3 g per day). The first was a crossover RCT that compared thiamine to placebo without a washout phase between treatments. The second was a non-randomized comparative trial, which gave AD subjects sequentially higher doses of thiamine followed by placebo, and reported statistically significant improvements in the cognitive score (ADAS) at 3 dosages of 4 g per day or more thiamine. Of note this thiamine supplementation study reported improvement in the scores of Clinical Global Impression of Change (CGIC), which is a physician rated assessment of overall change from the baseline. However the study reported results among fewer than 10 subjects for dosages of 7 g or more between 8 to 13 months.

Mimori 1996 was an uncontrolled cohort study that used fursultiamine, a thiamine derivative that is easily converted into an active form of thiamine in the body. They evaluated nine people with AD and normal mean levels of blood thiamine. The study reported a statistically significant improvement in cognitive function (MMSE) and the emotional component of the Gottfries-Brane-Steen (GBS) score after the intervention with fursultiamine. Blood thiamine levels increased markedly after fursultiamine intervention.

Interactions and Covariates. In its second experiment, Meador 1993 tested a maximum dose of thiamine for 13 months. The first phase of the study the thiamine dose was incrementally increased each month to a maximum of 6 g per day. In the second phase a “best dose” (defined as achievement of best ADAS scores) was used, and in the third phase a maximum dose of 8 g per day was achieved. However, after month 5, the results for the doses greater than 6 g of thiamine were reported only in a small subset of two to six subjects. High dose thiamine significantly reduced ADAS scores compared to baseline, but MMSE showed no significant change. No other interactions were reported in the studies. Only the uncontrolled cohort study reported blood levels of thiamine before intervention, and included subjects with normal blood levels. All the studies of thiamine intervention used tests of the same domain and measured global cognitive function as the outcome.

Summary. Five poor quality studies assessed in four articles reported data on the effect of thiamine intervention among people with probable or possible AD. Overall, two RCTs, and one cohort study reported improvement in cognitive function during short-term treatment of thiamine. However, without a comparable control group, it is difficult to assess the validity of the uncontrolled cohort study given the variable course of dementia over time. Two long-term studies failed to show any discernable differences in cognitive function as measured by MMSE. However the N-RCT that supplemented AD subjects with progressively higher doses of thiamine followed by placebo reported statistically significant improvement of cognitive score (ADAS) in the initial months with reduced deterioration at 11 to 13 months, suggesting some beneficial effect of high dose of thiamine intervention. This study also reported long-term clinical benefit from thiamine intervention.

Riboflavin (Vitamin B2) Intervention Studies

No prospective trial has evaluated the effect of vitamin B2 treatment on neurocognitive function.

Vitamin B6 Intervention Studies

Study Descriptions. Two RCTs assessed the effect of B6 intervention on cognitive function in humans.^{65, 66} They included 75 and 38 cognitively intact subjects in the intervention arm respectively. One RCT used 75 mg of B6 per day for 5 weeks while the other trial administered 20 mg of B6 per day for 12 weeks. Both doses are considerably higher than the US RDA dose of 1.3 to 1.7 mg daily. One study used digit symbol coding, vocabulary test, and digit span-backwards from WAIS III as well as the Stroop test for executive function and initial letter for verbal fluency. The second RCT applied the Associate Recognition Task and the Long Term Memory Storage to assess cognitive function. The trials were of moderate (B) and low (C) quality and had narrow and moderate applicability, respectively.

Overall Effect. Deijen 1992 reported a significant decrease in Long Term Memory Storage for the intervention group, which corresponds to memory improvement (P<0.03). However, no significant effect was recorded in the Associate Recognition Task for the intervention group in the same study. However, no formal comparison between intervention and placebo group was provided in the article and no information is given for the significance of the changes in the placebo group.

Interactions and Covariates. There is no adequate evidence to support any dose effect of B6 on the outcomes. There was no evidence across studies of differences in effect on tests of different cognitive domains. No other interactions were reported in the studies.

Summary. Two RCTs of moderate and low quality, with narrow and moderate applicability for cognitively intact populations investigated the effect of B6 intervention on cognitive function in humans. With treatment, a significant improvement was found with one cognitive function test. It is uncertain whether this change is of any clinical benefit. It is also unclear whether the changes with treatment were significantly different than changes in the control arm in the same study. No other significant effect was reported in the studies. Because of the very limited evidence no conclusions can be drawn for the effect of B6 on preventing cognitive function decline.

Cobalamin (Vitamin B12) Intervention Studies

Study Descriptions. Five RCTs, one N-RCT, and seven uncontrolled cohort studies assessed the effect of B12 intervention on cognitive function in humans.^{65, 67}–⁷⁸ Among the RCTs (Table 11) sample size ranged from 18 to 70 in the intervention arms. Two RCTs recruited cognitively intact participants with normal B12 while another two recruited cognitively intact subjects with low B12 levels. Low B12 levels was defined as 136 to 203 pg/mL in one RCT and less than 163 pg/mL in the other. The fifth RCT included subjects with cognitive impairment. The five RCTs used different doses and routes of B12 interventions ranging from 0.01 mg B12 per day orally to 1 mg 3 times per week B12, intramuscularly. Two trials used oral B12, the remainder intramuscular. B12 doses were higher than the US RDA of 2.4 μg per day. Almost all trials implemented more than one cognitive function test. MMSE was used by three trials, several parts of the Wechsler Adult Intelligence Scale (WAIS) by two trials while a number of miscellaneous tests were also reported (Table 12)., including the Cognitive Subscale of Cambridge Examination for Mental Disorders of the Elderly (CAMCOG), 12 word learning test - immediate and delayed - Stroop, initial letter from verbal fluency, memory quotient, lower and upper limit of retention span, visual and verbal memory, verbal and performance Intelligence Quotient (IQ). RCTs were of moderate to low quality (1A, 2B, 2C) and had narrow to moderate applicability.

Among the N-RCT and the cohort studies (Table 12), sample size ranged between 14 and 56. Three studies recruited demented participants, regardless of specific diagnosis. Two selected only participants with low serum B12 levels defined as less than 200 pg/mL in one trial and less that 300 pg/mL in the other trial. There were three additional trials that included subjects with low B12 levels (less than approximately 200 pg/mL); one recruited cognitively impaired individuals with B12 less than 203 pg/mL and the other included cognitively intact persons with B12 values equal or less than 203 pg/mL while the third one evaluated patients with AD or other dementias and B12 values less than 190 pg/mL. Two other studies included subjects with AD, regardless of B12 status. As with the RCTs, cobalamin doses were considerably higher than the RDA. Almost all trials implemented more than one cognitive function test (Table 12). MMSE was used by five trials, Mattis' Dementia Rating Scale (DRS) by two studies, WAIS by one trial, GBS by two trials while a number of miscellaneous tests were also reported, including CAMCOG, Interview for Deterioration in Daily living activities in Dementia (IDDD) -Initiative, IDDD-Performance, Revised Memory and Behavioral Problems (RMBPC) -Memory, RMBPC-Disruptive behavior, delayed verbal word learning test, verbal fluency, similarities, trail making test, Rivermead behavioral face recognition test, Hasegawa's Dementia Rating Scale (HDS), and Consortium to Establish a Registry for Alzheimer's Disease (CERAD). The N-RCT was of low quality and had moderate applicability. Cohort studies were (by definition) of low quality (C) and had narrow to moderate applicability.

Overall Effect. Only one RCT (Hvas 2004) found a significant difference in effect on 12 word learning test at 15 minutes between the B12 and the control groups among cognitively impaired subjects. With other tests, some significant changes were found either with or without B12 supplementation, but these changes were not significantly different than each other. Kwok 1998, which included cognitively intact participants with low B12 (<163 pg/mL) revealed a significant improvement for B12 group in performance IQ; however, this was not significantly higher than the change in the control group. No other significant effects were reported in the RCTs.

Among the seven uncontrolled cohort studies, Teunisse 1996 reported that MMSE score was significantly worse after treatment among demented individuals with low B12 levels (<200 pg/mL). Ikeda 1992 found a significant improvement in Mattis' DMR scale at 2 months among people with AD, which however, was not maintained 4 months after the end of treatment. In the same study, GBS and HDS were also significantly improved at the end of treatment, but their improvement was also not maintained 4 months later. Van Asselt 2001, in the N-RCT, found that after treatment, performance on the delayed recall of verbal word learning test, similarities and verbal fluency test was significantly improved among cognitively intact subjects with low B12 (<203 pg/mL). In contrast, Teunisse 1996 noted significant deterioration in several tests including IDDD-Initiative, IDDD-Performance, and RMBPC-Disruptive behavior. The other five cohort trials found no change in cognitive function after B12 treatment.

Interactions and Covariates. RCTs reached statistical significance less often than cohort studies. When they reported significant changes, RCTs usually implied an improvement of cognitive function. Significant changes for cohort studies were not consistent, some finding improvement, some decline of cognitive function scores. There is large heterogeneity among RCTs and cohort studies in terms of dose, route, and duration of treatment and it would be difficult to support any conclusion about a potential dose effect. However, the only significant changes in cognitive score were found when B12 was injected rather than given orally. Seal 2002 directly evaluated the effect of B12 oral supplementation in two intervention arms, one receiving double the dose of the other, and compared these groups with placebo. No significant change was found when MMSE score differences of the three groups were compared. In the cohort study by Mitsuyama 1988 five of the 14 demented participants were orally supplemented with 2 mg B12 daily while nine subjects had an additional B12 injection of 0.5 mg. Only those receiving B12 injections showed an improvement from baseline in GBS (P value not reported).

I to 2001 analyzed the results based on the severity of dementia of Alzheimer's type. Both subjects with questionable or mild AD and participants with moderate or severe AD had non-significant improvement in MMSE after treatment. Abyad 2002 analyzed the results according to treatment duration for dementia before the trial. Both subjects with short-term treatment and subjects with long-term treatment improved MMSE; short-term treatment participants had a significant increase. Martin 1992 evaluated separately patients with cognitive impairment of long duration and patients with cognitive impairment of short duration. Only the subjects with disease of short duration showed a significant improvement in Mattis' DRS.

There was no evidence across studies of differences in effect on tests of different cognitive domains.

Summary. Five RCTs of narrow to moderate quality, one non-randomized comparative study, and seven cohort studies assessed the effect of vitamin B12 intervention on cognitive function in humans. All studies had narrow to moderate applicability. They evaluated populations that included normal participants, subjects with cognitive impairment, dementia, or AD. Several studies recruited only individuals with low B12 levels, however the definition for low B12 levels varied. There was large heterogeneity among studies in terms of dose and route of intervention as well as the cognitive function assessment instrument. Although several of the studies showed small changes in cognitive function, few reached statistical significance. Across studies that assessed similar populations after implementing the same test, results were conflicting. Several cohort studies revealed significant improvement while a smaller number of cohorts reported a significant decline in scores for cognitive function. However, the interpretation of these studies is difficult because they analyzed subjects who may have had variable courses of dementia over time, without using a control group for comparison. Vitamin B12 was given intramuscularly in the only RCT that found a significant effect in the treatment group compared to controls. Similarly, only cohort studies that used intravenous or intramuscular vitamin B12 reported a significant effect on cognitive function scores. However, given the lack of data directly comparing oral and injected routes of vitamin B12 and the paucity of controlled trials limits any conclusions regarding the utility of different routes of administration. There was very limited evidence whether other covariates may interact with B12 supplementation. Most studies did not take into consideration potential factors such as disease duration that may interfere with B12 effect.

Folate Intervention Studies

Study Descriptions. Three RCTs^{65, 79, 80} and two uncontrolled cohort studies^{81, 82} reported data on folate supplementation and the effect on cognitive function or therapeutic benefit. Two studies were conducted among subjects with dementia, one among cognitive impaired, one among normal subjects, and the fifth among those with PD. All but one study included subjects whose mean ages were 60 years and above; the remaining study did not document mean age. The studies tested various doses of folic acid or folate ranging from 0.75 to 20 mg daily. These doses are all substantially larger than the US RDA of 0.4 to 1 mg per day. Study durations ranged from 5 to 17 weeks. The primary outcome for three studies was cognitive function measured with WAIS-R and/or other cognitive tests. One study also evaluated therapeutic benefit. About 130 subjects were tested in five studies. All studies were of moderate to poor quality (2 B, 3 C). Three studies have moderate, and two narrow applicability.

Overall Effect. Among the five studies that measured cognitive function in people with dementia, cognitive impairment, or no disease, one RCT found a trend towards worsening of neuropsychological scores in the folic acid treatment group, suggesting a negative benefit of high dose folate among subjects with dementia. Two other studies, one RCT and one uncontrolled cohort studies found statistically significant improvement with folic acid or folate treatment compared with placebo among demented, cognitive impaired, and normal subjects, but did not report any clinical benefit. The study of patients with PD found no therapeutic benefit.

Fioravanti 1997 compared the effects of folic acid intervention with placebo among cognitively impaired subjects who had serum folate levels below 3 ng/mL. Cognitive function was assessed using six components of the Randt Memory Test (RMT). Compared to baseline the folic acid treatment group showed significant improvement in Attention Efficiency score, one of the six components of RMT, after 60 days treatment. However comparison between folic acid and placebo groups showed significant changes in the actively treated group with improvement in four out of six cognitive components of RMT.

Bryan 2002 was a double blind, placebo controlled randomized trial conducted to assess the effect of vitamin supplementation and dietary intake among normal women. The study utilized a mixed factorial design with four treatment groups (B6, B12, folate, and placebo) and three age groups (younger, middle-aged, and older) to assess the effect on cognitive performance. Here we summarize the results from the subgroup of older women who were treated with a small dose of folate (0.75 mg). In post hoc comparisons of two measures of memory, Rey Auditory-Verbal Learning Test (RAVLT) immediate recall and recognition list, and verbal ability (verbal fluency-initial letter), older women in the folate treatment group identified significantly more words when compared to the placebo group. No significant effect was observed in other tests.

Sommer 2003, in a very small double blind RCT, compared 20 mg per day of folic acid to placebo in seven subjects who fulfilled the standard diagnostic criteria for dementia. There were small trends towards a negative effect on cognitive abilities with folic acid treatment in two of the cognitive tests, the Associated Learning subtest that measures verbal learning, and Trails B that measures perpetual motor speed.

Rapin 1988 recruited subjects with dementia and low red blood cell (RBC) folate levels. This study was an uncontrolled cohort study that treated dementia subjects with 50 mg per week of folinic acid for about 17 weeks. Treated subjects had a significantly improved performance in five of 16 cognitive tests. They also reported improved feeling of well-being.

The single study of patients with PD, by McGeer 1972, was a cohort study of 15 mg per day of folic acid among 18 subjects. The study found that the folic acid intervention provided slight to no therapeutic benefit assessed by subjective or objective change in PD symptoms.

Interactions and Covariates. The intervention trials did not provide adequate evidence to support any dose effect of folate on the outcomes. However, Fioravanti 1997 found that the cognitive improvement after folate intervention was correlated in a linear fashion with the low levels of folate at baseline. One RCT⁸⁰ and one uncontrolled study⁸² studied the effect of folic acid intervention on cognitive function among subjects with low serum or RBC folate levels, and one RCT among those with normal folate levels.⁷⁹. Two other studies did not provide data on blood folate levels before or after intervention. One RCT and one uncontrolled cohort study reported significant effects in the same cognitive domain.

Summary. A total of five studies of moderate to poor quality reported data on the effect of folate intervention among normal people or those with dementia, cognitive impairment, or PD. Overall, one RCT among subjects with dementia and normal folate levels found a trend towards worse performance in the cognitive function scores with folic acid intervention. Two studies, one RCT and one uncontrolled cohort study found statistically significant improvements in the cognitive scores in the actively treated groups among demented, cognitive impaired, and normal subjects and the last found no benefit among PD subjects. However, interpretation of the cohort studies is difficult without a comparable control group given the variable course of dementia over time. None of the studies provided data on clinically significant effect after the vitamin intervention.

Combined B Vitamins Intervention Studies

Study Descriptions. Six studies, including 3 RCTs⁸³–⁸⁵ and 3 uncontrolled cohort studies⁸⁶–⁸⁸ assessed the effects of a combination of B vitamins as interventions on cognitive function in elderly subjects. Four trials included subjects with AD and/or mixed dementia,^{83, 85, 86, 88} one with cognitive impairment,⁸⁷ and one without dementia.⁸⁴ All studies used different daily doses of various B vitamins in the ranges of 0.8 to 15 mg folic acid or folate, 3 to 80 mg B6, and 0.1 to 2 mg B12; all substantially over the US RDA. Three used a combination of folic acid, B6, and B12, two used folic acid and B12, and one used folate and B12. All but one study treated subjects with oral vitamin supplementation; Shaw 1971 used a combination of B12 injection and oral folate. All studies included subjects whose mean ages were greater than 70 years. The primary outcome for all studies was cognitive function. About 470 subjects were tested in six studies. All studies were of moderate to poor quality (2 B, 4 C). One study has broad applicability, four moderate, and two narrow applicability.

Overall Effect. Five of the six studies found no significant change in cognitive function, generally measured with MMSE or WAIS, after combination B vitamin supplementation.

The single study to find a significant improvement, Nilsson 2001, was an uncontrolled cohort study. This study found significant changes in the cognitive score performance among subjects with mild-moderate dementia and elevated plasma homocysteine concentration with a significant global clinical improvement after vitamin intervention. There was a four point increase in the mean MMSE score from a baseline score of 17. However the same study failed to show any improvement among patients with severe dementia or those with mild-moderate dementia and normal plasma homocysteine levels with a mean baseline score of 21 after combined vitamin intervention.

Lewerin 2005, a RCT, compared moderate doses of vitamins supplementation to placebo among ambulatory normal subjects, and measured cognitive function with a battery of nine tests. The placebo arm performed better in three of the tests compared to the actively treated group. Two other RCTs reported no significant changes in cognitive function in either intervention arms after vitamin supplementation. The remaining two cohort studies, Aisen 2003 and Lehmann 2003, reported no observable differences after treatment.

Interactions and Covariates. There was no dose related responses discussed in the studies. Only Nilsson 2001, an uncontrolled cohort study reported a difference in cognitive function with relation to severity of dementia, and with the levels of plasma homocysteine (<2.69 versus >2.69 mg/L) after combined B vitamin intervention. This study also reported a significant improvement in alertness, orientation in time and space, recent memory and fewer clinical fluctuations among subjects with dementia after combined vitamin intervention. One RCT⁸⁵ used a combination of B12 injection and oral folate among senile dementia with high and low RBC folate levels; remainder of the studies evaluated combined vitamin intervention among those with normal blood vitamin levels. There was no evidence across studies of differences in effect on tests of different cognitive domains.

Summary. Six studies of moderate to poor quality reported data on the effect of combined B vitamin intervention among normal people, cognitively impaired, and those with AD and/or mixed dementia. The three RCTs found no benefit in the actively treated arm compared to placebo. Only one of the uncontrolled cohort studies found both statistically and clinically significant large benefit. However, interpretation of the cohort studies is difficult without a comparable control group given the variable course of dementia over time. The two other uncontrolled cohort studies reported no benefit after intervention.

Longitudinal Studies

Study Descriptions. One prospective nested case-control study, three prospective cohort studies, and one retrospective case-control study examined the association between dietary intake levels of B vitamins and neurodegenerative diseases or cognitive function (Tables 15–17). Each study associated the B vitamin intake levels with different outcomes, including the risk of newly developed PD cases, rates of cognitive change per year, follow-up cognitive function examined by various cognitive tests, and a diagnosis of probable AD. Two studies were of good quality, one study was of moderate quality, and the other two studies were of poor quality.

Overall Effects. Chen 2004 conducted a nested case-control study to investigate whether intake of folate or related B vitamins involved in folate and homocysteine metabolism was associated with PD risk, using two large cohorts in the US - the Health Professionals Follow-up study (1986-2000) and the Nurse's Health Study (1980-1998).⁸⁹ The two cohorts were analyzed separately and then pooled analyses were also performed. Participants' dietary intakes were assessed by a food frequency questionnaire during the previous 12 months. It was found that controlling for age, smoking, alcohol consumption, caffeine intake and lactose intake, there were no significant associations found between the baseline intake of folate, vitamin B6, or vitamin B12 and relative risk of PD in either study. Several sensitivity analyses were also performed for different levels of folate intake. Individuals at either the low end (≤200 μg/day) or the high end (>1000 μg/day) of folate intake had a PD risk similar to that of people with normal folate intake, controlling for various possible confounders. Furthermore, in a separate analysis, supplemental intake of folate, vitamin B6, or vitamin B12 was also not related to the risk of PD. Compared with non-supplemented participants, individuals whose supplemental folate intake was more than 400 μg per day had a pooled RR of 1.0.

Morris 2005 conducted a prospective longitudinal cohort study to examine the associations of dietary folate and vitamin B12 with 6-year cognitive change in the participants of the Chicago Health and Aging Project.⁹⁸ Change in cognitive function measured at baseline, 3-year and 6-year follow-ups, using the average z score of four tests: the East Boston Tests of immediate and delayed recall, the Mini-Mental State Examination, and the Symbol Digit Modalities Test. The median dietary intake of folate ranged from 175 to 382 μg/day for first and last quintile respectively. The median intake of folate, from food and supplements, ranged from 186 to 742 μg/day for first and last quintile respectively. At baseline, it was found that persons with high intake of total folate (from food and supplements) tended to have a more favorable risk profile for cognitive change (more years of education, higher baseline cognitive scores, and greater consumption of vitamin E and vitamin C) than persons with low intake. After a median follow-up of 5.5 years, unexpectedly, high folate intake from food sources and/or supplements was associated with a faster rate of cognitive decline in a mixed models adjusted for multiple risk factors. Further sensitivity analyses showed no change in the effect estimates after restricting the analyses in persons who reported poor health status or with low baseline cognitive scores (bottom 15% of the distribution) at baseline. Intake of vitamin B12, with or without vitamin supplementation, was not significantly associated with cognitive change in the multivariate model or with adjustment for folate intake (data not shown).

Part of Tucker 2005 study examined the association between dietary B6, B12 and folate intakes and 3-year changes in cognitive measures in the Veterans Affairs Normative Aging, a longitudinal cohort consists of 321 aging men at baseline.⁹⁶ Cognitive function was assessed with the MMSE and on the basis of measures of memory, verbal fluency, and constructional praxis, which were adapted from the revised WAIS and the CERAD batteries at 2 time points. Improbable dietary intakes (total energy >16.75 or <2.51 MJ) were excluded from further analysis. Over a mean 3-year follow-up, changes in constructional praxis measured by spatial copying were significantly associated with dietary folate, B6 and B12. The mean dietary folate, B6 and B12 was 440 μg/day, 3.98 mg/day and 9.57 μg/day, respectively. Dietary folate remained independently protective against a decline in spatial copying score after adjustment for other vitamins and for plasma homocysteine. Dietary folate was also protective against a decline in verbal fluency. There was no other significant association found between dietary folate, B6, or B12 and changes in other cognitive measures. The major limitation of these analyses is that dependent biases might occur, since the dietary intakes and cognitive measures were both assessed by self-administered instruments.

Deijen 2003 conducted a prospective cohort study to examine the relation between nutritional intake and daily functioning in elderly people being evaluated in a psychogeriatric nursing home.⁹⁰ During the study 60 percent of the dropouts “became ill” compared to 34 percent of the subjects who completed the study. Participants' dietary intakes were assessed by a combination of a 3-day record and weighing-back methods at baseline, weeks 8, 16 and 24, recorded by nurses. An analysis was performed to test for associations with change in ZIG-scores. Two experimental groups were formed based on high (>1.0 mg/day) and low intakes (≤1.0 mg/day) of B6 compared to their median intakes at baseline. There were no interactions between intake groups and week, indicating that the high and low intake groups had the same pattern of ZIG-scores across the 6-month experimental period. There was deterioration in cognitive, physical and social functioning with time. However, this study had several limitations. Dropouts had worse health status than completers; it is unclear whether the nurses who assessed exposure and outcomes were the same; and the restriction to nursing home residents limits applicability.

Mizrahi 2003 conducted a retrospective case-control study compared patients with AD to healthy controls. (Mizrahi, 2003 2188 /id} Both proxy and surrogate respondents of cases and controls were asked to recall their food consumptions using a food frequency questionnaire during three age periods: 20 to 39, 40 to 59 and 60 or more years old. It was found that those with AD had lower mean dietary vitamin B6 and folate intake compared to controls in the over 60 years of age period (P=0.05 and 0.01, respectively), but not in younger age periods. No statistically significant correlations were found between homocysteine levels and dietary vitamin B6 and folate intake in the three age periods for those with AD and controls. This study was deemed to be of poor quality due to serious recall biases. Specifically, controls might have had more accurate recalls than cases, because cases' intakes were estimated by proxy or surrogate respondents, and all respondents might remember the food consumption during more recent age period better than distant age periods. Finally, the recall periods were too long to obtain accurate food consumption data.

Summary. No significant associations were found between dietary intakes of B6 or B12 and PD, AD, cognitive functioning, or cognitive decline across three studies. One additional study found higher dietary intakes of B6, B12, and folate were associated with improvements in some, but not all, cognitive function measures. In three separate studies, folate intake was not associated with PD or AD; however, in one study, higher folate intake from food sources and/or supplements was associated with a faster rate of cognitive decline after adjusting for multiple risk factors.

Cross-Sectional Studies

Study Descriptions. Five cross-sectional studies examined the association between the dietary intake levels of B vitamins and cognitive function or the risk of age-related neurodegenerative disease.^{65, 92}–⁹⁵ Of the five studies, three were of fair quality and two were of poor quality.

Thiamine (Vitamin B1) Level Association Studies

Study Descriptions. Eight cross-sectional studies examined the association of thiamine levels with diagnosis or cognitive function.⁹⁹–¹⁰⁶ Four of these studies measured serum, plasma, and RBC levels of thiamine among AD, non-AD, PD, and cognitively impaired subjects (Table 23). Four other studies reported the plasma, cerebrospinal fluid (CSF), and brain levels of thiamine diphosphate (TDP), the active form of thiamine in the brain (Table 24). The sample size for the cross-sectional studies ranged from 34 to 290. The two studies by Mastrogiacoma 1996 and Molina 2002, which reported brain or CSF levels of thiamine derivatives (free thiamine, total thiamine, thiamine monophosphate, thiamine diphosphate), compared the levels between AD and normal controls. Jimenez-Jimenez 1999 compared the CSF levels of thiamine derivatives between PD and controls. Snowdon 2000 compared the thiamine levels from stored blood within a subset of AD subjects who had histologically significant and non-significant lesions in the brain at autopsy. All studies were graded moderate to poor quality (3 B, 5 C). Three studies have broad applicability, two moderate, and three narrow applicability.

Overall Effect. Four studies evaluated plasma and/or RBC thiamine or thiamine pyrophosphate (TPP), two of which found significantly greater deficiency among AD subjects compared to PD or controls (Table 23). Four other studies evaluated thiamine or thiamine derivatives in CSF, plasma, and cerebral cortex (Table 24). Two of these reported significantly reduced levels of thiamine derivatives in plasma or cerebral cortex among AD subjects and one among subjects with PD.

Gold 1995 examined the thiamine levels in the plasma and RBCs and found significantly low plasma thiamine levels and a high prevalence of plasma thiamine deficiency, but not for RBC thiamine, among probable AD, compared to non-AD subjects. A subsequent publication by the same author (Gold 1998) evaluated thiamine levels in the plasma and RBC thiamine among AD and PD subjects, and reported similar significantly lower levels and higher prevalence of plasma thiamine among AD subjects. The plasma level of thiamine among PD subjects was normal. Assantachai 1997 estimated serum levels and reported a high, but non-significant prevalence of thiamine deficiency among cognitively impaired compared to normal elderly subjects. Scillepi 1984 reported normal mean thiamine levels among AD and non-AD subjects.

Jimenez-Jimenez 1999 compared CSF levels of thiamine and its phosphate esters among PD patients and normal controls. All thiamine derivatives in the CSF except free thiamine were normal among PD patients. Molina 2002 compared subjects with AD to normal controls. This study examined CSF and plasma levels of thiamine derivatives and reported significantly lower plasma levels of thiamine derivatives among AD subjects. Among a subset of AD patients annual mean decreases in MMSE score was assessed and there were no significant differences in mean MMSE scores with high versus low thiamine levels. Mastrogiacoma 1996 examined thiamine derivatives in autopsied cerebral cortex among AD and normal subjects. The mean levels of TDP in the temporal, parietal, occipital areas of cerebral cortex were slightly, but significantly reduced by 18 to 21 percent among AD subjects compared to normal subjects. The levels of other thiamine derivatives were normal among AD subjects. Snowdon 2000 described a negative, but non-significant correlation of serum thiamine levels with severity of cerebral cortex atrophy among AD subjects with histologically significant lesions. AD subjects with histologically non-significant lesions had a positive, but non-significant correlation of serum thiamine levels with severity of atrophy of the cerebral cortex.

Interactions and Covariates. Gold 1998 (Table 23) performed secondary analyses excluding four PD patients less than 60 years of age, matching the two groups for age. They found significant differences in both the plasma thiamine levels and the prevalence of thiamine deficiency (defined as plasma or RBC thiamine values below the range of normal for their respective age group) between AD, and PD subjects. Snowdon 2000 (Table 24) adjusted age as a potential confounder in the correlation analyses, but other studies did not adjust for potential confounders. No longitudinal studies evaluated the association of thiamine and cognitive function.

Summary. Overall, eight cross-sectional studies evaluated levels of thiamine among AD, cognitively impaired and PD patients. Three studies reported significantly reduced mean thiamine or TDP levels in the plasma and brain among AD subjects, and one reported similar reduction in mean levels of thiamine derivative among PD subjects. None of the studies that showed significant results adjusted for potential confounders. The remaining four studies found no differences between the investigated groups.

Riboflavin (Vitamin B2) Level Association Studies

Study Descriptions. Two studies ^{106, 107} reported cross-sectional data on riboflavin (B2) levels among AD, PD, and dementia subjects. One study compared blood levels of B2 in AD with control subjects, and the other compared plasma levels of B2 in PD with dementia subjects. Scileppi 1984 assessed blood levels of 12 vitamins among 55 AD subjects. The control group included a total of 58 normal, depressed, vascular and mixed dementia subjects. Coimbra 2003 assessed plasma levels of B2 among 31 PD subjects compared with 10 dementia subjects without stroke. These studies were assessed to be of poor quality (C) and moderate applicability.

Overall Effect. Scileppi 1984 reported no significant differences in the riboflavin levels between AD and the control groups. Coimbra 2003 found a mean plasma B2 level below normal among PD subjects, which was significantly lower than among subjects with dementia. Neither study adjusted for any confounders.

Summary. Two poor quality studies assessed the cross-sectional levels of riboflavin among AD, PD, dementia, and other control groups. One study reported no significant difference between AD and control subjects. The second reported lower plasma levels among PD subjects with a significant difference between PD and dementia subjects.

Vitamin B6 Level Association Studies

Study Descriptions. Ten studies examined the association of vitamin B6 serum levels with the diagnosis of dementia or cognitive impairment, and cognitive function, including three prospective longitudinal studies, one case control study, one retrospective cohort study, and five cross-sectional studies.^{18, 105}–¹¹¹ Pyridoxal-5′-phosphate (PLP), an active coenzyme form of B6, was used to estimate B6 serum levels in all but two studies; one did not report the method for estimating B6¹¹¹ and another study used a protozoological assay.¹⁰⁵ The longitudinal studies recruited 313 to 1,092 participants, the case control study 80 subjects while the sample size for the other studies ranged from 30 to 127. Studies had narrow to broad applicability. Two of the longitudinal studies, the case control study and the retrospective cohort were of moderate quality (B) while the third longitudinal study and the five cross-sectional studies were of low quality (C).

One longitudinal study recruited participants with normal cognitive function and investigated whether low baseline levels for vitamin B6 may be a risk factor for cognitive decline after about 3 years of follow-up (Table 26).¹¹² Cognitive decline was defined as 7 year cognitive change with at least a 9-point drop of the total cognitive score (the sum of all subtest scores). Another longitudinal study also recruited participants with normal cognitive function and tried to correlate cognitive function after almost 8 years of follow-up with the baseline levels of vitamin B6 (Table 26).⁹⁶ Cognitive function in this study was assessed by MMSE score, as well as by tests on construction praxis, language, working and recall memory. In the third longitudinal study, subjects with normal cognitive function were followed for a median of 8 years and new cases of AD or dementia were recorded (Table 26).¹⁸ The case control study compared the number of subjects with low PLP among patients with AD with that among individuals with intact cognitive function.¹⁰⁹ The retrospective cohort study correlated the number of lesions in autopsies of women with AD with B6 levels¹⁰⁵ while the rest of cross-sectional studies compared the mean B6 serum levels between different groups of participants, one comparing patients with AD, persons cognitively impaired but not demented, and normal participants; another study compared patients with AD and patients with other dementias; another comparing patients with AD and normal individuals; and the other two comparing patients with PD to patients with dementia, or to normal subjects (Table 27).

Overall Effect. Seshadri 2002, in the longitudinal study, reported that B6 serum levels were not correlated with the risk of AD or other dementias after adjusting for age, sex, and ApoE genotype; although specific results were not provided. Kado 2005 did not find any significant increase in the risk for cognitive decline for participants with very low vitamin B6 levels after adjusting for age, sex, education, baseline cognitive function, baseline physical function, and smoking. However, Tucker 2005, after adjusting for baseline cognitive measures, age, education, smoking, alcohol intake, BMI, diabetes, systolic blood pressure, time of second measure relative to folic acid fortification, time interval between the two cognitive measures, and serum creatinine, reported a significant correlation between higher levels of vitamin B6 at baseline and better performance in the figure copying test after 3 years of follow-up (β=0.38, P <0.05). No other significant association was described in the same study between vitamin B6 levels and the score of other cognitive tests.

None of the cross-sectional studies found any significant correlation between B6 levels and diagnosis of AD or number of AD lesions in brain autopsy, cognitive impairment, or dementia. In addition, two cross-sectional studies found no association between B6 level and diagnosis of PD, compared to either people with dementia or normal people.

Interactions and Covariates. Generally, results were non-significant irrespective of the study design. Woitalla 2004 examined the association of B6 levels and diagnosis of PD in patients with different MTHFR genotypes (CC, CT, TT). No significant difference was reported when the three genotype groups were compared to each other or to the non-PD participants. There was no evidence across studies of differences in association with tests of different cognitive domains.

Summary. Three prospective longitudinal studies, one case control study, one retrospective cohort, and five cross-sectional studies examined the potential correlation of B6 serum levels with the diagnosis of dementia or cognitive impairment, and cognitive function. The studies were generally of low quality and had narrow to broad applicability. Only one of the longitudinal studies described a significant correlation between higher levels of vitamin B6 at baseline and better performance in the figure copying test after 3 years of follow-up. However, a similar association was not found for other cognitive tests that were assessed in the same study. It is uncertain whether the improvement in performance in the figure copying test is clinically important. Additionally, no statistically significant correlations were reported between B6 levels and AD or number of AD lesions in brain autopsy, cognitive impairment, PD, or dementia. There is very limited, low quality evidence to allow conclusion for any association between B6 levels and prevention or regression of cognitive function decline.

Cobalamin (Vitamin B12) Level Association Studies

Study Descriptions. Seven longitudinal studies examined the potential association of serum B12 levels with the risk for developing dementia, AD, or cognitive decline (Table 28).^{18, 25, 96, 112}–¹¹⁵ Two of the longitudinal studies recruited participants with normal cognitive function and tried to correlate cognitive function after 3 and 8 years of follow-up to the baseline levels of vitamin B12.^{96, 114} Cognitive function in these two studies was assessed by MMSE score, as well as by tests on construction praxis, language, working and recall memory, and visual reproduction. Sample size ranged between 234 and 1092 among the studies. Six studies included cognitively intact subjects while the seventh also recruited patients with AD. Follow-up ranged between 20 months and 8 years. Studies were generally of moderate quality (5 B, 2 C) and broad applicability.

Four cross-sectional studies assessed the potential association between B12 serum levels and cognitive function (Table 29).^{83, 116}–¹¹⁸ Sample size ranged from 127 to 680. One study included patients with AD and frontotemporal dementia, another study had patients with PD and normal individuals while two studies recruited only subjects with AD. All studies implemented MMSE to assess cognitive function. In addition one study used ADAS test while another used Mattis' DRS. Studies were of low to moderate quality (2 B, 2 C) and narrow to broad applicability.

There were 22 additional studies that compared the B12 serum levels between different groups of participants.^{99, 106, 107, 111, 119}–¹³⁶ Two of the longitudinal studies also compared mean B12 levels with AD diagnosis and cognitive function at baseline.^{25, 113} Five studies estimated the odds ratio for AD or cognitive impairment at certain B12 serum levels (Table 30).^{120, 121, 123, 127, 130} Nine studies compared the percentage of participants with low serum B12 level among different groups including subjects with dementia, cognitive impairment, AD, vascular dementia, and cognitively intact individuals (Table 31).^{25, 119, 121, 122, 124, 125, 129, 133, 134} Definitions of low B12 level varied among studies, ranging from ≤139 mg/mL to <340.1 mg/mL. In addition, 19 studies compared the mean B12 serum levels between different groups of participants, such as subjects with dementia, cognitive impairment, AD, senile and vascular dementia, PD, and cognitively intact individuals (Table 32).^{99, 106, 107, 111, 113, 120, 122}–^{132, 134, 135} Sample size ranged from 41 to 939. All studies measured serum B12 levels except for one study that reported also CSF B12 levels.¹²⁴ Studies were of low to moderate quality (1 A, 7 B, 14 C) and narrow to broad applicability.

Overall Effect. None of the longitudinal studies showed any significant correlation between serum B12 levels and the risk for developing AD or dementia. However, Tucker 2005, after adjusting for baseline cognitive measures, age, education, smoking, alcohol intake, BMI, diabetes, systolic blood pressure, time of second measure relative to folic acid fortification, time interval between the two cognitive measures, and serum creatinine, reported a significant correlation between higher levels of vitamin B12 at baseline and better performance in the figure copying test after 3 years of follow-up (β=0.56, P <0.05). Similarly, Elias 2005, after adjusting for age, education, and gender as well as for Framingham stroke risk profile score, creatinine, alcohol consumption, coffee consumption, total cholesterol, BMI, and ApoE genotype described a significant correlation between higher levels of vitamin B12 at baseline and better performance in the global composite score for cognitive function after almost 8 years of follow-up (β=0.002, P <0.05). This study also reported significant correlation for immediate and delayed recall in the visual reproduction test (β=0.04, P<0.02 and β=0.04, P<0.03 respectively).

Among the other studies that evaluated cognitive function, no significant correlation was found between B12 serum levels and cognitive function test scores for any of the populations included. Among the studies that estimated OR, only Argyriadou 2001 found that the odds for low B12 serum level is two-fold higher in subjects with MMSE score low enough to show cognitive impairment than in cognitively intact individuals (P=0.03). The cutoff level for low B12 in this paper was <145 mg/mL. There were two more studies, which did not estimate an OR, but reported that significantly more subjects with AD or cognitively impaired presented with low serum B12 levels. Tripathi 2001 compared patients with AD and low B12 levels (<187 mg/mL) and subjects with other types of dementia (P<0.05) while Shahar 2001 compared cognitively impaired subjects presenting with lower B12 levels (<203 mg/mL) and cognitively impaired subjects with higher B12 levels (one group with B12 levels 150 to 250 mg/mL and another with greater than 250 mg/mL) (P=0.04 after adjusting for age).

Clarke 1998 compared mean serum B12 levels between patients with AD and cognitively intact participants and found no significant difference. However, when the same study was restricted only to the population with a histological diagnosis of AD, it found that significantly lower serum B12 levels were measured compared to the normal population (P<0.05). A subgroup of the participants in the same study, including 51 subjects with histological diagnosis of AD and 65 normal participants, was evaluated in a later publication by Refsum 2003. Serum holotranscobalamin levels, an active part of total serum B12, were measured and were found to be significantly lower in patients with AD than in controls (P<0.001).

Postiglione 2001 also found significantly lower B12 levels in patients with AD compared to normal individuals but the difference was not significant when the analysis was adjusted for age, serum creatinine, and duration of AD. In addition Religa 2003 reported significantly lower serum B12 levels in the AD group compared to cognitively intact participants (P<0.05). In the same study, no significant difference was reported when subjects with mild cognitive impairment were compared with the normal population. Tripathi 2001 also compared subjects with AD and patients with other dementias and found that the AD population presented with significantly lower B12 levels. In contrast Regland 1988 found that mean vitamin B12 levels for AD and cognitively intact groups were normal while the senile dementia group had a significantly lower B12 level (P=0.0002). None of the other studies including populations with AD or cognitive impairment supported any significant differences for B12 levels. Studies with subjects presenting with PD and vascular dementia did not report any significant results.

Interactions and Covariates. Prospective longitudinal studies reached statistical significance less often than cross-sectional studies. When they reported significant associations, both longitudinal and cross-sectional studies usually found that better cognitive function was related to higher vitamin B12 levels. Assantachai 1997 found that patients with AD and ApoE ε4 genotype had significantly lower B12 levels than patients with vascular dementia and ApoE ε4 genotype, while AD patients with other genotypes had similar B12 levels as those with vascular dementia.

Woitalla 2004 examined the correlation of B12 levels and diagnosis of PD in patients with different MTHFR C677T genotypes (CC, CT, TT). No significant difference was reported when the three genotype groups were compared to each other or to the non-PD participants. McCaddon 2004 found that holotranscobalamin was significantly higher in CC genotypes than the heterozygous CG genotype (P=0.04). Postiglione 2001 also found non-significant differences in B12 levels when they analyzed subcategories of AD patients or controls who were homozygous for MTHFR C677T or non-homozygous.

Postiglione 2001 also reported a statistically significant correlation between duration of disease in months and B12 (r = -0.460, P<0.05), supporting a decrease in B12 vitamin levels with longer AD duration. Anello 2004 reported that B12 levels were not influenced by other covariates such as the severity of dementia or age of onset of the disease. Refsum 2003 showed that low holotranscobalamin was associated with AD at high total homocysteine concentrations (OR 9.45), but not at low homocysteine concentrations.

There was no evidence across studies of differences in association with tests of different cognitive domains.

Summary. Thirty-three studies of low to moderate quality and narrow to broad applicability investigated a potential association between serum or CSF vitamin B12 levels and cognitive function, or diagnosis of several types of dementia and cognitive impairment. Most of the studies focused on AD. The threshold values to define low B12 levels varied across studies. Based on the very few longitudinal studies, vitamin B12 levels did not affect the risk for developing AD or dementia. However, one of the longitudinal studies reported a significant correlation between higher levels of vitamin B12 at baseline among cognitively intact subjects and better performance in the figure copying test after 3 years of follow-up. Another longitudinal study also described a similar correlation for the global composite score for cognitive function and the immediate and delayed recall in the visual reproduction test after almost 8 years of follow-up. However, the clinical importance of these results is unclear. The existing evidence from other studies, which implemented a cognitive function assessment instrument, did not support any correlation between vitamin B12 levels and cognitive function. Among cross-sectional studies, there was a trend for vitamin B12 levels to be lower in patients with AD or other types of dementia, which in certain studies reached statistical significance. However, this trend was not consistent. An inverse correlation between vitamin B12 levels and duration of AD was reported by one study. Besides that evidence for patients with AD or cognitive impairment, there was very limited evidence for populations with PD, and vascular dementia. Potential factors such as genetic mutations, or disease severity that may affect vitamin B12 levels were analyzed by few studies without a consistent effect. Considering also that most of the studies were cross-sectional, no causal relation between B12 vitamin and the developing or progression of dementia can be established.

Folate Level Association Studies

Study Descriptions. A total of 34 studies examined the role of folate vitamin levels with the diagnoses of age related neurocognitive disorder or with cognitive function. Twenty-three studies examined the mean folate level with cognitive function or PD,^{83, 99, 105}–^{107, 109, 111, 113, 116, 120, 122}–^{125, 127}–^{132, 137}–¹³⁹ 10 examined the prevalence of folate deficiency in normal and cognitively impaired or demented participants,^{25, 114, 119, 121, 122, 124, 125, 129, 140, 141} and 15 assessed the risk of AD, dementia, and cognitive impairment with folate levels.^{18, 25, 96, 105, 112}–^{115, 120, 121, 123, 127, 130, 140, 142} Ten of the studies were longitudinal, examining the association between folate levels and future cognitive function; the remainders were case-control or cross-sectional with single time-point analyses (Table 33). The folate levels were assessed in the RBC, plasma, serum, CSF, and blood. The sample sizes of the studies ranged from 30 to 1,100. The majority of the studies were graded moderate to poor quality (1 A, 15 B, 18 C). Fifteen studies had broad, 17 moderate, and 2 narrow applicability.

Overall Effect. All studies reported lower mean folate levels or higher prevalence of folate deficiency among subjects with AD, and cognitive impairment. Among studies that assessed the association between folate levels and cognitive function, four longitudinal studies and one case-control study reported a statistically significant association between lowest quantile of folate level and cognitive decline after adjusting for possible confounders. One other case-control study reported a significant inverse association of folate with cognitive function. Two studies reported folate levels within normal limits among PD subjects.

Fewer than one-third of the 23 studies that examined the mean folate levels and cognitive function reported significantly lower mean folate levels in AD compared with normal subjects (Table 33). Only two studies - Postiglione 2001 and Bowirrat 2002 - adjusted mean folate levels for possible confounders; they reported no significant differences in mean folate levels between AD and normal subjects.

All studies that examined the prevalence of folate deficiency compared subjects with AD or cognitively impairment with other types of dementia or normal subjects (Table 34). The range of prevalence of folate deficiency was between 0 and 67 percent. Across all studies, the prevalence of folate deficiency was higher among AD than controls. Only two studies, Joosten 1997 and Wang 2001 reported statistically significant prevalence of folate deficiency among AD subjects, and those with MMSE mean score ≤26, respectively.

Among 10 studies that examined the longitudinal decline of cognitive function (Table 35), Ravaglia 2005, Kado 2005, Tucker 2005, Maxwell 2002, and Clarke 1998 compared the lowest versus the highest quantile of folate. After adjusting for possible confounders, Ravaglia 2005, Kado 2005, and Tucker 2005 found an increased risk of cognitive function decline among those in the lowest quantile levels of folate. Tucker 2005 found independent effects of lowest folate level among men with mild cognitive impairment with longitudinal (mean of 3 year) and large decline in the score of constructional praxis (spatial copying). Maxwell 2002 examined the risk of adverse health outcomes including cognitive decline with association to the quartiles of folate during 5-year follow-up. This study noted that subjects with lowest folate levels were at greater risk for significant cognitive decline independent of possible confounders. In contrast, Seshadri 2002, Elias 2005 and four other longitudinal studies found no association between folate level and future cognitive function. Three case-control and one cross-sectional studies assessed the odds of AD in relation to high versus the low levels of folate, using a single time point analyses (Table 36). Of note, Quadri 2004 reported a consistently significant association of cognitive impairment with the lowest tertile of folate.

Snowdon 2000, a case-control study described a significant negative age-adjusted correlation of folate levels with severity of cerebral cortex atrophy among all AD subjects, as well as a subset of AD subjects with histologically significant cerebral cortex lesions at autopsy. The regression analysis adjusted for age and histological severity of AD found a 4.4 ng/mL decrease in serum folate was associated with a 1-point decrease for the MMSE score indicating a significant inverse association of folate with cognitive function.

Two studies reported folate levels in PD subjects (Table 33). Woitalla 2004 compared the mean folate level in PD with normal subjects. This study subdivided PD and normal subjects according to their methylene tetrahydrofolate reductase (MTHFR) genotypes CC, CT, TT. The folate levels did not differ between PD and normal subjects, but was significantly higher in the CT allele subgroup of PD compared with normal subjects. Coimbra 2003 reported folate levels within the normal range that did not differ between PD and dementia subjects.

Interactions and Covariates. About one-third of the studies reported associations of folate levels with cognitive function decline adjusted for possible confounders. Four longitudinal studies that adjusted for all possible confounders reported independent and statistically significant association between folate level and cognitive function. Across the studies, there was an association between folate level and global cognitive function. Only one study⁹⁶ reported significant effect on one other cognitive domain.

Summary. There are 34 studies of moderate to poor quality that have examined the role of folate vitamin levels with the diagnoses of age related neurocognitive disorder or with cognitive function. Overall, the studies reported lower mean folate levels or higher prevalence of folate deficiency among AD and cognitive impaired subjects. One-third of the studies adjusted for possible confounders. Of note, four of the 10 longitudinal studies reported consistent and statistically significant association between folate level and cognitive function score. Two case-control studies reported greater risk of cognitive function decline with lower folate levels among AD subjects. Two studies reported mean folate levels within normal limits among PD subjects.

B Vitamin-Homocysteine Interaction With Cognitive Function

Testing the hypothesis of a potential effect of the interaction between homocysteine and B vitamins on cognitive function was not the primary purpose for this report. Consequently, studies that did not consider B vitamins as independent predictors of cognitive function and limited their analyses to homocysteine or the interaction of homocysteine and B vitamins only, were not included in our review.

It is well established that B vitamin status (mostly folate, B12, and possibly B6) is a major determinant of homocysteine level.¹⁴³–¹⁴⁵ Among elderly subjects in the Framingham study, for example, two-thirds of cases of high homocysteine were associated with at least one vitamin concentration below the 70^th percentile.¹⁴³ In addition, studies have consistently found that homocysteine levels can be reduced with B vitamin treatment.¹²³ Furthermore, numerous studies - including several evaluated in this review of B vitamins - have demonstrated an association between higher homocysteine levels and worsened cognitive function. ^{18, 96, 114, 115} Elevated homocysteine levels putatively may cause cognitive decline through both neurotoxic and vasotoxic effects.¹⁴⁶ Of note, almost all trials of combinations of B vitamins - and several interventions with individual B vitamins - were designed with the intent of lowering homocysteine levels. Among studies that reported data on associations between cognitive function and both B vitamins and homocysteine, most found that the association of homocysteine and cognitive function was statistically independent of B vitamin levels; thus homocysteine levels were predictive of cognitive function even after correction for B vitamin deficiency. This statistical independence was found in both cross-sectional studies ^{84, 110, 128, 141} and longitudinal studies,^{18, 115} but does not imply a biological independence. Several studies, both cross-sectional^{112, 147} and longitudinal^{96, 114} found evidence of interactions between homocysteine and either folate, B12, or combined folate, B6, and B12.

Interpretation of the relative effect of different homocysteine levels on any interactions between B vitamin status and cognitive function is more problematic. Given that homocysteine is associated with cognitive decline and that B vitamins affect homocysteine levels, it is hypothesized that much of the B vitamins' role in preventing cognitive decline is through their effect on maintaining low homocysteine levels and thus any associations between B vitamin status and cognitive function would not be expected to be independent of homocysteine level.

However, among the five longitudinal studies that analyzed potential interactions between homocysteine and B vitamins on their association with cognitive decline, the MacArthur Studies of Successful Aging (MSSA),¹¹² the Conselice Study of Brain Aging (CSBA),¹¹⁵ and the Veterans Affairs Normative Aging Study (VANAS)⁹⁶ each found that baseline serum folate level (and also dietary folate in VANAS) was significantly associated with various measures of cognitive decline, independent of homocysteine or vitamins B6 or B12. Each set of authors conclude that the independent contribution of low folate to cognitive impairment may affect the development of cognitive impairment through mechanisms other than homocysteine's direct neurotoxicity. Of note, VANAS found independent effects of folate only for constructional praxis (figure copying), while MSSA analyzed a summary cognitive score based on several tests and CSBA analyzed diagnosis of AD. In contrast, both the Framingham Study¹⁸ and the Framingham Offspring Study¹¹⁴ found no association between folate level and cognitive function.