Introduction

Alzheimer’s disease (AD) is a neurodegenerative disorder, contributing to about two thirds of all dementias in the elderly population [1]. It causes progressive memory loss in mid-to-late adult life. Some individuals inherit this form of dementia before the age of 65 (known as early-onset or familial AD); but most often, AD occurs late in life [2]. It is a heterogeneous disorder with both familial (about 1 % of cases) and sporadic forms, results from selective damage of specific neuronal circuits in the neocortex, hippocampus, and basal forebrain cholinergic system. Affected regions show senile plaques, comprised of neurites displayed around extracellular deposits of beta-amyloid peptides, and many neurons develop neurofibrillary tangles, which reflect the local accumulation of abnormal intracytoplasmic filaments, composed of hyperphosphorylated isoforms of the tau protein [3]. AD is a multifactorial pathology resulting of the interaction of both genetics and environmental factors. Over 100 rare, highly penetrant mutations have been described in three genes (amyloid beta precursor protein, presnilin 1, and presnilin 2) for early-onset familial AD [4], and for late-onset AD, only the association with the apolipoprotein E (APOE) gene has been convincingly replicated. Several clinical and epidemiological studies show a relation between vascular disorders and late-onset AD [1, 5]. Vascular risk factors such as diabetes, hyperlipidemia, hypertension, heart diseases, and high serum homocysteine are also reported risk factors for AD [6, 7].

Elevated levels of homocysteine (Hcy) have been linked to AD [5, 8]. Folate is a cofactor in one-carbon metabolism, during which it promotes the remethylation of homocysteine. Numerous epidemiological and experimental studies have linked folate deficiency and resultant increased homocysteine levels with AD [5]. Methylenetetrahydrofolate reductase (MTHFR) is an important enzyme involved in the folate-dependent metabolism of homocysteine.

MTHFR enzyme mediates the irreversible conversion of 5,10-methylenetetrahydrofolate (5,10-MTHF) to 5-methyltetrahydrofolate (5-MTHF) [9, 10]. Several polymorphisms in the MTHFR gene have been identified, out of which, the most studied and clinically important polymorphism is C677T in exon 4, resulted in substitution of alanine amino acid by valine at position 222 in MTHFR protein [11]. MTHFR functions in dimeric form and flavin adenine dinucleotide (FAD) work as a co-factor, but variant MTHFR (222 valine) dissociates into monomers and its enzymatic activity reduces [12]. This substitution makes enzyme thermolabile with reduced enzymatic activity. The mean activity of TT variant enzyme was 40–50 % that of the CC variant enzyme at 37 °C [13].

The frequency of MTHFR C677T mutation varies among racial and ethnic groups of the world. T allele frequency ranges from 0.20 to 0.55 in Europeans, 0.11 to 0.35 in Americans, 0.063 to 0.094 in Africans, from 0.04 to 0.38 in Asian population, and 0.10 to 0.47 in Australians [14–18]. Several studies showed that the mutant T allele increases homocysteine levels particularly in folate deficiency state [11, 13, 19, 20]. Impact of MTHFR C677T polymorphism on development and pathogenesis of AD have been conflicting and inconclusive [21–28]. Hence, meta-analyses of all published case control studies investigating C677T polymorphism as risk factor for AD were carried out to shed some lights on conclusive role of MTHFR C677T polymorphism in AD.

Methods

Meta-analyses of published case control articles were carried out according to MOOSE guidelines [29].

Literature and Search Strategy

Electronic databases Pubmed, Springer Link, Elsevier, and Google Scholar were searched up to January 2015 for suitable articles using keywords “MTHFR and Alzheimer’s disease,” “folate metabolism and Alzheimers disease,” and “MTHFR C677T polymorphism and Alzheimers disease.” The references from the eligible articles were also reviewed to find other potential articles. If more than one study by the same author using the same case series was published, either the studies with the largest sample size or the most recently published study was included.

Inclusion and Exclusion Criteria

The following inclusion criteria were set for the meta-analysis: (i) each study should be an independent case-control study, (ii) the purpose of all the studies and statistical methods should be similar, (iii) studies should reported enough information to calculate the odds ratio with 95 % confidence interval (CI), and (iv) inclusion of the patients should be done according to the standard diagnosis parameter (Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) and Mini Mental State Examination (MMSE)). The exclusion criteria were: (i) only case studied; (ii) review papers, editorial, and letter to editor; (iii) containing overlapping data; and (iv) study not providing enough information to estimate odds ratio (OR) with 95 % CI (incomplete raw data) and not well described.

Data Extraction

The following information was extracted from each study: the first author’s family name, year of publication, country, race, sample size, outcome, characteristics of controls, case and control diagnostic criteria, genotyping method, and genotype distribution in cases and controls. Detailed information, wherever not available, was collected by contacting authors.

Statistical Analysis

The strength of association between the MTHFR C677T polymorphisms and AD risk was evaluated by OR and 95 % CI according to allele contrast (T vs C), homozygote (TT vs CC), heterozygote/co-dominant (CT vs CC), recessive (TT vs TC + CC), and dominant (TT + CT vs CC) models. A chi square-based “Q” test defined by Cochran was used to assess the heterogeneity (between study variability) in the meta-analysis [30]. Since the Q statistics is only useful for testing the existence of heterogeneity qualitatively but not quantitatively, another index “I 2,” calculated as the percentage of the total variability in a set of effect sizes due to true heterogeneity, was used to quantify the degree of heterogeneity [31]. A tentative classification of I 2 values proposed by Higgins and Thompson has been used to interpret the magnitude, viz., 25, 50, and 75 %, which corresponds to low, medium, and high heterogeneities, respectively [32]. In the absence of significant heterogeneity determined by the results of Q test, the Mantel-Haenszel fixed effect model (Peto method) was used for the combination of data, while in the presence of significant heterogeneity, the Dersimonian-Laird random effect model (DL method) was used for combining the data [33, 34]. High-resolution plots (forest plots) were generated to estimate the pooled odds ratio corresponding to 95 % CI and the p value. Stratified analyses were performed by ethnicity. Sensitivity analysis was performed to evaluate the stability of the results by removing the studies not in Hardy-Weinberg equilibrium (HWE), studies with very high p value, and studies with small sample size. Cumulative meta-analysis was also done to observe the effect of subsequent addition of each study. Hardy-Weinberg equilibrium and chi-squared test methods were used to test the distribution of genotypes in the control group of each study. For this purpose, data was analyzed using calculator available at http://ihg.gsf.de/cgi-bin/hw/hwa1.pl.

Publication Bias

Publication bias was investigated by using the funnel plots, viz., funnel plot of standard error by log odds ratio and funnel plot of precision by log odds ratio. Different statistical tests such as Begg and Mazumdar rank correlation [35] and Egger’s regression intercept [36] were adopted to assess and quantify the publication bias and its impact on the analysis. P < 0.05 was considered to indicate statistical significance, and all the p values were two sided. All statistical analyses, forest, and funnel plots were performed by the computer program MIX version 1.7 [37].

Result

The preliminary search resulted in 119 publications from Pubmed, Google Scholar, Elsevier, and Springer Link. Out of which, 51 were irrelevant for the present meta-analysis, which includes reviews, book chapters, case reports, editorials, and articles mentioning other genes. After initial exclusion, a total of 68 article publications were identified. Out of which, 16 articles were irrelevant for the present meta-analysis, and in 11 articles only, cases were discussed. Thus, a total of 41 articles were included in present meta-analysis. The search workflow was shown in Fig. 1.

Fig. 1
figure 1

Flow diagram of selection of studies included in the meta-analysis

Full size image

Study characteristics were summarized in Table 1. Forty-one articles that investigated the association between C677T polymorphism and AD were found suitable for the inclusion in the present meta-analysis [3, 21–28, 38–68]. One author [25] studied two different population; their data were included as two independent studies. A total of 42 studies were found suitable for the inclusion in present meta-analysis.

Table 1 Details of studies included in the present meta-analysis
Full size table

The studies were published between 1998 and 2014. All these 42 studies were performed in different countries—Brazil [49, 51], China [27, 44, 45, 47, 48, 50, 56, 57, 59, 60, 62, 65], Egypt [68], Germany [26], India [66, 67, 69], Iran [52], Ireland [41], Israel [21, 39], Italy [22, 23, 25, 40, 43, 63, 64], Japan [38, 42, 46], Korea [55], Poland [24, 53, 54, 58], Sweden [3, 61], Tunisia [70], and USA [25].

Summary Statistics

In all 42 studies, total cases were 4888 with CC (1676), CT (2290), and TT (922), and controls were 6142 with CC (2389), CT (2719), and TT (1034) genotypes. In control genotypes, percentages of CC, CT, and TT were 38.90, 44.27, and 16.80 %, respectively. In total cases, genotype percentages of CC, CT, and TT were 34.29, 46.85, and 18.86 %, respectively. Frequency of CC genotype was highest in both cases and controls (Table 2). In cases and controls, the allele C was the most common. Except two studies [44, 62], the genotype distributions among the controls of all 39 studies followed Hardy-Weinberg equilibrium. Out of 42 studies, ORs of 12 studies [3, 22, 23, 26, 39, 43–45, 50, 53, 56, 57] were below one; i.e., these studies did not show any association between MTHFR C677T polymorphism and AD.

Table 2 The distributions of MTHFR C677T genotypes and allele numbers in Alzheimer’s cases and controls
Full size table

Allele Contrast Meta-Analysis

The main results of meta-analysis and the heterogeneity test were shown in Table 3. Allele contrast meta-analysis showed significant association with fixed effect model (allele contrast model ORT vs C = 1.20, 95 % CI 1.13–1.27, p < 0.0001, I 2 = 71.58 %, P heterogeneity < 0.0001, P Pb = 0.25) and random effect model (ORT vs C = 1.29, 95 % CI = 1.07–1.56, p = 0.0003) (Table 3 and Fig. 2). High significant heterogeneity was found, so random effect model was adopted. In cumulative meta-analysis using random effect model, the association of T allele with AD turned statistically significant with the addition of study of Liao et al. [47] and remained significant thereafter.

Table 3 Summary estimates for the odds ratio (OR) of MTHFR C677T in various allele/genotype contrasts, the significance level (p value) of heterogeneity test (Q test), and the I 2 metric and publication bias p value (Egger test) in total, Asian, and Caucasian studies
Full size table
Fig. 2
figure 2

Forest plots for the association between MTHFR C677T polymorphism and Alzheimer’s disease for allele contrast model (T vs C) with random effect model

Full size image

Genotype Contrast Meta-Analysis

Table 3 summarizes the ORs with corresponding 95 % CIs for association between MTHFR C677T polymorphism and risk of AD in dominant, recessive, homozygote, and co-dominant models. There was evidence of association of the MTHFR TT genotype with the risk of AD relative to the MTHFR CC genotype. Since there was no significant heterogeneity (p = 0.29, I 2 = 10.02 %) between the studies, the fixed effect pooled OR was considered (homozygote model ORTT vs CC = 1.31, 95 % CI = 1.16–1.48, p = 0.0002) (Fig. 3). Significant association was observed between C677T polymorphism and AD using co-dominant model (ORCT vs CC = 1.24, 95 % CI = 1.13–1.35, p < 0.0001). The overall analysis of the recessive model for T allele showed insignificant heterogeneity (p = 0.21, I 2 = 14.56 %) and showed significant association (ORTT vs CT + CC = 1.13, 95 % CI = 1.00–1.28, p = 0.04). However, the dominant model for the effect of T allele produced significant association overall (ORTT + CT vs CC = 1.29, 95 % CI = 1.19–1.40, p < 0.0001) (Fig. 4).

Fig. 3
figure 3

Forest plots for the association between MTHFR C677T polymorphism and Alzheimer’s disease for homozygote model (TT vs CC) with random effect model

Full size image
Fig. 4
figure 4

Forest plots for the association between MTHFR C677T polymorphism and Alzheimer’s disease for dominant model (TT + CT vs CC) with random effect model

Full size image

Subgroup Analysis

Subgroup analysis based on ethnicity was performed. In all eligible studies, 21 studies were conducted in Asians, 18 studies were conducted in Caucasians, and 2 studies were conducted in other population. In Asian population (number of studies = 21; 2588 cases/3630 controls), allele contrast meta-analysis showed significant association adopting both fixed (ORT vs C = 1.19, 95 % CI = 1.10–1.28, p = <0.0001, I 2 = 34.02 %, p value of heterogeneity (P hetero) = 0.06, p value of Eggers test (P Pb) = 0.46), and random (ORT vs C = 1.21, 95 % CI = 1.1–1.34, p = 0.0003) effect models. Combined mutant genotypes (dominant model) also showed significant association with fixed (ORTT + CT vs CC = 1.32, 95 % CI = 1.17–1.48, p < 0.0001) and random (ORTT + CT vs CC = 1.33, 95 % CI = 1.17–1.51, p < 0.0001) effect models. In this subgroup, heterogeneity between studies (I 2 = 11.54 %, P hetero = 0.03) was low and publication bias (P Pb = 0.16) was absent (Table 3 and Fig. 5).

Fig. 5
figure 5

Forest plots for the association between MTHFR C677T polymorphism and Alzheimer’s disease for homozygote model (TT vs CC) in Asian studies with random effect model

Full size image

In European population (number of studies = 18; 2227/2433 cases/controls), allele contrast meta-analysis showed significant association with both fixed (ORT vs C = 1.26, 95 % CI = 1.15–1.37, p < 0.0001) and random (ORT vs C = 1.34, 95 % CI = 1.01–1.68, p = 0.01) effect models with significant heterogeneity (I 2 = 83.19 %, P heterogeneity = 0.43). The combined mutant genotype (dominant model) showed statistically significant association with fixed effect model (ORTT + CT vs CC = 1.18, 95 % CI = 1.04–1.33, p = 0.008) with no heterogeneity and no publication bias (P Pb = 0.28) (Table 3 and Fig. 6).

Fig. 6
figure 6

Forest plots for the association between MTHFR C677T polymorphism and Alzheimer’s disease for homozygote model (TT vs CC) in Caucasian studies with random effect model

Full size image

Sensitivity Analysis

Sensitivity analyses were conducted to determine whether modification of the inclusion criteria of the meta-analysis affected the final results. In allele contrast meta-analysis, sensitivity analysis performed by exclusion of the studies in which control population was not in Hardy Weinberg equilibrium, studies with small sample size and studies with high p values. Control population of only two studies [44, 62] were not in HW equilibrium, and heterogeneity did not decreased after exclusion of these two studies (P hetero, I 2 = 72.59 %). Exclusion of ten studies involving small sample size, less than 50 (Chapman et al. [21], n = 49; Nishiyama et al. [38], n = 24; Zuliani et al. [23], n = 40; Bi et al. [44], n = 42; Fernandez et al. [49], n = 30; da Silva et al. [51], n = 43; Dorszewska et al. [54], n = 38; Yuan et al. [56], n = 30; Zhang et al. [59], n = 43; Elhaway et al. [68], n = 43), also did not decreased heterogeneity (P hetero < 0.0001, I 2 = 77.27 %).

Publication Bias

Begg’s funnel plot and the Egger’s test were conducted to estimate the publication bias of articles. Both the results of Begg’s and Egger’s test showed evidence of publication bias in allele contrast and co-dominant models, and absence of publication bias was observed in homozygote, dominant, and recessive genetic models (T vs C Begg’s test p = 0.02, Egger’s test p = 0.24; CT vs CC Begg’s test p = 0.02, Egger’s test p = 0.03; TT vs CC Begg’s test p = 0.18, Egger’s test p = 0.3; dominant model TT + CT vs CC Begg’s test p = 0.06, Egger’s test p = 0.06; and recessive model TT vs CT + CC Begg’s test p = 0.47, Egger’s test p = 0.61) (Fig. 7 and Table 3).

Fig. 7
figure 7

Funnel plots of total studies a precision versus OR for T vs C model, b standard error versus OR for T vs C model, c precision versus OR for TT vs CC model, d standard error versus OR for TT vs CC model, e precision versus OR for TT + CT vs CC model, and f standard error versus OR for TT + CT vs CC model

Full size image

Discussion

Results of present meta-analysis showed strong association between MTHFR C677T polymorphism and Alzheimer’s disease. Several epidemiological and experimental evidences have linked derangements of one-carbon metabolism to vascular, neurodegenerative, and neuropsychiatric diseases, including strokes [71]. Folate deficiency and thermolabile MTHFR enzyme (TT genotype) with low activity increase concentration of homocysteine. The deficiency of MTHFR causes an accumulation of 5,10-methylenetetrahydrofolate as well as the inhibition of 5-methyltetrahydrofolate synthesis. Reduced synthesis of 5-methyltetrahydrofolate will cause decreased homocysteine remethylation. Hyperhomocysteinemia, hypomethioninemia, and reduced S-adenosylmethionine occur frequently in severe MTHFR deficiency [72]. There is several evidences that MTHFR mutant TT genotype is in accordance with elevated homocysteine level, which in itself is an independent risk factor for vascular diseases and cognitive impairment [73]. Folic acid supplementation has a protective effect on homocysteine-induced oxidative stress by reducing intracellular superoxide levels and, to a lesser extent, quenching hydrogen peroxide [74].

Higher concentration of homocysteine could affect adult brain and cause degeneration in adult brain by several mechanisms like (i) it altered DNA methylation pattern [71] and affects gene expression; (ii) it impaired DNA repair [71]; (iii) it increased oxidative stress [74], generated free oxygen radicals, accelerated DNA damage, and eventually lead to neuron apoptosis [74–77]; (iv) it reduced antioxidant reserves of the cell; (v) it enhanced beta-amayloid peptide generation [78]; (vi) it sensitized neurons to amyloid toxicity [75]; (vii) it released inflammatory mediators such as nuclear factor (NF)-kappa B, interleukin (IL)-1beta, IL-6, and IL-8 [66, 79]; (viii) it modulated IL-6 genes [80], which is a proinflammatory cytokine and promoted neuronal expression of neurofilaments, tau protein, and beta-amyloid precursor protein; all are involved in pathogenesis of AD [66, 81]; and (ix) Hcy impaired vascular endothelial function [82] (Stuhlinger et al. 2003), stimulated vascular smooth muscle proliferation [83], breaks balance between coagulation and bleeding pathway [84], and mediated thrombosis. Ultimately, these adverse effects reduce blood supply to the brain and accelerated the neuron apoptosis [77]. Meta-analysis is the statistical analysis of a large collection of analysis results for the purpose of integrating the findings, and it is a powerful tool for systematic review of a focused topic in the literature that provides a quantitative estimate for the effect of a gene, treatment intervention, or exposure [85]. Numerous meta-analyses were published considering MTHFR C677T as risk for several diseases and defects like neural tube defects [10], Down syndrome [86], congenital heart defects [87], schizophrenia [88], bipolar [89], anxiety [89], depression [90], and cancer [91]. One meta-analysis was identified during literature search on the same topic [77]. Zhang et al. [77] pooled 19 studies and reported significant association between AD and MTHFR 677 T polymorphism (OR = 1.15, 95 % CI = 1.08–1.39) using allele contrast genetic model. There are several published studies available but not included in the previous meta-analysis. So, author conducted a comprehensive meta-analysis with the largest number of studies to date to investigate the possible relationship between maternal MTHFR C677T polymorphism and the risk of AD.

The results of this meta-analysis should be interpreted with some caution because there were few limitations in present analysis like (i) results were based on unadjusted OR values that lack the original data from the eligible studies, which could lead to relatively weak power to estimate the real relationship; (ii) sample size in few included studies was relatively small to investigate the association between MTHFR C677T polymorphism and AD risk; (iii) significant between-study heterogeneity was detected and may be distorting the meta-analysis; and (iv) present meta-analysis was based on single-factor estimates, which overlooked the interactions of gene-gene and gene-environment in the development of AD. Present meta-analysis had several strengths also. First, substantial number of cases and controls were pooled from different studies, which significantly increased the statistical power of the analysis. Second, no publication biases were detected, indicating that the whole pooled results may be unbiased.

In conclusion, present comprehensive meta-analysis indicates that there is a conclusive association between the MTHFR C677T polymorphism and the risk of AD, whereas significant heterogeneity was evident across the individual studies. The subgroup analysis also indicated that there is a significant association between MTHFR C677T gene polymorphic variation and AD patients in Asian and Caucasian population.