Lack of association between the MTHFR C677T variant and migraine with aura in an older population: Could selective survival play a role?

Abstract

Background:

Several studies, but not all, of primarily middle-aged or younger adults have suggested that the common MTHFR C677T variant is a genetic risk factor for migraine with aura (MA). Here, we consider whether this variant is associated with MA risk in an older non-clinical population (AGES-Reykjavik cohort).

Methods:

Participants are a sub-sample (n = 1976) of subjects from the Reykjavik Study (RS; mean age 50) and its continuation, AGES-RS (mean age 76). We estimated the relative odds of MA in TT versus CC carriers using multinomial logistic regression. As both MA and the TT genotype may be linked with modestly reduced longevity, we performed a simple simulation to illustrate the effect that selective survival may have had on our observed gene–disease association.

Results:

TT versus CC carriers were at marginally reduced odds of MA (OR_TT 0.55 (0.3–1.0), p = 0.07), significantly for women (OR_TT 0.45 (0.2–0.9), p = 0.03). Assuming the ‘true’ (e.g. mid-life) effect of the TT genotype is OR_TT 1.26, from a recent meta-analysis, our simulation suggested that if 25-year mortality had been (hypothetically) 13% higher in MA subjects with the TT versus CC genotype, the measured effect of the TT genotype on MA would have been attenuated to non-significance (e.g. OR_TT 1.00). Our observed protective effect was consistent with the most extreme selective mortality scenario, in which essentially all of the previously reported increased mortality in MA subjects was (hypothetically) found in CT or TT carriers.

Conclusion:

The MTHFR 677TT genotype was associated with marginally reduced risk of MA in our older population. Our simulation illustrated how even modest selective survival might obscure the apparent effect of a genetic or other risk factor in older populations. We speculate that some of the heterogeneity previously observed for this particular genetic variant may be due to age range differences in the studied populations.

Keywords

Migraine folate MTHFR association studies in genetics selective survival selection bias

Introduction

The MTHFR gene on chromosome 1p36.3 encodes 5,10-methylenetetrahydrofolate reductase (MTHFR), a key enzyme in the folate metabolic pathway. A common functional variant in this enzyme, C677T, has been described in many populations (1). Carriers of the TT genotype have approximately 50% reduced enzymatic activity and consequently moderately increased total homocysteine, one of the major players in the folate metabolic pathway (2). This variant (or elevated homocysteine) has been associated with risk of various neurological disorders including Alzheimer-type dementia, Parkinson’s disease, epilepsy, and migraine with aura (MA) (3 –5). In particular, the C677T variant is possibly the best studied and most replicated genetic risk factor for a common migraine type (MA) (6 –20), although there has been considerable heterogeneity in the results to date (21,22).

Both MA and the MTHFR C677T variant may be related to longevity. As recently reported in the Reykjavik Study (RS) cohort, subjects with mid-life MA were at approximately 18% increased risk of all-cause mortality over 25 years of follow-up compared with those without headache (23). The TT genotype has also been reported to be less prevalent in older compared with younger subjects in several populations (Figure 1) – indirectly suggesting an adverse effect on longevity – although it has been suggested this variant may be protective in populations with mandated folic acid fortification (24).

Figure 1.

Proportion of younger versus older subjects who are MTHFR 677TT carriers as reported in selected population-based studies.

We considered whether the MTHFR 677TT genotype was associated with increased risk of MA in a sub-sample of 1976 men and women who participated in the population-based RS (1967–1991; mean age 50), which was followed by the Age Gene/Environment Susceptibility-Reykjavik Study (AGES-RS; mean age 76). Anderson et al. recently proposed a general model to estimate SNP effect size erosion due to selective survival using three highly lethal diseases (intracerebral hemorrhage, ischemic stroke, and myocardial infarction) (25). Given the age of our study population, we wondered whether our late-life gene-association study might be influenced by selectively reduced participation of MA subjects with the TT genotype. These data led us to consider the broader question of the degree of selective survival that would be sufficient to erode or even reverse a ‘true’ genetic or other risk factor when measured in an older study population, even in a disease that may be associated with only modestly reduced longevity.

Methods

Study population

Detailed descriptions of the RS and AGES-RS have been published previously (26 –30). In brief, the RS is a population-based cohort study established in 1967 by the Icelandic Heart Association to prospectively study cardiovascular disease in Iceland (26). In 2002, the RS continued as the AGES-RS to examine risk factors, genetic susceptibility, and gene/environment interactions in relation to disease and disability in old age (27). In the present genetic sub-study, 1976 men and women (mean age 76, range 66–93) were genotyped for the MTHFR C677T polymorphism. The AGES-RS was approved by the Icelandic National Bioethics Committees (VSN-00-063, 00-063-V8+1), which acts as the Institutional Review Board for the Icelandic Heart Association, the Data Protection Authority, and by the Institutional Review Board for the US National Institute on Aging, National Institutes of Health. Written informed consent was obtained from all participants.

Headache classification

Subjects were asked about headache symptoms in the mid-life RS interview. The mid-life headache classification and associated survival by headache type for this cohort has been described in detail (23). Briefly, those reporting headache once or more per month were asked follow-on questions about five features typical of migraine; those with two or more non-aura features were designated as migraineurs without aura (MO); those reporting either visual symptoms or numbness before headaches were designated as migraine with aura (MA). Subjects were classified into four mutually exclusive groups: no headache (n = 1357), non-migraine headache (NMH) (n = 367), migraine with aura (n = 167) and migraine without aura (n = 85). Subjects meeting both criteria for migraine with and without aura were included in the migraine with aura group.

Genotyping

DNA was isolated from peripheral blood lymphocytes according to the method developed by Scotlab Bioscience (Kirshaws Road, Coatbridge, UK). The proprietary GoldenGate assay by Illumina Genotyping Services (Illumina Inc., San Diego, CA, USA) was used for genotyping.

Analysis

We determined whether late-life genotype frequencies were in Hardy-Weinberg equilibrium overall and by headache categories. We calculated adjusted odds ratios (AORs) as an estimate of the relative risk of headache (NMH, MO, MA) in those with the CT or TT genotype compared with those with the CC genotype, adjusting for age (at blood draw for genotyping), sex, and duration of follow-up, using multinomial logistic regression.

Selective survival simulation

The prevalence of mid-life migraine has been previously reported for the Reykjavik cohort (23). We assumed that mid-life genotype frequencies by headache groups, if we had measured them, would be consistent with two assumptions: 10% of controls would be TT carriers (1), and the ‘true’ (e.g. mid-life) association between MTHFR C677T genotype and MA was as reported in a recent meta-analysis (22) using the additive model.

Using the calculated hypothetical mid-life genotype frequencies as baseline, we assumed that 25-year mortality risk in the MA group was different for those who were TT versus CT versus CC carriers. We first assumed, in the absence of selective survival, that MA subjects (all genotypes) were at 18% increased risk of 25-year mortality compared with controls (RR_MA-All 1.18 (1.1–1.3), p < 0.05) as previously reported (23). This is a slightly lower age and sex-adjusted RR than the fully adjusted relative mortality estimate (RR 1.21 (1.1–1.3), p < 0.05) from this study, as here we want to estimate the hypothetical effect of selective survival per se without accounting for the potential explanatory or mediating factors such as the cardiovascular risk factors that were included in the fully adjusted models. We also note that the other published mortality risk estimates for MA, to our knowledge, were not appropriate estimates for this simulation as they were based on cause-specific mortality or did not provide a pooled estimate for women and men (31 –33). We assumed no increased risk of mortality in subjects with headache other than MA, as previously reported (23,34) and, for simplicity, we assumed no increased risk of mortality in subjects with the TT genotype without MA.

Using the observed mortality in controls as baseline, the observed age- and sex-adjusted risk of mortality in subjects with MA compared with controls (23), and the hypothetical mid-life genotype frequencies, we calculated late-life genotype frequencies that would result from various levels of selective survival in MA subjects based on genotype. We held the overall RR of mortality in MA subjects constant at 1.18 and assumed that the increased risk of mortality in CT carriers would be half that of TT carriers. We then calculated ORs based on the resulting late-life genotype frequencies.

Results

Our study population consists of 1976 surviving RS study participants (1143 women, 833 men). Observed late-life genotype frequencies (Table 1) were broadly consistent with results from other Caucasian populations (1) and were in Hardy-Weinberg equilibrium within each headache group (Controls p = 0.11; NMH p = 0.52; MO p = 0.96; MA p = 0.27) and overall (p = 0.18). Compared with those without headache, there was no difference in adjusted odds of MA in subjects with the TT genotype compared with those with the CC genotype (Table 2 AOR_TT 0.55 (0.3–1.0), 0.07). When stratified by sex, women with the TT genotype were at reduced odds of MA compared with those with the CC genotype (AOR_TT 0.45 (0.2–0.9), 0.03).

Table 1.

Observed late-life MTHFR C677T genotype frequencies by mid-life migraine status in the AGES-Reykjavik Study.

	MTHFR C677T genotype			T allele	Total
	CC	CT	TT	T allele	Total
Control	612 (45%)	579 (43%)	166 (12%)	34%	1357
NMH	175 (48%)	153 (42%)	39 (11%)	31%	367
MO	37 (44%)	38 (45%)	10 (12%)	34%	85
MA	79 (47%)	76 (46%)	12 (7%)	30%	167
Total	903 (46%)	846 (43%)	227 (11%)	33%	1976

p = 0.6 for distribution of genotype by headache categories.

NMH: non-migraine headache; MO: migraine without aura; MA: migraine with aura.

Table 2.

Adjusted odds of migraine in subjects with the CT or TT genotypes relative to the CC genotype in the AGES-Reykjavik Study.

MTHFR C677T genotype	N	Headache classification
MTHFR C677T genotype	N	NMH	MO	MA
1) Men and women
		N = 367	N = 85	N = 167
CC	903	1.00	1.00	1.00
CT	846	0.92 (0.7–1.2)	1.10 (0.7–1.8)	1.02 (0.7–1.4)
TT	227	0.81 (0.5–1.2)	0.99 (0.5–2.0)	0.55 (0.3–1.0)
2) Men
		N = 119	N = 12	N = 34
CC	389	1.00	1.00	1.00
CT	348	1.01 (0.7–1.5)	0.73 (0.2–2.6)	1.75 (0.8–3.7)
TT	96	1.13 (0.6–2.1)	1.39 (0.3–7.1)	1.07 (0.3–3.9)
3) Women
		N = 248	N = 73	N = 133
CC	514	1.00	1.00	1.00
CT	498	0.87 (0.6–1.2)	1.14 (0.7–1.9)	0.87 (0.6–1.3)
TT	131	0.66 (0.4–1.1)	0.87 (0.4–2.0)	0.45 (0.2–0.9)*

Comparison group consists of 1357 individuals (668 men, 689 women) without headache.

ORs from multinomial logistic regression. The outcome variable is migraine status (NMH, MO, MA) and the independent variable is genotype (CC, CT, TT). ORs are adjusted for age, sex, and follow-up. *p < 0.05.

NMH: non-migraine headache; MO: migraine without aura; MA: migraine with aura.

Results from our selective survival simulation are shown in Tables 3 and 4. Hypothetical mid-life genotype frequencies (see methods) are shown in Table 3. Based on our assumptions, 12% of MA subjects in mid-life would have been TT carriers versus 10% of others. Table 4 illustrates the effect of increasing levels of hypothetical selective survival in MA subjects by genotype based on the assumptions described in the methods. In our first most conservative assumption of no selective survival, odds of MA in those with the TT genotype relative to those with the CC genotype were by design the same as our hypothesized ‘true’ OR per Schürks et al., additive model (22) (e.g. OR_TT 1.26) (Table 4, Scenario A). A complete attenuation of the ‘true’ OR occurred when the relative risk of mortality in MA subjects with the TT genotype was RR_MA-TT 1.26 (keeping the average relative risk of RR_MA-All 1.18 in all MA subjects) (Table 4, scenario B). The reported protective effect for the TT genotype as reported from the Women’s Health Study (average age of 55 at genotyping, RR 0.79 (0.65–0.96), p = 0.02) (12) is consistent with the survival estimates by genotype in scenario C. The results based on our sample (average age of 75) are consistent with the most extreme selective survival estimates in scenario D, in which almost all of the increased mortality risk in migraineurs previously reported for this cohort was (hypothetically) in CT/TT carriers. Genotype frequencies in MA subjects remained in Hardy-Weinberg equilibrium for all scenarios.

Table 3.

Hypothetical mid-life MTHFR C677T genotype distribution in the original Reykjavik Study cohort.

	A	B: Hypothesized mid-life MTHFR C677T genotype distribution
	A	CC		CT		TT
Control	13071	6111	47%	5653	43%	1307	10%
NMH	3631	1698	47%	1570	43%	363	10%
MO	626	293	47%	271	43%	63	10%
MA	1397	605	41%	629	46%	163	12%
Total	18,725

Percentages are rounded.

A: Actual prevalence of migraine in 18,725 subjects from the Reykjavik cohort at average age 50 (range 33–65) (from Gudmundsson et al., 2010 (23)).

B: Calculated hypothetical mid-life genotype distribution by headache type based on two assumptions: The prevalence of the TT genotype in controls is 10% (Botto and Yang 2000 (1)) and the ‘true’ relative odds of MA in TT versus CC carriers is 1.26, the pooled OR from one dozen studies as previously reported (Schürks et al., 2010 (22)).

NMH: non-migraine headache; MO: migraine without aura; MA: migraine with aura.

Table 4.

Illustration of how selective survival in subjects with a chronic disease (migraine with aura) based on genotype would change the apparent effect of the genotype in late-life: the AGES-Reykjavik Study.

		A	B	C	D
Assumed relative risk of mortality in migraineurs with aura (MA) compared with controls by genotype	RR_MA-CC	1.18	1.13	1.10	1.05
	RR_MA-CT	1.18	1.20	1.22	1.25
	RR_MA-TT	1.18	1.26	1.34	1.44
	RR_MA-All	1.18	1.18	1.18	1.18
Calculated proportion of TT carriers in the population		10.1%	10.0%	9.9%	9.8%
Calculated proportion of TT carriers in MA group		11.7%	10.0%	8.4%	6.4%
Calculated relative odds of MA for TT versus CC carriers (OR_TT)		1.26^a	1.00	0.79	0.56

Explanation of Table 4.

For the purpose of this simulation, OR_TT 1.26 is assumed to be the ‘true’ measure of effect for the TT genotype on migraine with aura (MA) based on a recent meta-analysis (Schürks et al., 2010 (22)) (i.e. odds of MA are assumed to be 26% higher in TT versus CC carriers in younger populations, based on pooled OR 1.13 (0.97–1.32) per T allele). RR_MA 1.18 is the observed age and sex-adjusted 25-year mortality rate ratio in MA subjects relative to controls as previously reported for the Reykjavik Study (Gudmundsson et al., 2010 (23)).

In the baseline scenario (column A), survival in the MA group relative to controls is assumed to be unrelated to genotype (e.g. RR_MA-CC = RR_MA-CT = RR_MA-TT 1.18). In Scenario B, the indicated survival assumptions in the MA group by genotype produce a complete erosion of the evident effect of the TT genotype (Scenario B: OR_TT 1.0). Scenario C survival assumptions produce the marginally protective effect for the TT genotype reported in the Women’s Health Study, in which subjects were on average age 55 at genotyping (Schürks et al., 2008 (12)) Scenario D survival assumptions produce the effect for the TT genotype reported herein, in which subjects were on average age 75 at genotyping.

Discussion

Selection bias is an ‘error due to systematic differences in characteristics between those who are selected for study and those who are not’ (35,36). Selection bias due to selective survival is a particular concern when the study population is elderly. If the disease of interest is associated with reduced longevity, there will naturally be fewer subjects with the disease (cases) in older study populations. Similarly, if the risk factor of interest (say, genotype ‘X’) is associated with reduced longevity, then there will naturally be fewer subjects in the study population with genotype X. If the reduced survival in cases is independent of genotype (or equivalently if the reduced survival by genotype is independent of case status) then the selection bias is considered ‘non-differential.’ The greater threat to study validity occurs with differential selection bias, meaning in this example that survival is additionally or exclusively affected in subjects with both the disease and genotype X. In this latter case genotype X might be over-represented in younger cases (e.g. associated with increased risk) but under-represented in older cases (e.g. associated with decreased risk) (25).

Although the MTHFR C677T variant is one of the few replicated candidate genetic risk factors for a common migraine type (21), we found in this older population that the TT genotype was marginally protective for MA. A recent meta-analysis (22) suggested that the pooled measure of effect for the TT genotype on MA was in the OR 1.3 range (additive model). There was evidence of heterogeneity by study location, which was ascribed to differences in ethnicity, but may additionally reflect regional environmental or dietary factors such as folic acid supplementation that influence the expression of MTHFR C677T genotype (37 –40). Our study population includes European Caucasians and there is no mandatory folic acid fortification of food stuffs in Iceland.

We speculate that the different age ranges of the studied populations may also have contributed to this heterogeneity. Extant studies have generally not reported their results in a way that would allow a robust consideration of an age effect. For example, when considering the effect of selective survival it is not sufficient to consider only the mean age of the study population. The age distribution should also be considered as we would expect to find the greatest effect of differential survival in the older populations.

The recent reports from the Women’s Health Study (WHS) are of particular interest. That study of women (mean age 54 years, with about 10% 65 years or older) who had no history of cardiovascular disease or cancer showed at baseline that the TT genotype was less prevalent (e.g. was protective) in the women with MA compared with controls (12,41). Over 11 years of follow-up, women with MA had an increased risk of incident fatal and non-fatal cardiovascular disease (RR 2.06 (1.5–2.8)) (12). The risk was stronger in the women with MA and the TT genotype (RR 3.66 (1.7–7.9)) compared with women with MA and the CC (RR 2.39 (1.56–3.68)) or CT (RR 1.48 (0.89–2.47)) genotypes, although confidence intervals overlapped. These findings are consistent with our hypothesis of selective mortality or (additionally in this case) morbidity given their exclusion of women with prevalent cardiovascular disease. Arguing against this is the finding reported in a follow-up study suggesting that the protective effect of the TT genotype was most evident for women with less frequent attacks (42).

Anderson et al. recently proposed a general model to estimate SNP effect size erosion due to selective survival for three highly lethal diseases (intracerebral hemorrhage, ischemic stroke, and myocardial infarction) (25). Here we show how even modest selective mortality over a long period of follow-up, such as might occur in MA, could hypothetically erode or reverse the measured effect of a genetic or other risk factor when studied in an older population. In such cases, the degree to which the observed measurement of effect will generalize to younger populations, and vice versa, will depend not only on the age distribution of the target study population, but also on the magnitude of the association in younger populations, the natural history of the disease or risk factor, and whether participation is selectively affected in subjects with both the disease and risk factor of interest.

Clinical implications

The common MTHFR C677T variant has been linked with risk of migraine with aura in several studies. We found a protective effect for the TT genotype in our older population.

In a simple simulation based on reported mortality history in migraineurs from our cohort, we illustrated how even modest selective mortality (in migraineurs with aura who are TT carriers) could hide the ‘true’ effect of a risk factor when measured in an older population.

Footnotes

Acknowledgment

We would like to thank the participants of the study and the Icelandic Heart Association clinic staff for their invaluable contribution.

Funding

This study was supported by a grant from the National Institutes of Health (N01-AG-1-2100), National Institute on Aging Intramural Research Program, the Hjartavernd (the Icelandic Heart Association) and the Althingi (the Icelandic Parliament).

Conflict of interest

None declared.

References

Botto

Yang

. 5,10-Methylenetetrahydrofolate reductase gene variants and congenital anomalies: A HuGE review. Am J Epidemiol 2000; 151: 862–877.

de Bree

Verschuren

WMM

Bjorke-Monsen

. Effect of the methylenetetrahydrofolate reductase 677C->T mutation on the relations among folate intake and plasma folate and homocysteine concentrations in a general population sample. Am J Clin Nutr 2003; 77: 687–693.

Diaz-Arrastia

. Homocysteine and neurologic disease. Arch Neurol 2000; 57: 1422–1427.

Mattson

Shea

. Folate and homocysteine metabolism in neural plasticity and neurodegenerative disorders. Trends Neurosci 2003; 26: 137–146.

Scher

Tsao

. MTHFR C677T genotype as a risk factor for epilepsy including post-traumatic epilepsy in a representative military cohort. J Neurotrauma 2011; 28: 1739–1745.

Kara

Sazci

Ergul

. Association of the C677T and A1298C polymorphisms in the 5,10 methylenetetrahydrofolate reductase gene in patients with migraine risk. Brain Res Mol Brain Res 2003; 111: 84–90.

Kowa

Yasui

Takeshima

. The homozygous C677T mutation in the methylenetetrahydrofolate reductase gene is a genetic risk factor for migraine. Am J Med Genet 2000; 96: 762–764.

Lea

Ovcaric

Sundholm

. The methylenetetrahydrofolate reductase gene variant C677T influences susceptibility to migraine with aura. BMC Med 2004; 2: 3–3.

Oterino

Toriello

Valle

. The relationship between homocysteine and genes of folate-related enzymes in migraine patients. Headache 2010; 50: 99–168.

10.

Pezzini

Grassi

Del

. Migraine mediates the influence of C677T MTHFR genotypes on ischemic stroke risk with a stroke-subtype effect. Stroke 2007; 38: 3145–3151.

11.

Scher

Terwindt

Verschuren

. Migraine and MTHFR C677T genotype in a population-based sample. Ann Neurol 2006; 59: 372–375.

12.

Schürks

Zee

Buring

Kurth

. Interrelationships among the MTHFR 677C>T polymorphism, migraine, and cardiovascular disease. Neurology 2008; 71: 505–513.

13.

Todt

Freudenberg

Goebel

. MTHFR C677T polymorphism and migraine with aura. Ann Neurol 2006; 60: 621–622author reply 622–623.

14.

Liu

Menon

Colson

. Analysis of the MTHFR C677T variant with migraine phenotypes. BMC Res Notes 2010; 3: 213–213.

15.

Tietjen

Herial

Utley

. Association of von Willebrand factor activity with ACE I/D and MTHFR C677T polymorphisms in migraine. Cephalalgia 2009; 29: 960–968.

16.

Joshi

Pradhan

Mittal

. Role of the ACE ID and MTHFR C677T polymorphisms in genetic susceptibility of migraine in a north Indian population. J Neurol Sci 2009; 277: 133–137.

17.

Ishii

Shimizu

Sakairi

. MAOA, MTHFR, and TNF-beta genes polymorphisms and personality traits in the pathogenesis of migraine. Mol Cell Biochem 2012; 363: 357–366.

18.

Samaan

Gaysina

Cohen-Woods

. Methylenetetrahydrofolate reductase gene variant (MTHFR C677T) and migraine: A case control study and meta-analysis. BMC Neurol 2011; 11: 66–66.

19.

Ferro

Castro

Lemos

. The C677T polymorphism in MTHFR is not associated with migraine in Portugal. Disease Markers 2008; 25: 107–113.

20.

Kaunisto

Kallela

Hamalainen

. Testing of variants of the MTHFR and ESR1 genes in 1798 Finnish individuals fails to confirm the association with migraine with aura. Cephalalgia 2006; 26: 1462–1472.

21.

Maher

Griffiths

. Identification of molecular genetic factors that influence migraine. Mol Genet Genomics 2011; 285: 433–446.

22.

Schürks

Rist

Kurth

. MTHFR 677C>T and ACE D/I polymorphisms in migraine: A systematic review and meta-analysis. Headache 2010; 50: 588–599.

23.

Gudmundsson

Scher

Aspelund

. Migraine with aura and risk of cardiovascular and all cause mortality in men and women: Prospective cohort study. BMJ 2010; 341: c3966–c3966.

24.

Yang

Bailey

Clarke

. Prospective study of methylenetetrahydrofolate reductase (MTHFR) variant C677T and risk of all-cause and cardiovascular disease mortality among 6000 US adults. Am J Clin Nutr 2012; 95: 1245–1253.

25.

Anderson

Nalls

Biffi

. The effect of survival bias on case-control genetic association studies of highly lethal diseases. Circ Cardiovasc Genet 2011; 4: 188–196.

26.

Sigurdsson

Thorgeirsson

Sigvaldason

Sigfusson

. Unrecognized myocardial infarction: epidemiology, clinical characteristics, and the prognostic role of angina pectoris. The Reykjavik Study. Ann Intern Med 1995; 122: 96–102.

27.

Harris TB, Launer LJ, Eiriksdottir G, et al. Age, Gene/environment susceptibility – Reykjavik Study: Multidisciplinary applied phenomics. Am J Epidemiol 2007: kwk115.

28.

Qiu

Cotch

Sigurdsson

. Retinal and cerebral microvascular signs and diabetes: The age, gene/environment susceptibility-Reykjavik study. Diabetes 2008; 57: 1645–1650.

29.

Eiriksdottir

Aspelund

Bjarnadottir

. Apolipoprotein E genotype and statins affect CRP levels through independent and different mechanisms: AGES-Reykjavik Study. Atherosclerosis 2006; 186: 222–224.

30.

Saczynski

Jonsdottir

Garcia

. Cognitive impairment: An increasingly important complication of type 2 diabetes: the age, gene/environment susceptibility–Reykjavik study. Am J Epidemiol 2008; 168: 1132–1139.

31.

Kurth

Gaziano

Cook

. Migraine and risk of cardiovascular disease in women. JAMA 2006; 296: 283–291.

32.

Liew

Wang

Mitchell

. Migraine and coronary heart disease mortality: A prospective cohort study. Cephalalgia 2007; 27: 368–371.

33.

Kurth

Kase

Schurks

. Migraine and risk of haemorrhagic stroke in women: Prospective cohort study. BMJ 2010; 341: c3659–c3659.

34.

Schürks

Rist

Shapiro

Kurth

. Migraine and mortality: A systematic review and meta-analysis. Cephalalgia 2011; 31: 1301–1314.

35.

Last

A dictionary of epidemiology New York: Oxford University Press, 1995.

36.

Vineis

McMichael

. Bias and confounding in molecular epidemiological studies: Special considerations. Carcinogenesis 1998; 19: 2063–2067.

37.

De Bree

Verschuren

Blom

. Coronary heart disease mortality, plasma homocysteine, and B-vitamins: A prospective study. Atherosclerosis 2003; 166: 369–377.

38.

De Bree

Verschuren

Kromhout

. Homocysteine determinants and the evidence to what extent homocysteine determines the risk of coronary heart disease. Pharmacol Rev 2002; 54: 599–618.

39.

Holmes

Newcombe

Hubacek

. Effect modification by population dietary folate on the association between MTHFR genotype, homocysteine, and stroke risk: A meta-analysis of genetic studies and randomised trials. Lancet 2011; 378: 584–594.

40.

Tsai

Loria

Cao

. Clinical utility of genotyping the 677C>T variant of methylenetetrahydrofolate reductase in humans is decreased in the post-folic acid fortification era. J Nutr 2009; 139: 33–37.

41.

Rexrode

Lee

Cook

. Baseline characteristics of participants in the Women's Health Study. J Women's Health Gender-based Med 2000; 9: 19–27.

42.

Schürks

Zee

Buring

Kurth

. MTHFR 677C->T and ACE D/I polymorphisms and migraine attack frequency in women. Cephalalgia 2010; 30: 447–456.

43.

Hessner

Dinauer

Kwiatkowski

. Age-dependent prevalence of vascular disease-associated polymorphisms among 2689 volunteer blood donors. Clin Chem 2001; 47: 1879–1884.

44.

Matsushita

Muramatsu

Arai

. The frequency of the methylenetetrahydrofolate reductase-gene mutation varies with age in the normal population. Am J Hum Genet 1997; 61: 1459–1460.

45.

Brattstrom

Zhang

Hurtig

. A common methylenetetrahydrofolate reductase gene mutation and longevity. Atherosclerosis 1998; 141: 315–319.

46.

Galinsky

Tysoe

Brayne

. Analysis of the apo E/apo C-I, angiotensin converting enzyme and methylenetetrahydrofolate reductase genes as candidates affecting human longevity. Atherosclerosis 1997; 129: 177–183.

47.

Heijmans

Gussekloo

Kluft

. Mortality risk in men is associated with a common mutation in the methylene-tetrahydrofolate reductase gene (MTHFR). Eur J Hum Genet 1999; 7: 197–204.

48.

Harmon

McMaster

Shields

. MTHFR thermolabile genotype frequencies and longevity in Northern Ireland. Atherosclerosis 1997; 131: 137–138.

49.

Faure-Delanef

Quere

Chasse

. Methylenetetrahydrofolate reductase thermolabile variant and human longevity. Am J Hum Genet 1997; 60: 999–1001.