We read with interest the article on the methylenetetrahydrofolate reductase (MTHFR) C677T polymorphism by Frederiksen et al.1 We do not agree with the hypothesis of the authors that testing C677T MTHFR is an alternative approach to establish if hyperhomocysteinemia is a cause or simply an effect of the disease. Indeed, hyperhomocysteinemia is not a monogenic condition, but a number of acquired and inherited risk factors contribute to its development. Several polymorphisms in the genes coding for the enzymes involved in the methionine metabolism are associated with elevated homocysteine levels. Furthermore, vitamin status and renal function play a crucial role in affecting homocysteine levels.2
The conclusion of Frederiksen et al on the role of homocysteine levels does not appear evidence based for these reasons: (1) they determined homocysteine levels 2 to 4 years after the end of follow-up in only a subgroup of subjects (we do not know the criteria used to select this subgroup), so it is possible that some selection biases were introduced; (2) the authors do not report any information on the vitamin status of subjects enrolled, whereas it is well documented that the effect of MTHFR polymorphism on homocysteine levels is modulated by vitamin status; (3) the authors performed a multivariate analysis adjusting for several parameters, but not for the main determinant of homocysteine levels such as creatinine levels. They concluded that hyperhomocysteinemia associated with MTHFR homozygosity is not causally related to vascular disease based on the fact that carriers of MTHFR 677TT had 25% higher homocysteine levels without an increased vascular risk. This subgroup of patients had mean homocysteine levels of 14.7 ± 0.5 μM, but we do not know how many subjects were hyperhomocysteinemic. Carriers of MTHFR homozygosity had 25% higher homocysteine levels with respect to the other genotypes, but it is unlikely that they were all hyperhomocysteinemic. On this basis, it does not appear reasonable to conclude that homocysteine is an effect of disease.
As homocysteine seems to be implicated in several diseases, it might be plausible that it is simply a marker of disease or of a metabolic derangement in which the leading actor is another parameter still unknown, but the study of Frederiksen et al does not provide the demonstration.
Mendelian randomization suggests that vascular events cause hyperhomocysteinemia, rather than vice versa
We appreciate the comments by Marcucci et al to our recent article.1 In that study, we examined 9238 individuals from the Danish general population of which 1374 and 208 during 23 years of follow-up developed ischemic cardiovascular disease (ICD) and venous thromboembolism (VTE), respectively; there were an additional 2961 Danes with ICD also enrolled. Plasma homocysteine was elevated 25% in methylenetetrahydrofolate reductase (MTHFR) 677T homozygotes versus noncarriers and 19% in ICD/VTE cases versus controls; however, hazard ratios and odds ratios for ICD and VTE for homozygotes versus noncarriers did not differ from 1.0, either in prospective or in case-control studies. We therefore concluded that ICD and VTE might cause hyperhomocysteinemia, rather than vice versa. Although Marcucci et al accept that this conclusion may indeed be true, they do not believe our data provide this demonstration. We disagree.
Our conclusion is based on the concept of Mendelian randomization, that is, the use of genotype-disease associations to make inferences about environmentally modifiable causes of disease.2,3 Mendelian randomization is the term applied to the random assortment of alleles at the time of gamete formation, which results in population distributions of genotypes that are generally independent of environmental factors.2,3 This is exactly what we observed in our study for 11 classical cardiovascular risk factors, that is, that these risk factors were equally common in individuals with either of MTHFR C677T noncarrier, heterozygote, and homozygote genotypes (Frederiksen et al1 (Tab1)). Likewise, among noncarriers, heterozygotes, and homozygotes, 59%, 58%, and 59%, respectively, were daily taking vitamins (chi-square: P = .74), while 1.8%, 1.4%, and 0.8%, respectively, had creatinine levels higher than 120 μM (P = .15). Finally, other polymorphisms in genes involved in methionine metabolism most likely will also be equally distributed between individuals with the 3 different MTHFR C677T genotypes, like we observed for factor V Leiden genotype.1 Therefore, our data clearly demonstrate the key advantage of Mendelian randomization, that is, avoidance of confounding.2,3
Marcucci et al have a number of questions that we are happy to answer: (1) The subgroup of individuals in which homocysteine levels were measured was simply those attending the 2001 to 2003 examination of the Copenhagen City Heart Study. Any selection bias introduced between this and former examinations is unlikely to influence relative homocysteine levels between genotypes. (2) As the fraction daily taking vitamins did not differ between genotype groups (previous paragraph), vitamin status is unlikely to have confounded our results. (3) As creatinine levels did not differ between genotype groups (previous paragraph), differences in renal function also are unlikely to have confounded our results. (4) Finally, among those with the MTHFR C677T noncarrier, heterozygote, and homozygote genotypes, 15%, 15%, and 30%, respectively, had plasma homocysteine levels higher than 15 μM (Refsum et al4(p19)) (chi-square: P < .0001).
In conclusion, therefore, because of Mendelian randomization generally avoiding confounding,2,3 we believe that our data suggest that ICD/VTE causes hyperhomocysteinemia, rather than vice versa.
Correspondence: Børge G. Nordestgaard, Department of Clinical Biochemistry, 54M1, Herlev University Hospital, DK-2730 Herlev, Denmark; e-mail: brno@herlevhosp.kbhamt.dk
