MTHFR: the variant we refuse to report as a clinical finding

What MTHFR does

Folate, vitamin B9, comes in several chemical forms, and the body keeps converting one into another. Folate from food and folic acid from supplements and fortified flour both feed into the cycle; folic acid first needs an enzyme called DHFR to make it usable^[5]. MTHFR performs one step in that cycle. It turns 5,10-methylenetetrahydrofolate into 5-methyltetrahydrofolate, often shortened to methylfolate, which then gives up its methyl group, through vitamin B12, so that homocysteine can be recycled into the amino acid methionine^[1].

The two variants the market talks about change one amino acid each. C677T, which current naming calls c.665C>T, swaps an alanine for a valine at position 222 of the protein; A1298C, now c.1286A>C, swaps a glutamate for an alanine at position 429^{[2, 3]}. C677T is the better studied of the two and the one with a measurable effect on homocysteine, which is why most of this article is about it. In a raw data file both go by their reference numbers, rs1801133 for C677T and rs1801131 for A1298C.

There is a real MTHFR disease, and it is a different thing. Severe MTHFR deficiency comes from rare variants that cause enzyme deficiency and very high homocysteine; the OMIM catalogue lists ten of them^[6]. Rare genetic defects that push homocysteine that high do cause coronary heart disease^[7]. The common variants are not those, and the evidence for one does not carry over to the other.

Two copies of C677T are ordinary

Marketing copy tends to present an MTHFR result as a discovery. The population numbers say otherwise. gnomAD, whose genome set is built on about 76,000 sequenced people^[8], puts C677T on 23.5 percent of Finnish chromosomes and 33.7 percent of non-Finnish European ones^[2]. Counted as people rather than chromosomes, 303 of the 5,296 Finns sequenced there carry two copies, 5.7 percent, against 11.3 percent of other Europeans.

Finns carry C677T less often than other Europeans, and the admixed American and Ashkenazi Jewish groups carry it most often. None of these numbers marks a group at risk. They show how much of the world carries a variant that the marketing presents as rare bad luck.

A1298C follows its own map. It is commonest in South Asian genomes, on 40.8 percent of chromosomes, and Finns carry it about as often as other Europeans^[3].

The Finnish count can be checked with nothing more than the allele frequency.

Folate decides how much homocysteine moves

The one well-established effect of C677T is on homocysteine, and its size depends on how much folate a population eats. In 2012 Clarke and colleagues pooled 70 studies of genotype and homocysteine in 68,369 people and sorted them by probable folate status^[7]. Where food was not fortified, in East Asia, people with two copies had homocysteine about 25 percent higher than people with none. In the US, Australia and New Zealand after folic-acid fortification began, the gap was 7 percent, with a wide margin of error. One copy moves homocysteine only about a quarter as far as two^[7].

A second pooled analysis, by Holmes and colleagues in The Lancet, put the same pattern in micromoles per litre. Across 237 data sets, two copies of C677T raised homocysteine by 3.12 µmol/L in low-folate Asia and by 0.13 µmol/L in regions with high folate from fortification, a difference indistinguishable from zero^[10]. The variant matters most where the diet is short of folate, and folate, from food or folic acid, narrows the gap.

Higher homocysteine did not mean more heart disease

For a while, homocysteine looked like a cause of heart disease that could be fixed with a vitamin. Prospective studies linked higher levels to more coronary disease, and the lowering that folic acid achieves was associated with an 11 percent lower risk^[7]. C677T then offered a test of causation. People born with two copies have higher homocysteine all their lives, so if homocysteine causes heart disease, they should have more of it.

The published studies said they did: 86 of them gave an odds ratio of 1.15. Clarke's group suspected that null results had gone unpublished, and collected 19 data sets in which the variant had been genotyped for other reasons, with 48,175 coronary cases and 67,961 controls. Two copies of C677T carried an odds ratio of 1.02, with a 95 percent confidence interval of 0.98 to 1.07, and 1.01 in low-folate populations, where the homocysteine effect is largest^[7]. The authors concluded that lifelong moderate elevation of homocysteine has little or no effect on coronary disease, and that the published excess reflected publication bias or methodological problems.

The randomised trials agree. Ten large trials in 50,378 people lowered homocysteine by about a quarter with B vitamins for around five years, and coronary events were unchanged, a rate ratio of 1.02^[7]. For stroke the picture has one loose end. Thirteen trials found no clear benefit, a relative risk of 0.94 with a confidence interval of 0.85 to 1.04, but in low-folate Asia two copies of C677T were associated with an odds ratio for stroke of 1.68, and no trial has tested homocysteine lowering only in such a region^[10].

Clots, miscarriage and the ACMG statements

MTHFR testing spread in clinics through thrombophilia panels, the tests ordered after a blood clot or repeated pregnancy loss. The ACMG reviewed that practice in 2013. Its practice guideline says that recent meta-analyses had disproven an association between MTHFR status and venous thromboembolism, and between raised homocysteine and coronary heart disease, and it concluded that MTHFR polymorphism testing has minimal clinical utility and should not be ordered as part of a routine evaluation for thrombophilia^[4]. The same guideline covers recurrent pregnancy loss^[4].

In 2020 the ACMG board reaffirmed the document, and the college reclassified it as a clinical practice resource because it no longer met the college's criteria for an evidence-based practice guideline^[11]. A 2024 review lists the American College of Obstetricians and Gynecologists, the College of American Pathologists, the American Academy of Family Physicians and the American Heart Association among the bodies that discourage the test^[6].

Depression and ADHD: what genome-wide studies found

The newer claims are psychiatric: MTHFR as a cause of depression, anxiety or ADHD, usually with methylfolate as the remedy. They grew out of candidate-gene research, small studies that picked a gene for its biology and tested one variant in it. In 2019 Border and colleagues tested the 18 most studied candidate genes for depression in samples of up to 443,264 people. MTHFR's C677T was one of them. Its strongest result across eight models had a p-value of 0.034, short of the study's threshold of 0.05 divided by 16, and as a set the 18 genes were no more associated with depression than genes picked at random^[12].

Genome-wide studies ask the question without picking genes first, and they have grown large. A 2019 meta-analysis of 807,553 people found 102 variants associated with depression^[13]. The Psychiatric Genomics Consortium's 2025 study, with 688,808 people with major depression and 4,364,225 controls, found 697^[14]. The largest ADHD study, 38,691 people with ADHD and 186,843 controls, found 27 loci^[15]. We checked every lead variant on chromosome 1 in all three. In both depression studies the nearest to MTHFR is about 3.4 million bases away; the 2025 study's nearest on the other side is 4.3 million bases away, and the only ADHD locus on chromosome 1 is 32 million bases away^{[13, 16, 15]}.

A genome-wide study cannot rule out an effect too small to reach its threshold. It does rule out an effect of the size the candidate-gene papers reported: Border's team calculated that their samples had more than 99.99 percent power to detect one^[12]. A variant carried by four to six Europeans in ten, by our arithmetic on gnomAD, with no effect that studies of hundreds of thousands can detect, is a poor explanation for one person's depression.

C677T in half a million Finns

FinnGen links the genotypes of about half a million Finns to national health registries and publishes its results^[17]. We looked up C677T and A1298C in its twelfth release for the conditions the MTHFR claims name. The results are per copy of each variant, the way FinnGen reports them.

Nothing here comes near the genome-wide threshold of 5 × 10⁻⁸, and the two smallest p-values, both for venous thromboembolism, fail even a simple correction for the 18 tests in the table, 0.05 divided by 18, or 0.0028. That correction is our arithmetic. They also point in opposite directions: C677T was slightly rarer in people with clots, A1298C slightly commoner. Odds ratios of 0.97 and 1.03 per copy are too small to matter to anyone if they are real^[9].

Per-copy results average one copy and two, so they can blur an effect that needs two copies. That is the limit of this table. The two-copy question for heart disease is the one Clarke's 19 data sets answered directly, and the answer there was 1.02^[7].

The methylfolate pitch

The commercial end of the myth is a product swap. If MTHFR is slow, the argument goes, folic acid gets stuck before the MTHFR step, so people with the variant need methylfolate, which enters after it. The diagram at the top of this page shows where the idea comes from. The trouble is the premise: a slower enzyme is still a working enzyme, and the folate measurements show it.

What the swap looks like in practice is documented. A 2023 case series collected 12 infertility patients with MTHFR variants from six practitioners recruited through an MTHFR education site. All 12 had been told to stop folic acid and were prescribed methylfolate, alone or with folinic acid, at an average daily maximum of 2,325 µg^[5]. The series had no comparison group, and its own authors write that the absence of substantive clinical research in humans limits how far the evidence transfers to practice^[5].

Pharmacogenomics has looked at the pairing too. CPIC, the consortium that writes gene-based prescribing guidelines, lists MTHFR with l-methylfolate at evidence level B/C and publishes no recommendation for it, as it does for MTHFR with methotrexate and carboplatin^[18].

What our report does with MTHFR

We read MTHFR where a file allows it and say what we found, without promoting it to a clinical finding. Where each piece of MTHFR appears depends on the file.

Excluding the whole gene has a cost. A rare variant of the kind behind severe MTHFR deficiency would not be reported in the clinical-findings section either. That is a gap in the report, and this page is where we say so.

# rsid      chromosome  position  genotype
rs1801131   1           11854476  TG
rs1801133   1           11856378  AG

The sample report shows the Frontier card in full, and the free file check tells you what kind of file you have before anything is paid for.