Hemochromatosis in your raw DNA: C282Y, H63D and the penetrance puzzle

What HFE does, and what C282Y breaks

The body controls its iron at the point of entry. The liver makes a hormone, hepcidin, that decides how much iron is absorbed from food and how much is released from storage^[4]. It acts on ferroportin, the only known exit for iron from a cell, and reduces how much iron it lets out^[5]. HFE is one of the proteins the liver uses to sense how much iron is already circulating, and it helps switch hepcidin production on^[4].

C282Y swaps one building block of the HFE protein, the cysteine at position 282, for a tyrosine, and the altered protein no longer reaches the cell surface where it works^[4]. With two copies, the liver makes too little hepcidin, ferroportin stays open, and more iron is absorbed from food than the body uses^{[4, 5]}. Every inherited form of hemochromatosis comes down to the same fault: too little hepcidin, or a ferroportin that hepcidin cannot bind^[5].

The surplus collects mainly in the liver, and also in the pancreas and heart^{[8, 5]}. It accumulates slowly: men with the common adult form usually develop symptoms between 40 and 60, women usually after menopause^[1]. The first symptoms are vague ones, tiredness, joint pain and abdominal pain; later ones include arthritis, cirrhosis, liver cancer and diabetes^[1].

Two variants, six results

Everyone has two copies of HFE, one from each parent, so a file can show none, one or two copies of each variant. The two variants almost never sit on the same copy: in both FinnGen and UK Biobank, nobody with two copies of C282Y also carried H63D, and nobody with two copies of H63D carried C282Y^[9]. That leaves six results in practice, and our report gives each its own reading.

In a raw file the two variants are rows keyed by their reference numbers, rs1800562 for C282Y and rs1799945 for H63D^{[10, 11]}. A chip export gives the two letters it measured at each row. Searching a download for those two numbers finds them if the chip had them, and not every chip does.

# rsid      chromosome  position  genotype
rs1799945   6           26091179  CC
rs1800562   6           26093141  AG

H63D and the strand problem

DNA has two strands, and each letter pairs with its complement, A with T and C with G. A chip export lists two letters per row and no strand, and genotype data sets are often unclear about which strand they used^[12]. For most positions the letters settle it. C282Y changes a G to an A; read from the other strand, the same change is a C to a T. The four letters do not overlap, so a row at that position can only be read one way.

H63D changes a C to a G, and on the other strand that is a G to a C: the same two letters. A row reading CG is one copy of H63D on either strand. The other two rows are the problem. CC means no H63D if the letters are on the reference strand and two copies if they are on the other one, and GG means the reverse. Genotype software treats C/G and A/T positions as ambiguous for this reason^[12].

Because a chip export cannot show which strand its letters came from, our report reads H63D from chip data only when the row says CG, the one genotype that means the same on both strands: one copy. A CC or GG row is marked as not examined, never as absent. That matters most for one copy of C282Y. With H63D unread, a chip cannot tell one copy of C282Y from one copy of each variant, so the report says the compound result has not been ruled out. Two copies of C282Y stand on their own: H63D cannot change that result.

A sequencing file does not have the problem. Each VCF record carries the reference genome's own letter at its position, beside the alternative^[13], so a C to G record at the H63D position has only one reading. A plain VCF lists only differences, so a position with no record is read as the reference letter, which is usually right and never proven; the report labels it as an inference. We have written about what a missing line can hide.

#CHROM  POS       ID         REF  ALT  GT
chr6    26090951  rs1799945  C    G    0/1
chr6    26092913  rs1800562  G    A    0/1

Penetrance for two copies of C282Y: from under 1 to 56 percent

Penetrance is the share of people with a genotype who go on to have the condition. For two copies of C282Y the published figures are far apart, because the condition can be counted at several depths: an iron level above a cut-off, a diagnosis in a medical record, organ damage put down to iron, or symptoms. Each study below counted a different one, in a different group of people, at a different age. The lowest figure came from 41,038 people screened at a US health appraisal clinic, 152 of whom had two copies^[14].

Six years later a Melbourne cohort of 31,192 people of northern European descent, followed for an average of 12 years, found documented iron-overload disease in 28.4 percent of homozygous men^[15]. In 2024 a UK Biobank analysis of 451,270 people projected that 56.4 percent of homozygous men would have a hemochromatosis diagnosis by 80, and concluded that homozygotes had more illness than earlier studies had shown^[16].

Three things move the number. One is the definition: a raised ferritin is common in homozygous men, 88 percent in the HEIRS screening, while organ damage is much rarer^{[17, 16]}. Another is age. In UK Biobank, 12.1 percent of homozygous men had a diagnosis at recruitment, at a mean age of 57, and 56.4 percent were projected to have one by 80^[16]; counted a third way, as the share with a diagnosis anywhere in the records, the same cohort gives 24.8 percent of homozygotes^[9]. The last is who was studied: the clinic study left out people already diagnosed, and UK Biobank volunteers were somewhat healthier than the population they came from^[16]. A review pooling several sources put the lifetime risk of cirrhosis or liver cancer at roughly 1 in 10 homozygous men whose iron goes untreated^[19].

A diagnosis is also a different thing from harm, and UK Biobank measured harm separately. By 80, homozygous men had more liver disease than men without either variant, 20.3 against 8.3 percent, more joint replacements, 27.9 against 17.1 percent, and more deaths, 33.1 against 25.4 percent. Men not yet diagnosed at recruitment showed much the same excess. Homozygous women had more liver disease, 8.9 against 6.8 percent, and more joint replacements, and no excess of deaths^[16]. In women the disease comes later, because menstruation removes iron^[9], and symptoms usually begin after menopause^[1].

H63D measures milder than it sounds

H63D is the more common of the two variants and the weaker. A population study in Busselton, Western Australia, took 2,531 people without C282Y: 62 had two copies of H63D and 711 had one. H63D raised transferrin saturation, and men with two copies were more likely than men without it to have both a high saturation and a ferritin of 300 µg/L or more, 9 against 0.7 percent. None had clinically significant iron overload^{[20, 16]}.

UK Biobank's figures point the same way. A hemochromatosis diagnosis by 80 was projected for 1.9 percent of men and 0.6 percent of women with two copies of H63D, and for 5.4 and 2.7 percent of people with one copy of each variant. Neither group had an excess of deaths, liver disease or any other outcome studied that survived the study's correction for multiple testing^[16]. A 2022 working group of the BIOIRON Society, the international society for iron research, wrote that the compound genotype has "minimal or no clinical penetrance" and cannot be taken as diagnostic^[21].

Finland carries C282Y at about half the European rate

C282Y is a northern European variant, and it is not spread evenly even there. In the gnomAD database, 3.6 percent of Finnish chromosomes carry it, against 6.5 percent of non-Finnish European ones; among East Asians it is close to absent^[24]. FinnGen, with genotypes and health records for about 420,000 Finns, measured 3.7 percent, against 7.4 percent in UK Biobank^[9]. H63D goes the same way, 9.6 against 15.0 percent^[25].

Within Finland the blood service found C282Y in every region, most often in Åland and South Ostrobothnia and least often in Kainuu, the region where the same study found rare Finnish-enriched disease variants most concentrated^[27]. A Finn with two copies is about as likely to be diagnosed as a Briton with two: 20.2 percent of homozygotes in FinnGen had a registry diagnosis, against 24.8 percent in UK Biobank counted the same way^[9].

Diagnoses overall were rarer in Finland, 0.1 percent of FinnGen participants against 0.4 percent in UK Biobank, and the FinnGen authors read the gap as a sign that the condition may be under-recognised in Finland^[9]. The gap alone does not settle that. A recessive genotype at half the allele frequency is expected to be about four times rarer, because the frequency is squared, and four times is the gap observed. That reading is our arithmetic; the authors also drew on what earlier diagnoses predicted in their models^[9].

Where the genotype stops and the blood test starts

A genotype says what could happen to iron levels; blood measurements say what has. For people with two copies of C282Y, the European Association for the Study of the Liver's 2022 guideline treats a transferrin saturation above 45 percent with ferritin above 200 µg/L in women, or above 50 percent with ferritin above 300 µg/L in men and postmenopausal women, as enough to diagnose hemochromatosis. For other HFE genotypes it requires iron in the liver to be shown by MRI or biopsy^[8]. The same guideline describes phlebotomy, the regular removal of blood, as the treatment, and says early diagnosis and treatment can prevent cirrhosis, liver cancer, diabetes and joint disease^[8].

Where the genotype was found first, the iron was often already up. In the HEIRS screening study, 88 percent of homozygous men and 57 percent of homozygous women who had never been diagnosed had ferritin above the study's cut-offs^[17]. At Geisinger, a US health system whose biobank returns selected sequencing results to participants, 201 people learned they had two copies of C282Y. Fifty-seven already had a diagnosis; of the other 144, 53 turned out to have iron overload^[28].

What our report says, file by file

Every report built from a chip export, a VCF or a gVCF has an HFE card. What it can say depends on the file, and the card names the variant it could not read rather than leaving it out.

Beside the genotype the card gives the penetrance note and says plainly that the result is a risk factor, not a diagnosis and not a measurement of iron. The sample report shows the card in full, and the free file check tells you whether a file is a chip export or a sequencing file before anything is paid for.