Carrier screening for couples: the one-in-four arithmetic and what a test can find

Two carriers and four outcomes

Each parent passes on one of their two copies of a gene, chosen at random. For two carriers of the same recessive condition that gives four equally likely combinations, and only one of them puts a disease-causing copy from each parent into the child. The other three give a child with at least one working copy: two of them a carrier like the parents, one a child who carries nothing.

The one in four applies to each pregnancy separately. The chance does not run down over a family, so two carriers can have several children with the condition or none^[1].

Genes on the X chromosome follow other rules. A woman who carries a variant on one of her two X chromosomes passes it to each son with a 50 percent chance, and a son who inherits it can have the condition, whatever the father carries^[3]. The one-in-four figure needs both partners, and for X-linked genes it does not apply. Our page on G6PD, an X-linked gene on our carrier panel, shows how one copy behaves in men and in women.

From three conditions to a thousand genes

Carrier screening began about fifty years ago with single conditions in single communities: Tay–Sachs disease among Ashkenazi Jews in the 1970s, sickle cell disease in Black individuals. Cystic fibrosis was the first condition American guidelines recommended screening for regardless of ancestry, and spinal muscular atrophy followed^[4]. We have pages on Tay–Sachs, sickle cell disease and cystic fibrosis.

Sequencing made a hundred genes nearly as cheap to read as one. In 2017 the American College of Obstetricians and Gynecologists called ethnic-specific, panethnic and expanded screening all acceptable, and listed criteria a condition on an expanded panel should meet several of, among them a carrier frequency of 1 in 100 or more, a detrimental effect on quality of life and an onset early in life^[5]. ACOG reaffirmed the statement in 2025.

In 2021 the American College of Medical Genetics and Genomics (ACMG) replaced the nonspecific word expanded with four tiers. Tier 1 is cystic fibrosis and spinal muscular atrophy for everyone, plus whatever family history suggests. Tier 2 adds conditions with a carrier frequency of at least 1 in 100, Tier 3 lowers the line to 1 in 200, and Tier 4 has no lower limit. Its list for Tier 3 has 97 autosomal recessive genes and 16 X-linked ones, and the college recommends offering Tier 3 to everyone pregnant or planning a pregnancy^[4]. It also states the limit every screen shares:

Research programmes go further still. Mackenzie's Mission, a carrier screening study funded by the Australian government, chose its genes through review by 16 clinical geneticists and kept 1,300: conditions that are life-limiting or disabling with onset in childhood, or ones where early diagnosis changes the outcome^[6]. It left out genes for hearing loss without other features, which shows that a gene list is a set of judgements as well as a set of frequencies. Our carrier panel includes connexin-26 hearing loss.

How many couples a screen finds

Mackenzie's Mission is the largest measured answer. Health care providers across Australia offered screening before or early in pregnancy, and 9,107 of the 10,038 enrolled couples completed it^[7]. The study screened both partners at once and reported one result for the couple: low chance, or increased chance for a named condition. Each person's own carrier status was not reported, because with a list that long most people carry something, while around 98 percent of couples do not share a carrier gene^[8].

Most couples screened learn that they share no carrier gene on the list; a few learn something no family history would have shown.

Other figures come from population databases rather than from couples tested. They count how often two people picked at random would both carry a variant in the same gene, and they depend on the gene list, on which variants count, and on the database version:

The two Finnish estimates differ twofold. The newer one counts many more genes in a database about five times larger, and counts only ClinVar's pathogenic classifications where the older one also counted loss-of-function and founder variants^{[2, 10]}. The longer the list, the more couples a screen finds. Modelling the results of 346,790 screened people, Haque and colleagues found that for couples of Northern European ancestry the panels then recommended by guidelines would identify 55.2 hypothetical pregnancies per 100,000 at risk of a severe or profound condition, and an expanded panel of up to 94 conditions 159.2^[11].

Why Finland needs its own list

Which variants a population carries depends on its history, and our guide to the Finnish Disease Heritage tells Finland's. For couple screening the consequence is practical. Kandolin and colleagues found 18 heritage genes carried by at least 0.5 percent of Finns that are not on ACMG's Tier 3 list^[2]. A panel built on American frequencies would miss carriers of congenital nephrosis of the Finnish type or Salla disease, among others.

Mackenzie's Mission chose its list for Australian health care and Australian values^[8], and HUS put the number of genes with a Finnish carrier frequency above 0.5 percent at 104, of which only 26 belong to the heritage^[9]. A Finnish panel would need both: the heritage founder variants, and the commoner genes the heritage list was never meant to cover.

Finland's public system

Predictive carrier screening for couples is not offered in Finland. HUS said so when it published its estimate in November 2024, adding that the results allow Finland to consider whether carrier couples could be identified in advance here as well^[9]. What the public system does offer is carrier testing for families where a variant is already known. Finnish has a word for each: kantajatutkimus, a test of one gene because a family variant has been identified, and kantajaseulonta, screening of several genes at once in healthy people or couples. Researchers at the University of Helsinki and HUS drew that line in a 2025 survey asking people with recessive conditions and their relatives how screening should be organised, if it is one day offered to people of family-forming age^[3].

At HUS, a person whose family has not been studied before reaches the clinical genetics unit with a doctor's referral, which a health centre, a private doctor, occupational health or a specialist clinic can write. If a relative has already been studied there, the person can call the unit directly. A healthy relative at risk of carrying a known family variant can be offered a predictive gene test. There is a fee for the visit, the gene tests are not charged separately, patients are seen within the six months the care guarantee allows, and results can take several months^[13].

The government decree on screening lists three prenatal screens, an early ultrasound, a screen for chromosomal anomalies and an ultrasound for severe structural anomalies^[14]; none of them tests the parents' carrier status.

The limits of chip data

A consumer chip from 23andMe, AncestryDNA or MyHeritage reads a fixed list of positions chosen in advance, most of them common variants. Carrier variants are individually rare, and chips read rare variants badly. In UK Biobank only 16 percent of chip calls for variants rarer than 1 in 100,000 were confirmed by sequencing, and 20 of 21 consumer chip files the same study checked had at least one rare disease-causing variant called in error^[15]. Our guide to false positives in raw data works through the measurements.

A chip also cannot see the variants that clinical screens handle with separate assays. Mackenzie's Mission measured the copy number of SMN1 for spinal muscular atrophy and the repeat length in FMR1 for fragile X syndrome with dedicated tests^[8]; neither is a short change that a chip export or a standard VCF records, and our report lists both conditions as gaps. Comparing two chip files for a shared carrier gene therefore says little: on the UK Biobank figures a shared rare variant is more likely a calling error than a real one, and finding none covers only the few positions both chips read.

What Aimosti's partner view shows

The partner view joins two people's carrier results, each from their own whole-genome report. There are two ways in. Each partner can upload their own file to their own account and link the two by an invitation by email, or one partner can add the other's file after the other confirms by email; nothing of theirs is read before they do. Either partner can withdraw, and the view ends for both.

The table covers the 56 genes of the curated carrier panel, 26 of them Finnish Disease Heritage genes, and a wider screen of ClinVar's listed disease-causing variants in 1,363 more recessive and X-linked genes. Only findings with at least two-star review in ClinVar are joined; single-submitter matches and founder variants ClinVar has not classified stay on each person's own report.

Where both partners carry a single variant in the same recessive gene, the row states the one-in-four pattern and hands the question to a genetic counsellor. It gives no fraction where one partner has two variants in the gene, for an X-linked gene, or where the gene or variant is also linked to a dominant condition, because the one-in-four figure assumes two ordinary carriers. A summary page can be printed for a genetic counsellor.

It differs from Mackenzie's Mission in one design choice: each person's full report stays theirs, including carrier findings in genes the other does not share, and the partner sees only the carrier results. It also lacks what a clinical screen adds: it reads no SMN1 copy number and no FMR1 repeats, it does not classify variants ClinVar has not classified, and the Deep Read deletion screen for the Finnish deletions in TYROBP and CLN3 stays on the individual report. The couple pack is two genome-report credits bought together, one for each partner, and does not include Deep Read. A separate couple outlook, which both partners have to switch on, does arithmetic on harmless traits and carries no health result.