BRCA1 and BRCA2: the measured cancer risks, and what a DNA file can show

Two genes that repair broken DNA

A break through both strands of DNA is among the most harmful things that can happen to a chromosome, and dividing cells meet them routinely while copying their DNA. The most accurate way to mend one is homologous recombination: the cell uses the intact copy of the same stretch, made when the chromosome was duplicated, as a template to rebuild what was lost. BRCA1 and BRCA2 work in this one pathway at different points. Roy, Chun and Powell's review describes BRCA1 as a protein with many roles in the damage response, from checkpoint signalling to repair, and BRCA2 as the mediator of the core recombination step^[1].

One working copy of each gene is enough for a cell to repair these breaks. A person who inherits a pathogenic variant still has a working copy in every cell; what the inherited variant removes is the spare, so a single later hit to the remaining copy can leave a cell without this repair route. The inheritance pattern is autosomal dominant: one altered copy is what raises the risk, and a parent who carries it passes it to each child with a probability of one half^[4].

Breast and ovarian cancer risk, measured prospectively

A cancer risk can be reconstructed from family histories, or measured by following carriers forward in time and counting what happens. Kuchenbaecker and colleagues did the second. They pooled 6,036 BRCA1 and 3,820 BRCA2 carriers recruited between 1997 and 2011 through three consortia of family cancer studies, most of them in the UK, the Netherlands and France, and counted the cancers that appeared during a median of five years of follow-up^[2].

The age pattern differed between the genes. Breast cancer incidence rose quickly in early adulthood, levelling off from about age 30 to 40 in BRCA1 carriers and from about 40 to 50 in BRCA2 carriers, after which it stayed roughly constant at 20 to 30 cases per 1,000 women a year up to age 80^[2].

Two things moved the estimates inside the same cohort. Risk rose with the number of close relatives who had had breast cancer: women with two or more such relatives had about twice the rate of women with none, a hazard ratio of 1.99 for BRCA1 and 1.91 for BRCA2. And where in the gene the variant sat mattered. Variants outside a middle stretch of each gene, positions c.2282 to c.4071 in BRCA1 and c.2831 to c.6401 in BRCA2, carried a higher breast cancer risk, by a hazard ratio of 1.46 for BRCA1 and 1.93 for BRCA2^[2]. A single number for all carriers of a gene is an average over people whose risks differ.

How carriers were found matters too. A study of women not selected for family history compared 48,826 women with breast cancer and 50,703 controls from 30 population-based studies. Protein-truncating variants raised the odds of breast cancer by a factor of 10.57 for BRCA1 and 5.85 for BRCA2^[6]. An odds ratio is a different measure from a cumulative risk, and one cannot be turned into the other without more data. In that analysis BRCA1 and BRCA2 had the two highest odds ratios of any gene tested.

Other cancers, and men

The genes were named for breast cancer, but men inherit these variants as often as women do^[4], and for years the risks to them rested on small studies. Li and colleagues used data on 3,184 BRCA1 and 2,157 BRCA2 families from the CIMBA consortium to estimate the risks of the other cancers that had been linked to the genes^[3].

A large relative risk on a rare cancer can still be a small absolute one. Male breast cancer is uncommon, so a 44-fold increase in BRCA2 carriers starts from a low base. The study's absolute estimates to age 80 ranged from 0.4 percent for male breast cancer to about 2.5 percent for pancreatic cancer in BRCA1 carriers, and from about 2.5 percent for pancreatic cancer to 27 percent for prostate cancer in BRCA2 carriers^[3].

Founder variants: three Ashkenazi, a Finnish set

Thousands of different variants that raise risk are known across the two genes^[7]. In a population that grew from a small group of ancestors, a few of them turn up in many families who do not know they are related, because each traces back to one person.

Together the three were carried by about 2.5 percent of the Ashkenazi Jews tested, roughly 1 in 40^[8]. When the FDA authorised 23andMe's first BRCA report in 2018, it covered only these three, and the agency noted that they are found in about 2 percent of Ashkenazi Jewish women and rarely in other groups^[9].

Finland's population history left a similar pattern. Sarantaus and colleagues traced the recurrent Finnish BRCA1 and BRCA2 variants to common ancestors who lived 7 to 36 generations ago, roughly 150 to 800 years, and found the families carrying each one clustered in the parts of the country that history would predict^[10]. In a 2022 study of 1,356 unselected Finnish breast cancer patients, 1.8 percent carried a pathogenic BRCA1 or BRCA2 variant, and 70.8 percent of those carriers had one of 10 recurrent variants^[11].

Our report does not use a separate Finnish BRCA list. It looks for every pathogenic BRCA1 and BRCA2 variant ClinVar classifies with enough review, founder or not, and under its Finnish view the hereditary breast and ovarian cancer card says that it has no reliable Finnish figure to cite. The Finnish pattern our report does cover is in the Finnish disease heritage, which is a different, recessive set of conditions.

BRCA rows in chip raw data

A consumer DNA test reads a fixed list of positions with a genotyping chip. Some of those positions sit in BRCA1 and BRCA2, and they reach customers in the raw-data download. At these rare positions the chips are unreliable. In UK Biobank, where chip results could be checked against sequencing, 889 participants had a chip call for a pathogenic BRCA variant and sequencing confirmed it in 37; the chips also missed 70 of the 107 carriers whose variants sat at positions they tested^[12]. We go through why, with our own check of chip files against the same people's genomes, in why most rare chip calls are wrong.

A validated report is a different product from the raw file. 23andMe's FDA-authorised BRCA report started with the three Ashkenazi variants. In October 2023 it grew to 44 variants, which the company says account for about 30 to 40 percent of cancer-related BRCA variants in people of African American, non-Ashkenazi European and Hispanic or Latino descent, and it states that the report does not include all BRCA risk variants^[7]. The company reports more than 99 percent agreement with Sanger sequencing for each of the 44^[7]. A BRCA row elsewhere in the same raw file has had only a general quality review^[13].

Uncertain variants, and classifications that change

Sequencing also finds variants that are neither clearly harmful nor clearly harmless, and the classification of a variant can change as evidence accumulates. In 1.45 million people tested at one commercial laboratory, 7.7 percent of the unique variants first reported as of uncertain significance were later reclassified, and 91.2 percent of those moved down to benign or likely benign. Variants first reported as pathogenic or likely pathogenic were reclassified into another category 0.7 percent of the time^[14]. Inside the regions our own report reads, ClinVar changed the classification of 951 records between its May and July 2026 releases; 114 became pathogenic or likely pathogenic and 105 stopped being^[15]. Our report does not show uncertain variants as findings, and the guide to variants of uncertain significance covers what one means and how they move.

What our report reads, file by file

BRCA1 and BRCA2 sit on our clinical-findings panel, which follows the ACMG's list of genes whose secondary findings are worth reporting from a clinical genome, version 3.3, 84 genes^[16]. The guide to the ACMG list explains how genes get onto it. Each variant in a sequenced file is matched to ClinVar by its exact position and letters, and the rules below decide what reaches the page.

Two limits apply. Our report does not call deletions or duplications of whole exons in these genes, a kind of change that a list of small variants can miss entirely, as the guide to genome files shows and the Finnish MLH1 deletion behind Lynch syndrome illustrates. And a finding is only as current as the ClinVar release it was read against, which is why the report names that release.