Most of this case study is also published in the Blue Eyes Directory, which is a site of its own and which recomputes its figures from the register at every build. The science, the deafness evidence, the hearing data, the prevention control and the breed table are there and are kept current. What is published only here is the registry policy material: the forty-four breed policies, the thirteen registries requiring a PAX3 test, and FIFe's Maine Coon rule. Read it in the Blue Eyes Directory.

One: dominant blue eyes, six PAX3 variants, and the registry rules written for them

Dominant blue eyes provide a well documented test of whether new registry requirements are coherent. Registries acted on the trait quickly and in public across 2025 and 2026, and the genetics were described in the peer-reviewed literature over the same period, so the rules and the evidence can be compared directly.

What the science says about dominant blue eyes

Six dominant blue eye variants have been identified, all in the PAX3 gene. Four are published. DBE^CEL (Roxi) is a retroviral insertion that arose in Kazakhstan and was used to found the Céleste breed. DBE^RE is a nonsense variant in a Maine Coon line. DBE^ALT is a second and different retroviral insertion, in the population behind the Altai breed, which is bred today under a World Cat Federation standard. DBE^AGO is a frameshift in a further Maine Coon line. Two more were identified in Chinese domestic shorthairs during 2025 and are not yet published: DBE^cDc in August 2025, since detected in the Abyssinian, and DBE^xHW in September 2025. The laboratory that identified both sells a panel covering five variants, and DBE^xHW is not among them. DBE^xHW is therefore identified but not commercially available, and it is the only one of the six variants in that position.

What those three terms mean. They describe three different ways of breaking the same gene, and the distinction matters because a test for one does not detect another.

• Retroviral insertion. A block of viral DNA, left in the cat genome by an infection in some distant ancestor, has copied itself into the middle of the gene. It is physically large, several hundred letters of DNA, and it interrupts the gene's instructions. DBE^CEL (Roxi) and DBE^ALT are each an insertion of this kind, at the same gene but not the same insertion.
• Nonsense variant. A single letter of DNA is changed, and the change happens to create a stop signal in the middle of the instructions. The cell stops reading there, so the protein is made only up to that point. DBE^RE is of this kind.
• Frameshift. DNA is read in groups of three letters, each group naming one building block of the protein. A frameshift is an insertion or deletion of a number of letters that is not a multiple of three, so everything after it is read in the wrong groupings and the remainder of the instruction becomes nonsense. DBE^AGO is of this kind.

All three have the same practical effect, that one of the cat's two copies of PAX3 no longer produces a working product, which is why a single copy is enough to change the eye colour. They are different events at the level of the DNA, so each needs its own laboratory assay. That is the reason no single test covers all six, and the reason a rule requiring only a PAX3 test does not specify what has actually been looked for.

All six identified variants are in PAX3. That is a fact about the six variants that have been found so far, and it does not mean PAX3 is the cause of every blue eye. In the nine unmapped lines nobody has yet found a variant at all, so nobody knows which gene is involved in them. It may be PAX3 and it may be a different gene entirely.

The consequence for a breeder is direct. A clear PAX3 panel on a cat from one of those lines does not explain why that cat has blue eyes, and it does not mean the cat will not pass the trait on. It means only that the cat does not carry the six variants that have been identified.

At least six further lines segregate variants that have not been identified: among them the Nadeya, Igor, Marusya, Malvina, Pillowtalk-Foreigner and Gosha lines, distributed across British, Scottish, Selkirk, Persian, Exotic, Siberian, Maine Coon, Devon Rex and Munchkin cats.

The six variants at a glance

The same facts are set out for each variant in the same order, so that any two can be compared directly. Green marks a variant for which a test can be bought.

Reading the table across rather than down gives three results.

• Fyodor, a black self with a white tail tip and blue eyes, is not the same animal as Nikita'l Seymour, the silver tabby of the Topaz after whom the British carriers of DBE^ALT are named as a lineage in Abitbol et al., Animals 2024;14:1845, Table 2. The two are separate cats carrying one variant. DBE^CEL (Roxi) and DBE^ALT arose in the same town, in the same decade and in the same breeding programme: the two cats these variants are traced to, Roxi Darlin and Fyodor, were both in the Darlin cattery at Ust-Kamenogorsk in the late 1990s. Because both cats sat in one programme, the two variants have been confused ever since. That is the origin of the belief that the Céleste carries the Altai variant.
• Three of the four published variants are named after the line or population they arose in. One is not. DBE^RE takes its name from the Rociri Elvis line and DBE^AGO from the Agostino line, each being the line the variant appeared in. DBE^ALT takes its name from the Altai, the population it appeared in. The exception is DBE^CEL (Roxi), which is named after the Céleste, a breed begun in 2015 from the descendants of the cat the variant came from and recognised by LOOF on 1 January 2023. The variant is therefore named after something that did not exist when it arose, which is why a Céleste in the pedigree cannot be inferred from a DBE^CEL (Roxi) result. Case study two sets out the consequences.
• What two copies does is known for two of the six and unknown for the other four. Four homozygotes have been genotyped, and all four were white. Three carried DBE^ALT, two Altai and one Nanotigr, viable, with more white than their parents and deafness in some of them, in Abitbol et al., Animals 2024;14:1845, Table 2. One carried DBE^CEL (Roxi), a kitten born to two Céleste parents that died at birth of cleft palate, in Abitbol, Couronné et al., Animal Genetics 2024;55:670-675, Table 1. DBE^RE is inferred embryonic lethal from the absence of homozygotes in pedigrees, and none has been observed. For DBE^AGO, DBE^cDc and DBE^xHW there is nothing at all, and DBE^AGO has been published since June 2025.

Waardenburg syndrome, and why the subtype does not fit the gene

Dominant blue eyes in cats is routinely described as a feline Waardenburg syndrome. Waardenburg syndrome in humans is not one condition but a set of them, separated into numbered types by which gene is mutated and distinguished clinically by which signs are present. The type a registry names therefore carries a claim about the gene, whether or not it intends to.

The rows this case study turns on are marked. Types 1 and 3 are the PAX3 types, and both carry dystopia canthorum. Type 2A is an MITF type, and it is defined by the absence of that same sign. Sources for the classification: Waardenburg Syndrome, StatPearls, NCBI Bookshelf; and the NIH Genetic Testing Registry entry for Waardenburg syndrome type 2A, which gives the condition as pigmentary abnormality of hair, skin and eyes, congenital sensorineural hearing loss, and “the absence of dystopia canthorum”.

Human Waardenburg syndrome arises from mutations in PAX3, MITF, SNAI2, SOX10, EDNRB and EDN3. The point that matters here is that the numbering is not a severity scale. It is a gene classification. Naming a type names a gene, and PAX3 accounts for types 1 and 3 and for no other type.

Waardenburg syndrome entered FIFe's material through its Maine Coon Breed Council. The Health and Welfare Commission paper Dominant Blue Eyes: A Call for Responsible Breeding, March 2025, quotes the Council's recommendation that genetic testing of blue-eyed Maine Coons is “strongly encouraged to identify potential risks early such as deafness and Waardenburg Syndrome”. A FIFe Breed Council is a FIFe body established to serve the federation in an advisory capacity on health, so this is an advisory opinion carried in a FIFe publication rather than a FIFe rule. The paper names no type, and the test it names is for PAX3. Those are two different claims and the paper does not separate them. Waardenburg syndrome is not one condition but a set of them classified by gene: PAX3 gives types 1 and 3 and no other type, while subtype 2A is caused by MITF and is defined by the absence of dystopia canthorum, the sign that distinguishes type 1. Naming the syndrome without naming the type leaves the gene open, in the same sentence that has already fixed it. Six feline variants have been identified, four of them published, and every one is in PAX3.

Two further strands of evidence bear on which classification the feline sign pattern fits.

• Wide-set inner eye corners, dystopia canthorum, have been reported in cats. Abitbol et al., Animals 2024;14:1845, record the sign in some dominant blue eye Maine Coons. It is a type 1 sign and not a 2A one. The paper does not say the variant causes it: “to date, we cannot link this defect to DBE alone or to the interaction between DBE and a specific face morphology”. The face shape has been accentuated by selection in certain Maine Coon lines, so the sign may come from the head the breed is bred for rather than from the variant. The same paper records that “dystopia canthorum was not observed in adult DBE cats from the Altai, Celestial, British, Sphynx, and Siberian genetic backgrounds”.
• A second craniofacial sign is reported in the two unmapped lines. The Igor and Marusya lines show a very small, pinched nose leather. It was reported to this survey by the breeder and is verified here from photographs. The corresponding sign in human type 1 is incomplete nostril development. One limit applies: both lines run through breeds bred for a short face, the British and the Persian and Exotic, so the nose shape may come from the breed rather than from the variant. It is recorded because these are the two lines for which no variant has been found, and a sign they share is a reason to sequence them.

One caution, from the paper that identified the first feline variant. Rudd Garcés et al., G3 2024, record that the mouse model of PAX3 loss of function does not reproduce the human picture: “Unlike WS1 human patients, Sp+/− mice do not show alterations in auditory function and ear morphology when compared with wild-type animals”. The same gene gives different signs in different species. A cat resembling a human subtype therefore need not carry a mutation in the gene that causes that subtype in people, and the human numbering may not transfer to cats at all.

The conclusion is the one this case study reaches about breed names and line names: a name is not a genotype. The Waardenburg type numbers classify human genes, so naming a type makes a claim about which gene is mutated. Four feline variants have been described and all four are in PAX3. For those four the accurate description is the variant itself. For the Igor and Marusya lines no type can be claimed at all, because no variant has been found in them.

A related point of law, if the resemblance is pressed. Type 2A in humans is described as carrying no deleterious effect other than deafness, and deafness alone is enough to engage a ban: European companion animal legislation provides that deaf cats must not be deliberately bred. Sources for the human classification: Pingault et al., Human Mutation 2010;31:391–406; Tachibana et al., Pigment Cell Research 2003;16:448–454; and for the equivalent equine work, Hauswirth et al., PLoS Genetics 2012;8(4):e1002653.

The blue-eyed lines for which no variant has been identified

The lines below segregate a dominant blue eye trait for which no causal variant has been found. No test can exist for any of them until one is. They are the reason a clear panel result narrows the risk without clearing the cat.

The dispersal, by date

Everything above is organised by variant, by breed or by registry. Set out by date it makes a different point, and it is the point this case study exists to make: the trait was reported for 119 years before anyone could test for it, and five registries acted within twenty-six months of the first variant being published. Rows are graded: published means the peer-reviewed literature, registry means an organisation's own document, and breeder means a breeder or breed record not verified here.

A cat is found in the random-bred population A breeding programme begins A variant is published A registry or a legislature acts A starred event has a section of its own below

Three things are visible in the dates and in nothing else. The gap is 119 years. From the first published report of permanently blue-eyed kittens in 1905 to the first identified variant in April 2024. Everything in between was managed by eye. The dispersal front-ran the science by about a decade. Between 2015 and 2022 the trait entered pedigree breeding in at least ten breeds across nine countries, and every one of those decisions was taken with no test in existence. And the regulatory response is almost entirely younger than the science. Of the registry actions above, the GCCF's variant codes, TICA's show rule, FIFe's obligatory test and the GCCF's British refusal all fall within twenty-six months of the first publication. The exceptions are the two that came first: the WCF's 2018 standard and LOOF's 2019 single-founder rule, both written before anyone knew there was more than one variant.

The lines, the variants, and which breeds hold more than one

A pedigree names catteries, not variants. This section is the bridge between the two at the level the argument needs: which named lines exist, which of them a variant has been identified in, and which breeds hold more than one. The animal-by-animal listing that supports it is kept separately, in the Blue Eyes Directory, which is in two parts: how it is compiled and what the numbers show, and the register itself. The reason for keeping it separate is given at the end of this section.

Lines and cattery prefixes only here. A prefix is a trade name and appears on every pedigree the cattery issues. No individual is named here who is not already named in the published literature. Declaration of interest: the author of this survey breeds British Shorthairs carrying DBE^ALT under the Belle Ayr prefix, from the Roddifer founders of the Nanotigr line, and Maine Coons carrying DBE^RE under the same prefix. The interest is direct, and it applies to case study one rather than to the survey as a whole. Those are the lines that case study examines: cats bred under the Belle Ayr prefix appear in its register and in its findings, both clinical series it reports were conducted on those cats, and the author stands to be affected by the conclusions it reaches about them. The author is a co-author of Abitbol M, Cloquell A, Kaczmarska A, Holmes K, Lühken G, Macaulay K. Dominant blue eyes in Maine Coon cats: new PAX3 variant and updated phenotypic data. Animal Genetics 2025;56:e70020, the BAER follow-on study, credited with investigation, resources and project administration. Cats of the author's own lines were among those sampled for the two 2024 papers in which the variants were identified, in which the author is acknowledged rather than an author: Abitbol M, Couronné A, Dufaure de Citres C, Gache V. A PAX3 insertion in the Celestial breed and certain feline breeding lines with dominant blue eyes. Animal Genetics 2024;55(5):670-675. doi:10.1111/age.13433; and Abitbol M, Dufaure de Citres C, Rudd Garces G, Lühken G, Lyons LA, Gache V. Different Founding Effects Underlie Dominant Blue Eyes (DBE) in the Domestic Cat. Animals 2024;14(13):1845. doi:10.3390/ani14131845. Two clinical series reported in this case study are the author's own and are not peer reviewed: the BAER hearing series in fourteen British cats of a DBE^ALT line, recorded at the University of Glasgow on 17 and 24 September 2024 and reported in a dated summary statement, Abitbol M, Cloquell A, Kaczmarska A, Holmes K, Macaulay K. BAER hearing screening in British Shorthair and British Longhair cats carrying DBE^ALT: Summary Statement. Belle Ayr Cats, 3 June 2025; and the blood pressure trial in forty-three cats of the same line, reported in a dated summary statement, Abitbol M, Macaulay K. Corin Levels vs Blood Pressure Trials in British Shorthair / Longhair Cats: Summary Statement. Belle Ayr Cats, 10 September 2024. Both form part of an ongoing research programme and are intended for publication. The registry under which those cats are registered changed during the work, because a policy change at the federation previously used closed registration to the line concerned, and the certification on which the descent claims rest was obtained from The International Cat Association thereafter. That is the author's own instance of the mechanism set out in the survey under the heading what a registry certifies, and what a ban removes, and it is declared for that reason as well as for the breeding interest. That line appears in the directory on the same terms as every other, and the interest is declared here as well as with the co-authorship declared at the end of this case study. Every row carries its grade of evidence in the cell, on the same four-point scale used throughout this case study:

• Genotyped. A cat of this line has a laboratory result for the variant named.
• Wild type. Cats of this line were tested for the known variants and found to carry none of them. This is a positive finding of absence, not a failure to test.
• Pedigree. Descent traced to the founder of the line, with no genotype on the individual cat.
• Reported. A breeder account or a self-published compilation at messybeast.com, not verified here. These rows are leads for checking and not findings.

Lines in which a variant has been identified

The Selkirk Rex is recorded against the Seymour line by pedigree rather than by genotype, and appears in the breed table on that basis. Prefixes given as not established mean this survey could not trace the line to a named cattery, and not that none exists.

Lines in which no variant has been identified

These are the lines for which no test can exist until the variant is found. Two of them have been tested against every known variant and carry none, which is a stronger statement than untested and is graded accordingly.

Which variants are in which breed

What this table does not show is direction. A breed may be the origin of one variant and the recipient of another, and a mark here does not distinguish the two. The Maine Coon is the clearest case: DBE^RE and DBE^AGO both arose in it, DBE^CEL (Roxi) arrived by outcross, and DBE^ALT is reported in it as well, which is why no single test clears a cat of that breed. The breed table later in this case study separates arrival from origin row by row.

Breeds down the side, variants across the top, and the tick tells you how good the evidence is. A black tick means a cat of that breed has been genotyped for that variant. A grey tick means it is reported somewhere and this survey has not verified it. The Maine Coon against DBE^ALT is the only grey tick in the table and its source is a breeder-compiled genealogy headed “blue-eye line created using Roxi and Seymour genes”: Roxi is the DBE^CEL (Roxi) founder and Nikita'l Seymour the DBE^ALT founder, so on the compiler's own account the line was founded on both.

Set against it, no genotype for DBE^ALT in a Maine Coon has been published, and the laboratory breed list for that variant runs Altai, British Shorthair and Longhair, Exotic Shorthair and Longhair, Persian, Ragdoll and Sphynx, with the Maine Coon absent from it. The tick stays grey for that reason and the cell is the strongest case in the table for a test.

If the Maine Coon blue-eye line was founded on both variants, cats of that line can carry one copy of each. A cat with one copy each of two different variants is a compound heterozygote, which is the state in which deafness and neonatal death are documented in this case study, and no test on sale reports both at once.

An empty cell means nothing has been recorded, which is not the same as nothing being present. The count on the right is what matters: a breed with two or more ticks is a breed in which two different variants can meet without leaving the breed.

A blank cell is weaker evidence than it looks, and here is the measure of it. Three Sphynx carriers had no blue-eyed cat anywhere in their recorded pedigree. A breed can therefore hold a variant with nothing in its papers to suggest it, which is what latency means in practice and why absence from this table cannot be read as absence from the breed.

✓ black tick, genotyped in a cat of this breed  ·  ✓ grey tick, reported somewhere and not verified here  ·  empty cell, nothing recorded. The ticks are deliberately not green: a tick here records the presence of a variant that causes harm when two of them meet, and a colour that reads as approval would read backwards. Colour is carried by the count instead, where it signals how many variants a breed holds.

Six breeds and breed groups already hold two or more variants, and the four largest hold four, five, five and six. In every one of them a cross between two different variants can be made without leaving the breed and without breaking any registry rule anywhere in the world. Against that, only three breeds in the table are held to a single variant by an instrument rather than by chance: the Céleste by LOOF's rule that every cat descend from one founder, the Azul by a standard requiring one variant per breed, and the Pharaoh by a standard permitting one outcross partner.

The prefixes and the animals behind these counts

The animal-by-animal record supporting this section is held separately, in the Blue Eyes Directory, which carries one entry per cat on the same four grades of evidence used above, with views by variant, by breed, by cattery prefix and by litter. This case study argues at the level of the line and the variant; the directory enumerates animals, and it grows with every pedigree supplied. Two of its findings belong in the argument rather than in the listing.

• Thirty-seven of the three thousand six hundred and fifty-six cats in the whole documented record have a laboratory result of their own. The rest are attributed by descent from a named founder, or asserted with neither a test nor a traced descent. That is the state of the evidence this case study is obliged to work from. This count is generated from the register at each build and is not a figure typed into the text, so it moves when the register moves; a cat whose owner holds a laboratory result that has not been supplied to this survey is graded on what the survey holds, not on what exists.
• A large part of the record cannot be keyed on a cat's name, because the cats have no name. On one breeder-compiled genealogy of a founder line, sixteen of twenty-four kittens are recorded only as "Girl 1" or "Boy 2", two of them share a label, and every animal of the second generation is unnamed. Those are the animals whose descendants cannot now be traced, and they are the reason the recommendation later in this case study is addressed to registries, which hold the only records in which such animals are individually identified.

Where the Rociri Elvis line actually begins, and why it is not a spontaneous mutation. The variant is generally described as having arisen spontaneously in Rociri's Elvis, born in the Netherlands in 2015. The pedigree record does not support that. Genotyping traced it back as a latent variant carried by white cats in which it could not be seen, to a cat recorded as Justcoons Ha Shisch, and then, after skipping several generations, to a cat recorded as JW Olmocabe's Alastor, beyond which the trail is cold. Elvis is the cat in which the trait first became visible, which is a different thing from the cat in which the variant first occurred. This is the same distinction drawn for Roxi and Lazuli in case study two, and the same one that makes a line name an unreliable guide to what a cat carries.

Two attributions circulating for this line should be treated with care. Elvis's blue eyes have been attributed to a mutation of the KIT white spotting gene, which the published genotyping does not support: the causal variant is in PAX3, and none of the fifty-eight cats sampled carried the dominant white KIT variant. And one prefix reported in this line is recorded on social media as descending from a red and white cat, while the breeder states the line descends from Joyfield cats; a third account names a stud of another prefix again. None of the three has been confirmed.

A possible explanation for the deafness disagreement. The two published series on DBE^RE disagree sharply, eight of eight hearing-tested cats abnormal in the first and none of six in the second. DBE^AGO had not been described when the first series was collected. If cats in that series carried DBE^RE and DBE^AGO together, they would have had no fully functioning copy of PAX3 at all, which would explain a severe result in one series and not the other. This is a hypothesis and not a finding: no cat has been reported genotyped for both. It is recorded because it is testable on stored samples, and because it would be a compound heterozygote effect of the kind this case study is otherwise about.

The Ojos Azules, and what the one closed programme actually established

The Ojos Azules is cited constantly in this subject, usually as a warning, and the citation is almost always second hand. It is set out here because it is the only dominant blue eye programme in the world that ran to a conclusion, and because three separate threads in this case study lead back to it.

The line is traced to one cat. A dark tortoiseshell with a white throat locket, white toes on each hind foot and brilliant blue eyes was found in an animal shelter in New Mexico and named Cornflower. She was bred to unrelated non-blue-eyed males from 1984 and the kittens inherited the eye colour, which established the trait as dominant. TICA recognised the breed in 1991. By 1992 ten cats were registered. The breed is registration only today, held in TICA's Category VI for experimental breeds. The route by which it got there is less tidy than it is usually told: a motion of 28 August 2014 returned it to registration only, and motions of the following day reconsidered that and reinstated its preliminary new breed status. The decision that effected the final demotion has not been located. Cats not descended from Cornflower are not of the breed, whatever they are called.

• It had a characteristic facial type, line-traced. Every descendant of Cornflower is recorded as having wide-set eyes, distinct cheekbones and plump whisker pads, with a white tail tip regardless of coat colour, and a flattened or kinked tail tip treated as an indicator of the gene. This is the only line in this survey for which a facial phenotype is traced to a single founder rather than asserted.
• Latent carriers were identified here first, and they were not harmless. Cornflower's granddaughter was a red mackerel tabby with deep amber eyes and the breed's facial type; bred to an unrelated cat she produced four kittens, three blue-eyed and one odd-eyed, proving that a cat could carry the trait without showing it. The programme's own figures then recorded that 9.5 per cent of the amber-eyed latents had birth defects. Harm was therefore not confined to the homozygote, which is a possibility no registry rule in this survey addresses.
• Homozygotes died before birth, with cranial deformities. They are recorded as having cranial deformities, white fur and a small curled tail, and as dying in the womb. The deformities were described as resembling those of double merle dogs. It was never established whether the mutation itself or an adjacent linked gene caused them.

The programme ended for reasons of breeding economics rather than biology. The same applies to every breed built on this trait. The defects were avoidable, by never mating two carriers, and the safe mating produced roughly half non-blue-eyed kittens. Breeders left the programme because half of every litter had to be sold as pets while buyers wanted the blue-eyed cats, and because outcrossing to keep the trait safe would lose the facial type within a few generations. The instrument that prevented the harm was the same instrument that made the breed commercially unviable. That is the tension the Céleste and the Azul standards now contain, arrived at from the other direction.

The breed is treated as extinct, although cats descended from the geneticist's own animals are recorded as still being bred, with white spotting and cornflower-blue eyes and no health problem reported to date. The genetic work was never published: the only genetic study of the line was carried out in the late 1990s and early 2000s and could not be found in open sources by a later researcher, who suggests it was lost on the geneticist's death. So the account above, detailed as it is, remains a breeder and breed record rather than published evidence, which is why this survey does not treat the cranial defects as established for any other line.

Two threads that lead out of it

First, a candidate gene the survey did not have. Merle in dogs is a mutation at SILV, also called PMEL, and double merle dogs show deafness, visual defects, a high degree of white and the pattern known as lethal white. The Ojos Azules homozygote defects were described as resembling double merle. Independently of that, a laboratory study for one British Shorthair line reported an approximately 100-nucleotide deletion in PMEL alongside three variant positions in PAX3. Two unconnected pointers to the same gene is not a finding, and neither has been published. It is recorded because PMEL does not appear anywhere in the four published feline variants, and because a cat carrying a PMEL variant would return a clear result on every DBE test sold today.

Second, the Ojos Azules became a registration of last resort, and this can now be dated. Because it was for many years the only recognised breed in which a blue-eyed, non-white, non-colourpointed cat could be entered, unrelated cats were registered under its name. Three instances are recorded in this survey: the founder of the Igor line, whose pedigree was registered as Ojos Azules “simply because there were no other DBE breeds at that time”; a blue-eyed tabby found in Norway and registered as Ojos Azules in 2012, at a time when that was “the only way to register blue-eyed non-white/non-colourpoint cats”; and a British Shorthair of the Igor line registered as Ojos Azules to circumvent the British registration bar.

The sequence matters. TICA reduced the breed to registration only in 2014, which removed the breed committee, the standard and any health rule, but did not remove the registration route. A cat entered under that name after 2014 therefore enters a breed with no committee to ask questions of it. Two of the three instances above post-date 2014.

The Barnaul and Aktau lines, and the one place two unidentified variants were deliberately crossed

These two lines are set out separately because between them they answer, in a single documented sequence, most of the questions this case study asks about what happens when a variant cannot be tested for. Both were found in street cats, both were investigated by the same person, both reached pedigree breeding through the same cattery, and the two were then mated together.

Prefixes recorded in the Barnaul dispersal, in order of transfer: Slavicat to Meowrigi, then to a German breeder, to Rehaloon and Steph Mzzn in France for Minuets, to an untraced holder in Ukraine, where it founded a Scottish Fold line, to Aldmeris in Uzbekistan, to Felicious in Kazakhstan and onward to the United States, to Schastie Doma in France as a foundation cat for British chinchillas, and to Tabassum in Uzbekistan. Several of these cats carry EMS codes ending 61 or 63, the blue-eyed and odd-eyed codes, so they are registered animals with pedigrees.

A deaf stud, and the rule that would have prevented it. The deaf kitten of that litter was used for breeding and sired many kittens. The source records the contrast itself, that in Europe and the United States a deaf cat would not be bred from. That is not a matter of custom: European companion animal legislation provides that deaf cats must not be deliberately bred, and German law prohibits breeding animals with defective hearing. The animal in question was in Uzbekistan, where no such provision was located. A rule that binds in one jurisdiction does not bind the cat, and the cats of this line have since moved to Germany, France and the United States.

Why these lines exist at all is worth stating, because it bears on how many more there are. Both founders were street cats, found within a few years of each other in the same region of central Asia and noticed only because breeders in those countries had begun developing local cats into recognised breeds. The most likely explanation offered in the source is not that new variants are appearing but that they have always been present in the random-bred population and nobody was looking. That is consistent with the GCCF genetics committee's own expectation that more variants will be found, and it is the reason a test panel covering the known variants cannot be treated as clearing a cat.

A recommendation: trace the descendants, because they are the only cohort that could identify the unmapped variants

Nobody records which pedigrees still carry descendants of these matings. No registry records the variant, and for an unmapped line there is no variant to record, so the only trace that exists anywhere is the pedigree itself. That is not a gap. It is an opportunity that is closing, and this survey recommends it be taken.

Why the pedigree is the instrument here, and not a poor substitute for one

For a variant that has been identified, a pedigree is inferior to a test: a test reports what the cat carries and a pedigree reports only what it might have inherited. For a variant that has not been identified, the relation reverses. There is no test to be inferior to. Descent from a named founder is the only available evidence that a cat may carry the variant, and a pedigree recording that descent is therefore the primary record rather than a proxy for one.

Gene discovery needs exactly what a traced pedigree supplies. The requirement, as stated by one of the breeders working on it, is two defined groups: cats that carry the variant, and control cats in which it is certainly absent. If the carrier group turns out to hold several different variants with a similar appearance, the search for a shared difference fails, because there is no shared difference to find. A cohort assembled by pedigree from a single founder is the one way to be confident, before any sequencing is attempted, that every cat in the carrier group carries the same variant. That is a requirement about who is in the sample and not about how cats are bred: it asks that the animals compared all carry one variant, so that the difference the analysis looks for is there to be found. It has nothing to do with breeding two carriers together, which this case study recommends against throughout.

The cohorts that already exist, and what each would answer

These are not hypothetical groups. Each is defined by named cats, and in each case the animals either exist now or existed recently enough for samples and pedigrees to be recoverable.

What a pedigree cannot answer and a genome can

The cohorts above would identify the variants that have never been mapped. A second set of questions concerns the variant that has been mapped, DBE^ALT, and none of them can be reached from pedigrees at all. All three have established methods and definite answers.

• How old is the variant. Not how old the founder was, which is a different question and one the pedigrees answer only approximately. The age of a variant is estimated from the length of the chromosome segment still shared around it by unrelated carriers. Recombination shortens that segment at a known rate each generation, so a long shared segment indicates a recent origin and a short one an old origin. It requires sequence from carriers in several unrelated lines and it returns an interval with a stated error. The pedigrees give only a floor: the variant was transmitting by 1999.
• Whether it arose in a breed or arrived in it through an outcross. The question is the same for every breed on the list and the test is marker sharing. A variant that arose in a breed sits on that breed's own haplotype in a background of that breed's ancestry. A variant that arrived sits on a foreign haplotype in a background that is otherwise the receiving breed's. The two predict different data and the difference is measurable, so the question is answerable breed by breed rather than argued.
  Worked as an example on the case the survey has most material for: if DBE^ALT arose in the British silver lines and reached the Kazakh population, cats descended from Fyodor should carry British ancestry beyond the variant itself. If it arose in the random-bred Kazakh population and entered the British, the British carriers should show the Kazakh haplotype on an otherwise British background. The same pair of predictions can be set for the Persian, the Ragdoll, the Sphynx, the Munchkin, the Lykoi, the Exotic and the Selkirk Rex, and each is a separate answerable question.
• Did the breeds carrying it get it once or several times. DBE^ALT is genotyped in the Altai, the Topaz, the British Shorthair and Longhair, the Persian, the Ragdoll, the Sphynx, the Munchkin, the Lykoi and random-bred domestic cats, is held by descent in the Selkirk Rex, and the commercial breed list adds the Exotic Shorthair and Longhair. If every one of those carries the same haplotype, they took it from one source and the dispersal is a single event traced through pedigrees. If the haplotypes differ, the variant entered more than once, and the breed lists in circulation are describing several introductions as though they were one.

Shared ancestry across the whole genome is what separates origin from transfer. A variant carried on a foreign haplotype in an otherwise pure background arrived; a variant carried on the background's own haplotype arose there. That is the only evidence that can settle where this variant came from, and no amount of pedigree work substitutes for it, because a pedigree records what was declared and cannot exclude what was not.

All three require samples from carriers in several countries, and some of those lines are now closed to registration in the registries that hold their records. A cat a registry will not register is a cat that registry holds no record of. Nobody can then say who its parents were, which line it belongs to, or where to find its relatives, and a sample without that information cannot be used: the comparison depends on knowing which line each animal comes from. So the study cannot be assembled from any one country's cats, and the lines it most needs are the ones that have been closed to registration. That is what the research registration class recommended elsewhere in this survey is for.

Why this is urgent rather than merely desirable

The records are being lost now, and four instances of it are documented in this survey.

• A founding cattery's records were stolen. On the invasion of Ukraine the cattery that founded the one programme deliberately built on two variants left for Poland; the computer holding its breeding records was taken, and most of its cats were required to be sterilised and homed. Cats described as that breed are still offered for sale.
• Independent lines cannot reliably be kept apart, and the reason is the trait rather than the cattery. One cattery working several DBE lines at once has recorded losing track of which cat descends from which, and the passage is set out in full below because the cause it gives is general.
• A founding cattery has closed and cannot be contacted. The line the survey's own author breeds from traces to a cattery that gave up breeding, with the note that nobody retains any connection with it and the ancestor of the line is unknown.
• A completed genetic study has already been lost. The only genetic work ever done on the Ojos Azules was carried out in the late 1990s, could not be found in open sources by a later researcher, and is believed to have been lost on the geneticist's death. That is the precedent: this has happened once already in this exact subject.

Every year that passes removes cats, closes catteries and disperses records, and the cohorts above shrink accordingly. The Aktau founder is already castrated, the Topaz lines are already sterilised, and one founding cattery is already unreachable. A cohort that can be completely ascertained today can only be sampled in five years and may not be reconstructable at all in ten.

What the survey recommends

This is a recommendation and not a finding. It follows from the material above rather than being established by it.

1. Registries should be asked to run the search they alone can run. Any registry holding the pedigrees can identify every registered descendant of a named founder with a database query. None has been asked to. The request is specific and answerable: how many registered cats descend from this cat, in which breeds, and in which countries.
2. Assemble each cohort from the pedigree before sampling. A carrier group defined by descent from one founder carries a single variant by construction; one assembled by appearance does not, and appearance is what has defeated every attempt so far.
3. Prioritise the crosses over the clean lines. A cat of known pedigree carrying one of two unidentified variants is worth more than a hundred cats carrying a variant already described, because it resolves two unknowns at once.
4. Sample the deaf cat's line first. A recorded phenotype, a known pedigree, an unknown genotype and many offspring is the rarest combination in this subject, and it is the only route to distinguishing homozygous harm from compound harm in an unmapped line.
5. Record the result against the individual animal, and carry it with the cat. That is the control this case study argues for throughout, and a variant identified but not recorded on the pedigree would leave the position exactly where it is.

The wider point is about method rather than about this trait. Every question in this case study that remains open, the deafness rate of one variant, the homozygote outcome of another, whether a facial sign is attributable, whether two variants interact, is open because no cohort was defined at the time and none has been assembled since. The events are documented, the animals are named, and the pedigrees exist. What has never been done is the arithmetic of joining them.

DBE^cDc and DBE^xHW are confined to China, and why their confinement is a warning rather than a reassurance

DBE^cDc and DBE^xHW have been recorded only in certain regions of China: in the domestic cat population there and, for DBE^cDc, in Abyssinians within that population. Neither has been reported outside it.

For breeders working in Europe, North America or Australasia there is at present no practical implication for a cat bred within those populations. Neither variant has been recorded in one, so testing for either is not currently indicated. That is a statement about these two variants only. An Australian breeder holding a blue-eyed cat of the Queensland line is not assisted by it, because no test exists for a variant that has not been identified, and a negative panel result on such a cat means only that the six described variants have been excluded.

The exception is an imported cat. A cat imported from China, or descended from one, is the single case in which testing for DBE^cDc is indicated now, and it is also the route by which the wider position would change. DBE^cDc can be tested for and DBE^xHW cannot, so an imported cat can be cleared of one of the two and not of the other.

The confinement of these two variants to China carries a different significance for a registry writing a rule. Both were identified within a single year, and each of the four older variants was similarly confined to one population before it dispersed: DBE^CEL (Roxi) began in one Kazakh male and is now genotyped in six populations, five of them breeds, and reported without verification in a further twelve rows of the breed table below. A requirement that names the variants it covers will therefore fall out of date, which is why the protocol proposed at the end of this section provides for annual review against the published variant list rather than a fixed schedule of tests.

Which laboratories test for which variant

Seven laboratories sell a dominant blue eye test. No two of them sell the same thing, so what a result means depends on which laboratory produced it.

Three of the seven cover DBE^RE alone. One covers DBE^ALT alone. One covers two variants, one covers three, and one covers five. No laboratory covers all six, because DBE^xHW is on no panel anywhere.

Two of the products are complementary rather than overlapping. GENINDEXE in France tests DBE^ALT and DBE^CEL (Roxi), neither of which the German laboratories detect, and the German laboratories test DBE^RE, which GENINDEXE does not. A cat cleared in one country is untested in the other. The laboratory directory records the full offering of each.

A registry rule requiring only that a PAX3 test be performed is satisfied by any of the seven, including any that does not cover the variant present in the animal being tested. A rule of that wording therefore establishes that a test was bought, and not that the relevant variant was looked for.

Two of the products are complementary rather than overlapping: a French laboratory tests the two variants a German laboratory does not cover, and the German laboratory tests the variant the French one does not. The laboratory directory records the coverage of each.

The harm arises in the mating rather than in the individual

A cat with one copy of one variant is, on present evidence, unaffected. What is recorded below is what happens when a kitten inherits two. The two rows of the table are the two ways that can occur.

Reading the table has two consequences for a breeder. Every row in the lower section involves two different variants, which is the combination a breeder working carefully within a single line will not produce, and which arises when cats from two separate lines are brought together. And the empty cells matter as much as the filled ones: for three of the six variants nobody has recorded what two copies does, so a mating that would produce one cannot be assessed at all.

The mating to be avoided is therefore variant to variant, whether the two variants are the same one twice or two different ones. As the lines mix across breeds, that mating becomes easier to make unintentionally.

The Topaz, a breed founded on two variants at once

The clearest illustration of that principle is a breed that no longer exists. The Topaz was founded deliberately on two different dominant blue eye variants. Its two founding studs came from the same Kazakh breeding programme: Roxi Darlin, carrying DBE^CEL (Roxi), and Nikita'l Seymour, carrying DBE^ALT. The published description records that the breed was created using both, thereby combining two DBE lineages. The Topaz has since been discontinued. The full history of those two cats, and of the breeds created from them, is the subject of case study two.

A breed founded on two variants will produce compound heterozygotes in successive generations by design rather than by accident. A compound heterozygote is a cat with one copy each of two different variants, and it is the state in which harm is documented. Two kittens genotyped as compound heterozygous for DBE^CEL and DBE^ALT were white and deaf. A further kitten presumed to carry DBE^CEL and DBE^RE died at birth with limb contractures and an abnormal head.

The compound heterozygote is evidenced, not inferred. Abitbol et al., Animals 2024;14:1845 reports a five-kitten litter born to two DBE parents with DNA held for both parents and four kittens: a latent dam of the Seymour line heterozygous for DBE^ALT, a sire heterozygous for DBE^CEL, one kitten heterozygous for each allele, and “the two white and deaf kittens were compound heterozygous for both DBE^CEL and DBE^ALT alleles”. The paper's own conclusion is that such cats “are viable but are at high risk of deafness” and that “breeders are warned that mating two DBE cats from the Roxi and Seymour origins may be deleterious”.

The dam was latent. She carried DBE^ALT and did not show the blue iris, so a rule triggered by eye colour would not have reached her, and the mating that produced two deaf kittens would have been permitted. And the DBE^ALT variant was itself identified from the whole-genome sequence of one of those two deaf kittens.

Six black-eyed kittens: the clearest reported instance of the compound state

In later generations of the Topaz, kittens were born whose irides were so dark they read as black, and on close inspection were a very dark olive green. Six were born, from particular parent combinations. All six died of infections that were successfully treated in their normal littermates. The account concludes that they carried a defect of the immune system alongside the eye colour, and records that the lines producing the trait were sterilised.

The homozygote explanation was considered and set aside on stated grounds. None of the six carried the white distribution characteristic of homozygotes, none was deaf, and no other indication of homozygosity was present. The cause given is a mix of the two genes, DBE^ALT and DBE^CEL.

This is the strongest reported instance of compound heterozygote harm in this survey, and it is stronger than the deaf kittens on three counts. It carries a visible marker, the black iris, so the state can be recognised without a test. It carries a distinct clinical outcome, failure to survive infections that littermates survived, which is not deafness and not neonatal death. And the competing explanation was examined and excluded with reasons given rather than left unaddressed.

What it is not. It is a breeder account and not a laboratory result. None of the six was genotyped, the lines are sterilised, and the breeding records were lost with the computer that held them. Six animals is a small number and infection has many causes. The survey records it as reported and does not adopt it as a finding.

Two things nonetheless follow that do not depend on accepting the cause. The observation was made before any of the variants had been named, so it cannot have been shaped by the literature. And the same account asks the question this survey asks, in its own words, of the high-white cats: which of the two genes are they homozygous for, “could they be homozygous for both genes?” A breeder working without a test had reached the compound heterozygote question from the phenotypes alone.

Source: messybeast.com, Blue-Eyed Breeds: Topaz (Ukraine), the Seymour DBE^ALT and Roxi DBE^CEL genes, 2016 to 2026, supplied to this survey on 15 August 2026. A breeder account throughout; no genotype appears in it.

No published source states why the Topaz was discontinued, and this survey does not assert a cause. What can be stated is that the mechanism was present in the breed's foundation and that the documented harm arises in precisely that pairing.

The record is missing. The cats are not

The Topaz genealogy continues, through the British Shorthair and Longhair, the German Longhair, the Maine Coon, the Persian, the Siberian, the Sphynx and Elf, the Ragamuffin, the Highlander, the Munchkin and Minuet and the Lykoi. This survey evidences the cats leaving Ukraine between 2014 and 2018, well before the invasion. The loss of the records in 2022, and of the cats that were still in Ukraine, did not end the line.

The records were lost to the invasion of Ukraine. The breeder fled to Poland with her cats and dogs; the computer holding the breeding records was stolen; and the Polish authorities required most of the cats, of which there were a large number, to be sterilised and homed as pets or sold. The source records that she was unwilling to see the cats taken up by other breeders who would have the work for nothing. The original Topaz is therefore either lost or suspended and many of the records with it, and cats described as Topaz still appear for sale online. That is why the line cannot be followed the ordinary way: there is no surviving breeding record to consult. Where the cats appear after that point, they appear in whatever register would take them: chart C6 annotates both Lazuli and Kaena, the two cats the Céleste was founded on, “Registered as European / XSH”, and Roxi, Nikita'l Seymour, Nikita'l Pheer-Ca and Nikita'l Markiss stand as XXH, XSH, XSH and XLH. Those codes are the successor records, not the original ones.

The survey identifies 174 records, 34 catteries and 11 breeds in which the Topaz lineage continues to progress, the most recent cat born in 2024. Nine were born in 2023 or later. The line is not historical.

Why no one has recorded it. The breeding records were stolen with the breeder's computer when she fled the invasion, so the usual method fails at the first step: there is nothing to query. The cats went out under the prefixes of the catteries that bought them and under whatever breed codes those catteries could get them admitted under, so the descent is spread across surviving registers in fragments that do not name the breed they came from. The only way to follow the line is to piece it together by name and by pedigree, one animal at a time, from charts, from individually held certificates and from later registrations elsewhere, which is what this register does and why the count can be given at all.

If the association holds, the Topaz has evidential value disproportionate to its size, because it represents a cross that could not ethically be designed. It indicates that the risk in these variants lies not in carrying one but in holding two within the same population. The rule that follows is neither a ban on blue-eyed cats nor a ban on carriers, but a ban on combining two variants, whether the same variant twice or two different ones.

That rule cannot be enforced with a single-variant test. To know whether a proposed mating puts two variants together, a breeder has to know which variants both cats carry, and a test that looks for one variant reports only on that one. A cat clear for DBE^RE may still carry DBE^ALT: the mating that produced the two deaf kittens would have passed a DBE^RE test on both parents. The seven products on sale cover between one and five variants each, so they are not interchangeable, and a rule that does not say which variants must be covered has not specified the test it is asking for.

A DNA test cannot establish whether a cat is deaf

One of the seven products is sold under the name Deafness in cats with DBE. It is a DNA test. It reports whether the cat carries the DBE^RE variant. It does not report whether the cat can hear, and no DNA test can, because hearing is measured by a BAER recording at a veterinary practice and not by a cheek swab in the post.

So a breeder who buys this test to find out whether a cat is deaf gets no answer to that question. A breeder who buys it to find out whether a cat carries DBE^RE gets the right answer, under a name that describes something else. The name assumes the very link the test is being sold on, and that link has since been qualified in the peer-reviewed literature. It rests on a series of eight affected cats reported when the variant was first described, in Rudd Garcés et al., G3 2024. A later study, Abitbol et al., Animal Genetics 2025, tested ten more cats carrying the same variant, from different lines, and found that all ten heard normally in both ears. The product name reflects the first paper and not the second.

A registry writing a requirement has to say which of the two it wants, because “a DBE test” and “a hearing test” are different things with different costs, different providers and different answers. A rule that asks for the wrong one will be satisfied by a result that does not address the risk it was written for.

Why two blue-eyed cats of the same breed can carry different variants

Two breeders may each advertise a blue-eyed British Shorthair, using the same breed name and describing the same eye colour, while one cat carries DBE^CEL (Roxi) and the other DBE^ALT. Nothing a buyer can see, and nothing on either advertisement or either pedigree, distinguishes the two.

This happens because breeding cats move between catteries. A variant enters a breed once, through a single outcross, and kittens from that line go to other breeders. A second variant enters the same breed later, through a different outcross made by a different breeder, and kittens from that line go to other breeders too. Within a few generations the breed holds both variants, in different catteries and often in different countries, described in identical terms. The same confusion now occurs one level down. A single named line may contain more than one variant, and the breeder-created names built on the trait, among them Azul, British Ice, Heavenly Eyes, Pantherina and Pharaoh, are names for an appearance rather than for a variant, so two cats sold under one of those names need not carry the same thing either.

The first consequence is evidential, and it governs how the breed table in this survey should be read. PAX3 DBE by breed, later in this case study, records the variants that have been found somewhere in each breed. It does not state that the breed is uniform. The British Shorthair row carrying DBE^ALT does not mean that blue-eyed British Shorthairs are DBE^ALT cats; it means at least one was. This also explains how competing breed lists may both be accurate: if DBE^CEL (Roxi) and DBE^ALT are both present in the British Shorthair in different catteries, a list naming either is correct and a list naming only one is incomplete.

The second consequence, and the technical heart of this case study

How these variants are governed turns on the distinction below.

1. Two copies of one variant is a risk a careful breeder can already avoid. A breeder working within a single line knows which variant runs in that line. If it carries DBE^ALT, mating two carriers risks a homozygote, a known and avoidable pairing. This is the risk existing rules are written for.
2. One copy each of two different variants is a separate risk, and it is the one documented to cause harm. The two kittens born deaf carried DBE^CEL (Roxi) and DBE^ALT, one copy of each. The kitten that died at birth is presumed to have carried DBE^CEL (Roxi) and DBE^RE, one copy of each. No homozygote was involved in either.
3. The mating that produces the harm is between two cats of the same breed carrying two different variants, each with a single copy. A cat with one copy of one variant is, on present evidence, unaffected, so both parents are healthy animals.
4. Whether that mating is avoidable depends entirely on whether the cats show the trait, and it divides into two cases.
   Blue eye to blue eye. Both cats visibly carry something. The risk is apparent to the breeder, the pairing is already prohibited by every registry rule written for this trait, and a breeder who makes it anyway is not acting carefully. Eye colour is a sufficient trigger here, and this is the case the existing rules address.
   Blue eye to latent carrier, or latent carrier to latent carrier. A latent carrier has the variant and shows no blue eye, sometimes with no white marking either. The breeder cannot know, the mating does not present as a blue-eyed pairing at all, and no degree of care would prevent it. This is the case in which nobody has been careless and the harm still occurs.
5. In the latent case, nothing in current practice detects the pairing. Knowledge of one's own line does not, because the two variants sit in two different lines and each breeder knows only their own. Eye colour does not, because the point of a latent carrier is that there is no blue eye to see. The pedigree does not, because almost no pedigree records which variant a cat carries. A single-variant test does not, because a cat clear for DBE^RE may carry DBE^ALT. Only genotyping both parents, across the full panel, reaches it.

The harm therefore arises from a mating that neither party can see coming, between two individually healthy cats, in a breed where both variants are present and nobody records which cat has which. The only instrument that closes that gap is a genotype recorded against the individual animal and transferred with it when it is sold, which is why that requirement appears twice in the protocol at the end of this case study.

Why a requirement written on appearance cannot work

Such a requirement fails in both directions, and both failures are documented rather than hypothetical.

It misses carriers. A latent cat carries a variant and never displays a blue eye, in some cases with no white marking either. Latent cats are recorded by genotype in Abitbol et al., Animals 2024;14:1845, Table 2, two in the Seymour lineage, three in the Nanotigr and one in the Roxi, and in Abitbol, Couronné et al., Animal Genetics 2024;55:670-675, Table 1, one Céleste. One latent queen was the dam of the two deaf compound heterozygotes. Under a requirement triggered by eye colour she would have bred without restriction.

It restricts cats that carry nothing. Blue eyes arise independently through the white spotting allele and through the colourpoint allele. The working rule used by many breeders, that a blue eye in a coloured coat indicates DBE while a blue eye in a white cat indicates white spotting, is incorrect: a documented mating between two cats carrying no DBE variant, one of them carrying white spotting, produced a blue-eyed offspring with parentage confirmed by DNA.

A ban on blue or odd eyes therefore excludes cats that present no risk while leaving the carriers that do. It regulates the appearance of the problem rather than the problem.

It also removes the evidence that would settle the question. Registration is the only independent certification of parentage a pedigree cat has, and a claim about descent that rests on a breeder's own record does not meet the standard applied at the higher tiers of peer review. Closing registration to a trait therefore ends the certified record of the line at that generation, while the cats themselves continue to be bred. The restriction is adopted because the evidence about the trait is incomplete and it operates to keep it incomplete. The mechanism, its two consequences and the registration class that would resolve it are set out in the survey under the heading what a registry certifies, and what a ban removes.

The deafness evidence for DBE^RE, and what it does and does not establish

The deafness evidence is set out in full below because the misstatement of it is now more widely circulated than the finding itself, including by a commercial testing company and by the general purpose chatbots most breeders will consult first.

The DBE^ALT hearing data, reported here for the first time

Deafness is the reason dominant blue eyes is treated as a welfare question rather than as a colour. Hearing is established by a BAER test, a brainstem auditory evoked response recording, which measures the response of the hearing nerve to a stimulus in each ear separately. It yields a result in a kitten or in an animal that will not co-operate with other forms of hearing assessment, and it is the only method that detects unilateral deafness.

For DBE^ALT no such testing had been carried out. The published papers list the DBE^ALT lines with hearing recorded as assumed normal, and state the reason: no BAER recording existed for any of those animals. The literature therefore carried an assumption in place of a measurement, for the variant most widely distributed in the British Shorthair.

A series is reported here for the first time. Fourteen British Shorthair and British Longhair cats of a DBE^ALT line, nine carriers and five wild type relatives, were BAER tested at the Neurology and Neurosurgery service of the Small Animal Hospital at the University of Glasgow on 17 and 24 September 2024, on a TruTrace EMG system, each ear recorded separately at 80 and 100 dB. All fourteen heard normally in both ears. The carriers spanned silver and golden backgrounds and ages from two months to three years, and one of the nine was a latent cat whose carrier status was established only by DNA testing. To the author's knowledge this is the first BAER series in DBE^ALT British cats.

The design of the series limits what it shows. All fourteen animals were genotyped. Nine carriers were heterozygous and five relatives were wild type, so the result applies to the heterozygous state only, which is the state in which the breed is ordinarily maintained. No homozygote occurred in the cohort. Only three DBE^ALT homozygotes have been described, two Altai and one Nanotigr, all white, with deafness in some of them, in Abitbol et al., Animals 2024;14:1845, Table 2. The series is silent on that genotype. The cohort is small and drawn from a single breeding programme, so the result is preliminary.

A latent carrier bred to a carrier, and what the litter showed

The prevention control proposed at the end of this case study rests on one proposition: that a cat can carry a PAX3 variant, show no blue eye at all, and pass the variant on. Every registry rule this survey coded triggers on the colour of the eye, so none of them can see such an animal. The proposition is established by genotype in Abitbol et al., Animals 2024;14:1845, Table 2 and Abitbol, Couronné et al., Animal Genetics 2024;55:670-675, Table 1. It is set out here because a mating in a British Shorthair programme demonstrates it directly, and because the mating was designed rather than observed.

The published position first. Latent carriers are a genotyped fact and not an inference. In the DBE^ALT lines, three Nanotigr cats and two of the Seymour lineage were found heterozygous for the insertion while showing no blue eye. One Celestial cat is likewise heterozygous for DBE^CEL (Roxi) without the phenotype, and in 2025 two latent cats were identified in the DBE^RE Maine Coon line, where none had been reported before. Latency appears in three separately arisen variants, so it is a property of the trait rather than a peculiarity of one line.

The mating. A latent British Shorthair queen, a black smoke carrying no blue eye, was bred in May 2024 to a carrier male of the same line. Of four kittens, two carry the trait and two do not. A cat that no eye-colour rule could have identified produced blue-eyed offspring.

The design matters: it is the control operating voluntarily. On present evidence the harm arises when two variants meet in one animal, so a carrier-to-carrier mating is the pairing the proposed control exists to prevent. Pairing a latent carrier with a carrier answers the transmission question without making that pairing. The queen's own certified pedigree carries no eye code, and neither does the sire's, though his sire is coded 61 and his great-granddam 69: a registry reading either document would see nothing at all.

Two limits apply and are stated rather than left to be found. This is one litter of four in one programme, so it demonstrates that transmission from a latent carrier occurs and establishes no rate. And the queen's carrier status was known only because she had been DNA tested; had she not been, the litter would have appeared as blue-eyed kittens from an ordinary-eyed mother, which is how such litters are ordinarily recorded and why the register holds so many untested possible carriers.

The DBE^ALT series is cited as a released summary statement: Abitbol M, Cloquell A, Kaczmarska A, Holmes K, Macaulay K. BAER hearing screening in British Shorthair and British Longhair cats carrying DBE^ALT: Summary Statement. Belle Ayr Cats, 3 June 2025. It is not peer reviewed. Its method and its cohort, cat by cat, are reproduced in the sources. The ten Maine Coon results from the same testing sessions are published: they are the BAER data reported as Table 2 of Abitbol et al. 2025, on which the author is a named co-author.

The mechanistic passage above is the assessment of Prof Marie Abitbol, VetAgro Sup, who described four of the six variants. The three sentences in quotation marks are her wording; what follows them, including the genealogy effect, is a summary of her position and not a quotation. Rudd Garcés et al., G3 2024, 10.1093/g3journal/jkae131, Results, clinical investigations. Abitbol et al., Animal Genetics 2025, 10.1111/age.70020, Table 2 and pages 3 to 4. The 2025 BAER cohort is the author's own; see the declaration of interest in the accompanying paper.

PAX3 DBE by breed, and what can be tested

Dominant blue eyes is not a single entity. It comprises at least six genetic variants, each of which arose in a different animal, and which have dispersed between breeds by outcrossing. The table lists every breed and population in which one of them has been recorded, the route by which it arrived, and whether a commercial test is currently available. One row per breed. Colour indicates testability.

How the breed table above is constructed

The three middle columns separate origin from arrival. Arose in this breed means the variant was first described in a cat of that breed. Arrived from elsewhere means it was introduced by outcrossing, either as a named variant or as a line whose variant has not been identified. Also reported records claims put to this survey by a breeder or by a self-published compilation at messybeast.com and not verified here; those entries are leads for checking rather than findings.

Each entry carries its own grade of evidence. One breed can rest on two grades at once, so the grade belongs to the entry and not to the row. An entry in the first two columns rests on a genotyped cat of that breed, unless it is marked (pedigree), which means descent traced to the founder of a named line whose variant has not been identified. Every entry in the third column is unverified. The Devon Rex row shows both: Gosha line (pedigree) under arrival, and DBE^CEL (Roxi) under report.

Where two documents give the same cat different codes, the certificate governs. A registration certificate carries the code in full, as the registry issued it. A pedigree database entry is a transcription of that certificate and does not always reproduce it: eye codes in particular are dropped. Where this survey holds both, the certificate is recorded and the database variance is noted against it. Where only the database entry exists, it is used and named as the source.

Four breeds and populations originated a variant. The Maine Coon for DBE^RE and DBE^AGO, the population behind the Altai for DBE^ALT, a Kazakh domestic shorthair for DBE^CEL (Roxi), and Chinese domestic shorthairs for the two variants described in 2025. The Céleste is not among them: the breed carries the variant's name, but the variant was introduced to it by outcross. All other entries appear under arrival, and for most breeds that placement is an inference from the founding-effect analysis and from the variant's name rather than a route traced animal by animal.

The British Shorthair is the exception, where the route is traced. It carries DBE^ALT in the Seymour and Nanotigr lines and the published analysis follows both. Fyodor, the random-bred cat noticed at Ust-Kamenogorsk in 1995 or 1996, is the founding cat of the Altai. Nikita'l Seymour, is a silver tabby of the Topaz and one of its two founding studs. They are two different animals carrying the same variant, and the line named for the second is the one that reaches the British. The Nanotigr line was founded by a male named Oliver, from a Russian cattery, registered as a British Shorthair and probably descended from early Altai cats. One of the three DBE^ALT homozygotes recorded in Abitbol et al., Animals 2024;14:1845, Table 2 was a Nanotigr cat. The British is therefore a substantial holder of the variant rather than an incidental one.

Two of the six variants are confined to China. DBE^cDc and DBE^xHW, both identified in 2025, have been recorded only in certain regions of China, in the domestic cat population there and in Abyssinians within it. They are included because they are identified, and in one case testable, and not because a carrier has been found outside that population. Each of the other variants was similarly confined at an earlier point.

A row is not a statement about a breed as a whole. Different catteries within one breed may hold different variants, so a row records what has been found somewhere in that breed and does not indicate uniformity. The British Shorthair row demonstrates the point: DBE^ALT is testable, while Nadeya, Igor, Oliver and Marusya are not. One route into the breed can be tested for and four cannot, which is why a clear panel result does not clear a British cat.

A reported row is an invitation rather than a conclusion. Anyone able to supply a genotype result or a traced pedigree for one of them is invited to make contact, and the row will be reclassified with the evidence recorded against it. One pairing was withdrawn before inclusion: DBE^RE was proposed as confined to the Agostino lineage, which cannot be correct, because Agostino is the DBE^AGO line and DBE^RE is the Rociri Elvis line. It is recorded because that class of transposition is what a published table exists to detect.

Two distinct reasons a test may not identify what a blue-eyed cat carries

Any cat can be swabbed and run against a panel. What cannot always follow is a definitive answer about why that particular cat has blue eyes, and there are two quite different obstacles to it. They are frequently conflated and they call for different remedies.

First, the variant is identified but is not offered commercially. This applies to one variant. Five of the six identified variants can be tested today. The sixth, DBE^xHW, was identified in September 2025 and has not been added to any panel. The remedy is the addition of a marker whose sequence is already known.

Second, the variant has not been identified. At least six blue-eyed lines, among them Nadeya, Igor, Marusya, Malvina, Karagaisky and Gosha, have no identified variant. There is no target on which to build an assay, so no laboratory can offer one. The remedy is research, and it is the least costly research identified in this survey.

The distinction between those two obstacles determines what a registry may reasonably say. For five of the six identified variants a test is commercially available, so the absence of a test is not among the reasons. Where a line's variant has not been identified, a clear panel result narrows the risk without clearing the animal, and the accurate statement is that the work has not been done.

On a reported row, a green cell indicates that the claim is checkable rather than confirmed. The variant named has a commercial test, so a single swab would resolve the entry either way.

The four published variants are available from more than one laboratory. The laboratory that identified the two Chinese variants sells a panel of five, and sells it by number rather than by name, so a result arrives as DBE2 positive rather than as the name of a variant. The numbers are the laboratory's own product codes and carry no scientific meaning. Its published list gives the mapping, which is set out here because a result cannot otherwise be interpreted:

The panel is one variant short, and the missing one is the newest. It carries DBE1 to DBE5 for DBE^CEL (Roxi), DBE^RE, DBE^ALT, DBE^AGO and DBE^cDc. DBE^xHW has no panel code and is not on the panel at all, so a cat carrying it returns a clear result on every product sold. The code for each variant is given in the table at the head of this case study, which is why the mapping is not repeated here.

A result reported as DBE2 positive is therefore a Rociri Elvis result. The sixth variant, DBE^xHW, has no code because it is not included, and it is marked red in the table above for that reason rather than for want of an identified target.

Thirteen breeds and populations rest on a genotyped cat, four on a traced pedigree, and approximately thirty on unverified report. Four of the six identified variants have already crossed breed boundaries; the remaining two are at present confined to a single population, which is the position the others occupied before they dispersed. The list is open.

Genotyped and pedigree rows: Abitbol M, CORIN and DBE variants in cats, presentation to the World Cat Federation 2026; Abitbol et al., Animal Genetics 2024 and Animals 2024; Rudd Garcés et al., G3 2024; Abitbol et al., Animal Genetics 2025. Reported rows: Blue-Eyed Breeds, messybeast.com, a long-running independent effort to document the dominant blue eye lines and to correct the record where breed names and variant names have been conflated, cited by Abitbol in the presentation above. It is not peer reviewed, which is a statement about publication route rather than about rigour: the site is compiled independently, with no commercial interest in the outcome, works from pedigrees and breeder records, and marks its own uncertainty, and much of the chronology of these lines is recorded nowhere else. That is distinct from a claim made in the marketing material of a company selling the test. Its page on the Australian and Asian lines carries several populations together and they are separate: DBE^cDc and DBE^xHW were identified in Chinese domestic cats, and an untested Australian line in tropical north Queensland is recorded on the same page as a distinct case. The page does not place either Chinese variant in Australia, and this survey does not either.

Three populations recorded as leads rather than as variants

Three further populations appear in the messybeast.com compilation and none of them is an identified variant. They are listed here so that the difference between a lead and a finding stays visible.

• The north Queensland line, Australia. A litter of four with three blue-eyed kittens and one odd-eyed, the odd-eyed kitten nearly solid black, which is what rules out white spotting as the explanation and is why the line is of interest. Samples were submitted for DNA testing and the working question is whether it is the same variant as an imported Céleste from the Chatterie d'Aerlin line or a further one at the same locus. Nothing has been genotyped. Reported.
• The Chengge mao, China. A local variety selected for mouse-catching, white with a coloured tail, which began producing blue-eyed cats. Blue-eyed solid colour cats are bred and sold in parts of China with no control over the breeding and the genes involved are unidentified. One breeder reported half the kittens of a blue-eyed to blue-eyed mating solid white and dead at or before birth, which is consistent with two incompatible blue eye variants meeting but is equally consistent with other causes, and no test was done. Reported.
• Blue-eyed Burmillas, Burmese and Tonkinese. Blue eyes occur in these lines and are attributed to the colourpoint route rather than to PAX3: DNA testing has identified the colourpoint allele in some Burmillas, and aquamarine eyes appear where the Burmese sepia allele meets the colourpoint allele. A line of blue-eyed Tonkinese continues to produce blue-eyed sepias and minks. This is the clearest ordinary demonstration of the point made throughout this case study, that a blue eye is not a variant. Not a dominant blue eye.

Source for the three: Blue-Eyed Breeds: Australian mutation, Chinese mutations, blue-eyed Burmillas, Burmese and Tonkinese, messybeast.com. The two Chinese variants on that page were identified by PetGeno, DBE^cDc in Chinese domestic cats and in a line of DBE Abyssinians from the BEAST cattery in China, and DBE^xHW in Chinese domestic cats.

Four coding systems, and the one thing only the GCCF's can do

The GCCF is the only registry in the world that records which variant a cat carries on its own registration record. It does so through GEMS, the coding system that assigns every registered cat a string identifying its breed, its coat colour and its eye colour. The breed is therefore already in the code. A variant code is added to a record that says which breed the cat is, so the question of which breed a variant is in can be answered for any individual cat the GCCF has coded.

The eye-colour codes are 61 blue, 63 odd-eyed, 67 Siamese eye colour, and, added for this trait, 68 dominant blue eye odd-eyed and 69. Note code 67: the GCCF does distinguish colourpoint blue from dominant blue eye blue at the coding level, which is more than most registries attempt. Alongside those sit four variant codes: 161 Unknown, 162 Celestial (DBE^CEL (Roxi)), 163 Altai (DBE^ALT), 164 Rociri Elvis (DBE^RE). Three named variants and a catch-all. They are classification codes and not counts: 162 identifies a cat as carrying DBE^CEL (Roxi), it does not say how many such cats there are. The tracking went live in June 2025, the IT report to Council recording that work to introduce it “is now live”.

So the failure is not that the scheme omits the breed. It is that the scheme cannot name three of the six variants.

Other registries code the trait, so the GCCF's scheme is distinctive rather than unique. Pedigree certificates obtained by this survey show three further systems in use. ARBC issues code 60 and glosses it on the certificate itself, in English, as “dominant blue eyes”, with 08 for the accompanying white spots. WFA issues a 68 series, with 68b for odd eyes and 68c to 68e for latent carriers, the same series as the TopCat and Felinological Federation of Russia Altai standard. REFR marks BE on the registration number, expressly including kittens that show no blue eye. And TICA has no code at all: its registration form records the eye as free text in a box, so a TICA pedigree cannot be queried for this trait in any way.

The distinction that survives is a narrow one and it is the one that matters. ARBC, WFA and REFR record the phenotype, or the fact of carriage. Only the GCCF records which variant. A code meaning “this cat has a dominant blue eye” cannot prevent the mating this case study is about, because the harm arises when two different variants meet and every one of those cats carries the same code.

A latent carrier is codable, and one registry codes it. WFA issues 68c to 68e for a cat carrying the variant without a blue eye, and a certificate held by this survey codes an entire litter of three as 68e. The consequence is that the animal hardest to identify by eye, and therefore the one most likely to disperse the variant unnoticed, is the animal for which a code already exists and is almost never used. This survey recommends that any registry recording the trait adopt a latency code and require it, since one is available to copy rather than to invent.

DBE^AGO was published in June 2025 and has no code, and neither do the two Chinese variants identified later that year. A cat coded 161 Unknown may carry a variant that has never been identified, or a variant that has been identified and has not yet been assigned a number, and the record cannot distinguish those two cases from one another. The coding list is dated August 2025, two months after DBE^AGO was published.

The gap in the coding scheme reaches the GCCF's own import rule. Its import requirement, adopted at Council on 23 February 2025, asks for either identification of the causal gene or a clear result for “all known DBE mutations”. That is a moving target which the registry's own code list does not enumerate: at the date of the list, all known DBE mutations numbered six and the scheme could record three.

One of the four names in the list is also wrong in kind, and for the reason set out in the section above. Codes 163 and 164 are Altai and Rociri Elvis, the population and the line in which those variants arose, which follows the convention. Code 162 is Celestial, a breed created years after the variant it names. The reference is at least unambiguous, since Céleste, Celestial and Celestial British are one breed and the code therefore points at a single population; the fault is in kind rather than in reference, naming a breed where the convention names the founder. The one departure from the naming convention has therefore been carried out of the literature and into a registry database, where correcting it is more onerous than correcting a paper, because every cat already coded would have to be recoded.

The codes also make a count possible for the first time. Every cat the GCCF has coded 162 is a recorded DBE^CEL (Roxi) cat and every cat coded 163 is a recorded DBE^ALT cat, so counting them is a database query. No other registry is in a position to run it, the tracking has been live since June 2025, and no figure has ever been published.

How one registry's forty-four breed policies handle the same trait

An audit of all breed registration policies at the Governing Council of the Cat Fancy, where policy is set by breed committees rather than centrally, returned the following distribution.

Six mutually inconsistent sets of exemption operate simultaneously. The British Shorthair policy exempts epistatic white, colourpointed and colourpointed-and-white cats; the Maine Coon policy exempts white cats only; the Bengal policy exempts colourpointed cats and not white ones; the Russian policy exempts nothing. The consequences differ as widely as the triggers. The same registry refuses to register a Maine Coon showing the trait, places a British Shorthair on a reference register from which it can neither progress nor be bred, annotates the pedigrees of a Persian's test-clear grandchildren, refuses a blue-eyed Russian White outright, and awards challenge certificates to a Sphynx.

What the thirteen registries requiring a PAX3 test have in common

Thirteen organisations make a PAX3 variant test compulsory. Nine make the HCM DNA test compulsory. One makes both, and no organisation anywhere in the world requires blood group typing. All three tests run on the same buccal swab, at the same laboratory, at comparable cost.

A comparison of that kind is only sound if it is like for like. Echocardiography is not comparable with a DNA test: it requires a cardiologist, costs several times more, must be repeated through the animal's life, and is unavailable to breeders in much of the world. A registry that does not mandate the scan can point to cost and access. Those reasons do not apply to a DNA test. Twelve of the thirteen organisations that compel a PAX3 test do not compel the HCM DNA test, and none of the thirteen compels blood group typing, although each is a single result from the swab already being taken.

The two classes of test also perform different functions, and the difference does not favour either position neatly. PAX3 testing is preventative: it does not detect an affected animal, it prevents the mating that would produce one, and a result never requires repeating. Cardiac screening is surveillance of disease already present in the population, which is why it must be repeated and why a single juvenile result establishes little. A registry could reasonably conclude that the preventative test is the better first measure. What has not been explained by any registry is why the preventative test became binding while the equally inexpensive, equally single, equally preventative blood group test was left to recommendation, in the breeds in which neonatal isoerythrolysis kills whole litters.

What FIFe's Maine Coon rule specifies, and what it leaves open

FIFe made a PAX3 test obligatory for blue-eyed and odd-eyed Maine Coons from 1 January 2026. The requirement exists, which is more than most registries have done, and it is drafted with unusual care in places: the cat must be identifiable by microchip, the microchip number must appear on the test certificate, results must be registered and published under § 3.5.1, and the breeder must inform buyers if the parents are carriers. Rule § 6.15 provides that all Maine Coons with blue eyes or odd eyes “must have a DNA test for the PAX3 gene, unless both parents are proven to be free from the PAX3 mutation”, and permits only PAX3 free to PAX3 free and PAX3 free to PAX3 carrier matings, with carriers usable only until 31 December 2026.

The rule names the gene and not the variant. It requires a DNA test for the PAX3 gene and speaks of a cat being PAX3 free, in the one breed that holds more variants than any other. DBE^RE and DBE^AGO both arose in the Maine Coon, DBE^CEL (Roxi) entered it by an outcross from the Topaz line, and DBE^ALT is reported in it. No single product on sale covers all four.

The consequence is that a Maine Coon can be certified PAX3 free on a test that looked for one variant, while carrying another. A cat tested only for DBE^RE, which is the product three of the seven laboratories sell, may carry DBE^AGO and be recorded as free. The permitted mating PAX3 free to PAX3 free then has no fixed meaning, because what free covers depends on which laboratory the breeder used.

The rule triggers on eye colour, so a latent carrier falls outside it unless a parent is known; that it applies breed by breed, with six incompatible sets of exemption across the registries examined here; and that it imposes no equivalent duty for the other tests available on the same swab.

---