The Global Cat Health Survey 2026

The findings in brief

Three things stand out from the coded data. The diseases that dominate discussion of pedigree cat health are, in regulatory terms, almost entirely optional; the requirements that do exist are often unverified; and nobody outside the registries is checking any of it.

European Union law recognises, approves and inspects breed societies for cattle, pigs, sheep, goats and horses. Its definition of “animal” is an exhaustive five-species list, so a cattle breed society is supervised more intensively than any cat registry on earth.

Where to start

Cat registry health requirements

Scored out of 100 across thirteen domains, then discounted according to whether compliance is actually verified before a pedigree is issued. Click any row for the requirement-by-requirement breakdown and its source. The organisation name is a separate link, shown in blue with an arrow, and opens that organisation's own website in a new tab rather than the breakdown. Each breakdown ends with a button that builds a printable report for that one organisation, which can be saved as a PDF; the report states on its face what it carries and what it leaves to the survey. Organisations publishing nothing retrievable are shown grey rather than red, because the absence of a rulebook is not a rulebook saying nothing is required.

Membership of an international federation predicts very little. Among the 37 FIFe bodies surveyed, scores range from 0.0 to 73.8 out of 100, with a mean of 34.7. The federation's own minimum scores 55.2 out of 100, so most of its national members require less than the federation they belong to. Grouping is by the federation an organisation belongs to. Bodies that answer to no federation, including several that name a federation only to record that they are not in it, are gathered at the end.

COI and inbreeding: three different logics

The coefficient of inbreeding, or COI, is the probability that the two copies of a gene in a kitten are identical because they descend from the same ancestor. A full sibling or parent to offspring mating produces a COI of 25 per cent, a half sibling or grandparent mating 12.5 per cent, and a first cousin equivalent 6.25 per cent. Registries, legislators and the published evidence each manage COI in a different way, and the three are not versions of the same rule. Which logic an organisation uses explains more than comparing its number with anyone else's.

Comparison matings

The coefficient each close mating produces, on a common scale. Every figure below is arithmetic and not a policy position, and each is the coefficient for that mating alone, before any inbreeding already carried by either parent is added to it.

The EU Regulation adopted 28 April 2026 prohibits the first four by name and states no coefficient. The fifth is the highest that remains available under it.

Not one instrument in the world states an evidential basis for its numerical threshold. A handful of registries publish a figure; none publishes a derivation.

The Governing Council of the Cat Fancy is the only body in this survey that varies the permitted figure with the depth it is calculated over, reducing it for shallower pedigrees. A coefficient without a depth is uninterpretable, and a figure calculated over five generations is not comparable with the same figure over twelve. Two other instruments state a depth without scaling to it, and both are set out below. The GCCF's numbers, though, are the highest found anywhere. The General Breeding Policy of December 2023 treats 1 to 25 per cent over eight to twelve generations as perfectly acceptable, permits more than 25 per cent where an experienced breeder has a specific reason, and sets its outer bound at 40 per cent, which it calls highly inadvisable to exceed. Forty per cent is well above a full sibling mating. It is also worth noting that 25 per cent, the figure usually quoted for this registry, is not its ceiling but the point above which a breeder is expected to be able to justify the decision. None of it binds anyone: the language throughout is should, inadvisable and would generally not be considered sound practice.

Because COI is arithmetic, every rule can be put on one scale whatever logic it uses. A registry that bars half-sibling and grandparent matings has an implied COI ceiling of 6.25 per cent whether or not it says so. One that bars only parent to offspring and full sibling matings leaves a COI of 12.5 per cent available. One with no rule at all leaves 25 per cent available. On that scale eleven registries already meet the evidence standard without ever stating a COI, and twenty-five permit a full sibling mating by saying nothing.

Over how many generations? The question that decides what a figure means

A coefficient of inbreeding is not a property of a cat. It is the output of a calculation run over a stated number of generations of pedigree, and the deeper the pedigree the more shared ancestors the calculation finds. The same pair of cats can return 8 per cent over five generations and 18 per cent over twelve. A figure quoted without its depth therefore cannot be compared with any other figure, cannot be checked, and cannot be enforced, because a breeder challenged on it need only recalculate over a shallower pedigree.

Three kinds of rule appear in this survey and only one of them needs a depth. The distinction matters, because a reader scanning the rules will see the word generations in many of them and conclude the question is settled.

• Relationship rules need no depth. A rule forbidding full siblings or parent to offspring names the mating and states no coefficient. Most binding limits in this survey are of this kind. Nothing is left undefined.
• Level rules cannot work without one. These state a percentage. Only three registries in the world state one at all, and they are set out below.
• Ancestor-count rules state a window, but they are not COI rules. Seven organisations require a minimum number of different ancestors within a stated number of generations. They give a depth, but the thing being measured is a headcount and not a coefficient, so they cannot be compared with a percentage.

Every instrument in the survey that states a COI figure

Six do, worldwide. Three are registries and appear with their depths in the interactive table further down this section. The other three do not appear there because they are not registries: one is binding law and two are advisory documents from bodies with no power to require anything. All six are brought together here because this is the only place they can be compared.

So of the six instruments in the world that state a number, two state the depth it is calculated over, one is unresolved, and three state none. The two that do are a British registry's non-binding guidance and a Belgian regional government's binding regulation. The two documents most often cited in policy discussion, the EU guide and the Animal Welfare Committee opinion, both state a figure without stating what it is calculated over.

Three different European figures, and why they are not alternatives

Three numbers are attributed to the European Union in discussion of this subject, and they come from three different places. They are set out here because they are routinely quoted as though they were one rule.

The arithmetic behind the third row is worth following, because it is the reason EU law is stronger than the EU guidance despite naming no number. A half-sibling mating and a grandparent to grandchild mating each produce a coefficient of 12.5 per cent, and a parent to offspring or full sibling mating produces 25 per cent. The Regulation prohibits all four by name, so the highest coefficient still available under Union law is the next step down, the first cousin equivalent at 6.25 per cent. Union law therefore arrives, without stating a number, at precisely the ceiling the evidence supports, while Union guidance states a number three times higher.

Whether the two are in conflict. They are not, and the reason is that they do not measure the same quantity. The Regulation controls the relationship between two parents: it asks how closely related this pair is. The guide controls the accumulated coefficient of one cat: its words are that cats with a coefficient above 20 per cent should not be bred from. A cat can carry a coefficient of 19 per cent from generations of accumulation and be mated to a completely unrelated cat, producing kittens whose coefficient from that mating is near zero. The Regulation permits that mating and the guide forbids the cat. A cat with a coefficient of 2 per cent mated to its half-brother is the reverse. Neither instrument overrides the other, and the guidance in any event binds nobody and predates the law by six years.

What the pair does reveal is a gap that neither closes. The Regulation is blind to accumulation. Every mating in a breed can comply with Article 8(4) while the breed climbs steadily towards 40 per cent, because avoiding half-siblings and grandparents says nothing about ancestry shared further back than that. The guide is the only European instrument that addresses the accumulated level, and it has no force. The Union has therefore made binding the control that catches the single mating, and left advisory the control that catches the trend.

A note on the EU guide, which is easily misread. The guide contains a five-generation clause: selective breeding should not proceed where the relationship of the pair is unknown over at least five generations. That is a rule about pedigree completeness, meaning what the breeder must know before mating, and it is not a calculation depth for the 20 per cent and 10 per cent figures, which sit in a separate clause. The two are independent: a pedigree can be complete to five generations while the coefficient is computed over twelve, and a coefficient computed over five generations says nothing about whether those five generations are complete. This survey has worked from its own record of the guide and not from the text, which it could not obtain, so the finding is that no depth is recorded here rather than that the guide contains none.

The seven ancestor-count rules, and why their depths are not equivalent

These organisations require a minimum number of different ancestors within a stated window. They state a depth, and they are frequently read as though they were all the same rule. They are not, in two respects: the thresholds differ, and the phrase three generations denotes two different windows.

Ten different ancestors out of a possible fourteen is a materially stricter rule than ten out of a possible twenty-eight, and both are written as three generations. The Ukrainian rule states a floor with no window at all, which leaves it open to exactly the objection that defeats a coefficient without a depth: the same pair of cats passes or fails according to how far back the counting goes.

The two tests applied here

Two different questions are asked, and they are not asked of the same registries. The evidential test applies everywhere, because the biology does not change at a border. The legal test applies only where the law itself reaches.

Test one: against the evidence standard, all registries

Test two: against Article 8(4), EU and EEA registries only

Every organisation below links to its own website, so you can go and check what it currently publishes rather than take this survey's word for it. Only registries the Regulation reaches are listed below. A registry in the United States, Australia, Japan, Russia or anywhere else outside the Union and the European Economic Area does not appear here and is not marked against this test, because the instrument does not apply to it. Nothing should be read into an organisation's absence from this table beyond its geography.

What a registry certifies, and what a ban removes

A registry does two things at once, and only one of them is usually discussed. It sets requirements, which is what the rest of this survey measures. It also issues the only independent, contemporaneous certification of parentage that exists for a pedigree cat. The second function is what makes research on inherited traits possible, and no registry rule examined here takes account of it.

Why the certificate, and not the cattery's own record, is the evidence

Work on an inherited variant has to establish how it is transmitted, whether it is fully penetrant, and whether the effects attributed to it travel with the variant or with the background of the family that carries it. Answering any of those requires several generations and a substantial cohort. A programme spanning five generations and more than a hundred animals is an ordinary size for such a question, not an exceptional one.

Every animal in such a cohort has to have its parentage certified as it is born. The record cannot be assembled afterwards, because the point of it is that a third party recorded the parentage at the time and had no interest in the result. A breeder's own record, however careful, is the account of the person making the claim. A registry certificate is the account of an organisation that was not making it. Journals applying the higher tiers of scrutiny treat those two as different in kind, and a line traced only through a breeder's own notes will not carry a claim about descent.

The registry is not a bystander to research on inherited traits. It is the custodian of the evidence, and it is the only custodian.

What reaches publication, and what a description on its own cannot do

A description of a cat is not evidence of anything. A trait that cannot be evidenced does not reach publication, and the reason is not editorial caution but that nothing in a description can be checked by a reader: the animal is not produced, the observer is not named, and the same words fit a cat with the trait and a cat without it. What raises a description to evidence is one of three things: a genotype recorded against a named animal; certified parentage, which allows the claim to be tested against the next generation; or an eyewitness account by an identified and competent observer who examined the animal and whose account can be relied on.

The third is weaker than the first two and it is not nothing. It is what the fancy press was doing in the 1970s, and doing properly: an editor who had spent ten years travelling to look at reported blue-eyed black cats published the finding that he had never verified one, because the eyes proved green or bluish green when he examined them himself. That is an eyewitness account by a competent observer and it was publishable. The letters he printed alongside it, describing cats nobody independent had seen, were printed as letters and not as findings.

The distinction survives into this survey unchanged. A breeder's account of a cat is recorded as reported. A laboratory result for a named animal is recorded as genotyped. A line traced through certified pedigrees to a named founder is recorded as line-traced. Nothing is promoted from the first to the third by being repeated, and the grade attaches to the animal rather than to the line, because a line is only as good as its weakest certificate.

One further test applies to dates, and it is arithmetic rather than judgement. A founder's reported date can be checked against the dates of the cats he is recorded as siring. This survey takes a tom's breeding life as eighteen months to eight years. Where a reported date puts a founder outside that window for a mating the survey holds on record, the mating governs and the reported date is not carried forward. The reported date stays on the record as what was reported. What changes is the date the survey works from, which becomes the window the breeding itself fixes. The same test is applied wherever a reported claim and the survey's own evidence cannot both be true.

A registration code is not a claim about an animal's origin, and this survey does not reconcile the ones that disagree. Where a trait moved between programmes, breeders and clubs registered the cats under whatever code would admit them, which is why one animal can appear as XSH on a database, as a Siberian on a genealogy and as a Topaz in a cattery's own account, and why fifty-three cats in this register carry an experimental code beside a recognised breed name. There is no one correct version, because the practice was ad hoc. The register therefore carries what each document says and does not choose between them. The same applies to cattery prefixes, which are gained on entering a programme and dropped on leaving it, so neither a breed code nor a prefix is evidence of identity or of descent. What settles either is a genotype for the variant and a certified pedigree for the parentage.

The chain has two links, and a registry holds both

An evidence trail for an inherited trait needs two things certified, not one. Identity answers whether the animal that was tested is the animal named on the result. Descent answers whose offspring that animal is. A microchip supplies the first: an implanted number read by the veterinarian at the moment of testing, which ties a BAER recording, a blood pressure reading, a radiograph or a laboratory result to one animal and not to a name on a form. Registration supplies the second, and prints the two together, so that the number in the cat and the parents on the certificate are one record.

The health programmes that are taken seriously already work this way, which is the clearest demonstration that the requirement is not a formality. The PawPeds hip dysplasia programme requires the cat to carry a microchip or tattoo and requires the identification number to be “exposed into the radiograph, not added digitally”, so that the image itself carries the identity of the animal it is an image of. The CFSG Code of Practice on cat breeding requires breeding cats to be microchipped and test results to be “stored against the microchip number”. Neither would be written that way if a name and a pedigree were thought sufficient.

Break either link and the material stops being evidence. An unidentified cat produces a result that cannot be attached to an animal. An unregistered cat produces an animal that cannot be attached to a line. Both failures are invisible at the time and fatal at review, because a reviewer's question is not whether the work was done honestly but whether anyone other than the person reporting it can verify what was done and on which animals.

Microchipping is now compulsory for cats in several jurisdictions in the survey, so the identity link is increasingly supplied by law. The descent link is supplied by nothing but a registry, and no law requires it.

The effect of a ban, stated as cause and consequence

When a registry closes registration to a trait, refuses pedigrees to cats carrying it, or moves them to a register from which no further breeding is recognised, the chain of certification stops at that generation. The animals still exist and are still bred. What ends is the independent record of who their parents were. From that point the line can be described but not evidenced, and a description that cannot be evidenced does not reach publication.

Two consequences follow, and neither is intended by any registry. Neither is a criticism of any organisation's motives either. The registries that restrict most are frequently those that require most by way of health testing, and the restriction and the requirement come from the same concern for welfare. What is at fault is the design of the instrument, which treats the permission to breed and the record of parentage as one thing when they are two.

The first is that evidence quality runs opposite to strictness. The organisations that prohibit a trait are frequently those with the fullest records and the strongest verification. The organisations where the work can continue are those whose rules restrict least. So the strictest registries generate no evidence at all about the traits they prohibit, and the evidence that does exist is produced under the registries that ask for the least. A researcher is pushed towards the weakest verification regime available in order to obtain any verification at all.

The second is that the ban seals itself. A restriction of this kind is ordinarily justified by uncertainty: the trait may be harmful, and the evidence is incomplete. Removing registration removes the means by which that evidence would be completed. The uncertainty then persists, and the same uncertainty continues to justify the restriction. A rule adopted because a question is unresolved operates to keep it unresolved.

The provision that is missing

No rulebook coded for this survey contains a route by which animals may be registered for the purpose of an approved study. There is no research register, no research exemption, no study licence, and no class of registration that records identity and parentage while withholding whatever entitlement the registry does not wish to grant. Registration is treated throughout as a permission to breed and to show, so refusing it necessarily refuses the record as well, although the two are separable and are separated routinely for other purposes.

The machinery already exists. Registries operate experimental, reference and foundation registers for new breeds and for outcrosses, in which parentage is recorded fully while show and championship status is withheld until conditions are met. A cat registered as an experimental of its breed has its descent certified exactly as any other cat does. Nothing in that mechanism depends on the registry approving of the trait; it depends only on the registry being willing to write down what the parents were.

Three studies that exist only because the record was available

The mechanism is not hypothetical. Three pieces of work carried out alongside this survey illustrate it, two completed and one in progress. Declaration of interest: all three were conducted by the author of this survey, on the author's own cats, and are declared here for that reason. The published papers discussed alongside them are other people's work and are cited as such.

Hearing, and what the second cohort changed. BAER testing is the recording that establishes whether an ear responds to sound. Two cohorts carrying DBE^RE have been tested and they gave opposite results. They are not the same cats and not the same branch of the line, which is the whole reason the second one changes the picture.

• The first, in the paper that described the variant. Ten Maine Coons of the line, and all eight of the blue-eyed cats returned abnormal waveforms, three deaf in one ear and five in both. Read alone that is eight affected out of eight, and it is the figure now most widely repeated. Not the author's cats and not one of the three studies above.
• The second, four years later, and from a different part of the same line. Ten Maine Coons, six carrying DBE^RE and four not, drawn from two branches of that line, under the BigBoots and Joyfield prefixes, across three generations, and all ten heard normally in both ears. These are the author's own cats, and this is the first of the three declared studies.

The consequence is that the two results together, and not the first alone, are the evidence. Pooling them gives eight deaf among fourteen BAER-tested carriers, all from one line, which supports the statement that deafness is associated with the variant at an undetermined prevalence and does not support the statement that a single-copy carrier is at very high risk. That distinction matters to every registry rule written for this trait, and it exists only because a second cohort could be assembled. Assembling it required cats known to carry the same variant, with certified descent, still being registered and still being bred. The variant is restricted to one lineage, so the second cohort was necessarily of that lineage; what it had to be, and was, is a different branch of it. Had registration closed at the first result, the first result would have stood as the whole of the evidence. The full account, with both papers and the mechanistic objection to the causal claim, is in case study one.

Blood pressure and the CORIN copper variant. A trial proposed by the geneticist who described the variant examined whether blood pressure differs between cats carrying the CORIN copper allele and cats not carrying it. Forty-three British Shorthair and British Longhair cats were selected and forty-one were analysed, assembled so that all three genotypes were represented in both sexes and in entire and neutered animals: wild type, heterozygous carrier, and homozygous. Genotyping for CORIN:c.2425C>T was carried out at VetAgro Sup, Lyon. Readings were taken with an oscillometric monitor by an independent veterinary practice, on unmedicated and unanaesthetised cats, five readings each, and the data were signed off by an Official Veterinarian. The finding was that there is no statistical relationship between blood pressure and CORIN copper genotype.

Reported in Abitbol M, Macaulay K. Corin Levels vs Blood Pressure Trials in British Shorthair / Longhair Cats: Summary Statement. Belle Ayr Cats, 10 September 2024. The statement gives the instrument, the dates, the practice, the veterinary sign-off, the genotyping and the composition of the cohort. It is not peer reviewed and is cited as a released summary statement.

The cats in that cohort were drawn from three generations of one breeding programme and were genotyped for the Inhibitor locus, which produces the silver coat, as well as for CORIN and PAX3. Coat is recorded by genotype rather than by description throughout the work under this prefix. The purpose is set out in case study one: silver is the coat in which a blue eye has been recorded as green and in which white spotting is hard to see, so a cohort described as silver by eye carries both problems into its own data.

A negative finding of that kind is exactly what a registry considering a restriction on a variant needs, and it can only be produced from a cohort in which every genotype is represented and every animal's parentage is certified. Assembling that cohort requires the animals to be registrable. A registry that closes registration to the variant forecloses the study that would tell it whether the variant matters.

Parturition, across three breeds and two variants, in progress. A proactive feline parturition management protocol is being run over 2025 and 2026 across the British Shorthair, the Exotic and the Maine Coon. It sets out a clinical trigger framework, the points at which intervention is indicated during and after birth, together with treatment protocols for queens and for kittens, and it tracks PAX3 and CORIN genotype against lines carrying neither. The question it addresses is whether outcomes at birth differ by line and by variant, which is a welfare question of a kind registries are asked to legislate on and rarely have data for.

Work of that design has two registration dependencies at once. Outcomes must be attributable to lines, so parentage has to be certified for every queen and every litter as it occurs, and the comparison is across three breeds, so the same registry has to be willing to register all three and the lines within them, including lines carrying the variants under study. A ban applied by one registry to one of those breeds removes an arm of the comparison, and a comparison missing an arm is not published.

Sources: BAER results from the two peer-reviewed studies of the Rociri Elvis line, cited in full with their DOIs in case study one. Parturition protocol: programme in progress over 2025 and 2026, not yet reported, and recorded here as an instance of the registration dependency rather than as a finding. CORIN trial: unpublished summary statement, 10 September 2024, genotyping by Prof Marie Abitbol, Génétique, VetAgro Sup, Campus vétérinaire de Lyon; blood pressure readings recorded and signed off by Robertson & Orr Veterinary Practice, Maybole, United Kingdom. The summary is unpublished and is described here as such.

What a research register is not

A research register is a class of record and not a licence to create affected cats. Where a particular mating should not happen, the answer is a rule against that mating, which is what the protocol at the end of case study one sets out and what the survey recommends wherever the harm is established. The animals that already exist are a different question, and the answer to that one is that they should be recorded and followed rather than removed from the record.

The narrower claim is harder to argue with. The evidence that now justifies restricting a trait was produced from animals that were registered while the trait was still permitted. Osteochondrodysplasia in the Scottish Fold was characterised in Folds and in Fold crosses whose descent was recorded, which is how the homozygous phenotype was shown to persist through outcrossing while the ear phenotype was concealed by it. The disease burden in the Persian and the Exotic was measured in cohorts of registered cats of known breed and known age. Polycystic kidney disease was quantified as 46 per cent in the Persian and 54 per cent in the Scottish Fold because someone could say with certainty which cats were Persians and which were Folds. Every one of those figures is now used to argue for restriction, and every one depended on the animals being on a register at the time.

A registry that closes its record does not stop the trait being bred. It stops the trait being counted. The animals go on being produced outside the register, and the next generation of evidence, which is what would settle the question the restriction was adopted to address, is never collected.

The registration classes that already exist, and what they are for

The recommendation above asks for something registries have largely built already. A sweep of the coded rulebooks finds classes and approval routes of this kind in organisations across five federations and three continents. What none of them provides is the ground: every one of these exists to develop a breed or to place a cat of irregular origin, and not one is available because a study needs the record. Adapting an existing class would take a sentence in the rules rather than a new system.

None of these classes is for a breed. They attach to the status of the animal, not to any particular breed: a cat whose ancestry is incomplete, whose parents are unregistered, whose colour is unrecognised, or which is being used in a programme that has not yet produced a recognised breed. That is why the same class serves a Munchkin cross, an unrecognised colour in the British, and a household cat of unknown origin at the same time, and it is why a research route could be added to the same list without touching any breed's standard. The recommendation in this section applies to every breed in the survey and to cats of no breed at all.

Three things separate these from what research needs. The first group is the closest: an application is made, a plan is stated, and the organisation approves it. Felis Britannica requires objectives, cats and plan in writing to its Breeding, Health and Welfare Commission, and the Slovak federation asks a breeder to state the aim, the benefit, the health risks, the crossing plan and the methods of genetic selection. That is a research application in everything but name, and its stated purpose is breed development. The second group records faithfully and asks nothing, which serves the record but gives a study no standing and no approval to point to. The third records the animal and closes the line, which preserves identity and destroys the generational chain, and it is the commonest response to a trait a registry has decided against.

The instrument that research needs already exists in a dozen rulebooks. Only the purpose is missing, and a purpose is one clause.

The function this survey does not score

Every figure in this survey measures one function. The thirteen domains ask what an organisation requires, and nothing in them asks how complete, how open or how durable its record is. The result is an asymmetry. An organisation is credited for mandating a test and credited nothing for maintaining the record that makes a test result mean anything. A result attached to an unidentified cat of unrecorded parentage is not evidence about a line, a variant or a breed. The record is the precondition of the requirement, and only the requirement is scored.

Two organisations show what that costs, and they are not the only two. The International Cat Association scores 5.7 out of 100, which is 5.7 per cent of the marks available across the thirteen domains. It also registers cats of mixed or unrecorded origin as an experimental new breed and certifies their parentage from that point on, and the line-tracing on which case study one and its directory rest was obtained under exactly that class. The World Cat Federation scores 14.2 out of 100, or 14.2 per cent, and its own breeding rules make no health test obligatory; what it does is register, and cats failing the three-generation test for its full register go on its supplementary and experimental register rather than being turned away. For comparison, the highest score in the survey is 73.8 out of 100 and the FIFe federation minimum is 55.2 out of 100. On the scale used here both organisations therefore sit far below the federations that mandate more. On the function that produced the evidence in this survey neither is credited at all, because there is no column for it. The table above shows the same is true of others.

An open register is not the absence of a health policy. It is the infrastructure on which a health policy would have to be built, and it is currently supplied by the organisations this survey scores lowest.

Those organisations, named, are the ten scoring under ten out of 100 at the census date: Felis Belgica 0.6, Clube Portugues de Felinicultura 1.1, Felis Romania 1.1, the Co-ordinating Cat Council of Australia 1.5, Asociación Felina TICA del Plata 3.4, the Southern Africa Cat Council 4.7, TICA 5.7, the Governing Council of the Cat Fancy of Ireland 6.8, the Canadian Cat Association 7.8, and the Feline Association of South Australia and Catz Inc. of New Zealand, both 8.7. WCF sits a little above them at 14.2. The evidence in case study one was obtained under the registers of two organisations on that list, and a low score is not a good thing. It shows only that the survey's instrument does not measure the function those organisations perform.

The survey does not propose measuring it. No organisation examined here holds a record of the kind such a measure would score, and a scale has to have something to score before it is worth building. What the survey records is the asymmetry itself: the requirement is measured, the record is not, and the organisations supplying the record are at the bottom of the table because of what the instrument asks rather than because of what they do.

The choice this leaves, and which of the two an organisation is

Two positions are coherent, and an organisation may take either.

The first is the regulator's position. An organisation that restricts a trait on health grounds accepts a corresponding duty towards the evidence, and provides a route by which animals may be registered for a declared study. The restriction and the research register belong together, because a restriction is justified by a state of knowledge and a research register is what improves it.

The second is the registrar's position. An organisation states that its function is registration and not the regulation of health and welfare, that what it certifies is identity and parentage, and that the setting of health requirements belongs to breed clubs, veterinary bodies or the law. It therefore neither mandates tests nor closes its register against a trait. That position follows from its stated role. A record that is complete and reliable is a public good in itself, and it is the good on which everything in this section depends.

What is not coherent is the third position, and it is the one most commonly occupied: restricting a trait on health grounds while providing no route to the evidence that would show whether the restriction was warranted. That takes the regulator's power and declines the regulator's duty.

No organisation in this census states which of the two it is. The closest any comes to the registrar's position is the World Cat Federation, whose own breeding rules make no health test obligatory of any kind and state that they are general guidelines to be adapted as local circumstances allow, with enforcement devolved to member organisations. That is the effect of what those rules do and do not say rather than a declared role, and the difference matters, because a reader cannot tell an organisation that has chosen to be a registrar from one that intends to regulate and has not got round to it.

A declared role would change how every figure in this survey is read. An organisation that holds itself out as a registrar and nothing more is not failing when it requires no health testing; it is doing what it said it would do, and its score is a description of its role rather than of its performance. An organisation that holds itself out as setting health requirements is answerable for what it sets, and for whether the evidence underneath it can be obtained at all.

The line-tracing in case study one rests on registration certificates obtained from registries whose rules permitted the cats concerned to be registered. Where that certification was unavailable, the entries in the directory are graded as reported or line-traced rather than as documented, and the difference is visible in the evidence column against each animal. See case study one.

Do member clubs pass on the federation's rules?

FIFe's rules bind its member organisations directly, so a member does not need to restate them for them to apply. But a breeder does not read the federation's rulebook. They read their own club's material, and if an obligation is missing from that, they will not know it exists. This compares each member's own published material against the seven DNA tests and three health examinations the federation makes obligatory.

This is not a compliance verdict and must not be read as one. A test absent from a member's own published material is not evidence that the member fails to enforce it. It is evidence that a breeder reading that member's own rules would not learn of the obligation.

The federation funds research, and the fund cannot be applied to from outside

FIFe operates a research fund, the Feline Fund. It is established by the federation, is sponsored by it and carries donations from its members and from individuals, and its stated purpose is to support scientific research on feline health and on “cat breeds at risk of being banned due to misconceptions”. That second purpose is the federation's own framing and is recorded here as theirs.

This bears directly on the gaps the survey records. Case study one identifies dominant blue eye lines that have never been sequenced and sibling cohorts that have never been tested, and the standard objection to closing either gap is cost. The federation that makes the PAX3 test obligatory across its whole membership also holds a fund whose declared object is research of exactly that kind, so the obstacle to a mapping study is not, on the face of it, an absence of money within the federation.

What the fund does not publish is how to reach it. No application process, no eligibility criteria, no closing dates and no indication of award size appear on the federation's page for it. A fund whose route of application is unpublished can be applied to only by those already known to the organisation.

Recommendation. The Feline Fund should publish an application route, eligibility criteria, closing dates and an indicative range of awards. The sequencing of the unmapped dominant blue eye lines and the testing of the sibling cohorts identified in this survey fall within its declared purpose and could be put to it directly if that route existed.

Source: FIFe, Feline Fund, fifeweb.org/cats/feline-fund/, read at the census date. The fund was drawn to this survey's attention by a member of a FIFe breed committee and the account above is checked against the federation's own published page.

The same question asked of every other umbrella body

If a member is to be credited for adopting its federation's rules, the obvious next question is what those rules actually contain. The answer is not the same across the federations, and in most of them it is very little. The sweep was extended on 9 August 2026 to the WCF, TICA, CFA, LOOF, CCCA and ACF families so that no umbrella is credited, or discredited, on an assumption.

List of testing laboratories

This is a directory of the laboratories doing feline testing, what each one lists, and which laboratories offer a given test. No prices: they change too often to publish responsibly, and every laboratory quotes its own. Use it to see what exists and where.

Only organisations that operate a laboratory are listed. A brand, a sales office, a distributor and a laboratory are four different things, and a directory that conflates them tells a breeder nothing about who is actually reading the sample. Where an organisation sells testing performed by somebody else, the entry says so and names the performing laboratory where that could be established from a primary source.

Looking for who does a particular test? Use By test. Choose PAX3, HCM, PKD, blood type or any of the other 37 tests, and the table lists every laboratory that offers it, what each one calls it, and the gene. By breed bundle does the same for a whole breed at once.

Compiled from each laboratory's own current test list, read in its own language. Laboratories whose list could not be retrieved are shown with a dash and marked unverified rather than left out, since their absence from a comparison would be misleading. Nothing here is a recommendation, and inclusion is not endorsement.

Three things the directory shows

The first is that quality is largely unverifiable. Of the laboratories documented, only one publishes both ISO 17025 accreditation and a current ISAG comparison test certificate for the cat. There is no quality assurance scheme for feline genetic testing anywhere: no accreditation requirement, no proficiency testing programme, and no published method or error rate. The nearest thing is a searchable directory of which laboratory offers which test, carrying no quality information at all.

The second is that no registry tells you where to go. Registries accept results from any laboratory, and one states outright that it endorses none. That is defensible as neutrality, but combined with the absence of any accreditation scheme it means a breeder choosing a laboratory has nothing official to go on.

The third is that the market is consolidating. One genomics business covering several familiar feline products was sold in March 2026, which would place three of the better known consumer panels under a single owner, and at least one long-standing veterinary laboratory has closed. Any guidance naming specific laboratories dates quickly.

Multi-test DNA panels: the group testing question

Registries add tests one at a time, breed by breed, each with its own certificate and its own fee. The commercial laboratories long ago stopped working that way. This section asks whether a single panel would be a more sensible basis for regulation than a list of separate requirements, and sets out the case both ways.

What a panel now contains

The large feline panels screen in the region of 45 to 50 genetic conditions and 25 or more traits, identify breed composition, and include blood typing as standard, for roughly £100. A worked example, a Maine Coon queen tested in March 2026, illustrates the scope: 50 conditions screened and reported clear, around 40 trait variants covering coat colour, coat type, tail length and polydactyly, a DNA blood type of A with the genotype recorded as a carrier of the b variant, and a heterozygosity figure of 34 per cent against a breed reference range of 32 to 37 per cent.

How that compares with what registries demand

One panel covers six of FIFe's seven federation-wide obligatory DNA tests, and every mandatory active-register DNA test imposed by GCCF. The single exception in both systems is the PAX3 test for dominant blue eyes.

The six it covers are CEP290 for the Siamese and Oriental group, HEXB for the Burmese, GLB1 and HEXB for the Korat, GBE1 for the Norwegian Forest, COLQ for the Devon Rex and ALX1 for the Bombay. On the GCCF side it carries the Maine Coon A31P, spinal muscular atrophy and pyruvate kinase deficiency tests, the Ragdoll cardiac variant, PKD for the Persian, Exotic and British, and rdAc for the Cornish Rex.

It also delivers two things no registry in the world requires. Blood typing is the clearest case: the queen in the example is type A, so her own litters carry no neonatal isoerythrolysis risk, but she carries the b variant, so a daughter bred to another carrier could be type B and it is that daughter who would lose kittens. A registry that never asks about blood type cannot see this coming, and the information is already in a report the breeder has paid for. The second is a genetic diversity measure, which sits alongside the inbreeding rules several clubs already operate without ever requiring it to be measured.

The case for regulating by panel

Cost and administrative burden fall rather than rise, since a single test replaces four or five certificates. Coverage becomes consistent between breeds instead of depending on which committee has met recently. And there is a documented precedent that bundling works: GCCF recorded that a laboratory produced a Maine Coon testing bundle matching its breed committee's import requirements exactly, and that import numbers recovered as a result.

The case against, and how much of it holds

The most cited review of feline genetic testing argues against treating panels as a substitute for targeted testing. Two of its points are sound. The first is that array error rates are proprietary and unpublished, so neither false-positive nor false-negative rates can be independently assessed.

The second is about what a result means, and it is the one worth stating slowly. A variant is shown to cause disease in the breed it was studied in. That finding does not automatically carry across to a different breed, where the same variant may sit on a different genetic background, may be harmless, or may simply never have been looked at. A panel does not distinguish. It runs every variant it holds against every cat, so a cat is tested for variants that have never been studied in its breed, and the report comes back in exactly the same format either way. The problem is not that such a result is wrong. It is that nobody can say what it means, and it does not look any different from a result that has been validated.

Both of those points stand on their own merits and are treated here as sound. As with any publication that bears on which commercial product a registry should require, a reader weighing it should check the author affiliations and the declaration of interest printed with it, as they would for this survey.

One argument frequently drawn from it does not hold at all, and it is worth dismantling because it is used to justify inaction. It is said that panels cause harm by producing incidental findings: results flagging conditions not named in the relevant registration policy, creating pressure to act on findings nobody has validated for that breed. Registry offices do report receiving them.

But a variant found in a breed where it was not expected is a discovery, not a nuisance. This survey is full of examples: PKD1 has leaked into breeds nobody screened for it, the fold variant turns up in the American Curl, the Norwegian Forest Cat, the Minuet and the Munchkin, and a dominant blue eye variant appeared in three Sphynx with no blue-eyed cat anywhere in their pedigrees. Every one of those was an incidental finding before it was a fact. An administrative burden on a registry office is not a welfare harm, and a rule cannot sensibly be opposed on the ground that it would tell people things their existing policy has no box for.

The right response to an unexpected result is the one research uses: treat the panel as a first line, then confirm. Send a second swab to an independent laboratory and see whether the finding replicates. That is ordinary practice when a result matters, it costs one further test rather than a policy of not looking, and it turns the unvalidated-variant objection from a reason to avoid panels into a protocol for using them properly.

On the specific case that prompted this argument, a panel mandate would not have helped. Two of the four published dominant blue eye variants are retroviral insertions of several hundred base pairs, which are difficult to genotype on the array platforms the large panels use, and no large multiplex panel offers a DBE test at all. That is a real limitation of the current products rather than an argument against the approach.

The limit that matters most

A panel cannot replace an echocardiogram. The worked example's own report states that hypertrophic cardiomyopathy has been diagnosed in Maine Coons negative for A31P, so the variant is not the sole cause and a clear DNA result is not a clear heart. Since no registry anywhere requires cardiac ultrasound of all its breeding cats, moving to a panel would tidy the DNA testing while leaving the largest gap exactly where it is.

Where the idea stands

No registry anywhere mandates a panel in place of a list of separate tests. The closest is a proposal put to a GCCF committee in autumn 2025 for mandatory full panels on domestic outcrosses and imported cats, which had not gone to Council as at June 2026. TICA has a commercial partnership with a panel provider, but results are not reported to the registry automatically and submission is voluntary.

Method

Each organisation was coded from its own current rulebook, read in the original language. Sources were restricted to official regulations, statutes, registration policies and health programme documents. Breeder websites, forums and secondary summaries were used only to locate an official rule, never as evidence of one.

Mandatory or merely advised

This distinction is the pivot of the whole dataset. Phrases such as “breeders are advised to” or “we strongly recommend” are coded as recommendations throughout, however firmly they are expressed. A mandatory requirement earns the full domain weight, a breed-specific one 60 per cent, a conditional one 50 per cent, and a recommendation 25 per cent.

Weights

The enforcement discount

The weighted total is multiplied according to how compliance is established: full marks where results must be filed before a pedigree or litter registration is issued, 85 per cent where verification is partial, 70 per cent where compliance is self-declared, and 60 per cent where no verification mechanism exists at all. A rule nobody checks is not the same as a rule gated at registration.

What the score can and cannot see

The score counts what an organisation requires. It cannot see whether a requirement is drafted so that it works, and the two come apart more often than the numbers suggest.

The clearest worked example is in the dominant blue eye material, now held in the Blue Eyes Directory. FIFe's rule for the Maine Coon makes a DNA test compulsory for blue-eyed and odd-eyed cats, and the survey scores it for that. The same rule requires “a DNA test for the PAX3 gene” and asks whether a cat is PAX3 free, naming the gene rather than the variant, in the one breed that holds four different variants of it. A cat can therefore satisfy the rule while carrying a variant nobody looked for. Both things are true at once and only the first is scored.

A high score means an organisation asks for a lot. It does not establish that what it asks for is effective. Anyone using these scores to compare organisations should read the rule itself alongside the number, and the rule text is linked from every row of the table.

What the score assumes about an organisation's role

The thirteen domains measure health requirements, which assumes that the organisation being measured holds itself out as setting them. That assumption is not tested anywhere in the coding, because no organisation in the census states which of the two roles it occupies. An organisation whose function is registration and which leaves health requirements to others is not failing when it requires nothing; a low score for such a body describes its role rather than its performance. The score cannot tell that body apart from one that intends to regulate and has not done so, and neither can a breeder reading the rules. The consequences of that, and the recommendation that follows, are set out under what a registry certifies, and what a ban removes.

Limits of this work

Nineteen organisations publish nothing retrievable, so their position is unknown rather than established as absent. Thirty-nine jurisdictions have now been assessed for their legal framework. Canada and the United States were assessed federally at first. In two federal states where animal welfare is provincial or state business, that understates them badly; the federal records are kept, because each carries a finding of its own, and the substantive provinces and states now sit alongside them, as they already do for the United Kingdom, Belgium and Australia. Unverifiable cells score zero, which depresses the scores of organisations whose rules exist but could not be located, so the confidence figure should be read alongside every score. The weights were not derived from any epidemiological estimate of disease burden, and different weights produce a different ranking. The country accountability instrument gives equal weight to six heterogeneous questions and is best read as a structured description rather than a measurement.