Radiographic classes at the pre-purchase examination: where the percentages come from

The classes come from a consensus, not from a study

At the end of a pre-purchase examination there is often a Roman numeral. It sums up how far the radiographs depart from what counts as unremarkable: the lowest class for images without any deviation, the higher ones for increasingly marked changes. The scheme goes back to a guideline drawn up by the German-speaking veterinary profession. It is an ordering proposal by specialists, meant to regulate the conversation between the practice, the buyer and the seller. How many steps it contains has changed between versions of the guideline, and that alone makes the point: the numeral describes an image according to a convention, it does not measure a property of the horse.

What the numeral is not matters more. No study can be found in the medical literature databases that sorted horses into these classes, followed them over years and then counted how many in each class went lame. The percentage range that the trade pins to the middle classes, usually around five to twenty per cent for a later clinical problem, comes from the same consensus paper and not from an analysis of follow-up data. This is no reproach to its authors: deriving such a figure from data requires a cohort of sport horses run over years with an endpoint fixed in advance. That study does not yet exist.

Is the class therefore worthless?

No. It orders what can be seen on the images and creates a shared language for everyone involved. What it does not deliver is a prognosis with a tested hit rate. Anyone reading a class is reading a description of the image and the judgement of specialists, not a measured probability for the years ahead.

In healthy horses, findings are the rule

Whoever radiographs will find something. A Dutch study radiographed the stifles, hocks and fetlocks of 811 randomly selected warmblood yearlings. Contours were normal throughout in 60.7 per cent at the stifle, 68.6 per cent at the hock and 64.6 per cent at the fetlocks. Across all the joints examined together, only 30.5 per cent of the horses were free of any deviation. In this population the radiologically flawless warmblood is the minority.

In American Thoroughbred yearlings at the sales the pattern is the same. Vascular channels in the proximal sesamoid bones were present in 98 per cent of the animals, irregular channels or channels wider than two millimetres in 79 per cent. The findings that are reputed to affect load-bearing capacity, by contrast, were rare: a fragment on the dorsal margin of the pastern bone affected 1.6 per cent of forelimbs, the distal intermediate ridge of the tibia 4.4 per cent. The authors drew an uncomfortable conclusion from this: what is common is mostly harmless, and what is suspicious is so rare that its importance can hardly be established.

At the back, too, deviation is not the exception. In a comparison of 47 never-ridden yearlings with 55 trained Thoroughbreds without recognisable back pain, narrowed or touching spinous processes were present in roughly a third of the spaces examined. The trained horses had no more of them than the yearlings, which argues against the widespread account that such images only arise under the saddle.

What the few follow-up studies show

Only a few studies have tracked horses over years on the basis of their radiographs, and almost all come from racing. In the largest, 1,162 Thoroughbred yearlings were radiographed and followed through their two- and three-year-old seasons. Around eight in ten made at least one start. Most of the changes found were not associated with later performance. Three findings stood out, and even with them the majority ran: 14 of 24 horses with a marked lucency at the back of the cannon bone, 8 of 14 with bony proliferation on the sesamoid bones and 19 of 30 with changes at the carpus went to the start.

A more recent cohort of 2,508 yearlings assessed the sesamoid bones separately. Three groups of findings went along with a lower chance of a race start, and bony deposits on the outer surface additionally with total earnings more than halved. The practical message, though, lies in the counter-example: a common and long-debated flattening of the outer surface bore no relation to performance and had usually disappeared by the two-year-old sales. The authors therefore suggest reporting such findings separately instead of merging them into a single grade.

The same cohort supplies a figure that puts many a purchase decision in perspective. Lucencies beneath the cartilage of the medial femoral condyle, a finding over which sales fall through, were present in 9.65 per cent of the yearlings. At the most severe grade, 78 per cent of the horses made a start, against 84 per cent with an unremarkable stifle. On the performance of those that ran it had no measurable effect.

A Japanese analysis of 1,082 yearlings likewise found a statistical association between certain findings and a missing race start, then went one step further. For twelve of these horses the authors traced individually why they had never run. None had failed because of the finding seen on the images: the causes were pelvic fractures, a tendon inflammation, a laryngeal paralysis, economic reasons. Twelve cases prove nothing, but they show how quickly an association is turned into a cause.

The same grade means different things in different regions

Perhaps the single most important observation comes from an analysis of 510 American pre-purchase examinations carried out between 1991 and 2000. Radiographs were taken in 61.6 per cent of cases, most often of the front feet and the hocks. At the navicular bone and the pedal bone a high grade of finding went along with lameness. At the hock it was the other way round: there, high grades occurred together with lameness less often, and the great majority of the most severely changed horses were still in work at follow-up.

A British case series supports this special position of the hock. In 91 horses whose pain in the small hock joints had been confirmed by targeted anaesthesia of the joint, the severity of the radiographic changes was related neither to the degree nor to the duration of the lameness. The authors preface their work with the reminder that a healthy horse can show such signs as well, and treat the response to the anaesthesia, not the image, as the yardstick.

Why does the region matter so much?

Because the same description means different things in different places. In the American analysis a high grade at the navicular bone and the pedal bone went along with lameness, at the hock less often. A grade that lumps all regions together erases precisely this difference and suggests a comparability the data do not support.

Two specialists, two classes

A class is only as reliable as the reading that leads to it. With 60 navicular bones from post-mortem examinations, assessed by three specialists on radiographs and on cross-sectional images, mean agreement between the assessors was poor in both techniques. When the same person assessed the same images a second time, agreement was only moderate. The cross-sectional images increased the confidence with which the assessors held their judgement, without their agreeing any more often. These bones were presented without surrounding soft tissue and without clinical context, which should if anything make agreement easier.

For the back the picture is more favourable. In 100 healthy horses, three assessors agreed well on clearly delimited individual features, such as the width of the space or a circumscribed increase in density. As soon as all the abnormalities of one back were added into a total score, however, agreement fell to a moderate level. The authors state expressly that this has to be taken into account at a pre-purchase examination.

How wide the spread is in everyday practice is shown by a worldwide survey of 211 veterinary surgeons. They assessed the same image series of the navicular bone, the hocks and the stifles on a four-point scale. Mean concern rose, as expected, with the extent of the change, yet at every single step the range of answers stayed wide. The direction of the difference depended on the region: a marked navicular finding worried specialist veterinary surgeons more than the other respondents, a marked hock finding worried them less.

One often-quoted counterweight deserves to be placed in context. Three experienced surgeons who assessed 30 image series from a sales repository reached consensus on the risk grading for 97.1 per cent of the sites assessed. That figure covers all the sites judged, however, the great majority of them unremarkable, which pulls it arithmetically upwards. The authors' own statement is narrower: experienced assessors agree reliably when there is nothing to see, and disagree as soon as common findings have to be graded. The hock region was the most contentious.

A finding is not a fixed possession of the horse

One point is almost always lost in the debate about classes: findings come and go. In a Walloon study, 58 young sport horses were radiographed twice to a standardised protocol, on average at 407 and at 680 days of life. Of 36 osteochondrosis changes found, nine had altered between the two appointments, had newly appeared or had disappeared. Four horses carried their finding at only one of the two appointments. The widespread assumption that the picture is stable after the first year of life does not survive this test.

In very young horses the movement is greater still. Lusitano foals followed from birth to the age of 18 months showed signs of osteochondrosis at the hock or the stifle in 76.08 per cent of cases before their first month of life, and still in 16.20 per cent at 18 months. Regression could be observed up to the twelfth month, that is later than the age limits assumed until then would suggest.

In the racehorse cohorts, too, some of it dissolved. Of 31 yearlings with the mildest grade of a lucency at the medial femoral condyle, the finding had disappeared in eleven by the two-year-old sales, was unchanged in fourteen and had become more marked in six. Reading a class therefore always means reading a date as well. The same horse, radiographed six months earlier or later, can end up in a different class.

What the market makes of the finding

The finding acts on the price long before it acts on the horse. In the largest analysis of pre-purchase examinations so far, 13,603 three- and four-year-old Thoroughbreds were examined before Irish and British auctions. Some abnormality was present in 73.6 per cent, but only 12.0 per cent carried one classed as prejudicial to the intended purpose. The market reacted clearly: horses with such a note sold in 38.1 per cent of cases, horses without it in 77.6 per cent.

How much is found depends, moreover, on how closely one looks. In a British analysis of 133 pre-purchase certificates, 57.1 per cent of the horses carried a prejudicial finding. With the detailed five-stage examination it was 68.5 per cent, with the shortened two-stage examination 34.1 per cent. More expensive horses were examined in detail more often, radiographed more often and accordingly carried a note more often. The commonest reason for a reservation, incidentally, was not the radiograph but a lameness.

What the class can do and what it cannot

Behind the weak prognosis lies a fundamental problem of the field. A methodological review examined how equine orthopaedic research decides which horse counts as healthy. For osteoarthritis, 127 publications analysed used eleven different procedures; for tendon and suspensory ligament disease, 84 publications used eight, a median of three procedures per paper. In several studies the comparison horses called healthy had not even been examined for lameness. As long as the yardstick varies, no analysis can say what an image tells us about health.

The balance is similar for the testing of the search methods themselves. A systematic review of ultrasound in osteochondrosis notes at the outset that radiographic screening at purchase is routine, but then finds no dependable data on the diagnostic accuracy of ultrasound. That is a statement about ultrasound and not about radiography. It does show, however, how little tested the methods are with which the search at purchase is carried out.

What does this mean for the appointment at the yard? The class describes the images, it orders the severity of the findings and it makes the examination traceable. It is not a measure of durability, it holds for one particular day, and it depends on the region, on the extent of the examination and on the person assessing. Anyone who knows these four limits reads the finding as far as it is established.

What actually lies behind the common claims about radiographic findings at purchase

Original analysis

Limitations and uncertainty

• The core of the topic is not backed by studies: the class scheme and the percentage ranges tied to it come from a consensus paper of the veterinary profession that was never published as a study. All the papers cited here test neighbouring questions; none tests the classes themselves.
• The only longitudinal cohorts come from racing. Their endpoint is the race start or the earnings, not the durability of a sport or leisure horse, and the breed studied differs from the one usual in the central European trade.
• Sales populations are selected: horses that never reached the auction because of a finding are missing from every one of these analyses. Severe lesions are therefore under-represented there, and the proportions reported underestimate their occurrence.
• The editorial requirement to base at least half of the sources at the level of systematic review, meta-analysis or randomised trial cannot be met here: only two of the papers used reach that level. A randomised trial is not feasible on this question, since nobody can allocate findings.
• The frequencies reported depend on the imaging protocol. One of the papers cited showed this itself, as sedation altered the detection rate at the stifle; in another, a missing projection at the stifle probably led to findings being overlooked.
• Population figures for Switzerland could not be found in this research. The data come from the Netherlands, Great Britain, Ireland, Germany, Belgium, the United States, Japan and Brazil.

Open questions

• What would a ten-year cohort of warmbloods yield if they were graded at the pre-purchase examination and then followed with an endpoint fixed in advance?
• Which findings should be reported separately instead of being folded into an overall grade, and from what point does aggregation degrade the informative value?
• Do structured criteria and trained reading improve agreement between assessors far enough to change purchase decisions measurably?
• How often does a finding placed in a higher class regress during further growth, and what determines that course?

Frequently asked questions

My horse has a middle radiographic class. How high is the risk really?

That figure does not exist in tested form. The percentage ranges quoted with the classes come from a consensus paper of the veterinary profession and not from a cohort that followed graded horses over years. What does exist is follow-up data from racing with a different endpoint: there, around eight in ten radiographed yearlings later made at least one start, and most findings were not associated with performance. No percentage for your horse can be derived from this, neither upwards nor downwards.

Why does a second practice arrive at a different class?

Because the assessment itself varies, and this has been measured. With 60 navicular bones, mean agreement between three specialists was poor, and the same person reached the same judgement on a second reading only moderately often. In a worldwide survey, 211 veterinary surgeons graded the same image series, and at every step the range of answers stayed wide. Experienced assessors agree reliably above all when there is nothing to see. As soon as a common finding has to be graded, judgements diverge, most strongly at the hock.

Are radiographic findings more common in young horses?

In very young horses yes, and many disappear again. Lusitano foals showed signs of osteochondrosis in almost three quarters of cases before their first month of life, and still in about one sixth at 18 months, with regression continuing into the twelfth month. Afterwards the picture stays in motion as well: in 58 young sport horses radiographed twice about nine months apart, a quarter of the lesions found had changed, had newly appeared or had disappeared. A report therefore always describes one day, not a permanent state.

Is a horse without findings the safer horse?

The available data do not allow that to be said. First, a horse entirely without deviation is rare: across all the joints examined, only around three in ten Dutch warmblood yearlings were unremarkable. Second, cross-sectional images of showjumpers in full competition revealed changes that were visible on no radiograph. A clean image therefore means that nothing was found at the chosen sites with the chosen method. No statement about the coming years is attached to it.

Why are the hock and the navicular bone judged so differently?

Because the data there point in different directions. In an analysis of 510 pre-purchase examinations, a high grade at the navicular bone and the pedal bone went along with lameness, at the hock less often; of 21 horses with the highest hock grade, 20 were still in work at follow-up. In 91 horses whose hock pain was confirmed by anaesthesia, the severity of the images was related neither to the degree nor to the duration of the lameness. A single grade across all regions erases this difference.

Is a more detailed examination with more images worth it?

That is a question for the examining practice, not for the state of the evidence, but one figure belongs to it. In a British analysis of 133 pre-purchase certificates, 68.5 per cent of the horses examined in detail over five stages carried a prejudicial finding, against 34.1 per cent of those examined in the shortened two-stage form. More examination finds more, and whether what is found has consequences is not answered by that figure. In this analysis the commonest reason for a reservation was in any case not the radiograph but a lameness.

Sources

1. Wrangberg T, Kendall A. Who Is Healthy? A Review of How Equine Control Groups Are Defined in Clinical Orthopaedic Research 1999-2021. Veterinary and Comparative Orthopaedics and Traumatology, 2022 (Systematic review | Horse)DOI 10.1055/s-0042-1745756
   There is no standardised definition of the healthy horse: from 2,472 and 2,746 references screened there remained 127 papers on osteoarthritis with eleven and 84 papers on tendon and suspensory ligament disease with eight different procedures, a median of three per publication, and in several studies the comparison horses had not been examined for lameness.
2. Hoey S, Stokes D, McAllister H, Puggioni A. A systematic review evaluating the use of ultrasound in the identification of osteochondrosis in horses. The Veterinary Journal, 2022 (Systematic review | Horse)DOI 10.1016/j.tvjl.2022.105825
   The authors note that radiographic screening for osteochondrosis at purchase is routine, but because the studies are too few and too heterogeneous they find no dependable data on the diagnostic accuracy of ultrasound for osteochondral lesions.
3. van Hoogmoed LM, Snyder JR, Thomas HL, Harmon FA. Retrospective evaluation of equine prepurchase examinations performed 1991-2000. Equine Veterinary Journal, 2003 (Cohort study | Horse)DOI 10.2746/042516403776014325
   In 510 pre-purchase examinations with radiographs in 61.6 per cent of cases, a high grade of finding at the navicular bone and the pedal bone was associated with lameness, at the hock less often, and 20 of 21 horses with the highest grade in the small hock joints were still in work at follow-up; the authors call for interpretation in the light of the clinical examination.
4. Kane AJ, Park RD, McIlwraith CW, Rantanen NW. Radiographic changes in Thoroughbred yearlings. Part 1: Prevalence at the time of the yearling sales. Equine Veterinary Journal, 2003 (Cross-sectional study | Horse)DOI 10.2746/042516403776014280
   Across more than 1,100 limb regions examined, 98 per cent of the yearlings had vascular channels in the sesamoid bones and 79 per cent irregular or wide channels, while the findings regarded as worrying mostly lay below five per cent: changes are the norm, and the genuinely suspicious ones are too rare for their importance to be established easily.
5. Kane AJ, McIlwraith CW, Park RD, Rantanen NW. Radiographic changes in Thoroughbred yearlings. Part 2: Associations with racing performance. Equine Veterinary Journal, 2003 (Cohort study | Horse)DOI 10.2746/042516403776014307
   Of 1,162 radiographed yearlings, 946, that is 81 per cent, made at least one start; most findings did not affect later performance, and even for the three least favourable findings the majority of affected horses ran, which is why the authors want their results understood only as a supplement to the clinical impression.
6. Shelton AV, Tupper J, Bolt DM. Prejudicial findings regarding suitability for intended purpose during pre-purchase examinations in a mixed horse population-A retrospective observational study in the United Kingdom. Equine Veterinary Journal, 2024 (Cross-sectional study | Horse)DOI 10.1111/evj.14061
   In 133 pre-purchase certificates, 57.1 per cent of the horses carried a prejudicial finding, 68.5 per cent under a five-stage examination against 34.1 per cent under a two-stage one; the commonest reason was lameness rather than the radiograph, and more expensive horses were examined in detail more often.
7. Barrett E, Arkins S. Abnormalities detected at pre-purchase examination of National Hunt racehorses presented at sale. Equine Veterinary Journal, 2019 (Cross-sectional study | Horse)DOI 10.1111/evj.13164
   Among 13,603 Thoroughbreds examined before auction, 73.6 per cent had abnormalities but only 12.0 per cent abnormalities classed as prejudicial; the sales rate was 38.1 per cent with such a note against 77.6 per cent without, so the paper measures the market effect and not later health.
8. Wendling EM, Hellige M, Freise F, Lingens P. Prevalence and radiographic features of cyst-like lesions in German Warmblood horses presented for pre-purchase examination. Frontiers in Veterinary Science, 2026 (Cohort study | Horse)DOI 10.3389/fvets.2026.1809383
   Among 3,863 German warmbloods presented without lameness, 4.25 per cent had cyst-like lesions, most often in the pastern bone and the femoral condyles; agreement between two readings was very good only for objective features, and missing projections probably led to findings being overlooked.
9. Peat FJ, Kawcak CE, McIlwraith CW, Keenan DP. Radiological findings in the proximal sesamoid bones of yearling and 2-year-old Thoroughbred sales horses: Prevalence, progression and associations with racing performance. Equine Veterinary Journal, 2024 (Cohort study | Horse)DOI 10.1111/evj.14051
   Only certain sesamoid findings went along with a lower chance of a race start and with earnings roughly halved, while the common flattening of the outer surface showed no effect and had usually disappeared by the two-year-old sales; the authors suggest reporting such findings separately rather than in an overall grade.
10. Peat FJ, Kawcak CE, McIlwraith CW, Keenan DP. Subchondral lucencies of the medial femoral condyle in yearling and 2-year-old Thoroughbred sales horses: Prevalence, progression and associations with racing performance. Equine Veterinary Journal, 2023 (Cohort study | Horse)DOI 10.1111/evj.13945
    Lucencies beneath the cartilage of the medial femoral condyle were present in 9.65 per cent of 2,508 yearlings; at the most severe grade 78 per cent started against 84 per cent with an unremarkable stifle, with no effect on the performance of those that ran, and mild grades partly regressed, eleven of 31 disappearing by the two-year-old sales.
11. Argue BJ, Ahern BJ. Evaluation of agreement for radiographic lesions and risk for racing in thoroughbred yearling sale repository radiographs. Frontiers in Veterinary Science, 2024 (Cross-sectional study | Horse)DOI 10.3389/fvets.2024.1430993
    Three experienced surgeons reached consensus across 30 image series for 97.1 per cent of the sites assessed, a rate that covers overwhelmingly unremarkable sites; the authors' own conclusion is narrower, namely reliable agreement on the absence of a lesion and disagreement on the risk grading of common findings, most strongly at the hock.
12. Groth AM, May SA, Weaver MP, Weller R. Intra- and interobserver agreement in the interpretation of navicular bones on radiographs and computed tomography scans. Equine Veterinary Journal, 2009 (Cross-sectional study | Horse)DOI 10.2746/042516408x345125
    In 60 navicular bones from post-mortem examinations, mean agreement between three assessors was poor on radiographs as on cross-sectional images, and agreement of one person with themselves only moderate; cross-sectional imaging raised the confidence expressed without improving agreement, which leads the authors to question the diagnostic and prognostic use of such gradings.
13. Looijen MGP, Hanousek K, Maree JTM, Visser MC. Inter- and intra-observer agreement for grading radiographs of thoracolumbar spinous processes in healthy horses during a pre-purchase examination. The Veterinary Record, 2021 (Cross-sectional study | Horse)DOI 10.1002/vetr.1065
    In 100 healthy horses, agreement between three assessors was good for clearly delimited individual features but only moderate for the sum of all the abnormalities of one back; the authors call for these limits to be taken into account at the pre-purchase examination.
14. Byam-Cook KL, Singer ER. Is there a relationship between clinical presentation, diagnostic and radiographic findings and outcome in horses with osteoarthritis of the small tarsal joints?. Equine Veterinary Journal, 2009 (Case series | Horse)DOI 10.2746/042516408x345107
    In 91 horses with pain in the small hock joints confirmed by intra-articular anaesthesia, there was no relation between the severity of the radiographic changes and the degree or duration of the lameness; the authors recall that healthy horses also show such signs and treat the anaesthesia as the diagnostic yardstick.
15. Pressanto MC, Pepe M, Coomer RPC, Pilati N. Radiographic abnormalities of the thoracolumbar spinous processes do not differ between yearling and trained Thoroughbred horses without perceived back pain. Journal of the American Veterinary Medical Association, 2023 (Cross-sectional study | Horse)DOI 10.2460/javma.22.09.0419
    In 47 never-ridden yearlings and 55 trained Thoroughbreds without recognisable back pain, narrowed or touching spinous processes were present in roughly a third of the spaces, with no difference between the groups, which argues for a developmental origin and against one acquired through training.
16. Nagy A, Dyson S. Magnetic Resonance Imaging, Computed Tomographic and Radiographic Findings in the Metacarpophalangeal Joints of 31 Warmblood Showjumpers in Full Work and Competing Regularly. Animals, 2024 (Cross-sectional study | Horse)DOI 10.3390/ani14101417
    In 31 showjumpers competing regularly, 53 of 62 limbs showed increased density at the sagittal ridge and the condyles on cross-sectional imaging, which the authors interpret as adaptation to loading, whereas the bone resorption found beneath the cartilage was visible on no radiograph and its clinical meaning remains open.
17. Van Cauter R, Serteyn D, Lejeune JP, Rousset A. Evaluation of the appearance of osteochondrosis lesions by two radiographic examinations in sport horses aged from 12 to 36 months. PLoS One, 2023 (Cohort study | Horse)DOI 10.1371/journal.pone.0286213
    In 58 young sport horses radiographed twice, on average at 407 and 680 days of life, 9 of 36 lesions had changed between the examinations, had newly appeared or had disappeared, and four horses carried their finding at only one of the two appointments: the finding is therefore not stable after the first year of life.
18. Baccarin RY, Pereira MA, Roncati NV, Bergamaschi RR. Development of osteochondrosis in Lusitano foals: a radiographic study. The Canadian Veterinary Journal, 2012 (Cohort study | Horse)PMID 23543926
    Lusitano foals showed radiographic signs of osteochondrosis at the hock or the stifle in 76.08 per cent of cases before their first month of life and still in 16.20 per cent at 18 months, with regression up to the twelfth month and thus later than the age limits assumed until then would suggest.
19. Van Grevenhof EM, Ducro BJ, Van Weeren PR, Van Tartwijk JMFM. Prevalence of various radiographic manifestations of osteochondrosis and their correlations between and within joints in Dutch warmblood horses. Equine Veterinary Journal, 2009 (Cross-sectional study | Horse)DOI 10.2746/042516408x334794
    Of 811 randomly drawn warmblood yearlings, 60.7 per cent had entirely normal contours at the stifle, 68.6 per cent at the hock and 64.6 per cent at the fetlocks, but only 30.5 per cent across all joints together; flattened contours and fragments are statistically two different phenomena and should be assessed separately.
20. Miyakoshi D, Senba H, Shikichi M, Maeda M. A retrospective study of radiographic abnormalities in the repositories for Thoroughbreds at yearling sales in Japan. The Journal of Veterinary Medical Science, 2017 (Cohort study | Horse)DOI 10.1292/jvms.16-0425
    Of 1,082 yearlings, 82, that is 7.6 per cent, never made a start; there was a statistical association between certain findings and the missing start, yet among twelve horses traced individually none had failed because of the finding, which leads the authors themselves to question a causal link.
21. Esselman AM, Johnson SA, Frisbie DD, Barrett MF. Substantial variability exists in the interpretation of survey radiographs among equine veterinarians. Equine Veterinary Journal, 2024 (Cross-sectional study | Horse)DOI 10.1111/evj.14045
    211 veterinary surgeons surveyed graded the same image series on a four-point scale: concern rose with the extent of the change but stayed widely spread at every step, and the direction of the difference depended on the region, with more concern about the navicular bone and less about the hock among specialists.