HOHorseBmoderate
In 80 dressage horses referred for investigation, reported signs did not differ between horses with and without a lesion. What symptoms are really worth.
Executive summary
Behavioural signs say less about the stomach than their reputation suggests. In 80 dressage horses, the frequency of owner-reported signs did not differ between horses with and without a lesion on gastroscopy, and their disappearance did not predict healing of the mucosa. Two distinct diseases hide under a single name, the squamous and the glandular form. Only gastroscopy establishes a finding.
Key points
Anyone looking for information on gastric ulcers in the horse usually finds a single disease with a long list of symptoms attached. The specialist literature has separated two entities since 2015. The consensus statement of the European College of Equine Internal Medicine distinguishes disease of the squamous mucosa, that of the glandless upper part of the stomach, from disease of the glandular mucosa, which does carry glands. The two differ in how they arise, in their risk factors and in their response to treatment. The umbrella term that brings them together has been in use since 1999.
This separation is not a subtlety for specialists: it governs the interpretation of every single sign. A review from 2023 notes that a large share of the work on the glandular form appeared only after the consensus statement. Reading a symptom list that merges the two forms therefore means reading a list asked to answer two different questions at once.
Because in the living horse both mucosal surfaces can be seen only by endoscopy. The consensus statement names gastroscopy expressly as the only method that establishes a diagnosis. Everything that comes before it remains a suspicion. How solid that suspicion is can nevertheless be measured, and that is exactly what several research groups have done.
A sign is never worth more than the frequency of the disease it is meant to announce. And that frequency is high. In a Danish study of 201 horses that their owners considered healthy and that were not in race training, more than half carried a lesion judged significant: 53 per cent. An editorial from 2025 gathers, for sport horses, frequencies of 50 to 70 per cent for the squamous form and up to 75 per cent for the glandular form.
In horses in training the picture is sharper still. In a synthesis of seven randomised trials covering 566 horses, the untreated comparison groups showed ulceration in 75.3 and 87.2 per cent of cases depending on the subgroup. Finding a lesion in a horse in training is therefore the normal case, and not a discovery.
The obvious counter-calculation does not hold either. In an Icelandic cohort, 81 horses were examined with the camera within two weeks of moving from pasture into the stable: 71.6 per cent already carried a significant squamous lesion. Eight weeks later, in training and with conserved forage fed several times a day, the figure was 25.4 per cent. The authors attribute the difference expressly to the forage feeding and not to the behaviour of the horses.
The most direct test so far dates from 2025. In all, 80 dressage horses in competition were referred by their vets for investigation and examined with the camera; 55 carried a lesion, 25 did not. The comparison covered how often owners reported particular signs. Across the whole group, the most frequently named were girth sensitivity in 78 per cent, declining performance in 33 per cent and weight loss in 30 per cent. The frequency of these signs did not differ between horses with and without a lesion.
The result has two sides, and both count. Neither the reported signs nor a test for occult blood in the droppings reached a rate of correct answers sufficient to carry out a pre-selection. And in the 19 horses re-examined with the camera after treatment, the disappearance of the signs did not predict healing of the mucosa: the agreement between what owners reported and what the camera showed was poor.
No. As early as 2006, a Swedish study of 80 Standardbreds in training reached the same conclusion: age, sex, performance, behaviour and feeding habits gave only weak pointers to which horses were affected. Of all the characteristics examined, only training status was clearly associated with squamous lesions.
An Icelandic cross-sectional study of 211 riding horses from 21 premises examined a broader list in 2023. Neither body condition, nor behaviour in the box and the paddock, nor the intensity and frequency of work, nor age were clearly associated with the ulcer score. What was associated instead were management characteristics: the region, the sex, the time spent in the paddock and the proportion of non-structural carbohydrates in the ration. The same work does report, however, that horses showing the signs usually attributed to the stomach more often had a significant glandular and a severe squamous change. The signs are therefore not worthless: they are weak and unspecific.
No sign is tied to the stomach as firmly as sensitivity when the girth is tightened. The figures behind it are sobering. In an American case series, 37 horses were presented between 2004 and 2016 for girth sensitivity. In 12 of them it was a gastric ulceration, in 10 an orthopaedic problem of varying kind, in 3 a poorly fitting saddle, in 1 a tumour of the reproductive tract and in 10 other conditions, among them a liver abscess, an aneurysm of the caudal vena cava, sternal pain and a urinary tract infection. That list covers 36 of the 37 horses. The authors themselves note that the exact cause of girth sensitivity remains hard to establish.
The reverse direction is just as instructive. In the populations gathered by the 2025 editorial, 25 per cent of the horses carrying a squamous lesion and 31 per cent of those carrying a glandular lesion showed girth sensitivity. Put differently: about seven horses in ten whose gastric involvement was established did not react noticeably to girthing. A sign that is absent in the majority of affected horses and occurs at a similar frequency in the others does not separate the two groups.
A 2026 study of 52 horses examined with the camera asked which type of sign goes with which lesion. The pattern was clear: physical signs appeared above all with squamous involvement, behavioural signs and mixed pictures above all with glandular or mixed involvement. What did not appear was just as clear: between the severity of the lesions and the clinical signs there was no link at all. The discipline the horse was used for was likewise associated both with the type of signs and with the site of the lesion.
This work describes the distribution of signs in horses already diagnosed. There was no comparison group free of findings on gastroscopy, and therefore no statement either about how accurate a sign is. It is a single result, not replicated to date, and the authors write themselves that the relationship between the site of the lesion and the clinical picture remains poorly defined.
A study of 109 Thoroughbreds across eight training yards in Great Britain and Australia likewise found different patterns for the two forms, one of which contradicts the common image: horses that behaved aggressively towards people less often had a squamous lesion, and not more often. Horses with stereotypies had one more often. Work spread over five or more days a week went together with a clearly raised risk of the glandular form, as did racing below expectations. The estimates are imprecise; the direction is less so.
To attribute a sign to the stomach, one has to know how often it has other explanations. On this point the figures are solid. In a convenience sample of 60 riding horses whose owners considered them free of complaints, 73 per cent were lame on at least one limb, mostly mildly. Behaviour under saddle was explained in that study by two quantities: lameness and the skill of the person riding.
The most striking counter-test comes from 150 horses investigated for declining performance. After a musculoskeletal source of pain had been removed by diagnostic anaesthesia, in some cases together with a change of saddle, the behaviour score under saddle fell from a median of 9 to 2 out of a possible 24 points. Only 30.0 per cent of these horses were consistently lame, and 37.3 per cent wore a saddle judged to fit badly enough to affect performance. There was no link between the behaviour score and the degree of lameness. This work did not perform gastroscopy: it therefore says nothing about how many of these horses also carried a gastric lesion.
And even once lameness has been ruled out, the list stays long. In 259 non-lame Standardbreds examined for declining performance, equine asthma and gastric disease headed the diagnoses, ahead of exercise-induced pulmonary haemorrhage, dynamic narrowing of the upper airways, cardiac rhythm disturbances and exertional myopathies. The 2025 editorial further cites an analysis in which 84 per cent of 275 racehorses referred for declining performance had more than one possible cause after full investigation.
Gastroscopy is the yardstick against which everything else is measured, and it too is not free of error. In one study, six people assessed 60 anonymised video recordings. With the ordinal grading system of the EGUS Council, agreement between assessors was substantial for both mucosal surfaces, without being complete, and a little higher within the same person at two different times. A visual analogue scale performed clearly worse, especially for the glandular mucosa, and depended in addition on the experience of the person assessing.
This work says nothing about the correctness of the assessment, only about its repeatability, and it rests on video recordings rather than on examining the horse. In everyday practice, with varying image quality and preparation, the spread is likely to be greater. For the question asked in this article that means the yardstick against which behavioural signs fail is not itself a perfect yardstick.
The reverse route, measuring the pain rather than the lesion, has not proved a shortcut either. A validated facial expression scale did not separate, among 61 horses, the animals carrying gastric lesions from those without, and its value did not rise with severity. The authors expressly do not conclude that gastric ulcers are painless: they hold that this instrument does not capture that pain, and call for other scales to be tested. Lame horses and horses with systemic inflammation were excluded, that is, precisely the group most common in practice.
Part of the puzzle dissolves if behaviour is seen not as a simple readout but as something learnt. A clinical commentary from 2025 describes problem behaviours persisting in some horses after healing confirmed on gastroscopy: as learnt, anticipatory reactions to a pain once experienced. Conversely, part of the behaviour gives way after treatment. The two together explain why the disappearance of a sign does not prove a cure, and its persistence does not prove a failure.
There are studies pointing the other way, and they belong in the picture. In a clinical study of 63 sport horses carrying a glandular lesion, healing of the lesions was associated with the disappearance of the clinical signs; declining performance was the most common presenting sign in both groups, in 65.1 per cent overall. The authors conclude from this that the glandular form does indeed produce clinical signs. The reservation lies in the design: the clinical data came from questionnaires sent to owners who knew which treatment their horse had received.
Just as double-edged is a trial in nine show jumpers from the same yard. After twelve weeks on a low-starch ration instead of one rich in sugar and starch, the gastric score and the behaviour score under saddle fell together, and the two values were closely linked. The recordings were assessed blind. There was, however, neither a comparison group nor random allocation: twelve additional weeks of training explain part of the change on their own. And two values falling together says nothing about which of them moves the other.
Original analysis
| Sign | Where it was measured | Reported link with the finding | Other explanations documented in the source set | What does not follow |
|---|---|---|---|---|
| Girth sensitivity | 80 dressage horses; 37 horses in a case series; populations gathered by an editorial | Reported in 78 per cent of the whole group, with no difference between horses with and without a lesion; present in 25 to 31 per cent of horses whose involvement was established | Orthopaedic problem, poorly fitting saddle, tumour of the reproductive tract, liver abscess, sternal pain, urinary tract infection | That girth sensitivity points to the stomach |
| Declining performance | 80 dressage horses; 63 sport horses with a glandular lesion; 259 non-lame Standardbreds; 109 Thoroughbreds | Reported in 33 per cent of the dressage group, with no difference between groups; the most common presenting sign in 65.1 per cent of glandular cases | Equine asthma, exercise-induced pulmonary haemorrhage, dynamic narrowing of the upper airways, cardiac rhythm disturbance, exertional myopathy | That the cause has been found as soon as a gastric lesion is present |
| Weight loss | 80 dressage horses | Reported in 30 per cent of the whole group, with no difference between horses with and without a lesion | Not set against other causes in the studies used here | That weight loss points to the stomach |
| Aggression towards people | 109 Thoroughbreds in race training | Inverse link: less often with a squamous lesion, and not more often | Not broken down further in the same dataset | That aggression grounds a suspicion of the stomach |
| Stereotypies such as crib-biting or weaving | 109 Thoroughbreds in race training; 211 Icelandic riding horses | In the Thoroughbreds more often associated with a squamous lesion; in the Icelandic horses, behaviour in the box and the paddock was not associated with the ulcer score | Management, feeding and the training yard itself emerged as factors in their own right | That a stereotypy is caused by the gastric lesion |
| Facial expression at rest | 61 horses, seven assessors | No difference between horses with and without a lesion, no rise with severity | Lame horses and horses with systemic inflammation were excluded | That gastric ulcers are painless |
| Behaviour under saddle | 60 riding horses believed free of complaints; 150 horses with musculoskeletal pain; 9 show jumpers | In the 9 show jumpers the behaviour score fell at the same time as the gastric score, with no comparison group | Lameness and the skill of the person riding; the score fell from 9 to 2 out of 24 after musculoskeletal pain was removed | That a high behaviour score points to the stomach |
| Behaviour in the box and the paddock | 211 Icelandic riding horses from 21 premises | Not significantly associated with the ulcer score | Region, sex, time in the paddock, proportion of non-structural carbohydrates and a change of pasture within four weeks | That calm behaviour in the stable rules out a lesion |
Not reliably. In the study of 80 dressage horses referred for investigation, the frequency of the signs owners reported did not differ between horses with and without a lesion, and a test for occult blood in the droppings gave no usable separation either. As early as 2006, a Swedish study of 80 Standardbreds held that age, sex, performance, behaviour and feeding habits give only weak pointers. The 2015 consensus statement names gastroscopy as the only method that establishes a diagnosis. Whether such an examination is indicated is decided at the veterinary practice.
Not necessarily. In a case series bringing together 37 horses presented for girth sensitivity, 12 had a gastric ulceration, 10 an orthopaedic problem, 3 a poorly fitting saddle, 1 a tumour of the reproductive tract and 10 other conditions. Conversely, in the figures of a 2025 editorial, only 25 to 31 per cent of the horses whose involvement was established showed girth sensitivity at all. The sign is therefore absent in the majority of affected horses and at the same time has many other causes. It justifies an investigation, but grounds no diagnosis.
From the signs alone, with no certainty at all. A 2026 study of 52 horses examined with the camera did find a pattern: physical signs rather with squamous involvement, behavioural signs rather with glandular or mixed involvement. That result comes, however, from a small group without healthy comparison horses and has not been replicated to date. The same work also shows something else: the severity of the lesions was not associated with the clinical signs. In the living horse, only gastroscopy shows which mucosa is affected.
The evidence suggests that the stomach is neither the only candidate nor regularly the first. Among 60 riding horses whose owners considered them free of complaints, 73 per cent were mildly lame, and behaviour under saddle was explained by lameness and the skill of the person riding. Among 150 horses investigated for declining performance, the behaviour score fell from a median of 9 to 2 out of 24 points after a musculoskeletal source of pain had been removed. The order in which an investigation proceeds belongs to the veterinary practice.
One does not follow from the other. In the 19 horses re-examined with the camera after treatment, the improvement owners reported agreed poorly with the gastroscopic result. A clinical commentary from 2025 further describes some behaviours persisting as a learnt, anticipatory response even though the lesion has healed. Conversely, part of the behaviour does give way after treatment. Signs that have gone therefore prove no cure, and signs that persist prove no failure.
Not automatically. Among 201 Danish horses their owners considered healthy, 53 per cent carried a lesion judged significant, and in the untreated comparison groups of seven randomised trials in horses in training the figure was 75 to 87 per cent. At that frequency, almost every gastroscopy finds something. A 2025 editorial warns expressly against moving from a lesion found to the cause of the behaviour. In one analysis cited, 84 per cent of 275 racehorses referred for declining performance had more than one possible cause.
Terms defined in this text
Equine Health
The percentage ranges attached to radiographic classes come from a veterinary consensus, not a follow-up study: what the data actually show.
HOHorseBmoderate Open access
Movement & Recovery
Of 750 horses at competitions, 44 per cent left no finger under the noseband: what the taper gauge, pressure sensors and a cadaver head show.
HOHorseBmoderate Open access
Equine Health
Among 100 jumping horses in active competition, not one had a normal back on radiographs: what a kissing spines finding says, and what it cannot predi
MUMultiple speciesBmoderateT1hypothesis only Open access
Once a month: new syntheses, checked studies, and what has changed in the state of the evidence.