Haemangiosarcoma is a malignancy arising from vascular endothelial cells or their bone marrow-derived precursors, and its biology makes it one of the most feared diagnoses in small animal practice. Tumours typically develop in the spleen, heart or skin, and with the exception of the cutaneous form they behave aggressively, metastasising early and producing vague, intermittent clinical signs that are easily mistaken for less serious conditions. Previous work has suggested the disease accounts for 40 to 64 percent of primary splenic disease and 46 percent of cardiac tumours in dogs. Because internal bleeding from a ruptured splenic mass is frequently the first visible sign, many affected dogs reach the clinic only in crisis. The new study’s finding that annual incidence and 12-month period prevalence were relatively similar, at 0.036 and 0.046 percent respectively, underscores how short survival typically is: dogs rarely live long enough with the disease for prevalence to accumulate far beyond incidence.

The sheer scale of the dataset allowed the researchers to do something earlier studies could not. Much of the existing literature on canine haemangiosarcoma comes from referral hospitals or insured populations, both of which skew toward owners who can afford extensive diagnostic workups. That selection bias limits how well findings generalise to the broader dog population. By contrast, VetCompass captures the everyday caseload of first-opinion practices, where most dogs actually receive care and where many diagnoses rest on clinical judgement rather than histopathology. The team ran two parallel nested case-control analyses: one including all clinically diagnosed cases regardless of laboratory confirmation, and a second restricted to the 273 cases supported by histopathological evidence. Comparing the two offered a way to test whether the broader clinical diagnoses still carry epidemiological signal, and reassuringly, the risk factor patterns proved broadly consistent across both models.

Breed emerged as the dominant risk factor, and the results contain genuine surprises. Ten breeds showed significantly increased odds of clinical haemangiosarcoma compared with crossbreeds, led by the Dogue de Bordeaux with an odds ratio of 9.56, followed by the Flat Coated Retriever at 8.33 and the German Shepherd Dog at 6.35. The Hungarian Vizsla also featured prominently at 5.49. In the histopathologically confirmed subset, the Flat Coated Retriever’s odds soared to 17.5 times those of crossbreeds, with the Vizsla, German Shepherd Dog, Beagle and Boxer also at elevated risk. Notably, the Flat Coated Retriever, Hungarian Vizsla and Beagle had not previously been flagged at this level of confidence in large-scale studies, a gap the authors attribute to the statistical power afforded by their enormous sample. Six breeds, including the Jack Russell Terrier and Yorkshire Terrier, showed reduced odds, offering a counterpoint that may help clinicians weigh risk in small breeds presenting with splenic disease.

The breed findings carry weight beyond the consulting room because they point toward genetic predisposition. When the researchers replaced breed with derived variables in their models, dogs in the lowest bodyweight category under 7.5 kilograms had markedly reduced odds, while risk climbed steadily with size, reaching an odds ratio of 4.42 for dogs weighing 37.5 kilograms or more in the clinical model and 5.17 in the histopathological model. Genotype-based breed grouping told a similar story, with the GG1 group, which includes the German Shepherd Dog, Portuguese Water Dog and Doberman Pinscher, showing roughly threefold increased odds of clinical diagnosis. The authors caution, however, that the prominence of German Shepherd Dogs within these groupings may skew the results, and assuming every breed in a high-risk group shares the risk would be an ecological fallacy. Still, the consistent signal across Kennel Club pastoral and working groups and ancestral mountain-type breeds strengthens the case that inherited factors shape susceptibility, opening a potential route to risk reduction through altered breeding practices and outcrossing in high-risk lines.

Age produced a pattern that was strikingly non-linear. Compared with dogs aged seven to under nine years, those aged 11 to under 13 had the highest odds of clinical diagnosis at 2.83, while dogs aged five to under seven had only 0.22 times the odds. Yet in the histopathological subset, peak odds shifted younger, to the nine-to-eleven-year bracket, and univariable comparisons showed that the probability of obtaining histopathological confirmation actually declined with age. The researchers suggest this reflects owner decision-making rather than tumour biology: older dogs may be less likely to undergo the surgery needed to obtain tissue, whether because of cost, anaesthetic risk or insurance status. There is also a worrying possibility that symptoms in very elderly dogs are attributed to old age, delaying veterinary visits and allowing the disease to kill before it is ever diagnosed, a scenario the authors urge veterinarians to counter by helping owners of senior dogs recognise warning signs.

Reproductive status, socioeconomic context and geography all left fingerprints on the data, albeit more subtly. Neutered males showed increased odds of clinical diagnosis compared with entire females at 1.49, with neutered females at 1.28, a pattern directionally consistent with an Italian register-based study of visceral haemangiosarcoma, though the effect was smaller here and reproductive status dropped out of the histopathological model entirely. Dogs attending clinics in the least deprived areas of the UK had 1.66 times the odds of clinical diagnosis compared with those in the most deprived areas, an association the authors interpret cautiously, noting it may reflect differences in insurance uptake and owners’ willingness to pursue definitive diagnostics rather than true differences in disease occurrence. Intriguingly, dogs attending clinics in mixed urban-rural areas had reduced odds of 0.71 compared with urban clinics, a novel finding given that previous investigations of environment and pollution exposure had been inconclusive.

The study is not without limitations, and the authors are candid about them. Because the data were collected for clinical rather than research purposes, some cases may have gone unrecorded, and conversely some clinically diagnosed cases lacking histopathological confirmation may have been false positives, particularly given the risk of confirmation bias when veterinarians encounter breeds already known to be predisposed. Roughly two-thirds of incident clinical cases were diagnosed without laboratory confirmation, which is simply the reality of primary care where invasive or costly diagnostics are often inappropriate for dogs with advanced disease. The histopathological analysis, with only 273 cases, is likely underpowered for less common breeds, and the restriction of risk factor modelling to dogs aged five and older limits generalisability to younger animals. None of these caveats, however, undermines the central value of the work: a population-level baseline that referral-based studies structurally cannot provide.

For practicing veterinarians and dog owners, the practical payoff is earlier suspicion. Visceral haemangiosarcoma presents with signs so vague and intermittent that any tool sharpening clinical triage matters, and knowing that a middle-aged Flat Coated Retriever, German Shepherd Dog or large-bodied crossbreed faces several-fold elevated odds should prompt a lower threshold for imaging and targeted workup. For researchers, the newly identified high-risk breeds offer fresh leads for the genetic studies needed to unravel a disease whose pathogenesis remains poorly understood despite decades of investigation. And for breeders, the evidence that predisposition clusters within breed lines and genotype groups supports serious consideration of outcrossing strategies in the most affected breeds. With 2.25 million dogs represented and the full dataset made openly available, this study sets a new reference point for canine cancer epidemiology in the UK and a template for how primary-care data can illuminate diseases that are all too often diagnosed only in their final, fatal act.

Subject of Research: Epidemiology and risk factors for canine haemangiosarcoma in UK primary veterinary care

Article Title: Epidemiology of haemangiosarcoma diagnosed in dogs under primary veterinary care in the United Kingdom: Frequency and risk factors

Article References: Epidemiology of haemangiosarcoma diagnosed in dogs under primary veterinary care in the United Kingdom: Frequency and risk factors. (n.d.). https://doi.org/10.1186/s44356-025-00041-4

Image Credits: AI Generated

DOI: 10.1186/s44356-025-00041-4

Keywords: haemangiosarcoma, dogs, veterinary oncology, VetCompass, epidemiology, breed risk, cancer, risk factors, UK, primary veterinary care, genetics, splenic tumours

The retrospective cohort study searched electronic patient databases for all dogs diagnosed with lymphoma or lymphosarcoma between 1 January 2000 and 31 December 2021, with case status reviewed in May 2023. To be included, dogs required a diagnosis confirmed by cytology or histology and consistent clinical findings, reviewed by a board-certified veterinary oncologist. Dogs that had received prior chemotherapy, radiation or immunotherapy were excluded, although those previously treated with glucocorticoids were retained. The final cohort of 1366 dogs represented 1.6 percent of the hospital population over the study period, with 451 cases from the university teaching hospital and 915 from a private referral centre. More than 105 breeds were represented, alongside crossbred dogs.

The demographics of the cohort revealed a disease of middle-aged and older dogs. Ages ranged from just over two months to 17.7 years, with a median of 8.5 years, and median body weight of 23.2 kilograms. Neutered animals dominated the population, accounting for 84.5 percent of cases. The overwhelming majority of dogs, 72 percent, presented with multicentric or nodal disease, and 93.7 percent of those were at stage III or higher at diagnosis, underscoring how frequently lymphoma is advanced by the time it is recognised. Cutaneous lymphoma was the next most common form at 10.2 percent, followed by gastrointestinal disease at 6.1 percent and mediastinal disease at 2.9 percent. Where immunophenotyping was performed, B-cell tumours slightly outnumbered T-cell tumours, 53 percent to 45 percent.

The researchers also mapped the interconnections between laboratory abnormalities and clinical presentation, uncovering a web of statistically significant associations. Anaemia was more frequent in dogs with advanced stage, gastrointestinal disease, substage b, thrombocytopenia, hypoalbuminaemia and reduced appetite. Hypercalcaemia clustered with mediastinal disease, T-cell immunophenotype and advanced stages. Reduced appetite, recorded consistently in the medical records independently of substage, was associated with nearly every adverse marker, including anaemia, thrombocytopenia, hypercalcaemia and hypoalbuminaemia. These patterns suggest that a constellation of systemic derangements travels together in canine lymphoma, and that appetite may serve as a simple bedside indicator of disease burden.

Treatment patterns reflected the realities of veterinary oncology practice. The most common initial approach was a CHOP-based combination protocol, built on cyclophosphamide, doxorubicin, vincristine and prednisolone, used in 480 dogs, with a further 62 dogs receiving T-cell specific regimens. Other dogs received single-agent doxorubicin, mitoxantrone or lomustine, glucocorticoids alone, surgery, or a variety of other combinations. Notably, 355 dogs, or 26 percent, received no recorded treatment, and 26 were euthanised at diagnosis, figures that highlight the difficult decisions owners face when confronted with an aggressive cancer and the cost and commitment that intensive chemotherapy demands.

The core outcome analysis focused on the 542 dogs treated with CHOP-based protocols, of which 532 had sufficient pre-treatment data. The overall response rate was 87.5 percent, with 67.4 percent achieving complete remission and 20 percent partial remission. The median progression-free survival time was 200 days, and the median overall survival time was 252 days, with one-year survival estimates of roughly 25.6 percent for progression-free survival and 37.2 percent for overall survival. These figures place Australian dogs squarely within, though at the lower end of, the ranges reported from large European and North American studies, a reassuring finding given concerns that quarantine-driven genetic isolation, different breed demographics and environmental factors might produce divergent outcomes.

Perhaps the most striking result was the identification of the treating hospital itself as an independent prognostic factor for both progression-free and overall survival, despite the two centres sharing similar climates, breed profiles and referral-based management. The researchers could not identify an obvious difference in clinical management, other than less frequent steroid use at one hospital, and speculate that unmeasured differences in case mix or unidentified biases may be responsible. The finding carries a practical message for the field: when multi-centre veterinary studies pool cases to boost numbers, the treating clinic should be treated as a variable in the analysis, something rarely done in veterinary oncology to date.

Two further novel or rarely reported prognostic variables emerged. Reduced appetite at diagnosis independently predicted shorter progression-free and overall survival, a factor that has been investigated in only two prior studies and confirmed in just one, possibly because appetite is usually folded into substage rather than analysed on its own. And in a subset analysis restricted to breeds with ten or more cases, breed significantly influenced overall survival: compared with crossbred dogs, boxers, golden retrievers and Labrador retrievers had significantly shorter survival, with hazard ratios of 1.98, 2.09 and 3.02 respectively. Intriguingly, all three breeds are also recognised as being at elevated risk of developing lymphoma in Australia, hinting that the genetic factors which predispose certain breeds to the disease may also shape how their tumours respond to treatment.

The study also weighed in on a long-running controversy: the role of glucocorticoids within combination chemotherapy. Prior glucocorticoid exposure before chemotherapy showed no significant effect on outcome in any group, allowing clinicians some reassurance that short-term palliative steroid use while awaiting chemotherapy is unlikely to compromise later results. However, dogs whose CHOP protocols included prednisolone had significantly shorter progression-free and overall survival and lower remission rates than those whose protocols omitted it, even after multivariate adjustment. Because steroid inclusion was at clinician discretion, bias towards treating sicker dogs with steroids cannot be excluded, and the authors caution that this finding requires independent confirmation before practice changes.

The consistent prognostic anchor across nearly all analyses was response to initial treatment, confirming its status as the most reliable predictor of survival in canine lymphoma, yet one that cannot be known before therapy begins. The authors argue that identifying an accessible, cost-effective predictor of response at diagnosis remains one of the field’s most pressing needs, and would transform the conversations veterinarians have with owners contemplating treatment. For Australian dogs and their veterinarians, the broader message is one of connection: outcomes in this isolated population mirror those reported internationally, meaning that evidence from global studies can be applied to Australian patients, and Australian data, now substantially richer, can inform the world.

The scale of the new study deserves emphasis in context. Large veterinary lymphoma cohorts exceeding 200 treatment-naïve dogs are rare in the global literature, and the Australian dataset of 1366 cases, accumulated over 22 years, now ranks among the most extensive regional collections published anywhere. This matters because most of what veterinarians know about canine lymphoma prognosis derives from North American and European institutions, where breed distributions, referral patterns and owner attitudes toward chemotherapy may differ substantially from those in an island continent with strict quarantine controls shaping its dog population for generations.

The methodological grounding of the work also strengthens its conclusions. Survival endpoints were calculated in alignment with the Veterinary Comparative Oncology Group consensus criteria, the standardised response framework adopted across veterinary oncology since 2010, with earlier records retrospectively mapped onto those definitions. Progression-free and overall survival were measured from the commencement of treatment, and dogs still in remission at last examination were right censored rather than counted as failures. Overall survival was chosen over lymphoma-specific survival because, even when another cause of death was recorded, the authors could not exclude lymphoma as a contributing factor, a conservative decision that avoids inflating favourable outcomes.

The definition of a modified CHOP protocol also clarifies how real-world practice diverges from textbook regimens. Substituting mitoxantrone for doxorubicin in dogs with cardiac disease, adding agents such as L-asparaginase, or extending protocols to 19 or 25 weeks all qualified, reflecting the clinical judgement that shapes individualised care. By analysing these dogs alongside standard CHOP recipients, the study captures the heterogeneity of referral practice that pure prospective trials often exclude, at the cost of some protocol standardisation.

Several limitations temper interpretation. As a retrospective cohort, the study depended on owner willingness and financial capacity to pursue complete staging, meaning some recorded stages represent minimum estimates, particularly for dogs clinically at stage III. Immunophenotyping was not performed in every case, and the glucocorticoid findings rest on clinician discretion rather than randomisation, leaving residual confounding by case severity an open possibility. The hospital effect, while statistically robust, remains biologically unexplained.

For practicing veterinarians, the practical takeaways are concrete. Appetite at diagnosis, already routinely recorded, merits attention as an independent prognostic marker rather than a mere component of substage. Multi-institutional studies should treat the treating hospital as an analytical variable. And the finding that Australian outcomes sit within international ranges, albeit at the lower end, provides a defensible evidence base for applying global trial results to Australian patients while encouraging local participation in future multi-centre studies.

Subject of Research: Presentation, treatment outcomes and prognostic factors for canine lymphoma in Australian dogs

Article Title: A retrospective study on the presentation and outcomes for Australian dogs with lymphoma: 1366 cases

Article References: Bennett, P., Williamson, P., & Taylor, R. (2026). A retrospective study on the presentation and outcomes for Australian dogs with lymphoma: 1366 cases. Veterinary Oncology, 3(1), Article 19. https://doi.org/10.1186/s44356-026-00070-7

Image Credits: AI Generated

DOI: 10.1186/s44356-026-00070-7

Keywords: canine lymphoma, dogs, chemotherapy, CHOP protocol, prognostic factors, veterinary oncology, survival, glucocorticoids, breed risk, Australia, B-cell lymphoma, T-cell lymphoma