The motivation behind the study is compelling. Cancer is a leading cause of death in companion dogs, and its incidence climbs steeply with age. A 2025 survey by the veterinary imaging company HT Vista found that 62 percent of masses seen in US veterinary clinics go undiagnosed, underscoring a glaring diagnostic gap. Because dogs also serve as valuable spontaneous models for human oncology, a cheap, scalable screening method built on data that veterinarians already collect every year, the complete blood count and serum biochemistry panel, would be transformative. The central hypothesis was that although individual blood parameters are uninformative on their own, subtle multivariate patterns hidden across dozens of values might harbor a presymptomatic signature of malignancy that machine learning could unlock.

To test that hypothesis rigorously, the study harnessed the Golden Retriever Lifetime Study, a large prospective observational cohort of 3,044 purebred Golden Retrievers enrolled between 2012 and 2015 and followed for their entire lives. Annual visits generate full physical examinations, owner and veterinarian questionnaires, and biospecimens processed for complete blood counts and chemistry profiles on standardized laboratory instruments. Cancer diagnoses were assembled from two complementary sources covering all three tiers of diagnostic confidence, from histologically confirmed cases down to presumptive clinical suspicions, yielding a final cancer cohort of 659 unique subjects. The most common tumor types were hemangiosarcoma at just over a quarter of cases, followed by mast cell tumors and lymphoma.

The resulting analytical dataset of 22,460 veterinary visits presented a formidable statistical challenge: only 6.3 percent of visits were cancer-positive, a degree of class imbalance notorious for derailing machine learning algorithms. The study design deliberately embraced real-world messiness. Visits were labeled positive only if they occurred on or after the recorded diagnosis date, and for the 41.4 percent of cancer subjects with no post-diagnosis laboratory work, only the final pre-diagnosis visit was labeled positive. Crucially, the model was trained without discriminating for treatment status, meaning it was exposed to laboratory results from both treated and untreated periods after diagnosis, exactly as real-world screening data would be.

The methodological framework was exhaustive. Six base algorithms were compared, spanning logistic regression, random forests, two gradient boosting methods, a neural network, and Naïve Bayes. These were crossed with three feature selection strategies, including recursive feature elimination and a manually curated panel of fifteen biomarkers tied to paraneoplastic syndromes such as anemia, thrombocytopenia, and hypercalcemia, and with six resampling techniques for handling imbalance, from SMOTE to class weighting. Every one of the 126 pipelines underwent five-fold cross-validated grid search optimized for the Matthews Correlation Coefficient, a metric well suited to imbalanced data. Data were split at the patient level to prevent information leakage, ensuring that no dog’s visits appeared in more than one partition and that the held-out test set provided a genuinely unbiased estimate of generalization.

The winning pipeline, a logistic regression model with class weighting and recursive feature elimination, achieved a respectable area under the ROC curve of 0.815 on the test set, with a 95 percent confidence interval of 0.793 to 0.836. That figure means a randomly chosen cancer-positive visit had an 81.5 percent chance of receiving a higher risk score than a randomly chosen cancer-negative visit, proof that a genuine signal exists in the data. But ranking ability is not the same as classification, and here the model failed decisively. Its F1-score was just 0.25, and its positive predictive value a mere 0.15, meaning that of all visits flagged as high risk, only 15 percent were actual cancer cases while 85 percent were false alarms.

The precision-recall curve told the clinical story plainly. Sitting only slightly above the no-skill baseline set by the 6.3 percent cancer prevalence, it showed that achieving any meaningful sensitivity would drive precision to levels that would flood clinics with false positives. The model’s high negative predictive value of 0.98 initially suggested promise as a rule-out test, but that hope collapsed under scrutiny of its recall: at 0.79, the model would miss roughly one in five dogs with cancer, an unacceptable false-negative rate for any screening application. The study’s conclusion was unambiguous: the model, and likely any model built solely on this data modality, is unsuitable for clinical deployment.

Explainable AI analysis using SHAP, a game-theory-based method that quantifies each feature’s contribution to individual predictions, revealed why the model fell short. The single most powerful predictor was patient age, followed by markers of anemia, such as lower hemoglobin, and systemic inflammation, such as elevated band neutrophils and a higher neutrophil-to-lymphocyte ratio. These are biologically plausible features, but they are profoundly non-specific. In effect, the model had learned to identify older dogs showing signs of chronic illness rather than a distinctive signature of cancer, functioning more as an old-dog detector than a malignancy biomarker. Cancer-induced blood changes are simply too entangled with those caused by common geriatric conditions, from chronic kidney disease to inflammatory bowel disease, for a bloodwork-only model to separate them.

A further and arguably more insidious problem was time-varying confounding by treatment. Because the dataset included post-diagnosis visits, the model could associate the iatrogenic effects of therapy with the cancer label. A dog with lymphoma treated with a multi-agent chemotherapy protocol will develop a stress leukogram and possible hyperglycemia; a dog with a mast cell tumor on a tyrosine kinase inhibitor may develop hypoalbuminemia from gastrointestinal toxicity. The model thus learned a confounded causal pathway, from cancer to diagnosis to treatment to blood changes, rather than the desired direct pathway from cancer to blood changes. Addressing this will require advanced causal inference methods such as marginal structural models, the study notes.

The study’s limitations are candidly acknowledged. Grouping dozens of histologically distinct malignancies into a single multi-cancer label, a concession to limited case numbers for any one tumor type, inevitably biased the model toward generic markers of systemic illness. The population was restricted to a single breed with known genetic predispositions to certain cancers, limiting generalizability, and the model lacked access to physical exam findings, comorbidities, and owner-reported signs. Yet these constraints are precisely what make the benchmark valuable: it quantifies, under realistic conditions, the maximum performance achievable from routine laboratory data alone. The path forward, the author argues, lies not in more complex algorithms but in multi-modal data integration, combining bloodwork with imaging, medical records, molecular diagnostics, and clinical context to emulate the reasoning of an expert clinician, validated in large, diverse, external cohorts before any clinical use.

Subject of Research: Machine learning assessment of routine laboratory data for early canine cancer detection on an imbalanced dataset

Article Title: Assessing the feasibility of early cancer detection using routine laboratory data: an evaluation of machine learning approaches on an imbalanced dataset

Article References: Assessing the feasibility of early cancer detection using routine laboratory data: an evaluation of machine learning approaches on an imbalanced dataset. (n.d.). https://doi.org/10.1186/s44356-025-00054-z

Image Credits: AI Generated

DOI: 10.1186/s44356-025-00054-z

Keywords: veterinary oncology, machine learning, canine cancer, early detection, imbalanced data, routine laboratory data, Golden Retriever Lifetime Study, predictive modeling, SHAP, class imbalance, bloodwork screening, logistic regression

Mast cell tumours are the most frequently diagnosed cutaneous malignancy in dogs, accounting for an estimated 16 to 21 percent of all skin neoplasms. They arise from mast cells, immune cells rich in inflammatory mediators such as histamine and heparin, and typically present as masses in the skin or the tissue beneath it. Their biological behaviour spans an extraordinary range: some are cured permanently by simple surgical excision, while others ulcerate, recur locally and spread to lymph nodes, liver and spleen. Paraneoplastic effects driven by the release of mast cell granules can produce itching, bruising, skin swelling and gastrointestinal signs, complicating the clinical picture. Two histological grading systems, the three-tier Patnaik scheme and the two-tier Kiupel scheme, are widely used to predict behaviour, and both have been shown to be consistently prognostic, particularly when applied together.

Golden Retrievers have repeatedly been identified as a breed at increased risk of MCTs, and germline genetic studies have pinpointed risk variants in genes including GNAI2 and a hyaluronidase gene in US Golden Retrievers. Yet until now, no epidemiological study had specifically characterised MCT frequency, clinical features and survival within the breed in the United States. The GRLS, a prospective cohort of 3,044 Golden Retrievers enrolled between June 2012 and April 2015 and balanced by sex across five geographic regions, offered a uniquely powerful opportunity. Dogs were followed annually with owner and veterinarian questionnaires, biological samples and a dedicated biopsy submission pathway, with histopathology reviewed by two blinded veterinary pathologists. The study’s primary cancer endpoints include haemangiosarcoma, lymphoma, osteosarcoma and high-grade MCT, with high grade defined as Patnaik grade 3 or Kiupel high.

The headline numbers are striking. Of the 3,044 dogs, 164, or 5.39 percent, were diagnosed with a total of 234 mast cell tumours, yielding an incidence rate of 5.58 per 1,000 dog years at risk. A lifetable analysis revealed that young dogs were rarely affected: 99.1 percent of the cohort reached five years of age without an MCT diagnosis. Annual incidence risk peaked at 1.39 percent in dogs aged greater than eight to nine years, before declining in the oldest age groups. The authors note that this decline mirrors patterns seen in other veterinary and even human cancers, where incidence falls in extreme old age, possibly reflecting cellular senescence, stem cell exhaustion or a cancer-resistant phenotype, though diagnostic bias in geriatric patients may also contribute.

Compared with incidence figures from a mixed-breed insured population in Sweden, the Golden Retriever rate appears up to tenfold higher, which the researchers interpret as evidence of a genuine and substantial breed predisposition rather than an artefact of geography or study design. The practical message for owners is clear: vigilance for skin masses, particularly as dogs enter middle and older age, is warranted, and any new lump should prompt veterinary investigation rather than a wait-and-see approach. The median age at first diagnosis in the cohort was just over eight years.

The tumours themselves were predominantly low grade. Cutaneous tumours accounted for 70.5 percent of events and subcutaneous tumours for 16.7 percent, with the remainder being metastatic entries, mucosal or visceral primaries, or digital masses. The most common anatomical locations were the torso, limbs and head and neck, consistent with distributions reported in multi-breed populations. Histopathological grading was available for nearly all cutaneous tumours, and the most frequent grade was the intermediate-low P2KL category. Overall, only 47 of the 234 tumours, or 20.1 percent, met the study’s high-grade definition, a proportion consistent with previous reports of 22 to 26 percent high-grade disease in broader populations. Notably, the findings do not support earlier suggestions that Golden Retrievers are disproportionately prone to high-grade tumours.

Recurrence and spread were comparatively uncommon. Most dogs, 84.2 percent, had a solitary tumour, while 15.9 percent developed new de novo masses at distinct sites, with a median interval of 368 days between the first and subsequent tumours. Local recurrence occurred in just under five percent of cases, and documented metastasis in 9.8 percent, most often to lymph nodes. Critically, every dog with metastatic disease had at least one high-grade tumour; no dog with only low-grade MCTs was found to have spread. The authors suggest this could influence clinical staging recommendations, since extensive staging appears to have low diagnostic yield in low-risk patients, although the absence of standardised staging protocols means some metastases may have gone undetected.

The survival analysis produced the study’s most reassuring result. Median lifespan was 11.68 years in dogs diagnosed with MCT and 11.75 years in the remaining cohort, a difference that was not statistically significant. Median survival time from first MCT diagnosis to death from any cause was 1,367 days, and only 19 dogs, or 11.6 percent, had MCT recorded as their cause of death. Overall survival probabilities after diagnosis were 81.7 percent at one year, 70.0 percent at two years and 40.6 percent at five years. When the analysis was stratified, however, the prognostic weight of tumour biology became evident: the presence of metastasis, a high histological grade, or local recurrence were each associated with statistically significantly reduced survival, while the development of additional de novo tumours was not. Nearly 90 percent of the dogs that died from their disease had at least one high-grade tumour.

The researchers acknowledge several limitations. Cohort recruitment relied on word-of-mouth and snowball sampling, dogs were required to have three-generation pedigrees, and participating owners were highly engaged, with subsidised histopathology likely boosting diagnostic rates relative to the wider population. Treatment details were largely unavailable, precluding analysis of therapeutic impact, and the number of MCT-related deaths was too small for multivariable modelling. Nevertheless, the study’s general-practice setting gives it a realism that insurance-claim and referral-based studies lack. Its overall message is one of measured optimism: for most Golden Retrievers, a mast cell tumour diagnosis is not life-limiting, low-grade disease carries an excellent prognosis with prompt and appropriate management, and the findings justify early investigation and treatment of skin masses rather than fatalism. As the GRLS cohort continues to yield data, it is cementing its role as one of the most valuable resources in comparative oncology, with implications that may ultimately extend to human mast cell disorders and cancer research more broadly.

Subject of Research: Epidemiology, clinical features and survival of mast cell tumours in Golden Retrievers enrolled in the Golden Retriever Lifetime Study

Article Title: Mast cell tumours in the Golden Retriever Lifetime Study: a cohort study assessing frequency, clinical features and survival

Article References: Stratton, Z. V., Brodbelt, D. C., Guillén, A., O’Neill, D. G., Labadie, J., Swafford, B., & Taylor, C. (2026). Mast cell tumours in the Golden Retriever Lifetime Study: a cohort study assessing frequency, clinical features and survival. Veterinary Oncology, 3(1), Article 10. https://doi.org/10.1186/s44356-026-00060-9

Image Credits: AI Generated

DOI: 10.1186/s44356-026-00060-9

Keywords: Golden Retriever, mast cell tumour, canine cancer, veterinary oncology, Golden Retriever Lifetime Study, dog lifespan, histological grading, Kiupel, Patnaik, metastasis, incidence, survival analysis