Introduction

We developed a test to detect circulating hemangiosarcoma-associated cells in dogs prior to disease onset. Dogs were assigned into risk categories for hemangiosarcoma development, allowing for the rational deployment of eBAT, a drug targeting cancer-stem cells and the tumor niche, as a means for cancer prophylaxis.

Methods

125 dog samples (28 hemangiosarcoma; 29 other cancers; 27 benign vascular pathology; and 41 healthy) comprised a training set for machine learning. 10-fold cross-validation established parameters for hemangiosarcoma-associated cell detection and determined sensitivity and specificity. We applied these algorithms to a prospective cohort of 209 clinically healthy golden retrievers, boxers, and Portuguese Water Dogs as a validation set to assess test performance.

Results

Classification accuracy was 85% for healthy dogs and up to 89% for hemangiosarcoma. The test has outstanding specificity (95%) and acceptable sensitivity (89%) for early detection: the false negative rate at 6-months after testing was <1% (1/99). Twenty-one dogs in the prospective cohort developed cancer or another chronic condition; nineteen with a prediction of cancer-associated pathology were diagnosed with cancer (90%). Two of these 21 dogs (10%) were misclassified. One had a cancer prediction but died from a non-malignant condition. The other had a “healthy” prediction but died of cancer.

Clinical Significance

To our knowledge, this is the first test that accurately assigns risk for hemangiosarcoma development in clinically healthy dogs, providing rationale for cancer chemoprophylaxis. Our study also provides proof-of-concept for prospective trials for early cancer detection in companion dogs.