How the test was developed
The canine lymphoma blood test resulted from many years work with veterinary oncologists, first opinion practices and mathematicians in the UK, USA and the Netherlands.
The canine lymphoma blood test resulted from many years work with veterinary oncologists, first opinion practices and mathematicians in the UK, USA and the Netherlands. The work comprised of two phases:
- The discovery of new biomarker tests for canine lymphoma.
- The refinement of these tests using mathematical models capable of quantifying and monitoring disease progress in dogs diagnosed with lymphoma.
We have previously reported that mass spectrometry was able to discover multiple biomarkers capable of differentiating between dogs diagnosed with lymphoma from healthy dogs and those diagnosed with diseases presenting in a similar manner to lymphoma. Subsequently we demonstrated that one of the biomarkers identified by mass spectrometry was the acute phase protein Haptoglobin (HAPT).
Other publications have also reported that CRP levels increase in canine lymphoma and CRP tests are now routine in the diagnostic work up of Non-Hodgkins Lymphoma in Humans. However, CRP alone lacks specificity to assist in the diagnosis of lymphoma in dogs. By adopting a multivariate analytical approach to develop algorithms employing both CRP and Hapt values, we have been able develop a new test which overcomes the limitations found by measuring single APPs.
By working with the University of Leicester, a multivariate algorithm was developed which draws on a database of samples (control and lymphoma), using the relative levels of C-RP and HAPT, the age and sex of the animal and the presence of lymphadenopathy. All samples in the database were confirmed as positive either by cytology or histopathology and all negative samples were confirmed to be free of the disease for a minimum of 6 months after the serum had been taken. The algorithm uses three distinct analytical processes to assess the risk of lymphoma in each dog being presented to it. The algorithm combines decision trees (as shown in Figure 1 below), k nearest neighbour analysis (KNN) and probability density function estimation (PDFE) [Scott, 1992]. Further information is available upon request or you can view online at http://arxiv.org/abs/1305.4942.
The resultant algorithm was used in a blinded retrospective study to investigate remission and recurrence of lymphoma in dogs receiving chemotherapy. The objective was to compare clinicians’ assessment using palpation and cytology to the results of serum biochemical tests for haptoglobin (HAPT) and C-reactive protein (C-RP). We called this multivariate method the canine Lymphoma Blood Test (cLBT)