Introduction
Feline lymphoma is one of the most common cancers for cats with an estimated incidence rate of roughly 50–200 per 100,000. The most common treatments for the disease include the combinations therapies, namely CHOP (cyclophosphamide, doxorubicin, vincristine, prednisolone) and ChlP (chlorambucil, prednisolone). Response and survival can vary significantly across the patients even when treated by the same regimen. Up to 18-fold difference in the progression-free survival (PFS), for example, was observed among the patients treated with ChlP therapy. Given the variability in treatment outcome and plurality of regimens, there is an unmet need for a predictive technology assisting optimization of treatment for feline lymphoma.
Methods
We conducted flow cytometric (FC) analysis of the fine needle aspirates taken from 141 feline patients diagnosed by cytology with lymphoma. Expression levels of the cell surface molecules—CD4, CD5, CD8, CD14, CD18, and CD21—were measured and used together with the forward and side scatter values for immunophenotyping and clustering. The latter was done by employing the machine learning algorithm (KMeans) with k ranging from two to six.
Results
Unsupervised clustering of the FC data enabled risk stratification of the patients with respect to ChlP or CHOP therapies. The overall median PFS with respect to the two regimens were 492 and 120 days with the hazard ratios of 5.1 and 2.7 between the high and low risk subgroups, respectively.
Conclusion
We developed the predictive technology that classifies each feline lymphoma patient as low or high risk with respect to the disease progression when treated by ChlP or CHOP therapy.
Clinical significance of the results
The proposed risk-stratification can contribute to personalized treatment selection regarding ChlP and CHOP for feline lymphoma. Retrospective analysis suggests that a superior PFS may be achieved for each patient by selecting a therapy associated with low risk of disease progression.