Mary Frye1, Crystal Soto1, JessicaWest1, George Courcoubetis1, Ilona Holcomb1, and Sungwon Lim1
1ImpriMed,Inc., 3980 Fabian Way, Palo Alto, CA 94303, USA
Relapsed canine B-cell lymphoma has no standard of care, leaving clinicians without prospective guidance among reinduction and salvage protocols. A multimodal machine learning model integrating flow cytometry and ex vivo drug-sensitivity data predicts response to ten chemotherapeutics; a prior study reported better outcomes under concordant treatment,¹ but cost implications are unquantified.
In the published 60-dog cohort, two equal concordance groups (high vs. low adherence) were adopted from the prior analysis. Per-patient cost combined drug costs and the machine learning prediction report, drawn from the price list of one confidential metropolitan referral hospital; visits, tests, and exams were excluded. Doses were standardized to a 30kg dog to normalize body-weight variation. Cost per day was the total cost divided by progression-free survival days, summarized per concordance group by mean and median.
Mean cost per day was $117.09 (high concordance) versus $188.53 (low concordance), a $71.44. per-day (40.2%) difference favoring higher concordance. High-concordance dogs also had longer progression-free survival (median 49 vs. 21 days) versus low concordance and received fewer drugs (190 vs. 336)despite longer follow-up.
Concordance with model predictions was associated with lower drug cost per day and substantially longer progression-free survival, supporting inclusion of the prediction report in treatment planning. Because the per-day metric is sensitive to survival and total cost rose with longer treatment, economic value should be interpreted alongside survival.
1.Callegari AJ, Tsang J, Park S, et al. Multimodal machine learning modelsidentify chemotherapy drugs with prospective clinical efficacy in dogs withrelapsed B-cell lymphoma. Front Oncol. 2024;14:1304144.doi:10.3389/fonc.2024.1304144.
