Comparison of Serological and Molecular Assays for Bartonella Species in Dogs with Hemangiosarcoma.

Currently, a gold standard diagnostic test for Bartonella infection in dogs is lacking. This represents a critical limitation for the development and evaluation of new diagnostic tests, as well as for the diagnosis of, and research on, bartonellosis in dogs. This retrospective observational study aims to compare the results of commonly performed and newly-reported Bartonella spp. diagnostic tests in banked clinical specimens from 90 dogs with hemangiosarcoma (HSA) using composite reference standard (CRS) and random effects latent class analysis (RE-LCA) techniques. Samples from each dog were tested using six serological or molecular diagnostic assays, including indirect fluorescent antibody (IFA) and Western blot (WB) for the detection of antibodies in serum, and qPCR and droplet digital PCR (ddPCR) in blood and fresh frozen tissue biopsy samples (mainly splenic HSA tumors and histopathologically normal spleen or skin/adipose tissue). Bartonella infection prevalence was estimated to be 78% based on the CRS (parallel testing with all six assays), and 64% based on the RE-LCA model. The assay with the highest diagnostic accuracy was qPCR performed on fresh frozen tissue biopsy samples (sensitivity: 94% by RE-LCA and 80% by CRS; specificity: 100%). When comparing newly-reported to traditional Bartonella diagnostic assays, ddPCR was more sensitive for the detection of Bartonella DNA than qPCR when testing blood samples (36% vs. 0%, p < 0.0001). Dogs that were positive on serological assays alone with negative molecular assays were highly unlikely (<3%) to be classified as infected by the RE-LCA model. These data indicate that Bartonella spp. DNA can be PCR amplified from fresh frozen tissues from a majority of dogs with HSA using both qPCR and ddPCR, supporting the use of these methods for future controlled studies comparing the prevalence of Bartonella spp. DNA in the tissue of dogs with HSA to that of unaffected controls.