Abstract: Background: Bone marrow (BM) lesions in dogs may be spatially heterogeneous, yet single-site BM aspiration and/or core biopsy remain common, potentially increasing diagnostic misclassification. HYPOTHESIS/ Aims: We hypothesized that single-site BM sampling increases the probability of site-related diagnostic misclassification relative to a within-dog “truth-set” and aimed to estimate the probability of a truth-consistent interpretation when sampling 1–4 sites. ANIMALS: Sixteen client-owned dogs undergoing BM evaluation for hematologic disease. Methods: Dogs had their BM aspirates and trephine cores collected in a randomized sequence from four sites (bilateral proximal humeri and iliac crests) using a powered device. Two masked pathologists independently reviewed aspirate smears, and two other masked pathologists independently evaluated core biopsy sections. Diagnostic “truth-sets” (i.e., the within-dog modal interpretation[s] across sites) were derived, and the probability of a truth-consistent interpretation when sampling 1–4 sites was estimated under three agreement rules. Results: For cytology, single-site sampling was truth-consistent across the three rules in 50.0%–81.7% of cases; sampling two sites increased accuracy by 17.2%–23.3% (p < 0.002), with only marginal gains beyond two sites; inter-pathologist agreement was moderate (κ = 0.574), and no single site was superior. For core biopsy, moving from one to two sites increased the probability of truth-consistent interpretation from 76.6% to 94.8% and from 28.1% to 47.9% across increasingly stringent agreement rules; 13.3% of reads were nondiagnostic, and inter-pathologist agreement was modest (κ = 0.30; 42%). Conclusions: Reliance on a single BM site carries meaningful misclassification risk; sampling two sites provides the largest clinically relevant reduction while maintaining practicality.
.jpg "Arnon Gal, DVM PhD DACVIM (SAIM) DACVP (AP) photo")