Analytical variability of estimated platelet counts on canine blood smears.

Abstract

The analytical variability of estimated platelet counts in dogs has not been reported. The purpose of this study was to assess the magnitude of analytical imprecision of platelet estimates and the possible impact of this imprecision on clinical decisions. Three independent observers counted the number of platelets in 3 different areas (LE=lateral edge; CM=central monolayer; FE=feathered edge) of 30 canine blood smears with different instrumental platelet counts. The coefficient of variation (CV) for each observer was calculated in different areas of each smear (intra-observer variability), among different regions of each smear (inter-area variability), and among different observers in each area (inter-observer variability). The influence of these variabilities on the classification of platelet estimates as adequate, increased, or decreased was also assessed. The CVs recorded in the different areas by each observer ranged from 8% to 88% and were negatively correlated (P<.001, r=-.65) with the mean number of platelets per field. The mean platelet number was significantly lower in the FE and significantly higher in the CM compared with the LE, but the magnitude of this difference varied with the operators. The concordance among operators regarding platelet estimates was fair (k=0.36) to substantial (k=0.71) depending on the area. The overall inter-area concordance was moderate (k=0.59). Platelet estimates suffer from high variability that could lead to patient misclassification. Therefore, guidelines to standardize the platelet estimate are needed.