A rapid manual technique does not have adequate sensitivity and specificity to reliably discriminate between neutropenic and non‐neutropenic dogs prior to administration of chemotherapy

Abstract

ObjectivesTo develop and determine the accuracy of a rapid manual technique for the detection of pre‐treatment neutropenia (<1.50 × 109/L) in dogs receiving chemotherapy.Materials and MethodsTwenty canine blood smears with known neutrophil counts between 1.00 × 109/L and 3.00 × 109/L were reviewed by two internal medicine clinicians and linear regressions performed to determine a cut‐off value for a manual neutrophil count equating to >1.50 × 109/L. Consecutive blood samples from dogs undergoing chemotherapy were processed through an automated haematology analyser (VetScan HM5, Abaxis), and prospective blinded manual review by the same two observers assessed whether the manual technique could accurately detect dogs with neutropenia.ResultsLinear regression analysis found a cut‐off of >26 neutrophils per 10 low power fields at the monolayer to be equivalent to a neutrophil count of >1.5 × 109/L. A total of 183 blood samples from 43 dogs were reviewed. Automated techniques detected neutropenia in 16 of 183 (9%) blood samples. Using the manual cut‐off technique, 13 of 16 (81%) and 11 of 16 (69%) of neutropenic samples were correctly identified by observer 1 and observer 2, respectively. Twenty‐three of 167 non‐neutropenic dogs (14%) were incorrectly classified as neutropenic by observer 1, and 27 (16%) by observer 2. Inter‐observer agreement was 92%. Sensitivity was 81% (95% confidence interval 54% to 96%) for observer 1 and 69% (95% confidence interval 41% to 89%) for observer 2. Specificity was 86% (95% confidence interval 80% to 91%) for observer 1 and 84% (95% confidence interval 77% to 89%) for observer 2.Clinical SignificanceManual estimation resulted in up to five of 16 (31%) neutropenic samples being incorrectly classified. A full automated differential cell count remains preferable.