Background: In addition to differential cell counts, modern haematology analysers generate suspect flags if abnormal cells are detected. Reports on validation of suspect flags are scarce. We have routine experience with the Abbott Cell-Dyn 4000 analyser for over 5 years and have previously demonstrated the utility of the blast flag. Here we report a similar study on the performance of the analyser's Variant Lymphocyte (VL) flag.
Aim of the study: Evaluation of the diagnostic performance of the Cell-Dyn 4000 VL flag, as compared with lymphocyte morphology in blood smears. In addition, we investigated the usefulness of the numerical VL flag confidence index as provided by the analyser.
Materials and methods: All samples generating a VL flag were reviewed over a 5-month period. We also reviewed smears from patients with known lymphoid disorders, even if the analyser did not flag the sample. Two experienced investigators assessed lymphocyte morphology independently.
Results: In total, 187 samples were included in the study, of which 183 had a VL flag and four had not. Of the 183 flagged samples, 83 appeared to have abnormal lymphocyte morphology and 100, normal lymphocyte morphology. The sensitivity of the VL flag for detecting abnormal lymphocytes was 0.95 and the positive predictive value was 0.44. Using ROC analysis of the VL flag confidence index, the area under the ROC curve was 0.58 (95% confidence interval 0.50-0.65).
Conclusions: The Cell-Dyn VL flag has reasonable sensitivity but a high false-positive rate. In addition, its performance is insufficient for detecting clinically relevant abnormal lymphocytes. As the VL flag appeared to rely mainly on numerical criteria, it has no added value over numerical criteria defined by the laboratory.