The CELL-DYN 3000 (Unipath Corp., Mountain View, CA) differential was evaluated in a tertiary care hospital using samples with a broad range of distributional and morphologic abnormalities. Particular attention was directed to the performance of the instrument-generated suspect flags that occur as an aid to identify samples with abnormal leukocytes, as well as the estimates of abnormal cells that are made by the instrument. The CELL-DYN 3000 showed excellent quantitative results for the white blood cell differential compared with a 400-cell manual differential, in which morphologic abnormalities were absent or occurred in low numbers (< or = 5%). Specificity of the BLAST, VARIANT LYMPH, NRBC, WBC, or DIFF suspect flags (with the requirement that the blast estimate and variant lymphocyte estimate by the instrument be > or = 1%) was 82.6%. Sensitivity of these flags to detection of more than 5% "abnormal" leukocytes (blasts, malignant lymphoid cells, grossly dysplastic neutrophils, nucleated red blood cells, or reactive lymphocytes) or significant platelet clumping was 81.6%. The primary deficiency was the inability of the CELL-DYN 3000 to flag samples with small numbers (< or = 5%) of nucleated erythrocytes, lymphoid blasts, or hairy cells, or more than 5% reactive lymphocytes. Specificity of the IG flag (with immature granulocyte estimate > or = 3%) for immature granulocytes (metamyelocytes, myelocytes, or promyelocytes) was 94.9%. Sensitivity of the immature granulocyte flag varied from 41.7% for identifying IG > or = 1% to 100% for the three samples with immature granulocytes > or = 3%. Calculation of sensitivity and specificity to varying percentages of bands showed poor flagging performance, with many false-positive and false-negative results at all levels.
View details for Web of Science ID A1992KB87700012
View details for PubMedID 1462958