The white blood cell differential (WBC-diff) technologies

Transcription

1 The New Hematology Analyzer Sysmex Performance Evaluation of a Novel White Blood Cell Differential Technology Katharina Ruzicka, MD; Mario Veitl, MD; Renate Thalhammer-Scherrer, MD; Ilse Schwarzinger, MD Context. The new hematology analyzer Sysmex XE (TOA Medical Electronics, Kobe, Japan) has a novel, combined, white blood cell differential technology and a special reagent system to enumerate nucleated red blood cells. Design. Performance evaluation of both technologies of the Sysmex according to the H20-A protocol of the National Committee for Clinical and Laboratory Standards and comparison of the results with those for the hematology analyzer Sysmex (TOA Medical Electronics). Specimens. Five hundred forty-four blood samples randomly chosen from various inpatient and outpatient departments of the Vienna University hospital. Results. Five-part white blood cell differential counts on the revealed excellent correlation with the manual reference method for neutrophils, lymphocytes, and eosinophils (r.925,.922, and.877, respectively) and good correlation for monocytes and basophils (r.756 and.763, respectively). The efficiency rates of flagging for the presence of 1% abnormal white blood cells were 83% () and 66% (). The correlation of automated and microscopic nucleated red blood cell counts was excellent (r.97). Conclusions. From the present evaluation and our former experience with other types of Sysmex analyzers, we conclude that the new white blood cell differential technology of the represents a further development toward more efficient flagging of abnormal white blood cells. (Arch Pathol Lab Med. 2001;125: ) The white blood cell differential (WBC-diff) technologies of hematology analyzers can be roughly divided into electric and optical methods. 1 In the electric impedance method, cells are classified on the basis of a combination of cell size data from direct-current resistance information and intracellular data from alternating-current capacitance information. The optical method discriminates between cells on the basis of forward- and side-scattered light. To optimize the WBC-diff capacities, these technical methods may be combined with chemical alterations of either blood cells and/or reagents. Flow-cytochemical differential instruments classify cells using a combination of the optical method and enzyme cytochemistry. 2 The Abbott Cell Dyn 4000 combines the optical method with a special reagent that allows for discrimination of nucleated red blood cells (NRBCs) from WBCs by staining NRBC nuclei with a specific fluorochrome dye. 3 The Sysmex hematology analyzer (SE-9000) combines the electric impedance method with special reagents that disrupt the mature WBCs but fix immature WBCs, which are then identified in a special immature cell (IMI) channel. 4 The new hematology analyzer Sysmex performs the WBC-diff count by combining 3 kinds of optical information (forward-scattered light, side-scattered light, and side fluorescence) with the preestablished imped- Accepted for publication September 25, From the Department of Laboratory Medicine, University of Vienna, Vienna, Austria. Reprints: Ilse Schwarzinger, MD, Department of Laboratory Medicine, University of Vienna, Währinger Gürtel 18-20, A-1090, Vienna, Austria ( ilse.schwarzinger@univie.ac.at). ance/imi method. 5 Furthermore, the has a special channel to enumerate NRBCs. The present evaluation focuses on the flagging efficiency of this new combined technology for abnormal WBCs and on the accuracy of the analyzer s NRBC enumeration. The results are presented in comparison with those from our current routine hematology analyzer, the Sysmex. MATERIALS AND METHODS Hematology Analyzers Sysmex. The Sysmex (TOA Medical Electronics, Kobe, Japan) has a throughput of 150 samples per hour and provides 32 parameters, including reticulocyte and NRBC counts. 5 Measurement of WBCs is performed by flow cytometry using a semiconductor laser to detect forward- and side-scattered light information. Red cell lysis is performed by a reagent that selectively suppresses the degranulation of basophils, resulting in their separation from other forms of WBCs (Figure 1, A). In the DIFF channel, WBCs are permeabilized to enable staining of their DNA and RNA with a fluorescence dye. Cells are then categorized according to their side-scattered light and fluorescence intensity characteristics. A 4-part WBC-diff is created from the WBC populations: lymphocytes, monocytes, eosinophils, and neutrophils plus basophils (Figure 1, B). In the IMI channel, a special reagent acts on the lipid pattern of the cell membrane to selectively protect immature WBC against disruption, whereas mature leukocytes are disrupted (Figure 1, C). 4 After this reaction, cells are categorized by direct-current resistance and alternating-current capacitance information. The presence of abnormal WBCs is indicated by the suspect messages: Blasts?, Immature Gran?, Left Shift?, Atypical Lympho?, and Abn Lympho/ L Blasts?. The suspect messages are generated by combining pattern abnormalities in the 4-DIFF and IMI scattergrams; the Arch Pathol Lab Med Vol 125, March 2001 Hematology Analyzer Sysmex Ruzicka et al 391

2 392 Arch Pathol Lab Med Vol 125, March 2001 Hematology Analyzer Sysmex Ruzicka et al

3 areas of abnormal WBC locations in the respective scattergrams are shown in Figure 2, A and B. The quantifies NRBCs by using a reagent that, after red cell lysis, simultaneously denucleates, shrinks, and slightly stains the nuclei of NRBCs. The reagent does not alter the shape of WBCs but stains their intracytoplasmic organelles and nuclei. The difference in size and staining intensity allows discrimination of NRBCs from WBCs and enumeration of NRBCs (Figure 3). The NRBCs are indicated as absolute numbers per 100 WBCs. The WBC count is automatically adapted according to the results of the NRBC channel, by subtracting the NRBC count from the WBC count measured in the WBC/BASO channel. The software version used in the present evaluation was version 12. Sysmex. On the Sysmex (TOA Medical Electronics), the traditional direct-current technique is used for cell sizing, whereas information on the nuclear size and density are gained by radiofrequency detection. Complex algorithms are used to determine the optimum discriminator placement for separation of each cell population. 6 The software version used for the present study was version 13. The creates the suspect WBC messages Blasts?, Left Shift?, Immature Gran?, and Atypical Lymph? The areas used to define cells of the Left Shift and Immature Gran categories are separated in the 4-DIFF and IMI scattergrams of the, whereas on analyzers of the NE-series, the Left Shift area is integrated in the Immature Gran area. Thus, the XE might create both flags concomitantly, whereas the suppresses the Left Shift flag in the presence of the Immature Gran flag. The WBC suspect flag Abn Lympho/L Blasts? is only created by the and assists in the detection of lymphoid blasts. Specimen Collection Whole blood was collected in Vacutainer K3-EDTA tubes (Becton Dickinson, Mountain View, Calif) and analyzed within 4 hours after collection. Samples for evaluation were randomly chosen from various inpatient and outpatient departments of the Vienna University hospital, and included specimens from patients with normal hematologic profiles, specimens from patients with reactive hematologic abnormalities, and specimens from patients with known hematologic disorders. Reference Method Blood films were prepared by the manual wedge technique and stained according to a modified Wright technique. Reference differential counts were performed in accordance with the National Committee for Clinical Laboratory Standards (NCCLS) H20-A protocol. 7 Briefly, 2 manual, 200-cell WBC-diff counts were performed from each film by 2 independent, qualified medical technologists. Films were considered positive for pathologic WBCs if they showed more than 10% band forms and/or 1% metamyelocytes (corresponding to Left Shift), or 1% myelocytes and/or promyelocytes (corresponding to Immature Gran), or 1% blast cells, or 5% atypical lymphocytes. Lymphocytes were classified as atypical if they exhibited either nucleoli or nuclear shape abnormalities, reactive morphology, or plasmacytoid morphology. The NRBCs were counted outside the percentage count of WBCs and were indicated as absolute numbers per 100 WBCs. Table 1. Regression Analyses for Comparison of White Blood Cell Differential Counts Cell Type Comparisons* r Slope Intercept Neutrophils Lymphocytes Monocytes Eosinophils Basophils * Comparisons were made with counts determined by the manual reference method. Automated WBC-Diff Counts The 5 automated WBC-diff parameters performed by the 2 analyzers were compared with the manual reference counts. Automated WBC-diff counts were considered positive if they contained any of the suspect WBC flags described previously. To account for the different strategies used by the analyzers to create a Left Shift flag, this flag was not analyzed separately, but was combined with the Immature Gran flag under the heading Myeloid Precursors. After the suspect flags were compared with the results of the microscopic reference method, they were classified as true positive (TP), true negative (TN), false positive (FP), and false negative (FN). Statistics Correlations of the 5 WBC-diff parameters were estimated by regression analyses. The predictive value of instrument flagging for the presence of abnormal WBC after clinical review was expressed by the following parameters 8 : Sensitivity (%) [TP/TP FN] 100 Specificity (%) [TN/TN FP] 100 Efficiency (%) Percentage of Subjects Correctly Classified [TP TN/TP FP FN TN ] 100 RESULTS Correlation of Automated 5-Part WBC-Diff Parameters Correlations of the 5 WBC-diff parameters provided by the automated analyzers with the manual reference counts are shown in Table 1. Only samples with complete, automated, 5-part WBC-diff counts were included for regression analyses. Correlation coefficients for neutrophils, lymphocytes, and eosinophils were slightly better and correlation coefficients for monocytes and basophils were clearly better for the. Flagging of Abnormal WBCs Four hundred eighty-six differential counts were compared to assess the sensitivity and specificity of the analyzers suspect WBC flags. The predictive values of instrument flagging for detection of 1% pathologic WBCs (ie, either blasts and/or myeloid precursor cells and/or atypical lymphocytes) are shown in Table 2. One hundred forty-nine samples exhibited 1% abnormal WBCs on microscopic examination. The flagging efficiencies were 83% for the and 66% for the. This difference Figure 1. Figure 2. Figure 3. A, White blood cell/basophil (WBC/BASO) scattergram. B, 4-DIFF scattergram. C, Immature cell (IMI) scattergram. A, Abnormal cell locations on a 4-DIFF scattergram. B, Abnormal cell locations on an immature cell (IMI) scattergram. Nucleated red blood cell scattergram. Arch Pathol Lab Med Vol 125, March 2001 Hematology Analyzer Sysmex Ruzicka et al 393

4 Table 2. Sensitivity, Specificity, and Efficiency of Instrument Flagging for Detection of 1% Abnormal WBCs* Analyzer All Samples (n 486) * WBC indicates white blood cell /L (n 72) WBC Count /L (n 275) /L (n 139) Table 3. Sensitivity, Specificity, and Efficiency of Instrument Flagging for Detection of 1% Blasts and Myeloid Precursor Cells and 5% Atypical Lymphocytes Analyzer Blasts Myeloid Precursors Immature Gran Atypical Lymph was mainly due to the higher flagging specificity of the. Both analyzers produced comparable FN results, but the produced almost threefold the number of FP results produced by the. Flagging efficiencies were best for samples with normal WBC counts throughout all categories of abnormal WBC flags. On both analyzers, the flagging specificity decreased with WBC counts of /L, an effect that was markedly less pronounced on the. Conversely, the flagging sensitivity was lowest among samples with WBC counts of less than /L, but again, this effect was less marked on the (Table 2). Blasts. Twenty-three of 486 (5%) blood films showed 1% blasts on microscopic examination (Table 3). The rates of FN results were 1% for the and 2% for the. Among the 5 samples that were missed by the, 2 were flagged by the Immature Gran flag, and 2 were flagged by the Atypical Lymph flag; the fifth sample did not produce a suspect WBC flag. All FN samples exhibited WBC counts below /L. The NE gave twice as many FP results as the (11% vs 5%, respectively). Among the 25 samples with an FP result determined by the, 13 contained 1% cells of the Immature Gran category, 8 contained singular ( 1%) myeloid precursor cells, 1 exhibited singular ( 1%) blasts, 1 contained 5% atypical lymphocytes, and 2 did not show any pathologic WBC on microscopic examination; 16 samples exhibited WBC counts that exceeded /L, 5 had normal WBC counts, and 4 were leukocytopenic. All FP samples were additionally flagged by the Immature Gran flag. Immature Gran/Left Shift. One hundred thirty of 486 (27%) samples showed 1% myeloid precursor cells on microscopic examination. Eighty-four blood films contained 1% cells of the Immature Gran category. The efficiency rates for detection of myeloid precursor cells were 84% for the and 62% for the (Table 3). The was highly sensitive in detecting cells of the Immature Gran category: the rate of FN results was only 1%, compared with 12% on the. Among the 7 samples with an FN result determined by the, 4 were leukocytopenic, and 3 exhibited normal WBC counts. The rates of FP results were 16% on the and 9% on the. Twelve of 75 FP results of the were truly negative, without any pathologic WBCs on microscopic examination; 37 samples contained singular ( 1%) cells of the Immature Gran category, with or without additional cells of the Left Shift category; 17 samples contained increased numbers of bands and/or metamyelocytes but no cells of the Immature Gran category; 4 samples exhibited blasts; and 5 samples contained atypical lymphocytes. Atypical Lymph. Twenty-two of 486 (5%) samples contained 5% atypical lymphocytes. The rates of FP and FN results were identical for both analyzers (Table 3). Abn Lympho/L Blasts?. The Abn Lympho/L Blasts? flag occurred in 31 of 486 (6%) samples. None of these samples contained blast cells. Thirteen samples contained atypical lymphocytes; 7 of those were also flagged by the Atypical Lymph flag. Enumeration of NRBCs NRBC counts were determined for 544 samples, including the 486 samples that were analyzed for flagging of abnormal WBC. One hundred six samples contained NRBCs on microscopic examination (range, /100 WBCs). The analyzer gave a positive NRBC result in 91 samples (77 TP and 14 FP). In 453 samples, the counted zero NRBC (424 TN and 29 FN). Among the 29 FN samples, 23 showed 1 NRBCs /100 WBCs, and 6 showed 2 NRBCs /100 WBCs on microscopic examination. The correlation between microscopic NRBC counts and NRBC counts provided by the was excellent (r.97) for 540 samples containing fewer than 100 NRBCs/100 WBCs (Figure 4). There was a trend toward lower NRBC counts on the for samples containing more than 15 NRBCs /100 WBCs. Four samples exhibited 428, 340, 214, and 213 NRBCs/100 WBCs; the respective counts on the were 312, 258, 96, and 60. The analyzer flagged all NRBC counts above 100/ Arch Pathol Lab Med Vol 125, March 2001 Hematology Analyzer Sysmex Ruzicka et al

5 Figure 4. Correlation data of NRBC counts ( vs manual). The correlation coefficient (r) is.97, the slope of the linear regression is 0.62, and the y-intercept is WBCs, as well as the corresponding WBC and WBC-diff counts. NRBC Counts and Abnormal WBC Flags Among the 91 samples with positive automated NRBC counts, 83 exhibited at least 1 additional abnormal WBC flag. Among the remaining 8 samples without additional abnormal WBC flags, 5 were TP and 3 were FP. Four of the TP samples were from severely anemic patients; the fifth sample was obtained from a patient with known hairy cell leukemia. COMMENT The Sysmex is a new hematology analyzer that provides 32 parameters, including reticulocyte and NRBC counts. 5 Data on precision, reproducibility, linearity, carry over, and stability have been provided The present study focused on the performance of the novel WBC-diff technology that combines an optical laser/fluorescence method with the impedance/imi method, and on the accuracy of NRBC enumeration. The results were compared with those of standardized, microscopic 400-cell counts and with results obtained from the established hematology analyzer. Correlation coefficients for neutrophils, lymphocytes, monocytes, eosinophils, and basophils were consistently better for the than for the. The difference was marked for monocyte counts, which have been reported to be less accurate on the, 12,13 and for basophil counts, which have generally been reported to poorly correlate with manual reference methods. 12,14,15 The sensitivity of flagging 1% pathologic WBC was 76% on the, which is similar to the flagging sensitivities that we have previously observed for Sysmex analyzers of the NE, SF, and SE series. 13,16 The flagging specificity was markedly better for the when compared with the in the present study and also best when compared with previously established specificity rates for the SE-9000 and the SF ,16 Thus, among all hitherto evaluated Sysmex analyzers, we obtained the best flagging efficiency with the. We have previously shown that the flagging efficiencies of hematology analyzers are associated with total WBC counts. 12,16 The flagging sensitivities were rather poor in leukocytopenic samples and highest in samples with more than WBC/L. Conversely, the flagging specificities markedly decreased with rising WBC counts. This trend was also recognizable on the. However, for samples containing fewer than WBC/L, the sensitivity rate of the was markedly better than the rates observed with the other analyzers. 12,16 Thus, the XE seems more suitable for screening for pathologic WBC in leukocytopenic samples. Furthermore, when compared with all other tested analyzers, the exhibited a better flagging specificity in samples with leukocytosis. 12,16 The flagging sensitivity of the for blasts was excellent for samples with normal and elevated WBC counts. However, the presence of 1% blasts was missed in 5 samples with WBC counts of less than /L. This observation further stresses the importance of microscopic clarification of samples with unexplained leukocytopenia. 16 Most of the FP samples exhibited elevated WBC counts, with myeloid precursor cells on microscopic examination, most probably representing reactive conditions. Combined evaluation of the Blast flag with the Abn Lympho/L Blasts flag on the did not increase the sensitivity of blast cell flagging but resulted in an increase of FP results. The sensitivity of flagging of Immature Gran was significantly higher for the than for the in the present study and also significantly higher than sensitivities we obtained on other hematology analyzers. 12,16 The detection of Immature Gran was least sensitive in samples with low WBC counts. The better sensitivity of the was slightly outweighed by a decreased specificity at a positivity threshold of 1% pathologic cells. However, among the 75 FP samples, only 12 were truly negative, without any pathologic WBC on microscopic examination, whereas the majority of samples contained either few numbers of Immature Gran or pathologic WBCs of other categories. Thus, the Immature Gran flag on the seems to be highly sensitive for the presence of low numbers of myeloid precursor cells. The sensitivity of flagging for the presence of 5% atypical lymphocytes was poor on both analyzers. This observation corresponds to our previous experiences with other hematology analyzers and might best be explained by the rather broad definition of an atypical lymphocyte. 12,16 To screen for abnormalities of the peripheral blood lymphoid system, it might thus be more effective to check the absolute lymphocyte numbers. The new Abn Lympho/L Blasts flag detected few additional samples with atypical lymphocytes that had not been flagged by the Atypical Lymph flag. The is the second hematology analyzer, after the Abbott Cell Dyn 4000 (Abbott Diagnostics Division, Mountain View, Calif), that provides numerical NRBC results. 3 In the present evaluation, the correlation of microscopic and automated NRBC counts was excellent for samples with fewer than 100 NRBCs/100 WBCs. The observed trend toward lower NRBC counts in samples with Arch Pathol Lab Med Vol 125, March 2001 Hematology Analyzer Sysmex Ruzicka et al 395

6 more than 15 NRBCs/100 WBCs turned into a marked difference for the 4 samples with NRBC counts of more than 200/100 WBCs. A similar trend has been reported for NRBC counts on the Cell Dyn Thus, the value of automated NRBC enumeration in terms of concomitant correction of the WBC count is limited, since the necessity for correction of WBC counts increases with increasing NRBC counts. The clinical importance of automated NRBC enumeration has to be established. Beyond the neonatal period, the presence of NRBCs in the peripheral blood is highly indicative of a pathologic condition. The automated NRBC method offers a tool to screen for such pathologic samples. However, the additional request for an automated NRBC count on the would double the cost of each complete blood count. Thus, the automated enumeration of NRBCs cannot reasonably be considered for screening as long as the costs overshadow the possible benefits of this method. Moreover, in the present study, 83 of the 91 samples with positive automated NRBC counts exhibited at least 1 additional abnormal WBC flag and would therefore have been further investigated anyway. The remaining 5 TP samples without an additional WBC flag exhibited numerical complete blood count abnormalities that would also have prompted further diagnostic workup. Automated NRBC enumeration might be beneficial in neonatology, to facilitate the estimation of WBC counts on the one hand and to provide prognostic information on the other hand Monitoring the course of hemolytic disorders and the treatment of renal anemia with erythropoietin and iron in patients undergoing hemodialysis might be additional fields of application for automated NRBC counts In summary, our data suggest that when compared with other analyzers of the Sysmex series that we have tested, the new combined WBC-diff technology of the is superior in terms of predicting the presence of abnormal WBCs. We have already noted that, relative to the NE series, the SE and SF series have greater sensitivity for the detection of abnormal WBCs 12,16 ; the represents further improvement in terms of specificity. The performance of automated NRBC counts on the compares very well with that reported for the Cell Dyn The authors wish to thank our laboratory staff for excellent technical assistance. References 1. Tatsumi N, Tsuda I, Furota A, Takubo T, Hayashi M, Matsumoto H. Principle of blood cell counter development of electric impedance method. Sysmex J Int. 1999;9: Ornstein L, Ansley HR. Spectral matching of classical cytochemistry to automated cytology. J Histochem Cytochem. 1974;22: Kim YR, Yee M, Metha S, Chupp V, Kendall R, Scott CS. Simultaneous differentiation and quantitation of erythroblasts and white blood cells on a high throughput clinical haematology analyser. Clin Lab Haematol. 1998;20: Buttarello M, Bulian P, Temporin V, Rizotti P. Sysmex SE-9000 hematology analyzer: performance evaluation on leukocyte differential counts using an NCCLS H20-A protocol. Am J Clin Pathol. 1997;108: Inoue H. Overview of automated hematology analyzer TM. Sysmex J Int. 1999;9: van Wersch JWJ, Bank C. A new development in haematological cell counting: the Sysmex, automation for cell count and physical five-part leukocyte differentiation. J Clin Chem Clin Biochem. 1990;28: National Committee for Clinical Laboratory Standards. Reference Leukocyte Differential Count (Proportional) and Evaluation of Instrumental Methods. Villanova, Pa: National Committee for Clinical Laboratory Standards; Approved standard H20-A. 8. Swaim WR. Laboratory and clinical evaluation of white blood cell differential counts. Comparison of the Coulter VCS, Technicon H-1 and 800-cell manual method. Am J Clin Pathol. 1991;95: Briggs C, Harrison P, Grant D, Staves J, Chavada N, Machin SJ. Performance evaluation of the Sysmex TM, automated haematology analyser. Sysmex J Int. 1999;9: Gould N, Connell B, Dyer K, Richmond T. Performance evaluation of the Sysmex TM, automated hematology analyzer. Sysmex J Int. 1999;9: Tsuruda K, Tsuji T, Usui T, et al. Evaluation and clinical usefulness of the automated hematology analyzer, Sysmex TM. Sysmex J Int. 1999;9: Korninger L, Mustafa G, Schwarzinger I. The haematology analyser SF- 3000: performance of the automated white blood cell differential count in comparison to the haematology analyser NE Clin Lab Haematol. 1998;20: Goossens W, van Hove L, Verwilghen RL. Monocyte counting: discrepancies in results obtained with different automated instruments. J Clin Pathol. 1991;44: Vives-Corrons JL, Besson I, Jou JM, Gutierrez G. Evaluation of the Abbott Cell-DYN 3500 hematology analyzer in a university hospital. Am J Clin Pathol. 1996;105: Picard F, Gicquel C, Marnet L, Guesnu M, Levy JP. Preliminary Evaluation of the new hematology analyzer COULTER(r) GEN.S TM in a university hospital. Clin Chem Lab Med. 1999;37: Thalhammer-Scherrer R, Knöbl P, Korninger L, Schwarzinger I. Automated five-part white blood cell differential counts. Arch Pathol Lab Med. 1997;121: Yeruchimovich M, Mimouni FB, Green DW, Dollberg S. Nucleated red blood cells in healthy infants of women with gestational diabetes. Obstet Gynecol. 2000;95: Buonocore G, Perrone S, Gioia D, Gatti MG, Massafra C, Agosta R. Nucleated red blood cell count at birth as an index of perinatal brain damage. Am J Obstet Gynecol. 1999;181: Hanlon-Lundberg KM, Kirby RS. Nucleated red blood cells as a marker of acidemia in term neonates. Am J Obstet Gynecol. 1999;181: Korst LM, Phelan JP, Ahn MO, Martin GI. Nucleated red blood cells: an update on the marker for fetal asphyxia. Am J Obstet Gynecol. 1996;175(pt 1): Beguin Y, Loo M, R Zik S, et al. Quantitative assessment of erythropoiesis in haemodialysis patients demonstrates gradual expansion of erythroblasts during constant treatment with recombinant human erythropoietin. Br J Haematol. 1995; 89: Bhandari S, Norfolk D, Brownjohn A, Turney J. Evaluation of RBC ferritin and reticulocyte measurements in monitoring response to intravenous iron therapy. Am J Kidney Dis. 1997;30: Besarab A, Amin M, Ahsan M, et al. Optimization of epoetin therapy with intravenous iron therapy in hemodialysis patients. J Am Soc Nephrol. 2000;11: Arch Pathol Lab Med Vol 125, March 2001 Hematology Analyzer Sysmex Ruzicka et al