Comparison of four automated hematology analyzers

Comparison of four automated hematology analyzers

Koenn, Mary E

OBJECTIVE: To compare four automated hematology analyzers for efficiency and sensitivity.

DESIGN: Four automated hematology analyzers were compared in a side by side study: Bayer ADVIA 120 (Bayer Diagnostic Division, Tarrytown, NY), Beckman Coulter GEN S (Beckman Coulter, Brea, CA), Abbott CELL DYN 3500 and CELL DYN 4000 (Abbott Diagnostics, Santa Clara, CA). 164 specimens were analyzed for cell counts, indices, and the automated WBC differential (DLC). Tallies were kept of all interventions, defined as any parameter necessitating examination of a stained blood smear by a clinical laboratory scientist. A 400-cell manual differential was performed on each specimen and used as the reference to prepare truth tables for each type of WBC.

PATIENTS: Specimens comprised regular runs from this tertiary care teaching hospital. These included inpatients, outpatients, and oncology patients, including bone marrow transplant patients.

MAIN OUTCOME MEASURES: Results from the truth tables were used for calculating sensitivity and efficiency for each analyzer. Each DLC parameter was analyzed for variance using the one-way ANOVA test.

RESULTS: No intervention was required for 103 of 164 specimens for the CELL DYN 3500; the ADVIA gave 70 reportable DLCs without intervention, the GEN S provided 91 and the CELL DYN 4000 resulted in 117 of 164 DLCs without intervention. Agreement or efficiency was 65% for the CELL DYN 3500,41% for the ADVIA, 58% for the GEN S, and 79% for the CELL DYN 4000. Sensitivity was 67% for the CELL DYN 3500, 86% for the ADVIA, 76% for the GEN S, and 71% for the CELL DYN 4000. Probability of significant variation was as follows for each parameter: % neutrophil 0.8747, % lymphocyte 0.8830, % monocyte 0.0296, % eosinophil 0.7903, and % basophil

CONCLUSION: The analyzers tested were acceptable for routine laboratory work. Selection would depend on individual need with respect to sensitivity and efficiency. The clinical significance of disagreement between the DLC and the manual differential remains to be determined.

ABBREVIATIONS USED: DLC = differential leukocyte count; FN = false negative; FP = false positive; HCT = hematocrit; HGB = hemoglobin; MCH = mean corpuscular hemoglobin; MCHC = mean corpuscular hemoglobin concentration; MCV = mean cor

puscular volume; MPV = mean platelet volume; PLT = platelet count; RBC = red blood cell; RDW = red cell distribution width; TN = true negative; TP = true positive; WBC = white blood cell.

INDEX TERMS: automated hematology analyzer; automated leukocyte differential; hematology analyzer efficiency; hematology analyzer sensitivity; hematology analyzer specificity.

The effectiveness and value of clinical hematology automation in obtaining a complete blood count (CBC) and differential leukocyte count (DLC) are documented by a reduction in actual personnel time spent in reporting CBC and DLC results. For the automation to be effective, hematology analyzer results should be precise and accurate, require little intervention for verifying, and be readily released as derived. These features have been ordinarily available for some time for a CBC, but less obtainable and accessible for a DLC. Evaluation and adoption of a new automated DLC method centers around documentation of improved precision, ensuring less operator intervention and investigation of abnormal results.

For more than 25 years, multi-channeled automated hematology analyzers have provided a reliable and accurate CBC in large numbers on a daily basis and most often with immediate results.1-3 This historically has been accomplished using direct current impedance to enumerate cellular elements: red blood cell count (RBC), white blood cell count (WBC), and platelet count (PLT). Automated DLC implementation is a more complicated process and has been through several different technologies. Flow cytometry with its capabilities for swift enumeration, high throughput, and laser light scatter differentiation of WBCs has evolved as the primary technology in an automated DLC.3 On some instruments flow cytometry is also incorporated into the CBC measurement. There it provides a cross-check on the RBC and PLT count.4 The adoption process for a hematology analyzer primarily focuses on the automated DLC, the numbers of stained smears and manual differentials eliminated, and the reviews of flagged results required. The improved precision and speed of the automated DLC over the 100-cell manual differential is well documented.3,5-10 The imprecision and inaccuracy of the manual differential results from limited cells viewed and maldistribution of cells in wedge preparations.3 The manual differential is the accepted reference method for evaluating an automated DLC but poses problems in evaluating the automated monocyte and basophil results. The correlation between the automated and manual differential for neutrophils, lymphocytes, and eosinophils is described as good, but because of low numbers in manual counts, there is poor agreement between the manual and automated basophil and monocyte counts.’

Another concern in automation evaluation is the reliability of flags generated when performing CBCs and DLCs on pathological samples. There is less publication in this area and inconsistency and inefficiency among analyzers when it has been studied.’11 The review of numerous flags with pathological samples slows down the automated reporting and results in time consuming investigations. The sensitivity of flags readily denotes such samples as abnormal but the specificity often needs to be validated.

The study presented evaluates four automated hematology analyzers in a side by side investigation of their CBC and DLC. The number of interventions required to operate each analyzer was of particular interest along with the sensitivity and specificity of the CBC and DLC.

MATERIALS AND METHODS

Hematology analyzers

This study evaluated four automated hematology analyzers: Bayer ADVIA 120(Bayer Diagnostic Division, Tarrytown NY), Beckman Coulter GEN S(Beckman Coulter, Brea CA), and the Abbott CELL DYN 3500 and CELL DYN 4000(Abbott Diagnostics, Santa Clara CA).

Automated hematology analyzers are very complex and incorporate multiple technologies within one instrument. Thus a variety of measurement methods are utilized by these four instruments and often more than one method is used in reporting a single parameter. Light scatter readings from a flow cytometer have enhanced hematology analyzers providing greater accuracy and speed.3 Each of these four instruments incorporates a flow cytometer in its routine operation but the flow cytometric technology is used in a variety of ways to obtain the CBC and DLC.

Impedance is still the primary measurement methodology utilized by the Beckman Coulter GEN S system. The CBC is derived solely with impedance and these readings are further clarified with a WBC scattergram and RBC and platelet distribution curves. Impedance is also included in the DLC determination; WBCs are differentiated into various types by volume, conductivity, and scatter (VCS). DLC VCS technology uses impedance to measure WBC volume, flow cytometry for laser light scatter measurements, and conductivity (high frequency current or radio waves) to determine WBC opacity.3,6,9,10,12,-14 Intellikinetics technology automatically adjusts reagent reaction temperature, reagent delivery volume, and exposure time to optimize the DLC determination. A contour gating algorithm (AccuGate software) analyzes the raw data and delineates WBC populations, especially differentiating the small leukocytes.15

The ADVIA 120 technology does not include any impedance measurements; all results are derived from flow cytometric technology. The light scatter enumeration of RBCs and PLTs is improved with a two-dimensional platelet analysis. Both the volume and refractive index of PLTs are determined from the light scatter readings; the resulting scattergrams assist in differentiating RBCs, PLTs, and other interfering substances in the platelet size range.16 The ADVIA DLC uses light scatter readings of cell size and peroxidase activity after exposure to a peroxidase/chromagen reagent to enumerate the neutrophils, eosinophils, monocytes, and lymphocytes. The basophils are counted in a different flow cytometric operation; a lysing reagent is added, stripping the other WBCs of their cytoplasm, and leaving the basophils intact and countable.3,17 Another cluster of abnormal WBCs, designated as large unstained cells (LUC), sometimes results. These are peroxidase-negative cells that are either large lymphocytes, virocytes, or stem cells.17,18

The Abbott CELL DYN 3500 also enumerates the RBCs and PLTs with impedance measurement. The WBC count, however, is obtained with two measurements, an impedance channel for WBC number and volume designated as WIC, and an optical channel, a flow cytometry measurement of WBC number, the WOC.14,19,20 For the reported WBC count, the WIC/WOC data are compared and ordinarily the WOC count is used unless a significant difference occurs between the WOC and the WIC. A flag alerts the operator for an investigation of the difference. To improve the WOC in the presence of resistant RBCs, found in newborns, an optional RBC extended lyse mode is available.7,19,20 The DLC flow cytometry measurement is obtained using CELL DYN Multi-Angle Polarized Scatter Separation (MAPSS) technology, differentiating WBCs based on light scatter readings at three angles and an additional reading with polarized light. These readings are plotted on scattergrams as a function of cell size, cell complexity and surface morphology, nuclear shape, and cytoplasmic granularity, and the per cent neutrophils, lymphocytes, monocytes, eosinophils, and basophils obtained.5’6,9,14,19,21

To improve the RBC/PLT count, the Abbott CELL DYN 4000 has added an optical measurement to the CBC determination, similar to the two-channel WBC count on the CELL DYN 3500. Comparison of the optical and impedance RBCs and PLTs improves data for these two parameters by providing checks on the measurements and detecting the presence of sample interferences.13 This instrument also enumerates the nucleated red blood cell (NRBC) with the addition of a fluorescence measurement in the RBC light scatter measurement. The fluorescent dye strips NRBC cytoplasm away and stains the nuclear DNA.4,13,21,22 Additionally, any non-viable WBCs and fragile WBCs present have cytoplasm removed, DNA stained, and are enumerated.4,13,21-23 Flow cytometric MAPSS technology is again used to determine the DLC, but with refinement of angle that determines cell surface morphology and complexity.4

The study assessed the four analyzers by performing a side by side comparison using 164 specimens from the usual population of patients of this tertiary care teaching institution. Specimens comprised regular laboratory runs from hospital inpatients and outpatients, including oncology patients.

The analyzers were installed and operated according to manufacturers’ guidelines. Each specimen was analyzed first on the CELL DYN 3500 because it was the analyzer in use in the laboratory. Specimens were then immediately run in random order on each of the three analyzers being evaluated for possible purchase. The testing was done by seven clinical laboratory scientists (CLSs). Each specimen was evaluated for the following parameters: WBC, RBC, hemoglobin (HGB), hematocrit (HCT), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), red cell distribution width (RDV), platelets (PLT), mean platelet volume (MPV), and the automated WBC differential.

A 400-cell manual differential was performed on each specimen following a modification of the protocol established by the National Committee for Clinical Laboratory Standards for leukocyte differential counting, Approved Standard H20-A.24 The manual differential was considered the reference method for statistical interpretation of the DLC results.

Those data presented in this paper focus on evaluation of the parameters verifiable by conventional microscopy. Each analyzer was assessed for frequency of intervention, defined as any result that would require the examination of a stained slide by a CLS before the acceptance of analyzer data. Results requiring intervention included flags for quantity not sufficient for differential (low WBC), no data reported, RBC morphology, variant lymphocytes, platelet flags, platelet counts less than 30,000/(mu)L, and suspect WBC populations (increased numbers of bands, immature granulocytes, and blasts).

DLC data was classified into four categories: 1) true positives (TP) in which instrument flags were verified by conventional microscopic examination; 2) false positives (FP) in which the instrument flag was not verified by microscopy; 3) true negatives (TN) in which there were no instrument flags and no abnormalities were noted by microscopy; and 4) false negatives (FN) in which microscopy revealed abnormalities not flagged by the instrument.

DLC data was further interpreted by calculation of the false positive ratio, the false negative ratio, sensitivity, and efficiency or agreement using the following formulae:

false positive ratio (%) = FP/(FP + TN) x 100

false negative ratio (%) = FN/(FN + TP) x 100

sensitivity (%) = TP/(TP + FN) x 100

agreement (efficiency) (%) = (TP + TN)/(TP + FN + FP +TN) x 100

A one-way ANOVA statistical analysis was used to evaluate the DLC vs manual differential. This analysis which compares the differences in the means of three or more groups of data (four instrument DLCs) is used when the groups differ with only one factor at a time (DLC cell classification).

RESULTS

Results for the four analyzers were examined for consistency and the analyzers were found to be comparable with respect to WBC, RBC, HGB, HCT, MCV, MCH, MCHC, RDW, PLT, and MPV with one exception. The ADVIA required interventions for the WBC in 19 instances.

Of 164 specimens, the CELL DYN 3500 produced 103 DLCs that were reportable without intervention. The ADVIA produced 70, the GEN S produced 91, and the CELL DYN 4000 resulted in 117. Slides for manual examination were necessary for 37% of the specimens when run on the CELL DYN 3500, 57% for the ADVIA, 45% for the GEN S, and 26% for the CELL DYN 4000. These results are summarized in Table 1 with a breakdown of the required interventions.

False positive ratios were 36% for the CELL DYN 3500, 64% for the ADVIA, 44% for the GEN S, and 24% for the CELL DYN 4000. False negative ratio for the CELL DYN 3500 was 33% while false negative ratios were 14% for the ADVIA, 24% for the GEN S, and 29% for the CELL DYN 4000. Sensitivities were as follows: CELL DYN 3500 67%; ADVIA 86%; GEN S 76%; and CELL DYN 4000 71%. Efficiencies or agreement, representing the percentage of items correctly classified, were as follows: 65% for the CELL DYN 3500; 41% for the ADVIA; 58% for the GEN S; and 79% for the CELL DYN 4000. These results are summarized in Table 2.

The results of the one-way ANOVA statistical analysis of the DLC vs manual differential are summarized in Table 3. Two of the DLC parameters displayed statistically significant difference (p

DISCUSSION

Number of interventions is an important criterion evaluated in the selection of an automated hematology analyzer. Less technologist time spent in microscopic slide examination results in increased efficiency of work flow and less cost. The fewest interventions were required by the CELL DYN 4000 (26%), followed by the CELL DYN 3500 (37%) and the GEN S (45%). The most interventions were required by the ADVIA with more than half (57%) of the specimens requiring slide examination and/or WBC count intervention. Performance of a complete manual differential was required least often by the CELL DYN 4000 (22 of 164), followed by the ADVIA (39 of 164), and the CELL DYN 3500 (41 of 164). The GEN S required the greatest number of complete manual differentials (53 of 164).The ADVIA and the CELL DYN 4000 were equal in failure to provide data (1 of 164) while the CELL DYN 3500 failed to provide data in 5 of 164 and the GEN S provided no data for 12 of 164 specimens. RBC morphology was flagged least by the GEN S (0 of 164). The CELL DYN 3500 required RBC morphology scans for 4 of 164 while the CELL DYN 4000 was comparable with 5 of 164. The ADVIA generated the most RBC flags with 15 of 164. There were no significant differences in platelet flags among the four analyzers. Based on total numbers of interventions, the performance of the CELL DYN 4000 would appear superior to the other analyzers when considering laboratory efficiency.

Selection of an automated hematology analyzer also requires evaluation of the clinical sensitivity and specificity. It is important not only that the analyzer provide reportable results with few interventions, but also that these results be reliable. It is necessary to weigh the increased efficiency of work load against the possibility of missing clinically significant parameters. Construction of truth tables assists in this evaluation of clinical sensibility. Numbers approaching 100% for sensitivity and agreement would be the ideal. It must be recognized that limitations posed by CLS subjectiveness and by the small numbers of cells examined in the traditional manual differential necessitate cautious interpretation of sensitivity and efficiency. A false negative does not necessarily indicate that a clinically significant problem was missed by an analyzer. Interpretation of a false positive is limited by the small cell population examined on a smear. More study is required to assess the clinical significance of disagreement between automated methods and microscopic examination of a stained smear.

The truth tables demonstrated that the ADVIA had the most false positives and the fewest false negatives. This resulted in a calculated sensitivity higher than that of the other analyzers (86% compared to 67% for the CELL DYN 3500, 71% for the CELL DYN 4000, and 76% for the GEN S). Efficiency as calculated demonstrated a higher value for the CELL DYN 4000 (79% compared to 65% for the CELL DYN 3500, 58% for the GEN S, and 41% for the ADVIA). Efficiency or agreement represents the percentage of specimens that are correctly classified. Individual laboratories will place different emphasis on the importance of each of these statistics depending upon patient population, staffing, and work load.

The one-way ANOVA statistical analysis used to evaluate the DLC vs manual differential resulted in two statistically significant probabilities, p

CONCLUSION

Selection of an automated hematology analyzer depends upon the needs of the individual laboratory. These four analyzers were generally comparable in performance. There were differences in sensitivity and efficiency. Whether these differences are clinically significant would require more investigation. The ADVIA was most sensitive at 86% followed by the GEN S (76%), the CELL DYN 4000 (71%), and the CELL DYN 3500 (67%). The CELL DYN 4000 was most efficient at 79% followed by the CELL DYN 3500 (65%), the GEN S (58%), and the ADVIA (41%). While the manual differential remains the standard against which automated DLC accuracy is measured, subjectiveness, and the small number of cells counted compared to the cells counted by an analyzer cast doubt on the appropriateness of this method of validation. Most importantly, research needs to concentrate on the clinical significance of the false positives and negatives generated by automated hematology analyzers.

ACKNOWLEDGEMENT

We wish to acknowledge Gerald Hobbs PhD, WVU Health Science Center Statistical Consultant, for his assistance in preparing the statistics for this paper.

REFERENCES

1. Springer W, Prohaska W, Neukammer J, and others. Evaluation of a new reagent for preserving fresh blood samples and its potential usefulness for internal quality controls of multichannel hematology analyzers. Am J Clin Pathol 1999;111:387-96.

2. Hogan J, Raik E. An internal laboratory evaluation of peripheral blood film abnormalities not detected by an automated blood cell counter. Pathology 1996;28:287.

3. Bentley SA, Johnson A, Bishop CA. A parallel evaluation of four automated hematology analyzers. Am J Clin Pathol 1993;100:626-31.

4. CELL-DYN 4000 System Operator’s Manual. Santa Clara CA: Abbott Diagnostics, 1997.

5. Vives-Corrons JL, Besson I, Jou JM, and others. Evaluation of the Abbott Cell-DYN 3500 hematology analyzer in university hospital. Am J Clin Pathol 1996;105:553-9.

6. Stroop DM, Triplett RC, Perrotta G, and others. Comparison of the Abbott Cell Dyn 3000 SL and the Coulter STKS hematology analyzers. Ann Clin Lab Sci 1994;24:250-8.

7. Sakamoto C, Yamane T, Ohta K, and others. Automated enumeration of cellular composition in bone marrow aspirate with the CELL DYN 4000 automated hematology analyzer. Acta Haematol 1999; 101:130-4.

8. Buttarello M, Bulian P Temporin V, and others. Sysmex SE-9000 hematol- ogy analyzer: performance evaluation on leukocyte differential counts using an NCCLS H20-A protocol. Am J Clin Pathol 1997; 108:674-86.

9. Jones RG, Faust AM, Matthews RA. Quality team approach in evaluating three automated hematology analyzers with five-part differential capability Am J Clin Pathol 1995;103:159-66.

10. Thalhammer-Scherrer R, Knobl P, Korninger L, and others. Automated five– part white blood cell differential counts. Efficiency of software-generated white blood cell suspect flags of the hematology analyzers Sysmex SE-9000, Sysmex NE-8000, and Coulter STKS. Arch Pathol Lab Med 1997;121:573-7.

11. Gulati GL, Kocher W, Schwarting R, and others. Suspect flags and regional flags on the Coulter-STKS. Lab Med 1999;30:675-80.

12. Cornbleet PJ, Myrick D, Levy R. Evaluation of the Coulter STKS five-part differential. Am Jour Clin Pathol 1993;99:72-81.

13. Bowen KL, Procopio N, Wystepek E, and others. Platelet clumps, nucleated red cells, and leukocyte counts: a comparison between the Abbott CELL– DYN 4000 and Coulter STKS. Lab Hematol 1998;4:7-16.

14. Fournier M, Gireau A, Chretien M, and others. Laboratory evaluation of the Abbott Cell DYN 3500 5-part differential. Am J Clin Pathol 1996;105:286-92.

15. Picard F, Gicquel C, Marnet L, and others. Preliminary evaluation of the new hematology analyzer COULTER GEN-S in a university hospital. Clin Chem Lab Med 1999;37:681-6.

16. Kunicka JE, Fischer G, Zelmanovic D. Improved platelet count accuracy: two-dimensional platelet analysis. Am Clin Lab 1998;Nov-Dec:6-7.

17. Kutter D. Prevalence of myeloperoxidase deficiency: population studies using Bayer-Technicon automated hematology. J Mol Med 1998;76:669-75.

18. Thirup P. LUC, what is that? Clin Chem 1999;45:1100.

19. Sanzari M, DeToni S, D’Osualdo A, and others. Complete analytical and diagnostic performances of the Abbott Cell Dyn. 3500. Panminerva Med 1998;40:116-25.

20. Dorner K, Schulze S, Reinhardt M, and others. Improved automated leukocyte counting and differential in newborns achieved by the haematology analyzer CELL-DYN 3500. Clin Lab Haematol 1995;17:23-30.

21. Mentz F, Baudet S, Maloum K, and others. Quantification of apoptosis by the Abbott CD4000 hematology analyzer. Hematol Cell Ther 1998;40:183-8.

22. Kim YR, Yee M, Metha S, and others. Simultaneous differentiation and quantitation of erythroblasts and white blood cells on a high throughput clinical haematology analyzer. Clin Lab Haematol 1998;20:21-9.

23. Goossens W, Scott CS, Walsh A, and others. First basic performance evaluation of the CELL-DYN 4000. Lab Hematol 1996;2:151-6.

24. National Committee for Clinical Laboratory Standards. Reference leukocyte differential count (proportional) and evaluation of instrumental methods, NCCLS Document H20-A. Villanova, PA, 1992.

MARY E KOENN, BEVERLY A KIRBY, LINDA L COOK, JULIE L HARE, SHARON H HALL, PAULA M BARRY, CHERYL L HISSAM, STEPHANIE B WOJCICKI

Clin Lab Sci 2001;14(4):238

Mary E Koenn MS is Assistant Professor, Medical Technology Program; Beverly A Kirby MA is Assistant Professor, Medical Technology Program; Linda L Cook MD FCAP FASCP is Assistant Professor, Department of Pathology, all at West Virginia University, Morgantown WV

Julie L Hare SH is ClinicalApplications Specialist, Hemostasis and Chemistry, at Dade Behring, Integrated Services Division, Deerfield IL.

Sharon H Hall is a Technical Specialist in Hematology; Paula M Barry is a Specialist in Hematology; Cheryl Pam is a Specialist in Hematology; Stephanie B Wojcicki is a Specialist in in Hematology, all at West Virginia University Hospitals Inc, Morgantown WV.

Address for correspondence: Mary Ellen Koenn MS, Medical Technology Program, West Virginia University, Room 2163C, PO Box 9211, Morgantown WV26506-9211. (304) 293-1632, (304) 293– 6249 (fax). MKoenn@hsc.wvu.edu

Copyright American Society for Clinical Laboratory Science Fall 2001

Provided by ProQuest Information and Learning Company. All rights Reserved