Detection and Classification of Acute Leukemia by the Coulter STKS Hematology Analyzer

Transcription

1 HEMATOPATHOLOGY Detection and Classification of Acute Leukemia by the Coulter STKS Hematology Analyzer JAMES D. HOYER, MD, 1 CATHERINE P. FISHER, MD, 1 VICKI M. SOPPA, BS, 1 KAY L. LANTIS, SH, MT(ASCP), AND CURTIS A. HANSON, MD 1 This study evaluated the ability of the Coulter STKS Hematology Analyzer (Coulter, Hialeah, FL) to detect and classify acute leukemias involving the peripheral blood. One hundred ten acute leukemias cases with circulating blasts were studied: 7 acute myeloid leukemias (AML) and 38 acute lymphoblastic leukemias (ALL). The leukemias were divided into "high count (>11. X 1 9 /L)" (8 AML, ALL) and "normal/low count (<11. X 1'/L)" (44 AML, 16 ALL) categories. Most cases in the high count group elicited the blast suspect flag. The remaining cases of AML and ALL in both the high and normal/low count group were detected by the blast flag, other suspect flags, and/or definitive flags. Only one case of AML-M6 was initially missed using these flag combinations; a subsequent analysis elicited the blast flag. The blast populations in the high count group were localized into characteristic myeloblast or lym- In the past two decades there has been a continuous improvement in automated hematology analyzers. In addition to providing multiparameter quantitative information, recent technologic advances have led to the successful automation of the five-part white blood cell () differential. Using complex computer softwaregenerated algorithms, the current generation of analyzers have the ability to detect abnormal cell populations and provide semi-interpretative "flags." The Coulter STKS Hematology Analyzer system (Coulter, Hialeah, FL) identifies either quantitative abnormalities (ie, definitive flags, or qualitative abnormalities, or suspectflags).definitive flags are defined by the individual laboratory user and can be adjusted to the degree of trust and experience that the laboratory has with automated differential counts. The suspect flags are internally de- From the ' Department of Laboratory Medicine and Pathology, Division ofllematopathology, Mayo Clinic. Rochester, Minnesota; and Department of Pathology, Hematology Laboratory. University of Michigan Hospitals. Ann Arbor. Michigan. Manuscript received February 6, 1996; revision accepted March 13, Address reprint requests to Dr. Hoyer: Mayo Clinic, First Street SW, Rochester, MN phoblast regions of the scatterplot in 8.1% of AML and 63.6% of ALL cases, respectively. However, the remaining cases had indeterminate or aberrant scatterplot patterns, such that an accurate leukemia classification was not possible. The scatterplot pattern also was not helpful in differentiating AML FAB subclasses. The authors conclude that using a combination of appropriate suspect and definitive flags to trigger checking criteria and microscopic review, the Coulter STKS Hematology Analyzer will be successful in detecting virtually all cases of acute leukemia involving the peripheral blood. Although the scatterplots may give useful information, the patterns obtained are not sufficiently distinctive to aid in classifying acute leukemias. (Key words: Coulter STKS; Hematology analyzer; Automation; Acute leukemia) Am J Clin Pathol 1996;16: fined via computer software by the manufacturer and are based on abnormalities in light scatter, conductivity, and electrical impedance measurements. A particular flag is raised when a significant cell population falls outside the normal distribution in this three-dimensional array. These instrument-generatedflags,used in conjunction with laboratory-established criteria for quantitative or distributional abnormalities, can be used to develop morphologic "checking criteria." These criteria are critical in determining which cases require a microscopic review or a manual differential. The potential for time and cost savings with such an automated differential system is significant. As always, maximizing the effectiveness of such a screening system without compromising patient care is key. Acute leukemia brings a special problem to the clinical laboratory as related to automated differential counts. The purpose of our study was to specifically evaluate: (1) the ability of the Coulter STKS to detect the presence of blasts in an acute leukemia; and () the ability of the instrument to aid in the distinction of acute lymphoblastic leukemia (ALL) from acute myeloid leukemia (AML). Additionally, we were interested if FAB subclassification of AML cases could be accomplished by evaluating the Coulter STKS scatterplot data. 35

2 HOYER ET AL. 353 Coulter Hematology Analyzer and '. Classification of Acute Leukemia TABLE 1. COULTER STKS SUSPECT FLAGS* Flags prompting a morphologic review Immature granulocytes/bands- (IGB) Variant lymphocytes (VL) Blasts (B) Nucleated red blood cells (NRBC) Platelet clumps (PltCl) Giant platelets (GPU) Review slide (RS) Rags not prompting a morphologic review Immature granulocytes/bands-1 (IGB1) Dimorphic RBC population (DiRBC) Micro-RBCs/RBC fragments (Mf-RBC) RBC agglutination (Agg) * The abbreviations represent Coulter STKS instrument notation. MATERIALS AND METHODS One hundred ten cases of newly diagnosed acute leukemia with circulating blasts were included in this study. These included 84 cases from Mayo Medical Center, Rochester, Minnesota (August 1993-December 1994) and 6 cases from the University of Michigan Hospitals, Ann Arbor, Michigan (October 1991-March 1993). In all cases, the complete blood count (CBC) results from a Coulter STKS automated hematology analyzer was available. This included the multiparameter complete blood count (CBC) and five-part automated white blood cell () differential. Also included were the white blood cell scatterplot printouts and accompanying instrument-generated leukocyte suspect flags and user-defined definitive flags. The cases were part of the normal daily workload of the laboratories and the EDTA peripheral blood specimens were processed and data obtained during routine operation of the STKS according to both standard hematology laboratory procedures and manufacturer's instructions. Wright-Giemsa-stained peripheral blood smears with a manual differential count were available in all cases. 1 Bone marrow aspirates and/or biopsy specimens taken at or near the time of the peripheral blood specimens (within days) and any ancillary studies (cytochemical results and flow cytometric immunophenotyping) that were performed were used to classify the leukemic process. These cases were all categorized according to the French-American-British (FAB) system. " 5 For the purposes of the study, all cases of acute leukemia were further divided into a "high" count group (> 11. X 1 9 /L) and a "normal/low" count group (^11. X 1 9 /L). The first part of the study evaluated the ability of the Coulter STKS to detect acute leukemias using the instrument flagging capabilities. For example, all suspect flags used in the study required at least a morphologic review of the slide (Table 1). Additionally, high and low defini- tive flags could also generate a morphologic review depending on the type of abnormality that was present (Table ). The leukemic cases were then subdivided into three categories based on flagging results: (1) those cases in which the suspect flag included a blast flag; () those cases that had suspect flags other than a blast flag (with or without definitive flags); or (3) those cases that had only definitive flags and no suspect flags. These cases were then analyzed, based on the types of flags present, to determine which cases would or would not have generated a morphologic review. Although many of these cases were suspected clinically, we wished to determine whether any case would have been missed if based exclusively on Coulter STKS data. The second portion of the study evaluated the usefulness of the scatterplot data in classifying acute leukemias as AML or ALL and for the subclassification of AML. The normal/low count group was excluded from this portion of the study due to the relatively small number of circulating blasts. In the high count group, the scatterplots were reviewed independently by two observers (CPF/JDH) and the type of leukemia was predicted based on the appearance of the scatterplot. The cases were divided into three groups, based on the location of the apparent blast population: (1) "typical," in which the blast population localized to characteristic regions of the scatterplot (ie, lymphoblasts near the lymphocyte region, myeloblasts near the monocyte and/or neutrophil regions); () "indeterminate," in which the blast population extended through both lymphoblast and myeloblast regions or a discrete population localized to an intermediate or overlapping position; or (3) "aberrant," in which blasts localized to an unexpected region (ie, lymphoblasts in neutrophil region). Additionally, the high count AML cases were further analyzed to TABLE. COULTER STKS DEFINITIVE FLAGS: LOW/HIGH LIMITS Parameter Low High s <1.X1 9 /L >99.9X1 9 /L Neutrophils <1.X1 9 /L >1.X 1 9 /Lor>99.9% Lymphocytes <.5 X 1 9 /L >5.5 X 1 9 /Lor>99.9% Monocytes >.X 1 9 /Lor>% Basophils >.5 X 1 9 /Lor>5% Eosinophils >1.5 X 1 9 /Lor>99.9% Platelets <4 X 1 9 /L >999X1 9 /L RBCs <.X1' /L >8.X1 I /L Hgb <5.g/dL >.g/dl RDW >17.5 MCV <7.fL > 15. fl = white blood cells; RBCs = red blood cells: Hgb = hemoglobin; RDW = red cell distribution width: MCV = mean corpuscular volume. Vol. No. 3

3 354 HEMATOPATHOLOGY TABLE 3. ACUTE LEUKEMIAS: CLASSIFICATION MO Ml M M3 M4 MS M6 M7 Acute myeloid leukemia High s(>l 1. X 1 9 /I_) Normal/low s (<; 11. X 1'/L) (n = 8) (n = 44) (n = 7) B-Precursor TCell Burkitt's/L3 Acute lymphoblastic leukemia High s (>11. X 1 9 /L) Normal/low s (<; 11. X 1 9 /L) s = while blood cells. (n = ) (n=16) (n = 38) determine whether any FAB subtypes produced distinctive patterns that would aid in their subclassification. RESULTS The 11 acute leukemia cases studied included 7 cases of AML and 38 cases of ALL. Table 3 lists the AML FAB subtypes and the immunologic subtypes of ALL. The number of respective cases with a high count or a normal/low count are also listed. Detection of Acute Leukemia Acute myeloid leukemia (AML). All AML cases with a high count (n = 8) generated a blast flag. The percentage of blasts identified morphologically in these cases ranged from 7% to 99%. Other suspect flags were present in 17 of these 8 cases and included immature Acute myeloid leukemia High s(> 11.OX 1 9 /L) Normal/low s (si 1. X 1 9 /L) Acute lymphoblastic leukemia High s (>11. X 1 9 /L) Normal/low s (<;11. X 1 9 /L) High s (>11. X 1 9 /L) Normal/low s ( 11.OX 1 9 /L) TABLE 4. ACUTE LEUKEMIAS: COULTER STKS FLAGS (n = 8) (n = 44) (n = 7) (n = ) (n=16) (n = 38) (n = 5) (n = 6) (n = 11) 3 granulocytes/bands (IGB) (n=1), variant lymphocytes (VL) (n = 7), and nucleated red blood cells (nrbc) (n = ). The 44 remaining AML cases had normal or low counts; the percentage of blasts identified morphologically in these cases ranged from % to 7%. Twenty-four of these 44 (54.5%) AML cases elicited a blast flag. Nonblast suspectflagswere generated in 15 of the remaining cases that did not generate a blast flag; only a small number of circulating blasts were identified morphologically in these cases. The nonblast suspect flags included VL (n = 8), nrbc/platelet clumps/giant platelets (n = 5), IGB (n = 1), and review slide (n = 1). Four more of these latter AML cases were identified by definitive flags only; three of these four had neutropenia and one case had anemia, thrombocytopenia, and an increased RDW. The one remaining case in this group, although Blast* 8/8 (1.%) 4/44 (54.5%) 5/7 (7.%) 17/ (77.3%) 9/16 (56.3%) 6/38 (68.4%) 45/5 (9.8%) 33/6 (55.%) 78/11(7.9%) 6 6 Other Suspect-]- 15/44 (34.1%) 15/7 (.8%) 3/ (13.6%) 4/16 (5.%) 7/38 (18.4%) 3/5 (6.%) 19/6 (31.7%) /11(.%) Definitive Flag(s) Only$ 4/44 (9.1%) 4/7 (5.6%) / (9.1%) 3/16 (18.8%) 5/38 (13.%) /5 (4.%) 7/6 (11.7%) 9/11(8.%) s = white blood cells. * Blast suspect flag present ± other suspect and definitive flags. t Leukocyte suspect flags other than "Blast" ± other definitive flags. X Only definitive flag(s) present; no leukocyte suspect flags. A.J.C.P.'September 19%

4 HOYER ET AL. 355 Coulter Hematology Analyzer and the Classification of Acute Leukemia suspected clinically for leukemia, would not have been initially detected based on our flagging and review criteria. Th'is case was subsequently classified by bone marrow examination as an acute erythroleukemia (AML, FAB-M6). Although the patient had pancytopenia, the magnitude of his cytopenias were not enough to trigger a review based on how the definitive flags had been defined. The peripheral blood smear differential showed 18% blasts; however, because the was.5 X 1 9 /L, a low absolute number of circulating blasts was present. The only suspectflag that was elicited initially in this case was an IGB1flag,which was not sufficient for a morphologic review. However, a blast flag was seen on a subsequent sample days later, before the start of chemotherapy. Acute lymphoblastic leukemia. Seventeen of ALL cases (77.3%) with high counts showed a blast flag, often in conjunction with the VLflag (n = 6), the IGB flag (n = 5), or both VL and IGB flags (n = 5). Only one case had the blastflagalone. In addition, three cases (13.6%) were identified by the VL flag alone. The remaining two cases (9.1%) generated only definitive flags: (1) leukocytosis and basophilia; and () leukocytosis, anemia, and thrombocytopenia. Nine of the 16 ALL cases (56.3%) with normal or low counts elicited a blast flag. In all but one of these nine cases, a blast flag was accompanied by the VL suspect flag. Four cases (5.%) were identified by other nonblast suspectflags,the most common of which again was the VL flag (3 of 4 cases). The remaining three (18.8%) ALL cases were identified by quantitative abnormalities alone. These included (1) anemia; () an elevated RDW and thrombocytopenia; and (3) neutropenia, lymphocytosis, and increased RDW. Summary. The summary of all AML and ALL cases in this study are shown in Table 4. More than 9% of cases generated a blast or other suspect flags. The blast flag was commonly accompanied by other suspect flags, often the VL and/or IGB flag. In both ALL categories and in the normal or low count AML, there were a small minority of cases that elicited only definitive flags and no suspect flags. The definitive flags in these cases prompted further morphologic evaluation in which the leukemic cells were then detected. As stated previously, only one case of AML in the normal/low count group would not have prompted a microscopic review based onflaggingcriteria. Leukemia Subclassification We also evaluated the potential role of the scatterplots in classifying the acute leukemias. Figure 1 shows the normal distribution of peripheral blood elements in a VBC V O L u M E MONOCYTE LYMPHOCYTE nrbc/plt/debris PMN EOSINOPHIL DF1 FIG. 1. Scatterplot showing normal distribution of peripheral blood elements. Coulter STK.S scatterplot. Table 5 summarizes the appearance of the scatterplots from the high count AML and ALL cases in our study. The normal/low count group was excluded from this portion of the study due to the relatively small number of circulating blasts. Most AML, 3 of 8 (8.1 %), and ALL, 14 of (63.6%), would have been accurately classified based on the scatterplot appearance (Fig. ). However, there were significant numbers of both AML and ALL cases that did not fit into a classic scatterplot pattern (Fig. 3). In the AML group, both cases with aberrant scatterplots that localized to the lymphocyte region were classified as AML-M1. The three cases of AML demonstrating indeterminate patterns were classified as AML-M 1, -M, and -M5. Two cases of ALL had aberrant scatterplot patterns that were primarily confined to the neutrophil area; both were classified as B-precursor ALL. The six cases of ALL which gave indeterminate patterns usually had a large variability in the size of the blasts and were classified as B-precursor ALL (n = 3), T-cell ALL (n = ), and ALL-L3(n= 1). FAB subclassification of the AML cases based on the peripheral blood scatterplots was not consistent or reliable in most cases. All four cases of acute promyelocytic leukemia (AML-M3) with a high count gave a reproducible scatterplot pattern (Fig. B). This was not disease-specific as many cases of AML-M 1 or -M had a similar scatterplot appearance. In many cases of AML, the localization of the blasts to the monocyte or neutro- Vol. 16-No. 3

5 356 HEMATOPATHOLOGY TABLE 5. SCATTERPLOT/MORPHOLOGY CORRELATION IN HIGH WHITE BLOOD CELL COUNT ACUTE LEUKEMIA AML(n = 8) ALL(n = ) Typical Indeterminate Aberrant 3/8(8.1%) 14/(63.6%) 3/8(1.7%) 6/ (7.3%) /8(7.1%) /(9.1%) AML = acute myeloid leukemia: ALL = acute lymphoblastic leukemia; Typical = blast population localized to characteristic regions of the scatterplot (i.e.. lymphoblasts near lymphocyte region, myeloblasts near monocyte and neutrophil regions); Indeterminate = blast population extending through both lymphoblast and myeloblast regions or a discrete population localized to an intermediate or overlapping position; Aberrant = blasts localized to unexpected region (i.e.. lymphoblasts in neutrophil region). phil region seemed to depend on what mature cells predominated in the peripheral blood (Fig. 4). This was particularly true of cases classified as AML-M4 and -M5. Specifically, six cases of AML-M4 and -M5 had the leukemic population placed within the monocyte area, whereas the otherfivecases were localized mainly within the neutrophil area of the scatterplot. This discrepancy was divided equally among the AML-M4 and -M5 cases. DISCUSSION V L U H E A B ' Clinical hematology laboratories have undergone significant changes in the automation arena due to continued technological advances. The era of manual CBC determinations has passed, with automated blood cell analysis now a routine part of virtually all hematology laboratories. The greatest change in the hematology laboratory has been the development of more sophisticated and accurate leukocyte differential counters as part of the hematology analyzer instrumentation. Automated differential leukocyte counts as performed by modern hematology analyzers are more accurate, more precise, more economical, faster, and safer than traditional manual differential counts. Numerous evaluations of various hematology analyzers have been published with most showing excellent correlation between automated differential results and those obtained by manual differential counts. 6 " 15 Although the goal of any automated differential leukocyte counting system is to replace the manual differential count, this has not been fully accomplished in the clinical laboratory. The automated leukocyte differential count is superior to the manual differential count in many aspects, but in some cases it fails to provide important morphologic detail that only the microscopic review can provide. The detection of acute leukemia by a hematology analyzer is of special importance. It is critical that specimens with suspected or possible acute leukemia be evaluated for leukemic cells. As a part of instrument evaluation studies, most authors have included some acute leukemia cases in their study sets, although generally in small numbers."" 15 These studies have demonstrated the difficulties in detecting lymphoblasts, recognizing small numbers of blasts, and distinguishing blasts from variant lymphocytes. Only three previous studies have specifically addressed the evaluation of acute leukemias on automated hematology analyzers. 16 " 18 Two of these studies used the Technicon H-l instrument. Ka-., jbjjgi;.;- i= : :.?:?«":.!.;... '.;!' Mp- ' HHHK':*: '.' SjjJS *J$j}. : : : : $ ' : ' '. ',-... * " " OF : 'flfhwh...ljj$bs&3v T \ '..'"'. ; < ijv i"v*' ' ^ OF 1 FIG.. Examples of typical acute leukemia scatterplots from patients with a high. A. B-precursor ALL ( = 7. X 1 9 /L); blasts are within the lymphocyte region; B, AML, FAB-M3 ( = 86. X 1 9 /L); promyelocytes are within the neutrophil region; C, AML, FAB- M5a ( = 98. X 1 9 /L); blasts are within the monocyte region. A.J.C.P.«September 1996

6 HOYER ET AL. 357 Coulter Hematology Analyzer and the Classification of Acute Leukemia : '; ft- ;*ss Iff EL,*'*'.. gjgj Eg:;/ ' '*ifl.y^sls,- l '^'-^*ffl.' -r^lll :* - : M^M ;..;:JsgiSSj ' ' ("''v^iw^'"' '! - } :.. ' I 1 ft' I fe; r,i.''. «r* *. * Of 1 FIG. 3. A. Example of a B-precursor ALL demonstrating an aberrant scatterplot pattern ( =.4 X 1 9 /L); blasts are within the neutrophil region; B, Example of a T-cell ALL showing a indeterminate scatterplot pattern. ( = 59. X 1 9 /L); blasts are distributed over a wide area from the lymphocyte to the monocyte region. warabayashi 16 evaluated that instrument's ability to detect leukemic blasts based on the blast suspectflaggenerated by the instrument. Krause and colleagues 17 evaluated the instrument capability in making the distinction between AML and ALL based on the histograms obtained from the instrument and the blast suspect flag. Neither of these studies incorporated the blast flag with other suspect flags or numerical CBC abnormalities to evaluate the hematology laboratory's ability to detect acute leukemia. Only one previous study evaluating acute leukemias has used the Coulter technology. 18 This study was limited due to the fact that only the VCS portion (that portion concerning the automated leukocyte differential) was evaluated and did not utilize the entire Coulter STKS instrument, such as the red and white blood cell counting parameters. Using both the suspect and definitiveflaggingcapabilities of the STKS, we found that virtually all cases of acute leukemia in a retrospective series would have generated a microscopic evaluation of the blood smear had the acute leukemia been clinically unsuspected. In cases of acute leukemia with normal or low peripheral counts, the detection of leukemia was sometimes more difficult as a blast flag was not always generated. However, other suspect and definitive flag abnormalities led to further morphologic analysis of these "minimally leukemic" cases. The importance of incorporating quantitative criteria (ie, definitive flags), in addition to suspect flags was born out by a small number of cases (5.6% AML and 13.% ALL) that were detected only on the basis of abnormal quantitative CBC parameters. It is important to emphasize that the scatterplots obtained in cases of acute leukemia are not by themselves diagnostic. For example, other entities such as a leukemic phase of malignant lymphoma or prolymphocytic leukemia could potentially give a scatterplot similar to acute leukemia with generation of a blast or other suspect flags. Subsequent evaluation of suspect and defini- V L U H E A " : /» I'^a ski '* ittv'wt; P* - IP':.* ' w JF t.. OF 1 FIG. 4. Two cases of AML, FAB-M4, demonstrating differences in scatterplot gating: A, The blast population is within the monocyte region ( = 19.4 X 1 9 /L; 45% blasts on manual differential); B, The blast population is within the neutrophil region ( = 91.7 X 1 9 /L; 83% blasts on manual differential). i Vol. 16-No. 3

7 358 HEMATOPATHOLOGY tiveflagsby morphologic review will clarify these potentially ambiguous cases. The variant lymphocyte (VL) flag was very useful in this study, particularly in regard to detecting ALL. The majority of ALL cases had both the blast and VL flags occurring together. However, in a minority of cases, the blastsflaggedonly as VL. In AML cases with a normal or low count, there were also a small number of cases in which a VLflagwas present, but not a blastflag.subsequent microscopic review identified a small number of circulating blasts. Although blasts sometimes mayflagas variant lymphocytes, we feel this is not a problem as long as this phenomenon is recognized. For this reason, it is our opinion that all VLflagsshould be followed by a peripheral blood smear review. Our evaluation showed that the scatterplots obtained by the Coulter STKS were not sufficiently precise to accurately distinguish AML from ALL. Although within each group the majority of cases gave a similar distinct pattern, there were still a significant minority of cases in each group (17.8% AML and 36.4% ALL) that gave a confusing or indeterminate scatterplots. This precludes using scatterplots in giving preliminary diagnostic information that would help direct therapy. The scatterplots were likewise not sufficiently distinctive to subclassify FAB subsets of AML. The STKS computer software uses floating discriminators between the monocyte, granulocyte, and lymphocyte regions in determining where to draw the lines between the different cell populations. In some cases, the determination of whether the blast population was in the monocyte or granulocyte scatterplot region seemed to depend on what, if any, other cell types comprised the remainder of the differential (Fig. 4). This appears to be the main problem hampering the use of these scatterplots in the subclassification of AML. In summary, the Coulter STKS hematology analyzer, when combined with appropriate morphologic checking criteria based on suspect and definitiveflags,will be successful in prompting microscopic review of virtually all new cases of acute leukemia. Our study indicates that the increasing sophistication of the Coulter STKS, as well as other hematology analyzers, should result in further decreased reliance on the ordering of manual differential counts. However, current automated leukocyte differential instrumentation will not totally replace morphologic review of the peripheral blood smear. Likewise, clinical suspicions will always play an important diagnostic role, and must be an acknowledged part of the morphologic review process. Our study demonstrates that a carefully designed automated leukocyte differential approach that uses appropriate microscopic review will detect virtually all acute leukemia cases that present to the clinical hematology laboratory. Acknowledgments. The authors thank Ms. Sara Brackett for her diligent assistance in preparing this manuscript. REFERENCES 1. National Committee for Clinical Laboratory Standards. Reference leukocyte differential count (proportional) and evaluation of instrumental methods; Approved Standard, NCCLS document H-A(1SBN ) Villanova, PA: NCCLS, Bennett JM, Catovsky D, Daniel MT, et al. Proposal for the classification of the acute leukemias. French-American-British (FAB) Cooperative Group. Br J Haematol 1976;35: Bennett JM, Catovsky D, Daniel MT, et al. Criteria for the diagnosis of acute leukemia of megakaryocyte lineage (M7): A report of the French-American-British Cooperative Group. Ann Intern Med 1985; 13: Bennett JM, Catovsky D, Daniel MT, et al. Proposed revised criteria for the classification of acute myeloid leukemia: A report of the French-American-British Cooperative Group. Ann Intern Med 1985; 13: Bennett JM, Catovsky D, Daniel MT, et al. Proposal for the recognition of minimally differentiated acute myeloid leukemia (AML-MO). Br J Haematol 1991;78: Warner BA, Reardon DM. Afieldevaluation of the Coulter STKS. Am J Clin Pathol 1991 ;95: Swaim WR. Laboratory and clinical evaluation of white blood cell differential counts: Comparison of the Coulter VCS, Technicon H-l, and 8-cell manual method. Am J Clin Pathol 1991;95: Robertson EP, Lai HW, Wei DC. An evaluation of leukocyte analysis on the Coulter STKS. Clin Lab Haematol 199; 14: Verheul FEAM, Spitters JMC, Bergmans CHF. Evaluation and performance of the Coulter STKS. Eur J Clin Chem Clin Biochem 1993;31: Buttarello M, Gadotti M, Lorenz C, et al. Evaluation of four automated hematology analyzers: A comparative study of differential counts (imprecision and inaccuracy). Am J Clin Pathol 199;97: Fialon P, Becker M, Boisseau MR. Evaluation of the white blood cell differential given by the new analyzer Coulter STKS in both hematologic and non-hematological patient groups. Pathol Biol 199;38: Cornbleet JP, Myrick D, Levy R. Evaluation of the Coulter STKS five-part differential. Am J Clin Pathol 1993,99: Bentley SA, Johnson A, Bishop CA. A parallel evaluation of four automated hematology analyzers. Am J Clin Pathol 1993; 1: Fournier M, Adenis C, Fontaine H, et al. Evaluation and use of the white blood cell differential provided by the Coulter'" STKS in a children's hospital. Clin Lab Haematol 1994; 16: Moreau E, Philippe J, Klint J, et al. Performance of the Coulter STKS and the Sysmex NE-15 in a paediatric population. Eur J Clin Chem Clin Biochem 1991;3: Kawarabayashi K, Tsuda I, Tatsumi N, et al. Leukemic blasts detected by the Technicon H-l blood cell counter. Am J Clin Pathol 1987;88: Krause JR, Costello RT, Krause J, et al. Use of the Technicon H-l in the characterization of leukemias. Arch Pathol Med 1988;11: Bruno A, Del Poeta G, Venditti A, et al. Diagnosis of acute myeloid leukemia and system Coulter VCS. Haematologica 1994; 79: A.J.C.P.- September 1996