Toward a reference method for leukocyte ... - Wiley Online Library

85 downloads 12786 Views 565KB Size Report
Jun 26, 2012 - a reference differential count in method comparisons or validations (5). However, ..... mode files were sent to an FTP server. Access to the list mode ..... ond tube dedicated to nRBCs identification with CD45 and. DRAQ5.
Original Article

Toward a Reference Method for Leukocyte Differential Counts in Blood: Comparison of Three Flow Cytometric Candidate Methods Mikael Roussel,1* Bruce H. Davis,2 Thierry Fest,1,3,4 Brent L. Wood,5* on behalf of the International Council for Standardization in Hematology (ICSH)

1

CHU de Rennes, Laboratoire d’He´matologie, Pole de Biologie, Rennes, F-35033 France

2

Trillium Diagnostics, Bangor, Maine

3

INSERM, UMR U917, Rennes F-35043, France

4

Universite´ Rennes 1, F-35043 Rennes, France

5

Laboratory Medicine, University of Washington, Seattle, Washington

Additional Supporting Information (MIFlowCyt) may be found in the online version of this article. Received 7 October 2011; Revision Received 1 June 2012; Accepted 1 June 2012 Grant sponsor: International Council for Standardization of Haematology (ICSH) *Correspondence to: Mikael Roussel, Laboratoire d’He´matologie, P^ole de Biologie, CHU de Rennes, 2 rue Henri Le Guilloux, 35033 Rennes Cedex 9, France. E-mail: [email protected] or Brent L. Wood, Laboratory Medicine, University of Washington, Seattle Cancer Care Alliance, Mail Stop G7-800, 825 Eastlake Ave. E., Seattle, WA 98109 Email: [email protected] Published online 26 June 2012 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/cyto.a.22092 © 2012 International Society for Advancement of Cytometry

Cytometry Part A  81A: 973 982, 2012

 Abstract A Complete Blood Count performed by an automated hematology analyzer frequently requires a microscopic slide review. Recently, we and others have proposed combinations of monoclonal antibodies for an extended leukocyte differential by flow cytometry. The aim of this study was to compare the performance of these proposals. Ninetytwo samples were analyzed at 2 sites to compare the accuracy of three published methods. Reference methods used were i) cell counter for leukocyte count and ii) microscopic review as defined by CSLI H20-A2 for cell subsets. Comparison of flow cytometers from 2 manufacturers (FC500 and CANTO/LSRII) was performed. Published protocols were adapted to three different models of flow cytometer and each provided similar results in leukocyte subset enumeration, although some discrepancies were noted for each protocol in comparison with the reference method. The conclusion is that each protocol carries advantages and disadvantages and there is no clear ‘‘winner’’. This study supports the fact that flow cytometry is a candidate to become a reference method for the leukocyte differential. None of the tested protocols clearly demonstrated superiority and each had demonstrable deficiencies. Additional work to develop a consensual 8 to 10 color panel is concluded to be necessary for a satisfactory reference method. ' 2012 International Society for Advancement of Cytometry  Key terms flow cytometry; leukocyte differential; complete blood count; reference method; laboratory instrumentation

THE

complete blood count (CBC) with automated leukocyte differential is an important test for the diagnosis or evaluation of a wide spectrum of diseases from inflammatory states and sepsis to malignancies. Currently the CBC and leukocyte differential is provided by an automated cell counter and of these 10-50% of specimens generate leukocyte-related abnormal cell flags; this often requires a manual microscopic review of a blood film (1–3). In routine practice, manual review consists of a 100–200 nucleated cell microscopic count and morphologic classification of the different types of leukocytes present. The normal blood smear has five principal leukocytes types, i.e., neutrophils, eosinophils, basophils, lymphocytes and monocytes, but other cells can be present such as reactive lymphocytes, immature granulocytes, blasts, nucleated red blood cells, and less frequently plasma cells. In addition, cell morphology must strive to identify abnormal cells (4). Manual differentials are timeconsuming, require experienced operators yet nevertheless rely on subjective interpretations. A 400-cell count (200 cells by two observers) is currently recommended by Clinical and Laboratory Standards Institute (CLSI) H20-A2 as means to generate a reference differential count in method comparisons or validations (5). However, the accuracy and statistical precision of a reference method surveying only 400 cells, compounded by the inherent subjectivity in the distinction of some cell types, seems

ORIGINAL ARTICLE Table 1. Flowchart of the study FLOWCHART

Flagged samples (n) Reference method Cell counter Manual Diff (CLSI H-20) Flow Cytometer model FC500 CANTO LSRII Rejected samples (n)

SEATTLE

RENNES

50

50

XE-2100 Yes

LH-755 Yes

Yes No Yes

Yes Yes No

0

8

Hematology cell counters used where LH755, Beckman Coulter, Miami, FL and Sysmex XE2100, Kobe, Japan, respectively, for Rennes and Seattle. Additionally a 400-leukocyte count was performed as the reference method as recommended by the H20-A2 document from the CLSI (5).

absurd given current technologies that typically count tens of thousands or more cells with a high degree of accuracy and precision (6). Multiparameter flow cytometry (FCM) seems to present an attractive alternative to a microscopic differential. Indeed, hundreds of thousands of cells can be analyzed in a standardized manner giving the opportunity to detect not only the populations recognized in a blood smear, but also other subsets of cells including various blast and lymphocyte sub-populations. Recently, we and others demonstrated the feasibility of the flow cytometry approach using combinations of monoclonal antibodies allowing the detection and quantification of up to 16 cell subsets (7–10). Currently at least three different flow cytometry protocols have been published with the potential application of becoming a reference method for cell differential counting. The International Council for Standardization of Haematology (ICSH), as part of their mission to identify improved methods of standardization for laboratory hematology diagnostic assays, deemed it important to assess the proposed panels against each other. In this attempt to define an improved reference method for the leukocyte differential by flow cytometry, an international working party was formed by ICSH. After a definition of requested cell subsets to be recognized by the candidate methods, this study was undertaken to understand the relative merits and weakness among the three protocols in hopes of defining a reliable reference method for the validation and calibration of present and future CBC analyzers.

MATERIALS AND METHODS Patients Selection and Study Design The hematology laboratories from the University Hospital of Rennes (France) and the University of Washington Medical Center (Seattle, WA) participated in this study. In each laboratory, peripheral blood samples flagged by the hematol974

ogy analyzers were randomly selected (50 at Seattle and 50 at Rennes) in an attempt to enrich for the presence of abnormal or immature cell populations (Table 1). Flags included qualitative (suspicion of immature granulocytes, blasts, atypical lymphocytes) as well as quantitative abnormalities. Normal samples were not included in this study. Residual peripheral blood samples following CBC analysis in our routine laboratories were used according to institutional review board and national guidelines. The study protocol was approved by the hospital ethics committees. All specimens were anticoagulated with K2 EDTA and were tested within 9 h after blood collection. As comparative methods, samples were run on the routine cell counter available (LH755, Beckman Coulter, Miami, FL and Sysmex XE2100, Kobe, Japan, respectively, for Rennes and Seattle) following manufacturer recommendations. Additionally, a manual microscopic 400-leukocyte differential count was performed (50 at Seattle and 50 at Rennes) as the reference method as recommended by CLSI H20-A2 (5). Each sample was then further analyzed by flow cytometry using the three protocols described below. Analyses were performed in parallel by both LSRII (Becton-Dickinson, San Jose, CA) and FC500 (Beckman-Coulter, Miami, FL) at Seattle and both CANTO (Becton-Dickinson, San Jose, CA) and FC500 at Rennes (Table 1). Antibodies and Flow Cytometry ICSH mandated that populations to be identified by FCM were neutrophils, basophils, eosinophils, monocytes, lymphocytes, immature granulocytes, blasts, and nucleated red blood cells (nRBCs). Furthermore, the total number of colors was limited to 5 so the method might be widely performed in clinical laboratories. Published candidates protocols were not all in agreement with these goals, hence each of the three tested protocols were modified to meet these aims. The three candidate protocols were referred to by the country of origin of the authors (7–10). Dr. Bjo¨rnsson (Malmo¨, Sweden) proposed to use CD123 instead of CD203c in the Swedish (SWE) protocol to allow more accurate basophil detection (9). Therefore, the SWE combination derived from the Bjornsson’s publication included CD36, CD123, CD138, DRAQ5, CD16, CD56, CD45 in a single 5 color tube; Table 2). The U.S. combination derived from Dr. Cherian’s publication included 2 tubes in 5 colors, respectively, US1: CD19, CD16, CD123, HLA-DR, CD33, CD64, CD45 and US2: Syto16, CD34, CD117, CD45, CD33, CD64, CD38 (8; Table 2). Finally, a second tube was designed to complete the French (FR) protocol as the published one could not identify some requested cell types such as nRBCs or myeloid blasts (7,10). Therefore, the FR combination derived from Faucher’s publication included 2 tubes in 5 colors with respectively FR1: CD36, CD2, CD294, CD19, CD16, CD45 and FR2: Syto16, CD13, CD20, HLA-DR, CD34, CD117, CD45 (10; Table 2). Fluorochromes where then adapted to the clinical models of cytometers provided by two major instrument manufacturers using fluorescein isothiocyanate (FITC), phycoerythrin (PE), PE-cyanine 5 (PC5), and PE-cyanine 7 (PC7) for all panels and PE-Texas Red (ECD) or APC-H7 for, respectively, FC500 or CANTO/LSRII models Leukocyte Differential by FCM: Comparison of Three Methods

Cytometry Part A  81A: 973 982, 2012

FR2

FR1

SWE

US2

US1

FR2

FR1

SWE

NH4Cl/formal No wash NH4Cl/formal No wash Optilyse C/BC No wash Versalyse/BC No Wash Versalyse/BC No wash

NH4Cl/formal No wash NH4Cl/formal No wash OptilyseC/BC No Wash Versalyse/BC No wash Versalyse/BC No wash

LYSIS

FL2 (PE)

CD19/BC/J4.119 CD123/BD/G3 and CD16/BC/3G8 Trucount/BD Syto16*/Invitrogen CD34/BD/8G12 and CD117/DAKO/104D2 Flow count/BC CD36/BC/FA6.152 CD123/BD/9F5 and CD138/BC/B-A38 Flow count/BC CD36/BC/FA6.152 CD2/BC/39C1.5 and CD294/BC/BM16 Flow count/BC Syto16*/Invitrogen CD13/BD/WM15 and CD20/BD/2H7

Trucount/BD

FL1 (FITC except*)

CD19/BC/J4.119 CD123/BD/7G3 and CD16/BC/3G8 Trucount/BD Syto16*/Invitrogen CD34/BD/8G12 and CD117/DAKO/104D2 Flow count/BC CD36/BC/FA6.152 CD123/BD/9F5 and CD138/BC/B-A38 Flow count/BC CD36/BC/FA6.152 CD2/BC/39C1.5 and CD294/BC/BM16 Flow count/BC Syto16*/Invitrogen CD13/BD/WM15 and CD20/BD/2H7

Trucount/BD

BEADS

BLUE DETECTOR D (PE)

BLUE DETECTOR B (PE-Cy5)

HLA-DR/BC/ Immu-357

CD19/BC/RJ4.119

CD45/BC/J.33

HLA-DR/BC/ Immu-357 CD45/BC/J.33

FL3 (PE-TxR)

CD34/BC/581 and CD117/BC/104D2D1

CD33/BC/D3HL60.251 and CD64/BC/22 CD33/BC/D3HL60.251 and CD64/BC/22 CD16/BC/3G8 and CD56/BC/N901(NKH-1) CD16/BC/3G8

FL4 (PE-Cy5)

HLA-DR/BD/L243 CD34/BC/581 and CD117/BC/104D2D1

HLA-DR/BD/L243 CD33/BC/D3HL60.251 and CD64/BC/22 CD45/BD/2D1 CD33/BC/D3HL60.251 and CD64/BC/22 DRAQ5*/Biostatus CD16/BC/3G8 and CD56/BCN901(NKH-1) CD19/BD/SJ25C1 CD16/BC/3G8

RED DETECTOR A (APC-H7 EXCEPT*)

MANUFACTURER/CLONE

CD45/BC/J.33

DRAQ5*/ Biostatus CD45/BC/J.33

CD38/BD/HB7

CD45/BC/J.33

FL5 (PE-Cy7 except*)

CD45/BC/J.33

CD45/BC/J.33

CD45/BC/J.33

CD38/BD/HB7

CD45/BC/J.33

BLUE DETECTOR A (PE-Cy7)

Cells were labeled with fluorochromes adapted to each flow cytometer (LRSII/CANTO or FC500). Two antibodies labeled with the same fluorescence may be mixed (indicated by ‘‘and’’). FL: Fluorescence Line. BC: Beckman Coulter; BD: Becton Dickinson; NH4Cl: Ammonium Chloride 10.25% formaldehyde.

FC500

US1

LSRII/CANTO

US2

METHOD

FLOW CYTOMETER

BLUE DETECTOR E (FITC EXCEPT*)

Table 2. Panels of antibody (clone/manufacturers) and dyes used

ORIGINAL ARTICLE

975

976

105–125 490–530 49–64 49–64 2–4 1–10 3–4 1.5–2.5 5.7–9.7 15–25 9.7–11.7 3–4 10.3–14.3 20–30 12–16 3–4 2.5–8.5 9–13 9.1–11 13–20 85–125 90–110 116–148 113–153

METHOD

US1-2 SWE FR1 FR2

US1-2 SWE FR1 FR2

CANTO/LSRII

FC500

Samples Ninety-two flagged samples were fully analyzed after the exclusion of 8 cases in Rennes due to technical problems (Table 1). Technical problems included platelet aggregation as well as visible non-specific staining notably with Syto16. All together, the mean reference leukocyte count was at 7.1x109 cells/L (range: 0.12-65.2 3 109 cells/L) including leukopenia (n 5 22) and leukocytosis (n 5 45) samples. Of note, 11 patients had more than 1% blasts on manual review (range from 2.5% to 63.5%); 19 had more than 2% of immature

FLOW CYTOMETER

RESULTS

Target channel used for standardization with 8 peaks, Rainbow beads (Spherotech) and FlowSet Fluorosphere (Beckman Coulter) for respectively CANTO/LSRII and FC500. Number of the peak targeted is indicated in brackets. FSC: Forward Scatter; SSC: Side Scatter; FL: Fluorescence Line.

FSC

FSC

64–104 500–550 450–480 328–428

FL5 FL4 FL3 FL2 SSC

FL1

31,500–33,500 (1) 12,000–14,000 (1) 12,000–14,000 (1) 12,000–14,000 (1) 7500–8500 (3) 3500–5500 (3) 27,000–29,000 (2) 7500–8500 (2) 20,500–22,500 (2) 8500–10,500 (1) 20,500–22,500 (2) 20,500–22,500 (2) 23,000–25,000 (3) 15,500–17,500 (3) 6500–7500 (3) 15,500–17,500 (3) 22,000–24,000 (3) 15,000–17,000 (3) 15,000–17,000 (3) 22,000–24,000 (3)

BLUE DETECT B (TARGET PEAK) RED DETECT A (TARGET PEAK) SSC

BEADS TARGETS RANGE

BLUE DETECT D (TARGET PEAK) BLUE DETECT E (TARGET PEAK)

Data Collection and Statistical Analysis In order to share data analysis, all flow cytometric list mode files were sent to an FTP server. Access to the list mode data can be obtained upon request to corresponding authors. Listmode files were analyzed by Kaluza (Beckman Coulter) and Woodlist (written by Brent Wood) software respectively at Rennes and Seattle. Gating strategies previously described were employed for each of the protocols (8–10). All results were exported to an internal Microsoft Excel table. Pearson and Spearman coefficients, Bland and Altman plots, and descriptive analyses were obtained using the GraphPad 5.0 statistics program (Prism software). For normal populations, Pearson coefficients were used for comparison between slide review or cell counter counts and flow cytometry. For abnormal cells (immature granulocytes, blasts and nRBCs) without Gaussian distribution, Spearman coefficients were used for comparison of groups. Bland and Altman plots were used to evaluate the bias between techniques.

Table 3. Targets used for flow cytometry standardization

(Table 2). Antibody cocktails were titered, centrally prepared (BLW, Seattle) and distributed. The choice was made to include in this comparison antibodies and reagents from various manufacturers (Table 2). Each leukocyte-flagged sample was processed using a ‘‘lyse-no wash’’ protocol. Briefly, 100 microliters of blood was labeled with 100 microliters of antibody cocktail. After a 15 min incubation, red blood cells were lysed with Versalyse, Optilyse or NH4Cl buffer (Table 2) and after a further 10 min the samples were ready for analysis by flow cytometry. The lysing reagents used were selected by the authors of the published methods for use in these modified protocols. Counting beads were also added to the prepared sample as indicated (Table 2). Before analysis, flow cytometer settings were standardized (i) FC500 from Rennes and Seattle and (ii) CANTO from Rennes and LSRII from Seattle) using Rainbow Beads (Spherotech, Liberty, IL) or Flow Set Fluorosphere (Beckman Coulter) for respectively for CANTO/LSRII and FC500. Beads were used principally to standardize the settings between similar instruments at the two sites. No attempts were made to standardize intensities between instrument platforms for a given assay other than to perform global instrument optimization prior to defining settings. For each panel, targets defined in the Table 3 were used. For each sample, 100,000 nucleated cells were analyzed after cell debris exclusion.

BLUE DETECT A (TARGET PEAK)

ORIGINAL ARTICLE

Leukocyte Differential by FCM: Comparison of Three Methods

ORIGINAL ARTICLE

Figure 1. CD45/SSC scattergram obtained with both cytometers and protocols. Gating strategies were previously described. A sample analyzed in parallel on FC500 and CANTO with the SWE, US and FR protocol is shown. Neutrophils are labeled in red, eosinophils in orange, basophils in black, monocytes in green and lymphocytes in blue. CD45 in on X-axis whereas SSC and SSC-A parameters respectively for FC500 and CANTO are on the Y-axis. Files were reanalyzed with the Kaluza software. [Color figure can be viewed in the online issue which is available at wileyonlinelibrary.com]

granulocytes (range from 2% to 36%), and 11 samples had a nRBC count from 1.25 to 12 out of 100 leukocytes. Finally, lymphocytosis was detected in 14 patients (range from 4 to 8.47 3 109 cells/L). Comparison Among Flow Cytometers All samples were run in parallel on instruments from Beckman Coulter (FC500) and Becton Dickinson (LSRII or CANTO). As an example, we show a CD45/SSC scattergram for a single patient analyzed with the SWE, US and FR protocols on a FC500 and a CANTO (Fig. 1). The first analytical focus was to identify bias between the flow cytometers (Fig. 2). We found no statistical difference (paired t-test) between the protocols when comparing nucleated cell count results obtained on the FC500 to those on the LSRII, either for normal or abnormal cells. The correlations were high for leuCytometry Part A  81A: 973 982, 2012

kocyte, neutrophils, eosinophils, basophils, lymphocytes, and monocytes (r correlation coefficient respectively at 0.976, 0.975, 0.975, 0.847, 0.966, and 0.972; Fig. 2). Similar results were also found when comparing FC500 to CANTO (data not shown). Finally, using Bland-Altman analysis, we found no systematic bias between cytometers (data not shown). Therefore we chose to use data from the FC500 at Rennes and from the LSRII at Seattle for further comparisons. Comparison of the Leukocyte Counts Between Protocols In a first attempt to evaluate the protocols, the leukocyte count obtained on cell counters (LH755 at Rennes and Sysmex XE2100 at Seattle) was compared to that obtained from the US, SWE and FR protocols (from FC500 at Rennes and LSRII at Seattle). All together, 92 samples were available for the three 977

ORIGINAL ARTICLE

Figure 2. Correlation between flow cytometers. The entire dataset from Seattle was analyzed simultaneously on FC500 and LSRII. Results obtained with the SWE, US and FR protocols on both cytometers are shown. Leukocyte count correspond to the sum of all subsets while neutrophils (neutro), eosinophils (eosino), basophils (baso), lymphocyte (lympho) and monocyte (mono) correspond to the cell count defined using Trucount (BD).

techniques. Leukocyte counts obtained by flow cytometry were the sum of all detected subsets. We found a Pearson correlation coefficient (r) of 0.987, 0.975, and 0.989 (Table 4). The slope was 0.994, 0.905, and 1, respectively, for US, SWE, and FR protocols (Table 4). Using Bland-Altman analysis, we found no systematic bias for the tested methods (Fig. 3). Upper and lower 95% limit of agreement for the difference between tested and reference methods were, respectively, 5.888, 3.561, and 3.5; and 24.325, 23.728, and 23.167 for the SWE, US and FR protocols. Besides this, we found good correlation between the protocols (r at 0.981, 0.989 and 0.989, respectively, for US vs. SWE, US vs. FR, and SWE vs. FR; Table 5). We conclude that good agreement between the methods for leukocyte counting was achieved. Comparison of the Standard and Abnormal Leukocyte Subsets Between Protocols We then evaluated the tested protocols regarding the standard leukocyte subsets (Fig. 4 and Table 4), comparing 978

results from flow cytometry and the reference manual count. Correlations were high for neutrophils, eosinophils, lymphocytes, and monocytes (Pearson correlation coefficient respectively [0.973, [0.912, [0.907 and [0.935 for all three protocols), but were lower for basophils (r range from 0.303 to 0.417). The relatively poor correlation is likely in large part due to the small number of basophils sampled and detected by manual microscopic counting, a factor that also likely plays a role with other less frequent cell populations such as blasts, immature granulocytes and nRBCs (see below). As an example, patients with 1 or 4 basophils counted on 400 cells had, respectively, clusters of 287 and 1437 events by FCM. Similarly, slopes were between 0.876 and 1.074 for neutrophils, eosinophils and monocytes; between 0.727 and 1.019 for lymphocytes and from 0.189 to 0.393 for basophils (Table 4). Using Bland-Altman analysis no systematic bias for the tested methods was found (Fig. 4). Comparison between protocols found good correlation (r [ 0.887) except for basophils (r range from 0.646 to 0.733; Table 5). Leukocyte Differential by FCM: Comparison of Three Methods

ORIGINAL ARTICLE Table 4. Descriptive analysis for the three immunocytometric methods compared to reference manual cell counts CELL TYPE

Leukocytes

Neutrophils

Eosinophils

Basophils

Lymphocytes

Monocytes

I Grans

Blasts

nRBCs

Table 5. Performance comparison among tested protocols CELL TYPE

PROTOCOL

CORRELATION COEFFICIENT

SLOPE

INTERCEPT

US SWE FR US SWE FR US SWE FR US SWE FR US SWE FR US SWE FR US SWE FR US SWE FR US SWE FR

0.987 0.975 0.989 0.978 0.984 0.973 0.936 0.935 0.912 0.303 0.413 0.347 0.950 0.907 0.921 0.934 0.925 0.947 0.572 0.356 0.558 0.671 0.256 0.533 0.474 0.325 0.401

0.994 0.905 1.000 0.915 0.926 0.876 0.941 1.019 0.885 0.213 0.393 0.189 1.019 0.727 0.834 1.074 0.892 0.972 0.693 0.764 0.876 1.085 0.809 1.052 1.088 0.844 0.815

0.150 0.350 20.167 0.119 20.195 20.073 0.053 0.042 0.036 0.018 0.028 0.022 0.015 0.213 0.126 0.090 0.151 0.087 0.048 20.053 0.003 20.003 0.020 0.037 0.021 0.027 0.016

A Pearson correlation (r) coefficient was used for leukocytes, neutrophils, eosinophils, basophils, monocytes and lymphocytes; whereas Spearman correlation (r) coefficient was used for nonGaussian populations (Immature granulocytes, blasts and nRBCs). Slope and intercept value from a Passing-Bablock analysis is shown.

The three protocols were then evaluated for enumeration of immature granulocytes, blasts and nRBCs (Fig. 5 and Table 4). The Spearman coefficient correlation (r) ranged from

Leukocytes Neutrophils Eosinophils Basophils Lymphocytes Monocytes I Grans Blasts nRBCs

US VERSUS SWE (r VALUE)

US VERSUS FR (r VALUE)

SWE VERSUS FR (r VALUE)

0.981 0.989 0.887 0.722 0.958 0.971 0.804 0.620 0.849

0.989 0.988 0.934 0.646 0.967 0.973 0.954 0.741 0.732

0.989 0.995 0.958 0.733 0.950 0.966 0.811 0.349 0.633

A Pearson correlation (r) coefficient was used for leukocytes. neutrophils. eosinophils. basophils. monocytes and lymphocytes; whereas Spearman correlation (r) coefficient was used for nonGaussian populations (Immature granulocytes, blasts and nRBCs).

0.356 to 0.572 for immature granulocytes; from 0.256 to 0.671 for blasts and for 0.325 to 0.475 for nRBCs. Slopes were between 0.693 and 0.876 for immature granulocytes, between 0.809 to 1.085 for blasts and from 0.815 to 1.088 for nRBCs (Table 4). The relatively poor correlation between morphology and flow cytometry for nRBCs appears to be due in part to the relatively small numbers of cells identified by both methods, as well as the relatively poor discrimination provided by the nucleic acid binding dye Syto16 for this population. Draq5 appears to provide better visual discrimination for nRBCs and the apparent reasonable correlation for nRBCs between methods is likely spurious due to the small number of samples with nRBCs evaluated. Notably, abnormal cells, i.e., blasts, immature granulocytes and nRBCs were correctly detected in all specimens in which they were identified microscopically. During this study, one acute promyelocytic leukemia was tested and promyelocytic blast cells were counted as immature granulocytes by all the three protocols.

DISCUSSION A CBC with leukocyte differential is provided by automated cell counters, but in 10-50% of the cases results need to

Figure 3. Correlation between cell counters and flow cytometry for leukocyte count. Bland and Altman difference plot were used to compare flow cytometry and cell counter for leukocyte count. Leukocyte counts obtained by flow cytometry were the sum of all detected subsets. Each dot represents one of the 92 samples, either from Rennes or Seattle. Difference between tested method (US, SWE, or FR) and reference method were plotted versus the average of both methods. Upper and lower 95% limits of agreement are represented (dashed line), as well as the line of identity (thin line). SWE: Swedish protocol, US: US protocol, FR: French protocol.

Cytometry Part A  81A: 973 982, 2012

979

ORIGINAL ARTICLE

Figure 4. Correlation between cell counter and flow cytometry for neutrophil, eosinophil, basophil, monocyte, and lymphocyte counts. Bland and Altman difference plot were used to compare flow cytometry and cell counter for neutrophils- (neutro), eosinophils- (eosino), basophils- (baso), lymphocyte- (lympho), and monocyte- (mono) count. Each dot represents one sample, either from Rennes or Seattle. Difference between tested method (US, SWE, or FR) and reference method were plotted versus the average of both methods. Upper and lower 95% limits of agreement are represented (dashed line), as well as the line of identity (thin line). SWE: Swedish protocol, US: US protocol, FR: French protocol.

be confirmed by a manual review of a blood smear (2,3). This scheme remains unchanged since the advent of cell counters during the second half of the twentieth century. In the last two decades, flow cytometry has been proposed as the reference method for the detection of various cells in blood, for instance, immature granulocytes, lymphocytes, monocytes and dendritic cells (11–15). However, all these studies were based on single cell type identification per antibody combination and were not suitable for a full differential count. 980

Recently, efforts have been made to develop relevant combinations of antibodies for a five part differential (7–10). This study reports the first systematic comparison of these candidates published protocols for the establishment of a reference method for extended leukocyte differential. During the writing of this manuscript a new interesting proposal has been published and reported on 100 normal and 102 abnormal samples (16). These authors proposed a 10 antibody 5 color cocktail using CD3, CD4, CD16, CD56, CD19, Leukocyte Differential by FCM: Comparison of Three Methods

ORIGINAL ARTICLE

Figure 5. Correlation between cell counters and flow cytometry for immature granulocytes, blasts and nRBCs counts. Bland and Altman difference plot were used to compare flow cytometry and cell counter for immature granulocytes- (I Grans), blasts- and nRNCs count. Each dot represents one sample, either from Rennes or Seattle. Difference between tested method (US, SWE or FR) and reference method were plotted versus the average of both methods. Upper and lower 95% limits of agreement are represented (dashed line), as well as the line of identity (thin line). SWE: Swedish protocol, US: US protocol, FR: French protocol.

CD14, CD45, CD34, CD71, and CD138 combined with a second tube dedicated to nRBCs identification with CD45 and DRAQ5. Excellent correlation was found between flow cytometry and manual count except for monocytes and basophils on normal blood samples. The main advantage of this proposal compared to those evaluated in our study is the detection of lymphocytes subsets (T-, B- and NK-); unfortunately this combination also has no markers for basophils and eosinophils. The SWE proposal has the advantage of a ‘‘1 tube, 6 antibodies, 1 DNA stain’’ combination, with the ability to discriminate platelets, as well as the ICSH requested cells. We found good and similar correlation to other protocols, except for immature granulocytes [r 5 0.356 vs. 0.572 (US) and 0.558 (FR)] and blasts [0.256 vs. 0.671 (US) and 0.533 (FR)], probably due to the lack of CD34 and CD117. The US protocol is the largest one with 2 tubes, 13 antibodies and 1 DNA stain. The advantage of this panel is the redundancy of markers allowing multiple gating strategies for difficult samples with antigenic modulation, the ability to identify miscellaneous cells including specific identification of basophils (see below), and markers of cell activation states such as CD64 expression on neutrophil, or HLA-DR on T-lymphocytes (17). The FR protocol is a ‘‘2 tube, 12 antibodies, 1 DNA stain’’ with the Cytometry Part A  81A: 973 982, 2012

advantage that the first tube has been extensively evaluated in multiple centers on thousands of samples and with a CEmarked product now used in some routine hematology laboratories. The use of CD2 is an advantage in case of T-lymphoblasts lacking surface CD3. A common disadvantage for both the SWE and the FR protocols is the use of CD36 that has been clearly shown to activate platelets and enhance their aggregation on leukocytes (neutrophils and to a lesser extend lymphocytes) leading to a theoretical overestimation of monocyte and underestimation of neutrophils (18). However in our hands, after reanalyzing random samples from our database, aggregation does not significantly modify the neutrophil or lymphocyte count (data not shown), even though modification of gating may be necessary. In addition to these findings, miscellaneous populations were detected by the different tested protocols. Concerning the lymphocyte subsets, the SWE protocol detected the sum of NKand cytotoxic T- cells; the US detected B-, cytotoxic T- and reactive T- cells; whereas the FR protocol detected B-, T non cytotoxic and the sum of NK- and cytotoxic T- cells. Concerning the blast subsets, the US protocol detected B- and myeloidsubtypes, whereas the FR detected B-, T- and myeloid subsets. Plasma cell and plasmacytoid dendritic cells were detected by the SWE and the US protocols. CD161 monocytes were 981

ORIGINAL ARTICLE detected by all the protocols. The ability of a protocol to reliably detect miscellaneous populations will facilitate studies to address the clinical relevance of such cells. To date, analysis of cell subsets (CD4/CD8 lymphocytes, dendritic cells, CD34 progenitors for instance) remains a specialized test, although their enumeration may be clinically relevant and have a direct impact on patient care if they were routinely available (19,20). In the current study, a limited set of patients and clinical conditions were analyzed. No normal samples were included, i.e., from healthy donors, as all samples were flagged for abnormal leukocyte by the automated hematology cell counter. This was intentional, as variants of all three methods have previously been shown to perform well on normal samples and represent a relatively trivial and irrelevant clinical situation (7–10). It is the identification of infrequent and abnormal cell populations that are requiring standardization of methodology and investigation of clinical utility. By flow cytometry, blasts are routinely detected at a low level in normal healthy samples that are not consistently seen by morphology. Indeed, we as well as Faucher et al. found a 95th percentile at, respectively, 0.06 and 0.012 3 109/L blasts in healthy samples (7,10). Given that the reference method used in this study is morphology, this study focuses on the correlation between morphology and flow cytometry. In the current study, any significant blasts (>1%) were also only detected by flow cytometry. This is the first multicenter study in the field to compare all the available protocols at the time of study design, including modifications to allow them to be used with both major commercial flow cytometers. As a result, each blood sample generated 10 FCS files to be analyzed and the study represents a substantial body of work. In summary, this was a pilot study where a multicenter evaluation has assessed published protocols for the leukocyte differential by FCM. While the results demonstrate and reiterate the feasibility of the immunodifferential by flow cytometry, each method had some minor disadvantages. We conclude that there is no clear winner in the accuracy and precision of the tested protocols. Several points have emerged: (i) the use of a DNA stain results in difficulties to standardize the protocols and a lack of reproducibility still not acceptable for a reference method; (ii) beyond the classical 5-part differential it seems clinically desirable to have access to enumeration of blood cell progenitors and B-, T- and NK- subpopulations. To address both of these concerns, we propose the use a dedicated reagent combination for nRBCs detection separate from that used for white cell subset enumeration; these to both minimize the problematic impact of the DNA binding dye and allow more detailed assessment of white cell subpopulations. Ideally, the leukocyte differential would use reference antibodies for each requested lineage including lymphocyte subsets and use markers with little variation in pathological conditions. It can also be concluded that to establish a reference method for nucleated cell differential counting with a 5 or 6

982

color method is still suboptimal and the use of a greater number of simultaneous fluorochromes will be required. Accordingly, a novel eight to ten color consensus method is now being evaluated that should meet the performance specifications set forth by ICSH.

ACKNOWLEDGMENTS The authors thank the International Council for Standardization of Haematology (ICSH) for funding part of this study by an unrestricted grant to Drs. Brent L. Wood and Mikael Roussel. The authors appreciate the assistance of Dr. Tony Avril, Centre Eugene Marquis, Rennes.

LITERATURE CITED 1. Barnes PW, McFadden SL, Machin SJ, Simson E. The international consensus group for hematology review: Suggested criteria for action following automated CBC and WBC differential analysis. Lab Hematol 2005;11:83–90. 2. Novis DA, Walsh M, Wilkinson D, St Louis M, Ben-Ezra J. Laboratory productivity and the rate of manual peripheral blood smear review: A College of American Pathologists Q-Probes study of 95,141 complete blood count determinations performed in 263 institutions. Arch Pathol Lab Med 2006;130:596–601. 3. Briggs C, Longair I, Slavik M, Thwaite K, Mills R, Thavaraja V, Foster A, Romanin D, Machin SJ. Can automated blood film analysis replace the manual differential? An evaluation of the CellaVision DM96 automated image analysis system. Int J Lab Hematol 2009;31:48–60. 4. Bain BJ. Diagnosis from the blood smear. N Engl J Med 2005;353:498–507. 5. Clinical and Laboratory Standards Institute. Reference leukocyte (WBC) differential count (proportional) and evaluation of instrument methods; Approved standard. 2007;27. 6. Rumke CL. Imprecision of ratio-derived differential leukocyte counts. Blood Cells 1985;11:311–4,315. 7. Roussel M, Benard C, Ly-Sunnaram B, Fest T. Refining the white blood cell differential: The first flow cytometry routine application. Cytometry A 2010;77A:552–563. 8. Cherian S, Levin G, Lo WY, Mauck M, Kuhn D, Lee C, Wood BL. Evaluation of an 8color flow cytometric reference method for white blood cell differential enumeration. Cytometry B Clin Cytom 2010;78B:319–328. 9. Bjo¨rnsson S, Wahlstro¨m S, Norstro¨m E, Bernevi I, O’neill U, Johansson E, Runstro¨m H, Simonsson P. Total nucleated cell differential for blood and bone marrow using a single tube in a five-color flow cytometer. Cytometry B Clin Cytom 2008;74B:91–103. 10. Faucher J-L, Lacronique-Gazaille C, Fre´bet E, Trimoreau F, Donnard M, Bordessoule D, Lacombe F, Feuillard J. ‘‘6 markers/5 colors’’ extended white blood cell differential by flow cytometry. Cytometry A 2007;71A(11):934–944. 11. Hubl W, Tlustos L, Erath A, Andert S, Bayer PM. Proposed reference method for peripheral-blood monocyte counting using fluorescence-labeled monoclonal antibodies. Cytometry 1996;26:69–74. 12. Fujimoto H, Sakata T, Hamaguchi Y, Shiga S, Tohyama K, Ichiyama S, Wang FS, Houwen B. Flow cytometric method for enumeration and classification of reactive immature granulocyte populations. Cytometry 2000;42:371–378. 13. Finn LS, Hall J, Xu M, Rutledge JC. Flow cytometric validation of automated differentials in pediatric patients. Lab Hematol 2004;10:112–118. 14. Della Bella S, Giannelli S, Taddeo A, Presicce P, Villa ML. Application of six-color flow cytometry for the assessment of dendritic cell responses in whole blood assays. J Immunol Methods 2008;339:153–164. 15. Giannelli S, Taddeo A, Presicce P, Villa ML, Della Bella S. A six-color flow cytometric assay for the analysis of peripheral blood dendritic cells. Cytometry B Clin Cytom 2008;74B:349–355. 16. Van De Geijn G-J, Van Rees V, Van Pul-Bom N, Birnie E, Janssen H, Pegels H, Beunis M, Njo T. Leukoflow: Multiparameter extended white blood cell differentiation for routine analysis by flow cytometry. Cytometry A 2011;79A:694–706. 17. Cid J, Aguinaco R, Sanchez R, Garcia-Pardo G, Llorente A. Neutrophil CD64 expression as marker of bacterial infection: A systematic review and meta-analysis. J Infect 2011;60:313–319. 18. Ouk C, Jayat-Vignoles C, Donnard M, Feuillard J. Both CD62 and CD162 antibodies prevent formation of CD36-dependent platelets, rosettes, and artefactual pseudoexpression of platelet markers on white blood cells: A study with ImageStream(R). Cytometry A 2011;79A:477–484. 19. Riccardi F, Della Porta MG, Rovati B, Casazza A, Radolovich D, De Amici M, Danova M, Langer M. Flow cytometric analysis of peripheral blood dendritic cells in patients with severe sepsis. Cytometry B Clin Cytom 2011;80B:14–21. 20. Saenz JJ, Izura JJ, Manrique A, Sala F, Gaminde I. Early prognosis in severe sepsis via analyzing the monocyte immunophenotype. Intensive Care Med 2001;27:970–977.

Leukocyte Differential by FCM: Comparison of Three Methods