Clin. Lab. Haem.

2000, 22, 345±350

New quantitative parameters on a recently
introduced automated blood cell counter ±
the XE 2100TM
C. BRIGGS, Department of Haematology, University College London Hospital,
P. HARRISON, London, UK
D. GRANT,
J. STAVES
S.J. MACHIN

Summary The XE 2100 (Sysmex Corporation) is a cell counter that furthers the technology of
¯uorescent ¯ow cytometry developed from the earlier range of Sysmex analysers. The
new diagnostic features are a nucleated red cell count (NRBC), the ability to measure
platelets by impedance as well as an `optical' platelet count using a ¯uorescence dye
and an immature granulocyte (IG) count.
The NRBC count was highly correlated (r ˆ 0.97) with the manual reference count.
For counts below 100 ´ 109/l the `optical' method and the immunocount gave good a
correlation (r ˆ 0.97) optical and impedance counts were also well correlated
(r ˆ 0.89). The use of the `optical' platelet count signi®cantly improves the reliability
of low platelet counts. The IG count correlated with visual counts (r ˆ 0.81) and
allows the detection of immature cells at an earlier stage in the laboratory process.
The introduction of ¯uorescent ¯ow cytometric analysis allows extended quanti®cation
of additional cell populations and so potentially improves screening and monitoring of
various pathological conditions.
Keywords Automated blood cell counter, nucleated red cell count, immature granulocyte count,
platelet counts, ¯ow cytometry

throughput and minimize cost per test by utilizing

Introduction
appropriate re¯ex testing.
Automated blood cell counters have changed dramatically The XE 2100 (Sysmex Corporation, Japan) is a recently
over the last 20 years. Reliable and accurate platelet introduced fully automated blood cell counter that utilizes
counts, a ®ve part white cell differential count and red cell new technology to extend the range of quantitative cell
indices including a reticulocyte count are all now analyses to include a nucleated red cell count, an
standard. In addition they provide a fast throughput (at immature granulocyte count (metamyelocytes, myelo-
least 100 samples per hour) and some potential for re¯ex cytes and promyelocytes) and an optical ¯uorescent
testing so that reagents/procedures that are more expen- platelet count in situations where an impedance count
sive are only performed when indicated. All instruments may be unreliable. The analyser is a development from
are also able to alert users to abnormal qualitative and earlier Sysmex systems, particularly the SE 9000 and
quantitative cell populations although many of the 1 9500 (Besson 1996; Pollard et al. 1999), the R series of
algorithms lack speci®city and/or are over sensitive. automated reticulocyte counters (Tichelli et al. 1990) and
The challenge for instrument manufacturers is to increase the SF 3000 (Korninger et al. 1998).
the range of quantitative cell analysis, yet to maintain The nucleated red blood cells (NRBC) are measured in a
speci®c NRBC channel. The cell membranes of the NRBC
Accepted for publication 19 May 2000 and red cells are lysed whilst the white cells are slightly
Correspondence: Carol Briggs, Department of Haematology, Univer-
perforated but remain intact, to allow quick in¯ux of a novel
sity College London Hospital, 25 Grafton Way, London WC1E 6AU,
UK. Fax: +44 207380 9886; E-mail: carolbriggs@hotmail.com propriety dye. Nuclear material is speci®cally stained and
the cells are hydrodynamically focused through a ¯ow cell 345

Ó 2000 Blackwell Science Limited

346 Automated blood cell counter quantitative parameters

cytometer. Fluorescence intensity and forward scatter light biomedical scientists. Immature granulocytes were de®ned
intensity analysis enables leucocytes and nucleated red as the sum of metamyelocyte, myelocyte and promyelo-
cells to be separated and counted. Leucocytes bind a larger cyte counts. Blast cells were not included in this count,
amount of the dye compared to the NRBC and conse- nor were stab cells or band cells, which were classi®ed as
quently show higher ¯uorescence intensity; the smaller neutrophils. NRBC were counted visually by two observers
volume of the NRBC nuclei gives a second differentiation viewing 200 WBC. This made the lower limit of detection
criterion and allows visible isolation of the NRBC cluster. 0.5 NRBC per 100 white cells. Initially 465 samples were
The impedance platelet count on the XE 2100 is used in for the study, 100 from apparently healthy normal
measured using the same technology as the SE 9500, adults, the rest pathological samples selected to cover a
with an appropriate ¯ag if the platelet histogram graph is wide range of haematological abnormalities. A further
outside preset limits. The `optical' platelet count is study of platelet counts was undertaken with an addi-
measured in the reticulocyte channel. A new patented tional 139 selected samples with thrombocytopenia,
polymethine ¯uorescent dye is used to stain the RNA/DNA known platelet disorders or potential interfering substan-
of reticulated red cells. It also stains the platelet membrane ces. These were analysed on the XE 2100 for both
and granules. In the ¯ow cell, each single cell is passed impedance and `optical' counts and compared to the
through the beam of a semiconductor diode laser, more platelet count measured by our proposed reference ¯ow
than 30 000 cells are counted for each sample, and this 3 cytometric RBC ratio method (Harrison et al. 2000). In
gives an optical ¯uorescent count. this technique, platelets are identi®ed immunologically
The immature granulocyte (IG) count is derived from with a monoclonal antibody, Anti-CD61-FITC (Beckton
the DIFF channel by ¯uorescence ¯ow cytometry using the 4 Dickinson). The platelet count is derived from the ratio of
same technology as the SE 9500. The combination of side ¯uorescent platelet events to collected red cell events
scatter (inner complexity of the cell), forward scatter measured in a ¯ow cell cytometer. The red cell count is
(volume) and ¯uorescence intensity of the nucleated cells derived from a calibrated impedance cell counter. This
characterizes each cell detected and of the different immunological platelet counting method has since been
leucocyte populations (clusters). Abnormal and immature used as the basis of a new international reference platelet
cells, with their larger nuclear volume show much higher counting assay under the auspices of the International
¯uorescence intensity than normal cells, and are distin- Committee of Standardization in Haematology (Harrison &
guishable in the DIFF scattergram. Ault 2000). Our original method has been slightly
The aim of this study was to evaluate the performance modi®ed and yet unpublished data will show that our
characteristics of the XE 2100 with particular attention to reference method and the new consensus reference
the new parameters. All standard full blood count, method show excellent correlation.
differential and reticulocyte parameters on the XE 2100
showed good correlation with the SE 9500 (Briggs et al.
Results
2000) so were not further examined in this study. The
new parameters were speci®cally compared in a selection
Comparison of visual NRBC count with the XE 2100
of normal samples and pathological material from the
daily workload on the routine laboratory SE 9500 with None of the 100 apparently healthy normal samples
RAM-1Ô, using manual ®lm inspection and other refer- contained NRBC on either the XE 2100 or by visual
ence techniques where appropriate. counting. One hundred and forty-nine pathological sam-
ples were selected by diagnosis as likely to have NRBC in
peripheral blood, these included sickle cell disease, thalas-
Materials and methods
saemia major, neonates, haematological or other malig-
The XE 2100 analyser tested at University College London nancies, HIV and malaria. Of these samples, 107 had
Hospital was a preproduction model with software version NRBC on the XE 2100 ranging from 0.1 to 855/100
9.0. All specimens were sampled in the open mode over a WBC. The lower limit of detection on the XE 2100 is 0.1
3-month period in the spring of 1999. The anticoagulant NRBC/100 WBC.
2 used was K3EDTA in vacutainers; (BD Haemoguard) and In two samples the XE 2100 did not detect NRBC when
all samples were processed within 4 h of venesection. they were seen on the ®lm. One sample was from a patient
A visual 400 cell WBC differential count was performed with myelodysplasia (1.5 NRBC/100 WBC) and the other
according to NCCLS H20-A protocol (1992). Two 200-cell from a neonate (1 NRBC/100 WBC). Twenty-nine (27%)
differentials were performed by at least two quali®ed had no detectable NRBC by blood ®lm examination. Two
Ó 2000 Blackwell Science Ltd., Clin. Lab. Haem., 22, 345±350
 C. Briggs et al. 347

sets of data had very high NRBC by both methods (200 samples. The platelet counts ranged from 5 to 1485 ´
and 900 NRBC/100 WBC). These were excluded from the 109/l. Each correlated with the other (r ˆ 0.99).
subsequent analysis to avoid undue bias. In the remaining One hundred and thirty-nine selected samples with
76 samples where NRBC were detected by the XE 2100 thrombocytopenia, known platelet disorders or potential
and on the blood ®lm the correlation between the two interfering substances were then analysed on the XE
methods was highly signi®cant (r ˆ 0.97). The Altman 2100 and by our ¯ow cytometric RBC ratio method.
Bland analysis of paired differences showed that the visual There was a close correlation between the counts by
count overestimated NRBC compared with the XE 1200 at the two methods (r ˆ 0.99) but in those with a platelet
high counts (Figure 1). count below 100 ´ 109/l there was a slightly less good
In nine samples with Plasmodium falciparum infection correlation (r ˆ 0.88) between the XE 2100 impedance
and one of Plasmodium vivax (parasitaemia counts of count and the reference immunocount. In some
< 0.1%±10.3%) NRBC were detected in four samples by samples (Table 1) there were signi®cant discrepancies
both the XE 2100 and the blood ®lm examination ®lm. between the platelet counts by impedance. If the
The performance of the NRBC ¯ag and action message platelet transfusion threshold was set at 20 ´ 109/l
generated when the analyser is run in CBC and DIFF mode (Figure 2) then the impedance count would have
only was assessed in an additional 62 samples in NRBC resulted in eight unnecessary transfusions and seven
mode to establish whether NRBCs were present and the would have been denied appropriate transfusions. If the
same samples then analysed in CBC and DIFF mode only. transfusion threshold were set at 10 ´ 109/l then only
The presence or absence of the NRBC ¯ag was then noted one patient would have been transfused unnecessarily
and compared with the NRBC count by both XE 1200 and and two patients under transfused. The `optical' platelet
blood ®lm estimates. Three samples (5%) gave a false count would not have resulted in any unnecessary
positive NRBC ¯ag which appeared to result from the transfusions (Figure 3) but at a transfusion threshold of
presence of large platelets. Six samples (9.5%) had false 20 ´ 109/l there would have been ®ve patients that
negative ¯ags even though NRBC were detected (0.5±4.0/ would not have received transfusions when the immu-
100 WBC) by both methods. nocount indicated that would have been appropriate.

Comparison of the SE9500 and XE 2100 impedance Comparison of the XE 2100 immature granulocyte
platelet count and XE 2100 `optical' platelet counts count with the visual reference count.

The XE 2100 impedance and `optical' platelet counts were A total of 465 samples, including 100 normal samples
compared with the SE 9500 impedance count in 415 were analysed on the XE 2100, in the DIFF mode, and by
a visual 200 cell differential count by at least two
observers. 278 out of these (60%) gave an IG count on
the XE 2100, range 0.01±5.1 ´ 109/l. The lower limit of

Table 1. Examples of impedance platelet counting errors with a

correct `optical' count when compared to the immunocount from
the RBC ratio

SE XE XE RBC
impedance impedance `optical' ratio
count count count count
Diagnosis (´ 109/l) (´ 109/l) (´ 109/l) (´ 109/l)

NHL 75 98 66 68
ITP 68 49 99 104
ITP 41 46 102 94
ITP 46 42 70 73
ITP 49 51 64 63
ITP 18 18 24 24
ITP N/A 61 88 101
Figure 1. The difference in NRBC between the XE 2100 and MDS N/A 10 16 15
visual counts versus the mean NRBC/100 WBC in 76 patholo- HIV 81 80 63 66
gical samples (Altman Bland plot).
Ó 2000 Blackwell Science Ltd., Clin. Lab. Haem., 22, 345±350
348 Automated blood cell counter quantitative parameters

Figure 2. The effect of XE 2100 im-

pedance platelet counts on platelet
transfusion decision making. Two differ-
ent transfusion thresholds are illustrated
at platelet counts of either 10 or
20 ´ 109/l. Overestimates and under-
estimates compared with the red cell ratio
estimate are indicated.

detection was apparently preset at 0.1%. 175 (38%) were

Discussion
in the range 0.1±0.4% and 103 (22%) were in the range
0.5±34.3% 129 of these samples (28%) gave a visual IG The aim of this study was to evaluate the performance
count range 0.01±3.94 ´ 109/l. The lower limit of characteristics of this new automated haematology ana-
detection was 0.5%. 21 of these 129 samples gave a lyser, and in particular examine the three new quantita-
visual IG count 0.04±1.05 ´ 109/l which was not detec- tive parameters, which are generally not all available from
ted by the XE 2100. other manufacturers. The NRBC count reported by the XE
In the 108 samples contained IG on both the XE 2100 2100 gives excellent correlation with the visual reference
and the visual counts there was a signi®cant correlation counts. Visual counts tend to give an overestimate when
between the two (r ˆ 0.81) but there was a bias towards NRBCs are increased. It is likely that the XE 2100 count is
on overestimate in the visual count at higher levels the more reliable and precise due to the greater number of
(Figure 4). cells being analysed. An accurate NRBC count should

Figure 3. The effect of XE 2100 `optical' platelet counts on

platelet transfusion decision making. Two different transfusion
thresholds are illustrated at platelet counts of either 10 or
20 ´ 109/l. Overestimates compared with the red cell ratio are Figure 4. The difference in immature granulocyte counts
indicated, there were no underestimates. between the XE 2100 and visual estimates (Altman Bland plot).
Ó 2000 Blackwell Science Ltd., Clin. Lab. Haem., 22, 345±350
 C. Briggs et al. 349

be important in the diagnosis and therapeutic monitor- analyser we evaluated neither of the two above features
ing of many clinical states such severe anaemias and were fully de®ned. New software is currently being
haemoglobinopathies particularly b thalassaemia major evaluated and it is hoped that this will improve both the
and sickle cell disease in haemolytic disease of the switching algorithm and platelet ¯ag from the impedance
newborn) and haematological and other malignancies count indicating when an `optical' count should be
where the number of NRBCs is often disproportionately performed.
high for the degree of anaemia. An automated NRBC A good correlation was observed between the XE 2100
should also be important to the busy routine haematology immature granulocyte and visual counts despite the low
laboratory to eliminate the need for laborious visual number of cells seen and the subjectivity of the identi®-
differential counts to be performed solely to correct the cation. It is anticipated that an automated immature
total white cell count. granulocyte count will help in monitoring leukaemoid
The NRBC ¯ag on the XE 2100 has been shown to be reactions, such as severe and chronic infections, in¯am-
effective at detecting their presence. This would allow the mation and tissue necrosis, neoplasms of all types and
routine running of the analyser to be run in CBC and DIFF myeloproliferative diseases. This will be helpful to those
mode. Samples generating the NRBC ¯ag would need to be laboratories that previously had to rely on immature
rerun with the analyser in NRBC mode. granulocyte ¯ags, which have notoriously high false
The XE 2100 impedance platelet count provides an positive/negative rates, to select samples for ®lm review.
acceptable count for the majority of clinical samples. A reliable immature granulocyte count will be detected at
However the limitations of impedance technique have an early stage in the laboratory processing without having
become increasingly recognized, particularly at low levels to wait for the preparation, examination and reporting of a
(0±50 ´ 109/l) and in pathological states such as TTP, manual blood ®lm.
ITP, myelodysplastic conditions, haemolytic anaemia, The introduction of ¯uorescent ¯ow cytometric analysis
other conditions with red cell fragmentation and aplastic in a routine cell counter allows quanti®cation of additional
bone marrow. When large platelets are present the cell populations and so potentially improves screening and
impedance count may exclude them and give a falsely monitoring of pathological conditions. With reliable
low count. Conversely where cell debris or red cell reporting of these new parameters there should be a
fragments are present the impedance count may include reduced number of visual differentials performed in a
these based on overlapping size and overestimate the laboratory, con®dence in an accurate platelet count in all
platelet count. This is particularly relevant in severely thrombocytopenic samples and clinical detection of
thrombocytopenic samples (Rowan 1991; Hammerstrom abnormal cells in a peripheral blood sample at an early
1992; Dickerhoff & von Ruecker 1995; Springer et al. pathological stage. Future analysers should allow even
1998). better quantitative differentiation of precise cell popula-
The XE 2100 `optical' platelet count shows excellent tions and make even more use of the ¯ow cytometric
correlation with the reference ¯ow cytometric method, capabilities now being developed.
particularly with platelet counts below 100 ´ 109/l. It is
far more reliable than the impedance count at this level
References
and this will allow more appropriate clinical decisions to
be made, particularly with regard to platelet transfusions. Ancliff P.J. & Machin S.J. (1998) Trigger factors for prophylactic
platelet transfusion. Blood Reviews. 12, 234±238.
This is especially so with platelet counts below 10 ´ 109/l,
Besson et al. (1996) Differential Leukocyte Performance Evalua-
where prophylactic platelet transfusions are regularly tion of the Sysmex SE-9000 Haematology Workstation.
required (Ancliff & Machin 1998). Furthermore improved Laboratory Haematology. 2, 22±25.
accuracy of platelet counts at this level may allow the Briggs C., Harrison P., Grant. D., Staves J., Chavda N. & Machin
platelet transfusion threshold to be reduced to 5 ´ 109/l S.J. (1999) Performance Evaluation of the Sysmex XE 2100Ô
without any increase in the risk of spontaneous bleeding an Automated Haematology Analyser. Sysmex Journal. 9, 113±
119.
(Harrison et al. 1999).
Dickerhoff R. & von Ruecker A. (1995) Enumeration of platelets
This assumes that the XE 2100 will always report the by multiparameter ¯ow cytometry using platelet speci®c anti-
most accurate platelet count on all samples, whether bodies and ¯uorescent platelet particles. Clinical & Laboratory
impedance or `optical', and even more importantly that Haematology. 17, 163±172.
when the analyser is not being run in reticulocyte mode Hammerstrom H. (1992) Spurious platelet counts in acute leu-
kemia with DIC due to cell fragmentation. Clinical & Laboratory
it generates the appropriate ¯ag to alert the user to
Haematology. 14, 239±243.
perform an `optical' platelet count. On the preproduction
Ó 2000 Blackwell Science Ltd., Clin. Lab. Haem., 22, 345±350
350 Automated blood cell counter quantitative parameters

Harrison P. & Ault K. (2000) An interlaboratory study of a count of the Sysmex SE-9500 to re®ne apheresis timing in
candidate reference method for platelet counting. American peripheral blood stem cells. British Journal of Haematology.
Journal of Clinical Pathology, in press. 106, 538±544.
Harrison P., Horton A., Grant D., Briggs C. & Machin S.J. (2000) Rowan R.M. (1991) Platelet counting and assessment of platelet
Immunoplatelet Counting. A Proposed New Reference Proce- function, In: (Koepke J.A. ed.) Practical Laboratory Haematology
dure. British Journal of Haematology. 108, 228±235. p.157. Churchill Livingstone, New York.
Korninger L., Mustafa G. & Schwarzinger I. (1998) The Hae- Springer W., von Ruecker A. & Dickerhoff R. (1998) Dif®culties
matology Analyser SF-3000: performance of the automated in Determining Prophylactic Transfusion Thresholds of Plate-
white cell differential count in comparison to the Haematology lets in Leukemia Patients. Blood. 92, 2183±2184.
Analyser NE-1500. Clinical & Laboratory Haematology. 20, Tichelli A., Gratwohl A., Driessen A. et al. (1990) Evaluation of
81±86. the Sysmex R-1000. An automated reticulocyte analyser.
Pollard Y., Watts M.J., Grant D., Chavda N., Linch D.C. & American Journal of Clinical Pathology. 93, 70±78.
Machin S.J. (1999) Use of the Haemopoietic progenitor cell

Ó 2000 Blackwell Science Ltd., Clin. Lab. Haem., 22, 345±350