Scandinavian Journal of Clinical & Laboratory Investigation, 2014; 74: 673–680

ORIGINAL ARTICLE

Automated white blood cell counts in cerebrospinal fluid using

the body fluid mode on the platform Sysmex XE-5000

AIHONG LI1, ELISABETH GRÖNLUND2 & GÖRAN BRATTSAND1

1Departmentof Medical Biosciences, Clinical Chemistry, and 2Department of Pathology, Umeå University Hospital,
Scand J Clin Lab Invest Downloaded from informahealthcare.com by Nyu Medical Center on 06/18/15

Umeå University, Umeå, Sweden

Abstract
Background. The Sysmex XE-5000 offers automated quantification of red blood cells and white blood cells (WBCs) in
body fluids, with differentiation of polymorphonuclear cells (PMNs) and mononuclear cells (MNCs). Methods. We
evaluated automated WBC counting in cerebrospinal fluid (CSF) using the body fluid mode on the Sysmex XE-5000,
comparing it with flow cytometry as the reference method, and also with manual counting by microscopy. Experimental
analysis for linearity and limit of detection was performed by diluting isolated WBCs in cell-free CSF. To study the ability
to discriminate between PMNs and MNCs, samples were spiked using MNCs separated from peripheral blood. Com-
parison of WBC counts between a counting chamber and the XE-5000 was performed for 198 CSF samples. Results. In
the experimental set-up, within-run (CV 19%) and between-day imprecision (CV 15.3%) in quantitating total number of
WBC on XE-5000 was acceptable for WBC counts ⱖ 25 ⫻ 106/L. Compared with expected cell counts, mean bias was
For personal use only.

⫹ 2.6% for flow cytometry, ⫹ 5.5% for XE-5000 and ⫺ 73.2% for manual counting. Differentiation between PMNs and
MNCs was in concordance with flow cytometry. In comparisons of clinical CSF samples, overall agreement between the
XE-5000 and manual counting was observed in 81% of the samples, but mean difference in WBC differentiation was higher
for PMN (51.1 ⫻ 106/L) than for MNC (7.95 ⫻ 106/L). Conclusion. Despite limited precision at low WBC counts, XE-5000
could be a favourable alternative to the labour-intensive, time-consuming and less reliable manual counting and cuts
turnaround times in routine CSF-based diagnosis.

Key Words: Leukocyte count, neutrophils, monocyte, cerebrospinal fluid, automated exam

Introduction would delay the reporting of result. Other disadvan-

In the majority of laboratories in Sweden, cells in
cerebrospinal fluid (CSF) have traditionally been
analyzed in Türk’s staining solution (Gentian violet/
acetic acid) by manual microscopic counting in a
Bürker chamber as described by Berggren Söderlund
et al. [1]. The method provides simultaneous differ-
ential counting of white blood cells (WBCs) without
cytospin preparation. In contrast, document H56-A
of the Clinical and Laboratory Standards Institute
(CLSI) recommends that erythrocytes and WBCs
should be counted unstained in a counting chamber,
while morphological examination of nucleated cells
is performed on air-dried and stained cytospins [2].
Morphological examination is no doubt better using
the CLSI procedure, but this tedious procedure
would be very difficult to use as a 24-h service, and
tages of microscopic evaluation are the high degree
of imprecision and the requirement for skilled
personnel [3–9]. Since differential counting of
WBCs in CSF into polymorphonuclear cells (PMNs)
and mononuclear (MNCs) may aid in clinical
decision-making regarding antibiotic therapy, it is
important to report CSF-WBC counts together with
MNC and PMN counts to provide valuable diagnos-
tic information in various medical conditions.
The XE-5000 offers a protocol for the quantifica-
tion of red blood cells (RBCs) and WBCs in body
fluid mode, and also differentiation between PMNs
and MNCs. Moreover, there is no need for pretreat-
ment of samples. Thus, the XE-5000 is an attractive
method for automated analysis of WBCs, PMNs and
MNCs in most body fluids. Only a few studies have

Correspondence: Aihong Li, M.D., Ph.D., Department of Medical Biosciences, Clinical Chemistry, Building 6M, ground floor, Umeå University Hospital,
Umeå University, SE-90185 Umeå, Sweden. Tel: ⫹ 46 (0) 90 7852507. Fax: ⫹ 46 (0) 90 778197. E-mail: aihong.li@medbio.umu.se

(Received 6 February 2014 ; accepted 26 June 2014 )

ISSN 0036-5513 print/ISSN 1502-7686 online © 2014 Informa Healthcare

DOI: 10.3109/00365513.2014.939994
 674 A. Li et al.
evaluated the body fluid mode on the XE-5000
[9–13]. Boer et al. were the first to demonstrate the
general good agreement between manual counting
and XE-5000 [10]. However, high imprecision at low
WBC counts (⬍ 20 cells/mm3) was observed. There-
fore, a manual count of CSF samples with WBCs ⬎ 5
and ⬍ 20 per μL was recommended to increase diag-
nostic accuracy [10]. De Jonge et al. also found a
high imprecision with WBC counts ⬍ 10 cells per μL
and poor discrimination of MNCs from PMNs [11].
Paris et al. studied several different types of body
fluids and found a good correlation between auto-
mated and manual counting methods for all types of
fluids except CSF [12]. These studies suggested that
XE-5000 can be suitable for automated quantifica-
Scand J Clin Lab Invest Downloaded from informahealthcare.com by Nyu Medical Center on 06/18/15
tion of WBCs in a defined diagnostic setting.
However, a recent study by the German Society
for Clinical Chemistry and Laboratory Medicine
(DGKL) evaluated cell analysis on Sysmex XT-4000i
and XE-5000 with 20 DGKL CSF controls [14].
Conformity of WBC counts was obtained with
Fuchs-Rosenthal chamber as the reference. How-
ever, significantly increased PMN counts (2.2-fold
on average) were demonstrated with Sysmex
XT-4000i and XE-5000 compared to flow cytome-
try, indicating that WBC differentiation into MNC
For personal use only.
and PMN is unreliable [14]. Thus, the XE-5000
body fluid system has not been recommended as a
suitable alternative to manual differential cytologic
workup for CSF [14]. Results of examination with
the XE-5000 in 65 CSF samples with tumor cells
and 126 CSF samples without tumour cells sug-
gested that measurement of high-fluorescence cells
is not an appropriate diagnostic test for intrathecal
tumor cells [15]. The disparity between these previ-
ous reports indicates that more requires to be known
regarding interpretation of results for routine usage.
In the present study, with the combined use of
manual counting and flow cytometry, which was
used as the reference method, we evaluated the
performance of the body fluid mode on the Sysmex
XE-5000. Use of an experimental set-up and also
clinical samples showed that the XE-5000 could be
a favourable alternative for routine quantification of
WBCs, PMNs and MNCs in CSF.
Materials and methods
Sample preparation
WBCs were obtained using RBC lysis buffer from
whole blood samples newly taken from healthy
donors. We randomly selected WBCs in the range of
5–9 ⫻ 109 cells/L which consisted of approximately
60% PMNs and 40% lymphocytes/monocytes. One
millilitre whole blood was mixed gently for 5–10 min
with 50 mL lysis buffer (8.26 g NH4Cl, 1 g KHCO3,
and 33.6 mg EDTA in 1 L deionized water), followed
by centrifugation at 400 g for 5 min. Cell pellets were
washed twice in phosphate-buffered saline (PBS)
and resuspended in PBS. The cell suspensions in
different experiments had WBC counts in the range
2–6 ⫻ 109 /L as measured in the normal mode of the
XE-5000.
Normal CSF routine diagnostic samples (RBC
⬍ 100 ⫻ 106/L, PMN ⬍ 1 ⫻ 106/L and MNC ⬍ 5 ⫻ 106
/L), with no turbidity, were selected for preparing
cell-free CSF. Supernatant after centrifugation
(2000 g 5 min) was transferred into a new tube for
each sample and kept at ⫺ 20°C until use. A pool
of cell-free CSF was prepared by mixing about
30 supernatants for further use of dilution.
Between May of 2009 and April of 2010, 198
CSF samples were examined for routine diagnostic
purposes by WBC counting using the XE-5000 and
the manual counting method. At least 1 mL CSF was
collected in plain sterile tubes and the samples were
requested to arrive in the laboratory within 30 min
after collection. After arrival in the laboratory, all
automated and manual analyses were completed as
soon as possible preferably within 1 h. First, cell
counts were performed by the manual method as
described below. Then, if enough material was left
over (130 μL), automated cell counts were performed
in the open body fluid mode of a Sysmex XE-5000
haematology analyzer.
Results of WBC counts from patient CSF sam-
ples were divided into three subgroups including a
range of normal clinical reference (0–5 ⫻ 106
WBCs/L), low WBC count (6–20 ⫻ 106 WBCs/L)
and high WBC count (⬎ 20 ⫻ 106 WBCs/L). Overall
agreement between manual counting and XE-5000
was calculated with consideration of above reference
cut off level (WBC ⬎ 5 ⫻ 106 WBCs/L).
Microscopic analysis
Microscopic cell counting was performed by an
experienced technician using a Bürker counting
chamber (Paul Marienfeld GmbH & Co. KG Lauda-
königshofen, Germany) in the case of the perfor-
mance experiments.
Cell counts in clinical samples were performed in
clinical routine by technicians who have received
special training but had different degrees of experi-
ence. For each sample, 10 microscopic fields
(A-square) were examined with a ⫻ 40 objective lens.
50 μL CSF and 50 μL Türk’s solution (Merck,
Darmstadt Germany) were mixed and 10 μL of the
mixture was analyzed in a counting chamber. Ten
fields were evaluated for PMNs and MNCs. It often
takes at least 10 min for one sample.
Automated measuring on the XE-5000
The Sysmex XE-5000 haematology analyser (Sysmex
Corporation, Kobe, Japan) incorporates a body fluid
mode that allows instant quantification of WBCs and

other highly fluorescent cells in CSF, pleural fluid,
ascites, and synovial fluid without any treatment or
pretreatment.When switching from ‘full blood mode’,
the system performs background checks and a sepa-
rate washing step to avoid cross-contamination. It
takes approximately 195 s for one CSF sample on
the XE-5000. At least 130 μL CSF was required for
analysis. All samples were assessed in the open mode
in the system’s DIFF channel.
Flow cytometric analysis
Flow cytometric analysis was performed on a
FACSCanto II Flow Cytometer (BD Biosciences,
San José, CA, USA). In this analysis, we used a
Scand J Clin Lab Invest Downloaded from informahealthcare.com by Nyu Medical Center on 06/18/15
special test tube: Trucount (BD Biosciences). These
tubes contain a known number of particles, which
acts as a reference to determine the absolute number
of residual leukocytes in blood components. The
number of leukocytes was determined from a calcu-
lation based on the number of particles and the
sample volume. One hundred microliters of CSF was
required for analysis. Differentiation between PMNs
and MNCs on the FACSCalibur (BD Biosciences)
was analyzed using CD45/SSC gating according to
For personal use only.
Kleine et al. [16].
Performance studies
Limit of blank (LoB). The LoB was established in the
background analysis by measuring 20 different
samples of cell-free CSF, twice for each sample.
Functional sensitivity. To evaluate within-run impreci-
sion, we diluted WBCs into cell-free CSF at the
expected counts of 5, 25 and 50 ⫻ 106 WBCs/L.
WBC counting was determined on the XE-5000 by
repeatedly measuring samples 20 times and the CVs
at the expected counts were calculated. Manual
counting was also performed but samples could be
only evaluated 3–7 times at each of the three levels
due to repeatedly continuous microscopying, the
examiner was unable to properly evaluate the subse-
quent samples.
To test between-day accuracy, we performed
analysis by diluting the Sysmex control (e-CHECK)
in saline solution at an expected count of 30 ⫻ 106
WBCs/L and measuring WBC counts in the open
body fluid mode on the XE-5000 over 19 consecu-
tive days.
Linearity. WBCs isolated from blood were diluted in
a pool of cell-free CSF at eight different expected cell
counts between 5 ⫻ 106/L and 5000 ⫻ 106/L (5, 10,
25, 50, 250, 500, 1000 and 5000 ⫻ 106/L). WBC
counts were evaluated on the XE-5000, the FACS-
Canto and manual counting. The expected WBC
counts were plotted against the measured cell counts
Automated WBC counts in CSF on XE-5000 675
on FACSCanto, XE-5000 and manual counting.
Because of the short half- life of the leukocytes we
also tested a later time point at 3 h in the experimen-
tal set-up to find whether there are differences of
WBC counts at different levels between these two
time points.
WBC differentiation. Peripheral blood (PB) from a
healthy donor was used for the experiments.
WBCs were prepared as described above and diluted
at four expected WBC counts (10, 20, 100 and
1000 ⫻ 106/L). Cell analysis was performed on the
XE-5000, flow cytometry and by manual counting.
To further study differentiation between PMNs
and MNCs, mononuclear cells were separated by
centrifugation on a density gradient (Lymphoprep,
Pharmacia, Uppsala, Sweden) and washed twice
with PBS, and resuspended in PBS before dilution.
These cells were further diluted into a pool of
cell-free CSF. To diminish the effect of pH on cell
stability, a sterile-filtered cell-free CSF pool (pH 7.2)
was prepared by incubation at 37°C overnight in 5%
CO2. Four expected WBC concentrations (5, 10, 50
and 500 ⫻ 106/L) were prepared for cell analysis
using the XE-5000, flow cytometry and manual
counting.
Statistical analysis
Bias was calculated as [(mean obtained values ⫺
expected values)/expected values] ⫻ 100. For calcu-
lation of imprecision, the Excel 2003 was used for
data analysis. For linearity, Passing-Bablok regresson
anlysis was used and data were plotted using the
Method Validator (version 1.1). In comparison
between two methods in clinical CSF samples, mean
difference was obtained using Method Validator
(version 1.1). The Spearman correlation test was
used for correlation between two methods by SPSS
(version 18).
Results
Limit of blank (LoB)
In the background analysis, 20 cell-free CSF samples
were evaluated on the XE-5000. In three of 20 sam-
ples, 1 ⫻ 106 PMNs/L were detected, whereas the
remaining samples had no WBCs. This corresponds
to LoB of ⬍ 2 ⫻ 106/L for WBCs.
Functional sensitivity
Figure 1 shows the within-run imprecision on
XE-5000 in quantitating total number of WBCs at
three expected cell counts, with a CV of 32% at
5 ⫻ 106 WBCs/L, 19% at 25 ⫻ 106 WBCs/L, and
15.8% at 50 ⫻ 106 WBCs/L. Thus, the functional
 676 A. Li et al.
sensitivity on XE-5000, defined as the lowest level
with total CV of less than 20%, was ⱖ 25 ⫻ 106
WBCs/L.
The total WBC counts by manual counting
showed lower numbers of WBCs compared with
expected cell counts and the results from the
XE-5000. High imprecision was shown by manual
counting, with a CV of 35% for 5 ⫻ 106 WBCs/L,
38% for 25 ⫻ 106 WBCs/L and 40% for 50 ⫻ 106
WBCs/L.
Between-day accuracy was tested by diluting the
Sysmex control (e-CHECK) at an expected count of
30 ⫻ 106 WBCs/L. Median WBC count on the
XE-5000 was 30 ⫻ 106/L (range 24–39 ⫻ 106/L)
and the mean count was 31 ⫻ 106/L. Between-day
Scand J Clin Lab Invest Downloaded from informahealthcare.com by Nyu Medical Center on 06/18/15
imprecision was 15.3%.
Linearity
Linearity was tested at expected concentrations of
WBCs from 5 ⫻ 106/L to 5000 ⫻ 106/L. There
were good agreements between expected WBCs
and measured cells on FACSCanto (y ⫽ 0.86x ⫹ 3.9,
R2 ⫽ 0.998) and XE-5000 (y ⫽ 0.92x ⫹ 0.2,
R2 ⫽ 0.996) regarding total number of WBCs
For personal use only.
(Figure 2A, 2B). The mean bias was ⫹ 2.6% for
FACSCanto and ⫹ 5.5% for XE-5000 compared to
expected cell counts. However, manual counting
generally gave lower counts than expected and the
cell counts from the XE-5000 and FACSCanto. A
low slope value by Passing-Bablok regression analysis
was shown in Figure 2C (y ⫽ 0.37x – 2.0, R2 ⫽ 0.988).
At low level of WBC counts ⬍ 50 ⫻ 106/L, manual
counting failed to detect any cells (Figure 2C). The
mean bias was ⫺ 73.2% for manual counting.
In the experimental set-up, we found that the
total number of WBCs was lower at 3 h than at 0 h
on the XE-5000 and FACSCanto (Figure 2D and
2E). The mean bias was ⫺ 7.4% for FACSCanto
and ⫺ 34.4% for XE-5000. Interestingly, manual
counting showed slightly higher WBC counts
(slope ⫽ 0.39, Figure 2F) at this time point than at
Figure 1. Within-run imprecision for WBC counts on the XE-5000
in the experimental set-up. WBCs isolated from peripheral blood
were diluted at three expected counts of 5, 25, and 50 ⫻ 106
WBCs/L.
0 h (slope ⫽ 0.37). The mean bias was ⫺ 52.9% for
manual counting.
Differentiation between MNCs and PMNs
Table I summarizes the results of WBC differentia-
tion on the FACSCanto, XE-5000 and manual
counting. FACSCanto and XE-5000 showed lower
bias for MNC count in both PB dilutions and
isolated MNC dilutions than manual counting when
compared with the expected MNC counts. A high
bias, especially at low concentrations of MNC counts,
was found when microscopic analysis was applied.
Manual counting improved at expected MNC count
of 46 ⫻ 106/L.
In PB dilutions, PMN analysis on the FACS-
Canto showed preferable bias (ⱕ 11%) at all four
expected PMN counts tested. On the XE-5000, high
bias (33%) was shown only at low expected PMN
count (12.8 ⫻ 106/L). Bias varies from 25–56% in an
expected range of PMN counts (6.4–64 ⫻ 106/L)
when manual counting was used. In isolated MNC
samples, PMN analysis showed high bias in the
majority of PMN counts despite of these methods
used in the experimental set-up. However, the main
purpose using isolated MNC from blood was to test
accuracy of MNC analysis. Figure 3 shows a scatter
plot from the XE-5000 (panels A and B) and from
the FACSCanto (panels C and D) of the MNC and
PMN cell populations before and after purification
of MNCs.
The results raised the question of whether it
might be easier to miscount PMNs at low quantities
of WBCs by microscopic analysis.
Comparison of patient samples
Comparison of results from chamber counting and
from the XE-5000 was performed in 198 CSF
samples. WBC counts was significantly correlated
between manual counting and XE-5000 (r ⫽ 0.833,
p ⬍ 0.001). Mean difference was 40.8 ⫻ 106/L
(range 14.5–67.1). WBC count by the XE-5000 was
classified to the same subgroup as manual counting
in 161 out of 198 samples (81%) as shown in
Table II. We identified 37 samples with discordant
results. Seven samples showed lower WBC counts on
the XE-5000 than by manual counting, whereas
30 samples had elevated WBC counts on the
XE-5000.
Cell analysis in WBC differentiation between
manual counting and XE-5000 in clinical CSF
samples showed that mean difference was higher
for PMNs (51.1 ⫻ 106/L, range from ⫺ 11.6 to 114)
than for MNC (7.95 ⫻ 106/L, range from ⫺ 14.6 to
30.5, data not shown).
 Automated WBC counts in CSF on XE-5000 677
Scand J Clin Lab Invest Downloaded from informahealthcare.com by Nyu Medical Center on 06/18/15

Figure 2. Linearity of WBC counts on the FACSCanto, the XE-5000 and manual counting. Agreement between expected WBC count
and measured cell count is shown by scatter plot (Passing-Bablok regression analysis). The identity line (y ⫽ x) is drawn. The x-axis
represents expected WBC counts. The y-axis represents measured cell counts. Cells were immediately analysed after dilution on the
FACSCanto (A), the XE-5000 (B), and by manual counting (C). Cell counts were performed 3 h after dilution on the FACSCanto (D),
the XE-5000 (E), and by manual counting (F). Diluted cells were kept at room temperature.
For personal use only.

Discussion differentiation between MNCs and PMNs, com-

In the present study, we found that the open body pared to manual counting. In experiments we found:
fluid mode of the Sysmex XE-5000 was a favourable (1) that within-run and between-day accuracy was
method for determination of WBC counts and for acceptable at WBC counts of ⱖ 25 ⫻ 106/L, (2) that

Figure 3. Scatter plot of cell populations before and after separation of mononuclear cells. Differential counts of PMNs and MNCs on
the XE-5000 before isolation of MNCs from peripheral blood (panel (A) normal mode on DIFF channel) and after isolation of MNCs
from peripheral blood (panel (B) body fluid mode on DIFF channel). Differential counts on the FACSCanto before (C) and after (D)
isolation of MNCs from peripheral blood. R2 represents the population of lymphocytes, R3, represents the population of monocytes and
R4 represents the population of granulocytes. This Figure is reproduced in color in the online version of The Scandinavian Journal of
Clinical & Laboratory Investigation.
 678 A. Li et al.

Table I. WBC differentiation by FACSCanto, XE-5000 and manual counting in the experimental set-up.

MNCs ⫻ 106/L (bias %) PMNs ⫻ 106/L (bias %)

Expected Expected
MNCs ⫻ 106/L FACSCanto XE-5000 Manual PMNs ⫻ 106 /L FACSCanto XE-5000 Manual
PB*
3.6 4.2 (17) 3.5 (⫺ 3) 0 (⫺ 100) 6.4 6.8 (6) 8 (25) 8 (25)
7.2 7.3 (1) 6.5 (⫺ 9) 0 (⫺ 100) 12.8 12.9 (1) 17 (33) 20 (56)
36 37.9 (5) 32 (⫺ 11) 24 (⫺ 33) 64 67.3 (5) 55 (⫺ 13) 90 (41)
360 366.5 (2) 337 (⫺ 6) 340 (⫺ 6) 640 711.5 (11) 646 (1) 594 (⫺ 7)

MNCs#
4.6 4.8 (4) 5.5 (20) 0 (⫺ 100) 0.4 0.7 (75) 5 (1150) 0 (⫺ 100)
9.2 7.9 (⫺ 14) 8.5 (⫺ 8) 0 (⫺ 100) 0.8 0.7 (⫺ 13) 1.5 (88) 0 (⫺ 100)
46 46.4 (1) 42 (⫺ 9) 50 (9) 4 0.5 (⫺ 88) 4 (0) 10 (150)
460 450 (⫺ 2) 395 (⫺ 14) 456 (⫺ 1) 40 4.5 (⫺ 89) 24 (⫺ 40) 34 (⫺ 15)

MNCs, mononuclear cells; PMNs, polymorphonuclear cells. *PB, peripheral blood before dilution had 6.1 ⫻ 109 WBCs/L: 64% PMNs
Scand J Clin Lab Invest Downloaded from informahealthcare.com by Nyu Medical Center on 06/18/15

and 36% MNCs. The expected MNCs were obtained by diluting PB in cell-free CSF. #MNCs, mononuclear cells including monocytes
and lymphocytes were separated from the PB above using Lymphoprep and had 895 ⫻ 106 WBCs/L: 8% PMNs and 92% MNCs after
separation. The isolated MNCs were diluted in cell-free CSF.

linearity showed a good agreement between expected
number of cells and measured cell counts, and (3)
that differentiation between MNCs and PMNs was
in good concordance with the reference method
(flow cytometry) except at low PMN counts. In com-
parisons of clinical CSF samples, there was good
For personal use only.
sensitivity for the manual counting method was not
investigated systematically, preliminary results
showed high imprecisions (CV varied from 35–40%)
at all three expected WBC counts tested in this study.
In a recent report, a slightly lower CV of 12.5% for
the XE-5000 compared with 15.2% for the manual

overall agreement between the XE-5000 and manual
counting in the majority of cases. In discordant sam-
ples, WBC counts were often higher on the XE-5000
than by manual counting.
Requirements for WBC counts are based on the
lowest result with clinical significance. According to
CLSI H56-A, this was in the range of 6–10 ⫻ 106/L in
adult lumbar CSF samples [2], which has been widely
used by laboratories and accepted in clinical practice.
In this study, we found that the LoB (defined as ⫹ 2
SD of blank) for WBC (MNCs/PMNs) on the
XE-5000 was ⬍ 2 ⫻ 106/L. Linearity for WBC count-
ing in the experimental system covered the expected
cell counts in the majority of clinical CSF samples. An
acceptable agreement was found in most experiments
between the expected number of WBCs and the
results from the XE-5000 and FACSCanto, while
manual counting in general gave fewer WBCs. We
believe that our method of manual counting may suf-
fer from an unpredictable amount of under-recovery.
Data on the analysis of within-run accuracy
revealed an acceptable CV (19%) at a WBC count
of 25 ⫻ 106/L, but poor precision (CV ⫽ 32%) at
a WBC count of 5 ⫻ 106/L. Although functional
Table II. Agreement between manual and automated (XE-5000)
WBC counting in 198 clinical cerebrospinal fluid samples.
0–5 ⫻ 106 6–20 ⫻ 106 ⬎ 20 ⫻ 106
Manual/XE5000 n WBCs/L WBCs/L
0–5 ⫻ 106 WBCs/L 121 96 25
6–20 ⫻ 106 WBCs/L 25 6 14
⬎ 20 ⫻ 106 WBCs/L 52 0 1
Total 198 102 40
Fuchs-Rosenthal chamber was demonstrated [13].
However, high imprecision with WBC counts of
⬍ 10/μL on the XE-5000 was reported [15]. Boer
et al. reported total precision on the XE-5000 (a CV
of 14.9% for a mean WBC count of 37/mm3) [10].
In this study, we found a similar degree of between-
day imprecision of 15.3% for a mean WBC count of
31 ⫻ 106/L.
With consideration on stability of WBCs in CSF
we tested WBCs diluted in CSF at room temperature
from 0–6 h and found a tendency of a decrease in
total number of WBCs mainly due to loss of PMNs
(data not shown). In the experimental set-up, we
showed decreases of WBC counts after 3 h. However,
our experimental set-up of the stability test is of lim-
ited value because patient samples would be prefer-
ably used. One previous study found that the decay
in WBCs in CSF reached 40% after 2 h at room
temperature [17]. In another study, after 90 min,
lymphocyte counts were reduced by 65% while decay
of monocytes and neutrophils was more rapid and
reached 90% [18]. The data collected indicate that
the immediate analysis of CSF is important to pro-
vide accurate cell counts and differentiation of WBCs,
MNCs, and PMNs.
Accuracy in discrimination between MNCs and
PMNs by XE-5000 has not been critically evaluated
previously. Boer et al. described that WBC counting
WBCs/L
on the XE-5000 provided the highest rate of ‘correct
0 negatives’ (or the lowest rate of ‘false positives’) of
5 all analyzers reported so far [10]. However, they rec-
51
ommended manually counting of all CSF samples
56
with WBC counts of ⬎ 5 and ⬍ 20 ⫻ 106/L due to the

high degree of imprecision at low WBC counts
(⬍ 20 ⫻ 106 WBCs/L). Our findings in the experi-
mental set-up showed that our manual method of
differentiation between MNCs and PMNs in this
range of WBC count could not improve the accuracy.
Furthermore, differential counting by microscopic
analysis in clinical CSF samples agree our experi-
mental finding. Paris et al. also found that there was
a poor correlation in WBC differential count between
the automated XE-5000 method and the manual
method [12]. de Jonge et al. reported that for PMNs
below 20 ⫻ 106 /L, agreement was poor with a large
positive bias using the XE-5000 compared with man-
ual counting [11]. However, good agreement was
shown with higher counts of WBCs, PMNs and
Scand J Clin Lab Invest Downloaded from informahealthcare.com by Nyu Medical Center on 06/18/15
MNCs (⬎ 20 ⫻ 106 WBCs/L). Positive bias was only
seen for the lower WBC counts (⬍ 20 ⫻ 106/L) [11].
This may be explained by some debris or fragments
counted as PMNs. Thus, it was not recommended
that differential analysis (PMNs, MNCs) should be
reported to the clinicians if the WBC count was
below 10 ⫻ 106 WBCs/L [11]. Regarding cell counts
obtained with flow cytometry, some low MNC counts
and some high PMN counts were observed on the
Sysmex XT-4000i and the XE-5000 in 20 DGKL
CSF controls [14]. Furthermore, poor likelihood
For personal use only.
ratios, both positive (3.0) and negative (0.29), were
recently demonstrated in detection of tumour cells
in CSF by the XE-5000, suggesting that evaluation
by microscope seems mandatory [15]. Our experi-
ments support that microscopic analysis is necessary
in distinguishing between malignant and non-malig-
nant cells, because a high bias between expected and
automated MNC and PMN counts on the XE-5000
was found at low WBC counts below 10 ⫻ 106/L.
Manual microscope examination, preferably using
stained cytospin preparation, is no doubt required
for malignant CSF samples.
When comparing CSF samples from patients, a
recent study found that agreement between the man-
ual method and the XE-5000 method reached 95%
for the lowest level of WBCs (ranging from 0–10
cells/μL) and it was 100% for highest level (⬎ 200
WBCs/μL) [19]. However, MNC counts in the low-
est category was overestimated by the automated
method [19]. We observed an overall agreement
between manual counting and counts from the
XE-5000 in 81% of CSF clinical samples. We also
observed a high frequency of elevated WBC counts
by automated counting compared to manual count-
ing. The discordant WBC counts may have been due
to the explanations given above, and to the perfor-
mance of manual counting by several technicians,
some of whom were very experienced and some of
whom were less experienced. The lack of agreement
may also have been a result of the time delay between
manual counting and analysis on the XE-5000, which
possibly affected cell sedimentation or cell survival.
WBC counts (MNCs, PMNs) on the XE-5000 often
Automated WBC counts in CSF on XE-5000 679
tended to be higher than by manual counting.
This is in agreement with the observation of under-
recovery with manual counting seen in the experi-
mental set-up.
The reason for our manual counting method
apparently showing under-recovery is unclear. We
believe that cells may stick to the walls of tips or
tubes in the process of sample dilution in Türk’s
solution and transfer to the counting chamber. Why
recovery for the manual method in the experimental
set-up appears to have been better in samples
after 3 h of storage is not easy to explain but we
speculate that the deteriorating cells in these samples,
particularly PMNs, may be less prone to activation,
aggregation and stickiness. One limitation of our
study is that we analyzed artificial samples. Cells in
patient samples are neutrophils, lymphocytes and
monocytes from blood. Manipulation of leukocytes
in RBC lysis buffer and transfer to cell-free CSF
may, however, affect cell integrity, cell adherence,
and recovery. The buffering capacity of cell-free CSF
is low, and the pH will increase as CO2 in solution
evaporates [20], something that may affect recovery.
On the other hand, if this kind of handling really
does affect recovery, the robustness of the manual
counting method must be called into question.
Automated analysis of a single CSF sample on
the XE-5000 only takes approximately 195 s,
including switching mode and washing steps. A
recent study found that CSF cell analysis using the
XE-5000 gives a 7.5-fold reduction in turnaround
time compared to counting with the Fuchs-
Rosenthal counting chamber [13]. This may be an
important additional advantage for clinicians who
often have to make therapeutic decisions as quickly
as possible.
In summary, the results of our experiments and
comparisons in clinical CSF samples indicate that
the XE-5000 may offer a relatively robust and accu-
rate method for automated routine quantification of
WBCs, MNCs and PMNs in CSF in a diagnostic
setting. Despite its limitations, the accuracy of
XE-5000 is superior to that of the manual counting
method.
Acknowledgements
We thank Anita Hurula and all the technicians at the
Department of Clinical Chemistry, Umeå University
Hospital, Sweden, for performing the analyses.
Declaration of interest: The authors report no
conflict of interest. The authors alone are responsible
for the content and writing of the paper.
This study was supported by grants from the
County Council of Västerbotten, Umeå, Sweden
(ALF) Grant number (ALF no 7000468).
 680 A. Li et al.
References
[1] Berggren Söderlund M, Blom A, Carlson J, Dahlbäck B,
Hansson L, Harper P, Hillarp A, Hultdin J, Jeppsson J,
Ljungdahl L, Malm J, Sterner G, Tryding N, Vinge E,
Eriksson S, Christensson A, van Hage M. Akuta
likvoranalyser. Laurells Klinisk Kemi i praktisk medicin
2003;8:e:364–5.
[2] Szamosi DI, Bautista JM, Cornbleet J, Glasser L, Rothe G,
Sandhaus L, Key M, Meloni-Ehrig A, Culp NB,
Dougherty W. Body fluid analysis for cellular composition;
Approved Guideline. H56-A. 2006;26.
[3] Aune MW, Sandberg S. Automated counting of white
and red blood cells in the cerebrospinal fluid. Clin Lab
Haematol 2000;22:203–10.
[4] Mahieu S, Vertessen F, Van der Planken M. Evaluation of
ADVIA 120 CSF assay (Bayer) vs. chamber counting of
cerebrospinal fluid specimens. Clin Lab Haematol 2004;
Scand J Clin Lab Invest Downloaded from informahealthcare.com by Nyu Medical Center on 06/18/15
26:195–9.
[5] Salinas M, Rosas J, Iborra J, Manero H, Pascual E.
Comparison of manual and automated cell counts in EDTA
preserved synovial fluids. Storage has little influence on the
results. Ann Rheum Dis 1997;56:622–6.
[6] Andrews J, Setran E, McDonnel L, Kussick S, Wood BL,
Sabath DE. An evaluation of the cell-dyn 3200 for counting
cells in cerebrospinal and other body fluids. Lab Hematol
2005;11:98–106.
[7] Aulesa C, Mainar I, Prieto M, Cobos N, Galimany R. Use
of the Advia 120 hematology analyzer in the differential
cytologic analysis of biological fluids (cerebrospinal, perito-
neal, pleural, pericardial, synovial, and others). Lab Hematol
For personal use only.
2003;9:214–24.
[8] Barnes PW, Eby CS, Shimer G. An evaluation of the utility
of performing body fluid counts on the coulter LH 750. Lab
Hematol 2004;10:127–31.
[9] Sandhaus LM, Ciarlini P, Kidric D, Dillman C,
O’Riordan M. Automated cerebrospinal fluid cell counts
using the Sysmex XE-5000: is it time for new reference
ranges? Am J Clin Pathol 2010;134:734–8.
[10] Boer K, Deufel T, Reinhoefer M. Evaluation of the XE-5000
for the automated analysis of blood cells in cerebrospinal
fluid. Clin Biochem 2009;42:684–91.
[11] de Jonge R, Brouwer R, de Graaf MT, Luitwieler RL,
Fleming C, de Frankrijker-Merkestijn M, Sillevis Smitt PA,
Boonstra JG, Lindemans J. Evaluation of the new body fluid
mode on the Sysmex XE-5000 for counting leukocytes and
erythrocytes in cerebrospinal fluid and other body fluids.
Clin Chem Lab Med 2010;48:665–75.
[12] Paris A, Nhan T, Cornet E, Perol JP, Malet M, Troussard X.
Performance evaluation of the body fluid mode on the
platform Sysmex XE-5000 series automated hematology
analyzer. Int J Lab Hematol 2010;32:539–47.
[13] Zimmermann M, Ruprecht K, Kainzinger F, Heppner FL,
Weimann A. Automated vs. manual cerebrospinal fluid cell
counts: a work and cost analysis comparing the Sysmex
XE-5000 and the Fuchs-Rosenthal manual counting cham-
ber. Int J Lab Hematol 2011;33:629–37.
[14] Kleine TO, Nebe CT, Lower C, Geilenkeuser WJ,
Dorn-Beineke A. Cell analysis in cerebrospinal fluid (CSF)
using Sysmex® hematology analyzers XT-4000i and
XE-5000: evaluation with CSF controls of the Joint German
Society for Clinical Chemistry and Laboratory Medicine
(DGKL). Cytometry A 2012;81:255–64.
[15] Zimmermann M, Otto C, Gonzalez JB, Prokop S,
Ruprecht K. Cellular origin and diagnostic significance of
high-fluorescent cells in cerebrospinal fluid detected by the
XE-5000 hematology analyzer. Int J Lab Hematol 2013;
35:580–8.
[16] Kleine TO, Nebe CT, Lower C, Lehmitz R, Kruse R,
Geilenkeuser WJ, Dorn-Beineke A. Modifications of haema-
tology analyzers to improve cell counting and leukocyte
differentiating in cerebrospinal fluid controls of the Joint
German Society for Clinical Chemistry and Laboratory
Medicine. Cytometry A 2009;75:688–91.
[17] Chow G, Schmidley JW. Lysis of erythrocytes and leuko-
cytes in traumatic lumbar punctures. Arch Neurol 1984;
41:1084–5.
[18] Dux R, Kindler-Rohrborn A, Annas M, Faustmann P,
Lennartz K, Zimmermann CW. A standardized protocol for
flow cytometric analysis of cells isolated from cerebrospinal
fluid. J Neurol Sci 1994;121:74–8.
[19] Perne A, Hainfellner JA, Womastek I, Haushofer A,
Szekeres T, Schwarzinger I. Performance evaluation of the
Sysmex XE-5000 hematology analyzer for white blood cell
analysis in cerebrospinal fluid. Arch Pathol Lab Med
2012;136:194–8.
[20] Kristensen SR, Salling AM, Kristensen ST, Hansen AB.
Unrecognized preanalytical problem with the spectrophoto-
metric analysis of cerebrospinal fluid for xanthochromia.
Clin Chem. 2008;54:1924–5.