Hematopathology / Platelet Counting Accuracy in DIC and Acute Leukemia
Accuracy of Platelet Counting by Automated Hematologic
Analyzers in Acute Leukemia and Disseminated
Intravascular Coagulation
Potential Effects of Platelet Activation
Seon Young Kim, MD,1 Ji-Eun Kim, MS,1,2 Hyun Kyung Kim, MD, PhD,1,2 Kyou-Sup Han, MD, PhD,1
and Cheng Hock Toh, MD, PhD3
Key Words: Platelet count; Acute leukemia; Disseminated intravascular coagulation; Platelet activation
DOI: 10.1309/AJCP88JYLRCSRXPP
Abstract
Platelet counting in patients with acute leukemia
or disseminated intravascular coagulation (DIC) may
have a risk for erroneous counts owing to the presence
of nonplatelet particles or platelet activation. We
evaluated automated platelet counting methods using
the Abbott Cell-Dyn Sapphire (Abbott Diagnostics,
Santa Clara, CA), Sysmex XE-2100 (Sysmex, Kobe,
Japan), ADVIA 2120 (Siemens Diagnostics, Tarrytown,
NY), and Beckman Coulter LH 750 (Beckman
Coulter, Miami, FL) compared with the international
reference method (IRM). Automated platelet counting
methods were inaccurate compared with the IRM,
without evidence of interfering nonplatelet particles.
It is interesting that platelet activation markers were
associated with DIC severity and erroneous platelet
counting, suggesting that platelet activation is a
potential source of inaccuracy. Furthermore, the
artifactual in vitro platelet activation induced a high
degree of intermethod variation in platelet counts. The
inaccuracy of automated platelet counts increased the
risk for misdiagnosis of DIC. More attention needs to
be given to the accuracy of platelet counts, especially
in clinical conditions with florid platelet activation.
634 Am J Clin Pathol 2010;134:634-647
634 DOI: 10.1309/AJCP88JYLRCSRXPP
Measurement of platelet counts using automated hematol-
ogy analyzers is usually quite precise and accurate. However,
the accuracy of automated platelet counts can be compro-
mised when measuring very low platelet counts or in the
presence of interference from non-platelet particles or platelet
abnormalities.1 Recent studies, mainly focusing on the counts
of low levels of platelets, demonstrated that automated counts
were not as accurate in severely thrombocytopenic samples.2-5
These findings are of concern because current clinical guide-
lines lowered the prophylactic platelet transfusion threshold to
10 × 109/L for patients without additional risk factors.6
In addition to this limitation of the technology, automated
platelet counts can be inaccurate even at normal or high plate-
let ranges owing to the characteristics of blood specimens, eg,
in specimens with a substantial amount of interfering particles,
including WBC fragments, RBC fragments, immune com-
plexes, bacteria, lipid droplets, or protein aggregates. WBC
fragments can cause the spurious elevation of platelet counts
in patients with acute leukemia at diagnosis and during che-
motherapy.7-11 In 1 study, platelet-like fragments of leukemic
blasts were observed in 25% of acute leukemia cases.11 In addi-
tion, granulocyte fragments or microorganisms are possible
sources of platelet count overestimation in septic patients.12-14
RBC fragments, which are often observed in patients with
malignancies as part of microangiopathic processes, are also
recognized as a cause of erroneous platelet counts.1
To improve the accuracy of platelet counting and the
feasibility of comparing platelet counts between analyzers,
the International Council for Standardization in Haematology
(ICSH) and the International Society of Laboratory
Hematology (ISLH) proposed flow cytometry analysis of
monoclonal antibody–labeled platelets and the calculation of
© American Society for Clinical Pathology

platelet counts from the platelet/RBC ratio as a new interna-
tional reference method (IRM).15 After the establishment of
the IRM, multicenter studies comparing counting methods in
platelet concentrates were performed. These studies identi-
fied considerable variation between the different counting
principles and between different instruments using the same
counting principle; this variability was suggested to be related
to the increased proportion of activated small platelets.16-18
When platelets are activated, they become spherical with a
hypogranular cytoplasm and release small particles. This may
lead to the erroneous detection of platelets when using automat-
ed hematology analyzers owing to their deformed morphology.
Patients with acute leukemia or disseminated intravascular
coagulation (DIC) have an increased risk not only for interfer-
ence from nonplatelet particles but also for counting errors due
to platelet activation because platelet activation is inevitable
during the course of a disease in which high levels of thrombin
are generated and many inflammatory cytokines induce platelet
activation. Recognizing erroneous results of automated platelet
counts in these situations is especially critical for a consistent
decision in the diagnosis of DIC and for clinical decision mak-
ing regarding transfusion. The platelet count is an indispens-
able parameter in the DIC scoring system proposed by the
International Society on Thrombosis and Haemostasis Sub-
Committee of the Scientific and Standardization Committee on
DIC, in which platelet counts of less than 100 × 103/μL (100
× 109/L) and less than 50 × 103/μL (50 × 109/L) would score 1
and 2 points, respectively.19
The aim of this study was to investigate the accuracy of
platelet counting in patients with acute leukemia or suspected
of having DIC, clinical conditions with a risk for errone-
ous platelet counts owing to nonplatelet particles or platelet
activation. First, we compared 7 different automated platelet
counting methods from 4 widely used hematology analyzers,
Abbott Cell-Dyn Sapphire (Abbott Diagnostics, Santa Clara,
CA), Sysmex XE-2100 (Sysmex, Kobe, Japan), ADVIA 2120
(Siemens Diagnostics, Tarrytown, NY), and Beckman Coulter
LH 750 (Beckman Coulter, Miami, FL), with the IRM. Second,
to identify the factors responsible for inaccuracies in automated
platelet counting, the interference of nonplatelet particles was
examined in some specimens. Third, we analyzed the asso-
ciation of intermethod platelet count variation with platelet
activation status. Fourth, we explored whether artificial in vitro
platelet activation may induce intermethod variation of platelet
counts. Finally, we assessed the impact of the inaccuracy in
automated platelet counts on the diagnosis of overt DIC.
Materials and Methods
Study Population
EDTA-anticoagulated whole blood samples submitted
to the clinical laboratory for CBC counts were analyzed
© American Society for Clinical Pathology
Hematopathology / Original Article
for 2 cohorts of patients. The “acute leukemia” group con-
sisted of 44 samples from 21 patients with acute leukemia
who showed increased peripheral blasts (>20% of WBCs).
Besides 21 samples obtained at the initial examination of
each patient, 23 samples were repeatedly collected from 10
patients immediately after the initiation of cytotoxic che-
motherapy. These samples were included because previous
reports suggested an increased incidence of leukemic frag-
ments after chemotherapy.10 The “suspected-DIC” group
consisted of 159 samples from the same number of patients
who were clinically suspected of having DIC and had
undergone a battery of DIC screening tests. The DIC scores
were calculated according to the International Society on
Thrombosis and Haemostasis scoring system (scores calcu-
lated from platelet count, prothrombin time, fibrinogen, and
D-dimer), and “overt DIC” was defined as a score of 5 or
more.19 The suspected-DIC group was divided into 3 sub-
groups according to an arbitrary DIC score (scores calculated
from prothrombin time, fibrinogen, and D-dimer, regardless
of platelet count) as follows: patients with an arbitrary score
of 5 or more, definite DIC (overt DIC present, regardless of
the platelet count); patients with an arbitrary score of 3 or 4,
borderline DIC (the presence of overt DIC was dependent
on the platelet count); and patients with an arbitrary score
of 2 or less, unlikely DIC (no overt DIC, regardless of the
platelet count). The characteristics of the patients and details
of underlying disorders are shown in ❚Table 1❚.
Whole blood specimens from 15 healthy volunteers
(25-40 years old; platelet count, 125-344 × 103/μL [125-344
× 109/L]) were collected with written informed consent for
healthy control samples and an in vitro activation study. This
study was reviewed and approved by the Seoul National
University College of Medicine Institutional Review Board
(Seoul, Republic of Korea).
Automated Hematology Analyzers and the IRM
Seven different counting methods from 4 automated
hematology analyzers were used as follows: immunologic,
optical, and impedance methods using the Abbott Cell-Dyn
Sapphire (ImmPLT, CD-o, and CD-i, respectively); imped-
ance and optical methods using the Sysmex XE-2100 (XE-i
and XE-o, respectively); optical methods using the ADVIA
2120 (ADVIA); and the impedance method using a Beckman
Coulter LH 750 (LH750). The ImmPLT method using the
Cell-Dyn Sapphire is a fully automated method using a fluo-
rescein isothiocyanate (FITC)-conjugated CD61 monoclonal
antibody, and it generates direct reports of absolute counts.3
Each analyzer was calibrated according to the manufacturer’s
guidelines, and quality controls were performed according to
the standard procedures.
The IRM was performed according to the recommended
protocol of the ICSH/ISLH.15 Two monoclonal antibodies,
Am J Clin Pathol 2010;134:634-647 635
635 DOI: 10.1309/AJCP88JYLRCSRXPP 635
Kim et al / Platelet Counting Accuracy in DIC and Acute Leukemia

❚Table 1❚
Patient Characteristics*

Suspected DIC (n = 159)

Acute Leukemia Definite DIC Borderline DIC Unlikely DIC

Total (n = 180) (n = 21) (n = 41) (n = 64) (n = 54)

No. of samples collected 203 44 41 64 54

Age (y) 59 (2-88) 36 (2-79)† 59 (16-79) 62 (20-88) 62 (2-86)
Male/female (% male) 113/67 (62.8) 11/10 (52) 24/17 (59) 39/25 (61) 39/15 (72)
Platelet count (× 109/L)‡ 87.8 (3.7-339.0) 72.1 (6.7-330.6) 77.0 (8.8-304.0) 82.7 (3.7-339.0) 111.8 (16.3-317.5)
DIC score‡ 4 (0-8) 3 (1-8) 6 (5-8) 4 (3-6) 2 (0-4)
Underlying disorders
Hematologic malignancy 30 (16.7) 21 (100) 0 (0) 0 (0) 9 (17)§
Sepsis/infection 55 (30.6) 0 (0) 13 (32) 26 (41) 16 (30)
Solid tumors 42 (23.3) 0 (0) 11 (27) 19 (30) 12 (22)
Hepatic failure 26 (14.4) 0 (0) 14 (34)† 9 (14) 3 (6)
Organ destruction 9 (5.0) 0 (0) 1 (2) 3 (5) 4 (7)
Trauma 5 (2.8) 0 (0) 1 (2) 1 (2) 4 (7)
Vascular abnormalities 5 (2.8) 0 (0) 1 (2) 2 (3) 2 (4)
Severe immunologic reaction 5 (2.8) 0 (0) 0 (0) 2 (3) 3 (6)
Obstetric calamities 3 (1.7) 0 (0) 0 (0) 2 (3) 1 (2)

DIC, disseminated intravascular coagulation.

* Data are shown as the median (range) for continuous variables or number (percentage) for categorical variables unless otherwise indicated.
† P < .05 by Mann-Whitney U test (continuous variable) or χ2 test (categorical variable) compared with the “Unlikely DIC” group.
‡ Platelet counts and DIC score by the international reference method. Platelet counts are given in Système International units; to convert to conventional units (× 103/μL),

divide by 1.0.
§ Patients with lymphoma or leukemia who were in remission.

anti–CD61-FITC (Becton Dickinson, San Jose, CA) and
anti–CD41-FITC (Becton Dickinson) were used. Flow
cytometry analysis was performed using a FACSCalibur
with CellQuest software (Becton Dickinson). RBC/platelet
ratios were calculated as described elsewhere,15,20 with the
known RBC count determined as the average of each RBC
count obtained using the Cell-Dyne Sapphire impedance
cell counter, XE-2100, ADVIA 2120, and LH750. Precision
studies were performed on 3 to 10 replicates of 15 specimens
with platelet counts ranging from 12 to 414 × 103/μL (12-
414 × 109/L) and demonstrated a mean coefficient of varia-
tion (CV) of 5.4%. All procedures of platelet counting were
performed within 8 hours after venipuncture.
Investigation for the Presence of Interfering Nonplatelet
Particles
To determine whether WBC or RBC fragments or
endothelial microparticles were present, we examined periph-
eral blood smears stained with Wright-Giemsa in all samples.
We also performed flow cytometric analysis using the follow-
ing monoclonal antibodies (all from Becton Dickinson): anti–
CD45-phycoerythrin (PE), anti–CD33-PE, and anti–CD54-PE
for WBC fragments; anti–glycophorin A–FITC for RBC frag-
ments; and anti–CD62E-PE for endothelial microparticles.
Investigation of Platelet Activation Parameters
Platelet activation parameters, including mean fluores-
cence intensity of side scatter (MFI of SSC), mean platelet
component concentration (MPC), and percentage of small
platelets (pPLT-s), were determined in all specimens at the
same time platelet counts were measured. MFI of SSC in the
gated CD41+/CD61+ platelet events was set out from the
scattergram of flow cytometry in the IRM.21 MPC, the mean
refractive index of platelets, was obtained using the ADVIA
2120.22 These parameters reflect the internal complexity and
density of platelets and decrease when platelets degranulate
on activation.21 The pPLT-s was calculated by dividing the
number of small platelets by the total platelet count in the
ImmPLT method.17 The small platelets, which are defined
by CD61+ platelets below a fixed lower threshold in 7° light
scatter, include platelet microparticle fractions and increase
after platelet activation.17,23
Effect of In Vitro Platelet Activation on Intermethod
Variation
The 15 EDTA-anticoagulated whole blood specimens
from 10 healthy volunteers and 5 patients (platelet counts,
60-120 × 103/μL [60-120 × 109/L]; DIC scores, 2-4) were
divided into aliquots in 3 tubes. Two aliquots from each
specimen were incubated with 0.64 mmol/L fibrin inhibitor
(glycine-proline-arginine-proline, Sigma Aldrich, St Louis,
MO) for 5 minutes and were then stimulated with 3 or 10
μmol/L adenosine diphosphate (Sigma Aldrich) for 5 minutes
at room temperature. The remaining aliquot was kept without
manipulation. The platelet counts were obtained using the 7
automated methods and the IRM as described. The degree of
platelet activation was estimated by flow cytometric analysis
using anti–annexin V–PE and anti–PAC-1-FITC (Becton
Dickinson), as well as by the platelet activation parameters
obtained from the automated analyzers.

636 Am J Clin Pathol 2010;134:634-647 © American Society for Clinical Pathology

636 DOI: 10.1309/AJCP88JYLRCSRXPP
 Hematopathology / Original Article

Statistical Analysis
Data were compared by using the Mann-Whitney U
test and Kruskal-Wallis analysis of variance for continu-
ous variables and the χ2 test for categorical variables. The
comparison between the 2 methods was performed by using
Passing-Bablok regression analysis, the Pearson correlation
coefficient, and Bland-Altman analysis. Because the differ-
ence between the analyzers and the IRM increased as the
platelet counts increased, percentage difference plots were
used as suggested by Bland and Altman.24 The 95% limits
of agreement (95% LA) were calculated from the 95% range
(mean ± 1.96 SD) of the percentage differences to quantify the
degree of agreement between the 2 methods.
The percentage of erroneous results between each auto-
mated method and the IRM was arbitrarily defined by the per-
centage of specimens that were outside the ± 25% error limits
(± 25% difference from zero difference) of the platelet count.
This error limit was modified from the Clinical Laboratory
Improvement Amendments criteria for the acceptable per-
formance of the quantitative hematology test in proficiency
testing programs.25
The CV across the 7 automated methods and the IRM
(intermethod CV) were also calculated for each specimen.
The percentage of erroneous results across the quartiles of
each platelet parameter were analyzed with adjustments for
age, sex, and platelet levels by using linear regression, and the
test for linear trend of the percentage was performed on the
basis of logistic regression.
Statistical analyses were carried out using MedCalc, ver-
sion 9.6 (MedCalc Software, Mariakerke, Belgium) and SAS,
version 9.1 (SAS Institute, Cary, NC). P values less than .05
were considered statistically significant.
Results
Comparison of Platelet Counts Between Automated
Methods and the IRM
The Pearson correlation coefficients between each
automated method and the IRM ranged from 0.943 to 0.976
❚Figure 1❚ (upper panels). The correlation was the stron-
gest between ImmPLT and the IRM and weakest between
CD-o and the IRM. Despite these good correlations, the
slopes from Passing-Bablok regression were less than 0.90
(from 0.82 in XE-o to 0.89 in CD-i), except for ADVIA
(0.95) (Figure 1, upper panels), suggesting that there were
considerable proportional errors in platelet counts in the

❚Figure 1❚ Comparison of platelet counts between the A

Cell-Dyn Sapphire immunologic method (ImmPLT) (A),
optical method (CD-o) (B), and impedance method (CD-i)
300
(C); XE-2100 impedance method (XE-i) (D) and optical
method (XE-o) (E); ADVIA 2120 (ADVIA) (F) and Coulter LH
ImmPLT (× 109/L)

750 (LH750) (G) with the international reference method

200
(IRM) in all samples using Passing-Bablok regression (upper
panels) and Bland-Altman analysis (lower panels). Shaded
areas in the Bland-Altman plots indicate ± 25% error limits 100
(± 25% difference from zero difference) of platelet counts.
Upper panels: A, y = 0.867x – 1.187; r = 0.976. B, y =
0.844x – 1.901; r = 0.943. C, y = 0.890x – 2.892; r = 0.971. 0
D, y = 0.859x – 1.587; r = 0.969. E, y = 0.821x – 2.754; r 0 100 200 300
IRM (× 109/L)
= 0.970. F, y = 0.947x – 2.474; r = 0.968. G, y = 0.844x – 100
1.144; r = 0.971. Outside ± 25% error limits, lower panels:
% Difference (ImmPLT – IRM)

A, 19.7% (40/203). B, 39.9% (81/203). C, 14.7% (28/191).* 50

D, 22.7% (46/203). E, 29.0% (58/200). F, 13.4% (27/201). Mean + 1.96 SD
G, 28.0% (56/200). * For the CD-i, 12 platelet counts were 0
missing because the CD-i did not report platelet counts on
samples with a platelet count of <20 × 109/L. Results were Mean
–50
not given for 4 XE-o, 2 ADVIA, and 3 LH750 samples owing Mean – 1.96 SD
to insufficient volume. Platelet counts are given in Système
–100
International units; to convert to conventional units (× 103/μL),
divide by 1.0. IRM, international reference method. (Figure 1
–150
continues on next 2 pages.)
0 100 200 300
Mean of ImmPLT and IRM (× 109/L)

© American Society for Clinical Pathology Am J Clin Pathol 2010;134:634-647 637

637 DOI: 10.1309/AJCP88JYLRCSRXPP 637
Kim et al / Platelet Counting Accuracy in DIC and Acute Leukemia

B C

300 300
CD-o (× 109/L)

CD-i (× 109/L)
200 200

100 100

0 0
0 100 200 300 0 100 200 300
IRM (× 109/L) 100 IRM (× 109/L)
100
% Difference (CD-o – IRM)

% Difference (CD-i – IRM)

50 Mean + 1.96 SD 50
Mean + 1.96 SD

0 0

Mean
Mean
–50 –50 Mean – 1.96 SD

Mean – 1.96 SD
–100 –100

–150 –150
0 100 200 300 0 100 200 300
Mean of CD-o and IRM (× 109/L) Mean of CD-i and IRM (× 109/L)

D E

300 300
XE-o (× 109/L)
XE-i (× 109/L)

200 200

100 100

0 0
0 100 200 300 0 100 200 300
IRM (× 109/L) 100 IRM (× 109/L)
100
% Difference (XE-o – IRM)
% Difference (XE-i – IRM)

50 50
Mean + 1.96 SD
Mean + 1.96 SD

0 0
Mean
Mean
–50 Mean – 1.96 SD –50
Mean – 1.96 SD

–100 –100

–150 –150
0 100 200 300 0 100 200 300
Mean of XE-i and IRM (× 109/L) Mean of XE-o and IRM (× 109/L)

638 Am J Clin Pathol 2010;134:634-647 © American Society for Clinical Pathology

638 DOI: 10.1309/AJCP88JYLRCSRXPP
 Hematopathology / Original Article

automated methods compared with the count in the IRM,
except for ADVIA. The 95% LA on the Bland-Altman
plots of the ImmPLT, CD-o, CD-i, XE-i, XE-o, ADVIA,
and LH750 vs the IRM varied from –46% to +11%, –93%
to +37%, –40% to +26%, –54% to +27%, –54% to +24%,
–39% to +36%, and –61% to 26%, respectively (Figure 1,
solid lines, lower panels). The 95% LA was the widest in
the CD-o and narrowest in the ImmPLT. Erroneous platelet
count, defined by the percentage of specimens that were
outside the ± 25% error limits in the Bland-Altman plot,
ranged from 13.4% (ADVIA) to 39.9% (CD-o) (Figure 1,
shaded area, lower panels).
Investigation of Interfering Nonplatelet Particles
The examination of peripheral blood films revealed
no definite evidence of interfering nonplatelet particles in
any specimen. Flow cytometric analysis using anti-CD45
in 106 specimens from the acute leukemia (n = 31) and
suspected-DIC (n = 75) groups showed no WBC fragments
in the platelet-sized area on the scattergram of the IRM.
Analysis using anti-CD33 and anti-CD54 antibodies in 14
samples from the patients with acute myeloid leukemia
showed no blast fragments. There were no platelet-sized
fragments of endothelial cells labeled with anti-CD62E in
33 samples from the definite-DIC group. Flow cytometric
analysis with glycophorin A in 22 selected specimens also
did not show any platelet-sized fragments of RBCs.
Intermethod Variation of Platelet Counts in Terms
of Underlying Clinical Condition
The intermethod CV (mean ± SD) of platelet counts
was significantly higher in the acute leukemia group
(20.7% ± 12.2%) than in healthy control subjects (7.6%
± 1.7%; P < .001) ❚Figure 2A❚. Within the suspected-DIC
group, the intermethod CV gradually increased as the DIC
severity increased (unlikely DIC, 10.6% ± 4.5%; border-
line DIC, 13.4% ± 7.4%; definite DIC, 17.4% ± 13.3%
❚Figure 2B❚, ❚Figure 2C❚, and ❚Figure 2D❚. When the acute
leukemia specimens were divided into tertiles on the basis
of their platelet counts, the 95% LA tended to increase as
the platelet levels decreased across all automated methods,
and the 95% LA of CD-o was the widest among the auto-
mated methods (Figure 2A). Among the suspected-DIC
groups, the definite-DIC group showed the widest 95%

F G

300 300
LH750 (× 109/L)
ADVIA (× 109/L)

200 200

100 100

0 0
0 100 200 300 0 100 200 300
IRM (× 109/L) 100 IRM (× 109/L)
100
% Difference (LH750 – IRM)
% Difference (ADVIA – IRM)

50 Mean + 1.96 SD 50
Mean + 1.96 SD

0 0
Mean
Mean
–50 Mean – 1.96 SD –50
Mean – 1.96 SD

–100 –100

–150 –150
0 100 200 300 0 100 200 300
9
Mean of ADVIA and IRM (× 10 /L) Mean of LH750 and IRM (× 109/L)

© American Society for Clinical Pathology Am J Clin Pathol 2010;134:634-647 639

639 DOI: 10.1309/AJCP88JYLRCSRXPP 639
Kim et al / Platelet Counting Accuracy in DIC and Acute Leukemia

LA in all methods, and the unlikely-DIC group showed the the CD-o, XE-i, and LH750 showed the widest 95% LA
narrowest 95% LA (Figures 2B-2D). The ImmPLT dem- (Figure 2B). In the unlikely-DIC group, the 95% LA did
onstrated the narrowest 95% LA across all methods in the not show much variability across all methods and platelet
3 subgroups of suspected DIC. In the definite-DIC group, tertiles (Figure 2D).

A
60

40
% Difference (Automated Method – IRM)

20

–20

–40

–60

–80
Acute leukemia group
–100 Total (n = 44)
Platelets <60 × 109/L (n = 17)
–120 Platelets 60-120 × 109/L (n = 14)
Platelets >120 × 109/L (n = 13)
–140
Healthy control (n = 15)
–160
ImmPLT CD-o CD-i XE-i XE-o ADVIA LH750

B
60

40
% Difference (Automated Method – IRM)

20

–20

–40

–60

–80

–100
Definite DIC group
–120 Total (n = 41)
Platelets <60 × 109/L (n = 14)
–140 Platelets 60-120 × 109/L (n = 15)
Platelets >120 × 109/L (n = 12)
–160
ImmPLT CD-o CD-i XE-i XE-o ADVIA LH750

❚Figure 2❚ Bland-Altman plots showing the mean percentage differences and 95% limits of agreement of platelet counts in
different patient groups: acute leukemia (A), definite disseminated intravascular coagulation (DIC) (B), borderline DIC (C), and
unlikely DIC (D). The first line represents percentage differences of total specimens in each patient group, and the other lines
represent those of tertile subgroups based on platelet levels. Results for 15 healthy subjects are also illustrated in the acute
leukemia group (A).

640 Am J Clin Pathol 2010;134:634-647 © American Society for Clinical Pathology

640 DOI: 10.1309/AJCP88JYLRCSRXPP
 Hematopathology / Original Article

Association of Platelet Activation Markers With Erroneous the percentage of specimens with erroneous platelet counts
Platelet Counts and the Underlying Clinical Conditions (outside the ±25% error limits on Bland-Altman plots
To investigate whether platelet activation status had in Figure 1) was shown on the basis of the quartiles of
an effect on the erroneous measurement of platelet counts, platelet activation parameters in all specimens ❚Figure 3❚.

C
60

40
% Difference (Automated Method – IRM)

20

–20

–40

–60

–80
Borderline DIC group
–100
Total (n = 64)
–120 Platelets <60 × 109/L (n = 22)
Platelets 60-120 × 109/L (n = 23)
–140 Platelets >120 × 109/L (n = 19)

–160
ImmPLT CD-o CD-i XE-i XE-o ADVIA LH750

D
60

40
% Difference (Automated Method – IRM)

20

–20

–40

–60

–80
Unlikely DIC group
–100 Total (n = 54)
Platelets <60 × 109/L (n = 14)
–120
Platelets 60-120 × 109/L (n = 17)
–140 Platelets >120 × 109/L (n = 23)

–160
ImmPLT CD-o CD-i XE-i XE-o ADVIA LH750

Intermethod coefficients of variation (ranges) are as follows: A, 20.7% ± 12.2% (7.3%-61.6%); P < .0001;
B, 17.4% ± 13.3% (4.3%-64.0%); P < .0001; C, 13.4% ± 7.4% (4.4%-35.1%); P < .0002; D, 10.6% ± 4.5% (5.3%-24.4%);
P < .003. P by Mann-Whitney U test compared with the intermethod coefficient of variation of healthy control subjects (7.6% ±
1.7%; range, 4.6%-10.9%). ADVIA, ADVIA 2120; CD-i, Cell-Dyn Sapphire impedance method; CD-o, Cell-Dyn Sapphire optical
method; ImmPLT, Cell-Dyn Sapphire immunologic method; IRM, international reference method; LH750, Coulter LH 750; XE-i,
XE-2100 impedance method; XE-o, XE-2100 optical method.

© American Society for Clinical Pathology Am J Clin Pathol 2010;134:634-647 641

641 DOI: 10.1309/AJCP88JYLRCSRXPP 641
Kim et al / Platelet Counting Accuracy in DIC and Acute Leukemia

Specimens in the lowest quartile (Q1) of the MFI of SSC
and MPC, which represent hypogranular activated platelets,
showed significantly higher erroneous results than those
in the highest quartile (Q4) in the CD-o, XE-i, XE-o, and
ADVIA (Figures 3A and 3B). The CD-i and LH750 demon-
strated no apparent trends with these parameters. The pPLT-
s, the proportion of small and CD61+ platelets representing
activated small platelets, showed significant trends of high
error rates in the highest quartile (Q4) of pPLT-s for most
analyzers (Figure 3C). However, the platelet distribution
width showed no apparent increasing or decreasing trends
across the quartiles, indicating that platelet size variation
was not the main factor responsible for the erroneous plate-
let count (Figure 3D).
To investigate the activation status of platelets according
to the underlying clinical condition, the mean value of the
platelet activation parameter in each patient group was cal-
culated as shown in ❚Table 2❚. Patients with acute leukemia
showed significantly lower levels of the MFI of SSC and
MPC and significantly higher levels of pPLT-s than patients
with unlikely DIC. The definite-DIC group showed signifi-
cantly lower levels of MPC.
Effects of In Vitro Platelet Activation on the Agreement
of Platelet Counts
Fresh whole blood specimens were obtained from 10
healthy volunteers and 5 patients with suspected DIC and
were stimulated with a low and a high dose of adenosine
diphosphate. Platelet activation status was confirmed with
measurements of the percentage of annexin V+ or PAC-1+
platelets, MPC, and pPLT-s ❚Figure 4❚. As expected, the
intermethod CV was significantly increased after weak and
strong activation (control, 7.2% ± 1.4%; weak activation,
7.7% ± 1.4%; and strong activation, 9.9% ± 5.8%, Figures
4A-4C). Similarly, the 95% LA gradually became wider as
the intensity of platelet activation became stronger.

A
Samples Outside ±25% Error Limits (%)

MFI of SSC
*
Q1 Q2 Q3 Q4
50
†
†
40

*
30

20

10

0
ImmPLT CD-o CD-i XE-i XE-o ADVIA LH750
P for trend .105 .002 .142 .002 <.001 .615 .342

B
Samples Outside ±25% Error Limits (%)

MPC
Q1 Q2 Q3 Q4
50
‡ †
40
*
†
30

20

10

0
ImmPLT CD-o CD-i XE-i XE-o ADVIA LH750
P for trend .092 .150 .485 <.001 .003 .004 .378

❚Figure 3❚ Adjusted percentage of specimens outside the ± 25% error limits in platelet counts by each automated method are
presented on the basis of the quartiles (Q1-Q4) of platelet parameters: mean fluorescence intensity of side scatter (MFI of SSC)
(A), mean platelet component concentration (MPC) (B), the proportion of small platelets (pPLT-s) (C), and platelet distribution
width (PDW) (D).

642 Am J Clin Pathol 2010;134:634-647 © American Society for Clinical Pathology

642 DOI: 10.1309/AJCP88JYLRCSRXPP
 Hematopathology / Original Article

Impacts of Intermethod Variation in Platelet Counts
on the Diagnosis of Overt DIC
We investigated the clinical impact of the intermethod
variation in platelet counts on the diagnosis of overt DIC in the
borderline-DIC group (n = 64). Platelet counts measured by each
automated method were applied to the DIC scoring system and
compared with those from the IRM. There was a tendency for
all types of analyzers to overdiagnose overt DIC compared with
the IRM (false-positive, 3.2%-1.4%; false-negative, 0%-1.6%)
❚Table 3❚. The CD-i, XE-i, and ADVIA showed slightly higher
κ indices (≥0.90) than the other methods.
Discussion
This study demonstrated that automated platelet counting
methods were inaccurate in patients with acute leukemia or
DIC when compared with the IRM and that the intermethod
variation of the platelet count correlated with the platelet acti-
vation status. Furthermore, in vitro platelet activation induced
a higher intermethod variation of platelet counts. The inaccu-
racies of automated platelet counts have an impact on clinical
decision making in the diagnosis of overt DIC.
There has been some debate over which counting prin-
ciple, between the impedance and optical methods, measures
platelet counts more accurately. Some studies suggested that
the accuracy of the optical methods was superior for thrombo-
cytopenic specimens,2,26,27 while recent studies demonstrated
the impedance method to be more accurate for samples from
patients undergoing cytotoxic chemotherapy.5,28 Our data
showed that there was no consistent superiority of platelet
measurement in terms of the counting principle. The ImmPLT
and ADVIA gave better agreement with the IRM than the
other methods, while the CD-o demonstrated substantial inac-
curacy in patients with acute leukemia or severe coagulopathy.

C
Samples Outside ±25% Error Limits (%)

pPLT-s

50 Q1 Q2 Q3 Q4

40 * *

30 † †

‡
20 ‡ *
‡
‡ ‡
10
†

0
ImmPLT CD-o CD-i XE-i XE-o ADVIA LH750
P for trend .282 .045 .429 <.001 .001 .518 .677
Samples Outside ±25% Error Limits (%)

D PDW

50 Q1 Q2 Q3 Q4

40

30

20

10

0
ImmPLT CD-o CD-i XE-i XE-o ADVIA LH750
P for trend .126 .940 .155 .140 .678 .106 .131

All data were adjusted for age, sex, and platelet levels. * P < .05, † P < .01, and ‡ P < .001 compared with the highest
quartile (Q4). ADVIA, ADVIA 2120; CD-i, Cell-Dyn Sapphire impedance method; CD-o, Cell-Dyn Sapphire optical method;
ImmPLT, Cell-Dyn Sapphire immunologic method; LH750, Coulter LH 750; XE-i, XE-2100 impedance method; XE-o, XE-2100
optical method.

© American Society for Clinical Pathology Am J Clin Pathol 2010;134:634-647 643

643 DOI: 10.1309/AJCP88JYLRCSRXPP 643
Kim et al / Platelet Counting Accuracy in DIC and Acute Leukemia

❚Table 2❚
Levels of Platelet Parameters According to Underlying Clinical Condition in 203 Tested Specimens*

Acute Leukemia (n = 44) Definite DIC (n = 41) Borderline DIC (n = 64) Unlikely DIC (n = 54)

MFI of SSC 59.24 (8.47)† 70.84 (14.57) 74.71 (12.39) 73.70 (12.57)
pPLT-s (%) 2.57 (2.39)‡ 1.71 (2.12) 1.16 (1.36) 1.54 (1.54)
MPC (g/dL) 20.75 (1.08)§ 20.60 (1.05)† 21.42 (1.21) 21.64 (1.42)
PDW (%) 15.14 (2.12) 15.88 (2.56) 15.89 (2.23) 15.32 (1.64)

MFI, mean fluorescence intensity; MPC, mean platelet component; PDW, platelet distribution width; pPLT-s, percentage of small platelets; SSC, side scatter.
* Data are shown as the mean (SD). P values were calculated by using the Mann-Whitney U test and show comparison with the “Unlikely DIC” group.
† P < .001.
‡ P < .05.
§ P < .01.

A B

10 10

% Difference (Automated Method – IRM)

% Difference (Automated Method – IRM)

0 0

–10 –10

–20 –20

–30 –30

–40 –40
ImmPLT CD-o CD-i XE-i XE-o ADVIA LH750 ImmPLT CD-o CD-i XE-i XE-o ADVIA LH750

❚Figure 4❚ Bland-Altman plots showing the mean percentage

C
differences and 95% limits of agreement of platelet counts in
3 states of platelet activation. Fresh whole blood specimens
10 (n = 15) were stimulated without (control, A) or with 3
μmol/L adenosine diphosphate (ADP) (B) or 10 μmol/L ADP
(C). Intermethod coefficients of variation (ranges) are as
% Difference (Automated Method – IRM)

0 follows: A, 7.2% ± 1.4% (4.9%-10.4%); B, 7.7% ± 1.4%

(4.6%-9.6%); P < .164; C, 9.9% ± 5.8% (4.6%-28.9%); P
< .068. P by the paired t test compared with control. A,
Annexin V+ (%), 3.9 ± 0.5; PAC-1+ (%), 2.8 ± 0.8; MPC (g/
–10
dL), 24.3 ± 1.1; pPLT-s (%), 0.672 ± 0.502. B, Annexin V+
(%), 12.1 ± 4.2 (P < .001); PAC-1+ (%), 5.3 ± 1.7 (P = .005);
MPC (g/dL), 23.8 ± 1.0 (P = .083); pPLT-s (%), 0.748 ± 0.577
–20 (P = .048). C, Annexin V+ (%), 14.9 ± 1.6 (P < .001); PAC-1+
(%), 7.8 ± 4.9 (P = .048); MPC (g/dL), 23.6 ± 1.1 (P = .027);
pPLT-s (%), 0.786 ± 0.581 (P = .001). ADVIA, ADVIA 2120;
CD-i, Cell-Dyn Sapphire impedance method; CD-o, Cell-Dyn
–30
Sapphire optical method; ImmPLT, Cell-Dyn Sapphire
immunologic method; IRM, international reference method;
LH750, Coulter LH 750; MPC, mean platelet component;
–40 pPLT-s, percentage of small platelets; XE-i, XE-2100
ImmPLT CD-o CD-i XE-i XE-o ADVIA LH750 impedance method; XE-o, XE-2100 optical method.

644 Am J Clin Pathol 2010;134:634-647 © American Society for Clinical Pathology

644 DOI: 10.1309/AJCP88JYLRCSRXPP
 Hematopathology / Original Article

❚Table 3❚
Sensitivity and Specificity of Automated Methods Compared With the IRM for Diagnosis of Overt DIC in the Borderline DIC
Group (n = 64)

IRM Raw Platelet Count

DIC Score* ≥5 <5 FP (%) FN (%) SN (%) SP (%) κ Index

ImmPLT ≥5 24 6 9.4 0 100 85 0.81

<5 0 34
CD-o ≥5 24 8 12.5 0 100 80 0.75
<5 0 32
CD-i ≥5 24 1 1.6 0 100 98 0.97
<5 0 39
XE-i ≥5 24 3 4.7 0 100 93 0.90
<5 0 37
XE-o ≥5 23 4 6.3 1.6 96 90 0.84
<5 1 36
ADVIA ≥5 23 2 3.1 1.6 96 95 0.90
<5 1 38
LH750 ≥5 24 4 6.3 0 100 90 0.87
<5 0 36

ADVIA, ADVIA 2120; CD-i, Cell-Dyn Sapphire impedance method; CD-o, Cell-Dyn Sapphire optical method; DIC, disseminated intravascular coagulation; FN, false-negative;
FP, false-positive; ImmPLT, Cell-Dyn Sapphire immunologic method; IRM, international reference method; LH750, Coulter LH 750; SN, sensitivity; SP, specificity; XE-i,
XE-2100 impedance method; XE-o, XE-2100 optical method.
* Patients with a DIC score of ≥5 were diagnosed as having overt DIC.

The agreement of XE-i and XE-o with the IRM was variable
depending on the clinical condition or platelet level. Thus, in
our study, the optical counting principles were not superior to
the impedance counting principle.
The most frequently suggested mechanism of platelet
counting error is the inability to discriminate platelet-sized
interfering particles from true platelets.11 Because the speci-
mens included in our study were those having known disor-
ders related to “pseudoplatelets,” we investigated the possible
sources of nonplatelet particles using peripheral blood smears
and flow cytometric analysis of CD33, CD45, CD62E, and
glycophorin A, but we could not find any nonplatelet particles
of platelet sizes in patients with acute leukemia, in patients
immediately after chemotherapy, or in patients with DIC.
Thus, WBC or RBC fragments are unlikely to be the source
of the discrepancies in platelet counts between the counting
principles in our study samples.
In our results, as the proportion of small and hypogranular
platelets increased, the proportion of erroneous platelet counts
increased. This association seemed to be stronger for the opti-
cal methods than the impedance and immunologic methods.
Moreover, these platelet activation parameters, including the
MFI of SSC, MPC, and pPLT-s, showed a significant trend
for high levels in the acute leukemia and definite-DIC groups
compared with in the unlikely-DIC group. These results sug-
gest that platelet activation is a potential candidate for the
intermethod variation in platelet counts. We subsequently
confirmed that in vitro platelet activation induced the observed
intermethod variation of platelet counts. This is the first report
describing the association of platelet activation status with
intermethod platelet count variation in whole blood specimens.
The activated platelets are a minor component of the total
platelet count; however, when their proportion is significantly
increased in pathologic status, it needs to be considered as a
factor causing abnormal platelet counts. Although we could
not explore the exact mechanism of the association of platelet
activation and the intermethod variation, we could postulate
that automated analyzers with different counting principles
have different intrinsic detection limits for identifying degran-
ulated small platelets.
We finally assessed whether the intermethod variation
in automated platelet counts can cause misdiagnosis of overt
DIC. There was a potential to overdiagnose overt DIC com-
pared with the IRM when guided by the platelet counts from
all types of automated methods because the automated platelet
counts tended to underestimate compared with the IRM. One
multicenter study using only specimens with a platelet count
less than 20 × 103/μL (20 × 109/L) showed that most auto-
mated counting methods had overestimated platelet counts
at platelet levels of 5, 10, and 15 × 103/μL (5, 10, and 15 ×
109/L).5 However, other studies reported that automated plate-
let counts were underestimated at the higher levels of 20 or 50
× 103/μL (20 or 50 × 109/L), similar to the results described
herein.29-31 The inaccuracy of automated platelet counts at
higher levels could be a matter of serious concern in clinical
practice, not only because of the risk for misdiagnosis but also
because patients frequently have risk factors for hemorrhage
and higher transfusion thresholds.6
Despite these inaccuracies of routine automated methods
in some abnormal samples, it is not feasible to implement the
IRM for routine use as a necessity. Compared with automated
analyzers that are much quicker and cost-effective, the IRM is

© American Society for Clinical Pathology Am J Clin Pathol 2010;134:634-647 645

645 DOI: 10.1309/AJCP88JYLRCSRXPP 645
Kim et al / Platelet Counting Accuracy in DIC and Acute Leukemia
time-consuming and requires technical expertise and specific
facilities. The ImmPLT, another possible option, is not feasible
for routine use either owing to its high cost and the requirement
of 1 dedicated automated analyzer (Cell-Dyn). Therefore, to
improve platelet count accuracy, further efforts to develop a
simple and accurate standard method are necessary.
This study has some limitations. First, we performed this
accuracy study of platelet counts within 1 clinical laboratory
of a tertiary university hospital. Therefore, interinstitutional
variation was not investigated. Second, we used routine auto-
mated analyzers without advanced calibration of platelet
counts just before the study because our study was expected
to show intermethod variation only under the current situa-
tion of routine automated analyzers that were calibrated on a
regular basis according to each of the manufacturer’s guide-
lines. A universal calibrator would be expected to improve the
intermethod variation of platelet counts. Third, although we
focused on the readily available platelet activation parameters
generated by automated analyzers in this study, these param-
eters are less sensitive than flow cytometric platelet activation
markers such as anti-CD62p and PAC-1. Further analysis
using the actual counts of activated platelets measured by flow
cytometry would be able to give a more quantitative and spe-
cific interpretation.
Although hematology analyzers usually provide reliable
platelet counts, they may be inaccurate at enumerating platelets
in patients with acute leukemia or DIC, ultimately resulting in
the potential risk to make a wrong decision for DIC. It is inter-
esting that platelet activation markers were associated with
the severity of DIC and erroneous platelet counts, suggesting
that platelet activation is a potential source for the intermethod
variation in platelet counts. More attention needs to be given to
improve the accuracy of platelet counts, especially in clinical
conditions with high levels of platelet activation.
From the 1Department of Laboratory Medicine and 2Cancer
Research Institute, Seoul National University College of Medicine,
Seoul, Korea; and 3Department of Haematology, University of
Liverpool, Liverpool, England.
Supported by grant 2009-0075731 from the National
Research Foundation of Korea funded by the Korean Government
and by grant 2009-0093820 from the Priority Research Centers
Program through the National Research Foundation of Korea
funded by the Ministry of Education, Science and Technology,
Daejeon.
Address reprint requests to Dr H.K. Kim: Dept of Laboratory
Medicine, Seoul National University College of Medicine, 101,
Daehang-no Jongno-gu, Seoul 110-744, Republic of Korea.
References
1. Zandecki M, Genevieve F, Gerard J, et al. Spurious counts
and spurious results on haematology analysers: a review, part
I: platelets. Int J Lab Hematol. 2007;29:4-20.
646 Am J Clin Pathol 2010;134:634-647
646 DOI: 10.1309/AJCP88JYLRCSRXPP
2. Harrison P, Horton A, Grant D, et al. Immunoplatelet
counting: a proposed new reference procedure. Br J Haematol.
2000;108:228-235.
3. Kunz D, Kunz WS, Scott CS, et al. Automated CD61
immunoplatelet analysis of thrombocytopenic samples. Br J
Haematol. 2001;112:584-592.
4. Norris S, Pantelidou D, Smith D, et al. Immunoplatelet
counting: potential for reducing the use of platelet
transfusions through more accurate platelet counting. Br J
Haematol. 2003;121:605-613.
5. Segal HC, Briggs C, Kunka S, et al. Accuracy of platelet
counting haematology analysers in severe thrombocytopenia
and potential impact on platelet transfusion. Br J Haematol.
2005;128:520-525.
6. British Committee for Standards in Haematology, Blood
Transfusion Task Force. Guidelines for the use of platelet
transfusions. Br J Haematol. 2003;122:10-23.
7. Malcolm ID, Monks P, Katz M. Spurious thrombocytosis
in acute myelocytic leukemia [letter]. N Engl J Med.
1978;298:1260.
8. Ballard HS, Sidhu G. Cytoplasmic fragments causing
spurious platelet counts in hairy cell leukemia: ultrastructural
characterization. Arch Intern Med. 1981;141:942-944.
9. Hammerstrom J. Spurious platelet counts in acute leukaemia
with DIC due to cell fragmentation. Clin Lab Haematol.
1992;14:239-243.
10. Li S, Salhany KE. Spurious elevation of automated
platelet counts in secondary acute monocytic leukemia
associated with tumor lysis syndrome. Arch Pathol Lab Med.
1999;123:1111-1114.
11. van der Meer W, MacKenzie MA, Dinnissen JW, et al.
Pseudoplatelets: a retrospective study of their incidence
and interference with platelet counting. J Clin Pathol.
2003;56:772-774.
12. Krauss JS, Dover RK, Khankhanian NK, et al. Granulocytic
fragments in sepsis. Mod Pathol. 1989;2:301-305.
13. Hirokawa M, Manabe T, Ishimatsu S, et al. Leukocytic
fragments in blood smears. Acta Pathol Jpn. 1990;40:908-912.
14. Branda JA, Kratz A. Effects of yeast on automated cell
counting. Am J Clin Pathol. 2006;126:248-254.
15. International Council for Standardization in Haematology
Expert Panel on Cytometry and International Society of
Laboratory Hematology Task Force on Platelet Counting.
Platelet counting by the RBC/platelet ratio method: a
reference method. Am J Clin Pathol. 2001;115:460-464.
16. Johannessen B, Haugen T, Scott CS. Standardisation of
platelet counting accuracy in blood banks by reference
to an automated immunoplatelet procedure: comparative
evaluation of Cell-Dyn CD4000 impedance and optical
platelet counts. Transfus Apher Sci. 2001;25:93-106.
17. Hervig T, Haugen T, Liseth K, et al. The platelet count
accuracy of platelet concentrates obtained by using
automated analysis is influenced by instrument bias and
activated platelet components. Vox Sang. 2004;87:196-203.
18. Apelseth TO, Hervig T. In vitro evaluation of platelet
concentrates during storage: platelet counts and markers of
platelet destruction. Transfus Apher Sci. 2007;37:261-268.
19. Taylor FB Jr, Toh CH, Hoots WK, et al. Towards definition,
clinical and laboratory criteria, and a scoring system for
disseminated intravascular coagulation. Thromb Haemost.
2001;86:1327-1330.
© American Society for Clinical Pathology

20. Harrison P, Ault KA, Chapman S, et al. An interlaboratory
study of a candidate reference method for platelet counting.
Am J Clin Pathol. 2001;115:448-459.
21. Hickerson DH, Bode AP. Flow cytometry of platelets
for clinical analysis. Hematol Oncol Clin North Am.
2002;16:421-454.
22. Macey MG, Carty E, Webb L, et al. Use of mean platelet
component to measure platelet activation on the ADVIA
120 haematology system. Cytometry. 1999;38:250-255.
23. Springer W, von Ruecker A, Dickerhoff R. Difficulties in
determining prophylactic transfusion thresholds of platelets
in leukemia patients. Blood. 1998;92:2183-2184.
24. Bland JM, Altman DG. Applying the right statistics:
analyses of measurement studies. Ultrasound Obstet Gynecol.
2003;22:85-93.
25. Department of Health and Human Services. Centers
for Medicare and Medicaid Services Centers for Disease
Control and Prevention. Medicare, Medicaid, and CLIA
programs; laboratory requirements relating to quality systems
and certain personnel qualifications: final rule. Fed Regist.
2003;68:3639-3714.
26. Briggs C, Harrison P, Grant D, et al. New quantitative
parameters on a recently introduced automated blood cell
counter: the XE 2100. Clin Lab Haematol. 2000;22:345-350.
© American Society for Clinical Pathology
Hematopathology / Original Article
27. Chapman DH, Hardin J, Miers M, et al. Reduction of the
platelet review rate using the two-dimensional platelet
method. Am J Clin Pathol. 2001;115:894-898.
28. Sandhaus LM, Osei ES, Agrawal NN, et al. Platelet counting
by the Coulter LH 750, Sysmex XE 2100, and Advia 120: a
comparative analysis using the RBC/platelet ratio reference
method. Am J Clin Pathol. 2002;118:235-241.
29. Arroyo JL, Garcia-Marcos MA, Lopez A, et al. Evaluation
of a CD61 MoAb method for enumeration of platelets in
thrombocytopenic patients and its impact on the transfusion
decision-making process. Transfusion. 2001;41:1212-1216.
30. Keeney M, Brown W, Yee I. Platelet counts and flagging
rates from the LH 750 hematology analyzer compared with
the ICSH/ISLH platelet reference method and the Gen S
hematology analyzer. Lab Hematol. 2001;7:204-210.
31. Hong KH, Kim MJ, Lee KW, et al. Platelet count evaluation
using three automated haematology analysers compared with
the immunoplatelet reference method, and estimation of
possible inadequate platelet transfusion. Int J Lab Hematol.
2009;31:298-306.
Am J Clin Pathol 2010;134:634-647 647
647 DOI: 10.1309/AJCP88JYLRCSRXPP 647