Keywords
light transmission aggregometry - platelet function - precision - reproducibility
- reference ranges
Introduction
Platelets are essential in hemostasis.[1] Because congenital and, more often, acquired platelet disorders can cause bleeding,
assessing platelets is an important aspect in the laboratory work-up of patients presenting
with hemorrhages.[2] Whereas full blood counts and platelet parameters such as, median platelet volume,
are fully automated and therefore easy to deliver by the laboratory, platelet function
studies are requested much less frequently, often only after other hemorrhagic disorders
have been excluded.[3] This is most likely a consequence of the technical challenges associated with platelet
function studies, but probably also the lack of standardization, and the difficulties
in interpreting the results.[4]
[5]
[6]
[7]
[8] Although newer methods are in the market, many experts consider light transmission
aggregometry (LTA) as the gold standard for the assessment of platelet function, introduced
by Born and O' Brien in 1962.[9]
[10] LTA determines platelet aggregation in platelet-rich plasma (PRP) by assessing the
change in light transmission in response to added specific platelet agonists. Under
constant stirring, agonists, through the activation of specific receptors, can prompt
platelet granule secretion, activation, and aggregation. The reactivity pattern with
various agonists acting through different receptors and signaling pathways helps to
establish a diagnosis.[11] However, limitations of LTA are obvious: it requires large volumes of blood for
the preparation of PRP and platelet-poor plasma (PPP), which is necessary for calibrating
the measurement zero. Numerous preanalytical variables including, venipuncture, centrifugation,
adjustment of platelet count, and limited sample stability may affect the test outcome.[12]
[13]
[14] Instrument-related variables such as stirring intensity, physical properties of
the optical system, and software settings used to calculate the various parameters
of turbidity signature, as well as type, quality, and concentration of agonists further
affect the test outcome. In addition, LTA remains a time-consuming test method because
it includes multiple manual steps, which are also potential source of errors.
An expert panel of the Platelet Physiology Scientific and Standardization Committee
(SSC) of the International Society on Thrombosis and Haemostasis (ISTH) and others
have proposed steps for standardization.[14]
[15]
[16]
[17]
[18]
Like in other areas of laboratory testing, automation could play a key role in improving
the standardization of the procedure. A novel fully automated device for LTA, Thrombomate
XRA (TXRA), was developed recently with the goal of minimizing variables. A novel
feature is the ability of TXRA to measure LTA in PRP not only against autologous PPP
but also against a virtual reference. In this study, the performance characteristics
of the TXRA device were compared against a standard LTA instrument using fixed concentrations
of agonists of the identical source on both instruments.
Materials and Methods
Blood Samples
The study was conducted at Giessen University Hospital, Giessen, Germany. Anonymized
left-over material from healthy blood donors, patients with suspected hemostatic disorders,
and patients on aspirin and/or adenosine diphosphate (ADP) receptor antagonists was
used. Approval from the local ethics committee was obtained. There was no patient
selection process. If LTA testing was requested from the laboratory and appropriate
amounts of material were available to run both assays, subsequent samples obtained
from patients taking antiplatelet drugs or from patients with a suspected bleeding
disorder were included. Blood was drawn with Safety–Multifly cannulas (Sarstedt, 21Gx3/4''TW,
0.8 × 19 mm) into 10 mL Sarstedt Monovettes with 0.106 mol/L Na-citrate (9 vol blood + 1
vol anticoagulant). PRP was obtained by centrifugation for 10 minutes at 150 g, and
PPP was obtained by centrifugation for 20 minutes at 1,500 g. PRP was allowed to rest
for 30 minutes prior to analysis. All function studies were performed without prior
adjustment of platelet counts.
Instruments and Reagents
Test samples were run in parallel on two instruments, the automated TXRA (Behnk Elektronik,
Norderstedt, Germany) and the laboratory's standard instrument, PAP-8 (möLab, Langenfeld,
Germany). The same technical assistant performed all preparation and measurement steps
on both instruments throughout the study.
TXRA is a stand-alone analyzer for LTA, designed for the use either with or without
PPP as the reference. TXRA uses CE-marked reagent combinations provided as a unit.
In this study, ADP, arachidonic acid (ARA), epinephrine (EPI), collagen (COL, fibrillary
collagen from horse tendon), thrombin-receptor activating peptide (TRAP) and ristocetin
(RISTO) were used. Identical reagents and reagent concentrations were used on both
instruments. Concentrations were in accordance with the SSC/ISTH recommendations ([Table 1]).[16] Low RISTO (0.6 mg/mL) was included for the characterization of suspected type 2B
von Willebrand syndrome (VWS) or platelet-type VWS. TXRA automatically counts down
the recommended resting time of PRP after centrifugation.[16] Subsequently, PRP is inverted in the closed sample tube for gentle standardized
homogenization before it is automatically dispensed after cap-piercing into prewarmed
cuvettes. The automated addition of reagents starts the test. A mechanically added
steel ball mixes the sample. Using TXRA standard settings, the reaction is followed
simultaneously in five measuring channels for 6 minutes by high-precision bichromatic
(620 and 405 nm) light-emitting diode (LED) optics. The dual wavelength LED optical
system is also used for checking samples for potential interferences induced by hyperbilirubinemia,
lipemia, or hemolysis.
Table 1
Precision data of MA%
|
TXRA
|
PAP-8
|
|
Reagent
|
Mean CV
|
Range
|
Mean CV
|
Range
|
|
TRAP (10 µM)
|
2.1
|
1.1–3.0
|
2.1
|
1.1–3.0
|
|
ADP (2.5 µM)
|
1.6
|
0.9–2.0
|
3.7
|
2.6–4.2
|
|
COL (2 mg/mL)
|
1.4
|
0.7–1.9
|
3.0
|
1.0–5.0
|
|
ARA (1 mM)
|
4.6
|
2.6–6.7
|
3.3
|
1.3–6.2
|
|
EPI (5 µM)
|
2.2
|
1.3–3.9
|
2.8
|
1.9–3.7
|
|
RISTO (1.2 mg/mL)
|
2.1
|
1.2–3.1
|
1.7
|
0.6–2.8
|
|
CV (%)
|
2.8
|
|
3.3
|
|
|
SD
|
1.14
|
|
4.6
|
|
Abbreviations: ADP, adenosine diphosphate; ARA, arachidonic acid; COL, collagen; CV,
coefficient of variation; EPI, ephinephrine; MA%, maximum aggregation; PPP, platelet
poor plasma; RISTO, ristocetin; SD, standard deviation; TRAP, thrombin receptor activating
peptide (Ser-Phe-Leu-Leu-Arg-Asn); TXRP, Thrombomate XRA.
Note: Precision was calculated using five different donors and fivefold determination
for each reagent.
On TXRA, the maximum aggregation in percent for a sample is calculated according to
the formula:
with E
base = basic absorbance and E
PPP = absorbance of platelet-poor plasma.
The slope is an interpolated value that describes which aggregation value would be
reached if the aggregation was continued as at the point with the maximum slope.
The slope m of the two absorbance values Ea
and Eb
is calculated according to the formula Eslope
= E + (60 * m) with
where E is the mean absorbance (E) and Ta
and Tb
are the time points of the two measured values, whereas E is the measured value in the middle between Ta
and Tb
and can be calculated by searching the value with the nearest time point to Ta
+ (Tb
− Ta
)/2 is found.
Virtual Platelet-Poor Plasma Reference
The LED optical system on the TXRA was further employed for calculating the “virtual
platelet-poor plasma” (VPPP) optical properties by a novel proprietary algorithm based
on artificial intelligence. The VPPP value is used to “blank” the individual PRP,
thereby eliminating the requirement for autologous PPP. In this study, all measurements
were made against autologous PPP. The results of the VPPP method were calculated from
the stored PRP reaction data in the database.
Statistics
Precision testing was performed with five different healthy control samples, reference
values were assessed using 100 different healthy control samples, and a comparison
of patient data was performed using 47 different patient samples. Data were analyzed
with MS Excel and Abacus statistical software (LABanalytics GmbH, Jena, Germany).
Method comparisons were made with the method of Passing and Bablok or by linear regression.
Correlation coefficients were calculated according to Pearson.
Results
Precision
Precision testing comprised the fivefold analysis of PRP from five different individuals
with all reagents on both instruments. The results of the maximum aggregation in percent
(MA%) are summarized in [Table 1]. The mean CVs of the individual reagents ranged from 1.4 to 4.6% for the MA% value.
The highest CV for TXRA observed in a series of 6 × 5 tests on five individual PRP
samples with all reagents was 6.7% with ARA in one sample. The mean CV over all reagents
and samples in this series was 2.8% on TXRA and 3.3% on PAP-8. TXRA generates very
reproducible results with all reagents, but also the manual method performed by an
experienced technician and with the instrument and reagents used in this study indicates
high reproducibility. Precision values for slope and area under the curve (AUC) are
given in [Table 1] in the supplement. Mean CV over all reagents and samples was 6.9% for TRXA and 7.5%
on PAP-8 for slope, and 3.2 versus 2.9% for AUC, respectively, indicating similar
precision for these calculated results. Example readings from TXRA for all agonists
are presented in the supplement ([Fig. 1]).
Fig. 1 Maximum aggregation (MA%) of normal blood donors. (A–C) Distribution of MA% values obtained in (A) 100 healthy blood donors, separated in (B) n = 50 females and (C) n = 50 males on TXRA with PPP as a reference. Box-and-whisker diagrams show minimum,
lower quartile, median, upper quartile, and maximum values. Outliers are indicated
by circles. Crosses indicate data points outside the 2SD-range, (D) method comparison (Passing–Bablok analysis) for MA% values between TXRA and PAP-8
of 100 healthy blood donors (50 f/ 50 m).
Reference Range
The normal ranges for all reagents were tested in PRP samples from 50 female and 50
male healthy blood donors, aged between 18 and 65 years. All donors were free of medication,
had no history of hemostatic problems, and their full blood counts were normal. Their
results for MA% are summarized in [Table 2]. Mean and median values were similar for all reagents on TXRA. The MA% results on
TXRA obtained in male donors were slightly higher than in females, and also the distribution
was somewhat wider ([Fig. 1A-C]), but differences did not reach statistical significance. Results obtained on TXRA
showed upper (+2 standard deviation [SD]) and lower (−2SD) limits of the reference
range that were slightly higher than on PAP-8, but method comparison shows a linear
relationship with excellent correlation ([Fig. 1D]). The distribution for each platelet agonist is shown in [Fig. 2]. Mean and median values were very similar in normal blood donors. TRAP 10 µM lead
to almost complete aggregation on the TXRA in many normal samples. All other reagents
showed a normal distribution. Low RISTO (0.6 mg/mL) gave values between 2 and 11.5
MA% (2SD range) on the TXRA, but 0 to 5% (2SD range) on the PAP-8 (data not shown).
This probably reflects minor differences in reactivity by different mechanical features
of the two instruments or other factors.
Fig. 2 Normal range distribution histograms (TXRA, MA%). The distribution of MA% values
obtained on TXRA is shown for MA% for all reagents used in this study.
Table 2
Reference ranges for MA%. All healthy individuals (n = 100)
|
Thrombomate XRA
|
PAP-8
|
|
Reference: PPP
|
Reference: VPPP
|
Reference: PPP
|
|
All
|
All
|
|
Median
|
Mean
|
2SD range
|
Median
|
Mean
|
2SD range
|
Median
|
Mean
|
2SD range
|
|
ADP
|
92.2
|
91.6
|
82.2–101.0
|
87.8
|
87.2
|
82.7–91.6
|
82.0
|
81.5
|
69.1–93.9
|
|
ARA
|
85.7
|
86.1
|
72.3–99.8
|
82.4
|
82.8
|
77.5–88.0
|
82.5
|
82.8
|
70.5–95.0
|
|
COL
|
93.8
|
93.5
|
85.4–101.7
|
88.7
|
87.8
|
83.8–91.8
|
82.5
|
83.7
|
71.1–96.3
|
|
EPI
|
91.7
|
91.2
|
82.0–100.4
|
87.6
|
86.4
|
82.1–90.6
|
82.0
|
81.1
|
68.1–94.1
|
|
RISTO 0.6
|
6.4
|
6.7
|
2.0–11.5
|
4.4
|
4.7
|
2.8–6.8
|
1.0
|
1.7
|
0–4. 8
|
|
RISTO 1.2
|
95.8
|
95.4
|
86.3–104.5
|
90.1
|
89.5
|
85.8–93.1
|
80.0
|
80.5
|
70.5–90.4
|
|
TRAP
|
91.6
|
91.1
|
82.0–100.3
|
88.1
|
90.1
|
81.6–91.5
|
81.0
|
81.0
|
68.4–93.6
|
Abbreviations: ADP, adenosine diphosphate; ARA, arachidonic acid; COL, collagen; EPI,
ephinephrine; MA%, maximum aggregation; PPP, platelet poor plasma; RISTO, ristocetin;
SD, standard deviation; TRAP, thrombin receptor activating peptide (Ser-Phe-Leu-Leu-Arg-Asn);
VPPP, virtual platelet poor plasma.
Table 3
Reference ranges for MA%. Gender-specific normal ranges (TXRA, against PPP)
|
TXRA
|
Females (n = 50)
|
Males (n = 50)
|
|
Median
|
Mean
|
2SD range
|
Median
|
Mean
|
2SD range
|
|
ADP
|
92.4
|
91.8
|
83.2–100.4
|
91.8
|
91.5
|
81.2–101.7
|
|
ARA
|
85.1
|
86.2
|
72.6–99.7
|
86.0
|
86.0
|
72.0–100.0
|
|
COL
|
93.6
|
93.5
|
86.1–100.9
|
93.9
|
93.5
|
84.6–102.5
|
|
EPI
|
91.7
|
91.7
|
83.9–99.5
|
91.4
|
90.8
|
80.4–101.2
|
|
RISTO 0.6
|
6.2
|
6.7
|
1.5–11.9
|
6.5
|
6.7
|
2.5–11.0
|
|
RISTO 1.2
|
95.4
|
95.7
|
87.9–103.4
|
96.0
|
95.2
|
84.9–105.6
|
|
TRAP
|
91.4
|
91.1
|
81.5–100.6
|
92.0
|
91.1
|
82.3–100.0
|
|
PAP-8
|
|
ADP
|
80.0
|
80.6
|
69.4–91.8
|
82.5
|
82.5
|
69.2–95.8
|
|
ARA
|
81.0
|
81.4
|
72.4–90.3
|
83.5
|
84.1
|
69.7–98.6
|
|
COL
|
82.0
|
82.7
|
72.8–92.7
|
83.0
|
84.7
|
70.0–99.3
|
|
EPI
|
80.0
|
80.1
|
68.0–92.2
|
82.0
|
82.1
|
68.5–95.8
|
|
RISTO 0.6
|
1.0
|
1.5
|
0–4.6
|
1.5
|
1.8
|
0–4.9
|
|
RISTO 1.2
|
80.0
|
79.8
|
72.2–87.4
|
81.0
|
81.1
|
69.4–92.9
|
|
TRAP
|
79.
|
79.7
|
69.8–89.7
|
81.0
|
82.2
|
67.8–96.6
|
Abbreviations: ADP, adenosine diphosphate; ARA, arachidonic acid; COL, collagen; EPI,
ephinephrine; MA%, maximum aggregation; PPP, platelet poor plasma; RISTO, ristocetin;
TRAP, thrombin receptor activating peptide (Ser-Phe-Leu-Leu-Arg-Asn); TXRP, Thrombomate
XRA; VPPP, virtual platelet poor plasma.
Patient Data with Autologous Platelet-Poor Plasma
The individual patient data ([Fig. 3]) indicated principal agreement between TXRA and PAP-8 for both slope and MA% values
for all reagents. The closest correlation for MA% was achieved for ARA (r
2 = 0.978), and the correlation for COL, ADP, and RIS was also satisfactory (r
2 = 0.746, 0.604, and 0.516, respectively). EPI and TRAP showed some scatter in individual
paired data. As expected, RIS 0.6 showed only marginal responsiveness in all except
one patient. This patient with Von Willebrand disease type 2B showed significant reactivity
with RIS 0.6 (PAP-8: 81% MA, TXRA: 95.8% MA) on both systems, with 87 and 100% MA
with RIS 1.2. Of n = 20 patients taking aspirin, all 20 were identified on PAP-8 and TXRA. Their median
MA% was 2 (range, 0–7) on PAP-8 and 5.95 (range, 3.3–14.3) on TXRA for ARA. In total,
out of 47 patients considered to have a platelet function defect, 42 had at least
one test below the 2SD range on PAP-8, and 42 had at least one test below the 2D range
on TXRA. Of these, 40 patients were identified with both methods.
Fig. 3 Method comparison of individual reagents between TXRA and PAP-8 in 47 patients: MA%
and slope values. (A–F) maximum aggregation values in percentage, and (G–L) slope values (in instrument specific dimensions). Y-axis shows PAP-8, and X-axis shows TXRA results. R
2, Pearson's correlation coefficient. Results obtained with RISTO 0.6 are not shown
graphically because of the weak signal generated.
Passing–Bablok analysis demonstrates a linear relationship over the whole range of
MA% data.
[Fig. 4] shows the combined method comparison of normal and patients. Bland–Altman analysis
shows no identity, but very similar diagnostic information with both systems. Method
comparison for area under the curve and slope is presented in the supplement, [Fig. 2].
Fig. 4 Method comparison using MA% for n = 47 patients and n = 100 healthy individuals tested on TXRA against PAP-8 with identical reagents. (A) Passing–Bablok analysis and (B) Bland–Altman plot. Data represent 1,029 individual results from healthy controls
(n = 100, seven reagents) and 47 patients (n = 47, seven reagents).
Results Obtained with Virtual Platelet-Poor Plasma
TXRA employs artificial intelligence to reconstruct the optical properties of a fictive
autologous PPP, called the VPPP. This approach allowed to recalculate all results
that have been obtained against autologous PPP retrospectively against VPPP based
on PRP data stored in the database. This virtual PPP corrects the spectral properties
of a PRP an added reagent that it resembles autologous PPP by analysis of its spectral
properties according to a novel complex proprietary algorithm. The analysis of healthy
controls (n = 100) and patients with confirmed or suspected platelet function defects (n = 47, [Fig. 5]) demonstrates that there is little scatter between the set of data obtained with
PPP and VPPP. Calculating the reference ranges with VPPP in comparison to PPP shows
very similar ranges than obtained against PPP ([Fig. 5 B]), however with a few percent lower MA% values.
Fig. 5 Comparison between TXRA PPP (= traditional LTA) against the VPPP method (without
autologous PPP). Individual results from 47 patients tested with seven different reagents
(329 individual data points) are shown with either PPP or VPPP as the reference. Pearson's
correlation coefficient (R
2) is 0.849.
Discussion
TXRA is an automated LTA system reporting results that correlate well with an established
manual method. LTA is typically performed as a single determination, demanding high
precision. In our study, TXRA showed excellent precision with all reagents, comparable
to data obtained on the manual instrument, for which the coefficient of variability
(CV) was only slightly higher. The consistent low CV on the manual instrument was
probably related to the fact that one individual, very experienced technician operated
it throughout the study, strictly following standard operation procedures with respect
to PRP resting time, controlled PRP inverting before testing, and skilled pipetting.
Precision for manual LTA obtained in this study may not be representative for laboratories
in which several and/or less specialized technicians are involved, which likely leads
to slightly divergent approaches to the test. In contrast, the automated device offers
permanently standardized operation conditions. Another potential contributing factor
for excellent CVs may also be the use of preset, fixed concentrations of agonists,
which avoids variability due to predilution steps or freeze-and-thaw cycles. Good
precision was also reported for LTA automated on coagulation analyzers.[19] Our precision data confirm that LTA represents a highly reproducible method if external
influences are reduced to a minimum. A potential limitation of our data is that all
samples for the precision study were from healthy blood donors and did not include
abnormal platelet counts or clinically relevant, patient-related confounded variables
such as, hyperlipidemia, hyperbilirubinemia, hemolysis, or low von Willebrand factor.[20]
Our analysis of results focuses primarily on the most widely used parameter, maximum
aggregation in percent (MA%). Other calculated parameters of the aggregation signatures
were thoroughly inspected. In most cases, turbidity signatures were very similar,
with some variability in individual cases, which we especially observed with EPI and
TRAP (see supplement, [Table 1]). Unfortunately, comparing quantitative disintegration data was not expedient since
observation time at TXRA was preset (6 minutes in this study) and clearly shorter
than on the PAP-8 device (15 minutes). However, the observation can be prolonged also
on TXRA for detecting late disintegration. Rapid disintegration in a patient sample
is of course visible already in the 6 minutes observation period (see example in the
supplement, [Fig. 1]).
The normal range for MA% showed a tendency to slightly higher values obtained with
TXRA. MA% for all reagents showed normal distribution and minor variances between
males and females, not reaching statistical differences. Low RISTO (0.6 mg/mL) displayed
little differences between the TXRA and PAP-8. While PAP-8 did not generate measurable
aggregation in several normal samples, TXRA showed explicit values in the range between
2 and 11.5% MA. This may be related to the mechanical factors: while PAP-8 works with
round cuvettes and stir bars, TXRA has a flat cuvette, and mixing is achieved by lateral
movement of a steel ball by magnetic forces on a rail-like structure on the bottom
of the cuvette. The difference, however, could also be caused by the individual calculation
algorithms for MA% used in the two devices. It remains open if an apparently enhanced
resolution in this range by TXRA supports the analysis of patients with von Willebrand's
disease, specifically in type 2B or platelet type von Willebrand, or in Bernard Soulier
syndrome. Results in one patient with VWS type 2B with both systems were comparable.
Aggregometry results from unselected patients in general showed good correlation.
The degree of correlation varied depending on the agonists used. The agreement between
TXRA and the manual device with several agonists was very close for slope, AUC, and
MA%, while there was some scatter with EPI and TRAP. For these two agonists, results
showed a very similar trend but are not identical. This may reflect minor differences
between the mechanical factors, the optical systems, and software in the two systems.
Differences in EPI and TRAP between TRXA and the manual device were not observed in
healthy subjects and may reflect a different sensitivity to minor changes in platelet
reactivity. A time-dependent change of reactivity in the samples, when not tested
at exactly the same time on both devices, may contribute. A previous study between
two instruments of the same brand as the manual system used in this study showed differences
as well.[21] Since there is no established reference method for aggregometry, results cannot
be checked for “true” accuracy.
A novel aspect of this study is performing LTA without a reference by measuring PRP
against PPP. Interestingly, we could demonstrate here that TXRA enables LTA without
the use of autologous PPP by implementing an artificial intelligence-based approach
on TXRA. Results correlate strongly with the classical method with autologous PPP.
Beyond a relevant reduction in workload, this novel feature has also the potential
of reducing total blood volume, an advantage for children, but also for patients in
general.
In conclusion, this study shows that full automation of LTA with TXRA is feasible,
precise, and even possible in the absence of autologous PPP.
What is Known on this Topic?
-
Light transmission aggregometry is considered the gold standard for platelet function
testing.
-
It is a labor-intensive test involving multiple manual steps.
-
Additional platelet-poor plasma is required to calibrate the instrument.
What does this Paper Add?
-
A fully automated light transmission aggregometry device delivers precise measurements.
-
Automated aggregometry results correlate well with the manual method and make the
test almost user independent.
-
Instead of platelet-rich plasma, a virtual reference via artificial intelligence can
be used to calibrate the instrument and run the tests.
