Simultaneous assessment of blood coagulation and hematocrit levels in dielectric blood coagulometry

2.1.Sample blood

This study was approved by the Ethics Committee of Tokyo Medical and Dental University. Blood samples were drawn into collection tubes with 3.13% sodium citrate solution (9:1 in volume) from healthy volunteers who consented to the study. In a previous study [7], patients scheduled to undergo unilateral or bilateral total knee arthroplasty at the university hospital were considered for enrollment, and written, informed consent was obtained from all enrolled patients. Citrated blood samples were obtained in the morning just before the surgery and a day after it, and DBCM measurements and routine blood tests were carried out. Blood collection was done at the same time, with different collection tubes for DBCM measurement and the routine blood tests. Portions of these data were re-analyzed in the present work.

The clinical part of the present study using data corresponding to the patients was performed in Tokyo Medical and Dental University, and the nonclinical part of the study was carried out at Sony Corporation.

2.2.Ca²⁺ concentration dependences

The principle of DBCM is summarized elsewhere [3]. Briefly, dielectric measurements were done using an automated blood coagulation analyzer prototype (Sony Corporation, Tokyo, Japan). This apparatus consists of an automated blood dispenser, sample cartridge holders with temperature controller, an impedance analyzer board, and a computer. The system was calibrated by open, short, and load (50 ohm) circuits. An operator set disposable sample cartridges and blood collection tube in the apparatus after inversion mixing of blood, and started measurements. A whole blood sample of 180 µl was dispensed into each cartridge with titanium electrode inserts as a parallel plate type capacitor, which is electrically connected to the impedance analyzer board through electric contact pins and coaxial cables. The dispensed sample was mixed and stirred with the reagent previously pipetted into the cartridge. As reagents, 12 µl of calcium chloride solution with various concentrations from 50 to 400 mM (final concentration in blood from 3 to 25 mM) were tested, where blood samples were obtained from three healthy subjects (ages: 30, 34 and 42 years, respectively). The complex impedance data of the sample at 37°C (errors within 0 and −1°C) were recorded over time up to 60 min with an interval of one min and frequency range from 1 kHz to 10 MHz and converted to complex dielectric permittivity (ε′–jε″, where j is the imaginary unit), where the effects of series inductance and stray capacitance of the whole circuitry were corrected for by the method reported by Asami et al. [23].

The effect of Ca²⁺ concentration was also tested in a limited number of experiments by rotational thromboelastometry ROTEM^® delta (TEM group, Basel, Switzerland), where calcium chloride solutions of 200, 250, and 308 mM (final concentration in blood: 12.5, 15.6 and 19.3 mM, respectively) were used as reagents instead of the original one supplied by the manufacturer. The sample temperature was controlled to be 37°C during measurements by ROTEM^® delta.

2.3.Analysis of rouleau formation

Previously observed DS data for citrated whole blood from a healthy subject (age: 44 years) were re-analyzed to investigate the rouleau formation process [3]. In the cited work [3], a change in dielectric permittivity accompanying rouleau formation was observed at 37°C without recalcification, so that coagulation did not occur. Specimen blood dispensed into the sample cartridge was agitated by manual pipetting movements. Measurements started within 5 s after agitation ended, and the dielectric spectrum was recorded over time with a time interval of 3 s up to approximately 10 min. After that, agitation and dielectric measurements were repeated to confirm both repeatability and reversibility.

2.4.Dielectric permittivity and hematocrit level in clinical study

DBCM data at 37°C obtained in the recent clinical study on patients who underwent total knee arthroplasty [7] were re-analyzed to compare dielectric permittivity and the hematocrit level obtained by a routine blood count test. The mean age of the patients was 73.7 years (SD = 7.6, n (preoperative) = 21 and n (postoperative) = 24).

2.5.In vitro hematocrit handling

Blood samples with different hematocrit levels were prepared as follows. Whole blood in collection tubes from a healthy subject (age: 34 years) was centrifuged with a mild condition (300g, 10 min), platelet rich autologous plasma was added or removed, and then it was mixed again. Blood coagulation tests at 37°C were performed by DBCM and ROTEM^®, where 200 mM calcium chloride solution was used as a reagent to achieve a final concentration of Ca²⁺ in blood of 12.5 mM. The hematocrit level and platelet number of the samples were measured by a blood cell counting device, PoCH100i (Sysmex Corp., Kobe, Japan). In general, ethylenediaminetetraacetic acid (EDTA) is used as an anticoagulant for evaluation of hematocrit level using a blood cell counting device, where a blood sample is not diluted by EDTA because it is dry reagent. In the present experiments, however, we measured the hematocrit levels of citrated blood samples prepared for DBCM and ROTEM^® tests. Therefore, we corrected the hematocrit levels to take into account the 10% dilution by sodium citrate.

2.6.Repeatability test

The blood sample from a healthy subject (age: 45 years) was measured at 37°C by the prototype system of DBCM seven times to estimate repeatability of the method, where calcium chloride solutions of 200 mM (final concentration in blood: 12.5 mM) was used as a reagent. The hematocrit level was measured by PoCH100i and corrected in the same way described in the previous Section 2.5.

3.1.Effects of reagent concentration in whole blood coagulation tests

A typical dielectric spectrum of human whole blood is shown in Fig. 1(a), where dielectric permittivity ε′ decreases with increasing frequency and shows two major decays. The decay observed in the lower frequency region is due to electrode polarization that results from accumulation of charge carriers and formation of electric double layers near electrode surfaces [12,24]. The decay in the higher frequency region is attributed to the interfacial polarization of erythrocytes, known as β-dispersion [12,17–19,22]. The dielectric spectrum changed with blood coagulation [3–7]. Figure 1(b) shows dielectric permittivity normalized by the first measured data at 1 min (ε/εt=1min) as a function of time and frequency. Among that data, the normalized permittivity at around 10 MHz increased with blood coagulation. This feature is useful to evaluate blood coagulability [3].

Fig. 1.

A typical dielectric spectrum (dielectric dispersion curve) of whole blood from a healthy volunteer before the blood coagulation process proceeds (a), and the change in normalized dielectric spectra during the progression of blood coagulation (b). The normalization was done with the dielectric dispersion curve at the first time point.

Figure 2 shows permittivity change at 10 MHz through the blood coagulation process that proceeds with arrows to indicate the characteristic time of coagulation (tx). It was found that errors of tx attributable to measurement variability is not negligible but still overcome by the variability induced by Ca²⁺ reagent concentration (Fig. 3). The observed U- or V-shape change of tx was essentially the same as that reported in plasma samples [8], although only limited data points were presently available at the low Ca²⁺ concentration region where tx decreased with increase of Ca²⁺ in Fig. 3. We note that tx increases rapidly if Ca²⁺ approaches to zero concentration in Fig. 3, because citrated blood does not coagulate without addition of Ca²⁺ in general. Unintended changes of effective Ca²⁺ concentration in clinical tests, therefore, may lead to an incorrect diagnosis of blood coagulability.

Fig. 2.

Dielectric permittivity changes at 10 MHz with blood coagulation for the same blood sample with different final concentrations of reagent (Ca²⁺) at 6.3 mM (black circles), 12.5 mM (blue squares), and 21.9 mM (red triangles). The permittivity data are normalized by the minimum and maximum values of the permittivity, and tx was determined as the intersection of two extrapolated lines (dashed lines).

Fig. 3.

Reagent (Ca²⁺) concentration dependences of tx obtained by DBCM (asterisks and circles) or clotting time (CT) by ROTEM^® (squares). The colors of the symbols indicate three individual subjects, where blue asterisks and blue circles show the data for the same subject but on different dates of experiments.

3.2.Rouleau formation and permittivity changes in view of hematocrit estimation

It is known that rouleau formation markedly affects β-dispersion [12,19]. In particular, the relaxation strength increases and the relaxation frequency shifts toward low frequencies, as shown in Fig. 4(a). At the same time, one can see that the permittivity at around 2 to 5 MHz did not change very much with rouleau formation. The effects from the increase of the relaxation strength and the shift of the relaxation frequency canceled out in this frequency region. In fact, the permittivity change by rouleau formation at 3 MHz (Fig. 4(c)) was unnoticeable in comparisons with the cases at 10 MHz (Fig. 4(b)) and 1 MHz (Fig. 4(d)). In contrast, the permittivity at around 2 to 5 MHz did increase with increased hematocrit level in a previously reported study [21]. Therefore, it was expected that the permittivity at a few MHz would be a good indicator of the hematocrit level for human whole blood without the disturbing effects of rouleau formation.

Fig. 4.

Rouleau formation and dielectric response observed in a previous study [3]. Panel (a) shows the dielectric dispersion curves just after pipet mixing and 5 min after. Panels (b), (c), and (d) show changes of the normalized permittivity at 10, 3, and 1 MHz, respectively. The normalization was done with permittivity at the first time point. The measurements of these changes were started just after mixing was stopped and started again after resuspension, three times in a row. The arrows indicate the points at which mixing was stopped. (Rearranged from [3] under the open access license granted by American Chemical Society.)

3.3.Hematocrit level and dielectric response in clinical studies

It was shown in the last Section 3.2 that permittivity at 3 MHz was apparently less influenced by rouleau formation. Therefore, previously obtained data for patients who underwent total knee arthroplasty [7] were re-analyzed to discuss the relevance of permittivity at 3 MHz for hematocrit estimation by comparison with the traditional conductance method.

The relationship between conductivity of a blood sample and the hematocrit level is shown in Fig. 5. As expected, the coefficient of correlation was not high (R2=0.76), and plots deviated from the theoretical curve by Hanai’s model (Eq. (2)) with a constant plasma conductivity of σb=1.6 S/m at 37°C [25]. This is not surprising, because Hanai’s model in Eq. (2) presupposes spherical cells without aggregation, but that is far from the reality. Moreover, patients may show large variations in plasma conductivity, but it is difficult to measure without centrifugation. Therefore, it has been confirmed that hematocrit estimation on the basis of the traditional conductance method is not suitable for human whole blood.

Fig. 5.

Conductivity of blood samples (σs) at 50 kHz observed 1 minute from the start of DBCM measurement against the hematocrit level for patients who underwent total knee arthroplasty [7] (full square). The solid line shows the result of linear regression where the square of the coefficient of correlation (R2) is 0.76. The dashed curve shows the relationship of conductivity and cell volume fraction in Hanai’s theory (Eq. (2)) assuming constant conductivity of the supernatant as σb=1.6 S/m. A plot for a healthy subject is also presented (full diamond) with error bar (SD = 0.065 S/m, CV = 9.8%, n=7) to show the repeatability.

In good contrast with Fig. 5, Fig. 6 shows a better correlation between permittivity at 3 MHz and the hematocrit level (R2=0.83). Using such a relation, rough estimation of the hematocrit level will be possible simultaneously with blood coagulation tests.

Fig. 6.

Dielectric permittivity of blood samples at 3 MHz observed 1 minute from the start of DBCM measurement against hematocrit level for patients who underwent total knee arthroplasty [7] (full square). The solid line shows the result of linear regression where the square of the coefficient of correlation (R2) is 0.83. A plot for a healthy subject is also presented (full diamond) with error bar (SD = 64.0, CV = 6.8%, n=7) to show the repeatability. Note that there can be systematic errors in permittivity values mainly due to uncertainty of the correction for stray capacitance. In the worst case, for example, the presented permittivity may be underestimated by 5%. Nevertheless, this effect was not critical to the correlation, because always the same systematic errors appeared even in this case.

3.4.Effects of hematocrit level on whole blood coagulation tests

Blood samples with different hematocrit levels prepared by in vitro handling of identical blood were measured by DBCM and ROTEM^®. In these experiments, the concentration of the reagent (Ca²⁺) was constant, and it was 12.5 mM in blood. However, the concentration of the reagent in plasma was dependent on the hematocrit level, for example, it was 25 mM in blood with hematocrit of 50% if the effects of Ca²⁺ adsorption and uptake by erythrocytes are insignificant. Figure 7 shows prolonging behavior of the characteristic coagulation time, tx (DBCM) and CT (ROTEM^®), with an increasing hematocrit level. In the present concentration range of Ca²⁺, tx and CT increased with increasing Ca²⁺ concentration, as demonstrated in Fig. 3. Therefore, it was considered that the hematocrit dependences of tx and CT in Fig. 7 originated from the increase of Ca²⁺ in plasma with the increase of the hematocrit level.

Fig. 7.

The characteristic coagulation time, such as tx in DBCM (stars) and CT for ROTEM^® (squares), are plotted against the hematocrit level, where platelet counts for the hematocrit-manipulated samples were almost constant between 12.6 and 13.8 × 10⁴/µl. To show the repeatability, a plot for a healthy subject in DBCM is also presented (full diamond) with error bar (SD = 2.7 min, CV = 11%, n=7).

Anderson et al. reported that introduction of thromboelastometry significantly decreased the use of red cells and blood products in cardiac surgery [2]. At the same time, the hematocrit level can be drastically changed during a major surgery. Therefore, the test results may be affected by changes of reagent concentrations in plasma, as shown in the present work.