1.3.1.1 Optimization of the extraction solvent
We selected 15µL whole blood as the sampling volume due to its universality in clinical use. In addition, whole spot-cut DBS approach was performed to effectively eliminate the effects in the analysis of NOACs from different HCT levels, spot volume, and different punch sites on the same spot. To optimize the extraction recovery and develop a simple analytical method, the extraction solvent types and extraction time were discussed below. On the basis of previous studies, methanol and acetonitrile were considered as the ideal solvent to extract plasma because our target drugs are among
doi:10.6342/NTU201803315
22
low to medium lipophilicity (logP -2.4 to 2.71). We firstly compared the two water-miscible organic solvents to extract the DBS samples [18, 55-57]. The result showed that the aqueous solution was necessary for extraction, but extraction efficiency of dabigatran was only about 80% owing to its high polarity, and thus resulting in powerful bonding between hydrogen bond of dabigatran and cellulose in a DBS card.
Although a higher aqueous proportion of extraction solvent could improve recoveries for all compounds in our preliminary tests, using high percentage of water is a significant disadvantage of denaturing the protein completely. Comparing with other organic solvents, ACN (70%) was selected as the organic component because of its less viscosity for operating easily. Furthermore, formic acid was added in 70% ACN to a final concentration of 0.1% due to its ability of breaking the hydrogen bond, and also contributing to protonation on the compounds for improving mass sensitivity. Figure 1.4 (p.37) indicated the recovery of dabigatran significantly increased with the addition of 0.1% formic acid. From a previous study, sonication has been used for extracting the apixaban in the DBS sample for 10min [45], on the contrary, we applied a Geno grinder for up-down agitation vigorously at 1000 rpm to speed up the extraction process. In our preliminary study, we found that performing agitation for 3 or 5min showed similar extraction yields for three compounds. It has been reported that higher HCT levels may cause the difficulty for extraction, therefore, we finally chose 5min as the extraction
doi:10.6342/NTU201803315
23
time [26, 58]. Also, the volume of extraction solvent was investigated, and the volume of 1mL was the optimal choice because of easy operation and immersing the DBS disks thoroughly.
Interestingly, we found the increasing of storage time significantly decreased the extraction efficiency of NOACs form the DBS cards especially for dabigatran which tends to form hydrogen bond with the DBS cards (Table 1.2, p.42). The extraction recovery was improved after pre-rinsing the card by 0.1% formic acid solution before extracting by acetonitrile.
1.3.1.2 Development of UHPLC-ESI-MS combined with PCI-IS method
UHPLC-ESI-MS is considered as a favorable analytical system for quantifying the drugs in the DBSs. According to previous studies of quantification of NOACs in plasma [18, 19], a 50mm-Kinetex C18 column was selected as the analytical column. The composition of sample solution showed significant effect on peak shape and signal intensity. We found that more water in the sample solution led to a decrease in signal intensity and ACN caused double peaks due to its relatively high strength, and thus we finally selected methanol as the sample solution. This study used the first ion suppression zone to estimate blood volume on DBS spot [38], and the initial elution strength was adjusted to separate polar small-molecule components from salts in plasma.
We observed that the initial 1% ACN provided an ideal separation for salts and
doi:10.6342/NTU201803315
24
dabigatran. The remaining 0.9mins of 15% ACN was found to improve the peak shapes and separation for apixaban and rivaroxaban. Apixaban and rivaroxaban are high lipophilic compounds, and they are co-eluted at 100% ACN with phospholipids in plasma as shown in Figure 1.5 (p.37) using precursor ions of m/z 184 [59]. The co-elution with phospholipid resulted in significant matrix effect caused signal changes.
To eliminate the effect of residual phospholipids, the procedures of solid-phase extraction (SPE) or liquid-liquid extraction (LLE) have been considered as sample pre-treatment following the solvent extraction step [45, 60], however, these procedures increased cost and method complexity. This study used PCI-IS to correct the signal changes caused by the matrix effect and the PCI-IS strategy was additionally used to estimate and adjust blood volume on DBS cards. The solvent type of PCI-IS was also assessed for enhancing the sensitivity for analyzing three NOAC drugs. It was interesting that ACN was beneficial for analyzing these three compounds, and the intensity of apixaban could be even enhanced when using ACN as the PCI-IS solvent (Figure 1.6, p.38). The mass parameters were optimized for additionally improving method sensitivity. Figure 1.7 (p.38) showed the chromatograms obtained under optimal conditions. Three NOAC drugs showed good peak shape and the sensitivity could reach sub-ppb level. The high sensitivity facilitated its application in clinical analysis of NOAC drugs on DBS.
doi:10.6342/NTU201803315
25
1.3.1.3 Selection of the optimal PCI-IS
In this study, we used PCI-IS strategy to estimate blood volume and calibrate matrix effect cased signal changes in ESI. Ideal PCI-ISs for calibrating matrix effect are the structural analogs of the target compounds possessing the same hydrophilicity, lipophilicity and ionization ability as target analytes [61]. Isotopically labeled internal standards (SIL-ISs) of dabigatran, apixaban and rivaroxaban were used as the PCI-ISs for comparison their blood volume estimation and matrix effect calibration performances. Firstly, the infusion rate and concentration of the PCI-IS were optimized to acquire an appropriate sensitivity and could clearly revealed the matrix effect. We found that the sensitivity of analytes was decreased while increasing the flow rate.
Finally, we selected the concentration and flow rate of the PCI-IS as 15 ngmL−1 and 0.1 mLmin−1, respectively.
1.3.1.3.1 Using the different PCI-IS to estimate Blood Volume on DBSs
The PCI-IS was firstly evaluated for its performance in estimation blood volume.
Using PCI-IS method, we measured the extend of ion suppression and built a linear correlation between the blood volume from DBSs (X axis) and the reciprocal of the minimum responses at the first ion suppression zone (Y axis). The ion suppression zone in high aqueous phase was mainly caused by nonvolatile salts including sodium, potassium, and the chloride ion in the blood, so we could estimate the blood volume by
doi:10.6342/NTU201803315
26
salts-induced ion suppression on the PCI-IS chromatogram. The blood volumes on DBS cards are generally between 10 to 20µL. Therefore, we generated a linear regression from 10 to 35µL and found the high correlations of calibration curves for [13C,D3]-apixaban and [13C6]-rivaroxaban. However, using [13C6]-dabigatran as PCI-IS showed no significant salts-induced ion suppression along with differential blood volumes. To prove its universality estimation performance, seven DBS samples obtained from different individuals were used to calculate the blood volume estimation accuracy by the pre-constructed calibration curves. Our results showed that the accuracies of [13C, D3]-apixaban and [13C6]-rivaroxaban for volume estimation were within 80.3% to 123.8% and within 87.1% to 119.8%, respectively (Table 1.3, p.43).
Hence, we chose these two PCI-ISs to further evaluate their correction performances for matrix effect caused signal changes.
1.3.1.3.2 Selection of the PCI-IS for improving quantification accuracy
We compared the quantitative accuracy and precision (RSD%) of target
compounds with and without correction by PCI-ISs using three DBS samples. The
correction efficiency was evaluated at three concentrations, and the comparison of
correction performances is presented in Table 1.4 (p.44). Prior to PCI-IS correction, we
found that the precisions for all analytes were higher than 20% at medium
concentrations, and the accuracies of rivaroxaban and dabigatran at three concentrations
doi:10.6342/NTU201803315
27
were lower than 40% and higher than 150%, respectively. While applying [13C, D3]-apixaban as PCI-IS for correction, the accuracies for rivaroxaban and dabigatran were not noticeably improved at low concentration. On the other hand, using
[13C6]-rivaroxaban as PCI-IS for correction significantly improved the accuracies for all of the NOACs to approximately 83-110% and RSD was less than 12%. On the basis of
the above accuracy and precision results, [13C6]-rivaroxaban was considered as an effective PCI-IS to correct the matrix effects for three NOACs and estimate blood
volumes on DBSs simultaneously.
1.3.1.4 Evaluation of the HCT effect
It has been reported that different HCT levels could affect spot volume on the filter paper as well as the drug distribution between blood and plasma. In addition, the extraction efficiency of examined drugs would generate potential assay bias with varying HCT values. Accordingly, Youhnovski et al. reported the recovery of naproxen depending on HCT values [58], and Kosteret al. indicated that different HCT affected the recovery of sirolimus and everolimus [62]. Although many studies demonstrated that the internal standard could be spotted or sprayed before extraction, these procedures were more complicated to carry out for clinical use [32, 63].
To investigate the influence of the HCT in quantitative analysis, HCT levels
between 25%, to 65% were manually prepared according to the procedures described in
doi:10.6342/NTU201803315
28
the method section.
To study the effect of HCT on the extraction recovery, we used our optimized extraction method to test whether the recovery would fluctuate with varying HCT values. Our results revealed that the overall recoveries were ranged from 81-102% at two tested concentrations. We also observed that the matrix effect of rivaroxaban slightly decreased with increasing the HCT values (Table 1.5, p.45).
Before studying the effect of HCT on quantification of drug concentration, we
evaluated the quality of our method in preparing samples with different HCT values.
Firstly, using our DBS sample preparation method, the overall percentage RSD of drugs
distribution in the whole blood. Table 1.6 (p.45) showed the % RSD of response ratios
of three drugs at all HCT levels were all less than 15% indicating that this method was valid for the following experiments. After confirming the preparation method, the HCT
of 45% was set as the standard HCT values to compare with other HCT values because it is close to the population mean [40, 64] and the biases were acceptable within 15%
[65, 66].
HCT effect on quantification accuracy was evaluated at two concentrations.
Figure 1.8 showed dabigatran with HCT level of 55% was observed a highest deviation
of 14.98%. Apixaban and rivaroxaban showed the highest deviation of 11.63% and
12.91%, respectively (Figure 1.8, p.39) at HCT of 65%. Nevertheless, the overall bias
doi:10.6342/NTU201803315
29
data was within acceptable limit of 15% bias, and the RSD values of corrected mean
peak area among different HCTs were lower than 9% (Table 1.6, p.45). Therefore, we
concluded that HCT range among 25%-65% showed no significant HCT effect on
NOAC quantification.
Some studies reported that the distributive phenomenon of drugs depends on its permeability [50, 67], for example, the highly protein-bound drugs located only a low percentage in erythrocytes. Moreover, immunosuppressive drugs were predominantly distributed in erythrocytes (>70 % for all) through immunophilins, a specific proteins in RBCs [68]. On the other hand, Emmons and Rowland proposed some underlying factors might affect the blood-to-plasma distribution including saturable protein binding or blood composition change [48]. And these characters would influence the design of bridging studies. Previous studies indicated apixaban and dabigatran showed no specific distribution in erythrocytes [45], and our results are similar to the previous observations.
1.3.2 Method validation 1.3.2.1 Selectivity
Selectivity of this method was assessed by analyzing six blank DBS from different sources. Under optimized method, no endogenous interferences in extracts overlapped significantly with the target analytes and internal standards at the same retention time.
doi:10.6342/NTU201803315
30
1.3.2.2 Recovery, matrix effect, precision, and accuracy
The developed method was validated by extracting the mixture of NOACs on the
DBS samples at LLOQ, low, medium, and high concentrations. The results were summarized in Table 1.7(A) (p.46). For the recovery of the extraction procedures, the results were ranged between 95.5 1.5 and 114.0 6.9%. However, the matrix effects were between 22.5 4.4 and 86.8 12.3 % indicating a significant ion suppression.
Apixaban and rivaroxaban were co-eluted with blood lipids and were suffered from significant MEs. The MEs for apixaban and rivaroxaban at three concentrations were ranged among 22.0-36.37% and 23.5-25.0%, respectively.
The precision and accuracy were evaluated at LLOQ, low, medium, and high concentrations. The within-run and between-run accuracies of three NOACs in DBS were ranged within 88.9-104.3%. The RSD values of repeatability and intermediate precision were all below 10%. According to the results, we concluded that our method can provide accurate quantification for measuring dabigatran, apixaban and rivaroxaban in DBS samples within the concentration range shown in this study.
1.3.2.3 Linearity, LLOQ and LOD
The mixture of NOACs-spiked DBS samples was used to build calibration curves and the linear rage was designed according to the therapeutic range in human plasma [11, 16]. The linear regression data was showed in Table 1.7(B) (p.46). The coefficients
doi:10.6342/NTU201803315
31
of determination (R2) values were greater than 0.99 for all the calibration curves. From these three analytes in DBS samples, the LLOQ were 0.3 ng mL−1 for rivaroxaban, and 0.6 ng mL−1 for both dabigatran and apixaban. The LLOQ was sufficient for therapeutic concentrations of the three NOACs. LOD was 0.03 ng mL−1 for rivaroxaban and 0.015 ng mL−1 for both dabigatran and apixaban.
1.3.2.4 Stability
Previous studies reported that the stabilities for three NOACs in whole blood were stable within acceptable range, but the concentration of dabigatran slightly decreased after 2 hours storage at 25 °C (near 80%) [69]. And the freeze-thaw stability of NOACs in plasma showed significant decreased in concentration for dabigatran and apixaban.
The stability study of NOAC in whole blood showed only low concentration of dabigatran slightly decreased (~85%) after 4 hours storage at RT (Table 1.8(A), p.47).
The long-term stability of low and high concentration of NOACs in DBS samples revealed all compounds were stable after one month storage at all studying temperature (Table 1.8(B), p.47). We therefore could conclude DBS sampling strategy could improve the stability of NOAC in blood samples.
1.3.2.5 Clinical sample analysis
In order to find the conversion factors and establish the relationships between
doi:10.6342/NTU201803315
32
NOACs concentrations in DBS and plasma samples, we collected the paired plasma and DBS samples (n=269) from patients undergoing NOACs treatment. The correlation plots for dabigatran (n=61), apixaban (n=115) and rivaroxaban (n=93) in DBS and plasma were shown in Figure 1.9 (p.40). The correlation analysis revealed three NOACs showed good linear relationship between DBS and plasma concentrations with the Pearson correlation coefficient (R2) being 0.979, 0.970 and 0.981 for dabigatran, apixaban, and rivaroxaban. The slopes of three calibration curves were within 1.3-1.7 which indicated the drug concentrations in DBS samples were lower than their respective paired plasma samples. The conversion factors were therefore incorporated for translation DBS concentrations to plasma concentrations [70, 71]. Conversion factors calculated from the ratios between plasma and DBS concentrations of dabigatran, apixaban and rivaroxaban were 1.81, 1.59 and 1.31, respectively. Figure 1.10 (p.41) shows the Bland Altman plots for three drugs. The mean differences between DBS and plasma concentrations for dabigatran, apixaban and rivaroxaban were 10.7, -6.1 and 3.6 ngmL-1 respectively. Moreover, over 90% of the calculated concentrations were within the 95% confidence interval (CI) of bias. Emmons and Rowland have indicated most high plasma protein binding drugs were predominantly distributed in plasma and the erythrocytes were served as the diluent in DBS samples [48, 67]. Our observation showed agreement with their study that apixaban and rivaroxaban which present
doi:10.6342/NTU201803315
33
relatively high protein binding (approximately 87% and 92% respectively) have the high conversion factor. However, dabigatran with low protein binding (approximately 35%) showed the highest conversion factor among three NOACs. Moreover, the result also showed agreement with the clinical pharmacological reports that three NOACs distributed very little to erythrocytes [72-74]. Because the red blood cell-to-plasma ratio might have differences between capillary blood and venous blood, we should further evaluate this factor in the future [70].