Determination of piperaquine concentration in human plasma and the correlation of capillary versus venous plasma concentrations

Background A considerable challenge in quantification of the antimalarial piperaquine in plasma is carryover of analyte signal between assays. Current intensive pharmacokinetic studies often rely on the merging of venous and capillary sampling. Drug levels in capillary plasma may be different from those in venous plasma, Thus, correlation between capillary and venous drug levels needs to be established. Methods Liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS) was used to develop the method. Piperaquine was measured in 205 pairs of capillary and venous plasma samples collected simultaneously at ≥24hr post dose in children, pregnant women and non-pregnant women receiving dihydroartemisinin-piperaquine as malaria chemoprevention. Standard three-dose regimen over three days applied to all participants with three 40mg dihydroartemisinin/320mg PQ tablets per dose for adults and weight-based dose for children. Correlation analysis was performed using the program Stata® SE12.1. Linear regression models were built using concentrations or logarithm transformed concentrations and the final models were selected based on maximal coefficient of determination (R2) and visual check. Results An LC-MS/MS method was developed and validated, utilizing methanol as a protein precipitation agent, a Gemini C18 column (50x2.0mm, 5μm) eluted with basic mobile phase solvents (ammonium hydroxide as the additive), and ESI+ as the ion source. This method had a calibration range of 10–1000 ng/mL and carryover was negligible. Correlation analysis revealed a linear relationship: Ccap = 1.04×Cven+4.20 (R2 = 0.832) without transformation of data, and lnCcap = 1.01×lnCven+0.0125, (R2 = 0.945) with natural logarithm transformation. The mean ratio (±SD) of Ccap/Cven was 1.13±0.42, and median (IQR) was 1.08 (0.917, 1.33). Conclusions Capillary and venous plasma PQ measures are nearly identical overall, but not readily exchangeable due to large variation. Further correlation study accounting for disposition phases may be necessary.


Summary of changes
This method with a lower calibration range of 0.5-50ng/mL is a modification of the previous CPQA approved LC/MS/MS method: both methods use the same sample preparation and API5000-LC-MS/MS system; both utilize PFP-based analytical column and the same mobile phase solvents.
The changes are summarized in

Determination of Piperaquine in Human Plasma by LC-MS/MS on API5000 Assay Modified for Lower Calibration Range
Piperaquine (PQ) is an antimalarial drug used in preventive treatment and combination therapy, because of its long elimination half life (t 1/2 = 7days based on our unpublished data). It is a hydrophobic aromatic heterocyclic compound with multiple nitrogen atoms. PQ is not soluble in water and most organic solvents but soluble in acidified solvents [1].
Previously we developed a method to quantify PQ with API5000 LC-MS/MS system. The assay uses deuterated PQ (PQ-d 6 ) as the internal standard (IS), and 25 μL plasma sample was used for this analysis. The LLOQ could reach 0.5ng/mL. However, due to carryover issue, the LLOQ increased to 1.5ng/mL and calibration range was 1.5-250ng/mL [2]. During our initial analysis of 380 clinical samples from a malaria chemoprevention study with dihydroartemisininpiperaquine, 76 samples (20%) were below the LLOQ 1.5ng/mL, while only one sample was above 50ng/mL. Thus we determine to modify the method to a new calibration range of 0.5-50ng/mL to support this specific study.
According to CPQA guidelines[3], a partial validation should be performed if the concentration range is changed. The required validations are as follows: 1), three inter-/intra-day precision and accuracy, including the new LLOQ in the validation QC samples. 2), Matrix effect and recovery.

Principle of the Method
PQ and the internal standard (IS) PQ-d 6 are extracted from EDTA plasma by protein precipitation with 5% TCA in MeOH-water (1:1, v/v). The processed sample is injected onto a Pursuit PFP column (2.0 × 50 mm, 3 µm) eluted with 20 mM NH 4 FA 0.14% TFA (pH 2.96) and acetonitrile with 0.1% TFA in a gradient mode. APCI and multiple reaction monitoring (MRM) were used, and ion pairs 535/288 for PQ and 541/294 for the IS were selected for quantification, and 535/260 for PQ was used for confirmation. The retention times are typically 1.0(±0.2) min for PQ and 1.0(±0.2) min for the IS. Total run time is 3.0 min per sample. The LC-MS/MS system is operated at room temperature (25 ± 2 °C controlled by an air conditioner). Calibration curve standards and quality controls are prepared in blank EDTA human plasma from separately weighted and prepared PQ solutions. For calibration curves, spiked concentrations and peak area ratios of PQ to the IS are fitted by linear least squares regression, weighted by 1/x. Plastic sample vials should be used instead of glass due to absorption of PQ to glass surface.

Calibration Curve
Calibration standards, prepared in EDTA human plasma, consisting of 0.500, 1.00, 1.50, 5.0, 10.0, 25.0 and 50.0 ng/mL of PQ, were used to establish the calibration curve for assay validation. A single curve was plotted using concentration vs. mean peak area ratio obtained by injecting a set of calibrators in the beginning of the batch run. Linear least squares regression with 1/ x weighting of the calibration plot resulted in correlation coefficients [r] greater than 0.995 for all three assays [ Table 1]. The mean precision and percent deviation of the calibrations standards over 3 days, ranged from 2.4-14.6 % and (-8.0)-7.3%, respectively [ Table  1]. A representative calibration curve is shown in Figure 1. Representative chromatograms for blank plasma, the lower limit of quantification (LLOQ), and blank plasma after ULOQ are shown in Figure 2.
Inter-assay (between day) and Intra-assay (within day) Precision and Accuracy Precision is the degree of reproducibility; it characterizes the degree of agreement among a series of individual measurements. Precision is calculated as the coefficient of variation (%CV). Accuracy is the degree of correctness and is expressed as the percent deviation from the true concentration value. For inter-assay precision and accuracy, 6 replicates of validation samples, made from the same homogeneous matrix volume, at each of 3 different drug concentrations were analyzed on 3 separate days. A fresh calibration curve was used each day. The validation samples were made up with PQ concentrations of 1.50 ng/mL, 20.0 ng/mL, and 40.0 ng/mL. These were designated low, median, and high, respectively.
The inter-assay precision (CV %) of this method for PQ was 6.4, 5.1and 4.8 % for low, median, and high concentrations respectively. The overall accuracy (%dev) from nominal concentration value was 6.4,( -1.1), and 2.6% for low, median and high concentrations respectively [ Table 2].

Lower Limit of Quantification
Six replicates of validation samples at the lowest calibration concentration (0.500 ng/mL) were analyzed on 3 days to determine the inter-and intra-assay precision and accuracy of this lowest point on the calibration curve. The inter-assay precision (%CV) was 8.1 % and percent deviation was 6.8 %. The intra-assay %CV for the mean of these 3 replicate days ranged from 3.98-11.1% and the mean accuracy (%dev) from 2.30-10.0% [ Table 2].

Matrix Effect, Recovery, and Process Efficiency
Matrix effect According to CPQA guidelines Appendix F, five different lots of blank EDTA human plasma (#00023-35504, #00023-35491, #0002335479, #LS23-73766, #LS2356244) were used for matrix effect test. To evaluate matrix effect of hemolized plasma, 2 aliquots of plasma spiked with 1% and 5% whole blood (LS2356244 with 1% blood and LS2356244 with 5% blood) freeze and thaw for 2 cycles to lyse the red blood cells. Then matrix effect experiments were performed with the 5 normal plasma lots.
Four sets of samples at three concentration levels (low, median and high QC) were prepared and analyzed to determine matrix effect, recovery, and process efficiency. Un-processed PQ without matrix but with IS (PQ-d 6 ): 10 µL PQ (6, 80, or 160ng/mL) and 10 µL PQ-d 6 (8ng/mL) were spiked into 180 µL water-MeOH (1:1/v/v) containing 5%TCA. Set 2: Post extraction spike into extracted blank plasma (defines absolute and relative matrix effect). Each of the 7 lots of blank plasma samples (150 µL) was mixed with 600 µL water-MeOH (1:1/v/v) containing 5%TCA, vortexed for 10s and centrifuged for 3min at 25,000g. Aliquots of 180 µL supernatant were taken and spiked with PQ and IS (PQ-d 6 ) (10 µL each) to make the final concentration equal to those in set 1b. Set 3: Pre extraction spike into plasma then extract (defines recovery and overall process efficiency). PQ was spiked into plasma at 1.5ng/mL, 20ng/ml, and 40ng/mL in each of 7 lots of plasma, then extracted as described in SOP.
Set 1a and 1b consist of 7 replicates of injections at each concentration. Set 2 and 3 consist of 7 different sources of plasma at each concentration level. Injection was performed in the order of set1a-L1, set1b-L1, set2-L1, set3-L1, then repeat the order for matrix #2, 3, 4, 5, 6, and 7. Repeat the process for median-and high-validation samples. Data and comparisons are presented in table 3A-3D.
Relative matrix effect of the method for PQ and the I.S. can be evaluated by comparing the %CV from set 1b and set 2 ( Table 3A). The differences between CV% of peak areas from set 1b and 2 are -0.1, -3.3, and -1.7 at low, median, and high concentration levels, respectively; the corresponding values for IS are 0.2, -3.7, and -1.0, respectively. When comparing CV% from the peak area ratios (H-G), these values are 0.3, 0.1, and 0.9, all within 5%, suggesting that IS compensated for the variation. These results suggest no significant relative matrix effect.
Absolute matrix effect was evaluated with mean peak area values from set 1 and 2. A value of 100% means no matrix effect. At low, median, and high concentration levels, the matrix effect for PQ is 122, 120, and 117%, respectively. However, The IS exhibits the same trend of matrix effect (122, 112, and 114%, respectively). The difference (H-G) is less than 8% (Table 3B), suggesting IS compensates matrix effect for PQ. These results indicate that matrix effect in the method is well compensated by the deuterated IS. [Table3B].
Furthermore, slopes of lines connecting low, median, and high samples from each lot of plasma were calculated. The CV% from set 3 is 2.88% (<5%), confirming absence of significant matrix effect on quantification. [Table3C].
In summary, matrix effect on quantification for this assay is not significant and compensated well by the deuterated internal standard PQ-d6. Hemolyzed plasma won't affect quantification of PQ either.

Recovery (RE)
The recovery of PQ from plasma following sample preparation was assessed by comparing the peak areas from set 3 and set 2 [ Table 3B]. The recoveries for PQ are 107.2, 83.7 and 82.4% at low, median and high concentration, respectively, and the recovery for the IS ranged from 90.8-92.7%. The CV% of peak areas for recovery experiment was all within 10%. These results suggest the assay is highly reproducible across the concentration range. FDA guidelines state that recovery need not be 100% as long as it is consistent, precise, and reproducible.
Recovery, Process Efficiency and Matrix Effect were calculated with the following formulas: Recovery = 100 x peak area of pre-extraction spiked sample (set3) peak area of post-extraction spiked sample (set2) Matrix Effect = 100 x peak area of post-extraction spiked sample (set2) peak area of clean sample (set1b) Process Efficiency = 100 x peak area of pre-extraction spiked sample (set3) peak area of clean sample (set 1b) Set 1a: un-processed PQ without matrix and IS (defines coelution effect, not necessary in this case because IS will be present in all samples) Set 1b: Un-processed PQ without matrix but with IS Set 2: Post-extraction spike into extracted blank plasma (defines absolute and relative matrix effects) Set 3: Pre-extraction spike into blank plasma and then extracted (defines recovery and overall process efficiency)

Carryover evaluation
In the previously approved method, carryover from autosampler remained at a significant level: ~0.08% of ULOQ (250ng/mL), corresponding to ~0.2ng/mL PQ. To be conservative, the LLOQ for the assay was set at 1.5 ng/mL, meeting the criteria that blank signal should be <20% LLOQ signal. The calibration range was 1.5-250 ng/mL [2].
In this modified assay, since the LLOQ was lowered to 0.5 ng/mL, the ULOQ was lowered to 50 ng/mL accordingly. The residual peak in blank plasma after ULOQ was 17% LLOQ (Figure 2).
To further evaluate the effect of carryover on sample quantification. Calibrator #1 (LLOQ 0.5 ng/mL) was re-injected following calibrator #7 (ULOQ 50ng/mL), the determined concentration of the reinjected calibrator #1 has a mean accuracy (%deviation) of 6.1% and the percent difference from the control (calibrator#1) was 8.7% (Table 4). This result confirmed that carryover in this modified assay won't affect quantification of PQ.

Method
All samples were stored at -70 °C until prepared for analysis. Calibration standards and validation samples were prepared in plasma from separately weighed and prepared PQ solutions. Spiked concentrations and peak area ratios of PQ to IS for the calibration standards were fitted by linear regression with 1/x weighting to the equation. PQ concentrations were calculated from the regression parameters using peak area ratios.  Chromatograms of PQ in a blank plasma (blk), the lower limit of quantification (LLOQ), and a blank plasma after the upper limit of quantification (ULOQ). The lower panel was a chromatogram of the internal standard (PQ-d 6 ).