Using Mitra sampling to support first-in-human pharmacokinetic evaluations for PF-07059013

Nearly 90% of the world's sickle cell disease (SCD) population lives in three countries: Nigeria, India and the Democratic Republic of Congo. Individuals with SCD typically have signs of the disease during the first year of life, usually ~5 months of age [1–4]. In Africa, SCD is one factor in the very high rate of childhood mortality, 50–90%. Yet there is a lack of reliable, up-to-date information [5–9]. The use of dried blood microsampling can be beneficial for pediatric disease populations and in the developing world where centrifuges and dry ice may not be readily available. In the past decade, patient-centric microsampling innovations have been introduced and used in clinical trials. Dried blood spots (DBSs) have become a popular method in a variety of micro blood-sampling techniques in the life sciences sector, consequently competing with the field of conventional, invasive blood sampling by venipuncture [10]. There are several advantages to the DBS technique. The low blood volume requirement, ease of transportation and storage without need for special treatment make it a convenient blood sampling technique. On the other hand, although DBS sampling can eliminate the need for temperature controls, the sample processing and extraction procedures can be cumbersome. The greatest concern with DBS is that results may be affected by varying hematocrit levels affecting spot size and homogeneity.

Because of the advantages of volumetric absorptive microsampling (VAMS) over DBS, the Mitra™ microsampling device was chosen to be evaluated in selected groups following single and multiple dosing in the first-in-human study C4061001 (a phase I, randomized, double-blind, placebo-controlled, single- and multiple-dose escalation study to evaluate safety, tolerability, pharmacokinetics and pharmacodynamics of PF-07059013 in healthy adults; NCT04323124). PF-07059013 is a noncovalent stabilizer of the oxygenated state of hemoglobin S. In all the dosing groups of the first-in-human study where Mitra samplings were evaluated (2000-mg single dose and 3000-mg multiple-dose groups), parallel Mitra pharmacokinetic (PK) samples were collected by dipping the Mitra tips into the venous whole blood PK samples immediately after collection. Mitra DBS and liquid whole blood samples were collected at the same time point from the same healthy participant for comparison.

Using small-volume DBS for quantitative PK concentration measurements has several advantages over traditional liquid plasma samples [11–13]. However, it is known that different hematocrit levels may affect assay accuracy with DBS sampling [14,15]. VAMS addresses several of the issues with DBS. Most critically, it is not affected by the level of hematocrit because it is volumetric [16–18]. To demonstrate, the blood volumes absorbed by Mitra tips are free from the impact of different levels of hematocrit, assay accuracy was verified using quality control (QC) samples prepared in blood with a wide range of hematocrit levels. During the course of clinical development, multiple lots of Mitra tips could be used over several years. To ensure that assay accuracy was not impacted by different material lots, the same QC pool was deposited onto different lots of Mitra tips to evaluate assay variability introduced by different material lots.

Materials & methods

Chemicals & materials

PF-07059013 (lot no. 19-AP-00187 with potency of 98.5%) and its stable-labeled internal standard (SLIS) (d4-SLIS; PF-07221410-00, lot no. PF-07221410-00-0001 with purity of 99.0%) were provided by Pfizer Inc. (CT, USA). Supplementary Figure 1 shows the chemical structures for both compounds. Formic acid (certified grade), acetonitrile (HPLC grade), dimethyl sulfoxide (HPLC grade), isopropyl alcohol (certified grade), acetone (certified grade) and methanol (HPLC grade) were purchased from Fisher Scientific (Loughborough, UK). Ultrapure water (HPLC grade) was purchased from Fisher Scientific or obtained from in-house Milli-Q water purification system. Blank human whole blood with di-potassium (K2) EDTA as anticoagulant was purchased from Cambridge Bioscience or obtained from in-house donors (Cedar House, York, UK) stored under refrigerated conditions (2–8°C) and used without freeze–thaw cycles. Mitra Tips (cat. no. 100308-A) were purchased from Neoteryx (CA, USA).

Preparation of stock & working solutions

The two compounds, PF-07059013 and the SLIS, were accurately measured into Eppendorf tubes and dissolved completely in 1% formic acid in 50/50 (v/v) acetontrile/dimethyl sulfoxide with a final concentration of 10 mg/ml and 0.1 mg/ml, respectively. The calibration standard (STDs) and QC stock solutions were prepared from separate weighings and further diluted in 0.1% formic acid (in 50/50 (v/v) acetonitrile/water to produce working solutions. The SLIS working solution was prepared by diluting the SLIS stock solution to 250 ng/ml with methanol on each day of use. All stock solutions and working solution were subaliquoted and stored in Eppendorf tubes in a refrigerator set at approximately 4°C.

Preparation of STD & QC samples

Small volumes (<5% of total preparation volume) of stock solutions, or working solutions, were spiked into human whole blood for the preparation of liquid STD and QC samples. The theoretical concentrations of STDs were 50.0 (lower limit of quantitation [LLOQ]), 100, 500, 2000, 5000, 25,000, 40,000 and 50,000 (upper limit of quantitation [ULOQ]) ng/ml. The theoretical concentrations of QCs were 50.0 (LLOQ QC), 150 (low QC), 15,000 (mid QC), 37,500 (high QC) and 500,000 (dilution QC) ng/ml. STD and QC samples were thoroughly mixed, and 10 μl was deposited onto one Mitra tip (Figure 1A–F) by touching the tips to the surface of the liquid blood until the entire tip is filled with blood. This tip-wetting methodology replicates the collection of unknown samples.

Sample extraction

Samples were extracted by organic extraction from the Mitra tips followed by dilution. A 300-μl volume of SLIS working solution (or methanol for blanks) was added to the appropriate wells of a 1.2-ml 96-well extraction plate. A Mitra 96-autorack (Figure 1G) populated with samples in run order was then placed onto the plate and the sample tips submerged. The plate was sonicated in an ultrasonic water bath for 30 min and then vortex mixed at 1000 r.p.m. for 90 min to aid organic extraction of the test substances. Then 50 μl of each sample was aliquoted into a fresh 2-ml 96-well plate and diluted with 650 μl of 2.4 mM ammonium formate (acetonitrile-water [20:80 v/v]) and vortex mixed at 1000 r.p.m. for 5 min. The samples were further diluted by taking 40 μl into another fresh 2-ml 96-well plate and adding 560 μl of 2.4 mM ammonium formate (acetonitrile-water [20:80 v/v]) then vortex mixing at 1000 r.p.m. for 5 min and centrifuging at 5000 × g for 5 min at ambient temperature.

HPLC–MS/MS conditions

A Waters Acquity™ UPLC system (Waters Limited, Winslow, UK) was used to introduce samples and perform HPLC separation. Supplementary Table 1 summarizes the gradient elution. An Applied Biosystems API 5000™ triple-quadrupole mass spectrometer (AB Sciex UK Limited, Warrington, UK) operating under positive ion mode. Table 2 summarizes the MS/MS parameters used.

Data acquisition & processing

Applied Biosystems Analyst software version 1.6.2 was used to acquire total ion chromatograms. The same software was used to process MS data and to perform peak integrations of PF-07059013 and its SLIS. Regression and quantitation were performed with Watson™ LIMS version 7.4.1 using analyte-to-SLIS peak area ratios. Quadratic regression with 1/concentration weighting was used based on the concentration–response relationship observed in the accuracy and precision runs. The regression equation derived from the calibration curve was then used to back-calculate the measured concentrations of QC samples and unknowns. The back-calculated results were compared with the theoretical concentrations to obtain the accuracy and precision performance of QCs.

Clinical sample collection

VAMS using Mitra tips was evaluated from selected groups in both single ascending dose and multiple ascending dose portions of the first-in-human C4061001 study. In all the dosing groups where Mitra samplings were evaluated, parallel Mitra PK samples were collected by dipping the Mitra tips into the venous whole blood PK samples immediately after collection (Figure 1A–F). Mitra DBS and liquid whole blood samples were collected at the same sampling time point from the same subject for comparison. Sample size and method of data comparison were determined based on the available industry White Paper [19] due to the lack of specifics in current regulatory guidance.

Results & discussion

Method qualification

The method for quantitative determination of PF-07059013 was qualified in human dried blood (with K2EDTA as anticoagulant) according to the regulatory guidance [19,20], Pfizer internal standard operating procedures and the method qualification plan. The qualification was conducted to evaluate assay performances with respect to accuracy and precision, sensitivity, selectivity, specificity, carryover, matrix effect, hematocrit effect, dilution integrity, recovery and stability. The method is otherwise fully validated except for long-term ambient temperature sample stability. The stability evaluation was not conducted for two reasons. First, the Mitra assay feasibility evaluation was an exploratory study end point defined in the protocol and a method qualification would fit this purpose. Second, the clinical program was put on hold during the pandemic and later deprioritized. An internal decision was made to bring the assay to a fully validated state by conducting the long-term ambient stability if the program status changed.

Accuracy, precision & sensitivity

The assay accuracy and precision were established by three consecutive qualification runs. Each run included system suitability samples; a blank sample; a blank sample with SLIS; a set of STDs analyzed in duplicate; and six replicates of the LLOQ quality control (QC), low QC, mid QC and high QC. All three accuracy and precision runs met acceptance criteria (Table 1) with sufficient accuracy, precision and sensitivity demonstrated at assay LLOQ (Supplementary Figure 2A). Representative extracted ion chromatograms from blank matrix, ULOQ and LLOQ are shown in Supplementary Figure 2A–C.

Regression model (quadratic, 1/concentration weighing) was used because it adequately described the concentration–response relationship based on the three accuracy and precision runs. The correlation coefficient R² was greater than 0.999 in all qualification runs. including the three accuracy and precision runs. An example calibration curve can be found in Supplementary Figure 3.

Selectivity & specificity

To demonstrate selectivity and matrix effect, six normal blood lots were evaluated. No peaks were detected for both PF-07059013 and its SLIS at their retention time with their respective MRM transitions (Supplementary Figure 2B).

Potential ‘crosstalk’ between analyte and SLIS was also investigated owing to a common product ion being selected for both analyte and SLIS (Supplementary Table 2). A default pause time of 5 ms was added between MRM scans to avoid potential crosstalk.

To confirm the effectiveness of the pause time used, a PF-07059013 ULOQ sample was extracted without SLIS and analyzed for SLIS. There was no contribution from PF-07059013 to the SLIS MRM channel. Similarly, the contribution from SLIS to PF-07059013 was tested at the normal SLIS level used for extraction in six normal lots of blood. The contribution from SLIS to PF-07059013 MRM channel was confirmed to be zero as well. An example PF-07059013-extracted ion chromatogram from a blank dried blood sample extracted without SLIS is presented in Supplementary Figure 2B.

Carryover

Carryover was evaluated by injecting blank dried blood extracts (without SLIS) immediately after ULOQ calibrators. For all the qualification runs, the analyte PF-07059013 carryover was 0% of the mean SLIS peak area counts from all the STDs and QCs. They were all well within the preset acceptance criteria of 20 and 5%, respectively.

Matrix effect

To demonstrate no effects from matrix with different sources or properties, six normal blood lots were evaluated in triplicate using the matrix factor (MF) approach. Individual lots of each matrix type were extracted without SLIS following the sample processing procedure. The extracts were spiked with both analyte and SLIS to the final extract concentration of the QC L and the QC H, assuming 100% extraction recovery. Samples to measure response in the absence of matrix were prepared by spiking analyte and SLIS into diluent (20/80 v/v acetonitrile/water with 0.5% formic acid in 2.4 mM ammonium formate) to the final extract concentration levels of QC L and QC H. Individual and mean SLIS-normalized MF for each lot at each QC concentration level were calculated using the below formula.

For normal blood lots, the mean SLIS-normalized MF is 1.00 with minimal lot-to-lot variations (%CV = 2.3) at the low QC level. Similar results were obtained for the high QC with mean MF of 1.00 and interlot %CV of 2.0. The MF evaluation confirmed that minimal matrix effect was introduced by different sources and types of matrices.

Hematocrit effect

To evaluate the impact of different levels of hematocrit on assay accuracy, measurement and subsequent adjustment of %hematocrit in fresh whole blood was performed in the analytical lab following procedures described previously [21]. Hematocrit impact to assay accuracy is known for the DBS method because the spot size (blood spreadability on paper) is affected by the level of hematocrit. When part of the spot is excised using a fixed-diameter cutter, a variable volume of blood is used. However, VAMS material absorbs blood volumetrically. It has been demonstrated by Neoteryx that the volumetric absorption of blood is independent of hematocrit levels. Because this is the first time that Pfizer used VAMS material for whole blood sampling in a clinical setting, it was decided to evaluate accuracy of volumes absorbed by VAMS material by checking assay accuracy with different levels of hematocrit. The volumetric absorptive property of VAMS material is driven by the physical property of the material. The volume accuracy is not concentration dependent. Therefore, we decided to use mid-QC to verify the impact of hematocrit on accuracy of volumes by evaluating assay accuracy. The impact of different levels of hematocrit on assay accuracy was evaluated by preparing midlevel of QCs in human whole blood with varying amount of hematocrit, and then quantitate these evaluation QCs against a calibration curve that was made with blood that has 45% hematocrit. Assay accuracy ranged from -8.0 to 10.0% bias when hematocrit levels varied from 30 to 60% (Table 2). This evaluation confirmed the VAMS material used in Mitra tips can volumetrically absorb whole blood with little to no impact from the blood's hematocrit levels.

Mitra tip lot-to-lot variation

Clinical trials can last for many years with chronically dosed drugs. This results in multiple lots of sample collection material (Mitra tips, in this case) being used at the clinical sites and analytical labs. To evaluate the potential impact of the lot variations on assay accuracy, midlevel QCs in human liquid whole blood were prepared and deposited onto five lots of Mitra tips. These evaluation QCs were then quantitated against the same calibration curve. Interlot CV was 1.7% (Table 3), demonstrating minimal lot-to-lot variation.

Dilution integrity

Dilution of dried samples cannot be conducted with the traditional approach used for liquid samples. With liquid samples, dilution is typically achieved with either full volume dilution (i.e., use full assay volume) or partial volume dilution (i.e., use a portion of full assay volume) by adding additional volume of blank matrix prior to sample aliquoting into assay plate. With dried samples, dilution is achieved by adding blank extract to the final sample extract before reconstitution or analysis depending on sample processing procedures. In this qualification, dilutions with high QC (37,500 ng/ml, dilution factor = 10) and dilution QC (500,000 ng/ml, dilution factor = 20) were performed in one of the qualification runs by diluting the final QC extracts with blank dried blood extracts before analysis. Dilution integrity was demonstrated with both in- and out-of-range QCs with mean bias ranging from -2.4 to 1.3%, and CV was ≤13.7%.

Processed sample stability

A variety of storage conditions were evaluated to ensure the conditions encountered during each step of sample collection, shipment, storage and analysis did not impact sample integrity. The stability of PF-07059013 in liquid whole blood and dried blood at different conditions was thoroughly evaluated and is summarized in Table 4. The stability of PF-07059013 in liquid whole blood at ambient condition was established for at least 24 h to ensure no analyte loss during the drying step at sites. The processed extract stability (using high and low QCs) and reinjection reproducibility (using four levels of QCs: LLOQ QC, low QC, mid QC and high QC) of PF-7059013 in dried whole blood deposited on Mitra tips was established for 144 and 70.5 h, respectively at 6°C. Stability was assessed by comparing the mean concentration of the stability evaluation samples to the nominal concentration.

Sample analysis

A total of 363 Mitra DBS were collected. Each sample was collected using a four-tip clamshell device. This allowed the initial analysis, dilutions, repeats and mandatory repeats to be performed using the four tips from the same sample. All nonplacebo samples were analyzed in 15 runs with an overall run passing rate of 88.2%. An example chromatogram from a dosed subject is presented in Supplementary Figure 2D. Incurred sample reanalysis was conducted using 30 samples (approximately 10.4% of all the samples analyzed) with 28 samples (93.3%) meeting acceptance criteria (Supplementary Table 3). The excellent assay performance demonstrated assay robustness and consistent sample quality from tip to tip.

Samples were collected and dried at ambient temperature. Samples were also stored at ambient temperature at the clinical site and analytical lab before analysis. No temperature excursions were reported.

Results comparison

To demonstrate that the concentration results obtained from liquid blood and dried blood sampling techniques are equivalent, a cross-validation type of evaluation is preferred and expected by regulatory reviewers. However, a traditional cross-validation approach cannot be used because the same set of QCs (low, medium and high) cannot compared using the two methods that also need to be compared. If the QCs are prepared in liquid whole blood, then it is impossible to measure the liquid blood QCs using a dried blood Mitra method. If the QCs are prepared with Mitra probes as DBS, then it is impossible to measure the dried blood QCs using liquid whole blood assay. Therefore, the methodology used for bridging (not cross-validation) followed the recommendations from the Land O'Lake white paper [22]. Parallel liquid samples and DBS were collected from the same subject at the same time point. They were assayed with their respective liquid and dried blood assays. The concentrations and exposures were compared. The statistical approach required by ICH M10 on cross-validation (i.e., Bland–Altman analysis) was used for concentration comparison.

Concentration data obtained from liquid blood samples were compared with Mitra DBS for the same participant at the same time point. The percent difference was calculated using the following formula:

Of the 286 data pairs compared, 216 (75.5%) were within ± 20% of the mean. Correlation analysis (Figure 2) and Bland–Altman plot (Figure 3) demonstrated excellent correlation and minimal systematic bias.

In addition to the direct concentration comparison, PK parameters (area under the blood concentration curve and Cmax) calculated with the two datasets were also compared (Figure 4). Both PK exposure parameters from the single and multiple dose cohorts were generally comparable, although slightly larger variability for area under the curve was noted from the Mitra DBS for the 3000-mg MD cohort.

Conclusion

A sensitive and selective method for the determination of PF-07059013 in dried blood deposited on Mitra tips was developed, qualified and used in support of a clinical study with excellent assay performance. The concerns about hematocrit levels and material lot-to-lot variations were thoroughly evaluated in qualification, and it was determined that assay accuracy was not influenced by these factors.

Assay comparability of the liquid whole blood and Mitra dried blood assays was evaluated with 286 data pairs and demonstrated with three approaches. First, direct concentration comparison met the criteria of two-thirds of samples within ± 20% of the mean values. Second, statistical analysis using Bland–Altman plots and correlation analysis was conducted and demonstrated excellent assay correlation. Third, PK parameters derived from the two datasets produced comparable results.

Our initial evaluation with the Mitra VAMS microsampling device has built a solid foundation for potential implementation in efficacy trials. More evaluation is needed to bridge the venous and capillary blood sampling route, preferably in diseased populations. With a smaller sampling volume and easy storage and shipment without requiring dry ice, centrifuge and freezers, we see the excellent value of Mitra use in developing countries. With this easy-to-use microsampling device, we also recognize its potential use by patients themselves or assisted use by caregivers. In summary, all the advantages provided by the Mitra device can enhance patient trial experience and improve patient access to clinical trials by enabling sites that would otherwise not be able to conduct such trials due to limited access to dry ice and freezers.

Future perspective

At-home sampling will be broadly used in clinical trials beyond just PK to recruit and retain patients. Patient-centered microsampling will replace the traditional sampling for PK in special populations (such as pediatric) and in certain therapeutic areas (such as infectious disease and oncology). Industry best practice and regulatory guidance are likely to follow to set consistent standards across the world.