Simultaneous determination of trimethoprim and sulfamethoxazole in dried plasma and urine spots

Abstract

Background

Trimethoprim-sulfamethoxazole (TMP-SMX) is an antimicrobial drug combination commonly prescribed in children and adults. The study objectives were to validate and apply an HPLC-MS/MS method to quantify TMP-SMX in dried plasma spots (DPS) and dried urine spots (DUS), and perform a comparability analysis with liquid matrices.

Results

For TMP the validated range was 100–50,000 ng/mL for DPS and 500–250,000 ng/mL for DUS; for SMX, the validated range was 1000–500,000 ng/mL for both DPS and DUS. Good agreement was noted between DPS/DUS and liquid plasma and urine samples for TMP, while only modest agreement was observed for SMX in both matrices.

Conclusions

A precise, accurate, and reproducible method was developed to quantify TMP-SMX in DPS and DUS samples.

Introduction

Trimethoprim-sulfamethoxazole (TMP-SMX) is a combination of two antimicrobial agents that inhibit distinct proteins in the tetrahydrofolate synthesis pathway; TMP inhibits the enzyme dihydrofolate reductase and disrupts production of tetradyrofolic acid, whereas SMX mimics para-aminobenzoic acid (PABA) and prevents its conversion to dihydrofolic acid via dihydropteroate synthetase [1]. Inhibition of these proteins affects DNA bacterial synthesis and ultimately bacterial growth. When administered together, this drug combination has potent activity against aerobic gram-positive and gram-negative bacteria. In children and adults, TMP-SMX is prescribed to treat urinary, respiratory, or gastrointestinal tract infections [1]. Also, it is commonly prescribed to treat skin and skin structure infections caused by methicillin-resistant Staphylococcus aureus (MRSA) [2].

Simultaneous quantification of TMP-SMX in biological fluids has been performed using high performance liquid chromatography (HPLC) [3-7], tandem mass spectrometry [7,8], and capillary zone electrophoresis [9]. Often these methods have been applied to simultaneously quantify TMP-SMX in human plasma samples collected in adult pharmacokinetic (PK) studies. In pediatric PK studies, because of practical limitations regarding the number and volume of blood samples that can be collected ethically, dried blood spot (DBS) sampling and multi-drug assays have been proposed as novel tools to improve pediatric clinical trial designs [10]. The advantages of DBS sampling include significantly reduced blood volumes (10–25 μL), reduced biohazard risk, ease of storage (room temperature), and improved drug stability [11]. Measurement of drug concentrations in dried plasma spots (DPS) [12-14] and, less commonly, in dried urine spots (DUS) also have been reported [15]. However, measurement of TMP-SMX in DPS and DUS samples has not been reported previously, and measurement of drug concentrations in dried matrix samples in children often has focused on using DBS. Drug measurement in DPS samples has the additional advantage of avoiding the effect of varying hematocrit on sample homogeneity observed with DBS [16] and allows for easy reporting of results as the PK literature frequently focuses on plasma concentrations [12].

The objective of the analyses described herein was to develop and validate a high-performance liquid chromatography tandem mass spectrometry (HPLC-MS/MS) method for the simultaneous quantification of TMP-SMX in DPS and DUS samples collected in an opportunistic pediatric PK study. Clinical samples were then analyzed and a comparability analysis of the dried and liquid matrix samples was performed.

Experimental

Materials

Free base forms of the study compounds TMP (CAS No. 738-70-5, Batch SZB9352XV) and SMX (CAS No. 723-46-6, Batch SZBC124XV) were purchased from Sigma-Aldrich Corporation (St. Louis, MO, USA) (Figure 1). Stable isotope-labeled forms of the study drugs were used as internal standards (CDN Isotopes, Inc., Pointe-Claire, Quebec, Canada): [²H₃]-TMP (Lot: E395P36) and [²H₄]-SMX (Lot: M237P19). Control K₂ EDTA human plasma (BioChemed Services, Winchester, VA, USA) and urine (collected from human volunteers) was centrifuged for approximately five minutes at 4000 rpm prior to use. Whatman® FTA® DMPK-C were used for the DPS analysis and Whatman® FTA® DMPK-C IND dried matrix spotting cards were used for DUS analysis (Whatman Ltd Co., Middlesex, UK; GE Healthcare, Piscataway, NJ, Catalog No.WB120224).

Figure 1.

Chemical structure of study drugs and internal standards.

Standard solutions

For the calibration standards, eight concentration levels were prepared for both TMP and SMX in human plasma and urine: 100–50,000 ng/mL in human plasma for TMP; 500–250,000 ng/mL in human urine for TMP; and 1000–500,000 ng/mL in human plasma and urine for SMX. The following nominal concentrations were prepared for quality control (QC) samples for TMP/SMX in human plasma: 100/1000 (lower limit of quantification [LLOQ] and carryover assessments only), 300/3000, 4000/40,000, 40,000/400,000, and 100,000/1,000,000 (dilution linearity assessment only) ng/mL. In human urine, the following TMP/SMX concentrations were selected for QC samples: 500/1000 (LLOQ and carryover assessments only), 1500/3000, 20,000/40,000, 200,000/400,000, and 500,000/1,000,000 (dilution linearity assessment only) ng/mL.

For DPS analysis, stock solutions were prepared by accurately weighing the appropriate amount of TMP/SMX to dissolve in 1:1 (v/v) methanol: dimethyl sulfoxide (DMSO) to obtain 4 mg/mL and 40 mg/mL stock solutions of TMP and SMX, respectively. A combined stock solution was then prepared by combining equal volumes of each, resulting in a 2 mg/mL TMP and 20 mg/mL SMX solution. For DUS analysis, a similar procedure was followed, and a combined stock solution containing 10 mg/mL TMP and 20 mg/mL SMX was obtained. The stock solutions were stored at −70°C or below. Calibration standards and QC samples were made from these stock solutions.

Calibration standards and QC samples were prepared using human plasma or human urine that was thoroughly mixed, spotted on dried matrix spotting cards, and then dried. Calibration standards were extracted fresh daily. A 10 μL volume per spot was used, except for QCs that were prepared to test the effects of varying spot volume (5 and 15 μL). Calibrated mechanical pipets were used for all volumetric measurements. Calibration curves and QC samples were dried overnight on the bench top and then in a Minigrip zippered bag with desiccant for 24 hours prior to use.

Extraction method

Initially, 3 mm punches were used for the dried matrix sample analysis. However, analysis of 3 mm punches from the center and edge of punches indicated that the TMP and SMX concentrations on the spot were not homogeneous. Therefore, a 6 mm punch was used to sample the majority of the spot and minimize non-homogeneity issues. Methanol has been widely used as an extraction solvent for dried matrix sample analysis due to its ability to provide a relatively clean extract by both binding biological matrix to the filter paper on the DMS card and solubilizing the analyte. Extraction volumes of 100 μL, 200 μL, and 400 μL were tested for extraction volumes. The 400 μL volume provided the best extraction efficiency. The final extraction method involved punching a 6 mm spot into a microcentrifuge vial, adding 400 μL of internal standard in methanol, and vortexing for 5 minutes. Samples were then centrifuged for 13000 rpm for 5 minutes. An aliquot (50 μL) of each sample was added to a 96 well plate containing 50 μL of deionized water.

Liquid plasma and urine method sample preparation

An aliquot of sample (10 μL) was added to a sample container. Internal standard in methanol (70 μL) was added to each sample. Samples were then vortexed for 5 minutes and centrifuged at 4000 rpm for 10 minutes. Sample (25 μL) was then added to a 96 well plate containing 75 μL of deionized water and analyzed using a C8 HPLC column and tandem mass spectrometry.

LC-MS/MS

A detailed description of the equipment and settings is provided in Table 1. The Agilent 1200 series HPLC system and an Agilent 1290 auto-sampler were used (Agilent Technologies, Inc., Santa Clara, CA, USA). The ACE PFP, 2.1 × 50 mm, 3 μm (Advanced Chromatography Technologies Ltd, Aberdeen, Scotland) analytical column was used. The column temperature was 30°C. The injection volume, flow rate, and run time were 10 μL, 0.75 mL/min, and 3.5 minutes, respectively. A gradient mobile phase was used: water containing 0.1 % (v/v) formic acid (mobile phase A) and acetonitrile containing 0.1 % (v/v) formic acid (mobile phase B). During the first three minutes, the percentage mobile phase B increased from 5% to 45%; from 3.01–3.4 minutes, it was 100% B; and for the remainder of the run time (3.41–3.5 minutes), it was 5% B. The HPLC system was coupled with an Agilent 6460 series Triple Quadrupole system (Agilent Technologies, Inc., Santa Clara, CA, USA). The Agilent Mass Hunter software was used for data acquisition and quantitative analysis. A positive mode electrospray ionization interface was used. The following system settings were used: 350°C, gas temperature; 10 L/min, gas flow; 50 psi, nebulizer pressure; 10.1 L/min, sheath gas flow, 4000 V capillary voltage; 140/160 V TMP/SMX, fragmentor values; and 25/15 V TMP/SMX, collision energy values.

Table 1.

Description of equipment and settings

HPLC		MS/MS
Autosampler	Agilent 1260	Mass spectrometer	Agilent 6460 Series Quadrupole
Injection volume	10 μL	Ionization interface	Positive Mode Electrospray, Jet Stream
Chromatography system	Agilent 1200 series	Gas temperature	350°C
Flow rate	0.75 mL/min	Gas flow	10 L/min
Analytical column	Ace PFP, 2.1×50, 3.0 μm	Nebulizer pressure	50 psi
Column temperature	30°C	Shealth gas temperature	350°C
Run time/Data acquisition time	3.5 minutes/ 3.5 minutes	Sheath gas flow	10.1 L/min
Mobile phase A	Water containing 0.1% (v/v) formic acid	Capillary voltage	4000 V
Mobile phase B	Acetonitrile containing 0.1% (v/v) formic acid	Fragmentor value	160 V (sulfamethoxazole), 140 V (trimethoprim)
Injector wash	70:30 MeOH:Water	CE value	15 V (sulfamethoxazole), 25 V (trimethoprim)

Liquid plasma and urine concentrations were measured using validated LC-MS/MS methods. The Agilent 1200 Series HPLC system, Agilent 6410 Series Triple Quadrupole mass spectrometer, and Agilent Zorbax XDB-C8 analytical column (2.1 mm internal diameter × 30 mm length, 3.5 μm particle size) were used (Agilent Technologies, Santa Clara, CA, USA). A gradient mobile phase was made up of water containing 0.1% (v/v) formic acid and methanol containing 0.1% (v/v) formic acid. The method was validated according to the standards set forth by the U.S. Food and Drug Administration [17].

Method validation

The analytical methods were validated according to standards set forth by the U.S. Food and Drug Administration [17]. For both DPS and DUS samples, validation included assessment of standard curve fitting, specificity, within- and between-run accuracy and precision, recovery, matrix effect, linearity of dilutions, carryover, punch carryover, sample volume variation, reproducibility, and stability. For the latter, bench-top and post-preparative stability results are reported herein.

Specificity was assessed by analysis of samples prepared from human plasma and urine (six different lots each). The method was deemed specific if blank TMP and SMX responses in DPS and DUS samples were ≤20% of the average response at the LLOQ.

Recovery was assessed at three concentrations (low, middle, and high) by comparing extracted QC samples to unextracted QC samples that were prepared by spiking blank matrix post-extraction. Matrix effect was evaluated by comparing extracted and spiked solvent QC samples. Carryover was assessed by comparing five replicate injections of the lowest calibration standard followed by five replicate injections of the lower level calibration standard that had each been injected after an upper limit of quantitation (ULOQ) standard. In addition, punch carryover was assessed by evaluating analyte response with a blank DPS or DUS card that was punched immediately following punching of a card containing the ULOQ. To assess the linearity of dilutions, DPS and DUS samples that were two times greater than the ULOQ were prepared and diluted 10 times with dried plasma or urine spot extract for analysis. Five replicates were made for each matrix. Sample volume variation was evaluated by spotting mid-concentration QCs using a volume less (5μL) and greater (15 μL) than the validated spot volume of 10 μL. Five replicates were used for each sample volume.

To assess storage stability, low and high QC samples were stored at room temperature in Minigrip zippered bags with desiccant for two or eight days for DPS samples and two or 14 days for DUS samples. Post-preparative stability for DPS samples was assessed by injecting extracted QC samples (low and high QCs in replicates of five) stored at room temperature for one and nine days with fresh extracts of calibration standards. Similarly, extracted DUS QC samples (low and high QCs in replicates of five) were stored in the auto-sampler at room temperature for four days and injected with a fresh calibration curve. Samples were deemed stable if the mean values had an accuracy of within ±15% (i.e., 85 to 115%) and precision did not exceed a coefficient of variation (CV) of 15%.

Opportunistic pediatric study

DPS and DUS clinical samples were collected from pediatric patients enrolled in the Pharmacokinetics of Understudied Drugs Administered to Children per Standard of Care (POPS) trial (clinicaltrials.gov NCT01431326; protocol NICHD-2011-POP01), a multi-center (N=26), prospective, PK and safety study in children (<21 years of age). Children who received one of the targeted drugs of interest (including TMP-SMX) per standard of care as administered by their treating caregiver were eligible for enrollment [18]. Exclusion criteria included failure to obtain consent/assent or known pregnancy. PK samples were collected either at the time of routine clinical laboratory collections or, if the parent/patient consented, at a specific collection time for study purposes. Because this was a standard-of-care study, dosing and PK sample collection times varied between subjects.

Whole blood samples of 200–2000 μL were collected based on the participant’s age and weight. DPS samples were collected as an aliquot of liquid plasma samples. After plasma was separated via centrifugation ([2000 g] for 10 minutes at 4°C), a 10 μL micropipette was used to measure 10 microliters per spot, and two plasma spots were spotted per card. The cards were allowed to dry for at least two hours at room temperature. The card was then sealed tightly with two desiccant packs and a humidity indicator card in a gas impermeable (Minigrip) bag. The cards were stored and shipped desiccated at ambient temperature. A similar procedure was followed for DUS samples using an aliquot of urine.

Data analysis

For calibration standards, a plot of the analyte-to-internal standard peak area ratio versus analyte concentrations was created. A weighted (1/x²) power regression analysis was applied to the data. The concentrations of the analytes in the QC samples were determined using the respective calibration line; then, the accuracy and precision of the method was assessed.

An analysis was performed to compare DPS and DUS TMP and SMX concentrations against the established methods of analysis, liquid plasma (LPS) and liquid urine (LUS), respectively. The ratio of DPS to LPS samples was computed for each drug and plotted against the average concentration for each paired sample. Passing-Bablock regression analysis was used to characterize potential bias between matrices, with dried matrices plotted on the y-axis, and liquid matrices plotted on the x-axis as the reference standard. A Passing-Bablock regression is a non-parametric alternative to ordinary linear regression that assumes a constant ratio of variances, and is less sensitive to outliers when compared to other regression methods [19]. The presence of a systematic and/or proportional bias was noted if the y-intercept and/or regression slope significantly differed from zero and unity, respectively [20,21]. Bias also was assessed through calculation of the median percentage prediction error (MPPE), while imprecision was evaluated through calculation of the median absolute percentage prediction error (MAPE). Calculations of MPPE and MAPE were as follows:

$$\mathrm{MPPE}=\text{median}[100\%\times ({\mathrm{C}}_{\mathrm{DPS}}-{\mathrm{C}}_{\mathrm{LPS}})∕{\mathrm{C}}_{\mathrm{LPS}}]$$

$$\mathrm{MAPE}=\text{median}[100\%\times \mid ({\mathrm{C}}_{\mathrm{DPS}}-{\mathrm{C}}_{\mathrm{LPS}})\mid ∕{\mathrm{C}}_{\mathrm{LPS}}]$$

C_DPS and C_LPS denote the drug concentrations in DPS and LPS, respectively. Similar equations were used to calculate MPPE and MAPE for DUS and LUS samples. A value of < 15% for MPPE or MAPE was considered acceptable [21-23].

STATA 13 (College Station, TX) was used for statistical analyses, and MedCalc Statistical Software version 14.8.1 (Ostend, Belgium) was used to generate Passing-Bablock regression plots. Standard summary statistics were used to describe the demographics and laboratory characteristics of the study population.

Results and discussion

Standard curve fitting

Linear responses (using a power regression) in the TMP/internal standard peak area ratios were observed over the range of 100–50,000 ng/mL for DPS and 500–250,000 ng/mL for DUS samples. Also, linear responses (using a power regression) in the SMX/internal standard peak area ratios were observed over the range of 1000–500,000 ng/mL for DPS and DUS samples. The correlation coefficients were 0.995 or better for all runs. Accuracy values were within ±15% of the theoretical value for all runs at all concentration levels. Representative calibration plots for SMX and TMP in DPS and DUS are shown in Figures S1 and S2. Representative chromatograms of LLOQ samples for TMP/SMX in DPS and DUS samples are shown in Figure S3. Zero (internal standard added) and blank (no internal standard) samples were included with each validation run (Figures S4 and S5).

Specificity

The characteristic precursor [M+H]⁺ to product ion transitions, m/z 291 to 230 and 294 to 123 for TMP and its internal standard, and m/z 254 to 156 and 258 to 160 for SMX and its internal standard, were used as multiple reaction monitoring transitions to ensure that optimal selectivity was obtained. For TMP internal standard, the 294→123 transition was used because when we performed the compound optimization, the 123 daughter ion was more abundant and gave us more signal than the 230 daughter ion. The method was deemed to have adequate specificity based on the analysis of TMP and SMX responses in blank samples from six different lots (blank responses ≤20% average response at the LLOQ). No unacceptable interferences at the retention times of TMP, SMX, and the internal standards in DPS or DUS were noted with zero or blank samples. The response observed with LLOQ samples was assessed in three validation runs for both DPS and DUS samples. The following criteria were met for the LLOQ samples: the analyte response was at least five times the response obtained with blank samples (i.e., signal-to-noise ratio ≥ 5); peak response had a precision (i.e., coefficient of variation) of ≤ 20%; and accuracy was between 80% and 120% of the theoretical value.

Within- and between-run accuracy and precision

Mean within- and between-run accuracy and precision values for TMP-SMX DPS and DUS QC samples are shown in Tables 2 and 3, respectively. Accuracy was within ±15% of the theoretical value for all runs. Precision did not exceed 15% for any run.

Table 2.

Precision and accuracy for dried plasma spots

	Trimethoprim		Sulfamethoxazole
	Intra-run accuracy (%)	Intra-run precision (%)	Intra-run accuracy (%)	Intra-run precision (%)
Run 1
LLOQ	94.0	2.8	89.3	5.3
Low	91.6	2.8	86.4	3.9
Mid	96.6	0.8	91.6	2.0
High	95.5	2.8	90.2	4.7
Run 2
LLOQ	93.6	3.7	89.9	5.7
Low	92.3	3.7	88.3	3.7
Mid	99.5	2.1	97.1	1.9
High	95.6	2.0	89.9	3.0
Run 3
LLOQ	92.6	1.6	88.3	6.2
Low	95.4	1.7	93.3	2.1
Mid	96.4	2.5	93.0	3.7
High	94.2	2.7	90.4	1.6

	Inter-run accuracy	Inter-run precision	Inter-run accuracy	Inter-run precision
LLOQ	93.4	2.7	89.2	5.4
Low	93.1	3.2	89.4	4.5
Mid	97.5	2.4	93.9	3.6
High	95.1	2.4	90.2	3.1

Trimethoprim: LLOQ, 100 ng/mL; low, 300 ng/mL; mid, 4 μg/mL; high, 40 μg/mL.

Sulfamethoxazole: LLOQ, 1 μg/mL; low, 3 μg/mL; mid, 40 μg/mL; high, 400 μg/mL.

Accuracy and precision were assessed using five determinations per concentration.

Table 3.

Precision and accuracy for dried urine spots

	Trimethoprim		Sulfamethoxazole
	Intra-run accuracy	Intra-run precision	Intra-run accuracy	Intra-run precision
Run 1
LLOQ	92.7	2.5	96.0	6.1
Low	101.4	2.1	99.5	2.8
Mid	101.5	2.7	99.4	2.5
High	96.3	4.7	99.1	2.4
Run 2
LLOQ	94.0	2.9	95.1	2.9
Low	99.6	1.1	96.3	3.6
Mid	101.3	3.6	97.8	3.7
High	97.4	3.0	97.5	2.3
Run 3
LLOQ	99.3	2.6	97.4	5.1
Low	102.5	3.8	99.6	2.5
Mid	101.7	3.6	99.5	2.2
High	93.0	2.6	96.0	2.5

	Inter-run accuracy	Inter-run precision	Inter-run accuracy	Inter-run precision
LLOQ	95.3	4.0	96.2	4.6
Low	100.6	2.8	97.9	3.2
Mid	100.9	3.1	98.6	2.7
High	95.2	4.4	97.3	2.6

Trimethoprim: LLOQ, 500 ng/mL; low, 1500 ng/mL; mid, 20 μg/mL; high, 200 μg/mL.

Sulfamethoxazole: LLOQ, 1 μg/mL; low, 3 μg/mL; mid, 40 μg/mL; high, 400 μg/mL.

Accuracy and precision were assessed using five determinations per concentration.

Recovery, matrix effect, and linearity of dilutions

Recovery of TMP and SMX in DPS and DUS samples was consistent and reproducible (Table 4). No matrix effect was noted through comparison of the percent recovery observed in extracted QCs with that in spiked solvent samples (100% recovery). Dilution validation samples also met acceptance criteria based on accuracy and precision.

Table 4.

Percentage recovery for dried plasma spots and dried urine spots

	Trimethoprim		Sulfamethoxazole
	Dried plasma spots	Dried urine spots	Dried plasma spots	Dried urine spots
Extraction recovery (%)
Low QC	103.4	100.7	100.7	97.1
Mid QC	105.8	99.8	106.4	97.9
High QC	104.2	99.3	104.5	98.4

Trimethoprim (DPS): low, 300 ng/mL; mid, 4 μg/mL; high, 40 μg/mL.

Trimethoprim (DUS): low, 1.5 μg/mL; mid, 20 μg/mL; high, 200 μg/mL.

Sulfamethoxazole (DPS, DUS): low, 3 μg/mL; mid, 40 μg/mL; high, 400 μg/mL.

Extraction recovery was assessed using five determinations per concentration.

Carryover and sample volume variation

No carryover was noted for TMP and SMX in DPS and DUS samples when the lower level calibration standard was injected following injection of the upper ULOQ (five replicates each). When punch carryover was assessed, analyte response for blank samples was >20% of the mean LLOQ response, and thus the test failed. Punch carryover was then reassessed by use of three cleansing punches (using W903 paper) in between punching of the double blank samples. This resulted in a blank response <20% of the mean LLOQ response. Spot homogeneity was assessed for each matrix using a 3 mm punch size. Due to variability in the results, it was determined that a 6 mm punch must be used to sample the majority of the spot and minimize non-homogeneity issues. Last, to assess sample volume variation, 5 and 15μL spot volumes were compared with the validated spot volume (i.e., 10 μL). Accuracy and precision criteria were met for the 10 and 15 μL, but not for the 5 μL (116.3%).

Stability

TMP and SMX were stable in DPS when stored at room temperature for six days. In DUS, both molecules were stable at room temperature for 12 days. Post-preparative stability results indicated that processed DPS samples are stable for at least nine days at room temperature. Processed DUS samples were stable for at least four days at room temperature.

Comparability analysis

TMP and SMX concentrations measured in LPS and LUS were compared graphically with observations in DPS and DUS samples, respectively (Figures 2 and 3).

Figure 2.

Plot of dried plasma spot-to-liquid plasma sample ratios vs. the mean of the dried plasma spot and liquid plasma sample concentrations. The dark solid line represents the observed mean ratio, and the dashed lines reference the limits of agreement between the samples.

Figure 3.

Passing-Bablock regression of trimethoprim and sulfamethoxazole dried plasma and urine spots versus liquid plasma and urine sample concentrations (grey line represents line of unity; black line is the regression line; dashed lines represent 95% confidence interval).

DPS-to-LPS analysis

Forty-seven LPS and DPS paired samples were collected from 34 subjects (median [range] age, 6.9 years [0.2, 20.1]; weight, 26.1 kg [4.7, 139.4]; TMP dosing, 3.0 mg/kg/dose [0.6, 8.8] and 3.2 mg/kg/day [0.6, 17.6]; SMX dosing, 13.8 mg/kg/dose [3.1, 44.0] and 16.2 mg/kg/day [3.1, 88.1]). Four paired TMP samples were below the limit of quantification.

The mean TMP DPS to LPS ratio was 0.88 (95% CI: 0.84, 0.92), and the limits of agreement were 0.63 and 1.13. Passing-Bablock regression showed linear correlation between the TMP DPS and LPS concentrations, and exhibited negligible proportional and systematic biases between the two matrices (slope 0.93 [95% confidence interval {CI}: 0.91, 0.97]; intercept −0.04 [95% CI −0.07, −0.02]). The MPPE for the comparison of TMP DPS to LPS concentrations was −11.7%, and the MAPE was 12.1%.

The mean SMX DPS to LPS ratio was 0.81 (95% CI: 0.77, 0.86), and limits of agreement were 0.51 and 1.12. The Passing-Bablock regression showed modest agreement between the SMX LPS and DPS concentrations (slope 0.86 [95% CI: 0.80, 0.92]; intercept −0.76 [95% CI: −2.02, 0.25]), with evidence of proportional bias between the two matrices. MPPE for the comparison of LPS and DPS concentrations was −19.8%, and the MAPE was 20.3%.

DUS-to-LUS analysis

Twenty-one LUS and DUS paired samples were collected from 16 subjects (median [range] age, 13.9 years [2.3, 20.1]; weight, 43.2 kg [12.8, 86.0]; TMP dosing, 3.4 mg/kg/dose [0.6, 8.8] and 6.1 mg/kg/day [0.6, 17.6]; SMX dosing, 17.0 mg/kg/dose [3.1, 44.0] and 30.5 mg/kg/day [3.1, 88.1]). One paired SMX sample was below the limit of quantification.

The mean TMP DUS to LUS ratio was 0.94 (95% CI: 0.81, 1.06), and the limits of agreement were 0.40 and 1.47. Passing-Bablock regression showed linear correlation between the TMP DUS and LUS concentrations and exhibited no evidence of proportional or systematic bias between the two matrices (slope 0.96 [95% CI: 0.90, 1.08]; intercept −0.26 [95% CI: −2.78, 5.21]). The MPPE for the comparison of TMP DUS to LUS concentrations was −6.42%, and the MAPE was 17.52%.

The mean SMX DUS to LUS ratio was 0.81 (95% CI: 0.66, 0.97) and limits of agreement were 0.16 and 1.47. The Passing-Bablock regression showed modest agreement between the SMX LUS and DUS concentrations (slope 0.71 [95% CI: 0.58, 0.79]; intercept 1.16 [95% CI: −1.66, 8.56]), with evidence of proportional bias between the two matrices. MPPE for the comparison of SMX LUS and DUS concentrations was −18.66%, and the MAPE was 31.47%.

Conclusions

An accurate, precise, and selective method was developed to detect TMP and SMX in DPS and DUS samples. In the assay validation process, it was determined that three cleansing punches and a 6 mm punch size were optimal to prevent punch carryover and non-homogeneity issues. The developed method was applied successfully to measure TMP-SMX in PK samples collected from an opportunistic PK study. The lower limit of quantification of the assay was appropriate (100 ng/mL, DPS TMP; 500 ng/mL, DUS TMP; 1000 ng/mL, DPS/DUS SMX).

LPS and DPS yield similar concentrations of TMP as evidenced by a linear relationship on the scatter plot. While similar, however, TMP LPS and DPS concentrations were not identical as evidenced by a negligible deviation in the mean DPS/LPS ratio (0.88) and slope of the Passing-Bablock regression (0.93). Interpretation of the Passing-Bablock regression indicates that the differences between the two methods are both systematic and proportional, such that DPS concentrations are systematically lower than LPS values throughout the concentration range, but this difference is more marked at higher concentrations. When a correction factor of 0.88 (mean ratio of DPS/LPS) is applied to the DPS concentrations, the fit of the Passing-Bablock regression is improved slightly (slope 1.06, 95% CI: 1.03, 1.1; intercept −0.04, 95% CI: −0.08, −0.02). For TMP, DUS, and LUS concentrations of TMP were nearly interchangeable. Based on these results, dried matrix sampling is an appropriate method to assess TMP PK in clinical studies.

For SMX, LPS and DPS matrices yield modest agreement as evidenced by a relatively linear relationship on the scatter plot. However, SMX LPS and DPS concentrations deviate slightly from each other as evidenced by the mean DPS/LPS ratio and slope of the Passing-Bablock regression (0.86). Interpretation of the Passing-Bablock regression indicates the presence of proportional bias such that the relationship between DPS and LPS concentrations are more variable at higher concentrations. Application of a correction factor of 0.81 (DPS concentration was divided by the mean ratio of DPS/LPS) improved the fit of the Passing-Bablock regression, and proportional bias is no longer apparent (slope 1.07, 95% CI: 0.98, 1.14; intercept 0.94, 95% CI: −2.49, 0.30). A similar proportional bias was observed with SMX DUS samples, and application of a correction factor of 0.81 (mean ratio of DUS/LUS) improved the fit of the Passing-Bablock regression and decreased proportional bias, though this bias remains apparent (slope 0.88, 95% CI: 0.71, 0.98; intercept 1.43, 95% CI: −2.05, 10.57). However, due to the bias and imprecision observed between matrices, further study is needed before dried matrix sampling can be applied to assess SMX PK in clinical studies.

The present findings may be explained by a number of considerations. The physicochemical properties of TMP and SMX, and interaction with the filter paper, may contribute to differences in drug recovery from dried matrix versus liquid samples. The possibilities of inadequate recovery have been demonstrated with the drug, gabapentin. Its zwitterionic properties require special processes of drug extraction from the dried matrix as well as cleaning thereafter [24]. Previous investigations of everolimus concentrations in DBS compared to liquid plasma spots also have demonstrated a concentration-dependent level of drug extraction from the dried matrix [25]. In this study, in the clinical samples analyzed, extraction of drug at low concentrations was higher compared to extraction of drug at higher concentrations. This finding is similar to the proportionality observed in our evaluation of TMP and SMX concentrations, although in the method validation there was no evidence of concentration-dependent extraction. Proportionality has been improved previously by changes in sample preparation, including prolonged shaking time and ultrasonic vibration, as well as modifying methods of sample extraction [26]. Alternatively, we were able to demonstrate improvement in proportionality with the addition of a mathematical correction factor.

A drug’s degree of protein binding may mediate some interactions with the filter paper and the ability to extract drug from the dried matrix. The differential protein binding between TMP (40%) and SMX (70%) could potentially explain some of the increased variability observed in SMX compared with TMP evaluations [27]. Previous investigators have demonstrated that the addition of pasteurized plasma proteins to the filter paper used in DBS has been useful to deactivate the sites of absorption on the filter paper, making it easier to extract drug [28]. Chemicals also have been used to denature proteins and help promote drug extraction [29]. The benefit of these processes is the ability to potentially increase drug recovery; however, the addition of pasteurized proteins or other chemicals to the dried matrix also may contribute to matrix effects and interfere with the accuracy of the correlation of drug concentrations between liquid and dried plasma matrices.

The method of assessment of comparability has been a point of contention and discussion in the literature [21,30]. Previous authors have stressed the importance of consideration of linearity, normal versus non-parametric distribution of the data, and the potential for misclassification of method bias with the wrong method of comparability [21,30]. The methods used in this analysis are thought to be beneficial because of the lack of sensitivity to outliers and nonparametric methodology. In addition, the Passing-Bablock method has been used in previous comparisons of dried and liquid matrices, with good and consistent results [21,31]. Use of multiple methods in this analysis revealed similar findings for each drug, thereby confirming the appropriateness of the methodology.

Supplementary Material

Figure S1

Figure S1. Representative calibration curve plots for trimethoprim (top, ng/mL) and sulfamethoxazole (bottom, ng/mL) in dried plasma spots. Calibration standards and quality control samples are denoted as circles and triangles, respectively.

Figure S2

Figure S2. Representative calibration curve plots for trimethoprim (top, ng/mL) and sulfamethoxazole (bottom, ng/mL) in dried urine spots. Calibration standards and quality control samples are denoted as circles and triangles, respectively.

Figure S3

Figure S3. Representative chromatograms for lower limit of quantification (LLOQ) of trimethoprim and sulfamethoxazole samples. MRM = multiple reaction monitoring.

Figure S4

Figure S4. Representative chromatograms for blank samples. MRM = multiple reaction monitoring.

Figure S5

Figure S5. Representative chromatograms for zero samples. MRM = multiple reaction monitoring.

Future perspective.

Pediatric PK trials are challenging because of the limitations surrounding sparse sampling and blood volume requirements. Dried matrix sampling is an innovative approach to collect PK samples using limited blood volume. To be successful, dried matrix samples must be coupled with a highly sensitive analytical method (due to small sample volumes). When available, dried matrix PK samples can be collected to characterize drug disposition in children while minimizing the ethical concerns related to blood sampling in this very vulnerable population. Also, because of the ease of sample collection and processing with this innovative technology, samples can be collected and stored more effectively relative to wet matrices. For drugs where concentrations in dried matrix samples can serve as a surrogate for plasma concentrations, collection of DPS/DUS samples alone can be used to characterize the PK of drugs in children. DPS samples have the advantage of avoiding the “hematocrit effect” on blood spot homogeneity, matrix effects, and assay variability, which can be observed with DBS sample analysis.

Executive summary.

• A precise, accurate, and reproducible LC/MS-MS method was validated and applied to simultaneously quantify TMP-SMX in DPS and DUS samples collected in an opportunistic pediatric study.

• The LLOQ for TMP was 100 ng/mL in DPS and 500 ng/mL in DUS samples; for SMX, the LLOQ was 1000 ng/mL for both matrices.

• Three cleansing punches and a 6 mm punch size were optimal to prevent punch carryover and homogeneity issues, respectively.

• Dried matrix sampling is an accurate and precise option for evaluation of TMP concentrations in pediatric patients.

• Optimal bioanalysis conditions must be assessed and the application of this approach must be evaluated in clinical trials before there is widespread implementation of dried matrix sampling for drug measurement in children.

Defined key terms.

• Sulfamethoxazole: a bacteriostatic sulfonamide antimicrobial agent.

• Trimethoprim: a bacteriostatic antimicrobial agent belonging to the dihydrofolate reductase inhibitor drug class.

• Dried plasma spots: a sampling method whereby plasma is spotted on a collection card.

• Dried urine spots: a sampling method whereby urine is spotted on a collection card.