Therapeutic drug monitoring of tacrolimus and mycophenolic acid in outpatient renal transplant recipients using a volumetric dried blood spot sampling device

Abstract

Aims

Tacrolimus and mycophenolic acid dosing after renal transplantation is individualized through therapeutic drug monitoring (TDM). Home‐based dried blood spot (DBS) sampling has the potential to replace conventional TDM sampling at the clinic. A liquid chromatography–tandem mass spectrometry (LC–MS/MS) assay was developed to quantify tacrolimus and mycophenolic acid in DBS and clinically validated for abbreviated area under the concentration–time curve (AUC) monitoring using an innovative volumetric DBS sampling device.

Methods

Clinical validation was performed by direct comparison of paired DBS and whole blood (WB) (tacrolimus) and plasma (mycophenolic acid) concentrations and AUCs. Agreement was evaluated using Passing–Bablok regression, Bland–Altman analysis and DBS‐to‐WB predictive performance. TDM dosing recommendations based on both methods were compared to assess clinical impact.

Results

Paired tacrolimus (n = 200) and mycophenolic acid (n = 192) DBS and WB samples were collected from 65 kidney(–pancreas) transplant recipients. Differences for tacrolimus and mycophenolic acid were within ±20% for 84.5% and 76.6% of concentrations and 90.5% and 90.7% of AUCs, respectively. Tacrolimus and mycophenolic acid dosing recommendation differences occurred on 44.4% and 4.7% of occasions. Tacrolimus DBS dosing recommendations were 0.35 ± 0.14 mg higher than for WB and 8 ± 3% of the initial dose. Mycophenolic acid DBS dosing recommendations were 23.3 ± 31.9 mg lower than for plasma and 2 ± 3.5% of the initial dose.

Conclusions

Tacrolimus and mycophenolic acid TDM for outpatient renal transplant recipients, based on abbreviated AUC collected with a DBS sampling device, is comparable to conventional TDM based on WB sampling. Patient training and guidance on good blood‐spotting practices is essential to ensure method feasibility.

What is Already Known about this Subject

• Extensive pharmacokinetic variability necessitates immunosuppressant therapeutic drug monitoring (TDM)‐guided dose individualization after renal transplantation.

• Current immunosuppressant TDM is suboptimal, as patients need to visit the clinic and, in the case of abbreviated area under the concentration–time curve monitoring, remain there for several hours.

• The dried blood spot approach could enable home‐based TDM but sampling remains challenging.

What this Study Adds

• Tacrolimus and mycophenolic acid can be accurately quantified in dried blood spots for TDM purposes.

• Volumetric sampling combined with ʻwhole spotʼ bioanalysis is a feasible approach to circumvent haematocrit‐related dried blood spot sampling issues.

• Patient‐friendly volumetric dried blood spot sampling has the potential to enable the home‐based TDM of tacrolimus and mycophenolic acid in stable outpatient renal transplant recipients.

Introduction

The prevention of graft rejection after renal transplantation is currently typically established through lifelong immunosuppression with http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=6784 and http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=6832 with or without prednisolone, with less common regimens including cyclosporine, everolimus, sirolimus or belatacept 1, 2. As immunosuppressants show extensive inter‐ and intrapatient pharmacokinetic variability and a narrow therapeutic window, therapeutic drug monitoring (TDM)‐guided dose individualization for these agents is essential to limit sub‐ and supratherapeutic drug exposure and concurrent risk for graft rejection and toxicity 3, 4, 5, 6, 7. Although there is a general consensus that the full area under the concentration–time curve (AUC) represents the most accurate estimation of drug exposure 8, its time‐consuming sampling procedure renders this approach inoperable for routine TDM purposes. For tacrolimus, TDM is therefore based on trough concentration or abbreviated AUC as determined with a limited sampling strategy and maximum a posteriori (MAP) Bayesian estimation 8. Abbreviated AUC monitoring is generally considered an adequate marker for drug exposure, but the abiding absence of a conclusive prospective validation of its relationship with patient outcomes still raises controversy as to whether this approach should currently be applied in routine clinical care 8. For mycophenolic acid, abbreviated AUC monitoring is a more widely applied practice, but some controversy exists on this topic as well 9. At our centre, the available evidence for both tacrolimus and mycophenolic acid is deemed sufficient, and abbreviated AUC monitoring of both agents is incorporated in routine clinical care. For stable renal transplant recipients >1 year post‐transplantation, tacrolimus is monitored based on trough concentration (every 3 months) and abbreviated AUC (annually), and for mycophenolic acid abbreviated AUC monitoring is performed once every 1–2 years. For conventional abbreviated AUC and trough concentration monitoring, whole blood (WB) sampling takes place at the clinic by way of venepuncture. Unfortunately, this approach still presents a substantial patient burden as it necessitates the patient to visit the clinic, and in the case of abbreviated AUC monitoring, remaining there for several hours. Moreover, options regarding the frequency of monitoring and the choice of sampling time points are limited within this method, as frequent (long) visits to the clinic are patient unfriendly. The dried blood spot (DBS) approach combined with finger‐prick sampling poses a viable strategy to improve outpatient drug monitoring, as it enables the option of home‐based TDM 10, 11, 12. It thereby creates options for more frequent monitoring based on optimal sampling time points 12, at an overall reduced patient burden and higher cost efficiency 13 as compared with conventional TDM. Over the past decade, various methods for immunosuppressant determination in DBS samples have been described, with the most recent advances including multi‐analyte assays for simultaneous quantification of up to five immunosuppressants from one blood spot 14, 15, 16, 17, 18, 19, 20, 21. Unfortunately, only a few of these multi‐analyte assays have been clinically validated 17, 20, 21, 22, 23, 24. Additionally, interpretation of these studies often proves cumbersome because of a limited sample size, unsatisfactory statistical work or the absence of clinical acceptance limits. Aside from these issues, the bioanalysis of DBS samples still encompasses challenges. In particular, issues regarding sampling volume and haematocrit have proven difficult to resolve 10, 12, 25, 26. Although various strategies have been proposed and explored to circumvent or correct for these effects, the need for a straightforward and patient‐friendly solution remains, and this has hindered the widespread implementation of DBS in routine clinical care 10, 25. Here, we present a strategy combining an innovative patient‐friendly volumetric DBS sampling device with ʻwhole spotʼ bioanalysis to overcome the haematocrit challenge and pave the way for home‐based immunosuppressant TDM. The objectives of the current study were: (i) to develop a liquid chromatography–tandem mass spectrometry (LC–MS/MS) assay capable of quantifying tacrolimus and mycophenolic acid in DBS samples; and (ii) to perform a clinical validation of the method for abbreviated AUC monitoring, using a patient‐friendly DBS device for standardized volume blood sampling. Ultimately, we aimed to provide outpatient renal transplant recipients and their healthcare professionals with a straightforward and accessible option for home‐based immunosuppressant monitoring.

Methods

Patients and samples

Blood samples were obtained from patients enrolled in the Reducing Renal Function Deterioration (RRFD) study, which is currently being conducted at the Leiden University Medical Center (LUMC) in the Netherlands (Netherlands Trial Registry, NTR7256). Kidney(−pancreas) transplant recipients >1 year post‐transplantation with a creatinine clearance of >25 ml min⁻¹ [as estimated with the Chronic Kidney Disease Epidemiology Collaboration (CKD‐EPI) formula] 27, who were on a tacrolimus (Advagraf®)‐based immunosuppressive regimen were eligible to participate in the study. The study was approved by the medical ethics committee of the LUMC and patients gave written informed consent. At their baseline visit to the clinic, each patient provided four DBS samples obtained by finger prick and four paired WB samples obtained by venepuncture. The average time between WB and DBS sample collection was <5 min. Patients were assisted by a nurse practitioner with the sampling of the first spot, and the sampling of the remaining spots was performed by the patient. Sampling took place predose (C₀) and every hour over the course of 3 h after drug administration (C₁, C₂ and C₃). Prior to DBS sampling, the fingertip was cleansed with water, dried and punctured using a safety lancet (Sarstedt, Nümbrecht, Germany). The first drop of blood was discarded. DBS sample collection was performed using the HemaXis™ DBS sampling device, combining a microfluidic plate for standardized volumetric blood sampling (10 μl) and a conventional Whatman™ 903 protein saver card for blood collection (DBS System, Gland, Switzerland) 28. After collection, DBS samples were transferred to the laboratory of the LUMC Department of Clinical Pharmacy and Toxicology. Samples were dried for at least 24 h and stored at room temperature while awaiting bioanalysis. Prior to bioanalysis, the size, form and consistency of all spots were visually inspected by a laboratory technician, and in the case of a deviation approved or disapproved by a clinical pharmacist to ensure sample quality. Demographic information and clinical parameters at baseline were extracted from the electronic patient record, including gender; transplant type; immunosuppressant regimen and dosages; age; time post‐transplantation; estimated glomerular filtration rate (eGFR); and hematocrit. Tacrolimus and mycophenolic acid DBS concentrations; WB (tacrolimus) concentrations; plasma (mycophenolic acid) concentrations; and abbreviated AUC values and dosing recommendations based thereon were obtained from the electronic laboratory information system.

Bioanalysis

DBS assay

For the determination of tacrolimus and mycophenolic acid in DBS samples, a new assay capable of quantifying tacrolimus, sirolimus, everolimus, cyclosporine and mycophenolic acid in DBS simultaneously using two LC–MS/MS systems, was developed. This method was based on previously described assays for quantification of these agents in DBS 18, 20, 29. A more detailed description of the technical aspects and analytical validation of the developed assay are provided in the Supporting Information Appendix S1.

WB assay

Quantification of tacrolimus in WB samples was performed using a previously validated LC–MS/MS assay capable of determining tacrolimus, sirolimus, everolimus and cyclosporine simultaneously in WB 30. For the quantification of mycophenolic acid in plasma, a new LC–MS/MS assay was developed. This assay was based on two previously described methods for the quantification of mycophenolic acid in plasma, using high‐performance liquid chromatography ultraviolet detection 31 and LC–MS/MS, respectively 32. A more detailed description of the technical aspects and analytical validation of the developed assay are provided in the Supporting Information Appendix S2.

Statistical analysis

Statistics were performed in GraphPad Prism version 7.02 (GraphPad Software, San Diego, CA, USA) and Microsoft Office Excel (Microsoft Inc., Redmond, WA, USA) add‐in Analyse‐it statistics software (Analyse‐it Software, Leeds, UK). Sparsely sampled AUC MAP Bayesian estimation was performed using MW/Pharm 33 version 3.83 (Mediware, Groningen, the Netherlands), based on models for tacrolimus C₀, C_1, C₂ and C₃ yielding the estimated AUC from time zero to 24 h (AUC_0–24) 34 and mycophenolic acid C₀, C₁, C₂ and C₃ yielding the estimated AUC from time zero to 12 h (AUC_0–12) 35, 36, translated for application in MW/Pharm. Although limited sampling strategies with a higher predictive performance are available for the estimation of either tacrolimus AUC_0–24 or mycophenolic acid AUC_0–12, this four‐time‐point sampling schedule is the most practical and clinically feasible method for estimating the AUC of both agents based on the same sampling time points, while ensuring adequate estimation accuracy. Tacrolimus and mycophenolic acid dosing recommendations based on DBS, WB and plasma AUC were formulated by clinical pharmacists, with AUC values greater than ±20% off target resulting in a dose adjustment recommendation, as these exceed the extent of variation expected to arise from intrapatient pharmacokinetic variability. Both tacrolimus and mycophenolic acid exert considerable intrapatient variability (magnitude of around 20‐30% 37, 38), it would not be fruitful to perform dose corrections based on AUC fluctuations that arise from normal intrapatient variability. The minimum number of paired AUC values was set at 40 per drug, in accordance with the Clinical and Laboratory Standards Institute (CLSI) EP09‐A3 guidelines 39. Passing–Bablok regression analysis was performed to investigate any linear relationship between the methods 40. Method agreement was evaluated using Bland–Altman analysis 41. Analogous to the European Medicines Agency (EMA) guidelines on bioanalytical method evaluation, method agreement was considered sufficient if differences between the two methods were within ±20% of the average difference for ≥67% of the samples 42. Additionally, DBS to WB or plasma concentration predictive performance was assessed using the mean percentage prediction error (MPPE) for bias and the mean absolute prediction error (MAPE) for imprecision, as suggested by Sheiner et al. 43. Acceptance limits for the MPPE and MAPE were set at <15%. For method agreement on the AUC level, a clinical acceptance limit of ±20% around the ratio identity line was applied, as an AUC divergence greater than ±20% off target results in a dose adjustment recommendation and is therefore clinically relevant. Differences in dosing recommendations between the methods were evaluated to assess the clinical impact of any differences between the methods. Dose difference limits for tacrolimus and mycophenolic acid were set at 0.5 mg and 250 mg, as based on their respective lowest commercially available oral dosages.

Clinical feasibility

To identify possible feasibility issues for the clinical implementation of the method, any comments in the electronic patient file describing difficulties or anomalies regarding sampling or analysis of the DBS cards or the DBS sampling device were registered and evaluated. Evaluation of the intended DBS card logistic process in the home‐based setting, which will be essential for eventual clinical implementation of the method, was not possible within the present study as all sampling procedures were performed at the clinic.

Translation of DBS to WB concentration

In DBS, drug concentrations are determined in capillary WB, whereas conventional TDM of tacrolimus and mycophenolic acid is performed with concentrations determined in venous WB and plasma, respectively. As capillary blood is physiologically different from venous WB, and especially plasma, drug concentrations measured in these media differ as a result of differences in drug partitioning within the plasma and the various blood cells 10, 11. Hence, correction of DBS concentrations is often necessary to enable comparison of concentrations determined in DBS and WB or plasma 10, 11. Moreover, immunosuppressant trough and AUC target values currently applied in the clinic are mostly based on WB (tacrolimus) or plasma (mycophenolic acid) drug concentrations. As the use of different target values for different types of matrices is not clinically feasible, DBS concentrations need to be translated to WB concentrations to ensure correct clinical interpretation. The need for and extent of such a correction factor are assessed in a clinical validation study, yielding a drug‐specific DBS to WB conversion factor. Correction of DBS concentrations is performed based on the patient‐specific or average haematocrit, often in combination with the red blood cell to plasma partition ratio 25, 26. Translation strategies without haematocrit correction are performed based on the average DBS to WB or plasma ratio 21, 44, 45 or the DBS to WB or plasma concentration regression fit slope or equation 44, 46, 47, 48, 49. To investigate the need for and extent of correction of tacrolimus and mycophenolic acid DBS concentrations in the present study, a clinical pilot study was conducted from February to June 2016. The pilot study included 18 stable renal transplant recipients who were on a tacrolimus‐ (n = 9) or mycophenolic acid‐ (n = 9) based regimen. Sample collection, bioanalysis and statistics were performed according to the protocol of the main validation study. The pilot study was approved by the medical ethics committee of the LUMC and patients gave written informed consent. For tacrolimus, 32 paired DBS and WB samples were collected. The geometric mean DBS to WB concentration ratio was 1.10 [95% confidence interval (CI) 1.03, 1.16; range 0.79 to 1.45). Based on these findings, a conversion factor was deemed unnecessary. For mycophenolic acid, 36 paired DBS and WB samples were collected. The geometric mean DBS to plasma ratio was 0.68 (95% CI 0.65, 0.71; range 0.51 to 0.99). Correction of DBS concentrations based on this ratio showed a better predictive performance for plasma concentrations, with a lower variability [coefficient of variation (%CV)] compared with correction based on the patient‐specific or average haematocrit. This correction factor was therefore implemented in the main study by dividing mycophenolic acid DBS concentrations by 0.68, yielding the corrected mycophenolic acid DBS concentration (DBS_C).

Nomenclature of targets and ligands

Key ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to Pharmacology 50.

Results

Patients and samples

Between February 2017 and May 2018, 65 patients had completed their baseline visit. Patient characteristics with regard to gender, transplant type, immunosuppressant regimen, age, time post‐transplantation, renal function and haematocrit are presented in Table 1. In total, 200 and 192 paired tacrolimus and mycophenolic acid samples, respectively, were collected and 65 paired tacrolimus AUC_0–24 values and 49 paired mycophenolic acid AUC_0–12 values were estimated. Estimation of one tacrolimus DBS AUC_0–24 and one mycophenolic acid DBS_c AUC_0–12 were deemed unreliable because of divergent blood spot size, due to blood clotting within the sampling device capillary. In addition, the estimations of one tacrolimus AUC_0–24 pair and five mycophenolic acid AUC_0–12 pairs were deemed unreliable owing to the absence of a definite trough or peak level. These were excluded from the AUC analysis. In total, 63 paired tacrolimus AUC_0–24 and 43 paired mycophenolic acid AUC_0–12 values and concurrent dosing recommendations were included in the AUC analysis.

Table 1.

Summary of patient characteristics

Parameter	Mean	95% CI	Range
N	65
Gender (male/female)	44/21
Transplant type (KTx/KPTx)	53/12
Immunosuppressant regimen (CT/DT)	52/13
Age (years)	53.9	50.8, 57.0	23.4–76.4
Time post‐transplantation (years)	5.5	4.7, 6.3	1.0–7.3
eGFR (ml min −1 1.73 m −2 )	55.3	51.5, 59.1	27.0–84.0
Haematocrit (%)	0.41	0.40, 0.42	0.33–0.55

CI, confidence interval; KTx, kidney transplant; KPTx, kidney–pancreas transplant; CT, combination therapy of tacrolimus and mycophenolic acid (±prednisolone); DT, dual therapy of tacrolimus and prednisolone without mycophenolic acid; eGFR, estimated glomerular filtration rate (CKD‐EPI).

Tacrolimus

Tacrolimus WB and DBS concentrations and estimated AUC_0–24 values are presented in Table 2. Passing–Bablok regression analysis showed no proportional or constant bias, as shown in Table 3 and depicted in Figure 1A. Bland–Altman analysis showed a statistically significant absolute and ratio difference of tacrolimus DBS to WB concentrations, as presented in Table 3 and depicted in Figure 1B and Figure 1C, respectively. In total, 84.5% (169/200) of the paired concentrations fell within ±20% of the average ratio, thereby complying with the predefined minimum of ≥67%. MPPE (−7.38%) and MAPE (12.92%) were both within the acceptance limit of <15%, showing good predictive performance of DBS concentrations for WB concentrations. Additional analysis for evaluation of trough concentration monitoring compatibility (n = 63), showed a geometric mean C₀ DBS to WB concentration ratio of 0.93 (95% CI 0.88, 0.98; range 0.52 to 2.03). Furthermore, 22.2% of the C₀ DBS to WB concentration ratios exceeded the ±20% limits around the ratio identity line, which revealed that this method is not suitable for trough concentration‐based TDM of tacrolimus. Passing–Bablok regression analysis on the AUC_0–24 level showed some proportional bias and no constant bias, as shown in Table 3 and depicted in Figure 1D. Bland–Altman analysis showed a statistically significant absolute and ratio bias for tacrolimus DBS to WB AUC_0–24, as shown in Table 3 and depicted in Figure 1E and Figure 1F, respectively. In total, 90.5% (57/63) of the paired AUC_0–24 values fell within ±20% of the ratio identity line, thereby complying with the predefined minimum of ≥67% (Figure 1F). The MPPE (−7.04%) and MAPE (−9.92%) were both within the acceptance limit of <15%, showing good predictive performance of DBS AUC_0–24 values for WB AUC_0–24 values. The average daily tacrolimus dose at baseline was 4.03 mg (95% CI 3.51, 4.55; range 1.0 to 11.0). On 44.4% (28/63) of the occasions, a difference between dosing recommendations based on DBS AUC_0–24 and WB AUC_0–24 was observed. On average, dosing recommendations based on DBS AUC_0–24 were 0.35 mg (95% CI 0.21, 0.49; range −0.50 to 3.00 mg) higher than those based on WB AUC_0–24. To investigate the relative impact of the dosing recommendation differences, these were plotted against initial daily tacrolimus doses (Figure 2). On average, dosing recommendation difference to dose ratios were 0.08 (95% CI 0.05, 0.11; range −0.17 to 0.40).

Table 2.

Summary of sample characteristics

Parameter	N	Mean	95% CI	Range
Tacrolimus WB concentration (μg l −1 )	200	9.05	8.39, 9.71	1.26–22.56
Tacrolimus DBS concentration (μg l −1 )	200	8.35	7.71, 9.00	1.15–23.54
Tacrolimus DBS : WB concentration ratio	200	0.93	0.90, 0.95	0.50–2.03
Tacrolimus WB AUC 0–24 (μg h l −1 )	63	165.3	151.0, 179.6	44–336
Tacrolimus DBS AUC 0–24 (μg h l −1 )	63	152.7	139.1, 166.3	43–339
Mycophenolic acid plasma concentration (mg l −1 )	192	5.47	4.74, 6.20	0.38–34.84
Mycophenolic acid DBS concentration (mg l −1 )	192	3.46	2.99, 3.92	0.24–20.04
Mycophenolic acid DBS c concentration (mg l −1 )	192	5.08	4.40, 5.76	0.35–29.47
Mycophenolic acid DBS c : plasma concentration ratio	192	0.98	0.92, 1.03	0.46–4.75
Mycophenolic acid plasma AUC 0–12 (mg h l −1 )	43	42.8	37.5, 48.1	6–90
Mycophenolic acid DBS c AUC 0–12 (mg h l −1 )	43	41.2	35.9, 46.6	5–101

AUC, area under the concentration–time curve; CI, confidence interval; DBS, dried blood spot; DBS_c, corrected dried blood spot concentration; WB, whole blood

Table 3.

Summary of method agreement between tacrolimus and mycophenolic acid DBS vs. WB sampling

Parameter	Passing–Bablok		Bland–Altman absolute differences		Bland–Altman ratio differences
	Slope [95% CI]	Intercept [95% CI]	Bias [95% CI]	95% LOA (bias ±1.96 SD)	Bias [95% CI]	95% LOA (bias ±1.96 SD)
Tacrolimus DBS vs. WB concentration	0.96 [0.92, 1.00]	−0.23 [−0.53, 0.07]	−0.70 [−0.90, −0.50]	−3.51–2.11	0.93 [0.90, 0.95]	0.60–1.25
Tacrolimus DBS vs. WB AUC 0–24	0.91 [0.84, 0.99]	3.07 [−9.85, 12.82]	−12.6 [−17.2, −7.98]	−49.3–24.1	0.93 [0.90, 0.96]	0.72–1.14
Mycophenolic acid DBS c vs. plasma concentration	0.90 [0.86, 0.94]	0.05 [−0.05, 0.17]	−0.39 [−0.62, −0.16]	−3.55–2.77	0.98 [0.84, 1.12]	0.84–1.12
Mycophenolic acid DBS c vs. plasma AUC 0–12	0.96 [0.85, 1.10]	−0.78 [−5.10, 3.77]	−1.56 [−3.31, 0.19]	−13.01–9.90	0.97 [0.93, 1.00]	0.73–1.20

AUC, area under the concentration–time curve; CI, confidence interval; DBS, dried blood spot; LOA, limits of agreement; SD, standard deviation; WB, whole blood

Figure 1.

Evaluation of method agreement of tacrolimus monitoring based on dried blood spot (DBS) vs. whole blood (WB) sampling. (A) Passing–Bablok fit of tacrolimus DBS concentrations vs. WB concentrations (solid black line), with 95% confidence interval (dotted black lines) and line of identity (solid grey line). (B) Bland–Altman absolute difference plot of tacrolimus DBS concentrations vs. WB concentrations, with mean difference (solid grey line) and upper and lower limits of agreement (ULA; LLA: dotted grey lines). (C) Bland–Altman ratio difference plot of tacrolimus DBS concentrations vs. WB concentrations, with mean difference (solid grey line) and ULA and LLA (dotted grey lines). The shaded area represents the ±20% limits around the mean ratio. (D) Passing–Bablok fit of tacrolimus DBS area under the concentration–time curve from time zero to 24 hours (AUC_0–24) vs. WB AUC_0–24 (solid black line), with 95% confidence interval (dotted black lines) and line of identity (solid grey line). (E) Bland–Altman absolute difference plot of tacrolimus DBS AUC_0–24 vs. WB AUC_0–24, with mean difference (solid grey line) and ULA and LLA (dotted grey lines). (F) Bland–Altman ratio difference plot of tacrolimus DBS AUC_0–24 vs. WB AUC_0–24, with mean difference (solid grey line) and ULA and LLA (dotted grey lines). The shaded area represents the ±20% limits around the line of identity (DBS AUC_0–24 = WB AUC_0–24)

Figure 2.

Initial tacrolimus dosages, dried blood spot and whole blood dosing recommendations and concurrent dosing recommendation differences for each individual occasion. Occasions were sorted from lowest to highest initial tacrolimus dose to improve the readability of the figure. The shaded area represents the dosage range of the lowest commercially available oral tacrolimus dose

Mycophenolic acid

Average mycophenolic acid plasma, DBS and DBS_c concentrations and AUC_0–12 values are shown in Table 2. Passing–Bablok regression analysis showed some proportional bias and no constant bias, as shown in Table 3 and depicted in Figure 3A. Bland–Altman analysis showed a statistically significant absolute bias and no statistically significant ratio bias of mycophenolic acid DBS_c to plasma concentrations, as shown in Table 3 and depicted in Figure 3B and Figure 3C, respectively. In total, 76.6% (147/192) of the paired concentrations fell within ±20% of the average ratio, thereby complying with the predefined minimum of ≥67%. The MPPE (−2.48%) was within the acceptance limit of <15%, whereas the MAPE (18.66%) exceeded it, showing moderate predictive performance of DBS_c concentrations for plasma concentrations. Passing–Bablok regression analysis on the AUC_0–12 level showed no proportional or constant bias (Figure 3D). Bland–Altman analysis showed no statistically significant absolute or ratio bias of mycophenolic acid DBS_c to plasma AUC_0–12 values, as shown in Table 3 and depicted in Figure 3E and Figure 3F, respectively. In total, 90.7% (39/43) of the paired values fell within ±20% of the ratio identity line, thereby complying with the predefined minimum of ≥67% (Figure 3F). The MPPE (−3.50%) and MAPE (10.60%) were both within the acceptance limit of <15%, showing a good predictive performance of DBS_c AUC_0–12 for plasma AUC_0–12. The average daily mycophenolic acid dose at baseline was 1226 mg (95% CI 1109, 1344; range 500 to 2500). On 4.7% (2/43) of the occasions, a difference was observed between dosing recommendations based on DBS_c AUC_0–12 and dosing recommendations based on plasma AUC_0–12. On average, dosing recommendations based on DBS_c AUC_0–12 were 23.26 mg (95% CI –55.1 to 8.59; range −500 to 0.0) lower as compared with those based on plasma AUC_0–12. Dosing recommendation differences in relation to the initial daily mycophenolic acid doses are shown in Figure 4. Dosing recommendation difference to dose ratios were −0.02 (95% CI –0.06 to 0.01), on average, and 0 (n = 41) or −0.50 (n = 2).

Figure 3.

Evaluation of method agreement of mycophenolic acid monitoring based on dried blood spot (DBS) vs. whole blood sampling. (A) Passing–Bablok fit of mycophenolic acid corrected dried blood spot concentration (DBS_c) concentrations vs. plasma concentrations (solid black line), with 95% confidence interval (dotted black lines) and line of identity (solid grey line). (B) Bland–Altman absolute difference plot of mycophenolic acid DBS_c concentrations vs. plasma concentrations, with mean difference (solid grey line) and upper and lower limits of agreement (ULA; LLA: dotted grey lines). (C) Bland–Altman ratio difference plot of mycophenolic acid DBS_c concentrations vs. plasma concentrations, with mean difference (solid grey line) and ULA and LLA (dotted grey lines). The shaded area represents the ±20% limits around the mean ratio. (D) Passing–Bablok fit of mycophenolic acid DBS_c area under the concentration–time curve from time zero to 12 hours (AUC_0–12) vs. plasma AUC_0–12 (solid black line), with 95% confidence interval (dotted black lines) and line of identity (solid grey line). (E) Bland–Altman absolute difference plot of mycophenolic acid DBS_c AUC_0–12 vs. plasma AUC_0–12, with mean difference (solid grey line) and ULA and LLA (dotted grey lines). (F) Bland–Altman ratio difference plot of mycophenolic acid DBS_c AUC_0–12 vs. plasma AUC_0–12, with mean difference (solid grey line) and ULA and LLA (dotted grey lines). The shaded area represents the ±20% limits around the line of identity (DBS_c AUC_0–12 = plasma AUC_0–12)

Figure 4.

Initial mycophenolic acid dosages, dried blood spot (DBS) and plasma dosing recommendations and concurrent dosing recommendation differences for each individual occasion. Occasions were sorted from lowest to highest initial mycophenolic acid dose to improve the readability of the figure. The shaded area represents the dosage range of the lowest commercially available oral mycophenolic acid dose. WB, whole blood

Clinical feasibility

The feasibility of the DBS sampling device was evaluated on two domains: sampling and analysis. For the sampling process, some difficulties and anomalies arose. Eight patients (12.3%) had taken their medication prior to the visit, rendering some C₀ concentrations unreliable and concurrent AUC estimation and dosing recommendations impossible or questionable. Furthermore, the research assistants reported sampling difficulties related to the DBS device capillary for 15 patients (23.1%). For one patient (1.5%), excessive squeezing of the finger to generate a sufficient blood volume was reported. Additionally, three patients (4.6%) reported sampling difficulties due to impaired vision, not being able to see the DBS device capillary clearly. Issues regarding the analysis of the DBS samples resembled these sampling issues to some extent, with eight spots discarded owing to insufficient quality or size.

Discussion

In the present study, the clinical validation of a new DBS method for tacrolimus and mycophenolic acid monitoring in outpatient kidney(–pancreas) transplant recipients was presented. Differences between use of DBS and WB sampling with regard to the individual concentrations were within predefined acceptance limits for tacrolimus, showing adequate agreement. The predictive performance of tacrolimus DBS concentrations for their respective WB concentrations, as shown with the MPPE and MAPE, were consistent with these findings. For mycophenolic acid, the MPPE at the concentration level was within the predefined acceptance limit, whereas the MAPE exceeded this boundary, showing predictive performance of mycophenolic acid DBS_c concentrations for their respective plasma concentrations to be moderate at best. However, the clinical relevance of these results at the individual concentration level can be considered relatively low, as mycophenolic acid TDM is ideally performed based on AUC_0–12 and not on trough concentrations. By contrast, for tacrolimus, the results at both the individual concentration and AUC_0–24 level are clinically relevant, as tacrolimus TDM can be performed with both trough concentrations and AUC_0–24. For both agents, translation into AUC resulted in higher agreement between the methods, showing that use of the new DBS sampling method results in accurate estimation of tacrolimus and mycophenolic acid exposure. These findings were confirmed with the MPPE and MAPE results at the AUC level for both drugs. A higher level of agreement between the methods at the AUC level was expected. Firstly, because any differences between the paired DBS and WB or plasma concentrations directly contributed to the extent of method agreement in the single concentration comparison, whereas these differences are more or less flattened out by the remaining concentrations during AUC estimation and therefore only partially contributed to the extent of method agreement in the AUC comparison. Secondly, although DBS and WB sampling times were considered identical during the paired concentration comparison, small differences in sampling times could have increased the variability in concentration differences between the methods. This is especially true for mycophenolic acid, as its rapid absorption can result in swift concentration changes around the early sampling time points. Although the average time difference between DBS and WB sampling was less than 5 min, the time difference in 50 samples was >5 min, in seven samples was >10 min and in three samples was >15 min. This may have introduced additional variability in the individual concentration comparison, in contrast to the model‐derived AUC comparison, in which exact sampling times were incorporated in the AUC estimation and any effects resulting from sampling time differences were corrected for by the population pharmacokinetic model. This has a flattening effect on any concentration differences arising from time differences between WB and DBS sampling. Although agreement on the AUC level was sufficient for both drugs, some discrepancies in TDM dosing recommendations were observed. These discrepancies arose when differences between DBS AUC and WB AUC or plasma AUC resulted in one of them exceeding the ±20% deviation from the AUC target value while the other did not, leading to different dosing recommendations. For mycophenolic acid, differences occurred on two occasions. Further examination showed two cases of borderline deviation from the AUC target value, in which a small discrepancy in AUC between the methods resulted in a different dosing recommendation. Based on their low frequency and minor AUC deviation, this was deemed acceptable. For tacrolimus, dosing recommendation differences were observed on 28 occasions. For most of these occasions, differences between DBS AUC and WB AUC were within ±20% limits and dosing recommendation differences were small, relative to the initial tacrolimus dose. On average, differences in dosing recommendations were 8% of the initial tacrolimus dose, showing an overall limited and acceptable clinical impact. On some occasions, however, larger dosing recommendation differences were observed, occurring mainly during the first few months of the study. This could possibly be explained if these patients showed a divergent blood viscosity due to extremely high or low haematocrit values, affecting the flow rate in the DBS device capillary. However, the applied DBS sampling device has been validated for haematocrit values ranging from 0.26 to 0.62 28, while our patients showed values between 0.33 and 0.55. Therefore, it seems unlikely that these differences arose from haematocrit‐related issues. Another explanation could lie in the occurrence of deviations between reported sampling times and actual sampling times, as sampling times were recorded manually. As the model‐derived AUC also relies on the accuracy of the reported sampling time, such a deviation could influence the estimated AUC and concurrent dosing recommendation. Lastly, we suspect that some patients were subject to excessive squeezing of the finger, with concurrent capillary WB haemolysis and dilution through the addition of tissue fluids 12, resulting in a reduced DBS concentration, lower AUC estimate and higher dose adjustment advice. However, this could not be confirmed based on our data. Interestingly, the performance of the method was largely similar for both agents at the individual sample and AUC level, but discrepancies between the drugs were observed in the extent of TDM dosing recommendation differences. This might be partly explained by the fact that tacrolimus TDM at our centre is targeted on a specific AUC value rather than on an AUC target range, as used for mycophenolic acid. Additionally, the number of tablets administered to achieve the daily dose is usually higher for tacrolimus than for mycophenolic acid, resulting from its lower commercially available oral dosages. In general, this results in more options for dosing adjustments for tacrolimus than for mycophenolic acid. Hence, deviations between tacrolimus DBS and WB AUC translate more easily into differences in dosing recommendations than for mycophenolic acid. Although this method is compatible for abbreviated AUC tacrolimus monitoring, a clinically unacceptable degree of deviation was observed between DBS and WB tacrolimus trough concentrations. Hence, it is not recommended to apply this DBS strategy for trough concentration‐based tacrolimus TDM at present. Possible strategies to overcome this could include sampling trough concentrations in two‐ or threefold or application of a more sensitive MS system.

With regard to the correction factors, it was interesting to observe that mycophenolic acid DBS_c concentrations were still, on average, 2.5% lower than plasma concentrations. As the correction factor of 0.68 was based on pilot data from just 32 samples, correction by a factor of 0.66, as derived from the current study, might result in an even better estimation of plasma concentrations. A correction factor for tacrolimus could also be considered, as an average deviation of −7% between DBS and WB concentrations was observed. Dividing DBS tacrolimus DBS concentrations by 0.93, based on the average DBS to WB ratio in the current study, could present a plausible correction factor. This may give an even more accurate estimation of the actual tacrolimus exposure. Other strategies – for example, correction based on the Passing–Bablok fit equation or slope, or patient‐specific or average haematocrit – could also be (re)explored.

Regarding clinical feasibility, issues were often related to the DBS device or the quality and consistency of the samples. As the concept of DBS sampling was new to the patients, and training and instruction on the specific procedure and on general good blood‐spotting practice were rather brief, this may have resulted in excessive squeezing of the finger, with concurrent tissue fluid dilution and haemolysis 12, spot‐to‐spot carry‐over, finger‐to‐spot carry‐over or sample contamination through incorrect sample or device handling 10, 26. Furthermore, some patients experienced sampling difficulties due to impaired vision. As diabetic retinopathy is common in this population, especially in kidney–pancreas transplant recipients, this might pose a feasibility issue for further clinical implementation. Additionally, the feasibility of the intended DBS card logistic process for the home‐based TDM setting remains to be assessed, as evaluation of this aspect was impossible within this particular study design.

To our knowledge, the present study was the first to apply the HemaXis™ DBS sampling device for TDM purposes in a clinical setting. The device has previously been applied for cytochrome P450 phenotyping in healthy volunteers 51 and is currently being used in a venlafaxine trial 52. Additionally, this is one of the first studies to validate clinically a multi‐analyte DBS immunosuppressant assay by direct comparison of DBS and WB tacrolimus samples 17, 20, 21, 22, 23, 24, and the third for mycophenolic acid 21, 53. Moreover, it is the first to validate a mycophenolic acid DBS method for abbreviated AUC monitoring, and the second for tacrolimus 54. Lastly, the number of paired WB and DBS samples was higher than seen in any of the previous clinical validation studies for tacrolimus and mycophenolic acid quantification in DBS samples, which adds to the importance of the present study.

When placing our results in the perspective of previous studies on the subject of tacrolimus and mycophenolic acid determination in DBS, various similarities and discrepancies can be identified. Assessment of agreement of (abbreviated) AUC values based on DBS vs. WB concentrations has previously been reported in only one study for tacrolimus 54, by Cheung et al., whereas no such studies are available for mycophenolic acid. The latter study reported tacrolimus DBS concentrations to be, on average, 11% higher than tacrolimus WB concentrations (n = 106). Similar to our observations, they reported a levelling effect on the difference with the translation to AUC, resulting in DBS AUC_0–12 values to be, on average, approximately 8% higher than WB AUC_0–12 values (n = 36) 54. However, these results cannot be directly compared with those of the present study as Cheung et al. did not apply volumetric DBS sampling or ʻwhole spotʼ bioanalysis. The subsequent absence of correction for the haematocrit complicates interpretation of their results. In addition, these authors evaluated the extent of agreement based on DBS to WB concentration differences over the average tacrolimus concentration, rather than on DBS to WB concentration ratio differences. Application of this statistical methodology renders agreement more likely than it would be with a DBS to WB concentration ratio approach, as applied in the present study. Furthermore, Cheung et al. did not compare their findings with a predefined clinical acceptance limit, nor was a comparison of TDM dosing recommendations performed. Hence, the clinical significance of the differences between DBS and WB concentrations and AUC values reported in their study remains difficult to substantiate. With regard to the DBS to WB or plasma conversion factors, a few similar studies could be identified. For tacrolimus, the observed geometric mean DBS to WB concentration ratio of 0.93 was not in line with the conversion factor of 1.31 (n = 40) reported by Martial et al., who found higher tacrolimus concentrations in DBS than in WB 21. Various other studies (n = 18–172) in adult and paediatric solid organ transplantation have been performed on the subject, and reported a range of geometric mean differences of −0.70 to +0.29 μg l⁻¹ between tacrolimus DBS and WB concentrations 17, 20, 22, 23, 24, 55, 56. These findings show similarity with the geometric mean difference of −0.70 μl l⁻¹ reported in the present study. Based on the large physiological similarity between capillary and venous WB, a conversion factor for translation of tacrolimus DBS to WB concentration would theoretically not seem necessary. However, differences in sample handling, sample preparation and the analytical process of the applied assays may result in deviation. In the present study, the agreement between use of DBS and WB sampling was found to be acceptable without the need for tacrolimus DBS to WB concentration conversion. For clinical implementation, however, a correction factor could be considered to correct for the 8% difference between the tacrolimus DBS‐ and WB‐based AUC in an attempt to further reduce differences in dosing recommendations between the methods. For mycophenolic acid, the observed geometric mean DBS to plasma concentration ratio of 0.66 was in line with those described by Martial et al. 21 and Arpini et al. 53, who reported ratios of 0.77 (n = 38) and 0.61 (n = 77), respectively. Although the proposed conversion factors for both tacrolimus and mycophenolic acid seem plausible, as based on the present study, their considerable variability calls for further exploration, optimization and external validation.

The current study was subject to some methodological limitations. Firstly, DBS sampling was performed at the clinic, where the patient was assisted by a nurse practitioner with the sampling of at least one of the spots. This may have resulted in more consistent DBS sample collection than would be expected when patients perform the sampling independently at home. Additionally, feasibility issues arising in this clinical setting might not fully represent those in the home‐based setting. However, this was the most practical way to perform simultaneous DBS and WB sampling. Secondly, the intention was to draw DBS and WB samples at exactly the same time; however, owing to practical problems, a deviation of a few minutes sometimes occurred. This may have introduced additional variability between DBS and WB and plasma concentrations, as tacrolimus, and especially mycophenolic acid concentrations, may change over the course of minutes at the early sampling times. This could have been prevented by assigning two nurse practitioners per patient to perform the DBS and WB sampling simultaneously. Thirdly, two different LC–MS/MS systems were used for DBS and WB sample analysis. Although differences between the two systems do not exceed 10%, as monitored by repeated cross‐validation, this might have introduced additional variability. Contrarily, this set‐up did resemble the daily routine bioanalytical process in which these agents are also determined on different LC‐MS/MS systems. Hence, this might have resulted in a more accurate representation of the actual clinical performance of the method than that achieved by a validation on just one LC–MS/MS system. Fourthly, AUC estimation of both tacrolimus AUC_0–24 and mycophenolic acid AUC_0–12 was performed using a sampling strategy which does not show the highest predictive performance compared with some previously published sparse sampling strategies for the individual agents. However, the applied strategy was the most practical and clinically feasible method to estimate the AUC of both drugs based on the same time points, while ensuring adequate accuracy of the exposure estimation. Most importantly, it did not influence the results of the clinical validation, in terms of method agreement, as AUC estimation was performed identically for both methods. Lastly, one could argue whether application of the EMA guidelines for evaluation of method agreement is justified for tacrolimus and mycophenolic acid, considering their narrow therapeutic index, as this acceptance limit allows for approximately one‐third of the samples to deviate ±20% from the reference method. Understandably, it is essential to take the pharmacokinetic and pharmacologic characteristics of the concerning agents into consideration when formulating a clinical acceptance limit for a clinical validation study. However, as no other guidelines on this specific topic are currently available, and as this acceptance limit is also widely applied for bioanalytical interlaboratory cross‐validation, including agents with a narrow therapeutic index, application of this acceptance limit was deemed justified.

In the light of further optimization and clinical implementation of the presented DBS sampling strategy, an array of opportunities and challenges arises. With regard to the analytical procedures, determination of all agents in a single run would be a major improvement. Additionally, simultaneous quantification of a biomarker for renal function, as previously described for creatinine 17, 29 and iohexol 57, 58, could further reduce the number of hospital visits for the patient. In addition, automated online DBS card desorption, paper spray analysis or fully automated extraction could pose interesting strategies to increase DBS throughput 10 and facilitate a process scale‐up. For feasibility, the development of a DBS device which samples directly through the finger‐prick lancet might resolve sampling issues arising from impaired vision. Another option would be to use a different microsampling device without a capillary for these patients, as for instance the Mitra® device (Neoteryx, Torrance, CA, USA). A future study should be conducted to determine the ideal type of sampling device for various patient populations. In addition, the extent of variability arising from deviations in the sampling procedure, such as sampling from different fingers, should be explored. Moreover, automated sampling time registration through addition of a time recording chip could increase sampling time accuracy and reduce sample handling time, thereby increasing TDM accuracy and limiting sample contamination. Lastly, optimization of patient instruction and training with regard to the specific sampling procedure and general good blood‐spotting practice is of key importance to limiting sampling errors and concurrent variability 10, 26.

In conclusion, abbreviated AUC monitoring of tacrolimus and mycophenolic acid, collected using a DBS sampling device, is feasible and comparable to conventional abbreviated AUC monitoring based on WB sampling at the clinic. With viable options for further optimization of the analytical method and the sampling device, a transition towards home‐based outpatient immunosuppressant abbreviated AUC monitoring in the renal transplantation setting is no longer a distant prospect. It should be noted, however, that adequate patient guidance and training on the sampling procedures and good blood‐spotting practice remain essential for this approach to be successful in clinical practice.

Competing Interests

There are no competing interests to declare.

We appreciate the assistance of the analytical staff in the Department of Clinical Pharmacy and Toxicology of the LUMC, with special thanks to Trees Hessing, Erik Metscher and Annelies Kruijthof. Part of this study was financially supported by Astellas Pharma B.V.

Supporting information

Appendix S1 Dried blood spot bioanalysis

Table S1 Summary of within‐run and between‐run accuracy and precision for quantification of tacrolimus, mycophenolic acid, sirolimus, everolimus and cyclosporine in dried blood spots using liquid chromatography‐tandem mass spectrometry

Figure S1 Calibration curves, mass spectrograms and chromatograms for quantification of tacrolimus, sirolimus, everolimus and cyclosporine in dried blood spots using liquid chromatography‐tandem mass spectrometry

Figure S2 Calibration curves, mass spectrograms and chromatograms for quantification of mycophenolic acid in dried blood spots using liquid chromatography‐tandem mass spectrometry

Appendix S2 Whole blood bioanalysis

Table S2 Summary of within‐run and between‐run accuracy and precision for quantification of mycophenolic acid in plasma using liquid chromatography‐tandem mass spectrometry

Zwart, T. C. , Gokoel, S. R. M. , van der Boog, P. J. M. , de Fijter, J. W. , Kweekel, D. M. , Swen, J. J. , Guchelaar, H.‐J. , and Moes, D. J. A. R. (2018) Therapeutic drug monitoring of tacrolimus and mycophenolic acid in outpatient renal transplant recipients using a volumetric dried blood spot sampling device. Br J Clin Pharmacol, 84: 2889–2902. 10.1111/bcp.13755.