Dose prediction of tacrolimus in de novo kidney transplant patients with population pharmacokinetic modelling.
R.R. Press(1), B.A. Ploeger(2,3), J. den Hartigh(1), R.J.H.M. van der Straaten(1), J. van Pelt(1), M.Danhof(2,3), J.H. de Fijter(1), H.J. Guchelaar(1).
(1) Departments of Clinical Pharmacy and Toxicology, Nephrology and Clinical Chemistry. Leiden University Medical Center, Leiden, the Netherlands. (2) Leiden Amsterdam Center for Drug Research (LACDR), Leiden, the Netherlands. (3) Leiden Experts on Advanced Pharmacokinetics and Pharmacodynamics (LAP&P Consultants), Leiden, the Netherlands.
Introduction: The immunosuppressive drug tacrolimus (TRL) belongs to the group of calcineurin inhibitors. The calcineurin inhibitors TRL and ciclosporin A constitute the cornerstone of immunosuppressive therapy in solid organ transplantation. TDM of these drugs in kidney transplant patients (Tx) is mandatory since they have a narrow therapeutic index. TRL is responsible for liver toxicity as well as acute and chronic nephrotoxicity. Other complications of chronic therapy are cardiovascular and neurotoxicity, diabetes and several other clinical disorders [1]. Since a number of complications are related to the blood concentration of calcineurin inhibitors, therapeutic drug monitoring (TDM) became standard of care years ago. The golden target was established as the area under the (whole blood)-concentration time curve (AUC) of TRL [2]. Dosing guided by the target AUC has improved clinical outcome in terms of reduced toxicity and improvement of graft survival. TDM is especially important during high dose therapy shortly after Tx. Individual dose adjustments are made to achieve target exposure within days after start of the body weight based regime. However, frequent dose adjustments are often required resulting in under or over exposure for a considerable amount of time. As this could result in either lack of efficacy or toxicity it is important to reduce the frequency of dose adjustments by selecting an individualized optimal starting dose. This requires insight into the factors (i.e. covariates) that explain the variability in the pharmacokinetics (PK) of TRL.
Objective: This study aims at identifying mechanistically plausible and clinically relevant covariates that explain observed variability in the PK of TRL. In addition, simulations have been performed to identify the optimal dosing schedule for TRL.
Methods: De novo kidney Tx patients (n=33) were treated with basiliximab, mycophenolate mofetil (fixed dose), prednisolone and TRL. Patients received TRL either once or twice daily. TRL dose was adjusted according to a preset target AUC. PK samples were collected up to 12 hours after administration on week 2, 4, 6, 8, 10, 12, 17, 21, 26, 39 and 52 post Tx. Whole blood concentrations of TRL were measured with microparticle enzyme immunoassay (MEIA) on an IMx-analyzer.
The pharmacokinetic data were analysed using non-linear mixed effect modelling with NONMEM V. The effects of the potential covariates hematocrit, albumin, age, weight and genetic polymorphisms in CYP3A4, CYP3A5, P-glycoprotein (P-gp) [1,3] and the nuclear hormone receptor Pregnane-X-receptor (PXR) on TRL PK were studied. Potential covariates were selected by visual inspection of the possible relationship between random effects estimated with the base model and a covariate. Subsequently, the selected covariate relationships were evaluated with NONMEM by forward inclusion and backward deletion procedure. In addition, including a covariate effect should result in a reduction in the identified random variability and an improvement of the model fit.
Results: TRL PK was adequately described by a 2-compartment model with first-order absorption and first-order elimination from the central compartment. Two populations with a significant difference in central volume of distribution (Vc) were identified. Higher Vc values were identified for subjects with hematocrit values below 0.32 and a TRL dose of greater than 10 mg compared to subjects with higher hematocrit values (32 ± 13 L vs. 17 ± 6 L). In addition, 2 populations with different values for TRL clearance were identified. This bimodal distribution could be related to genetic polymorphisms. Pharmacogenetic differences were found between these populations with genetic polymorphisms (SNPs) in CYP3A5*3 (CL = 3 ± 0.5 vs. 4.5 ± 1.5 L/h) and PXR (CL= 3 ± 0.5 vs. 4 ± 1.3 L/h). SNPs in these enzymes are responsible for higher TRL clearance compared to the wild type. With this structural relationship the identified random inter-individual variability in CL is considerably reduced. Moreover, an association between the presence of promotor SNPs CYP3A4*1B (SNP responsible for increased CL) and ABCB T-129C (SNP responsible for decreased CL) and TRL CL was observed. Finally, there was a clear relationship between the apparent clearance (CL/F) and TRL-dose.
With the developed PK model we performed simulations to support our findings. We clearly demonstrate that the body weight based regimen should be replaced by either a standard starting dose or an individualized dose based on CYP3A5*3 or PXR genotype.
Discussion: TRL is subject to highly variable PK. In the present investigation the variability in TRL pharmacokinetics is described and the inter-individual variability relevant to individualised dosing is largely explained. A bimodal distribution was observed in TRL clearance. Genetic polymorphisms were found to be responsible for this observation. A SNP in PXR, next to SNPs in CYP-enzymes and P-gp, were demonstrated to be important in relation to TRL PK. The SNPs in CYP and P-gp are known to be associated with low through blood concentrations [3]. This study demonstrates the relationship with clearance. The transcription factor PXR modulates the expression of drug metabolising enzymes (i.e. CYP3A4) and transporters (P-gp). PXR is a low affinity and high capacity nuclear receptor for glucocorticoids, such as prednisolone [4]. Within 3 weeks after Tx high dose prednisolone (50 mg b.i.d.) is reduced to 10 mg o.d. Since the effect of prednisolone through PXR could potentially lead to increased TRL clearance differences in TRL CL in time would be expected. The SNP in PXR could be of use as a biomarker for TRL CL early after Tx when (high dose) prednisolone is administered to the patient.
In addition, for subjects with hematocrit values below 0.32 the TRL binding capacity of the red blood cells is lower after dose over 10 mg, resulting in a higher apparent volume of distribution, since more unbound TRL will distribute into peripheral tissue. This non-linearity could have great clinical implications (delayed graft function, liver toxicity, neurotoxicity) for the early period after Tx when patients with often low hematocrit values receive high dose TRL treatment.
During the analysis we observed a relationship between TRL clearance and dose. This can be explained by the fact that TDM is performed on basis of blood level monitoring. Patients with high blood levels have a low clearance and vice versa. Dose adjustments are made on basis of drug concentration therewith indirectly on basis of the drug clearance. Therefore, a higher clearance necessitates a higher dose. As this relationship was not observed after the first dose, but becomes visible after the first dose adjustment, this relationship is clinically introduced by TDM as was shown earlier for carbamazepine [5]. No relationship between bodyweight and the apparent CL of TRL was detected in the dataset. On theoretical grounds this was expected since TRL distributes extensively to tissues throughout the body and the blood fraction is bound to the red blood cell for more then 80%. This indicates that bodyweight is a poor predictor of TRL CL and could potentially lead to respectively under- or overdosing in patients with a low- or high bodyweight. The first TDM dose adjustment also corrects for this misassumption. Simulations with the PK model clearly show that a fixed dose regimen is preferred compared to dosing per kg body weight. Our findings show that a fixed dose in combination with TDM results in lesser dose adjustments. However, a more sophisticated way in improving dose predictions would be individualization on the basis of the covariate-adjusted population PK model with the implementation of genotyping and factors as hermatocrit as determinants in the dose selection of TRL.
Conclusion: Genetic polymorphisms in CYP3A5 and PXR are responsible for the observed variability in TRL clearance. In addition, for subjects with hematocrit values below 0.32 the TRL binding capacity of the red blood cells is lower in combination with a dose over 10 mg, resulting in a higher apparent volume of distribution. In combination with the poor predictive power of the bodyweight based regimen these factors should be considered in the TRL dosing strategy. First, giving a fixed dose rather than dosing per kg bodyweight would be an improvement. The starting dose could be further individualized by taking the presently identified effects of genotype and hematocrit on the PK of TRL into consideration. In this way, patients with either genotypes in CYP3A5 or PXR should receive a 1.5 times higher dose compared to patients with the wild-type gene. Therewith, the optimal dose for most patients could be reached earlier after Tx in order to balance optimally within the therapeutic window.
References:
[1] Staatz, C.E. et al. Clinical Pharmacokinetics and pharmacodynamics of TRL in solid organ transplantation. Clin. Pharmacokinet. 2004: 43 (10): 623-53.
[2] Scholten, E.M. et al. AUC-guided dosing of tacrolimus prevents progressive systemic overexposure in renal transplant patients. Kidney Int. 2005; 67: 2440-47.
[3] Hesselink et al. Genetic polymorphisms of the CYP3A4, CYP3A5 and MDR-1 genes and pharmacokinetics of the calcineurin inhibitors cyclosporine and tacrolimus. Clin. Pharmacol. Ther. 2003; 74: 245-54.
[4] Lambda, J. et al. Genetic variants of PXR and CAR and their implication in drug metabolism and pharmacogenetics. Curr Drug Metab 2005; 6: 369-383.
[5] Ahn, J.E. et al. Inherent correlation between dose and clearance in therapeutic drug monitoring settings: possible misinterpretation in population pharmacokinetic analyses. J PKPD 2005; 32 (5-6): 703-18.
