Population pharmacokinetics of rifampicin and 25-deacetyl-rifampicin in healthy Asian adults

Abstract

Introduction

Rifampicin is a critical component of the first-line regimen and the most active drug for treating drug-susceptible TB. Despite its extensive use, rifampicin dosing information to ensure optimal drug exposure remains limited. There are indications that standard doses of rifampicin may be insufficient for optimal antituberculous activity and that higher doses are required.¹

Many studies have documented high interindividual rifampicin pharmacokinetic variability. Population differences in rifampicin concentrations have been attributed, in part, to formulation type,² age,³ sex⁴ and highly variable absorption.^5–7 In addition, concomitant medication has also been cited as a contributory factor to between-subject variability of rifampicin pharmacokinetics,^8–10 with coadministration of isoniazid suggested as a factor worthy of investigation to increase therapeutic success. This is because rifampicin has been reported to undergo rapid decomposition in the presence of isoniazid under acidic conditions.^8,11

Rifampicin is primarily metabolized to 25-desacetyl rifampicin (25-deacetyl-rifampicin), which is partly active, via deacetylation by esterases or unidentified enzymes present in microsomal cells.¹² Rifampicin is also a substrate of the drug efflux pump P-glycoprotein, which is encoded by the ABCB1 gene.¹³ Therefore, the high between-individual variability in rifampicin may be due, in part, to differences in genes encoding drug-metabolizing enzymes or drug transporters. Several studies have reported the correlation between rifampicin exposure with genetic variations in the SLCO1B1 gene, which encodes the organic anion-transporting polypeptide 1B1.^14,15 In particular, the maximum plasma concentration and oral bioavailability of rifampicin in mutant carriers of the SLCO1B1 gene have been reported to decrease by 36% and 20%, respectively, relative to WT carriers. However, our knowledge about the pharmacogenetic determinants of rifampicin exposure remains incomplete. A recent report demonstrated that decreased expression of cytochrome P450 (CYP) 3A4 contributed to increased oral absorption of rifampicin, indicating a possible role of CYP3A4 in pre-systemic biotransformation of rifampicin.⁵ To our knowledge, the effect of variant alleles of the CYP3A4 isozyme on rifampicin exposure has not been quantified.

Recently, a significant effect of sex on rifampicin population pharmacokinetics was reported,¹⁶ but 93% of the between-subject variability remained unexplained. The effect of several of the variant alleles of SLCO1B1 and ABCB1 on plasma rifampicin concentrations in African TB patients has been investigated,¹⁵ but no studies exploring the relationships of these alleles with rifampicin population pharmacokinetics in Asian subjects exist. In addition, it is our understanding that the combined effects of multiple pharmacogenetic and biological factors in predicting 25-deacetyl-rifampicin population pharmacokinetics are yet to be explored. It would be desirable to compare with and to identify sources of rifampicin pharmacokinetic variability, especially in different populations, beyond those described in the published literature so as to decrease variable treatment outcomes. The population approach is the method of choice to estimate typical pharmacokinetic parameter values (fixed effects) in a given population and the associated variability (random effects).¹⁷

In this population pharmacokinetic analysis, we sought to determine the effect of coadministration of isoniazid as well as polymorphisms of CYP3A4, ABCB1, SLCO1B1, and drug transporters' transcriptional regulators pregnane X receptor (PXR) and constitutive androstane receptor (CAR) on the steady-state pharmacokinetics of rifampicin and 25-deacetyl-rifampicin in healthy Asian adults.

Patients and methods

Study subjects

This was a prospective, open-label, crossover study in healthy adults, stratified by CYP2B6 genotype as part of an ongoing study examining single-dose efavirenz pharmacokinetics in the presence of rifampicin.¹⁸ Some 830 people from two healthy volunteer cohorts were screened to identify those carrying CYP2B6 516 GG and TT. Twenty-four GG and 10 TT subjects were recruited. Clinical examination combined with liver and renal function tests were conducted to verify that the participants were healthy. The subjects abstained from medications including herbal preparations 1 week before and throughout the study period. All participants gave informed written consent prior to undergoing genotyping and pharmacokinetic evaluation. The study was reviewed and approved by the Institutional Review Board of the National Healthcare Group, Singapore.

Rifampicin and isoniazid treatment and sampling

In the main study, subjects were given a single dose of efavirenz (600 mg) followed by a washout period of 14 days. Subjects were then randomized 1 : 1 to either rifampicin (600 mg) or rifampicin (600 mg)/isoniazid (300 mg) daily for 14 days, followed by another single dose of efavirenz 600 mg, which was administered at the same time as the final dose of rifampicin or rifampicin/isoniazid, given in the fasting state (‘Treatment 1’). After another washout period of 14 days, subjects were switched over to rifampicin (600 mg)/isoniazid (300 mg) or rifampicin (600 mg) daily for 14 days, followed by another efavirenz single dose, administered at the same time as the final dose of rifampicin/isoniazid or rifampicin, given in the fasting state (‘Treatment 2’). For this study, blood samples were drawn pre-dose and at 1, 2, 4, 6, 8, 10, 12, 18 and 24 h after the final dose of rifampicin intake per ‘treatment’ period. Subjects returned to the study site every 2 days for observed dosing and to collect their rifampicin or rifampicin/isoniazid medications for the next day. Adherence to rifampicin/isoniazid was further assessed by pill counts and medication diaries.

HPLC–MS/MS analysis

The concentrations of rifampicin and 25-deacetyl-rifampicin in plasma were measured by an HPLC–MS/MS method validated in our laboratory.^19,20 Briefly, 20 μL of plasma samples was diluted with 80 μL of 0.1% (v/v) formic acid followed by extraction in 300 μL of methanol containing internal standards (rifampicin-d₃ and 25-deacetyl-rifampicin-d₃) using Captiva™ ND^Lipids (Agilent Technologies) solid-phase extraction. Serial dilutions were used to obtain linear calibration ranges for rifampicin (25–50 000 ng/mL) and 25-deacetyl-rifampicin (2.5–5000 ng/mL). The HPLC–MS/MS system consisted of an Agilent 1290 binary pump connected to an Agilent 6460 triple quadrupole mass spectrometer (Agilent Technologies, Waldbronn, Germany). Chromatographic separations were achieved on a ZORBAX Aq-SB HPLC column (Agilent Technologies) with gradient elution. The mass spectrometer was operated under positive ionization mode and the detection of rifampicin and 25-deacetyl-rifampicin was based on the multiple-reaction monitoring of m/z 823.3→791.5 and 749.3→399.1, respectively. The method has been validated according to FDA guidance for accuracy (89.9%–115.0%) and precision (relative standard deviation <13.7%) for each of the high-, medium- and low-quality controls. Stability of the analytes has also been assessed (mean recovered concentration of 88.5%–100.4%) after 48 h storage at 6°C autosampler and three freeze–thaw cycles. The lower limits of quantification for rifampicin and 25-deacetyl-rifampicin were 25 and 2.5 ng/mL, respectively.

Genotyping

Genomic DNA was extracted from the blood samples by using the Omega E.Z.N.A. Blood DNA Mini Kit (Omega Bio-Tek, Norcross, GA, USA). Real-time PCR using fluorescent probes for allelic discrimination was used for genotyping. Primers and Taqman probes were sourced from Applied Biosystems (Foster City, CA, USA). Absolute quantitative PCR mix was obtained from Thermo Fisher Scientific (Waltham, MA, USA). The PCR conditions for most SNPs began from an initial denaturation step at 95°C for 15 min, followed by 40 cycles of denaturation at 95°C for 15 s and then annealing and extension at 60°C for 60 s, with a plate read after each cycle. For SLCO1B1 rs4149056 (c.521T>C), the denaturation temperature was 92°C. Each PCR mixture contained 2 μL of genomic DNA (20 ng/μL), 12.5 μL of absolute quantitative PCR mix (2×), 1.25 μL of primer mix (20×) and 1.25 μL of Taqman probe mix (20×), made up to 25 μL with deionized water. The assays were run on an Applied Biosystems StepOnePlus real-time PCR detection system (Foster City, CA, USA). The following SNPs were genotyped for, based on previous reports of allele frequency and functional significance:²¹ ABCB1 [rs3842 (T>C)], SLCO1B1 [rs2306283 (c.388 G>A), rs4149015 (c.11187 G>A) and rs4149032 (c.85-7793 C>T)], PXR [rs2472677 (C63396T) and rs1523130 (T44477C)], CYP3A4 [rs4646437 (G>A)], CYP2B6 [rs35303484 (A>G) and rs35979566 (T>A)] and CAR [rs2307424 (G>A)]. All genotyping assays were performed in duplicate.

Population pharmacokinetic analysis

NONMEM 7.2 software in conjunction with Perl-speaks-NONMEM (PsN) 3.5.3²² was used to estimate the population typical parameters, interindividual (IIV), interoccasion (IOV) and residual (RUV) random effects. All estimations were carried out using the first-order conditional estimation method with η–ϵ interaction. Additional calculation, data manipulation and graphics were performed using Xpose 4.3.5²³ embedded in R 2.15.0 (http://cran.r-project.org/, open-source, S-based statistical software).

A log transformation of both sides approach was used, where the logarithms of observed plasma concentrations were fitted by the pharmacokinetic model to predict the log-transformed concentrations. As all subjects had pharmacokinetic sampling on two separate ‘treatments’ on top of a pre-dose sample prior to each respective ‘treatment’, IOV was evaluated for the model parameters. Accordingly, a total of four occasions was defined for the population pharmacokinetic analysis. Similar to previous population modelling of rifampicin pharmacokinetic data,^15,30 IIV and IOV were described using exponential models, assuming lognormal distribution, on each parameter and expressed as a percentage [%IIV = (ω²)^0.5 × 100; %IOV = (κ²)^0.5 × 100]. Correlation between random effects was also evaluated. To describe the RUV and its corresponding standard error in rifampicin and 25-deacetyl-rifampicin plasma concentrations, additive error, exponential error and additive-plus-exponential error models were evaluated.

The candidate covariates screened during population pharmacokinetic model building are listed in Table 1. First, the demographic covariates as well as the effect of coadministration of isoniazid were evaluated with respect to their impact on the IIV in the model parameters. Next, covariates expressing genetic polymorphisms were added to the pharmacokinetic model one by one. For pharmacogenetic covariates with few homozygous or heterozygous mutant subjects, carriers of the least frequent genotype were pooled together with subjects in one of the adjacent genotype categories. In the absence of intravenously administered drug, the absolute bioavailability could not be determined. However, the relative rifampicin bioavailability (F) was estimated for mutant subjects, while F was set to one for WT genotypes. Genetic polymorphisms were assumed to affect clearances of rifampicin and 25-deacetyl-rifampicin and/or F.

The pharmacokinetic model was selected based on goodness-of-fit plots, precision of estimates and the likelihood ratio test within NONMEM. In the case of non-nested models, the value of the Akaike information criterion was used. The final model was evaluated by plots of the observed and predicted individual plasma concentrations of rifampicin and 25-deacetyl-rifampicin and by plots of the population-predicted plasma concentrations and conditional weighted residuals (CWRES),³¹ which were obtained using Xpose software.

Qualification of the final population model was evaluated using an unstratified non-parametric bootstrap analysis and a visual predictive check (VPC) in PsN. For the bootstrap analysis, 1000 bootstrap replicates were generated and parameters obtained with the bootstrap replicates were compared with the estimates obtained from the original dataset. In addition, the percentage of successful convergence for the bootstrap replicates was also monitored. For the VPC, the predicted concentration–time profiles for 1000 datasets at each timepoint after administration of a 600 mg rifampicin dose at steady-state conditions were generated from the parameters and variances of the selected final model. The 95% CI for the median and 5th and 95th percentiles derived from the simulation were computed and visually compared with the observed rifampicin and 25-deacetyl-rifampicin concentrations.

Results

The final dataset was composed of 518 rifampicin and 548 25-deacetyl-rifampicin concentration observations from 34 subjects. Less than 1% of the concentration–time data were below the limit of quantification and these were excluded in the pharmacokinetic data modelling.

In the base model, a one-compartment model with first-order elimination best described the rifampicin data (Figure 1). Rifampicin oral absorption described by two transit absorption compartments proved superior over the other oral absorption models (zero- and/or first-order absorption models or a lag-time model). A two-compartment model linked to the central compartment of rifampicin best described the 25-deacetyl-rifampicin data. The structural model was parameterized in terms of oral k_a, apparent clearance of rifampicin (CL/F), apparent volume of distribution of the central compartment of rifampicin (V₄/F), apparent clearance of 25-deacetyl-rifampicin (CL_m/F), apparent volume of distribution of the central compartment of 25-deacetyl-rifampicin (V₅/F), apparent volume of distribution of the peripheral compartment of 25-deacetyl-rifampicin (V₆/F) and apparent intercompartmental clearance of 25-deacetyl-rifampicin (Q_m/F). F_MET was fixed to one in the model. The model included IIV on CL/F, F, k_a and CL_m/F and covariance between CL/F and CL_m/F. In addition, IOV on CL/F, F, V₅/F and CL_m/F was also included. The residual unexplained variability of both analytes, as selected by goodness-of-fit plots, CWRES versus time and decrease in OFV, was best described by an additive model on natural log-transformed data.

Figure 1.

Rifampicin/25-deacetyl-rifampicin compartment model. k_tr, transit rate constant of rifampicin; k_a, oral absorption rate of rifampicin; F, relative bioavailability of rifampicin; V₄/F, apparent volume of distribution of rifampicin; CL/F, apparent clearance of rifampicin; F_MET, fraction of rifampicin converted into 25-deacetyl-rifampicin; V₅/F, apparent volume of distribution of 25-deacetyl-rifampicin in the central compartment; CL_m/F, apparent clearance of 25-deacetyl-rifampicin; V₆/F, apparent volume of distribution of 25-deacetyl-rifampicin in the peripheral compartment; Q_m/F, apparent intercompartmental clearance of 25-deacetyl-rifampicin; RIF, rifampicin; dRIF, 25-deacetyl-rifampicin.

Open in new tabDownload slide

The η plots suggested potentially lower F in SLCO1B1 rs2306283 non-expressers (homozygous mutant type) compared with expressers (homozygous WT and heterozygous mutant type) and lower F in males compared with females. The shrinkage for IIV on CL/F, F, k_a and CL_m/F was 15%, 17%, 9% and 30%, respectively. The shrinkage for IOV on CL/F, F, V₅/F and CL_m/F was 30%, 16%, 18% and 26%, respectively. The η shrinkage estimates for CL/F, F and k_a were considered adequate and enabled the use of empirical Bayes estimates as a diagnostic tool to evaluate the predictive performance of the model.³²

The potential effects of demographic covariates, coadministration of isoniazid and genetic covariates were systemically evaluated in a univariate analysis. However, none of the investigated covariates altered the OFV significantly and, hence, they were not considered as significant covariates. As such, the pharmacokinetic parameters for the final model were similar to those for the base model (Table 2). Typical population parameters were CL/F 10.3 L/h (IIV 9.08% and IOV 17.0%), V₄/F 30.9 L, CL_m/F 95.8 L/h (IIV 39% and IOV 26.9%) and V₅/F 41.9 L (IOV 103%). The mean half-life of rifampicin was 2.08 h, which was physiological in nature and close in value to that reported by Shishoo et al.⁸ (2.37 h).

The basic goodness-of-fit plots (Figure 2) and VPC (Figure 3) for the final model show that model predictions were in reasonable agreement with the observed plasma concentrations of rifampicin and 25-deacetyl-rifampicin. In Figure 3, the observed median (continuous line) and 5th and 95th data percentiles (broken lines) were adequately captured by the corresponding simulation-based 95% CIs (dark and light grey areas) around the median and 5th and 95th prediction intervals. The final model was also internally validated by means of 1000 bootstrap runs (Table 2), which were successful in 93% of the runs and confirmed the parameter values.

Figure 2.

Diagnostic plots of the final rifampicin (RIF; left column) and 25-deacetyl-rifampicin (dRIF; right column) population pharmacokinetic model. First row: observations versus population predictions. Second row: observations versus individual predictions. Third row: CWRES versus population predictions. Fourth row: CWRES versus time after rifampicin dose.

Open in new tabDownload slide

Figure 3.

VPC for the final rifampicin (top panel) and 25-deacetyl-rifampicin (bottom panel) population pharmacokinetic model. In each subplot, the continuous line connects the observed median values, whereas the broken lines represent the observed 5th and 95th percentiles of the observations. The light grey areas indicate the 95% CIs of the 5th and 95th percentiles of the predicted values, whereas the dark grey area indicates the CI of the median.

Open in new tabDownload slide

Discussion

Our study enabled us to investigate the effects of several polymorphisms of drug-metabolizing enzymes, drug transporters and transcriptional regulators, sex and demographic variables, and coadministration of isoniazid on the population pharmacokinetics of rifampicin and its primary metabolite, 25-deacetyl-rifampicin in a population that is under-represented in the literature. A one-compartment model coupled with a transit absorption model adequately fitted the rifampicin data, while 25-deacetyl-rifampicin pharmacokinetic data were best described by a two-compartment model linked to the rifampicin model. We identified a positive correlation between IIVs of rifampicin and 25-deacetyl-rifampicin clearances (r = 0.52), which could be explained by an influence of body size on both of these pharmacokinetic parameters across our studied Asian subjects. Although η versus covariate plots suggested possible relationships between F and SLCO1B1 rs2306283 and sex, none of the covariates studied was found to significantly affect the disposition of rifampicin and 25-deacetyl-rifampicin.

The oral absorption of rifampicin was best described using a transit compartment model with two transit compartments. Our attempts to fit a transit compartment absorption model in which the number of transit compartments was estimated led to overparameterization of the model and resulted in a high relative standard error percentage in the model parameters.³³ This could be explained by the insufficient blood sampling prior to the attainment of the maximum concentration post-dose in our subjects. Our implementation of the model with two transit compartments resulted in a structural model with lower OFV as compared with a model with the absorption lag-time parameter, which was reported in previous rifampicin population pharmacokinetic models.¹⁶ A transit compartment model might also be preferred over a lag-time model because numerical instability could arise from the discontinuous feature of the lag-time absorption model.³³ In addition, inclusion of the transit model in our linked parent–metabolite model, while providing a better fit, did not affect the estimation of rifampicin and 25-deacetyl-rifampicin disposition parameters.

The typical parameter estimates for the apparent clearance and apparent volume of distribution of rifampicin are in line with values reported in ethnically different populations. The MTT [which is equal to (number of transit compartments + 1)/k_tr] in our study was 1.39 h, which is similar to those of previous studies that implemented a flexible transit absorption model [0.42 and 1.12 (female)–1.60 (male), respectively].^15,30 The apparent clearance of rifampicin in our study ranged from 7.49 to 12.0 L/h when body weight increased from 45.8 to 86.1 kg (range of body weight in our subjects) and these values are similar to the previously determined CL/F in South African, Mexican and Korean TB patients [11.0–19.2, 8.17 (female)–11.4 (male) and 6.1, respectively].^15,16,30,34 The apparent volume of distribution of rifampicin in our subjects ranged from 20.2 to 38.0 L, which overlaps with those reported in previous population pharmacokinetic investigations (35.0–76.3 L).^15,16,30,34

The findings from our study are in agreement with a previous study,¹⁵ which showed the lack of a significant effect of the ABCB1, CAR or PXR polymorphism on rifampicin pharmacokinetics. Despite the moderate to high frequency of SLCO1B1 polymorphisms in our subjects and reported significant effects of defective SLCO1B1 rs4149032 alleles on enzyme activity in African TB patients,¹⁵ we found no significant effect of SLCO1B1 variant alleles on rifampicin population pharmacokinetics. The contradictions in findings could possibly be explained by the higher concentrations of the drug influx transporter activity and/or expression found in patients coadministered with other anti-TB drugs.¹⁵ Considering the substantial incidences of SLCO1B1 rs4149032 and SLCO1B1 rs2306283 homozygous variant (24% and 15%) or heterozygous (56% and 53%) genotypes present in our subjects, a lack of effect of the SLCO1B1 genetic polymorphisms in the current analysis may need a confirmatory study using a larger dataset.

To our knowledge, the influence of CYP3A4 polymorphism on rifampicin or 25-deacetyl-rifampicin plasma levels has not yet been reported, despite widespread literature on the role played by intestinal and hepatic CYP3A4 enzymes in the bioavailability and metabolism of many drugs.³⁵ The CYP3A4 rs4646437 polymorphism was not identified as a significant covariate for rifampicin CL/F or F in the current study. Since the number of subjects in our population was relatively small, it might have reduced the chance of finding the CYP3A4 covariate relationship. Furthermore, given that this analysis was conducted in healthy volunteers, additional analyses to investigate effects of CYP3A4 polymorphisms on rifampicin pharmacokinetics and safety in TB patients are warranted.

In the literature, there are conflicting data on the role played by sex on rifampicin population pharmacokinetics. In this study, we could not identify any influence of sex on the apparent clearance and apparent volume of distribution of rifampicin, which corroborated the findings from Wilkins et al.³⁰ and Chang et al.³⁴ This is in contrast to findings from Chigutsa et al.¹⁵ and Milan Segovia et al.,¹⁶ which, respectively, reported higher apparent volume of distribution and MTT of rifampicin in males and higher apparent clearance and apparent volume of distribution of rifampicin in males. We did find a trend of lower relative bioavailability of rifampicin in males relative to females in this study. However, this trend was not statistically strong enough for inclusion in the final model. Although the difference in the influence of sex on rifampicin population pharmacokinetics between the current analysis and the literature are yet to be elucidated, taken together, it is speculated that increases in apparent clearance and apparent volume of distribution of rifampicin may be attributed more to reduced oral bioavailability than increases in clearance and volume of distribution.

Potential limitations to this work include prior stratification by CYP2B6 genotype and coadministration of single-dose efavirenz. However, CYP2B6 is not considered to be important in the elimination of rifampicin or isoniazid and we observed no difference in rifampicin pharmacokinetics between different CYP2B6 genotypes. A previous study failed to demonstrate any significant impact of efavirenz upon rifampicin pharmacokinetics.³⁶ Further, our data were derived based on one rifampicin dose level (600 mg). It has been reported that rifampicin exhibits non-linear kinetics (i.e. increased bioavailability and systemic exposure) with increased dose levels, most likely due to saturation of hepatic first-pass metabolism.^37,38 Hence, the derived population pharmacokinetic model may not be applicable to data sampled at rifampicin dose levels >600 mg.³⁹

In conclusion, our analysis developed a structural pharmacokinetic model that adequately described the disposition of rifampicin and 25-deacetyl-rifampicin in an Asian population. Our analysis showed that genetic polymorphisms observed for drug-metabolizing enzymes, transporters and transcriptional regulators were not significant predictors of the between-subject variability in rifampicin and 25-deacetyl-rifampicin pharmacokinetics. Although results from our covariate analysis are similar to those in other study populations, additional studies on patient populations are warranted to evaluate these factors.

Funding

This work was supported by: the Biomedical Research Council (BMRC 10/1/21/24/632) through its Joint Grant with the Medical Research Council, UK; the National Medical Research Council (NMRC/CSA/019/2010) through its Clinician Scientist Award to L. S.-U L.; and the Singapore Anti-Tuberculosis Association CommHealth grant.

Transparency declarations

None to declare.

Acknowledgements

We wish to thank: all volunteers, Samuel Hong and the Clinical Trials Research Unit, Changi General Hospital; and Serene Ng and the Investigational Medicine Unit, National University Health System for the conduct of this study.

References