The impact of CYP3A5*3 polymorphism on sirolimus pharmacokinetics: insights from predictions with a physiologically‐based pharmacokinetic model

Abstract

Aims

Sirolimus is an mTOR inhibitor metabolized by CYP3A4 and CYP3A5. Reported effects of CYP3A5 polymorphisms on sirolimus pharmacokinetics (PK) have shown unexplained discrepancies across studies. We quantitatively assessed the effect of CYP3A5*3 status on sirolimus PK by in vitro assessment and simulation using a physiologically‐based PK (PBPK) model. In addition, we explored designs for an adequately powered pharmacogenetic association study.

Method

In vitro metabolism studies were conducted to confirm individual CYP contribution to sirolimus metabolism. PK profiles were simulated in CYP3A5 expressers and non‐expressers with a PBPK model. The pre‐dose concentration predictions were used as the outcome parameter to estimate the required sample size for a pharmacogenetic association study.

Results

Sirolimus metabolism was inhibited by over 90% by ketoconazole, a CYP3A specific inhibitor. The PBPK model developed based on CL_int of recombinant CYP3A4, CYP3A5 and CYP2C8 predicted a small CYP3A5*3 effect on simulated sirolimus PK profiles. A subsequent power analysis based on these findings indicated that at least 80 subjects in an enrichment design, 40 CYP3A5 expressers and 40 non‐expressers, would be required to detect a significant difference in the predicted trough concentrations at 1 month of therapy (P < 0.05, 80% power).

Conclusions

This study suggests that CYP3A5 contribution to sirolimus metabolism is much smaller than that of CYP3A4. Observed discrepancies across studies could be explained as the result of inadequate sample size. PBPK model simulations allowed mechanism‐based evaluation of the effects of CYP3A5 genotype on sirolimus PK and provided preliminary data for the design of a future prospective study.

What is Already Known About This Subject

• CYP3A5 in addition to CYP3A4 contributes to sirolimus metabolism.

• The CYP3A5*3 non‐expresser polymorphism has been considered an important pharmacogenetic factor contributing to inter‐individual differences in sirolimus pharmacokinetics (PK).

• Clinical studies have reported contradicting results related to the effect of CYP3A5 genotype on sirolimus PK.

What This Study Adds

• A physiologically‐based pharmacokinetic (PBPK) model combined with in vitro metabolism data was used to simulate the effect of CYP3A5 genotype on sirolimus PK profiles and predicted a small difference in simulated sirolimus PK profiles between CYP3A5 expressers and non‐expressers.

• This finding was consistent with observations from in vitro metabolism studies using individual liver microsomes genotyped for CYP3A5*3.

• A power analysis using pre‐dose concentrations simulated with the PBPK model demonstrated the required sample size for a prospective pharmacogenetic association study.

Introduction

Sirolimus is a mammalian target of rapamycin (mTOR) inhibitor which has been used in solid organ transplantation and as a promising agent in cancer therapies 1, 2. Sirolimus has low oral bioavailability and large inter‐individual variability in its pharmacokinetics 3. Previous in vitro studies using recombinant P450 enzymes demonstrated that sirolimus is extensively metabolized through the CYP3A subfamilies, CYP3A4 and CYP3A5, and to a small extent through CYP2C8 4. CYP3A5*3 is a common polymorphism found in the gene encoding for CYP3A5 5 and this polymorphism has been of great interest as a potential factor to explain the observed inter‐individual differences in the pharmacokinetics (PK) and bioavailability. This single nucleotide polymorphism causes alternative splicing resulting in the absence of functional CYP3A5 5. The isoform is represented by ‘expressers’, those who carry at least one functional CYP3A5*1 allele (wild type), and by ‘non‐expressers’ who are homozygous for the CYP3A5*3 allele encoding for non‐functional CYP3A5. The CYP3A5*3 polymorphism is predictive of inter‐individual differences in the pharmacokinetics and bioavailability of CYP3A5 substrates like tacrolimus 6, 7, 8. Several studies have reported sirolimus exposure to be significantly higher in CYP3A5 non‐expressers. However, due to inconsistencies in reported results there is no consensus on the influence of the CYP3A5 genotype on sirolimus PK (Table 1) 9, 10, 11, 12, 13. Most studies did not observe a significant effect probably due to a combination of large PK variability and a small sample size.

Table 1.

Literature data related to effect of CYP3A5*3 on sirolimus pharmacokinetics

Reported study information						Reported outcome	Results of retrospective power analysis	Reference (Year)
Study population (Number of subjects)	Descriptive PK parameters compared	Concurrent CNI*	Total subject	Number of expressers (% of total)	Number of non‐expressers (% of total)	Difference in dose‐normalized C0 between expressers and non‐expressers	Probability of finding a true significant difference
Caucasian (47)	Dose normalized AUC, C 0, C max	None	47	6 (13%)	41 (87%)	Yes	29%	9 (2006)
		None (de novo)	21 (Sub‐group)†	3 (14%)	18 (86%)	Yes	14%
Chinese (47)	C 0 (ng ml–1)/Dose (mg kg–1)	None	47	21 (45%)	26 (55%)	Yes	57%	10 (2008)
Caucasian (140) African (4)	(C 0, ng ml–1)/(Dose, mg kg–1)	None (de novo)	51	13 (25%)	38 (75%)	No	49%	11 (2005)
Caribbean (5)		None (rescue)	69	11 (16%)	58 (84%)	Yes	48%
		CsA/Tac	29	7 (24%)	22 (76%)	No	29%
Caucasian (82) African (2)	(C 0, ng ml–1)/(Dose, mg kg–1)	Mix (n = 24, Tac)	85	7 (8%)	78 (92%)	No	34%	12 (2005)
South Asian (1)
		Tac	24 (Sub‐group)‡	3 (13%)	21 (88%)	No	13%
Caucasian (20)	(Dose, mg)/(C 0, ng ml–1)	None	20	4 (20%)	16 (80%)	No	17%	13 (2007)
	Dose normalized AUC, t 1/2, CLoral	None	10 (Sub‐group)	3 (30%)	7 (70%)	No	9%

^*CNI, calcineurin inhibitor such as ciclosporin A (cyclosporine A, CsA) and tacrolimus (Tac).

^†21 patients were de novo recipient out of 47 patients.

^‡24 patients were treated with a combined sirolimus and tacrolimus among the whole population of 85.

The possibility was calculated using trough concentration of sirolimus predicted by PBPK model with n = 1000 virtual healthy adult (each 500 subjects for the expresser and non‐expresser).

We previously developed a physiologically‐based PK (PBPK) model of sirolimus in adults based on in vitro metabolic activity data generated with recombinant CYP3A4, CYP3A5 and CYP2C8 enzymes 14. The purpose of this study was to assess quantitatively the effect of the CYP3A5*3 polymorphism on sirolimus PK by means of a simulation study using the PBPK model and in vitro enzymatic study results using human microsomes.

Methods

Materials

Sirolimus was purchased from LC Laboratories (Woburn, MA, USA). Ketoconazole and montelukast were obtained from Sigma‐Aldrich (St Louis, MO, USA). Zotarolimus was purchased from Molcan Corporation (Toronto, Canada). Potassium phosphate buffer (500 mm, pH 7.4) and the NADPH regenerating system were obtained from BD Bioscience (San Jose, CA, USA). Other reagents used in this study were commercially available and of analytical grade. Pooled human liver and intestinal microsomes were purchased from BD Bioscience. Individual human liver microsomes pre‐genotyped for CYP3A5*1 and CYP3A5*3 were purchased from Xenotech (Lenexa, KS, USA).

Enzyme assay

A substrate depletion assay was performed according to the method of Emoto et al. 14. Sirolimus was incubated with human microsomes (pooled or CYP3A5 pre‐genotyped single subject microsomes) in 100 mm potassium phosphate buffer (pH 7.4) in the absence and presence of P450 specific inhibitors at the concentration of 1 μm, ketoconazole for CYP3A inhibition 15 and montelukast for CYP2C8 inhibition 16, in a shaking water bath at 37°C. Final concentration of organic solvent in the reaction mixture was less than 1 (v/v) %. Sample collection was at serial time points, where the 0 min time point indicated the time when the NADPH regenerating system was added into the reaction mixture. Samples were extracted with methanol/0.2 m ZnSO₄ (80/20, v/v) including zotarolimus as an internal standard. After the removal of protein by centrifugation at 21 200 g for 10 min at 4°C, the supernatant was subjected to high performance liquid chromatographic separation with tandem mass spectrometric detection (LC‐MS/MS). Sirolimus was quantified by an integrated on‐line solid phase extraction‐high performance liquid chromatography (1200 series, Agilent Technologies, Santa Clara, CA, USA) ‐tandem mass spectrometry (API‐3000, AB SCIEX, Foster City, CA, USA) system according to the method previously reported. 14 The assay was linear over the concentration range of 10 nm (9.1 ng ml^‐1) to 3 μm (2.7 μg ml^‐1). The intra‐ and inter‐day coefficients of validation were 0.30–11% and 1.6–7.5%, respectively. The lower limit of quantification was 10 nm (9.1 ng ml^‐1). Data of in vitro intrinsic clearance (CL_int) estimates represent the average of triplicate determinations.

Estimation of intrinsic clearance and % CYP contribution

The peak area ratio of sirolimus to the internal standard, zotarolimus, was used for quantification. The ratio at each time point was expressed as a percentage of that at the 0 min starting point (defined as % remaining substrate in this study). The % remaining substrate values were plotted vs. incubation times on a semi logarithmic scale, and the slope was determined by linear regression analysis as the elimination rate constant (−k _el, min⁻¹). The apparent CL_int value was calculated by the following equation 17:

$${\mathrm{CL}}_{\mathrm{int}}=\frac{\left(k_{\mathrm{el}}\right)}{\mathit{CYP}\text{concentration}\mathrm{in}\text{reaction}\text{mixture}\left({\text{pmol}\mathrm{CYP}/\mathrm{ml}}^{‐1}\right)}$$

For the inhibition assay, percentage of contribution of each CYP isoform to the in vitro metabolism of sirolimus was calculated by the following equation 18:

$$\%\text{contribution}=\frac{\left(k_{\mathrm{el}}\right)-\left(k_{\mathrm{el},\text{inhibitor}}\right)}{k_{\mathrm{el}}}\times 100$$

where k _el and k _el,inhibitor were determined in the absence and presence of chemical inhibitor, respectively.

Linear regression analyses of in vitro sirolimus intrinsic clearances and individual metabolic activity for each CYP marker reaction were performed using GraphPad Prism (version 6.03, La Jolla, CA, USA).

PBPK model simulations

A PBPK model of sirolimus was previously developed using Simcyp software ver.12. 14 Two virtual genotype populations, one for CYP3A5 expressers (CYP3A5*1/*1 or *1/*3) and one for CYP3A5 non‐expressers (CYP3A5*3) were manually created from the built‐in healthy volunteer module. For CYP3A5 expressers the default abundance settings in Simcyp version 12 were used: hepatic CYP3A5 abundances in relation to CYP3A4 abundance defined as 137 pmol mg^–1 of hepatic microsomal protein (CYP3A5 = 0.39*CYP3A4 + 62.8 pmol mg^–1, with CVs of 41% for CYP3A4 abundance and 24% for the residual variability) 19 and intestinal CYP3A5 abundance as 24.6 nmol/total gut, with CV of 60%. In addition, intestinal CYP3A4 abundance was set at 66.2 nmol/total gut as a default setting (CV = 60%) 20, 21. For CYP3A5 non‐expressers, CYP3A5 abundance was modified to 0 pmol mg^–1 for liver and 0 nmol/gut for intestine. Sirolimus PK profiles were simulated for 500 virtual subjects and for each of the CYP3A5 expresser and non‐expresser populations. The maximum concentration (C _max), the area under the curve over 24 h after the last administration (AUC(0,24 h)), and the trough concentration after the last administration (C ₀) were evaluated for sirolimus when administrated orally at a 10 mg dose daily for 1 day, 1 week, 2 weeks, 1 month and 2 months.

Adequacy of sample size estimation

Trough concentrations of 500 virtual subjects for each of the CYP3A5 expresser and non‐expresser populations were first obtained from simulations with the PBPK model as mentioned above. These data were then used for sample size analysis as the source data set. Sample size analysis was conducted using R (version 3.1) by random sampling from the data set for two different conceivable scenarios of a clinical pharmacogenetics association study evaluating the effect of CYP3A5 genotype on trough concentrations (C ₀) as the endpoint. The random sampling from the concentration data set was repeated 30 000 times. The power to detect a difference was determined as the proportion of times a significant difference in means was observed. The Wilcoxon rank sum test was used at the significance level of P < 0.05 to determine differences in mean trough concentrations between CYP3A5 expressers and non‐expressers. Next, the probability of achieving significance was estimated for increasing numbers of subjects. The first scenario was based on random prospective enrolment of subjects. In this random enrolment scenario, the frequency of CYP3A5 expresser status in the population was set at 10%, 20% and 33% to reflect overall frequencies observed in the published studies listed in Table 1. The sample size analysis was performed using a binomial distribution of expressers (CYP3A5*1/*1 and CYP3A5*1/*3) or non‐expressers (and *3/*3) in the population. In the alternative scenario, an equal number of expressers and non‐expressers was assumed for enrolment to be achieved by genotyping prior to enrolment (enrichment).

In addition, a retrospective power analysis was conducted with the same method and by using the same simulated source data set to estimate probability of finding a significant difference in trough concentrations between CYP3A5 expressers and non‐expressers. In this analysis, the reported number of subjects enrolled as CYP3A5 expressers and non‐expressers in each study was used to estimate the post hoc probability of finding a true difference between genotype groups in each of the studies (Table 1).

Results

Assessment of CYP3A dependent in vitro sirolimus metabolism

The chemical inhibition study with pooled human liver and intestine microsomes using P450 specific inhibitors demonstrated the predominant role of CYP3A in sirolimus metabolism (Figure 1). Ketoconazole, a selective CYP3A inhibitor, inhibited sirolimus metabolism for more than 90% in both pooled liver and intestinal microsomes while montelukast, a selective CYP2C8 inhibitor, inhibited the liver and intestinal pathways by 5.3% and 6.8%, respectively. These results were consistent with those in individual liver microsomes (n = 10). Correlation analysis with individual liver microsomes revealed that sirolimus CL_int significantly correlated with the CYP3A marker reactions (testosterone 6β‐hydroxylation and midazolam 1′‐hydroxylation). However, the correlation with amodiaquine N‐dealkylation as a CYP2C8 marker reaction was not as good (Figure 2A–C). The correlation with testosterone 6β‐hydroxylation (r = 0.95) was higher than that with midazolam 1′‐hydroxylation (r = 0.74). Regression analysis by CYP3A5 genotypes resulted in improved correlation coefficients with midazolam 1′‐hydroxylation in both CYP3A5 expressers and non‐expressers (Figure 2D). In addition, amodiaquine N‐dealkylation did not correlate with the CYP3A marker reactions in this study. The wide range of individual sirolimus CL_int estimates appeared not to be different between CYP3A5 expressers and non‐expressers. No significant differences in sirolimus CL_int was detected between CYP3A5 expressers and non‐expressers in these numbers of samples. In vitro CL_int values were 4.1 ± 2.9 and 4.3 ± 2.8 μl min^–1 pmol^–1 P450 (mean ± SD, n = 5 for each), respectively.

Figure 1.

The inhibitory effect of specific P450 inhibitors on sirolimus metabolism in pooled human (A) liver and (B) intestinal microsomes. Montelucast and ketoconazole were used as CYP2C8 and CYP3A inhibitors at a concentration of 1 μm. Data are presented as mean ± SD of triplicate determinations. Closed circle, vehicle; open circle, montelukast; closed triangle, ketoconazole

Figure 2.

Correlation between in vitro intrinsic clearance (CL_int) of sirolimus and P450 marker reactions. The genotype data and the metabolic activity of each P450 marker reaction, (A) testosterone 6β‐hydroxylation and (B, D) midazolam 1′‐hydroxylation for CYP3A and (C) amodiaquine N‐dealkylation for CYP2C8, refer to the data sheet provided from the vendor. Linear regression analysis was conducted with data of total individual microsomes (A, B, and C) and separately in each CYP3A5 genotype (D, blue symbol and line for CYP3A5 expresser, black symbol and line for CYP3A5 non‐expresser) by Prism software. Data are presented as mean ± SD of triplicate determinations

Adequacy of sample size estimation

Sirolimus PK profiles were simulated with the PBPK model from a population with a realistic distribution of virtual CYP3A5 expressers and non‐expressers (500 subjects for each group, Figure 3). Mean simulation results showed a difference in C _max, AUC(0,24 h), and C ₀ between the two groups. The difference in PK parameter estimates in expressers and non‐expressers where expressed as a ratio (non‐expresser/expresser) increased over time and reached steady‐state after approximately 1 month of therapy (Figure 4). The non‐expresser : expresser ratios for C _max, AUC(0,24h) and C ₀ at steady‐state were 1.2, 1.3, and 1.4, respectively.

Figure 3.

Simulation of sirolimus PK profile by PBPK model with virtual CYP3A5 expresser and non‐expresser. Sirolimus PK profile was simulated by a PBPK model with the virtual CYP3A5 expresser (blue line) and non‐expresser (black line) subjects, which were created based on the healthy volunteer described in Methods. Lines are represented as mean of the simulation (n = 500) for each group

Figure 4.

The effect of CYP3A5 expression on sirolimus PK parameters. Sirolimus PK parameters, A) maximum concentration (C _max), B) AUC(0,24 h) (AUC for 24 h after a last administration) and C) trough concentration (C ₀, concentration at 24 h after a last administration), were estimated from PK simulations by PBPK model with virtual CYP3A5 expresser and non‐expresser subjects. D) Ratio (non‐expresser : expresser). The PK simulation was conducted for 1 day (D1), 1 week (1W), 1 month (1M), 2 months (2M) with multiple dosing of sirolimus (daily 10 mg administration). Each column is represented as mean ± SD

A power analysis performed using the simulated trough concentrations (C ₀) demonstrated that the optimal sample size was estimated to be 80 subjects (40 subjects for each group) to achieve 80% power for a study using enrichment based on pre‐enrolment genotyping (Figure 5A). On the other hand, for a study pursuing random enrolment and a design assuming CYP3A5 expresser allele frequencies of 10%, 20% and 33%, the adequate sample sizes to achieve 80% power were estimated to be 220, 125, and 90 subjects, respectively (Figure 5B).

Figure 5.

Power analysis with estimated trough concentrations of sirolimus by PBPK model with virtual CYP3A5 expresser and non‐expresser. The trough concentration (C ₀, concentration at 24 h after a last administration) was estimated from PK simulations by PBPK model with virtual CYP3A5 expresser and non‐expresser subjects. Power analysis was conducted in two scenarios, pre‐genotyped trial (enrichment, A) and randomly enrolled trial (B). As for the randomly enrolled trial, 33% (solid line), 20% (dashed line) and 10% (dotted line) were chosen as frequencies of CYP3A5 expressers. These values were intended to represent overall frequencies observed in clinical studies listed in Table 1 (ranged 8–45%). The frequency number of 33% is based on first reported number by Kuehl et al. 5

Retrospective analysis to assess probability of finding significance in reported studies

The probability of finding a significant difference in C ₀ between CYP3A5 expressers and non‐expressers in reported studies is summarized in Table 1. The calculated probabilities ranged from 9% to 57%, with all having less than 80% power to detect a difference. Two out of three studies which reported significant differences between CYP3A5 expressers and non‐expressers had a probability of detecting a true difference of 48 and 57%. The studies that did not find a significant difference had a calculated probability of finding a true difference of less than 49%.

Discussion

The aim in this study was to evaluate quantitatively the effect of the CYP3A5*3 polymorphism on sirolimus PK. The simulations were based on the differences in only CYP3A5 activity in virtual healthy subjects with no other contributors to inter‐individual variability such as disease progression, co‐medication and food effects which are common in clinical studies. Thus, the in silico simulation study was performed under the best possible conditions to detect CYP3A5 genotype mediated differences in sirolimus PK. Yet, the simulated sirolimus PK profiles showed relatively small differences between PK parameter estimates in CYP3A5 expressers and non‐expressers. This observation is corroborated by the preliminary in vitro results which seem to indicate comparable individual sirolimus intrinsic clearance estimates (CL_int) in liver microsomes genotyped as expresser and non‐expresser, despite large variability. Consistent with the current observation, Picard et al. 22 also did not show any significant differences in the sirolimus CL_int when using liver microsomes genotyped for CYP3A5*1/*3 and 3A5*3/*3. Showing this relatively small pharmacogenetic effect would require a large sample size and is therefore the most likely reason that more than half of clinical studies reported to date were not able to demonstrate a significant association as summarized in Table 1. In the simulation study, a significant effect was shown with 100 subjects receiving a 1 month long course of therapy. When PK parameters were expressed as the ratio of CYP3A5 non‐expressers to expressers values, ratios increased over time until steady‐state was researched at around 1 month after starting therapy.

The contribution of CYP3A5 to the sirolimus hepatic and intestinal metabolism in the expressers was predicted to be 24% and 14%, respectively, using P450 abundance data implemented in Simcyp as default setting, according to the method described previously. 14 In addition, PBPK model predicted F_g (the fraction of the drug entering the enterocytes that escapes first pass gut wall metabolism) values at steady‐state in CYP3A5 expressers and non‐expressers were predicted to be 0.23 and 0.26, respectively. F_h (the fraction of drug entering (or by‐passing) the liver that escapes first pass hepatic metabolism and biliary secretion) values were 0.88 and 0.90, respectively. These values were calculated as means from the simulation, with n = 500 virtual subjects for each group. Thus, the differences in the predicted F_g and F_h values between CYP3A5 expressers and non‐expressers were small. This indicates that the contribution of CYP3A5 on intestinal and hepatic availabilities of sirolimus should be small compared with that of CYP3A4.

In the regression analysis with individual liver microsomes, CL_int of sirolimus correlated better with testosterone 6β‐hydroxylation than with midazolam 1′‐hydroxylation. This difference may be due to the difference in contribution of each isoform as testosterone 6β‐hydroxylation is predominantly catalyzed by CYP3A4 rather than CYP3A5 (39‐fold difference) while midazolam l’‐hydroxylation is catalyzed equally by CYP3A4 and CYP3A5 23. This indirectly provides further evidence of a greater impact of CYP3A4 on sirolimus PK than of CYP3A5. Furthermore, we previously reported that the sirolimus CL_int of CYP3A4 was 2.4‐fold higher than that of CYP3A5 14. This is quite different from tacrolimus for which the in vitro CL_int of CYP3A5 was reported to be 2.7‐fold higher than that of CYP3A4. 24 This much higher CYP3A5 contribution may well explain the large number of positive studies showing significant association of CYP3A5*3 genotype with tacrolimus PK as summarized in the recent Clinical Pharmacogenetics Implementation Consortium (CPIC) guideline. 8.

In published clinical studies, sirolimus is often co‐administrated with ciclosporin (cyclosporine) (Table 1). Since ciclosporin predominantly inhibits CYP3A4 compared with CYP3A5 25, when co‐administrated with ciclosporin, the role of CYP3A5 in sirolimus metabolism is considered to be more pronounced. However, in a sirolimus phase III trial when patients also received ciclosporin, there was no significant difference in sirolimus mean trough concentrations between Black (n = 190) and non‐Black patients (n = 852) (Rapamune prescribing information; Pfizer/Wyeth‐Ayerst, http://labeling.pfizer.com/showlabeling.aspx?id=139) despite higher frequency of expressers in African‐Americans (60%) as compared with Caucasians (33%) 5. These observations also support that CYP3A5 contribution to sirolimus metabolism is much smaller than that of CYP3A4.

In addition to the CYP3A subfamily, we also examined the contribution of CYP2C8 by means of the chemical inhibition assay using human liver and intestinal microsomes, but no significant inhibition was observed. This is consistent with our previous study using recombinant enzymes showing that the CL_int by CYP2C8 (0.25 μl min^–1 pmol^–1 CYP) was much smaller than CYP3As (CYP3A4 9.33 μl min^–1 pmol^–1 CYP, CYP3A5 3.96 μl min^–1 pmol^–1 CYP). 14 These observations suggest that CYP2C8 has a minor contribution to sirolimus metabolism in the liver and intestine.

In conclusion, this study documents that CYP3A5 contribution to sirolimus metabolism is smaller than what has been observed for tacrolimus, resulting in relatively small CYP3A5 mediated genetic differences. PBPK model simulations allowed us to explain effects of CYP3A5 genotype on sirolimus clearance in a mechanistic fashion. An evaluation of sample size requirements using PBPK model generated data will facilitate the design of future prospective studies.

Competing Interests

All authors have completed the Unified Competing Interest form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare no financial relationships or activities with any organizations that could appear to have influenced the submitted work in the previous 3 years: C.E. had a salary from Otsuka Pharmaceutical Co. Ltd as an employee until August 2014, outside the submitted work.

Contributors

Wrote manuscript, Chie Emoto, Tsuyoshi Fukuda, Alexander A. Vinks. Designed research, Chie Emoto, Tsuyoshi Fukuda, Alexander A. Vinks. Performed research, Chie Emoto, Tsuyoshi Fukuda, Raja Venkatasubramanian. Analyzed data, Chie Emoto, Tsuyoshi Fukuda, Raja Venkatasubramanian. Contributed new reagents/analytical tools, Chie Emoto.

Emoto, C. , Fukuda, T. , Venkatasubramanian, R. , and Vinks, A. A. (2015) The impact of CYP3A5*3 polymorphism on sirolimus pharmacokinetics: insights from predictions with a physiologically‐based pharmacokinetic model. Br J Clin Pharmacol, 80: 1438–1446. doi: 10.1111/bcp.12743.