Impact of SLCO1B1 Genetic Variation on Rosuvastatin Systemic Exposure in Pediatric Hypercholesterolemia.

This study investigated the impact of SLCO1B1 genotype on rosuvastatin systemic exposure in hypercholesterolemic children and adolescents. Participants (8-21 years) with at least one allelic variant of SLCO1B1 c.521T>C (521TC, n = 13; 521CC, n = 2) and wild type controls (521TT, n = 13) completed a single oral dose pharmacokinetic study. The variability contributed by SLCO1B1 c.521 sequence variation to rosuvastatin (RVA) systemic exposure among our pediatric cohort was comparable to previous studies in adults. RVA concentration-time curve from 0-24 hours (AUC0-24 ) was 1.4-fold and 2.2-fold higher in participants with c.521TC and c.521CC genotype compared 521TT participants, respectively. Interindividual variability of RVA exposure within SLCO1B1 genotype groups exceeded the ~ 1.5-fold to 2-fold difference in mean RVA exposure observed among SLCO1B1 genotype groups, suggesting that other factors also contribute to interindividual variability in the rosuvastatin dose-exposure relationship. A multivariate model performed confirmed SLCO1B1 c.521T>C genotype as the primary factor contributing to RVA systemic exposure in this pediatric cohort, accounting for ~ 30% of the variability RVA AUC0-24 . However, of the statins investigated to date in the pediatric population, RVA has the lowest magnitude of variability in systemic exposure.

WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

☑ In children with hypercholesterolemia, response to 3‐hydroxy‐3‐methyl‐glutaryl Coenzyme A reductase inhibitors (statins) is variable. In adults, genetic variation of SLCO1B1 influences the pharmacokinetics of statins. To date, the mechanisms that influence statin disposition in a developing child is unknown.

WHAT QUESTION DID THIS STUDY ADDRESS?

☑ In the present study, we investigated the impact of SLCO1B1 c.521C>T genotype on rosuvastatin (RVA) systemic exposure in children and adolescents.

WHAT DOES THIS STUDY ADDS TO OUR KNOWLEDGE?

☑ Similar to adults, SLCO1B1 c.521 allelic variation impacts the pharmacokinetics of RVA in children, however, other unknown factors contribute to interindividual variability in the dose‐exposure relationship. There is less interindividual variability in RVA compared with pravastatin and simvastatin in the pediatric population.

HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

☑ This study highlights that less variability in systemic exposure was observed with children and adolescents dosed RVA compared with pravastatin and simvastatin.

Rosuvastatin (RVA) is a hydrophilic, synthetic inhibitor of 3‐hydroxy‐3‐methyl‐glutaryl Coenzyme A reductase, labeled to treat children 8 years and older with heterozygous familial hypercholesterolemia.1, 2, 3 RVA is administered in active acid form and undergoes hepatocellular uptake via the drug transporters, OATP1B1, OATP1B3, OATP2B1, and NTCP.1, 4 Minor cytochrome P450 (CYP)‐mediated metabolism via CYP2C9 leads to the formation of a minimally active metabolite, N‐desmethyl rosuvastatin (NDMRV; Figure 1).1, 2 RVA undergoes phase II metabolism via UGT1A3 leading to the formation of an inactive metabolite, rosuvastatin lactone (RVL).5 RVA is also a substrate for BCRP, responsible for hepatic clearance into the bile.6, 7 Due to its hydrophilic nature, RVA, similar to pravastatin, may experience less passive diffusion across the blood brain barrier8, 9 and skeletal muscle,1, 10, 11, 12, 13, 14, 15 making it a potentially safer statin alternative for maturing brains and myocytes in children compared with highly lipophilic statin agents (e.g., simvastatin).

Figure 1

Rosuvastatin (RVA) pathway. RVA undergoes CYP2C9‐mediated biotransformation to N‐desmethyl rosuvastatin and UGT‐mediated lactonization to form RVA lactone.

Two double‐blind, randomized, placebo‐controlled trials of rosuvastatin in children (ages 6–17 years) have demonstrated an ~ 35–50% reduction in low density lipoprotein cholesterol (LDL‐C), validating its efficacy in this age group.16, 17 However, considerable (approximately three‐ to fivefold) interindividual variability in drug response, as indicated by reductions in LDL‐C, has been observed.18 Additionally, ~ 35–60% of participants failed to achieve the target LDL‐C goal (< 130 mg/dL) despite documented adherence to the drug.16, 17 With this large degree of variability in response and unknown long‐term effects of chronic statin exposure, providing dose‐optimization (e.g., the lowest dose that achieves maximal response with minimal risk of toxicity) is essential to improving hypercholesterolemia treatment in the developing child.

In order to investigate the etiology of interindividual variability in statin response, however, we must determine if poor or no response is a function of inadequate drug exposure or altered drug target engagement due to diminished expression and/or function of the drug target proteins. One of the largest sources of variation in the dose‐exposure relationship is hepatic uptake of statins to the site of action within the liver. The SLCO1B1 gene encoding the hepatic drug transporter OATP1B1 has been shown to be a major contributor to RVA transport.4 Furthermore, the c.521T>C single nucleotide polymorphism (rs4149056) located in exon 5 of the SLCO1B1 gene affects the localization of the transporter on the basolateral membrane of the human hepatocyte.19 Functionally, this alteration leads to a nonsynonymous amino acid change (Val174Ala) that contributes to decreased human hepatocyte uptake4 and increased RVA systemic exposure.20, 21 In adults, Pasanen et al.20 demonstrated that a SLCO1B1 single nucleotide polymorphism (SNP; c.521T>C in exon 5) was associated with an ~ 1.5‐fold increase in RVA exposure compared with participants with the reference genotype (c.521TT). The association between SLCO1B1 c.521T>C gene variation and clinical response to statins (e.g., LDL‐C reduction, myopathy) has been demonstrated in several studies,22, 23, 24, 25 but in large part, the association regarding its impact on LDL‐C reduction remains ambiguous.26, 27, 28 The largest association of SLCO1B1 gene variation and myopathy has been observed with simvastatin acid and current Clinical Pharmacogenetics Implementation Consortium (CPIC) guidelines are available for simvastatin dosing according to SLCO1B1 genotype.29

Growth and development may modulate the magnitude of the genotype‐phenotype relationship in children, limiting the extrapolation of adult data to hypercholesterolemic children. In our recent pediatric genotype‐stratified, single‐dose, pharmacokinetic study of simvastatin in children > 8 years, the genotype‐phenotype relationship observed in adults was confirmed with each copy of the variant c.521C allele contributing to a 2.5‐fold increase in simvastatin acid systemic exposure.30 The magnitude of the genotype effect was twofold greater in children compared with adults, suggesting SLCO1B1 c.521T>C is more influential in children. In contrast, the magnitude of genotype effect in the same cohort of children receiving pravastatin was comparable to adults.31, 32 Of concern though, was the extent of interindividual variability in simvastatin and pravastatin systemic exposure that was observed within genotype groups (~ 8‐fold to 17‐fold range).31, 33 Currently, the influence of ontogeny on the RVA dose‐exposure relationship is not well‐established.

Understanding the impact of ontogeny on the genotype‐phenotype relationship for key drug transporters influencing statin disposition is of critical importance in optimal dose selection for pediatric patients. Thus, the primary goal of this investigation was to establish the role of genetic variation in SLCO1B1 on the rosuvastatin dose‐exposure relationship in hypercholesterolemic children and adolescents.

Methods

Subjects

Subjects meeting inclusion and exclusion criteria for this study (Supplemental Materials ) were recruited from the Children’s Mercy Hospital Cardiology Pharmacogenomics Repository, a living biorepository and patient registry designed to facilitate genotype‐guided clinical trial participant selection. Cardiology Pharmacogenomics Repository enrollees were invited to participate based on their SLCO1B1 c.521T>C genotype status. Reference genotype (c.521TT) and heterozygous variant populations (c.521TC and c.521CC) were age‐matched, ethnicity‐matched, and gender‐matched. The study protocol was reviewed and approved by the Children’s Mercy Hospital Institutional Review Board and the study conducted in accordance with US and international standards of Good Clinical Practice (US Food and Drug Administration (FDA) regulations 21 Code of Federal Regulations (CFR) 312 for Investigational New Drug studies and FDA guidance E6).

Genetic analysis

Genomic DNA was isolated from biospecimens using Sigma GeneElute Mammalian Genomic DNA Miniprep Kit (St. Louis, MO) or QIAamp DNA Blood Mini Kit (Valencia, CA) according to manufacturer protocols. All DNA samples were genotyped for the common SLCO1B1 SNPs −11187G>A (rs4149015), c.388A>G (rs2306283), and c.521T>C (rs4149056) using TaqMan SNP genotyping assays (LifeTechnologies, Carlsbad, CA) with KAPA Probe Fast qPCR Master Mix (2X) ABI Prism (KAPA Biosystems, Boston, MA) on a QuantStudio 12 k Flex Real‐Time PCR System (Applied Biosystems, Foster City, CA). Five to 20 ng of DNA was used per reaction. The cycling conditions for all assays were as recommended by the manufacturer. DNA samples from the Coriell Institute for Medical Research were used as controls. Twenty percent of samples were randomly selected and repeated for quality control. All reanalyzed samples were concordant with the original results.

To identify sequence variations in SLCO1B1, SLCO1B3, SLCO2B1, SLC10A1, ABCG2, CYP2C9, and UGT1A3 next‐generation sequencing (NGS) was performed. Briefly, a TruSeq Library was constructed according to manufacturer’s protocol (Illumina, San Diego, CA). A custom targeted capture sequencing panel (Integrated DNA Technologies, Coralville, IA) was used for enrichment. Samples were sequenced on an MiSeq instrument (Illumina) with paired end 200 base pair reads. Read coverage was ~ 500×. SLCO1B1 genotypes obtained by Taqman genotyping were 100% consistent with the NGS data.

CYP2C9 and UGT1A3 genotype calls were made as following: CYP2C9 genotype was called using Astrolabe, a bioinformatic tool that allows star allele calling from NGS data.34, 35 Astrolabe has been expanded to CYP2C9 and 2C19 (manuscript in preparation). For UGT1A3, an SNP report was generated and star alleles manually called using definitions per the UGT Nomenclature site at https://www.pharmacogenomics.pha.ulaval.ca/ugt-alleles-nomenclature/.

Study design

This was a single‐center, open‐label, genotype‐stratified, single oral dose pharmacokinetic study, comparing the disposition of rosuvastatin among hypercholesterolemic children and adolescents with one or more SLCO1B1 c.521C variant alleles to patients homozygous for the reference c.521TT genotype. Participants on current statin therapy withheld statin administration for a washout period of 7 days prior to the drug study visit. All participants had a screening physical examination (including Tanner Staging) performed by a licensed physician prior to drug administration.

Subjects ingested a single oral dose of rosuvastatin (ages 8–21 years: 10 mg tablet, Glenmark Pharmaceuticals, Lot Number FC0259) with 150 mL of water after an overnight fast and resumed meals no earlier than 3 hours after administration of the study dose. Serial venous blood samples (1.5 mL each) were drawn from an indwelling venous cannula before pravastatin administration (time 0), and at 0.5, 1, 1.5, 2, 2.5, 3, 4, 5, 7, 9, 18, and 24 hours post‐ingestion to measure plasma drug and metabolite concentrations. Samples were collected in a syringe, transferred to a tube containing potassium EDTA, gently mixed by inversion, and immediately centrifuged at 4ºC for 10 minutes at 600 g. Plasma was removed and stored at −80ºC until analysis. Predose urine and pooled postdose urine were collected through the duration of the study (24 hours).

Analytical methods

Plasma and urine concentrations of RVA, RVL, and NDMRV were measured on a Waters TQ‐S triple quadrupole tandem mass spectrometer with a novel ultra‐high pressure liquid chromatography‐tandem mass spectrometric method previously developed and validated in our laboratory.36 In brief, the dynamic range of the assay was from 0.5–100 nM for all analytes. The method was linear for all analytes in the concentration range 0.5–100 nM with intraday and interday precisions (as relative SD) of ≤ 10.3% and accuracy (as relative error) ranging from 97–111% at all quality control levels (1, 10, and 75 nM). Plasma samples were analyzed in duplicate. For instances where the coefficient of variation in the duplicates exceeded 20%, the samples were rerun (< 5% of samples).

Pharmacokinetic parameters

Power calculations to determine sample size for our study are detailed in Supplemental Materials . Pharmacokinetic analyses were conducted using Kinetica version 5.0 (ThermoFisher Scientific, Philadelphia, PA). Plasma concentration vs. time data for RVA, RVL, and NDMRV were curve fit using a peeling algorithm to generate initial monoexponential parameter estimates. Final estimates of the terminal elimination rate constant (λz) were determined from an iterative, linear least squares regression algorithm. A model‐independent approach was used and parameters of interest determined as follows. Individual peak plasma concentration (Cmax) and time to maximal concentration (Tmax) were obtained by direct examination of the plasma concentration vs. time profile. The area under the plasma concentration vs. time curve during the sampling period (AUC0–n) was calculated using the mixed log‐linear method, where n refers to the final sampling time with quantifiable drug or metabolite concentrations. Extrapolation of the AUC to infinity (AUC0–∞) was not performed; secondary peaking occurred in a significant number of subjects within the cohort and, therefore, insufficient data points to accurately capture the terminal elimination phase for RVA were available.

Statistical analysis

Pharmacokinetic data for the study cohort were examined using standard descriptive statistics in JMP version 14 (SAS, Marlow, UK). Pharmacokinetic parameters reflective of systemic exposure (Cmax and AUC0– ) were log‐transformed using the natural logarithm. The significance limit accepted for all statistical analyses was α = 0.05. SLCO1B1 genotype (c.521TT vs. c.521TC/CC) was treated as an independent variable. Pharmacokinetic parameters were compared between demographic and genotype groups, using Welch’s t‐test and Kruskal–Wallis test given the non‐normality of the dependent variables. For the pharmacokinetic parameters related to SLCO1B1 genotype groups (Table 2), Kruskal–Wallis test was utilized. For the pharmacokinetic parameters related to demographics (Table ), both Welch’s t‐test and Kruskal–Wallis test were utilized.

Table 2

Dose‐normalized pharmacokinetic variables of RVA and related analytes after a single dose of RVA in relationship to SLCO1B1 genotype

	SLCO1B1 c.521TT (n = 13)	SLCO1B1 c.521TC (n = 13)	SLCO1B1 c.521CC (n = 2)	P value
RVA
Cmax, ng/mL	4.3 (± 2.3)	6.2 (± 2.2)	10.1 (± 0.6)	0.01
Tmax, hour	5.0 (4.0–5.0)	4.0 (1.0–5.0)	4.0 (3.0–7.0)	N/A
AUC0–24, ng/mL*hour	36.0 (± 18.5)	49.2 (± 14.0)	79.7 (± 18.2)	0.01
RVA lactone
Cmax, ng/mL	0.8 (± 0.5)	1.2 (± 0.7)	1.4 (± 0.4)	0.09
Tmax, hour	5.0 (1.0–9.0)	4.5 (2.5–7.0)	5.0 (1.0–9.0)	N/A
AUC0–24, ng/mL*hour	11.8 (± 8.6)	17.2 (± 10.2)	18.5 (± 2.0)	0.17
N‐Desmethyl RVA
Cmax, ng/mL	0.8 (± 0.5)	1.0 (± 0.5)	1.0 (± 0.0)	0.50
Tmax, hour	4.0 (2.5–7.0)	4.0 (2.0–7.0)	5.0 (3.0–7.0)	N/A
AUC0–24, ng/mL*hour	5.4 (± 5.5)	5.8 (± 3.4)	5.7 (± 0.9)	0.44

AUC0–24, area under the curve from 0 to 24 hours; Cmax, peak plasma concentration; N/A, not applicable; RVA, rosuvastan; Tmax, time to maximal concentration.

Data expressed as mean (± SD), Tmax expressed as median (range). Kruskal–Wallis test utilized for all statistical analysis.

We used linear regression to examine associations between RVA AUCn and individual SNPs. For each SNP, we fit one model where the SNP was the only predictor of RVA AUCn, and a second model where sex, body mass index (BMI) percentile, and Tanner stage were added as covariates. For comparison, we also fit a base model with only sex, BMI percentile, and Tanner stage as predictors, allowing us to compute how much additional variability in RVA AUCn could be accounted for by adding each SNP variable to the base model. Tanner stage was computed as the average of the Pubic Hair Scale score and Breast Development/External Genitalia Scale score. CYP2C9 genotype was used to create two variables for use in modeling: number of *1 alleles (0, 1, or 2) and number of *2 alleles (0, 1, or 2). For all other SNPs, the count of variant alleles (0, 1, or 2) was computed for use in modeling.

Because administration of a fixed dose of RVA resulted in an almost fourfold range of weight‐based doses (0.07–0.28 mg/kg), pharmacokinetic exposure parameters (e.g., Cmax and AUC) were normalized to dose for each individual participant by dividing the exposure parameter value by actual mg/kg dose received then multiplied by the mean mg/kg dose for the entire cohort (e.g., participant AUC (ng*hour/mL)/ participant dose (mg/kg) * cohort mean dose (mg/kg)).

Results

Participant characteristics and adverse events

A total of 28 children and adolescents (15 male and 13 female participants) were enrolled in this investigation. The demographic and genetic constitution of the participant population is detailed in Table 1. Demographic parameters were similar between SLCO1B1 c.521T>C variant‐containing genotypes and c.521TT controls. Of note, the weight‐based dose received was similar between the genotype groups (Table 1). No adverse events were reported during the study.

Table 1

Characteristics of participants stratified by SLCO1B1 genotypes

	SLCO1B1 521 TT (n = 13)	SLCO1B1 521 TC (n = 13)	SLCO1B1 521 CC (n = 2)	P value
Age, yearsa	15.2 (± 3.0)	15.0 (± 2.8)	14.4 (± 4.3)	0.965
Weight, kga	86.0 (± 34.2)	87.0 (± 25.8)	61.8 (± 1.6)	0.314
Height, cma	164.2 (± 8.6)	165.6 (± 11.9)	157.5 (± 20.0)	0.659
BMI, kg/m2 a	31.6 (± 11.8)	31.1 (± 6.7)	25.5 (± 5.8)	0.695
Sexb
Female	6	6	1	0.994
Male	7	7	1
Ethnicityb
White, non‐Hispanic	7	7	1	0.951
White, Hispanic	5	5	1
African American	1	1	0
Tanner breast/testicularb
Stage 1	0	0	0	0.165
Stage 2	2	2	1
Stage 3	0	1	0
Stage 4	4	0	0
Stage 5	7	10	1
Pubicb
Stage 1	1	0	0	0.235
Stage 2	0	3	1
Stage 3	1	0	0
Stage 4	3	0	0
Stage 5	8	10	1
Dose, mg/kga	0.14 (± 0.06)	0.13 (± 0.04)	0.16 (± 0.00)	0.402

BMI, body mass index.

All data expressed as mean (± SD).

Designates Kruskal–Wallis test. bDesignates Chi‐square test performed.

Drug disposition profiles

The drug disposition profiles of RVA, RVL, and NDMRV (Figure 2 a–c) were consistent with first‐order absorption and elimination. RVA, RVL, and NDMRV were detected in all participants. Secondary peaking, suggestive of entero‐hepatic recirculation, was noted on the RVA profiles of 11 participants (39%) precluding accurate assessment of the mean terminal elimination rate constant across the population.

Figure 2

Rosuvastatin (RVA) and related metabolite pharmacokinetic profiles (a) mean ± SD plasma concentrations (nM) of RVA and dose‐normalized RVA and (b) RVA lactone and dose‐normalized rosuvastatin lactone (c) N‐Desmethyl rosuvastatin and dose‐normalized N‐Desmethyl RVA after a single dose of RVA in 28 healthy pediatric participants. Black, gray, and open white circles represent participants with the c.521TT (n = 13), c.521TC (n = 13), and c.521CC (n = 2) genotypes, respectively.

Overall, a relatively large degree of variability (approximately eightfold) was observed across the entire population with respect to measures of systemic exposure for RVA (median interquartile range (IQR), Cmax: 5.4 ng/mL (IQR 3.4–7.3 ng/mL); and AUC0–24: 42.7 ng*hour/mL (IQR 31.1–60.3; Table ). Significant variability was also observed for RVL (median IQR, Cmax: 1.1 ng/mL (IQR 0.4–1.5); AUC0–24: 12.9 ng*hour/mL (IQR 7.3–21.5; Table ) and NDMRV (median IQR Cmax: 0.8 ng/mL (IQR 0.5–1.1 ng/mL); AUC0–24: 4.7 ng*hour/mL (IQR 3.0–6.3); Table ).

Effect of SLCO1B1 genotype on systemic exposure

Maximum mean RVA concentrations (Cmax) were 2.3‐fold higher in the c.521CC (n = 2) group and 1.4‐fold higher in the c.521TC (n = 13) group relative to the c.521TT (n = 13) group (Figure 3a, Table 2). Mean RVA AUC0–24 was 2.2‐fold higher in the c.521CC group and 1.4‐fold higher in the c.521TC group relative to subjects with reference genotype (Figure 3b, Table 2).

Figure 3

Rosuvastatin (RVA) exposure (a) comparison of RVA peak plasma concentration (Cmax; ng/mL; P = 0.01) and (b) area under the curve from 0 to 24 hours (AUC0→24; ng*hr/mL) normalized for dose among SLCO1B1 c.521 genotypes (P = 0.01). Black, gray, and open white circles represent participants with the c.521TT (n = 13), c.521TC (n = 13), and c.521CC (n = 2) genotypes, respectively. Red circles represent potential “high outlier” participants (c.521TT, n = 4; c.521TC, n = 4).

RVL exposure was similar among the SLCO1B1 genotype groups (P = 0.17; Table 2). Similarly, NDMRV exposure was similar among c.521TC and c.521CC groups compared with c.521TT (reference) participants (P = 0.44; Table 2).

There was no relationship between RVA or related isomers and the presence of SLCO1B1 −11187G>A (rs4149015) or c.388A>G (rs2306283).

In addition to differences in RVA exposure among SLCO1B1 c.521 genotype groups, variability within genotype groups was observed. The largest range of variability occurred in the c.521TT group (5.5‐fold) compared with the c.521TC group (2.2‐fold). Overall, these data indicate that additional patient‐specific variables, in conjunction with SLCO1B1 genotype, influence RVA systemic exposure. RVA exposure was not normally distributed within the entire cohort or even within the c.521TT and TC genotype groups. In fact, there were four participants within each group that deviated from normality in the normal quantile plot (Figure ) and, subsequently, these were identified as a potential subgroup classified as “high outliers.”

Secondary post hoc analyses were conducted to seek insight into additional factors contributing to high systemic RVA exposure within SLCO1B1 genotype groups.

Effect of demographic and developmental factors on systemic exposure

Simple linear regression analyses of RVA, RVL, NDMRV AUC0–24 on independent variables (age, height, lean body weight, and BMI) revealed only a few weak, positive correlations (Table ). RVA and RVL AUC0–24 were similar among sex and ethnicities (Table ).

Effect of non‐SLCO1B1 sources of genetic variation on systemic exposure

Additional gene sequencing of hepatic transporters associated with RVA uptake (SLCO1B1, SLCO1B3, SLCO2B1, and SLC10A1) and efflux (ABCG2) was performed on all participants (Table ). Gene sequencing of SLCO1B1 and SLCO2B1 revealed no additional sequence variations implicating altered OATP1B1 or OATP2B1 expression and/or function. There were seven participants with the SLCO1B1 c.463C>A (rs11045819) known to result in increased OATP1B1 expression and function. However, two “high outliers” had the rs11045819 variant (one c.521TT and one c.521TC) and the other five participants (three c.521TT and two c.521TC) were in the lower half of their respective genotype. Thus, rs11045819 variation alone does not explain the dichotomy of the cohort within SLCO1B1 c.521T>C genotype groups. Eight participants had a nonsynonymous sequence variation in SLCO1B3 (c.767G>C; rs60140950) that alters transporter expression. The c.767G>C was detected in two “high outliers” in the c.521TC group and one subject in the c.521TT group. There was one additional “high outlier” in the c.521TC group with a nonsynonymous sequence variation in SLC10A1 (c.957G>A; rs149272163). In the entire cohort, three c.521TT participants had a sequence variation in ABCG2 (c.421C>A) with two of these participants being “high outliers.” Six participants had a sequence variation in ABCG2 (c.34G>A) including one c.521TC and one c.521TT “high outlier.” There were no participants having both SNPs, ABCG2 c.34G>A and c.421C>A. Comparison of each of these variants to wild type participants among the entire cohort and within SLCO1B1 c.521T>C subgroups did not achieve significance. Although these aforementioned sequence variations may potentially explain higher RVA exposure for six of the eight outliers, none were solely unique to the “high outlier” group.

Effect of CYP2C9 and UGT on systemic exposure

To determine if altered CYP‐mediated metabolism contributed to the extreme phenotype results within genotype groups, differences in NDMRV metabolite formation was assessed using postdose, pooled urine samples. However, the recovery of NDMRV as a percentage of the total analyte (RVA + RVL + NDMRV) in the aforementioned “high outliers” of both genotype groups was not as a whole lower compared with others. Of note, three “high outliers,” one c.521TC and two c.521TT participants, were within the normal quantile distribution plot (Figure ) with only one “high outlier” c.521TT participant in the lower group having SNPs associated with one nonfunctional CYP2C9 allele (*1/*3). Collectively, CYP2C9 genotype may have contributed to higher RVA systemic exposure for only one “high outlier,” but does not independently contribute to the variability within SLCO1B1 c.521T>C genotype groups.

Similarly, to determine if diminished lactonization in the “high outliers” led to increased RVA systemic exposure, plasma RVL as a percentage of the total analyte (RVA + RVL + NDMRV) was quantitated. However, the percentage RVL/total in the aforementioned “high outliers” of both genotype groups was not as a whole lower compared with others (Figure ). Of note, two “high outliers,” one c.521TC and one c.521TT participant within the lower end of the normal quantile distribution plot, were present and none of them had a sequence variation suggestive of diminished UGT1A3 function. Collectively, UGT1A3 genotype does not seem to contribute to the variability within SLCO1B1 c.521T>C genotype groups.

RVA multivariate model

We explored a number of multiple regression models (e.g., models for SLCO1B1 genotypes and demographic parameters), including multivariate models to quantify associations between the aforementioned SNPs and RVA AUC0–24, controlling for BMI percentile, sex, and Tanner Staging. The most significant SNP associated with RVA AUC0–24 outcome was SLCO1B1 c.521T>C genotype, yielding an R‐squared value of 0.39 in the multivariate model, as compared with an R‐squared of 0.10 when only the 3 covariates were included (Table ). Collectively, SLCO1B1 genotype seems to be the primary factor contributing to RVA systemic exposure among those evaluated in our analysis.

Discussion

The present study investigated the impact of SLCO1B1 genetic variation and developmental factors on RVA pharmacokinetics in children and adolescents. The magnitude of effect for the c.521TC group compared with the reference genotype in children was similar to that reported in adults (i.e., 1.4‐fold in the current study vs. 1.6‐fold as observed by Pasanen et al. 20. In contrast to the adult study where only a difference was observed between reference genotype and c.521CC, we observed a difference among each genotype group. We do recognize that our conclusion with regard to c.521CC genotype is limited given the small sample size (n = 2), this observation is, however, concordant with our previous simvastatin acid and pravastatin acid analyses.31, 33 In these studies, systemic exposure increased with each variant “C” allele that was present.31, 33 The magnitude of the SLCO1B1 genotype effect for RVA is less compared with that found for SVA but similar to PVA in this pediatric cohort.31, 33 Collectively, SLCO1B1 c.521T>C explained nearly three times more variability in RVA exposure compared with BMI percentile, sex, and Tanner Staging in our multivariate model.

Particularly striking was the observation of an ~ 5.5‐fold and 2.2‐fold range of RVA systemic exposures within the c.521TT and TC genotype groups, respectively (TT: 11.7–64.2 ng*hour/mL; TC: 35.3–78.9 11.7–64.2 ng*hour/mL; Figure 3 b). A similar observation has been described in adults with coefficients of variation for RVA AUC ranging from 36–51% within SLCO1B1 genotype groups.20, 21 An unexpected observation was the putative existence of two separate groups within each SLCO1B1 c.521T>C genotype group, as shown in Figure 3 b and Figure . Eight of 28 participants (~ 30%) were deemed as “high outliers.” Collectively, these data indicate that additional patient‐specific variables, in conjunction with SLCO1B1 genotype, influence RVA exposure and these need to be identified in order to effectively tailor RVA treatment to the individual patient.

Our previous investigation involving pravastatin observed that “high outliers” in the c.521TT group had the highest BMI values (Z‐score > +2.5) in the entire cohort, suggesting that liver adiposity may compromise statin transport.31 Previous investigations demonstrate a link between nonalcoholic fatty liver disease and diminished hepatic uptake transporter expression,37, 38, 39 subsequently altering simvastatin and pravastatin transport.39, 40 This association of obesity and diminished RVA transport (e.g., larger systemic exposure) was not observed in this similar pediatric cohort. However, we acknowledge that our studies were not specifically powered to answer this particular post hoc analysis. Collectively, the role of liver adiposity on pediatric statin disposition requires further elucidation.

RVA is a known substrate of several other transporters, including OATP1B1, OATP1B3, OATP2B1, NTCP, and ABCG21, 4 and, thus, additional gene sequencing was performed to ascertain if other sequence variants existed in the “high outliers” to explain the discordant results in our study population (Table ). There were seven participants that harbored the SLCO1B1 c.463C>A (rs11045819) SNP associated with enhanced expression and function. As shown in Table , two of our “high outliers” had rs11045819 and five in the “lower” group had the SNP. Thus, rs11045819 did not explain the dichotomy within c.521T>C groups.

Of the eight outliers, three were found to have nonsynonymous sequence variants in SLCO1B3 (c.767G>C; rs60140950) with one having homozygosity of the variant allele. This SNP was previously investigated by Schwarz et al. 41 and was found to cause diminished total protein and cell surface expression of OATP1B3, but was not implicated to result in diminished uptake of cholecystokinin‐8, a specific OATP1B3 substrate. Of note, they observed other sequence variants that were associated with diminished cholecystokinin‐8 and RVA uptake, but these were not found in our “high outliers.” The role of SLCO1B3 c.767G>C on RVA uptake was not investigated as a part of the aforementioned study but should be investigated in future analyses.

One additional participant in the “high outlier” group had a nonsynonymous SLC10A1 c.957G>A SNP. Previously, Ho et al. reported discordant results with the NTCP*2 allele (SLC10A1 c.800C>T), despite known to result in complete loss of function, having enhanced RVA uptake and NTCP*3 (SLC10A1 c.668T>C), known to alter cell surface trafficking, having diminished RVA uptake. The SLC10A1 c.957G>A SNP was not investigated as a part of this study, but its effect on RVA should be validated in future investigations.

ABCG2 (BCRP) is expressed on the apical membranes of a variety of tissues, including the liver and small intestine,42 and decreased activity of ABCG2 secondary to the c.34G>A and c.421C>A SNPs at the small intestine leads to enhanced absorption. This resultant increase in bioavailability leads to increased RVA systemic exposure, which has been confirmed in adult cohorts.43, 44 Two of the six participants with the ABCG2 c.34G>A SNP and two of three with the ABCG2 c.421C>A SNP were “high outliers.” Although this could lead to the higher exposure within each genotype group for some participants, these SNPs were not unique to the “high outliers” (i.e., were also found in 5 other participants) and, therefore, cannot be definitively concluded as the etiology leading to RVA exposure above the 99% for the cohort.

The role of CYP2C9 genotype on RVA systemic exposure was evaluated in our study cohort, implying that if RVA metabolism was impaired, it could result in higher RVA systemic exposure. There was one “high outlier” with an intermediate metabolizer CYP2C9 genotype. However, this finding did not explain the clear separation of the two groups within each SLCO1B1 genotype. Collectively, CYP2C9 and UGT1A3 genotypes do not seem to have an impact on the interindividual variability observed in the rosuvastatin dose‐exposure relationship in the pediatric population.

Limitations of Study

The primary goal of this investigation was to establish the role of genetic variation in SLCO1B1 on the RVA dose‐exposure relationship in hypercholesterolemic children and adolescents. Secondary post hoc analyses were conducted to ascertain if additional SNPs contributed to high systemic RVA exposure within SLCO1B1 genotype groups. We acknowledge that several of these tested SNPs had low variant allele frequency and may not have been observed in our cohort. In summary, we did observe some SNPs associated with altered RVA disposition, but no single sequence variant consistently accounted for the presence of unexpectantly high RVA AUC.

Conclusion

Collectively, our study demonstrates that SLCO1B1 genotype contributed to the variability in the dose‐exposure relationship observed in the pediatric cohort, yet, within each genotype group there exists variability that cannot be explained by SLCO1B1 genotype alone. Alternative proteins involved in RVA disposition may contribute to the separation observed within the genotype groups, however, no single SNP tested consistently accounted for the presence of expectantly high RVA AUC. Perhaps the most insight comes from the finding that of the statins investigated in the pediatric population, RVA had the lowest magnitude of variability of systemic exposure of individuals within genotype groups compared with pravastatin and simvastatin.31, 33 Moving forward, RVA may be the more ideal agent for future investigations that develop clinical decision support tools to optimize outcomes for children prescribed a statin for the treatment of hypercholesterolemia. Future analyses determining the impact of less variable systemic exposure, as seen with RVA, on the markers of short‐term and long‐term response, needs investigation prior to making it a preferred agent in pediatric populations.

Funding

This study was supported by the Marion Merrell Dow Clinical Scholar Award and the CPR is supported by Children’s Mercy Hospital Fellow Clinical Scholar Award and the Ward Family Heart Center. The author’s Statin Optimization in Pediatrics program is supported by a Clinical and Translational Science Award (CTSA) grant from National Center for Advancing Translational Sciences (NCATS) awarded to the University of Kansas for Frontiers: University of Kansas Clinical and Translational Science Institute (#KL2TR002367). The contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health (NIH) or NCATS.

Conflicts of Interest

The authors declared no competing interests for this work.

Author Contributions

J.W., S.R., A.G., R.G., V.S., and J.S.L wrote the manuscript. J.W., S.R, A.G., G.R., V.S., L.V.H, and J.S.L. designed the research. J.W., R.G., and L.V.H. performed the research. J.W., S.R., V.S, and J.S.L. analyzed the data.

Supplemental Figure 1.

Click here for additional data file.

Supplemental Figure 2.

Click here for additional data file.

Supplemental Figure 3.

Click here for additional data file.

Supplemental Tables 1-5.

Click here for additional data file.

Supplemental Materials 1.

Click here for additional data file.

Supplemental Figure Legends.

Click here for additional data file.