Assessing Pharmacogenomic loci Associated with the Pharmacokinetics of Vamorolone in Boys with Duchenne Muscular Dystrophy

Abstract

Human genetic variation (polymorphisms) in genes coding proteins involved in the absorption, distribution, metabolism, and elimination (ADME) of drugs can have a strong effect on drug exposure and downstream efficacy and safety outcomes. Vamorolone, a dissociative steroidal anti-inflammatory drug for treating Duchenne Muscular Dystrophy (DMD), primarily undergoes oxidation by CYP3A4 and CYP3A5 and glucuronidation by UDP-glucuronosyltransferases. This work assesses the pharmacokinetics (PK) of vamorolone and sources of inter-individual variability (IIV) in 81 steroid-naïve boys with DMD aged 4 to <7 years old considering the genetic polymorphisms of CYPS3A4 (CYP3A4*22, CYP3A4*1B), CYP3A5 (CYP3A5*3) and UGT1A1 (UGT1A1*60) utilizing population pharmacokinetic (PopPK) modeling. A one-compartment model with zero-order absorption (T_k0, duration of absorption), linear clearance (CL/F) and volume (V/F) describes the plasma PK data for boys with DMD receiving a wide range of vamorolone doses (0.25~6 mg/kg/day). The typical CL/F and V/F values of vamorolone were 35.8 L/hr and 119 L, with modest IIV. The population T_k0 was 3.14 hr yielding an average zero-order absorption rate (k₀) of 1.16 mg/kg/hr with similar absorption kinetics across subjects at the same vamorolone dose (i.e., no IIV on T_k0). The covariate analysis showed that none of the genetic covariates had any significant impact on the PK of vamorolone in boys with DMD. Thus, the PK of vamorolone is very consistent in these young boys with DMD.

Introduction

Duchenne Muscular Dystrophy (DMD) is a genetic muscle-wasting disease caused by variants of the DMD gene that leads to deficient or defective dystrophin production in muscle¹, resulting in progressive muscle deterioration and later loss of ambulation, and ultimately dilated cardiomyopathy, the need for assisted ventilation and premature death². It mainly affects males with symptom onset in early childhood (around 2–5 years old)³. Corticosteroids (CS) serve as the current standard of care in patients with DMD to help alleviate symptoms and slow disease progression. However, time- and dose-dependent side effects associated with CS, such as loss of bone density, weight gain, behavior changes, and stunting of growth, seriously affect patients’ quality of life⁴.

Vamorolone is a novel steroidal anti-inflammatory drug that is structurally differentiated from other corticosteroids by lacking the 11β oxygen moiety; the 11β position is a hydroxyl in active hormones and drugs (cortisol; active corticosteroid drugs), or a ketone in inactive hormones and drugs (cortisone; corticosteroid pro-drugs). As a result of loss of the 11β oxygen (replaced by a Δ9–11 carbon double bond), vamorolone uniquely loses substrate activity for the two drug/pro-drug modulatory 11β-hydroxysteroid dehydrogenase enzymes (HSD11B1, HSD11B2)⁵. The loss of the 11β oxygen also removes a contact site with the glucocorticoid receptor, changing the 3-dimensional structure of GR/ligand complexes at the obligatory co-factor binding site, with resulting changes in co-repressor (increased) and co-activator (decreased) binding⁶. Vamorolone is an antagonist of the mineralocorticoid receptor, with activities similar to eplerenone and spironolactone in pre-clinical models – this is also unique among the steroidal anti-inflammatory drugs⁷. Thus, while vamorolone binds to the same receptors as corticosteroids (i.e., glucocorticoid and mineralocorticoid receptors), vamorolone modifies the downstream activity of the receptors and shows beneficial efficacy/safety profiles versus other corticosteroids as seen in pre-clinical (mouse) studies^{7, 8}, and human clinical trials.

Vamorolone received regulatory approval for use in Duchenne muscular dystrophy in 2023 by both EMA (dissociative steroid label; age 4 years and older), and FDA (corticosteroid label; age 2 years and older). In a double-blind clinical trial of vamorolone vs. placebo and prednisone, vamorolone showed robust evidence of clinical benefit for all motor outcomes studied, and reduced safety concerns relative to prednisone (no stunting of growth, no deleterious changes in bone biomarkers)^{9, 10}. These findings were further supported by long-term open label studies vs. natural history comparators^{11, 12}.

Vamorolone is primarily metabolized by oxidation (CYP3A4, CYP3A5) and glucuronidation (UDP-glucuronosyltransferases). Corresponding genes contain polymorphisms that result in different enzymatic functionalities and thereby contribute to variability in pharmacokinetics (PK) of their substrates. One of the key single-nucleotide polymorphisms (SNPs) related to the functional variation of CYP3A5 is 6986A>G (rs776746), with the wild-type A nucleotide defined as *1 allele and the variant G nucleotide defined as *3 allele. Homozygous carriers of G (GG or CYP3A5*3/*3) produce very low or non-detectable levels of functional CYP3A5 protein¹³, and were reported to have significantly higher systemic exposure of tacrolimus and lower dose requirements as compared to CYP3A5*1/*1 (AA) or CYP3A5*1/*3 (AG) carriers among solid-organ transplant recipients¹⁴. CYP3A4*1B (rs2740574, −392A>G) is one of the most extensively studied CYP3A4 SNPs, which is linked to enhanced CYP3A4 activity^{15, 16}. Another relevant CYP3A4 SNP is CYP3A4*22 (rs35599367), with the homozygous GG being the wild type and the presence of AG or AA associated with decreased enzyme activity of CYP3A4^{17, 18}. The c.−3279 T > G mutation of UGT1A1 (UGT1A1*60, rs4124874) is more common in Caucasians than other ethnic groups^{19, 20}. This SNP results in reduced UGT1A1 activity and is associated with the toxic effects of irinotecan due to decreased glucuronidation of its active metabolite SN-38 by UGT1A1. A previous case report showed that the reduced methylprednisolone clearance in a male subject could not be explained by the CYP3A5*3/*3 genotype²¹.

Despite the established relationships of drug exposure and these pharmacogenomic gene loci in other disease states that have led to genotype-guided dosing of certain drugs^{17, 22–25}, no studies of pharmacogenomics (PGx) have been done with vamorolone. The PK of vamorolone and sources of variability were assessed by population PK (PopPK) analysis previously using data from parallel-group studies in 86 healthy men (fed and fasted, phase I, NCT02415439) and 48 boys with DMD (phase IIa, NCT02415439)²⁶. A one-compartment model with zero-order absorption, linear clearances (CL/F) and volumes (V/F) was applied, and covariates found to significantly affect the PK were food (high fat meal) and body weight (BW). This work builds upon this previous PopPK analysis to study 81 boys with DMD (27 from a phase IIa dose-ranging study [NCT02760264], and 54 from a more recent phase IIb study [NCT03439670]). We assess the impact of gene polymorphisms on vamorolone PK in boys with DMD (i.e., male, steroid-naïve, 4 to <7 years of age) undergoing these clinical trials with vamorolone.

Methods

Subjects and PK Sampling

Ethical standards committee approvals on the VBP15–002, VBP15–003, and VBP15–004 clinical trials were obtained for each of the 33 participating clinical trial recruitment sites, either for individual institutional or regional approvals. Written informed consent was obtained from all parents/guardians of participants in each trial (consent for research). The studies were conducted in accordance with the International Conference on Harmonization Guideline on Good Clinical Practice and applicable national or regional regulations or guidelines.

The patient populations assessed in the current PopPK analysis were previously steroid-naïve boys with DMD aged 4 to <7 years old from two vamorolone clinical trials (i.e., VBP15–002 and VBP15–004). VBP15–002 (NCT02760264) is a first-in-patient 2-week, open-label Phase IIa multiple ascending dose (0.25, 0.75, 2.0 and 6.0 mg/kg/day) study with 12 boys being assigned to each of the dose groups. Serial PK samples were collected on days 1 and 14 at 0 (pre-dose), 1, 2, 4, 6, and 8 hours post-dose. VBP15–004 (NCT03439670) is a randomized, double-blind, placebo- and prednisone-controlled Phase IIb trial carried out in 121 boys with DMD that included two 24-week treatment periods. For treatment period 1 (Weeks 1–24), participants were randomly assigned to the placebo (tablets or oral suspension), prednisone (0.75 mg/kg/day as tablets), and vamorolone (2 or 6 mg/kg/day as oral suspension) groups in a 1:1:1:1 ratio. A 4-week transition period (Weeks 25–28) followed the end of treatment period 1 during which all subjects continued to receive the same oral suspension (vamorolone or matching placebo) that they received during treatment period 1 and were tapered off their study medication tablets (prednisone or matching placebo). During treatment period 2 (Weeks 28–48), all participants received vamorolone at one of two dose levels (2 or 6 mg/kg/day) and blood samples were collected at 2 hours post-dose at the Week 2 Visit. Details of these clinical studies have been published^{9, 27}.

Single-Nucleotide Polymorphism (SNP) Genotyping and Metabolizer Phenotyping

All clinical trial participants in this current study provided informed consent (by their parents) for genetic studies of candidate gene loci. Blood samples for the genetic analyses were collected at clinical trial sites in EDTA tubes, then frozen after collection, and sent to the central laboratory. For the VBP15–002 trial, the blood samples were then sent to AGADA BioSciences (Halifax, Nova Scotia, Canada) where DNA isolation was carried out using QIAamp DNA Blood Maxi or Midi Kits. For VBP15–004, DNA isolation from frozen samples was carried out by the central laboratory, and DNA shipped to AGADA BioSciences. DNA samples were quantitated and adjusted to 100–350 ng/μl, and aliquoted by AGADA BioSciences, gridded onto 96-well microtiter plates. SNP genotyping was carried out by the Genomics Core of Icahn School of Medicine at Mt Sinai, NY using the Illumina GDA-PGx platform (GlobalDiversity Array with enhanced pharmacogenomics content). A data mask was utilized to only retrieve information related to the candidate genetic loci, thus obscuring possible identification of participants through their genetic profiles.

The specific SNPs assessed in this study included CYP3A4*22 C>T (rs35599367, C=CYP3A4*1, T=CYP3A4*22), CYP3A4*1B −392A>G (rs2740574; A=CYP3A4*1, G=CYP3A4*1B), CYP3A5*3 6986A > G (rs776746; A= CYP3A5*1, G= CYP3A5*3), and UGT1A1*60 c.−3279 T > G (rs4124874, T= UGT1A1*1, G=UGT1A1*60). For CYP3A4*22, GG allele genotype (CYP3A4*1/*1) is considered as expresser allele while AG (CYP3A4*1/*22) and AA (CYP3A4*22/*22) allele genotypes are considered as a decrease of function (DOF) alleles. For CYP3A4*1B, AA (CYP3A4*1/*1) is the wild type while GA (CYP3A4*1/*B) and GG (CYP3A4*B/*B) are DOF variants. For CYP3A5*3, individuals carrying at least one functional allele, i.e., AA (CYP3A5*1/*1) or AG (CYP3A5*1/*3), are considered as CYP3A5 expressers. UGT1A1*60 expressers carry at least one functional allele, i.e., AA (*1/*1) and CA (*1/*60) and nonexpressers are homozygous carriers of the mutant allele C (*60/*60).

As CYP3A4 and CYP3A5 exhibit substrate overlaps and comparable expression levels in some individuals^{13, 28}, the CYP3A4*22 and CYP3A5*3 genotypes are commonly combined to assess the interaction between CYP3A4 and CYP3A5 polymorphisms^{23, 29–31}:

Poor metabolizers (PM): individuals who are CYP3A5 nonexpressers (*3/*3) and carriers of at least one DOF allele of CYP3A4 (*1/*22; *22/*22). Normal metabolizers (NM): individuals who are CYP3A4 normal expressers (*1/*1) and CYP3A5 nonexpressers (*3/*3) or those who carry at least one CYP3A5 expresser allele (*1/*1; *1/*3) and at least one DOF allele of CYP3A4*22 (*1/*22; *22/*22). Ultra-metabolizers (UM): individuals with normal CYP3A4 activity (*1/*1) and at least one CYP3A5 functional allele (*1/*1; *1/*3).

The enzyme phenotype assignments based on the expresser status for the genotypes of individual or combined SNPs are listed in Table 1.

Table 1.

Summary of subject covariates assessed in this PopPK analysis.

Covariate		VBP15–002 (n=27)	VBP15–004 (n=54)	Total (n=81)
Body weight (kg)
Mean (CV%)		20.14 (13.75)	20.19 (16.42)	20.18 (15.5)
Median [Min, Max]		19.97 [15.1, 26.2]	20.15 [14.2, 26.2]	20.1 [14.2, 26.2]
Individual SNP genotype	Phenotypea	n (%)
CYP3A4*22 (C>T, rs35599367)
GG (*1/*1)	NM	23 (85)	44 (81)	67 (83)
AG (*1/*22)	PM	4 (15)	10 (19)	14 (17)
CYP3A4*1B (392A>G, rs2740574)
AA (*1/*1)	NM	23 (85)	50 (93)	73 (90)
GA; GG (*1/*1B; *1B/*1B)	UM	4 (15)	4 (7)	8 (10)
CYP3A5*3 (6986A > G, rs776746)
AA; AG (*1/*1; *1/*3)	NM	4 (15)	9 (17)	13 (16.1)
GG (*3/*3)	PM	23 (85)	45 (83)	68 (83.9)
UGT1A1*60 (c.−3279 T > G, rs4124874) a
AA; CA (*1/*1; *1/*60)	NM	22 (81.5)	42 (77.8)	64 (79)
CC (*60/*60)	PM	5 (18.5)	12 (22.2)	17 (21)
Combined SNP genotype
CYP3A4
*1/*1	NM	19 (70.4)	40 (74.1)	59 (72.8)
*1/*22; *1B/*22	PM	4 (14.8)	10 (18.5)	14 (17.3)
*1B/*1B; *1/*1B	UM	4 (14.8)	4 (7.4)	8 (9.9)
CYP3A4/5 b
3A4*1/*1 & 3A5*3/*3; 3A4 (*1/*22; *22/*22) & 3A5 (*1/*1; *1/*3)	NM	19 (70.4)	35 (64.8)	54 (66.7)
3A4 (*1/*22; *22/*22) & 3A5*3/*3	PM	4 (14.8)	10 (18.5)	14 (17.3)
3A4*1/*1 & 3A5 (*1/*1; *1/*3)	UM	4 (14.8)	9 (16.7)	13 (16.0)

^aIn NONMEM, NM, PM, and UM were indicated by 0, 1, and 2.

Population PK Analysis

The PopPK analysis was carried out using the first-order conditional estimation method with interaction (FOCEI) implemented in the nonlinear mixed effects modeling program (NONMEM) version 7.5.0 (ICON Development Solutions, San Antonio, TX, USA). NONMEM was used with Pirana (version 21.11.1), Perl-speaks-NONMEM (PsN, version 5.3.0), and Xpose (version 4). Data manipulation, plotting, and descriptive statistics calculation were performed using R (version 4.3.0).

All PK data were simultaneously described by a one-compartment model with zero-order absorption as was done previously²⁶:

$$V/F\bullet \frac{dC}{dt}=\mathit{\text{Input}}-CL/F\bullet C$$ (1)

$$\mathit{\text{Input}}=\left\{\begin{array}{c}\frac{\mathit{\text{Dose}}}{T_{k0}},t\le T_{k0}\\ 0,t>T_{k0}\end{array}\right.$$ (2)

where V/F is the apparent volume of distribution, CL/F is the apparent clearance, C is plasma concentration, Input is the zero-order absorption rate (k₀), and T_k0 is the duration of Input.

For individual i, the model parameters are:

$$T_{k0,i}=\widetilde{T_{k0}}{\bullet \text{e}}^{{\eta }_{T_{k0},i}}$$ (3)

$${V/F}_i=\widetilde{V/F}\bullet {\text{e}}^{{\eta }_{v,i}}$$ (4)

$${CL/F}_i=\widetilde{CL/F}\bullet {\text{e}}^{{\eta }_{CL,i}}$$ (5)

where bars indicate typical values of parameters; the IIV of model parameters were described using an exponential function and η values are the random-effect parameters that were assumed to be normally distributed with mean 0 and variance ω²; individual-specific parameters were assumed to have a log-normal distribution.

Residual variability was described using an additive plus constant coefficient of variation model:

$$Y_{ij}={\widehat{Y}}_{ij}+{\widehat{Y}}_{ij}\bullet {\epsilon }_{1ij}+{\epsilon }_{2ij}$$ (6)

where Y_ij and ${\widehat{Y}}_{ij}$ are the j^th observed and predicted concentrations for the i^th individual; ${\epsilon }_{1ij}$ and ${\epsilon }_{2ij}$ are the proportional and additive random residual effects. Both ε were assumed to be normally distributed (${\epsilon }_{1ij}~\text{N}\left(0,{{\sigma }^2}_{\text{prop}}\right)$, ${\epsilon }_{2ij}~\text{N}\left(0,{{\sigma }^2}_{\text{add}}\right)$).

The relationships between individual parameter estimates obtained by the base PopPK model and the genetic covariates listed in Table 1 were initially assessed by graphical evaluation. Next, potential relationships were evaluated using the Stepwise Covariate Modeling (SCM) automated method in PsN implemented by Pirana with the forward selection and backward elimination. For the forward selection, statistical significance was considered a decrease of the objective function value (OFV) of at least 3.84 (χ², p < 0 .05, df=1) and for backward elimination an increase of OFV of at least 10.83 (χ², p < 0.001, df=1). Functional forms used for the covariate analysis included a power model for continuous covariates, i.e., BW (Eq.7) and a linear function for categorical covariates, i.e., pharmacogenomic loci (Eq.8):

$$P_i=P_{pop}\bullet {\left(\frac{B{W}_i}{\mathit{\text{Median\hspace{0.17em}}}BW}\right)}^{{\theta }_{BW{P}_i}}$$ (7)

$$P_i=\left\{\begin{array}{c}{P}_{pop},NMs=0\\ P_{pop}\bullet (1+{\theta }_{PMs}),PMs=1\\ P_{pop}\bullet (1+{\theta }_{UMs}),UMs=2\end{array}\right.$$ (8)

where P_i represents the individual PK parameter; P_pop indicates the typical value of the parameter; θ_BWPi is the power coefficient reflecting the change in the natural log of individual parameter per unit change in the natural log of the ratio of individual BW_i and median BW (20.1 kg); θ_PMs and θ_UMs reflect the fold-changes in parameters for PM and UM as compared to those for NM.

The evaluation of model performance was based on goodness-of-fit (GOF) plots, precision of parameter estimates, the OFV and visual predictive check (VPC). Typical GOF diagnostic plots consist of observed concentration (DV) versus individual predicted concentration (IPRED), DV versus population predicted concentration (PRED), conditional weighted residuals (CWRES) versus PRED, and CWRES versus time since the first dose (TIME). The VPC was performed with 1000 simulated datasets that were obtained based on the parameter values from the final model, with the 5^th, 50^th, and 95^th percentiles of the predicted concentration-time data plotted and compared to those of the observed data to assess the agreement graphically.

Individual PK exposure values of vamorolone in boys with DMD, i.e., area under the concentration-time curve from time 0 to infinity (AUC_inf), were further obtained from dose divided by the PopPK model predicted individual CL/F.

Statistical Analysis

The statistical analysis was performed using R (version 4.3.0). Linear regression was applied to assess the relationships between individual PK parameters and continuous covariates. The Wilcoxon rank sum test was applied to compare the categorical data between two groups and the Kruskal-Wallis test was used when there were more than two groups. Statistically significant differences were concluded when p <0.05.

Results

A total of 169 participants with DMD were enrolled in the VBP15–002/003 (n=48) and VBP15–004 clinical trials (n=121). A total of 123 boys with DMD (47 from VBP15–002 and 76 from VBP15–004) who had evaluable vamorolone PK data are displayed in Figure 1. Missing PK data from VBP15–004 (n=45 missing) was due to missed visits due to the COVID pandemic, or refusal of the participant for PK sampling. Among these, 81 boys (27 from VBP15–002 and 54 from VBP15–004) with available PGx data were included. DNA samples that were missing were due to national ethical considerations (Israel) or refusal of the participant for blood draws for DNA. Concentrations below the lower limit of quantification (LOQ) (i.e., 2 ng/mL) were also included in the final modeling dataset. The value and feasibility of incorporating concentration data below the LOQ in PK analysis such as PopPK modeling have been demonstrated previously^{32, 33}. Covariates BW and PGx are summarized in Table 1. The BW distributions of boys with DMD from the two clinical studies are similar, with an overall median value of 20.1 kg (range: 14.2~26.2 kg). Among all included subjects, UGT1A1 NMs (*1/*1; *1/*60) and PMs (*60/*60) accounted for 79% and 21%. For the remaining CYP SNPs, the variant allele frequencies of CYP3A4*22 (A) and CYP3A4*1B (G) are comparable (17% and 10%) with more than 80% of boys carrying the wild-type alleles (*1/*1). In contrast, the majority (84%) of boys (68 out of 81) were CYP3A5 PMs carrying the non-functional homozygote GG (CYP3A5*3/*3) whereas 13 (16%) were NMs with at least one functional allele (i.e., AG or AA). For the combined CYP3A4 and CYP3A4/5 genotypes, 72.8% and 66.7% were designated as NMs, 9.9% and 16% were defined as UMs, and PMs accounted for 17.3% in both categories among the 81 patients with available CYP genotypes. White is the predominant race (~98%) in the clinical trials; the allele frequencies of the studied variants of CYP3A4, CYP3A5 and UGT1A1 in boys with DMD were generally consistent with those reported for the white population.^34–37

Figure 1.

Plasma vamorolone concentration-time profiles for boys with DMD from VBP15–002 and VBP15–004 studies. Closed circles are individual observations color coded by the indicated doses (unit: mg/kg/day). Lines connect mean values at each dose. Note: 2-mg/kg/day values from the VBP15–004 study are moved from 2 to 1.8 hr for better visualization.

In our preliminary assessments, all 123 boys with DMD with evaluable vamorolone PK data were included to examine the influence of BW on PK parameters applying a power function (Eq. 7). The BW of the 123 boys were summarized in Supplementary Table S1. The estimated power coefficients for CL/F and V/F were close to 0.75 and 1.0, which is consistent with the typical allometric relationship between BW and these PK parameters³⁸. Therefore, the power coefficients for CL/F and V/F in Eq. 7 were fixed to 0.75 and 1 (Figure 2) in the subsequent PopPK analysis involving the 81 boys with DMD who had PGx data, with the primary goal to assess the impact of genetic covariates on vamorolone PK.

Figure 2.

Individual estimates of clearances (CL/F) and volumes (V/F) versus BW. The solid red line indicates nonlinear regression line applying Eq. 7 with fixed power coefficients of 0.75 for CL/F and 1.0 for V/F.

The plasma PK profiles of vamorolone in the 81 boys with DMD receiving the 0.25~6 mg/kg/day doses were reasonably described by a one-compartment model with zero-order absorption, the same base PK model as previously²⁶. As shown by the PK plots from the 002 study (Figure 3 and Supplementary Figure S1), both individual and population predicted concentrations are in good agreement with the observed data. All model parameters were well estimated with low CV% values (Table 2). The duration of the zero-order absorption process (T_k0) was estimated to be 3.14 hr, yielding an average zero-order absorption rate (k₀) of 1.16 mg/kg/hr. No IIV was estimated on T_k0, indicating that the absorption kinetics of vamorolone in boys with DMD are similar at the same vamorolone dose given the comparable PK data. The apparent clearance (CL/F) and apparent volume of distribution (V/F) were estimated to be 35.8 L/hr (34.9 % IIV) and 119 L (42.8% IIV), which is 1.78 L/hr/kg and 5.92 L/kg given a median BW of 20.1 kg (Table 1). These PK parameter estimates are comparable to those obtained previously where the PK data for healthy men were included²⁶.

Figure 3.

Individual plasma concentration versus time plots for boys receiving 2 or 6 mg/kg of vamorolone from the 002 study. Closed circles are observations; solid and dotted lines indicate individual and population predictions by the PopPK model. The data for study day 1 and day 14 are indicated by red and blue.

Table 2.

Final pharmacokinetic parameter estimates obtained in the population analysis.

Parameter [Units]	Definition	Final Model		Bootstrap
		Estimate	RSE%	Median	95%CI
Population typical values
Tk0 [hr]	Duration of zero-order absorption process	3.14	4.4	3.14	(2.66, 3.68)
CL/F [L/hr]	Apparent clearance	35.8	17	35.83	(31.82, 40.51)
V/F [L]	Apparent volume of distribution	119	13.2	118	(98, 145)
Inter-individual variability
ω2-CL/F	Interindividual variability of CL/F	0.122	38.9	0.114	(0.043, 0.262)
CV%		34.9		33.8
ω2-V/F	Interindividual variability of V/F	0.183	37	0.184	(0.083, 0.312)
CV%		42.8		42.9
Residual variability
σ 2 - Prop	Proportional residual error variance	0.298	13.3	0.291	(0.215, 0.382)
σ 2 - Add	Additive residual error variance	7.02E-08	38.2	6.83E-08	(1.07E-08, 3.81E-07)

Typical GOF diagnostic plots (Supplementary Figure S3) revealed no significant systemic deviation or major bias, indicating the robustness of the model fits. There is a slight bias for individual plasma drug concentration greater than about 500 ng/mL (Supplementary Figure S3 panel A). The IIV on T_k0 was tested in our preliminary analysis, however, neither the IIV can be identified, nor the indicated bias can be improved. This might be either due to the lack of data at early time points or may be explained by similar absorption kinetics across subjects at the same vamorolone dose. Therefore, the IIV on T_k0 was not included in the current model. Supplementary Figure S4 displays the VPC plots stratified by dose. As shown, the 5^th, 50^th and 95^th percentiles of the observed data generally fall within their corresponding 90% prediction intervals of the simulated data generated by 1000 model simulations, representing a good predictive performance of the PopPK model.

For the covariate analysis, the individual PK parameter estimates obtained by the base PK model were plotted against the genetic covariates listed in Table 1 and statistical analysis was performed to examine any potential significant trends. As was shown, no statistically significant covariate-parameter relationship was identified for CL/F (Figure 4) or V/F (Figure S2). In line with the graphical assessment results, the automated SCM also showed that none of the genetic covariates exhibited any significant impact on either the CL/F or V/F of vamorolone in these subjects and thus the base PK model was accepted as the final model.

Figure 4.

Individual estimates of clearances (CL/F) versus genetic covariates. The boxes represent the 25^th to 75^th percentiles, with the whiskers extending to the 5^th and 95^th percentiles of the data and the asterisks representing outliers. The boxes are joined at the median values.

Bootstrap analysis was performed on the final model yielding a 96.4% convergence rate (i.e., the model converged in 482 out of 500 bootstrap runs). The bootstrap results are presented in Table 2. The original final model parameter estimates are close to the bootstrap medians and are included in the 95%CI based on the bootstrap procedure, suggesting good qualification of the final PopPK model in characterizing the PK data of vamorolone in boys with DMD.

The individual AUC_inf values of vamorolone in boys with DMD calculated from Dose divided by (CL/F) are summarized in Table 3. The predicted AUC_inf values in this work are comparable to those obtained by non-compartmental analysis (NCA) using the day-14 PK data for boys with DMD from the 002 study³⁹.

Table 3.

Summary of individual AUC_inf (ng×hr/mL) values predicted by the PopPK model.

Dose [mg/kg/day]	No. of Subjects		Mean (CV%)	Median [Min, Max]
	VBP15–002	VBP15–004
0.25	9	NA	132 (30.5)	120 [81, 230]
0.75	6	NA	608 (43.2)	589 [270, 1059]
2	6	21	1130 (13.0)	1150 [720, 1340]
6	6	33	3361 (8.8)	3340 [2320, 3990]

NA: not applicable

Discussion

It is increasingly recognized that human genetic variation (polymorphisms) in gene coding proteins involved in ADME processes can have a strong effect on drug exposure. In a study of over 1000 patients seen at the Mayo Clinic, 99% were found to have actionable variants (observable after drug dosing) in key PGx gene loci⁴⁰. A recent study of routine genotyping of 9 PGx loci (including CYP3A4, CYP3A5) found that preemptive screening offered medication improvement opportunities in 56% of participants for commonly used medications.⁴¹

Vamorolone has had ADME aspects defined through an extensive series of in vivo (mouse, dog, human) and in vitro (human liver cells and homogenates, cell lines) studies (https://www.accessdata.fda.gov/drugsatfda_docs/label/2023/215239s000lbl.pdf; https://www.ema.europa.eu/en/documents/product-information/agamree-epar-product-information_en.pdf). Additional aspects of tying drug dose to drug exposure includes drug-drug interactions (DDI), where metabolic pathways of vamorolone might be altered by concomitant medications leading to altered vamorolone exposure for a given dose, or vice versa (a strong inhibitor of CYP3A4, itraconazole, increased vamorolone Cmax and AUC by 8% and 44%) (https://www.accessdata.fda.gov/drugsatfda_docs/label/2023/215239s000lbl.pdf). As demonstrated by these studies, CYP3A4 and CYP3A5 are the major CYP isoforms involved in the metabolism of vamorolone. In addition, this nonionized drug is made more hydrophilic (as are all steroid hormones) by glucuronidation. Indeed, mass balance and metabolite studies of vamorolone in both adult volunteers and boys with DMD show that inactive glucuronides are at high concentrations in serum (72% of total drug-related product in sera). Glucuronidation is carried out by UGTs, and this is a large family of drug metabolism loci (as are the CYPs). The most important UGT loci for corticosteroids are UGT1A1 and UGT1A4⁴². Both these loci have impactful and relatively common PGx alleles. The most common genetic variant for UGT1A1 seen in the Caucasian population is UGT1A1*28 (rs8175347, (TA)7TAA), with approximately 10% carrying 2 copies of the UGT1A1*28 allele. The presence of the defective variant results in reduced UGT enzyme activity and is significantly associated with the risk of irinotecan toxicity due to accumulation of its active metabolites in the blood. Carriers of the homozygous allele UGT1A1*28/*28 are at increased risk for neutropenia following the treatment of irinotecan, and a reduction in the starting dose is recommended for these individuals^{43, 44}. Evidence has shown that there is a highly significant linkage disequilibrium that exists between UGT1A1*60 c.−3279 T > G and UGT1A1*28 in Caucasians¹⁹. For UGT1A4, there are key haplotypes that have been shown to have strong impact on UGT activity⁴², but these are relatively rare alleles making it challenging to develop PGx associations.

In this work, the gene polymorphisms of CYP3A5*3, CYP3A4*22, CYP3A4*1B, and UGT1A1*60 were examined and their impact on vamorolone PK were assessed in steroid naïve boys with DMD aged 4 to <7 years old utilizing PopPK modeling. The previous PopPK analysis utilized extensive data in healthy men to establish the base PK model (zero-order absorption, one-compartment PK, and linear clearances (CL/F) and volumes (V/F)²⁶, where the significant covariates were found to be food effect for all PK parameters (T_k0, CL/F and V/F) and BW for T_k0. Here, the same base PK model was also applicable for describing the more limited PK data in boys with DMD, where the IIV was estimated on CL/F and V/F but not T_k0. According to the results of the previous work²⁶ and given the relatively homogenous DMD patient population with a narrow age range assessed herein, only BW and gene polymorphisms of metabolizing enzymes relevant to vamorolone PK were included as potential covariates (Table 1) in this analysis. Preliminary assessments revealed typical allometric relationships between BW and PK parameters (i.e., CL/F and V/F) when all 123 boys with DMD who had evaluable vamorolone PK data were included. As such, this piece of information was incorporated into the PopPK model to solely assess the impact of gene polymorphisms on the PK for 81 boys who had both PK and PGx data. Based on the pre-defined criteria for covariate inclusion, our analysis indicated that none of the genetic covariates was considered to significantly affect the PK of vamorolone in these boys with DMD.” It is noteworthy that the PM genotypes of each of the genetic loci showed the expected decreased values of both vamorolone clearance (CL/F) and estimates of volumes (V/F) relative to NM genotypes. However, the wide variance in PK estimates, in part due to single sample estimates (2 hours) in the VBP15–004 study, which the majority (2/3^rd) of participants were from, make these findings preliminary.

While the multiple blood sampling from the −002 studies provides a robust basis for the PopPK, one limitation of this study and analysis is the single blood sampling in the −004 study at the 2 and 6 mg/kg vamorolone doses. This is not uncommon in PopPK studies, especially in children. Another limitation of this study is the relatively small number of participants (n=81) from the two trials (n=169 participants total) where both PK and DNA samples were available. This was due, in part, to some countries limiting or prohibiting DNA studies of trial participants, and limitations on patient study visits and missed sampling on participants due to the COVID-19 pandemic.

Conclusions

The gene polymorphisms of CYP3A5*3, CYP3A4*22, CYP3A4*1B, and UGT1A1*60 were examined and their impact on vamorolone PK were found to be negligible in steroid naïve boys with DMD aged 4 to <7 years old utilizing PopPK modeling. The results of this work adds to the evidence that vamorolone exhibits favorable PK profiles (linear and stationary with moderate variability in PK parameters) in the DMD population studied.

Data Sharing and Data Availability

Data are fully displayed in the manuscript and supplemental files and will be available to qualified individuals upon written request.