Population Pharmacokinetic and Pharmacodynamic Modeling for the Prediction of the Extended Amlitelimab Phase 3 Dosing Regimen in Atopic Dermatitis

ABSTRACT

Amlitelimab is a fully human, nondepleting, anti‐OX40 ligand monoclonal antibody being investigated for the treatment of moderate‐to‐severe atopic dermatitis (AD) in adults and adolescents. Population pharmacokinetic (PopPK) and pharmacokinetic/pharmacodynamic‐Eczema Area and Severity Index (PopPK/PD‐EASI) models were used to inform dosing regimen selection for amlitelimab phase 3 trials. The PopPK model was developed using phase 1 (healthy volunteers) and phase 2 (participants with AD) trial data, including individual exposure variables from the STREAM‐AD phase 2b trial following subcutaneous amlitelimab doses ranging from 62.5 to 250 mg given every 4 weeks (Q4W). The PopPK model was used to compute exposures for an extended dosing regimen of 250 mg Q12W (with 500 mg loading dose [+LD]). The PopPK/PD‐EASI model was developed from phase 2 trials to predict treatment responses (EASI values) with selected dosing scenarios. Finally, the dose for individuals with lower body weight (i.e., < 40 kg) was determined. Utilizing the PopPK model, the amlitelimab 250 mg Q12W + LD computed exposures were between the exposures of 62.5 mg Q4W and 250 mg Q4W + LD efficacious doses in the STREAM‐AD trial. Using the PopPK/PD‐EASI model, the simulated efficacy for dosing scenarios of 250 mg Q12W + LD regimen from initiation or 250 mg Q4W + LD for 24 weeks followed by Q12W to Week 60 was similar to continuous 250 mg Q4W. Simulations identified that a twofold dose reduction would allow participants < 40 kg to achieve amlitelimab exposures within the range observed in participants ≥ 40 kg on 250 mg Q4W or Q12W. These results support evaluation of a Q12W dosing regimen for adults and adolescents in phase 3 trials.

Steps of the PopPK/PD‐EASI model to inform amlitelimab extended dosing regimen in phase 3 clinical trials.

Study Highlights.

• What is the current knowledge on the topic?

  • ○
    Patients with atopic dermatitis (AD) can suffer from a high treatment burden. Extended dosing of treatments may reduce this burden. Amlitelimab, a fully human, nondepleting, anti‐OX40 ligand monoclonal antibody, is an investigational treatment that has demonstrated clinical improvements in adults with moderate‐to‐severe AD in a phase 2b trial when given subcutaneously every 4 weeks (Q4W).
• What question did this study address?

  • ○
    Population pharmacokinetic (PopPK) and pharmacokinetic/pharmacodynamic‐Eczema Area and Severity Index (PopPK/PD‐EASI) models were developed to evaluate whether extended dosing regimens beyond Q4W were applicable for amlitelimab phase 3 trials in adolescents and adults with moderate‐to‐severe AD.
• What does this study add to our knowledge?

  • ○
    The models can be used to predict drug exposures and simulate exposure and efficacy data to support extended dosing regimens to be utilized in the phase 3 clinical trials.
• How might this change drug discovery, development, and/or therapeutics?

  • ○
    The PopPK and PopPK/PD‐EASI models can be used to optimize extended dosing regimens during clinical development, which may address treatment burden for patients.

1. Introduction

Clinically characterized by recurrent eczematous lesions and intense pruritus, atopic dermatitis (AD) presents as a chronic inflammatory skin disease affecting children, adolescents, and adults [1]. It is associated with substantial impairment of quality of life, and available therapies involve a high treatment burden, often requiring daily, weekly, or biweekly dosing [1, 2, 3, 4, 5]. Moreover, many patients are unable to achieve adequate disease control with those medications [4, 5, 6].

Amlitelimab is an investigational fully human, nondepleting, anti‐OX40 ligand (OX40L) monoclonal antibody for the treatment of moderate‐to‐severe AD [7]. By blocking the interaction of OX40L with its receptor, OX40, amlitelimab may prevent the proliferation of pathogenic T‐cell populations and release of inflammatory cytokines underlying the highly burdensome symptomatology of AD [1, 8]. In a phase 1 clinical study in healthy participants, a supra‐dose proportionality pharmacokinetic (PK) profile of amlitelimab was demonstrated after a single intravenous (IV) administration of 0.006–0.05 mg/kg [9]. Furthermore, no major deviation from dose proportionality was observed for 0.15 mg/kg single IV dose and 0.075 mg/kg repeated doses every 4 weeks (Q4W). Clearance of amlitelimab remained stable at concentrations > 2 μg/mL. In the linear PK range, the terminal half‐life of amlitelimab ranged from 24 to 43 days [9]. Phase 2a and 2b clinical trials have demonstrated the safety and efficacy of amlitelimab in adults with moderate‐to‐severe AD [7, 10]. Both studies met the primary endpoint, showing improvements in Eczema Area and Severity Index (EASI) [7, 10]. The phase 2a trial included a 16‐week induction period, followed by a 20‐week withdrawal period for clinical responders [7], defined as those who achieved a Validated Investigator Global Assessment scale for Atopic Dermatitis (vIGA‐AD) score of 0 (clear skin) or 1 (almost clear skin) (vIGA‐AD 0/1) at Week 16. Clinical responders entered the withdrawal period, in which vIGA‐AD 0/1 was maintained 24 weeks after the last dose of amlitelimab in the majority of participants [7]. The STREAM‐AD phase 2b trial included a 24‐week induction period (Part 1) and a 28‐week maintenance/withdrawal period (Part 2) [10]. Clinical improvements were observed at Week 24 in Part 1 of the STREAM‐AD phase 2b trial with subcutaneous (SC) amlitelimab given Q4W compared with placebo. Furthermore, in Part 2 of the trial, the majority of clinical responders at Week 24—defined as participants who achieved a 75% reduction in the EASI (EASI‐75) and/or IGA 0/1—who were rerandomized to continue amlitelimab treatment demonstrated maintenance of clinical response at Week 52. This maintenance of clinical response at Week 52 was also observed in those rerandomized to amlitelimab withdrawal, which was 32 weeks after the last administered dose. The lowest dose (62.5 mg Q4W without a loading dose) demonstrated efficacy and safety similar to the highest dose (250 mg Q4W with a 500‐mg loading dose [+LD]), while reductions in AD‐related biomarkers were numerically greater with the 250 mg Q4W + LD regimen throughout the trial [10]. Taken together, the data support the evaluation of extended dosing regimens beyond Q4W with 250 mg amlitelimab. For patients with AD, the development of a treatment that facilitates both a reduced dosing frequency and the maintenance of durable clinical response addresses a significant unmet need.

The objective of the present study was to utilize population pharmacokinetic (PopPK) and PopPK/pharmacodynamic‐EASI (PopPK/PD‐EASI) models to establish extended amlitelimab dosing regimens and scenarios for future phase 3 clinical trials in adolescents (≥ 12 to 17 years old) and adults (≥ 18 years old) with moderate‐to‐severe AD.

2. Methods

2.1. Study Population

Data from five amlitelimab clinical studies were included: three phase 1 trials in healthy volunteers and two phase 2 trials (phase 2a and STREAM‐AD phase 2b) in adults with moderate‐to‐severe AD (Table 1). To be included in the analysis, each participant had to have at least two free amlitelimab serum concentrations above the lower limit of quantification (LLOQ; 0.0469 μg/mL). The PopPK/PD‐EASI model was developed using data from the two phase 2 clinical studies in participants with AD (phases 2a and 2b). EASI assessment is a bounded continuous score that incorporates body surface area and the intensity of lesional skin into one score ranging from 0 to 72 [11], and is commonly used as an efficacy endpoint in phase 2 and phase 3 AD studies. At the time of this analysis, all participants from the phase 2b trial had completed the 24‐week induction period (Part 1); however, only two participants had completed the 28‐week maintenance/withdrawal period (Part 2). All studies were conducted in accordance with Good Clinical Practice guidelines and the Declaration of Helsinki ethical principles. The protocols were approved by the institutional review board or ethics committee for each site, and all participants provided written informed consent.

TABLE 1.

Summary of clinical studies used to develop the PopPK and PopPK/PD‐EASI models.

Study	Phase	Model	Subjects	Number of PK samples	Number of EASI measurements	Route a	Amlitelimab dosing regimen
NCT03161288 (SAD/MAD study)	1	PopPK	48 healthy participants	808	NA	IV	Single dose: 0.006–0.05 mg/kg Multiple doses: 0.15 + 0.075 × 2 to 12.0 + 6.0 × 2 mg/kg Q4W (three doses)
NCT04449939 (absolute bioavailability study)	1	PopPK	24 healthy participants	386	NA	8 IV 16 SC	Single dose: 125 mg and 250 mg
NCT05796479 (drug comparability study)	1	PopPK	6 of 32 healthy participants	84	NA	SC	Single dose: 250 mg
NCT03754309 (proof of concept study)	2a	PopPK; PopPK/PD‐EASI	59 patients with AD	778	320	IV	200 mg loading dose + 100 mg Q4W up to 16 weeks; 500 mg loading dose + 250 mg Q4W up to 16 weeks
NCT05131477 (STREAM‐AD) (dose‐ranging study)	2b	PopPK; PopPK/PD‐EASI	302 patients with AD	2604	2157	SC	62.5 mg; 125 mg; 250 mg Q4W up to 24 weeks or 52 weeks (continue or withdraw); 500 mg loading dose + 250 mg Q4W up to 24 weeks or 52 weeks (continue or withdraw)

Abbreviations: AD, atopic dermatitis; EASI, Eczema Area and Severity Index; IV, intravenous; NA, not applicable; PD, pharmacodynamics; PK, pharmacokinetics; PopPK, population pharmacokinetic; Q4W, every 4 weeks; SAD/MAD, single ascending dose/multiple ascending dose; SC, subcutaneous.

^a IV was given as a 30 min infusion; SC volume of injection was 1 or 2 mL depending on the dose.

2.2. Development of the Models

Both PopPK and PopPK/PD‐EASI analyses were performed using a nonlinear mixed‐effects modeling (NONMEM) computer program (version 7.5.1) running on a Linux cluster of multiprocessor computers [12]. All runs were performed using the first‐order conditional estimation (FOCE) method with interaction option during the development of the models. Initially, the population parameters (fixed and random effects) were computed assuming no dependency between parameters and covariates. For each model, potential covariate effects were then explored using stepwise selection, including forward inclusion (alpha risk: 5%) and backward deletion (alpha risk: 0.1%) processes. Validation of the models was assessed by examination of the goodness‐of‐fit plots and individual concentration versus time curves, by estimation of several quality criteria such as bias, precision, and absolute average fold error. The models were further qualified by ascertaining their predictive ability through prediction‐corrected visual predictive checks (VPC) and assessing the uncertainty in parameter estimates using an iterative sampling importance resampling (SIR) method [13, 14].

2.3. PopPK Model

A two‐compartment structure with linear and nonlinear clearances, including an absorption process described by a first‐order rate for a SC route, a lag time, and a bioavailability factor, was selected based on a preliminary analysis.

Because of the specificity of the dataset, including adults with varying baseline body weights (40.5–148 kg), allometric scaling of linear clearance and volume parameters was privileged and tested before the covariate selection process. Potential PK covariates were investigated based on physiological and clinical relevance. Baseline demographic characteristics were considered, including body weight (except for the previously allometric‐scaled parameters), age, sex, race, renal function (creatinine clearance), and hepatic parameters (albumin, alkaline phosphatase, alanine aminotransferase, aspartate aminotransferase, and total bilirubin). The occurrence of anti‐drug antibodies coded as either binary covariate (at least once or never over study follow‐up) or longitudinal binary covariate was investigated. In addition, the clinical endpoints EASI and vIGA‐AD baselines, as well as patient status (healthy vs. AD) were investigated.

The final PopPK model was used to calculate the individual amlitelimab PK parameter estimates and derived exposure variables (C _max, C _trough, and area under the serum concentration‐time curve over the 4‐week dosing interval [AUC_4W]) for participants with AD treated with Q4W doses corresponding to the STREAM‐AD phase 2b study and to evaluate dosing regimens proposed for the future phase 3 clinical studies.

2.4. PopPK/PD‐EASI Model

For the PopPK/PD‐EASI analysis, direct and indirect response models (type I—inhibition of production)—in which free amlitelimab concentrations exerted an inhibitory effect on EASI production—were explored. The relationship between predicted free amlitelimab serum concentration and EASI was explored using the maximum drug‐induced inhibitory effect (I _max), power, or linear equations. The relationship between the individual estimates and the covariates was then investigated.

Similar to the PopPK model, baseline demographic characteristics—such as body weight, age, sex, race, renal function, hepatic parameters, and occurrence of anti‐drug antibodies—were tested as covariates. Responder status, which was a predefined clinical endpoint, was investigated as an additional covariate.

The final PopPK/PD‐EASI model was used to predict EASI and explore dosing regimens and scenarios in responder participants for future phase 3 clinical studies. The dosing regimens selected were based on the predicted efficacy (expressed in EASI percent change from baseline) across a broad range of exposures from 62.5 mg Q4W to 250 mg Q4W + LD in STREAM‐AD phase 2b, and the predicted maintenance of efficacy during the withdrawal of treatment for several weeks.

2.5. Simulations for Extended Dosing

The final PopPK model was used to compute exposures for the lowest (62.5 mg Q4W) and highest (250 mg Q4W + LD) dosing regimens evaluated in STREAM‐AD phase 2b, as well as exposures for an extended regimen of 250 mg Q12W + LD for a treatment duration of 24 weeks in all participants in STREAM‐AD phase 2b. Three dosing scenarios were simulated using the final PopPK/PD‐EASI model: (i) 62.5 mg Q4W and 250 mg Q4W + LD for 60 weeks as reference, (ii) 250 mg Q4W + LD for 24 weeks followed by Q12W maintenance for 36 additional weeks, and (iii) 250 mg Q12W + LD for 60 weeks. The simulations were performed in 1000 virtual responder participants per scenario. Body weight, EASI, and albumin ranges used for the simulations corresponded to the ranges observed in the participants with AD from the phase 2 studies; i.e., 40–150 kg, 16–72, and 37–55 g/L, respectively.

2.6. Simulations for Body Weight Dose Adjustments

Using the final PopPK model, simulations using different doses and dosing regimens were performed to generate amlitelimab exposures for participants with AD who have a body weight ≥ 25 and < 40 kg (i.e., adolescents with low to average body weight). The generated exposures were required to be within the exposure range observed in adults with AD weighing ≥ 40 kg (range: 40–150 kg) at the selected phase 3 clinical trial dosing regimens. Simulations in 1000 virtual participants with a body weight ≥ 25 and < 40 kg and in 1000 virtual participants with a body weight ≥ 40 kg were performed for 24 weeks after Q4W and Q12W regimens.

3. Results

3.1. Participant Characteristics

A total of 439 participants (78 healthy volunteers; 361 adults with AD) were included in the PopPK analysis (Table S1). The mean ± standard deviation (SD) age was 35.9 ± 13.7 years (range: 18–72 years), and the mean ± SD body weight was 75.5 ± 16.2 kg (range: 40.5–148 kg). Adults with moderate‐to‐severe AD constituted 82% (n = 361) of the total population, of which 84% were from the STREAM‐AD phase 2b trial.

3.2. PopPK Model

Prior to PopPK model selection, from the initial pool of PK samples (excluding the 1.69% values below LLOQ), outlier detection led to the deletion of 82 (1.73%) unrealistic PK samples and the exclusion of one patient with AD from the STREAM‐AD phase 2b trial who had only one concentration time point available at the time of the analysis. In total, 4660 concentration time points from 438 participants were included in the final dataset of the PopPK analysis.

The structure was a two‐compartment model with both a linear (first‐order process) and nonlinear (Michaelis–Menten) clearance from the central compartment. The SC absorption process was described by a first‐order constant of absorption from the SC compartment to the central compartment, a lag time, and a bioavailability factor (Figure 1). No important systematic deviations or major bias in any of the goodness‐of‐fit plots were observed (Figure S1). The performance and quality criteria of the model stratified by route of administration (intravenous [IV] or SC) or subject status (participants with AD or healthy volunteers) were comparable, and the predictive performances using the prediction‐corrected VPC technique and the evaluation of the uncertainties of the estimate values using the SIR approach qualified the model. Final PopPK parameters and their relationships to covariates are shown in Table S2. Significant covariates included body weight (allometric scaling on linear clearance and compartment volumes), EASI (increasing linear clearance), and albuminemia (increasing bioavailability). The quantitative impact of these covariates on exposure parameters in simulated typical participants is summarized in Table S3.

FIGURE 1.

Schematic representation of the final PopPK model and PopPK/PD‐EASI model. CL, clearance; Fsc, bioavailability after subcutaneous injection; IC₅₀, free drug concentration at 50% of maximum inhibitory effect; I _max, maximum drug‐induced inhibitory effect; IV, intravenous; K _in, first‐order production constant; K _out, first‐order dissipation rate constant; PopPK, population pharmacokinetic; PopPK/PD‐EASI, population pharmacokinetic/pharmacodynamic–Eczema Area and Severity Index; SC, subcutaneous.

3.3. PK Profile in Participants With AD

The final PopPK model was used to estimate individual exposure parameters after the first and last (Week 20) dose of the induction period (Part 1) in the STREAM‐AD phase 2b trial across all treatment arms (Figure 2). Amlitelimab exposure in adults with AD after Q4W injections increased in a dose‐proportional manner. An approximate twofold accumulation ratio was observed after Q4W dosing regimens without LD, which was consistent with the amlitelimab PK profile in steady‐state conditions following the third injection. When a LD was administered, steady state was reached after the LD injection with no additional accumulation. Linearity was observed from the lowest dosing regimen (62.5 mg Q4W) tested in participants with AD, consistent with saturation of the OX40L target in the dose range investigated in STREAM‐AD. During this linear window, the mean terminal half‐life was estimated to be 28 days. In a typical adult with AD who was randomized to treatment withdrawal during Part 2 of the STREAM‐AD phase 2b trial, nonlinearity occurred from 1 μg/mL with target‐mediated drug disposition (TMDD), representing approximately 20% of the total clearance. At the LLOQ (0.0469 μg/mL), the amlitelimab PK was nonlinear, with the TMDD representing 66% of the total clearance (Figure 3).

FIGURE 2.

Amlitelimab exposures by dose arm in participants with AD after first dose and last dose following Q4W administration in the STREAM‐AD phase 2b trial (Part 1; bars include 10th, 25th, 50th, 75th, and 90th percentiles, and individual values). AD, atopic dermatitis, AUC_4W, area under the serum concentration‐time curve from time 0 to Week 4 after dosing; C _max, maximum observed serum concentration; LD, loading dose; Q4W, every 4 weeks.

FIGURE 3.

Predicted clearance rates versus free amlitelimab serum concentrations in a typical patient with AD. Baseline body weight: 75 kg; baseline EASI: 27.5; LLOQ = 0.0469 μg/mL. AD, atopic dermatitis, CL, clearance; EASI, Eczema Area and Severity Index; LLOQ, lower limit of quantification.

Body weight, EASI, and albumin were found to influence PK model parameters. In participants with AD from the STREAM‐AD phase 2b trial, body weight effects resulted in a median steady‐state area under the serum concentration‐time curve from time 0 to Week 4 after dosing (AUC_4w) that was 42% to 58% higher across the four different doses in participants weighing < 74 kg versus ≥ 74 kg (Figure S2A). For EASI, the median steady‐state AUC_4w was at most 32% higher across the four different doses in participants with moderate AD (EASI ≤21) versus severe AD (EASI >21; Figure S2B). In contrast to those covariates, albumin was found to have a negligible increasing impact on exposure. Taken together, using the final PopPK model, body weight was identified as the main covariate impacting the amlitelimab exposures.

3.4. PopPK/PD‐EASI Model

Amlitelimab concentrations generated by the PopPK model were used to develop a PopPK/PD‐EASI model that could predict variations in EASI at any time. One patient was excluded because of limited data available (EASI data only for baseline and up to the first 4 weeks of treatment at the time of the analysis; i.e., number of available EASI data points ≤ 3). The PopPK/PD‐EASI model was developed using 2477 EASI evaluations collected from 269 participants (48 from phase 2a and 221 from STREAM‐AD), including data from the induction period (in all responder and nonresponder participants) and maintenance/withdrawal period (i.e., up to 32 weeks after the last dose) for the 61% of participants on amlitelimab who were clinical responders. At the time of the analysis, while two participants had completed the 28‐week maintenance/withdrawal period, 102 responders were on withdrawal for at least 12 weeks, as shown in Table S4. Baseline characteristics of all participants included in the model are shown in Table S5. The relationship between EASI and free amlitelimab concentration was best characterized by an indirect response model (type I—loss of production) in which amlitelimab concentration exerted an inhibitory effect on EASI production rate (Figure 1). The model was parameterized with K _out, the first‐order rate constant of EASI decrease; I _max, maximum drug‐induced inhibitory effect; and IC₅₀, the amlitelimab concentration at 50% of maximum drug inhibitory effect.

Model parameters and their relationships to covariates are shown in Table S6; only responder status was identified to have a significant effect on both I _max and K _out parameters. As with the PopPK model, no important systematic deviations or major bias in any of the goodness‐of‐fit plots were observed (Figure S3). The final model successfully predicted the change in EASI over time in responder and nonresponder participants (Figure 4).

FIGURE 4.

Predicted changes from baseline in EASI for (A) responder participants receiving amlitelimab from Week 0 and during the withdrawal period, (B) responder participants receiving amlitelimab from Week 0 and then not receiving amlitelimab during the withdrawal period, and (C) nonresponder participants in STREAM‐AD receiving amlitelimab from Weeks 0 to 24, and not entering the withdrawal period. Data presented as mean (dot) ± SD (error bars) of observed data. Dashed line represents mean of predicted responses. EASI, Eczema Area and Severity Index; LD, loading dose; SC, subcutaneous.

3.5. Simulations for Extended Dosing

Using the final PopPK/PD‐EASI model, simulated efficacy for the dosing scenario of 250 mg Q4W + LD for 24 weeks followed by Q12W as maintenance therapy for an additional 36 weeks was similar to that of the 250 mg Q4W + LD regimen given for 60 weeks (reference) (Figure 5a). The simulated efficacy of a 250 mg Q12W + LD extended dosing regimen from initiation for 60 weeks was comparable with that of either the 62.5 mg Q4W or the 250 mg Q4W + LD regimens evaluated in STREAM‐AD (Figure 5b). Using the final PopPK model, exposures for all participants in STREAM‐AD were computed with the extended dosing regimen of 250 mg + LD Q12W, with PK parameter estimates of C _max, AUC_12w, and C _trough falling between those for the lowest dose (62.5 mg Q4W) and the highest dose (250 mg Q4W + LD) (Figure 5c and Table S7).

FIGURE 5.

Simulated extended dosing regimen scenarios performed in 1000 virtual responder participants per scenario. (A) PopPK/PD‐EASI model‐predicted percent changes from baseline in EASI in responder participants using a 250 mg Q4W + LD for 24 weeks followed by the extended dosing regimen of 250 mg Q12W (last injection Week 48) in the range of 250 mg Q4W + LD (last injection Week 56; shaded areas include 10th to 90th percentiles, and lines represent the median). (B) PopPK/PD‐EASI model‐predicted percent change from baseline in EASI in responder participants for an extended dosing regimen of 250 mg Q12W + LD from initiation (last injection Week 48) in the range of the highest and lowest Q4W dosing regimens from STREAM‐AD (250 mg Q4W + LD and 62.5 mg Q4W; last injection Week 56; shaded areas include 10th to 90th percentiles, and lines represent the median). (C) Simulated exposures for an extended dosing regimen of 250 mg Q12W + LD from initiation and the lowest and highest dosing regimens in STREAM‐AD (62.5 mg Q4W and 250 mg Q4W + LD) simulated in all STREAM‐AD participants using the final PopPK model (bars include 10th, 25th, 50th, 75th, and 90th percentiles, and individual values). AUC_12W, area under the serum concentration‐time curve from time 0 to Week 12 after dosing; AUC_12‐24W, area under the serum concentration‐time curve from Week 12 to Week 24 after dosing; EASI, Eczema Area and Severity Index; LD, loading dose; PopPK, population pharmacokinetic; PopPK/PD‐EASI, population pharmacokinetic/pharmacodynamic–Eczema Area and Severity Index; Q4W, every 4 weeks; Q12W, every 12 weeks.

Overall, the PopPK/PD‐EASI modeling revealed that extended dosing regimens of amlitelimab at either 250 mg Q12W from initiation or 250 mg Q12W as maintenance therapy produce predicted reductions in EASI similar to the 250 mg Q4W regimen.

3.6. Exposure Simulations in Adolescents With AD

Individuals weighing < 40 kg are anticipated to have a 1.7‐ to 2.0‐fold (35–40 kg and 25–30 kg, respectively) higher median amlitelimab exposure (AUC_4W) than adults ≥ 40 kg at the same amlitelimab dosing regimen of 250 mg Q4W + LD, with 25% to 50% of participants above the target exposures (data not shown). Simulations were performed based on the final PopPK model to generate amlitelimab exposures in participants weighing < 40 kg (range: 25–40 kg). The objective was to identify dosing regimens that achieve exposures within the exposure range observed in adults with AD weighing ≥ 40 kg (range: 40–150 kg) at the selected dosing regimens of 250 mg Q4W + LD and 250 mg Q12W + LD. Simulations using the dosing regimens of a 250‐mg loading dose followed by 125 mg Q4W (Figure 6) or 125 mg Q12W (a twofold reduction in dosing) yielded amlitelimab exposures in participants weighing 25–40 kg that were comparable with those observed in adults weighing 40–150 kg.

FIGURE 6.

Predicted amlitelimab exposures by body weight band for adolescents < 40 kg (range: 25 to < 40 kg) compared with adults ≥ 40 kg based on a twofold dose reduction approach for Q4W dosing regimen. AUC_4W, area under the serum concentration‐time curve from time 0 to Week 4 after dosing; C _max, maximum observed serum concentration; LD, loading dose; Q4W, every 4 weeks.

4. Discussion

In the present study, the PK profile of amlitelimab was characterized in healthy volunteers and participants with moderate‐to‐severe AD using a PopPK model developed with free amlitelimab serum concentrations collected on treatment and during treatment withdrawal from several studies.

The PopPK model was utilized to compute the individual exposure parameters across all Q4W dose arms in the STREAM‐AD phase 2b trial, demonstrating that exposure increased in a dose‐proportional manner consistent with a 100% target engagement in the mean concentration range documented in STREAM‐AD (≥ 10 μg/mL) [10]. Nonlinearity occurred from 1 μg/mL with TMDD demonstrating a relevant contribution in the total clearance. This finding is consistent with the 1–2 μg/mL threshold for nonlinearity observed in healthy volunteers [9]. The difference in PK observed between healthy volunteers and participants with AD could be explained by a lower target expression in healthy volunteers. However, data on OX40L target were not available to well characterize the relationship between target occupancy and nonlinear clearance, and the difference between healthy and AD populations, representing a limitation of the PopPK model. From the PopPK model, body weight was identified as the main covariate affecting amlitelimab PK, which is frequently reported to affect monoclonal antibody concentrations [15]. An increasing effect of albuminemia on SC bioavailability of amlitelimab was identified, with a negligible impact on exposure. Interestingly, no clear difference in serum albumin levels was observed between healthy volunteers and participants with AD in this dataset (Table S3), whereas albumin has been reported as a significant covariate for the clearance of many monoclonal antibodies [16, 17]. This is an intriguing finding in the context of a chronic inflammatory skin disease such as AD and should be confirmed in future analyses with larger cohorts of individuals with AD receiving amlitelimab.

The amlitelimab concentrations generated by the PopPK model were used to develop the PopPK/PD‐EASI model to predict EASI improvement over time in participants with AD. Unbalanced data were collected between the induction period (clinical responders and nonresponders) and the maintenance/withdrawal period (clinical responders only). To better characterize responder and nonresponder participants, a mixture model commonly used was investigated (not shown here) and compared with the model that evaluated the responder status covariate. The covariate model outperformed the mixture model, showing a 120‐unit difference in the objective function. Furthermore, only 73% of participants were correctly classified by the mixture model in accordance with the clinical criteria as they were defined in the study protocols. In addition, the prediction‐corrected VPC clearly displayed a bias for the maintenance/withdrawal period in the mixture model, impairing its predictive ability in responder participants. As a result, the responder status covariate model was selected as the preferred approach to robustly simulate EASI responses under alternative dosing regimens of amlitelimab. Both the PopPK and PopPK/PD‐EASI models selected were fully qualified, and their predictive performance was validated.

Simulations for an extended dosing regimen of 250 mg Q12W + LD using the PopPK model predicted exposures between the range observed for the lowest (62.5 mg Q4W) and highest (250 mg Q4W + LD) doses in the STREAM‐AD phase 2b trial. Furthermore, the PopPK/PD‐EASI analysis revealed that extended dosing regimens of amlitelimab at 250 mg Q12W from initiation or 250 mg Q12W as maintenance therapy produced predicted reductions in EASI similar to what was observed in the STREAM‐AD phase 2b trial, which demonstrated all Q4W doses to be efficacious with an acceptable safety profile [10]. Although the PopPK/PD‐EASI predictions are limited to 32 weeks off therapy because of the available data from STREAM‐AD, the simulations from the PopPK and PopPK/PD‐EASI models support a 250 mg Q12W amlitelimab regimen. Based on these findings, both a continuous 250 mg Q12W regimen and a 250 mg Q4W regimen for 24 weeks followed by a Q12W regimen are being investigated in the phase 3 program. In Part 2 of the STREAM‐AD phase 2b trial, the majority of clinical responders maintained their clinical response following amlitelimab withdrawal for up to 32 weeks [10]. This further supports the viability of an extended Q12W dose. Severity of the disease and the time spent managing symptoms greatly contribute to the burden of the disease and quality‐of‐life impairment [18, 19]. Therefore, an extended dosing regimen that is efficacious would represent a substantial improvement in the treatment burden for patients with moderate‐to‐severe AD.

The STREAM‐AD phase 2b trial included only adults with moderate‐to‐severe AD. The inclusion of adolescents with AD in the phase 3 clinical trials would provide information to understand the risk–benefit profile of amlitelimab treatment in this patient population. Body weight was identified as the main covariate in the PopPK model; therefore, it was essential to determine if any dose adjustments were required for adolescents in the phase 3 clinical trial. The present analysis also examined simulated exposures in individuals with body weights from 25 to < 40 kg. Simulations revealed that the use of a twofold dose reduction in adolescents (body weight from 25 to < 40 kg) would result in amlitelimab exposures similar to those observed in adults ≥ 40 kg. The results support a regimen of 125 mg Q4W or 125 mg Q12W with a 250‐mg LD for adolescents weighing < 40 kg, which should have efficacy similar to that observed in the adult population based on exposure. For adolescents weighing ≥ 40 kg, the dosing regimens used in adults are appropriate, as the results show a positive risk–benefit profile with no exposure‐response relationship across a body weight range of 40.5 to 148 kg in the STREAM‐AD trial.

In conclusion, the PopPK and PopPK/PD EASI models successfully described the observed free amlitelimab concentrations and EASI. More specifically, the PopPK/PD EASI model successfully predicted EASI over 24 weeks in clinical responder and nonresponder participants, and up to Week 52 in responder participants for maintenance/withdrawal periods. Simulated exposure and efficacy data support the intended use of the extended regimen of Q12W with LD, both at initiation and after 24 weeks of Q4W dosing of amlitelimab, in the large phase 3 clinical trials in adolescents and adults with moderate‐to‐severe AD.

Author Contributions

G.T., K.P., T.B., S.W., L.B., C.‐H.L., J.T.O., K.Y., C.B., D.F., and F.H. wrote the manuscript. G.T., D.F., and F.H. designed the research. G.T., D.F., and F.H. performed the research. G.T., D.F., and F.H. analyzed the data.

Conflicts of Interest

Kim Papp has received honoraria and/or grants from, is a consultant, an investigator, or a scientific officer for AbbVie, Acelyrin, Akros, Alumis, Amgen, Arcutis, Bausch Health/Valeant, Boehringer Ingelheim, Bristol Myers Squibb, Can‐Fite Biopharma, Celltrion, Concert Pharmaceuticals, Dermavant, Dermira, DICE Pharmaceuticals, DICE Therapeutics, Eli Lilly and Company, Evelo Biosciences, Forbion, Galderma, Horizon Therapeutics, Incyte Corporation, Janssen, Kymab, Kyowa Hakko Kirin, LEO Pharma, Meiji Seika Pharma, Mitsubishi Pharma, Nimbus Therapeutics, Novartis, Pfizer, Reistone, Sanofi‐Aventis/Genzyme, Sandoz, Sun Pharma, Takeda, Tarsus Pharmaceuticals, UCB Pharma, and Zai Lab. Thomas Bieber is a speaker, consultant, and/or investigator for AbbVie, Affibody, Almirall, Amagma, AnaptysBio, AOBiom, Anergis, Apogee, Arena, Aristea, Artax, Asana Biosciences, ASLAN Pharma, Astria, Attovia, Bayer Health, Biofilm Control, BioVerSys, Boehringer Ingelheim, Bristol Myers Squibb, BYOME Labs, Connect Pharma, Daiichi Sankyo, Dermavant, DICE Therapeutics, Domain Therapeutics, DS Pharma, EQRx, Galderma, Galapagos, Glenmark, GSK, Incyte, Innovaderm, Janssen, Kirin, Kymab, LEO, LG Chem, Lilly, L'Oréal, MSD, Medac, Micreos, Nektar, Novartis, Numab, OM‐Pharma, Overtone, Pfizer, Pierre Fabre, Q32bio, RAPT, SamsungBioepis, Sanofi/Regeneron, TIRmed, UCB, Union Therapeutics, Upstream Bio, and Yuhan. He is the founder and chairman of the board of the nonprofit biotech “Davos Biosciences AG” within the international Kühne‐Foundation and founder of the consulting firm Bieber Dermatology Consulting. Stephan Weidinger has received institutional research grants from LEO Pharma, Pfizer Inc., and Sanofi Deutschland GmbH; has performed consultancies for AbbVie, Almirall, Boehringer Ingelheim, Eli Lilly and Company, Galderma, Kymab (a Sanofi company), LEO Pharma, Novartis, Pfizer Inc., Regeneron Pharmaceuticals, and Sanofi Genzyme; has lectured at educational events sponsored by AbbVie, Almirall, Galderma, LEO Pharma, Novartis, Pfizer Inc., Regeneron Pharmaceuticals, and Sanofi Genzyme; and is involved in performing clinical trials with many pharmaceutical companies that manufacture drugs used for the treatment of psoriasis and atopic dermatitis. Lisa Beck is a consultant for Allakos, Amgen, Arcutis, Arena Pharmaceuticals, AstraZeneca, Astria Therapeutics, Evelo Biosciences, Escient Pharma, Galderma, Incyte, Invea Therapeutics, Janssen, LEO Pharma, Merck, Nektar Therapeutics, Novartis, Numab Therapeutics, Pfizer, Rapt Therapeutics, Regeneron Pharmaceuticals Inc., Ribon Therapeutics, Sanofi‐Aventis/Genzyme, Sitryx Therapeutics, Stealth BioTherapeutics, Trevi Therapeutics, UCB Pharma, Union Therapeutics, and Xencor; and is an investigator for AbbVie, AstraZeneca, Pfizer, Regeneron Pharmaceuticals Inc., and Sanofi. Chih‐Hung Lee is a consultant/speaker for AbbVie, Eli Lilly, Janssen, Kyowa Kirin, LEO Pharma, Novartis, Pfizer, Sanofi, and Tanabe. Gilles Tiraboschi, John T. O'Malley, Karl Yen, Charlotte Bernigaud, David Fabre, and Fabrice Hurbin are/were employees of Sanofi and may hold stock and/or stock options in the company.

Supporting information

Data S1: psp470121‐sup‐0001‐Supinfo01.docx.

Data S2: psp470121‐sup‐0002‐Supinfo02.pdf.

Data Availability Statement

Qualified researchers may request access to patient‐level data and related study documents including the clinical study report, study protocol with any amendments, blank case report form, statistical analysis plan, and dataset specifications. Patient‐level data will be anonymized, and study documents will be redacted to protect the privacy of our trial participants. Further details on Sanofi's data‐sharing criteria, eligible studies, and the process for requesting access can be found at: https://www.vivli.org/.