Development of the Weight‐Tiered Q3W Regimen of Amivantamab in Combination With Carboplatin‐Pemetrexed in EGFR‐Mutated Non‐Small Cell Lung Cancer

ABSTRACT

Amivantamab, a bispecific antibody targeting epidermal growth factor and mesenchymal‐epithelial transition receptors, was initially approved under accelerated approval as a monotherapy for second‐line treatment of non‐small cell lung cancer (NSCLC) with exon 20 insertion mutations. Subsequently, amivantamab in combination with carboplatin‐pemetrexed demonstrated prolonged progression‐free survival compared to chemotherapy alone in the first‐line treatment of NSCLC with exon 20 insertion mutations and the second‐line treatment of NSCLC with exon 19 deletion or exon 21 L858R substitution mutations. To enable the combination with chemotherapy, the amivantamab regimen was changed from every 2 weeks (Q2W) to every 3 weeks (Q3W); both are weight‐tiered with a cutoff at 80 kg. Here we present the population pharmacokinetics (PK) and exposure‐response (E‐R) analysis supporting this dose regimen change. The population PK analysis demonstrated that the Q3W regimen achieved comparable trough concentrations as the approved Q2W regimen as well as between the two weight tiers. After accounting for confounding covariates, the E‐R relationship for PFS was flat in both indications. The safety profile of the Q3W regimen was generally manageable, despite trends of a positive E‐R relationship for adverse effects related to the mechanism of action. Besides weight, dose adjustment for other covariates was not warranted. These analyses supported the registration of the weight‐tiered Q3W regimen for amivantamab in combination with carboplatin‐pemetrexed in the new indications. This work demonstrates how the use of fit‐for‐purpose pharmacometrics analyses can support dose regimen changes without extensive dose selection clinical studies.

Study Highlights

• What is the current knowledge on the topic?

  • ○
    Amivantamab, an EGFR‐MET bispecific antibody, was first approved as an IV single agent in patients with Ex20ins mutated NSCLC. Amivantamab IV exhibited linear PK over a dose range of 350–1750 mg, with no dose‐limiting toxicities. The maximum tolerated dose was not identified in CHRYSALIS. The initial regimen approved for amivantamab is 1050 mg (1400 mg for patients ≥ 80 kg) weekly for 5 weeks and Q2W thereafter starting at week 7.
• What question did this study address?

  • ○
    To enable the combination with carboplatin‐pemetrexed, a Q3W amivantamab IV regimen was developed. How can a quantitative clinical pharmacology approach aid in the development and approval of the amivantamab IV Q3W regimen in EGFR mutated NSCLC?
• What does this study add to our knowledge?

  • ○
    The observed clinical PK, safety, and efficacy along with findings from the population PK and exposure‐response analysis suggest that a Q3W dosing regimen provided comparable amivantamab exposure to the approved Q2W regimen and demonstrated a favorable risk–benefit profile in EGFR mutated NSCLC.
• How might this change clinical pharmacology or translational science?

  • ○
    This study showcases how a quantitative clinical pharmacology approach can be taken to support a different dosing regimen and its successful development in clinical trials.

1. Introduction

Amivantamab is a fully human, bispecific antibody with immune cell‐directed activity, designed to target epidermal growth factor receptor (EGFR) and mesenchymal‐epithelial transition factor (MET) [1, 2]. The simultaneous inhibition of both pathways via ligand blocking, receptor downregulation as well as immune cell‐mediated cell killing contributed to the antitumor activity of amivantamab [3]. Amivantamab demonstrated robust and durable responses when administered as an intravenous (IV) monotherapy once every 2 weeks (Q2W) in the phase 1 CHRYSALIS study, prompting its accelerated approval as a monotherapy in patients with EGFR exon 20 insertions (Ex20ins) mutated non‐small cell lung cancer (NSCLC) whose disease had progressed on or after chemotherapy [2, 4]. The approved Q2W dose regimen is 1050 mg (1400 mg for patients ≥ 80 kg) weekly for 5 weeks and then Q2W thereafter starting at Week 7.

The combination of amivantamab with carboplatin‐pemetrexed (ACP) was initially evaluated in CHRYSALIS, demonstrating early antitumor activity and a tolerable safety profile [5]. To enable the combination with the once every 3 weeks (Q3W) carboplatin‐pemetrexed (CP) schedule, the regimen of amivantamab was changed from Q2W in monotherapy to Q3W in ACP with a higher dose for each weight tier, namely 1400 mg (1750 mg for patients ≥ 80 kg) weekly for the first 4 weeks and Q3W at 1750 mg (2100 mg for patients ≥ 80 kg) starting at Week 7. Each cycle was 21 days. Later, ACP was evaluated in two Phase 3 studies. In the phase 3 PAPILLON study, ACP significantly prolonged progression‐free survival (PFS) compared with CP alone (11.4 months vs. 6.3 months; p < 0.001) in participants with treatment‐naïve EGFR Ex20ins NSCLC [2, 6]. Based on these data, the accelerated approval of amivantamab monotherapy as a second‐line treatment for EGFR Ex20ins NSCLC was converted to full approval, and ACP was approved as a first‐line treatment for EGFR Ex20ins NSCLC. In the phase 3 MARIPOSA‐2 study, ACP significantly prolonged PFS versus CP in participants with EGFR exon 19 deletions (Ex19del) or exon 21 L858R substitution (L858R) mutated NSCLC after disease progression on osimertinib (6.3 months vs. 4.2 months; p < 0.001) [7]. Based on these data, ACP was also approved as a second‐line treatment for EGFR Ex19del or L858R NSCLC after disease progression on EGFR tyrosine kinase inhibitor (TKI) [2].

We previously reported the determination of the amivantamab monotherapy Q2W regimen in EGFR Ex20ins NSCLC as a second‐line therapy [8]. Here, we report the observed pharmacokinetics (PK) from CHRYSALIS ACP cohort which demonstrated that the amivantamab Q3W regimen achieved comparable trough concentrations as the initially approved Q2W regimen. We further report the population PK, exposure‐efficacy, and exposure‐safety analyses based on PAPILLON and MARIPOSA‐2 to support the registration of the amivantamab Q3W regimen in combination with CP in EGFR Ex20ins NSCLC as frontline therapy and in EGFR Ex19del or L858R NSCLC as a second‐line therapy.

2. Materials and Methods

2.1. Study Design

Details of the CHRYSALIS (NCT02609776), PAPILLON (NCT04538664), and MARIPOSA‐2 (NCT04988295) studies have been previously described and summarized in Table S1 [4, 6, 7]. Across all studies, participants included in the ACP arm received amivantamab IV at 1400 mg (1750 mg for patients ≥ 80 kg) weekly for the first 4 weeks and Q3W at 1750 mg (2100 mg for patients ≥ 80 kg) starting at Week 7. Consistent with the approved monotherapy Q2W regimen, the first weekly dose of amivantamab was split between 2 days to reduce the risk of infusion‐related reactions (IRR): 350 mg on Cycle 1 Day 1 and the remaining dose on Cycle 1 Day 2. IV pemetrexed (500 mg/m²) and carboplatin (area under the concentration‐time curve 5 mg/mL/min, maximum 750 mg) were administered on Day 1 of each 21‐day cycle. Carboplatin treatment was discontinued after 4 cycles.

All studies were conducted in accordance with current International Council for Harmonization guidelines on Good Clinical Practice, consistent with the principles of the Declaration of Helsinki, and applicable regulatory and country‐specific requirements. Each study was approved by an Independent Ethics Committee, and all participants provided written informed consent.

2.2. Amivantamab PK Quantification

The PK sampling timepoints in each study are described in Table S1. A validated, specific, and sensitive Meso Scale Discovery electrochemiluminescence immunoassay was used to measure amivantamab concentrations in serum.

2.3. Descriptive Analysis of PK Data

Descriptive analysis was performed on PK data obtained from CHRYSALIS, PAPILLON, and MARIPOSA‐2 participants who received amivantamab as monotherapy or in combination with CP. The serum trough concentration (C_trough) at the end of weekly doses (predose concentration on Cycle 2 Day 1, C_{trough, max}) and steady state (C_{trough, ss}) were summarized.

2.4. Population PK Analyses

The population PK analyses for PAPILLON and MARIPOSA‐2 were conducted separately once the data became available. Nonlinear mixed effects modeling software (NONMEM; version 7.4.0 and 7.5.0, ICON plc, Hanover, MD, USA) and the first‐order conditional estimation with interaction (FOCEI) option were used.

A 2‐compartment model with linear clearance was previously published to describe the amivantamab PK data from CHRYSALIS participants who received amivantamab monotherapy in the dose range of 140–1750 mg IV Q2W [8]. In this publication, parallel linear and nonlinear (i.e., Michaelis–Menten) clearance (Figure S1) was attempted to capture the faster clearance at the lowest dose (140 mg, only 3 participants), but this model was numerically unstable and therefore not selected. After a longer follow‐up of PK data from CHRYSALIS, a 2‐compartment model with parallel linear and Michaelis–Menten clearance converged successfully and performed better than the linear model (internal data). This model was used as the starting point for the present analysis and hereafter referred to as the prior CHRYSALIS model. The model was parameterized in terms of linear clearance (CL), volume of distribution in the central compartment (V₁) and in the peripheral compartment (V₂), intercompartmental clearance (Q), maximum velocity of the nonlinear clearance (V_max), and Michaelis–Menten constant (K_m). This model included baseline weight (WT) as a covariate on CL and V₁, and sex as a covariate on CL. Interindividual variability (IIV) was quantified for CL, V₁, V₂, and V_max. The residual unexplained variability (RUV) model was additive on the logarithmic scale (equivalent to a proportional model on the linear scale).

Initially, a Bayesian approach (MAXEVAL = 0) was applied to characterize the PK data from PAPILLON. The prior CHRYSALIS model was suboptimal; trends of covariate relationships were observed in ETA~covariate plots. Therefore, additional refinement of the model was pursued based on the pooled PK data from CHRYSALIS and PAPILLON. In total, 16,321 measurable amivantamab concentrations from 639 participants were included in this model update. Among them, 413 participants were from CHRYSALIS who received amivantamab monotherapy using the Q2W regimen (PK cutoff of 26 February 2021), 20 participants were from CHRYSALIS who received ACP (PK cutoff of 15 November 2021), 151 participants were from PAPILLON who received ACP, and 55 participants were from PAPILLON who received amivantamab monotherapy using the Q3W regimen in crossover. The PK cutoff for PAPILLON was 09 February 2023. Since the percentage of posttreatment samples below the limit of quantification (BLQ) was low (0.8%), the BLQ samples were omitted (M1 method) [9]. Baseline demographics (age, WT, sex, race), baseline serum albumin, and combination with chemotherapy (yes or no) were explored as covariates. Additional covariate relationships based on graphical analysis were added to the prior model (including existing covariates) to establish a full model. A backward elimination step was conducted to remove nonsignificant relationships (existing and newly added covariates) based on the objective function (OFV) value using a log‐likelihood ratio test with a p‐value of 0.001. Graphical analysis was repeated to confirm no residual covariate relationship. NONMEM control stream of the final model was provided in Supporting Information.

Subsequently, the updated model based on CHRYSALIS and PAPILLON data was used to characterize the amivantamab PK data from MARIPOSA‐2 using a Bayesian approach (MAXEVAL = 0). In total, 1106 measurable amivantamab concentrations from 123 MARIPOSA‐2 ACP participants (PK cutoff of 31 May 2023) were included.

2.5. PK Simulations

PK simulations were conducted to evaluate if the selected Q3W regimen provided comparable amivantamab exposure relative to the initially approved Q2W regimen. Additionally, simulated amivantamab exposure was compared across demographic and clinically relevant subgroups in forest plots.

Unless otherwise specified, the simulations were performed using the posthoc individual parameter estimates from the updated population PK model to reflect exposure in the target population and retain the correlations between covariates, such as weight and sex. Nominal doses were assumed. The first full dose was split across 2 days with the first infusion of 350 mg on Day 1 and the remaining portion of the dose on Day 2, following the clinical study design. Each dose was administered using a 2‐h IV infusion.

2.6. Exposure‐Efficacy and ‐Safety Analyses

Exposure‐response (E‐R) analyses for efficacy and safety were conducted in patients who received ACP in PAPILLON and MARIPOSA‐2. Due to the difference in patient population, the analyses were conducted separately for each study.

The E‐R efficacy analyses explored the relationship between simulated exposure and the primary efficacy endpoint PFS (using RECIST version 1.1 guidelines) assessed by blinded independent central review (BICR). Consistent with the previous analyses [8], the exposure metrics included C_{trough, 1st} (the trough concentration after the first full dose), C_{trough, max}, and C_{avg, 1st cycle} (the average concentration in the first treatment cycle). These early PK metrics, simulated based on patients' actual dosing records using posthoc parameter estimates from the updated PK model, were selected to reduce the number of participants excluded from E‐R analyses. Steady‐state PK metrics could be simulated for all participants assuming nominal dosing; however, not all participants could achieve steady‐state before dropout or disease progression; therefore, early PK metrics were chosen over steady‐state PK metrics. Kaplan–Meier plots and Cox proportional hazard (PH) models were used to evaluate the impact of exposure or covariates on PFS. Covariates evaluated in the E‐R efficacy analysis included age, WT, sex, race, ECOG, history of brain metastases, history of smoking, and EGFR mutation type (Ex19del vs. L858R for MARIPOSA‐2 only). The proportional hazard assumption was checked using Schoenfeld residuals test; none of the exposure metrics and covariates violated this assumption (p > 0.05) except EGFR mutation type in MARIPOSA‐2. Stratified COX‐PH modeling by EGFR mutation type was conducted to evaluate the impact of amivantamab exposure on PFS in MARIPOSA‐2. Example R codes for E‐R efficacy analyses were provided in Supporting Information.

The exposure‐safety analyses assessed the relationship between simulated C_{eoi, 1st} (end‐of‐infusion concentration after the first dose on Cycle 1 Day 1) and IRR. Adverse events (AEs) such as rash, paronychia, hypoalbuminemia, nausea, and constipation are related to the mechanism of action (MOA) of amivantamab. Thus, the correlation between these event rates with simulated C_{eoi, max} (maximum end‐of‐infusion concentration, i.e., end‐of‐infusion concentration of the 4th full dose) and C_{avg, 1st cycle} was investigated. The exposure‐safety relationships were explored graphically in R (version 3.6.2). The safety profile of ACP in PAPILLON and MARIPOSA‐2 was generally consistent with the approved amivantamab monotherapy and considered manageable and tolerable [4, 6, 7]. Most events were low grade; Grade ≥ 3 event rates for ACP were low [6, 7]: 11%, 7%, 4%, 1%, 0% in PAPILLON and 6%, 2%, 2%, 1%, 1% in MARIPOSA‐2 for rash, paronychia, hypoalbuminemia, nausea, and constipation, respectively. Rates of AEs leading to amivantamab dose reduction (36% in PAPILLON and 15% in MARIPOSA‐2) and discontinuation (7% in PAPILLON and 15% in MARIPOSA‐2) were low [6, 7]. Therefore, further statistical modeling of the exposure‐safety relationships for dose optimization was not pursued.

3. Results

3.1. Population PK Model Development

Demographics and baseline characteristics of the participants included in the population PK analyses stratified by study are summarized in Table S2.

The updated model based on pooled CHRYSALIS and PAPILLON data retained the structure model of the prior CHRYSALIS model, namely a 2‐compartment model with parallel linear clearance (CL) and Michaelis–Menten elimination (V_max and K_m) (Figure S1). In addition to WT on CL and V₁, and sex on CL identified in the prior model, the updated model included these additional covariate relationships: baseline albumin and age on CL, sex on V₁, and baseline WT on V₂ (using the same scaling exponent as on V₁). Other covariates, including race and the combination with chemotherapy (yes or no), were evaluated and had no significant impact on amivantamab PK.

All model parameters (Table S3), with the exception of K_m, were estimated with adequate precision (RSE < 30%). Amivantamab PK profiles were simulated based on the individual PK parameter estimates from the updated model. Based on the simulated PK profiles, both Q2W and Q3W regimens reached steady state by Week 13 (Figure 1). At steady state, the median contribution of the Michaelis–Menten pathway to total clearance was up to 30% for both Q2W and Q3W regimens at the end of the dosing interval (Figure S2). The lower precision of the K_m estimate (RSE 43.3%) could be explained by the limited contribution of the Michaelis–Menten pathway to the total clearance and limited PK data at concentrations around K_m.

FIGURE 1.

Comparison of simulated amivantamab exposure in Q2W versus Q3W regimen. In (A), the solid lines and shaded areas represent the simulated median and 90% prediction interval of simulated amivantamab concentrations; the vertical dotted lines indicate the trough concentration at the end of weekly induction doses C_trough,max (at Week 4 for Q3W and Week 5 for Q2W) or at steady state C_trough,ss (at Week 13 for both Q2W and Q3W). The simulation was conducted based on post hoc PK parameters from PAPILLON PK population.

Overall, the goodness‐of‐fit (GOF) plots (Figure S3) and prediction‐corrected visual predictive check (pcVPC, Figure S4) showed that the updated model adequately describes the central tendency and variability of the pooled amivantamab PK data from CHRYSALIS and PAPILLON. In external validation, this updated model also adequately described the amivantamab PK data from MARIPOSA‐2 (Figure S5).

3.2. Exposure Matching Between the Q2W and Q3W Regimens

The Q3W regimen was selected to match the trough concentrations of the Q2W regimen to ensure efficacy. During the dose escalation of amivantamab monotherapy in CHRYSALIS, no dose‐limiting toxicity was identified. At the selected Q3W dose, the peak exposure was expected to be lower than the highest escalated dose of amivantamab monotherapy in CHRYSALIS, 1750 mg QW for the first 4 weeks and Q2W starting at Week 5. The Q3W dose regimen was initially established to achieve comparable trough concentrations at the end of weekly induction doses (C_{trough, max}) and at steady state (C_{trough, ss}) to the approved Q2W regimen and demonstrated to be tolerable and efficacious, based on a limited number of ACP participants in CHRYSALIS (Table S4). When data from PAPILLON and MARIPOSA‐2 were available, the observed C_{trough, max} and C_{trough, ss} in ACP participants using the Q3W regimen were found to be comparable across studies. Concomitant CP did not affect amivantamab PK, which is consistent with the non‐overlapping metabolism pathways of amivantamab and CP and the covariate analysis. PAPILLON participants who received amivantamab monotherapy in crossover had comparable amivantamab trough concentrations to those who received ACP within the same study.

Consistent with observed results from CHRYSALIS, simulated C_{trough, max} and C_{trough, ss} based on posthocs of PAPILLON PK population (N = 206) were consistent between the Q2W and Q3W regimens (Figure 1). When stratified by weight tier, it appeared that the ≥ 80 kg weight tier had lower C_trough,ss for the Q3W regimen. A similar trend was observed for simulation based on the covariate distribution in PAPILLON PK population and estimated inter‐individual variability (Figure S6).

3.3. PK Exposure Subgroup Analysis

Figure 2 compared the simulated steady state exposure AUC_3weeks,ss (areas under the concentration‐time curve during the 3‐week dose interval at steady state) across subgroups of the PAPILLON participants (N = 206). Forest plot of C_trough,max (the most sensitive PK metrics in exposure‐response efficacy analysis) showed similar trends (Figure S7). The forest plot of AUC_3weeks,ss demonstrated that participants with WT < 80 kg and ≥ 80 kg had comparable exposure under the weight‐tiered Q3W regimen. Within the < 80 kg weight tier, higher WT had lower exposure. The 70–80 kg subgroup had 27% lower AUC_3weeks,ss than the < 60 kg group. Females had 32% higher AUC_3weeks,ss than males.

FIGURE 2.

Forest plot of AUC3weeks,ss in PAPILLON PK population. CI = confidence interval; ECOG = Eastern Cooperative Oncology Group performance status; GMR = geometric mean ratio; N = number of participants; PK = pharmacokinetics; WT = weight. Filled circles and lines represent GMR (90% CI); vertical dashed lines represent 80% (left) and 125% (right).

Other subgroups of clinical interest, including race (Asian vs. non‐Asian, Japanese vs. non‐Japanese, Chinese vs. non‐Chinese), mild or moderate renal impairment, mild hepatic impairment, history of brain metastases (presence or absence), ECOG performance status (0, ≥ 1), and other identified PK covariates (baseline serum albumin and age), all appeared to have comparable exposure, with the point estimate of geometric mean ratio (GMR) within 0.8–1.25.

Trends in the forest plot based on MARIPOSA‐2 PK population (N = 123) were consistent (Figure S8). The amivantamab exposure was consistent in Ex19del vs. L858R.

3.4. Exposure‐Response Efficacy and Safety Analysis in PAPILLON

Both efficacy and safety analyses included 151 PAPILLON patients who received ACP and had exposure estimates from the population PK analysis.

No significant differences in PFS were detected across the exposure tertiles of all 3 exposure metrics using log‐rank tests (p = 0.159, 0.056, and 0.094 for C_{trough, 1st}, C_{trough, max}, and C_{avg, 1st cycle}, respectively). The Kaplan–Meier plot of PFS by tertiles of C_{trough, max} (the PK metrics with the lowest p value) is shown in Figure 3. PFS appeared to be slightly better in the 3rd C_{trough, max} exposure tertile; however, the confidence intervals were overlapping with the lower exposure tertiles.

FIGURE 3.

Kaplan–Meier plot of PFS by BICR in ACP participants in PAPILLON and MARIPOSA‐2. ACP = amivantamab in combination with carboplatin‐pemetrexed; BICR = blinded independent central review; PFS = progression‐free survival; T = tertile. Shaded regions indicate 95% confidence interval.

Potential influence of covariates on the observed exposure‐response (E‐R) relationship of PFS was explored. The demographics and baseline characteristics of participants stratified by exposure tertiles are presented in Table S5. Consistent with the forest plot showing higher exposure in females, the proportion of female participants was higher in the 3rd tertile (41%, 42%, 84% females in 1st, 2nd, and 3rd tertiles, respectively). Females have been reported to have a reduced risk of death and benefit more from EGFR TKIs [10]. Brain metastasis was also shown to negatively influence PFS of patients with EGFR‐mutated NSCLC treated with first and second generations of EGFR TKIs [11]. Based on the PAPILLON data, sex and history of brain metastases were identified as significant covariates on PFS in univariable log‐rank screening tests (p = 0.015 and p < 0.001, respectively). Consistent with published reports [10, 11], females or absence of brain metastases were associated with better PFS (Figures S9 and S10). After accounting for the confounding effect of sex and brain metastases, further inclusion of C_{trough, max} did not significantly improve the model projection of PFS (Table S7). Stepwise covariate search starting from a COX‐PH multivariable model including all the covariates listed in Table S6 and C_trough,max (not allowed to drop) resulted in a final model of sex, brain metastases, and C_trough,max. The 95% confidence interval for the coefficient estimate of C_trough,max (Table S8) included zero, further confirming that there was no E‐R relationship between amivantamab exposure and PFS.

In addition, increasing amivantamab exposure did not associate with better PFS in males who had lower exposure and relatively worse PFS outcome than females (Figure S11). Within the < 80 kg weight tier, despite lower exposure in participants approaching 80 kg, PFS in 70 to 80 kg subgroup were comparable to PFS in < 60 kg subgroup (Figure S12). The ≥ 80 kg weight tier had comparable PFS as other weight groups (Figure S12), suggesting that the lower C_trough,ss for ≥ 80 kg in Q3W regimen relative to the approved Q2W regimen (Figure 1) was not clinically meaningful.

The E‐R relationships between amivantamab exposure and safety endpoints were explored graphically in ACP participants from PAPILLON. Participants with higher amivantamab C_{eoi, 1st} appeared to have less IRR (Figure 4). This counterintuitive observation could be explained by higher end of infusion concentration in participants who did not experience IRR and therefore less dose interruption of amivantamab infusion. Trends of increasing incidence of paronychia, constipation, and hypoalbuminemia with the increase of amivantamab exposure were observed (Figure 5).

FIGURE 4.

Bar plots of infusion related reaction rate stratified by amivantamab exposure quartile in ACP participants in PAPILLON and MARIPOSA‐2. ACP = amivantamab in combination with carboplatin‐pemetrexed; C_{eoi, 1st} = end of infusion concentration after the first dose (Cycle 1 Day 1 dose). Each bar represents the event rate per exposure quartile. The error bars represent the 95% CI for the event rate, calculated by the Wilson method. The numbers below the bars represent the number of participants with event over the total number of participants in the exposure quartile. A square bracket indicates the exposure is included in the quartile and a round bracket indicates the exposure is not included in the quartile.

FIGURE 5.

Bar plots of rash, paronychia, hypoalbuminemia, nausea, and constipation rate stratified by amivantamab exposure quartiles in PAPILLON ACP participants. ACP = amivantamab in combination with carboplatin‐pemetrexed; C_{avg,1st cycle} = average concentration in the first cycle; C_eoi,max, maximum end of infusion concentration; CI = confidence interval. Each bar represents the event rate per exposure quartile. The error bars represent the 95% CI for the event rate, calculated by the Wilson method. The numbers above the bars represent the number of participants with event over the total number of participants in the exposure quartile. A square bracket indicates the exposure is included in the quartile and a round bracket indicates the exposure is not included in the quartile.

3.5. Exposure‐Response Efficacy and Safety Analysis in MARIPOSA‐2

Both efficacy and safety analyses included 123 MARIPOSA‐2 participants who received ACP and had exposure estimates from the population PK analysis. The demographics and baseline characteristics of participants stratified by exposure tertiles are presented in Table S9.

No significant differences in PFS were detected across the exposure tertiles of C_{trough, 1st} and C_{avg, 1st cycle} (p = 0.22 for both metrics in log‐rank tests). Although there was a significant relationship between PFS and C_{trough, max} tertiles in the log‐rank test, it was the 2nd tertile (not the 3rd tertile with the highest exposure) that had the best PFS (Figure 3). Further univariable Cox‐PH analysis using the PK metrics as a continuous variable supported that C_{trough, max} did not affect PFS (p = 0.47). The impact of covariates on PFS was screened using log‐rank tests (Table S10). PFS was comparable in males versus females (Figure S14) and in 70–80 kg versus < 60 kg (Figure S15) despite lower exposure. The covariate that appeared to correlate with PFS was EGFR mutation type. Ex19del appeared to have better PFS than L858R mutation in MARIPOSA‐2 (Figure S13), consistent with reported outcomes of EGFR TKIs [12, 13]. After accounting for the impact of EGFR mutation type on PFS in multivariable Cox‐PH modeling, none of the PK metrics had a significant impact on PFS (Table S11). Due to the violation of the proportional hazard assumption for EGFR mutation type (p = 0.02), the impact of amivantamab exposure on PFS was further evaluated in COX‐PH modeling stratified by EGFR mutation type. Consistently with the previous multivariable COX‐PH modeling where EGFR mutation type was evaluated as a covariate, none of the amivantamab exposure metrics showed a significant impact on PFS (Table S12).

The E‐R relationships between amivantamab exposure and safety endpoints were explored graphically in ACP participants from MARIPOSA‐2. Consistent with PAPILLON, an apparent negative E‐R relationship was observed for IRR (Figure 4), and the incidence of hypoalbuminemia trended higher with the increase of amivantamab exposure (Figure 6).

FIGURE 6.

Bar plots of rash, paronychia, hypoalbuminemia, nausea, and constipation rate stratified by amivantamab exposure quartiles in MARIPOSA‐2 ACP participants. ACP = amivantamab in combination with carboplatin‐pemetrexed; C_{avg,1st cycle} = average concentration in the first cycle; C_eoi,max, maximum end of infusion concentration; CI = confidence interval. Each bar represents the event rate per exposure quartile. The error bars represent the 95% CI for the event rate, calculated by the Wilson method. The numbers above the bars represent the number of participants with event over the total number of participants in the exposure quartile. A square bracket indicates the exposure is included in the quartile and a round bracket indicates the exposure is not included in the quartile.

4. Discussion

Amivantamab IV was initially approved as a monotherapy for the second‐line treatment of NSCLC with Ex20ins mutated NSCLC using a weight‐tiered Q2W regimen. This Q2W regimen was selected based on the totality of PK, PD, safety, and efficacy data obtained from CHRYSALIS participants who received amivantamab monotherapy ranging from 140 to 1750 mg Q2W [8]. The approved Q2W dose falls in the linear range of amivantamab PK (350–1750 mg) and maintains complete saturation of soluble EGFR and MET during the dose interval [8]. Due to the impact of weight on amivantamab PK, a weight tier‐based Q2W dose regimen (higher dose for ≥ 80 kg) was implemented to reduce PK variability and provide comparable exposure for patients ≥ 80 kg [8]. The approved Q2W regimen demonstrated an objective response rate (ORR) of 40% with a durable response and a tolerable safety profile [8]. Further increasing the dose was not projected to improve ORR significantly, suggesting that the selected Q2W dose was at the plateaued response range for efficacy [8].

Subsequently, ACP was studied as the frontline treatment of NSCLC with Ex20ins mutations in PAPILLON and as the second‐line treatment of NSCLC with Ex19del or L858R mutations in MARIPOSA‐2. Developing a Q3W dose regimen that supports combining with the CP chemotherapy regimen was crucial for amivantamab's clinical utility and to improve overall patient convenience. Thus, a 2‐weight‐tier Q3W regimen was developed, matching the exposure of the approved Q2W regimen by increasing the dose for each weight tier (adding a vial to the weekly loading dose and 2 vials to the Q3W maintenance dose; each vial has 350 mg). This Q3W regimen was initially explored in 20 CHRYSALIS participants, demonstrating comparable observed trough concentrations to the approved Q2W regimen and acceptable tolerability. Both PAPILLON and MARIPOSA‐2 demonstrated that ACP, at the selected Q3W regimen, prolonged PFS compared to CP alone with a manageable and tolerable safety profile [6, 7]. The PFS benefit of ACP over CP was consistent across subgroups of age, sex, weight (< 80 vs. ≥ 80 kg), race, history of brain metastases, history of smoking, ECOG, and EGFR mutation type (Ex19del vs. L858R in MARIPOSA‐2) [6, 7].

Complementing the efficacy and safety data, E‐R efficacy and safety analyses supported a positive benefit–risk profile of the amivantamab Q3W regimen in combination with CP for each indication. In both PAPILLON and MARIPOSA‐2, after adjusting for confounding factors, no significant E‐R relationship was observed between amivantamab exposure and PFS, indicating that the response likely approaches a maximum at the chosen Q3W dose regimen. Trends of higher incidences of hypoalbuminemia, paronychia, and constipation with higher exposure of amivantamab were observed; these AEs were related to the MOA of amivantamab. However, the safety profile of ACP was generally consistent with amivantamab monotherapy and considered manageable and tolerable [4, 6, 7]. The AEs related to amivantamab were mostly of low grade, with low rate of AEs leading to dose reduction (36% in PAPILLON and 15% in MARIPOSA‐2) and discontinuation (7% in PAPILLON and 15% in MARIPOSA‐2) of amivantamab [6, 7].

The weight‐tiered Q3W regimen with a weight cutoff at 80 kg was supported by the population PK analysis and exposure‐efficacy analysis. Population PK simulations supported that the Q3W regimen provided comparable trough concentrations (C_trough,max and C_trough,ss) as the approved Q2W regimen. The 80 kg weight cutoff was predetermined from the Q2W regimen and further confirmed in the Q3W regimen. When stratified by the weight tier, one of the PK metrics, C_trough,ss for the ≥ 80 kg weight tier was lower than but overlapping with the Q2W regimen. A higher Q3W dose for the ≥ 80 kg weight tier was not warranted because comparable exposure and PFS were achieved between the two weight tiers [6, 7], and a flat E‐R relationship between amivantamab exposure and PFS was observed. For both PAPILLON and MARIPOSA‐2, within the < 80 kg weight tier, the 70–80 kg subgroup had lower exposure than the < 60 kg subgroup but comparable PFS, so lowering the weight cutoff was not warranted.

The population PK analysis revealed that weight, sex, baseline serum albumin, and age significantly affect amivantamab PK. Aside from weight‐tiered dosing, dose adjustments based on sex, baseline serum albumin, and age were not warranted. In PAPILLON, males had lower exposures and appeared to have worse PFS outcomes than females; however, a concern of inadequate exposure in males was not valid for the following reasons. PFS was not significantly correlated with exposure after accounting for the confounding effect of sex and brain metastases, and within males, higher amivantamab exposure was not associated with better PFS outcome. Overall, male participants who received ACP had better PFS than male participants who received CP [6]. Therefore, dose adjustment for males is not warranted for the PAPILLON population. In MARIPOSA‐2, males had similar PFS as females, and therefore no dose adjustment for males was warranted. In both populations, the other 2 PK covariates (baseline serum albumin and age) did not result in clinically relevant differences in amivantamab exposure (GMR within 0.8–1.25).

Taken together, the population PK and E‐R efficacy analyses presented here and the clinical efficacy and safety data supported the weight‐tiered Q3W regimen with a weight cutoff at 80 kg for amivantamab in combination with CP in the first line treatment of NSCLC with Ex20ins mutations or the second line treatment of NSCLC with Ex19del or L858R mutations.

One limitation of this work is that amivantamab was studied as an add‐on treatment to CP in PAPILLON and MARIPOSA‐2, where only one dose regimen of amivantamab was studied. However, the Q3W regimen was selected via PK bridging to the approved Q2W regimen for amivantamab monotherapy where the benefit–risk profile was well established through dose escalation. Despite trends of a positive E‐R relationship were observed for the AEs related to amivantamab's MOA, the safety profile of the studied Q3W dose in combination with CP was generally manageable, and the rate of AEs leading to dose modification of amivantamab was low in each population, so the current amivantamab Q3W dose is an appropriate dose.

Author Contributions

All authors wrote the manuscript. D.Y., Y.S., M.S., W.Z., and J.Z. designed the research. All authors performed the research and analyzed the data.

Funding

The studies and publication of this work are sponsored by Johnson & Johnson.

Conflicts of Interest

All authors are employees of Johnson & Johnson during the conduct of the research and may hold company stock.

Supporting information

Table S1: Study design and PK sampling.

Table S2: Demographics and baseline characteristics of CHRYSALIS, PAPILLON, and MARIPOSA‐2 participants included in population PK analysis.

Table S3: Parameter estimates of the amivantamab population PK model.

Table S4: Observed amivantamab trough concentrations in CHRYSALIS, PAPILLON, and MARIPOSA‐2.

Table S5: Demographics and baseline characteristics of PAPILLON ACP participants by amivantamab C_{trough, max} tertile.

Table S6: Summary of log‐rank tests for PFS stratified by covariates in PAPILLON ACP participants.

Table S7: Summary of Cox‐PH regression tests for PFS in PAPILLON ACP participants.

Table S8: Parameter estimates of the final Cox‐PH model for PFS in PAPILLON ACP participants.

Table S9: Demographics and baseline characteristics of MARIPOSA‐2 ACP participants by amivantamab C_{trough, max} tertile.

Table S10: Summary of log‐rank tests for PFS stratified by covariates in MARIPOSA‐2 ACP participants.

Table S11: Multivariable COX‐PH models for PFS in MARIPOSA‐2 ACP participants.

Table S12: Stratified COX‐PH models by EGFR Mutation Type for PFS in MARIPOSA‐2 ACP participants.

Figure S1: Amivantamab population PK model.

Figure S2: Contribution of nonlinear clearance at steady state.

Figure S3: Goodness of fit plots for the updated amivantamab population PK model based on CHRYSALIS (EDI1001) and PAPILLON (NSC3001).

Figure S4: Prediction‐corrected visual predictive check of the updated amivantamab population PK Model based on PK data from CHRYSALIS (EDI1001) and PAPILLON (NSC3001).

Figure S5: Prediction‐corrected visual predictive check of the updated amivantamab population PK Model based on PK data from MARIPOSA‐2 (NSC3002).

Figure S6: Comparison of simulated amivantamab exposure in Q2W vs. Q3W regimen based on the final model and estimated inter‐individual variability.

Figure S7: Forest plot of Ctrough,max in PAPILLON PK population.

Figure S8: Forest plot of AUC3weeks,ss in MARIPOSA‐2 PK population.

Figure S9: Kaplan–Meier plot of PFS by BICR stratified by sex in PAPILLON ACP participants.

Figure S10: Kaplan–Meier plot of PFS by BICR stratified by presence of brain metastases in PAPILLON ACP participants.

Figure S11: Kaplan–Meier plot of PFS by BICR stratified by exposure in male participants from PAPILLON who received ACP.

Figure S12: Kaplan–Meier plot of PFS by BICR stratified by weight category in PAPILLON ACP participants.

Figure S13: Kaplan–Meier plot of PFS by BICR stratified by mutation type in MARIPOSA‐2 ACP participants.

Figure S14: Kaplan–Meier plot of PFS by BICR stratified by sex in MARIPOSA‐2 ACP participants.

Figure S15: Kaplan–Meier plot of PFS by BICR stratified by weight category in MARIPOSA‐2 ACP participants.