Covariate effects and population pharmacokinetic analysis of the anti-FGFR2b antibody bemarituzumab in patients from phase 1 to phase 2 trials

Purpose A population pharmacokinetic (PK) analysis of the anti-fibroblast growth factor receptor 2b antibody, bemarituzumab, was performed to evaluate the impact of covariates on the PK and assess whether dose adjustment is necessary for a future phase 3 trial. Methods Serum concentration data were obtained from three clinical trials, with 1552 bemarituzumab serum samples from 173 patients, and were analyzed using nonlinear mixed-effects modeling. Results A two-compartment model with parallel linear and nonlinear (Michaelis–Menten) elimination from the central compartment best described the bemarituzumab serum concentration data. The final model estimated a typical linear clearance (CL) of 0.311 L/day, volume of distribution in the central compartment (Vc) of 3.58 L, distribution clearance (Q) of 0.952 L/day, volume of distribution in the peripheral compartment (Vp) of 2.71 L, maximum drug elimination by nonlinear clearance (Vmax) of 2.80 μg/day, and Michaelis–Menten constant (Km) of 4.45 μg/mL. Baseline body weight, baseline albumin, gender, and chemotherapy were identified as statistically significant covariates on the PK of bemarituzumab. Given the low interindividual variability of bemarituzumab key PK parameters (CL and Vc) and the small or modest effect of all statistically significant covariates on bemarituzumab exposure at steady-state, no covariate is expected to have clinically meaningful effects on bemarituzumab exposure. Conclusion No covariate had a clinically meaningful impact on bemarituzumab exposure. These results indicate that dose adjustment of bemarituzumab is not necessary, based on the aforementioned covariates, for a future phase 3 trial in gastric and gastroesophageal junction adenocarcinoma population with FGFR2b overexpression in combination with mFOLFOX6. Supplementary Information The online version contains supplementary material available at 10.1007/s00280-021-04333-y.


Introduction
Bemarituzumab (FPA144) is a first-in-class, recombinant, humanized, afucosylated immunoglobulin G1 (IgG1) kappa monoclonal antibody (mAb) directed against fibroblast growth factor receptor 2 IIIb (FGFR2b) and is being developed by Five Prime Therapeutics as a therapeutic for cancer. Bemarituzumab has two mechanisms of action: blocking the FGFR2b signaling pathway and enhancing antibody-dependent cell-mediated cytotoxicity (ADCC) against FGFR2boverexpressing tumors.
Bemarituzumab has been evaluated in three clinical trials {FPA144-001, FPA144-002, and FPA144-004 [FIGHT]}. FPA144-001 (NCT02318329), a phase 1, open-label, dosefinding trial evaluating the safety and pharmacokinetics (PK) of FPA144 in patients with advanced solid tumors including gastric and gastroesophageal junction adenocarcinoma (GEA) at six dose levels ranging from 0.3 to 15 mg/kg once every 2 weeks (Q2W), demonstrated evidence of monotherapy activity and acceptable tolerability of bemarituzumab in patients with FGFR2b overexpressing GEA [1]. FPA144-002, a phase 1, open-label, dose-finding trial evaluating the safety and PK of bemarituzumab in Japanese patients with advanced GEA at two dose levels, demonstrated that bemarituzumab can be administered safely to Japanese patients in doses up to 15 mg/kg Q2W. Since GEA tends to be highly heterogeneous within the same tumor, and when present, FGFR2b may not be uniformly distributed throughout the tumor specimen [2,3], it was projected that combining bemarituzumab with chemotherapeutic agents was likely to improve the clinical benefit over bemarituzumab alone. Therefore, a phase 2, randomized, double-blind, controlled trial evaluating the combination of bemarituzumab with modified FOLFOX6 (mFOLFOX6; leucovorin calcium, fluorouracil, and oxaliplatin) vs placebo with mFOLFOX6 in patients with previously untreated advanced GEA (NCT03343301 and NCT03694522, FIGHT) was conducted. The FIGHT trial using bemarituzumab at 15 mg/kg Q2W with 1 additional dose of 7.5 mg/kg on Cycle 1 Day 8 achieved clinically meaningful and statistically significant improvements across all three of its prespecified efficacy endpoints including objective response rate (ORR), progression-free survival (PFS), and overall survival (OS) in patients with FGFR2b-positive, non-HER2 positive frontline advanced GEA [4].
A population PK (popPK) analysis was originally conducted based on phase 1 FPA144-001 trial alone [5] to support the selection of the dose and schedule for bemarituzumab in the phase 2 FIGHT trial. However, the data available from the phase 1 FPA144-001 trial alone were insufficient for identifying key covariate relationships, such as combination with chemotherapy mFOLFOX6 (combotherapy), tumor type [gastric cancer (GC) vs gastroesophageal junction adenocarcinoma (GEJ)], prior gastrectomy, administration of a single dose of mFOLFOX6 prior to randomization, geographic region, race, and Japanese ethnicity. With the success of the FIGHT trial in patients with GEA, a new popPK analysis using PK data from three trials described above was conducted to evaluate the effects of available covariates on the PK of bemarituzumab in support of a future phase 3 trial in patients with GEA.
The objectives of the current analysis were to develop a popPK model for bemarituzumab to characterize the PK, understand the types of covariates and their magnitude effects on PK parameters, and assess whether dose adjustment for 15 mg/kg Q2W with 1 additional dose of 7.5 mg/kg on Cycle 1 Day 8 is necessary for a future phase 3 trial in combination with mFOLFOX6.

Bemarituzumab serum concentration assay in humans
Bemarituzumab serum concentrations in humans were quantitatively measured at ICON Laboratory Services, Inc. (Whitesboro, NY) with a validated ELISA [5].

Anti-bemarituzumab antibody assay in humans
The anti-drug antibody levels in study FPA144-001 were determined by a validated bridging electrochemiluminescence assay that utilized Meso Scale Discovery (MSD) technology [5]. The second method was an ELISA with affinity capture and elution pre-purification step to measure samples from studies FPA144-002 and FIGHT. The assay sensitivity was 30.2 ng/mL relative to the rabbit anti-bemarituzumab positive control. Drug tolerance was greater than 200 μg/ mL in the presence of 800 ng/mL of rabbit anti-bemarituzumab antibody. Anti-bemarituzumab levels were measured at ICON Laboratory Services, Inc. (Whitesboro, NY).

Study design and pharmacokinetic population
The popPK analysis was conducted using data from three clinical studies: FPA144-001, FPA144-002, and FIGHT (Supplemental Table 1). All studies were conducted in accordance with the Declaration of Helsinki and Guidelines for Good Clinical Practice; approval from Institutional Review Boards or Independent Ethics Committees was obtained for each trial [6,7]. Written informed consent was obtained from all trial participants.
Bemarituzumab was administered in a 30 min intravenous (IV) infusion at doses ranging from 0.3 to 15 mg/kg Q2W, with most patients receiving a dose of 15 mg/kg Q2W with or without 1 additional dose of 7.5 mg/kg on Cycle 1 Day 8. Patients were evaluable for PK analysis if they had received at least one dose of bemarituzumab and a corresponding PK sample collection after drug administration. Observations below the lower limit of quantitation (LLOQ) at 0.125 μg/ mL were omitted from the analysis.

Base model and random-effects model development
Based on known PK properties of bemarituzumab, the default structural model was a two-compartment model with parallel linear and nonlinear (Michaelis-Menten) elimination pathways from the central compartment using the differential equations below [5]: where A c and A p are the amounts of drug in central and peripheral compartments, respectively. V max represents the maximum drug elimination by nonlinear clearance, and K m , the Michaelis-Menten constant, indicates the drug concentration at 50% V max . CL and Q represent linear clearance and distribution clearance, respectively, while V c and V p represent volume of distribution in the central and peripheral compartments, respectively.
Alternative model structures (eg, two-compartment with time-varying clearance) were also explored, as appropriate. No time-varying CL was identified after bemarituzumab administration. The final base model was chosen based on the objective function value (OFV), goodness-of-fit plots, and reliability of model parameter estimates.
Assuming a log-normal distribution, the interindividual variability (IIV) in PK parameters was described by an exponential model: where θ i is the parameter for the i th subject, θ T is natural logarithm of the typical value of the parameter in the population, and ƞ i (ETA) is a random interindividual effect with mean 0 and variance ɷ 2 . The ɷ values are the diagonal elements of the IIV-covariance matrix (Ω), which was initially modeled as diagonal (DIAG option in the NM-TRAN $OMEGA record); thus, assuming no covariance between the random effects. A non-diagonal Ω matrix was finally implemented using the BLOCK option to estimate correlation between CL and V c based on the model fitness as well as ETA correlation results.
Residual error was described using an additive error model after log-transformation of the PK data: where y ij and ŷ ij represent the jth observed and predicted serum concentration, respectively, for the i th subject, and ε is the random residual effect, which is normally distributed with mean 0 and variance σ 2 . Other residual error models were explored if patterns were observed in the individual weighted residual (IWRES) versus individual predicted value (IPRED) plot.

Covariate model development
Following base model development, covariates likely to impact bemarituzumab PK were explored for a possible correlation with key bemarituzumab post hoc PK parameters ( Table 1). These covariates were selected based on physiological plausibility, clinical relevance, availability of data, as well as prior knowledge of similar compounds [5,12].
Given the previously known effect of body weight on the clearance and volume of other antibodies and bemarituzumab [5,12], the effect of body weight on PK parameters was tested first as part of the base model development. Significant body weight effects were incorporated into the base model based on improvements in model fit. Once the base model was developed, covariate screening was conducted by examining correlations between all other covariates and relevant PK parameters graphically, followed by linear regression (for continuous covariates) and analysis of variance (ANOVA) testing (for categorical covariates) using R. These analyses were conducted using the individual empirical Bayesian estimates (EBEs) of interindividual random effects of PK parameters (ETA values) obtained from the base model. Only those covariates that showed a significant (p < 0.05) correlation with the relevant PK parameters that could be meaningfully explained from both a clinical and scientific perspective were examined further in covariate modelling using NONMEM.
Continuous covariates were modeled using a power model as described in the following equation: Categorical covariates were modeled using the general equation: where θ i is the individual parameter value for the ith subject, θ T is the natural logarithm of the typical value of the parameter in the population, and ƞ i is an interindividual random effect with mean of zero and variance ɷ 2 . For a continuous covariate, Cov i is the individual covariate value for the i th subject, Cov pop is the population median or reference value of the covariate, and k cov is the coefficient describing the strength of the covariate effect on the parameter. For a categorical covariate, X i is the individual categorical covariate indicator for the i th subject (whose possible values are indicated by j), and k cov,j is the coefficient describing the strength of the covariate effect for category j (which is zero for the reference category).
Selection of the final covariate model (final popPK model) was determined for its significance based on the likelihood ratio test at the p < 0.01 level for forward inclusion and p < 0.001 for backward deletion.
pcVPC was used to visually assess the ability of the model to reproduce both the central tendency and variability of observed data over time [13]. A total of 1000 replicates of the original popPK dataset were simulated using observed covariates and dose regimens for each subject, the final model parameter estimates, subject-specific random effects, and residual error. The simulation-based 95% confidence interval (CI) (calculated as the 2.5th-97.5th percentile of the 1000 simulated trials) of the predicted median, 2.5th, and 97.5th percentiles of the concentration-time profiles and the corresponding observed data, normalized based on the median population prediction of each time bin, were overlaid to assess whether the model predictions can capture the observed median and spread of the concentration-time profiles.

Sensitivity analyses
The sensitivity analysis was performed for the final popPK model to examine the influence of statistically significant covariates on the predicted exposures including area under concentration-time curve at steady-state (AUC ss ), maximum serum concentration at steady-state (C max,ss ), and trough concentration at steady-state (C troug,ss ) after the target dose of 15 mg/kg Q2W for 52 weeks with 1 additional dose of 7.5 mg/kg bemarituzumab on Cycle 1 Day 8. Tornado plots were generated for different scenarios (10th and 90th values of continuous covariates or possible group of categorical covariates) to show the influence of each covariate on expected exposure compared with the reference value (the predicted exposure in a typical male patient on monotherapy with a body weight of 64 kg and albumin of 38 g/L after the target dose).

Population PK analysis dataset
The PK analysis dataset included 1552 bemarituzumab serum concentration-time data points from 173 patients (

Base model development and covariate assessment
Based on prior popPK analyses of bemarituzumab [5], a two-compartment model with parallel linear and nonlinear (Michaelis-Menten) elimination from the central compartment was chosen as the starting structural model. After the base model structure was established, the model was rerun after excluding outlier data points with absolute conditional weighted residuals (|CWRES|) > 5. Subsequently, given the previously known effect of body weight on the clearance and volumes of other antibodies and bemarituzumab [5,12], the effects of baseline body weight on CL, V c , Q, and V p were examined. Based on an examination of PK parameter-covariate relationships, gender, combotherapy/study, aspartate aminotransferase, albumin, a single dose of mFOLFOX6 prior to randomization, hepatic function based on NCI-ODWG, ECOG performance status, lactate dehydrogenase, total bilirubin on CL, gender, and combotherapy/study on V c had a statistical significance of p < 0.05 and were thus carried forward to the forward covariate search in NONMEM. In addition, tumor type (GC/GEJ/other) was also carried forward to the forward covariate search on both CL and V c for further evaluation. Screening of other covariates showed that age, race, eGFR based on Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [18], alkaline phosphatase, alanine aminotransferase, total protein, tumor size, prior gastrectomy, tumor type, FGFR2b status, and geographic region were not statistically significant covariates on bemarituzumab CL and V c .

Final population PK model
The covariate modelling process began with univariate testing, where each covariate effect that was found to be significant in the covariate screening was added to the base model one at a time. Testing of the covariates one at a time using a stepwise forward addition method in NONMEM showed that the effect of albumin and combotherapy/study on CL, and gender on V c , were significant (p < 0.01). The full popPK model included all significant covariate relationships. Covariates were then excluded from the full popPK model one at a time using a stepwise backward elimination method. The criterion for retention was a change in likelihood ratio > 10.83 for one parameter (p < 0.001). No covariate was removed in the backward elimination process.
The parameter-covariate relations in the final popPK model are described by the following equations: where CL i and V ci represent the linear clearance from the central compartment and volume of distribution of the central compartment of the ith individual; ƞ CL,i and ƞ Vc,i are the interindividual random effects of CL and V c of the ith individual; WT i , ALB i , Gender i , and Therapy i represent the body weight, albumin, gender, and combotherapy/study of the i th individual, respectively.
Parameter estimates for the final popPK model for bemarituzumab are presented in Table 2. The typical values of CL, V c , V p , Q, V max , and K m were precisely estimated, as evidenced by the small relative standard error (RSE) values (< 13%) and narrow confidence intervals from bootstrapping. The estimated coefficients of body weight, albumin, gender, and combotherapy/study effects were generally estimated with adequate precision (%RSE ranging from 6.46 to 23.2%). The IIVs and residual error were also well estimated, with %RSE ranging from 6.5 to 10.7% for IIVs and 5.19% for residual error.
For a typical male patient on monotherapy with a body weight of 64 kg and albumin of 38 g/L, the estimated CL was 0.311 L/day, V c was 3.58 L, Q was 0.952 L/day, V p was 2.71 L, V max was 2.80 μg/day, and K m was 4.45 μg/mL. The estimated linear clearance half-life was 14.9 days. IIVs on CL, V c , and V p were 29.2%, 14.9%, and 60.4%, respectively. The η-shrinkage for CL and V c was low (13.7% and 15.7%, respectively), suggesting their EBEs could be used to accurately describe the relationships between CL or V c and the relevant covariates. The larger IIV (97.4%) and relatively higher η-shrinkage (44.6%) for V max were expected given most patients were treated in the linear dose range. The ε-shrinkage for the residual error (ε) was 14.3%. Based on the estimated coefficients of covariate effects in the final model, a 10% decrease in body weight resulted in 7.06% decrease in CL and 3.81% decrease in V c (e.g., a 10% decrease from 64 kg resulted in CL of 0.289 L/day and V c of 3.45 L). A 10% decrease in albumin resulted in 7.17% General goodness-of-fit plots showed good agreement between predicted and observed concentrations of bemarituzumab (Supplemental Fig. 1a), with no apparent bias in the residual plots (Supplemental Fig. 1b). The distribution of IIV is centered at zero and is normally distributed for all parameters (Supplemental Fig. 2). Pairwise correlations between the ETAs showed a slight correlation between CL and V c , which is consistent with the Ω matrix structure of the model (Supplemental Fig. 3). These results confirm that the structural model as well as the IIV and residual error models described the observed data well. The parameter-covariate relationships in the final popPK model, namely the effect of body weight on CL and V c , albumin and combotherapy/ study on CL, and gender on V c are illustrated in Fig. 1. Collectively, these plots indicate that the covariate model accurately described the relationships between individual PK parameters and covariates in the final popPK model. The pcVPC plots (Supplemental Fig. 4) and NPC (data not shown) showed that the final popPK model could adequately reproduce the central tendency and variability of the bemarituzumab serum concentrations across all studies.

Sensitivity analysis
The sensitivity analysis showed that body weight was the most influential covariate on bemarituzumab exposure (Supplemental Fig. 5). Compared with a typical patient with a body weight of 64 kg, patients with body weight in the 10th percentile (45 kg) and 90th percentile (79 kg) of the GEA population were expected to have 13.4-18.0% lower and 7.0-12.1% higher steady-state exposures (AUC ss , C max,ss , and C trough,ss ). Baseline albumin had a modest effect on bemarituzumab exposure, patients with albumin at the 10th percentile (30 g/L) and 90th percentile (44 g/L) of the GEA population were expected to have 5.7-18.9% lower and 3.9-13.2% higher steady-state exposures. Patients administrated with combotherapy were expected to have 18.8% higher AUC ss , 8.6% higher C max,ss , and 28.6% higher C trough,ss compared to patients administered bemarituzumab as monotherapy. Compared Influence of body weight on CL 0.703 (4.00%) 0.695 (6.46%) 0.718 (0.518, 0.987) θ 9 Influence  to males, females were expected to have 0.9% higher AUC ss , 11.3% higher C max,ss , and 3.8% lower C trough,ss . Overall, the differences in exposure due to these significant covariates were within the overall variability of exposure in the GEA population, which was − 39.7% to + 66.0%, − 27.0% to + 45.4% and − 56.2% to + 92.4% for the 5th to 95th percentiles of the population relative to the typical values of AUC ss , C max,ss , and C trough,ss , respectively (Supplemental Fig. 5).
The geometric mean simulated AUC ss , C max,ss and C trough,ss in the lowest (Q1) or highest (Q4) quartile were up to 15.0% lower and 20.9% higher for body weight and 13.7% lower and 16.8% higher for albumin, respectively, compared with those of the GEA population. The impact of gender on V c resulted in 7.87% lower AUC ss , 2.03% lower C max,ss , and 8.68% lower C trough,ss in females (N = 50) compared to with those in males (N = 85), while the impact of combotherapy (FIGHT, N=76) on CL resulted in 37.0% higher AUC ss , 25.1% higher C max,ss , and 64.3% higher C trough,ss compared to monotherapy (FPA144-001 and FPA144-002, N = 59) (Fig. 2).
The predicted steady-state exposure metrics were also analyzed for the covariate subgroups to evaluate the need for dose adjustment in patient subgroups of interest (Fig. 3). The geometric mean simulated AUC ss , C max,ss , and C trough,ss in mild (N = 46) or moderate (N = 8) renal impairment were up to 14.6% lower or 23.8% lower, respectively, compared with those of patients with normal renal function (N = 80). Compared to patients with normal hepatic function (N = 103), patients with mild hepatic impairment (N = 23) were expected to have 17.9% lower AUC ss , 12.4% lower C max,ss , and 27.7% lower C trough,ss . The modest differences in the geometric mean AUC ss , C max,ss , and C trough,ss were predicted across geographic regions (− 10.1% to 19.6%), between ECOG performance status (− 9.99% to + 27.3%), Japanese and non-Japanese (− 1.86% to + 26.4%), and with or without a single dose of mFOLFOX6 prior to randomization (− 7.44% to 24.7%), compared with the GEA population geometric mean (Fig. 3). The relatively small differences in geometric mean simulated AUC ss , C max,ss , and C trough,ss were predicted between elderly patients ≥ 65 years and adults < 65 years (− 0.474% to + 1.54%), White and Asian (− 6.5% to + 16.7%), with or without prior gastrectomy (− 4.44% to + 2.55%), IHC FGFR2b ≥ 10% and FGFR2b 1-9% (− 1.41% to + 1.00%), and GC and GEJ (− 1.95% to + 15.8%), compared with the GEA population geometric mean (Fig. 3). These differences in exposures across covariate subgroups were small compared with the overall variability of exposures in the GEA population (− 27.0% to + 92.4%) (Supplemental Fig. 5).

Discussion
The final popPK model described the bemarituzumab PK data well, as assessed by diagnostic goodness-of-fit plots, individual fits, pcVPC, NPC, shrinkage, and non-parametric bootstrap results. The sensitivity analysis and population simulations support the decision to test the same dose and regimen used in the phase 2 FIGHT trial for a future phase 3 trial in the GEA population with FGFR2b over expression in combination with mFOLFOX6 without any dose adjustment.
Bajaj et al. summarized the most common covariates among 23 mAbs approved for use in at least one oncology indication were baseline body weight and gender on CL and V c as well as baseline albumin on CL [12]. In our current popPK model for bemarituzumab, body weight, baseline albumin, gender, and combotherapy/study were identified as statistically significant covariates which are consistent with what was previously reported, except for the addition of combotherapy/study. Body weight was identified as a significant covariate on CL and V c . Albumin and combotherapy/study were significant covariates on CL, and gender was a significant covariate on V c . In comparison with the popPK analysis reported previously [5], combotherapy/study was identified as a new significant covariate, as the previous popPK analysis did not include any samples from combination treatment or first-line patients with GEA [5]. Other covariates evaluated, including age, race, Japanese ethnicity, geographic region, eGFR, alkaline phosphatase, aspartate aminotransferase, alanine aminotransferase, total bilirubin levels, total protein, hepatic function based on NCI-ODWG, lactate dehydrogenase, ECOG performance status, tumor size, tumor type, FGFR2b status, prior gastrectomy, and administration of a single dose of mFOLFOX6 prior to randomization did not show a statistically significant impact on the PK of bemarituzumab. An immunogenicity impact assessment was not performed since no patients developed anti-bemarituzumab antibodies post-bemarituzumab treatment.
We previously compared the observed serum concentrations from combotherapy to simulated PK profiles from monotherapy and concluded that they are within a similar range [5]. However, considering the small patient population (12 patients), a covariate analysis between combotherapy and monotherapy was not performed in the previous analysis [5]. Seventy-six patients in the first-line setting were treated with bemarituzumab in combination with mFOLFOX6 in the phase 2 FIGHT trial, and this combotherapy/study was the range of data points within 1.5 times the interquartile range. The dashed red horizontal line represents the geometric mean of the GEA population. AUC ss area under curve at steady-state, C max,ss maximum concentration at steady-state, C trough,ss trough concentration at steadystate, n number of patients, Q quartile, GEA gastric and gastroesophageal junction adenocarcinoma identified as a covariate for CL. This finding of combotherapy as a covariate was unexpected because bemarituzumab and mFOLFOX6 have distinct mechanisms of actions, metabolism, and clearance mechanisms. Bemarituzumab is cleared from the body via target-mediated clearance (specific clearance) and proteolytic catabolism (non-specific clearance). Therefore, most likely, the finding resulted from different patient populations between monotherapy (FPA144-001 and FPA144-002) and combotherapy (FIGHT) studies. Studies FPA144-001 and FPA144-002 enrolled later line patients with solid tumor including GEA, and phase 2 of FIGHT enrolled patients with previously untreated advanced first-line patients with FGFR2b overexpressing GEA only except advanced gastrointestinal (GI) tumors for phase 1 patients. New data from future trials will provide more insights for us to understand the observation.
Since bemarituzumab has nonlinear clearance due to target-mediated CL based on the phase 1 dose-escalation study FPA144-001 (0.3-15 mg/kg), an important question is whether the expression level of FGFR2b has any impact on the PK of bemarituzumab. We previously reported that FGFR2b status at baseline (high vs others) in patients with late line GEA using bemarituzumab as monotherapy in the FPA144-001 trial was not a covariate for PK. Results from the FPA144-001 and FIGHT trials could not be combined to evaluate whether FGFR2b status was a covariate for PK due to slight differences in the IHC assays for each trial. Therefore, the data from the FIGHT trial were used to perform the analysis alone and showed that the V max -FGFR2b expression relationship was not significant at p < 0.05 and, therefore, was not carried forward to the forward addition covariate search. This observation could be confounded by the fact that patients enrolled in the FIGHT trial were administered a dose in the linear dose range. Since the target dose was in the linear dose range, this observation supports the fact that expression levels of FGFR2b will not require any dose justification, at least for the target dose tested in the FIGHT trial.
The sensitivity analysis showed that, compared with a typical male patient on monotherapy with a body weight of 64 kg and albumin of 38 g/L, patients with body weight and baseline albumin at the 10th percentile and 90th percentile of the GEA population were expected to have < 20% lower or higher steady-state exposures ( Supplementary  Fig. 5). Patients with combotherapy were expected to have 8.6-28.6% higher steady-state exposures compared to patients with monotherapy, while the effect of gender on bemarituzumab exposure was relatively small (0.9-11.3%).
Population PK simulation based on the final model for the GEA population showed that moderate exposure differences between covariate subgroups examined were all within ± 30% except for combotherapy/ study (Figs. 2, 3). Although the difference between