Model‐Informed Dosing Regimen of Sugemalimab for European Patients With Non‐Small Cell Lung Cancer: Bridging From Asian Clinical Data

ABSTRACT

Model‐informed drug development (MIDD) framework was employed to bridge sugemalimab dosing from an Asian population to European patients with non‐small cell lung cancer. We evaluated whether a fixed dose of 1200 mg every 3 weeks (Q3W) provides adequate exposure for European patients and, if not, which weight threshold and alternative dose would restore pivotal‐trial exposures and projected benefit. A population pharmacokinetic (popPK) model, developed from 1002 subjects (97.6% Asian) across six clinical trials, was externally validated using data from an extended‐interval higher‐dose regimen. Based on the validated popPK model, exposure simulations for high‐weight European patients (80–150 kg) under various dosing scenarios were then compared to exposures in the pivotal Asian study. Results indicated that while the 1200 mg Q3W dose provided adequate exposure for patients weighing up to 115 kg, those weighing 115–150 kg had lower exposures. To match the exposure‐efficacy profile of the pivotal study, a 1500 mg Q3W dose was proposed for this higher‐weight subgroup. Simulations from exposure‐response (ER) models confirmed that the 1500 mg Q3W dose for high‐weight patients would achieve comparable survival probabilities to the 1200 mg Q3W dose in Asian patients. The proposed regimen of 1500 mg Q3W for patients weighing over 115 kg ensures consistent therapeutic exposure, efficacy, and safety across diverse populations. The MIDD strategy for bridging dose regimens, substantiated by this study, enabled regulatory approval by the European Medicines Agency (EMA) and the UK Medicines and Healthcare Products Regulatory Agency (MHRA) without the need for additional dedicated clinical trials.

Study Highlights

• What is the current knowledge on the topic?

  • ○
    Sugemalimab, an anti‐PD‐L1 antibody, is approved in China at a fixed dose of 1200 mg Q3W for NSCLC, based primarily on Asian populations. While fixed‐dose immune‐checkpoint antibodies generally have modest exposure variability and wide therapeutic windows, body weight significantly impacts monoclonal antibody clearance. Heavier European patients might therefore experience substantially lower exposure than observed in Asians, raising concerns about dosing adequacy in this population.
• What question did this study address?

  • ○
    Can the 1200 mg Q3W sugemalimab dose, validated in an all‐Asian pivotal NSCLC study, achieve adequate exposure and clinical efficacy in heavier European patients? If inadequate, at what body‐weight threshold is an increased dose required, and what alternative dose restores optimal exposure?
• What does this study add to our knowledge?

  • ○
    The analysis confirmed that the approved 1200 mg Q3W dose achieves adequate exposure up to 115 kg, but heavier patients require dose adjustment. A dose increase to 1500 mg Q3W for patients weighing over 115 kg and restores exposures to within 80%–125% of pivotal trial levels. Simulated clinical outcomes predicted comparable efficacy between adjusted dosing in heavier patients and standard dosing in Asian populations.
• How might this change clinical pharmacology or translational science?

  • ○
    This work exemplifies the practical application of model‐informed drug development (MIDD) to bridge ethnic and demographic gaps without additional trials, supporting global regulatory decisions. The acceptance by EMA and MHRA highlights growing regulatory confidence in pharmacometric strategies, which could potentially accelerate global access to effective dosing regimens.

1. Introduction

Model‐informed drug development (MIDD) is a progressive approach that integrates mathematical models encompassing compound‐, mechanism‐, and disease‐level data for effective knowledge management, prediction, and decision making in pharmaceutical research and development [1, 2]. MIDD has emerged as a robust quantitative framework that facilitates optimal dose selection, justifies study design, and enables extrapolation to alternative dose regimens and diverse populations [3]. Over the past decade, collaboration between regulatory authorities and the pharmaceutical industry in advancing MIDD approaches has significantly increased [4]. These collaborative efforts have led to the publication of several guidelines [5], which ultimately aim to expedite the delivery of safer and more effective drugs to patients.

Sugemalimab, an anti‐PD‐L1 monoclonal antibody, has been approved by the China National Medical Products Administration for first‐line treatment in combination with chemotherapy for stage IV metastatic non‐small cell lung cancer (NSCLC) [6], following confirmation of its efficacy and safety in the pivotal GEMSTONE‐302 study [7]. Nevertheless, since the confirmatory study was conducted exclusively with Asian patients, concerns have been raised regarding the adequacy of the 1200 mg every 3 weeks (Q3W) dose regimen investigated in the pivotal study for European patients, who generally have higher average body weights than the Asian population [8].

The objective of this research was to assess whether the 1200 mg Q3W dose regimen of sugemalimab is adequate for European patients based on established population models. Additionally, the study aimed to identify a body weight threshold that would necessitate dosage adjustment and to determine the optimal dosage for patients exceeding this threshold. By incorporating MIDD techniques, this investigation sought to provide simulation‐based recommendations for optimizing sugemalimab dosing in European patients with NSCLC.

2. Methods

2.1. Data

The pooled data from two Phase I studies, two Phase II studies, and two Phase III studies were used to construct the population pharmacokinetic (popPK) model with detailed information provided in Table S1. Among the 1002 subjects included in the modeling, the majority (97.5%) received sugemalimab 1200 mg every 3 weeks (Q3W) dose, with most subjects (97.6%) being Asian. Additionally, only 75 subjects (7.5%) weighed more than or equal to 80 kg. In the confirmatory GEMSTONE‐302 study in patients with NSCLC, only 18 subjects (5.6%) out of the 320 patients in the sugemalimab group weighed 80 kg or more. In addition, PK data from 13 Asian colorectal cancer subjects (body weight: 41–95 kg) in the 1800 mg every 4 weeks (Q4W) arm were collected after the completion of the popPK analysis and therefore were not included in the model. We used their PK and relevant covariate data to externally validate the popPK model.

A cohort of 320 patients with stage IV NSCLC who received sugemalimab 1200 mg IV Q3W in the GEMSTONE‐302 study were employed to develop the exposure‐response (ER) models for efficacy. For the exposure‐safety evaluation, a total of 842 subjects from four clinical studies were included.

2.2. Established Population Models and Validation

2.2.1. PopPK Model

The population PK model for sugemalimab was previously identified as a two‐compartment disposition model characterized by time‐dependent elimination with weight, sex, albumin, disease status, and time‐varying ADA titer as statistically significant covariates. The model parameter estimates and their relative standard error were listed in Table S2. Model evaluation plots for the total 1002 patients and for the 75 patients weighing at least 80 kg were presented in Figures S1 and S2, respectively. However, the differences in model‐predicted exposures associated with these covariates were deemed to be not clinically meaningful, given that they fell within the 80%–125% acceptance criteria. The equations describing the typical values of the final model estimation, prior to accounting for interindividual variability, are as follows:

$$\begin{array}{c}\mathrm{C}{\mathrm{L}}_0=0.149\mathrm{L}/\mathrm{day}·{\left(\frac{\text{Weight}}{61.5\mathrm{kg}}\right)}^{0.77}·{\left(\frac{\text{Albumin}}{41.9\mathrm{g}/\mathrm{L}}\right)}^{-0.98}·0.904\hfill \\ \left(\text{if Female}\right)\hfill \end{array}$$

$$\begin{array}{c}\mathrm{C}{\mathrm{L}}_{\mathrm{T}}=0.108\mathrm{L}/\mathrm{day}·\left[1+0.092·\left(\log \left(\text{TITER}\right)-\log \left(0.5\right)\right)\right]\hfill \\ ·e^{-K_{\mathrm{des}}·t}·0.713\left(\text{if Female}\right)\hfill \end{array}$$

$$K_{\mathrm{des}}=\left\{\begin{array}{c}0.0188{\mathrm{day}}^{-1}\left(\text{if TTYPE}\ne \text{Lymphoma}\right)\\ 0.0465{\mathrm{day}}^{-1}\left(\text{if TTYPE}=\text{Lymphoma}\right)\end{array}\right.$$

$$Q=0.373\mathrm{L}/\mathrm{day}$$

$$V_c=3.42\mathrm{L}·{\left(\frac{\text{Weight}}{61.5\mathrm{kg}}\right)}^{0.446}·{\left(\frac{\text{Albumin}}{41.9\mathrm{g}/\mathrm{L}}\right)}^{-0.332}·0.858\left(\text{if Female}\right)$$

$$·0.896\left(\text{if TTYPE}=\text{Lymphoma}\right)$$

$$V_p=0.573\mathrm{L}·{\left(\frac{\text{Albumin}}{41.9\mathrm{g}/\mathrm{L}}\right)}^{-1.4}·3.21\left(\text{if TTYPE}=\text{NSCLC}\right)$$

where “TTYPE” is tumor type and “TITER” is time‐varying ADA titer. The total clearance (CL) is the sum of time‐independent clearance (CL₀) and time‐dependent clearance (CL_T).

2.2.2. Exposure‐Response Model

As sugemalimab exhibited time‐varying clearance, which indicates response‐driven exposure, the exposure metrics from Cycle 1 (AUC_C1, C_trough,C1, and C_max,C1) were used in the previous ER analyses as recommended in the literature to mitigate issues stemming from response driven ER [9, 10].

In the ER analysis for efficacy, both C_trough,C1 and AUC_C1 showed consistent exposure‐response patterns for progression‐free survival (PFS) and overall survival (OS), whereas C_max,C1 did not show a significant association. Among the exposure metrics, C_trough,C1 provided the best improvement in model fit and was therefore selected as the exposure metric for ER models. A modified Gompertz model was employed to characterize the relationship between sugemalimab exposure (C_trough,C1) and PFS, as evaluated by both investigator‐assessed and blinded independent central review (BICR), while an exponential hazard model was applied for OS [11]. All models incorporated a linear exposure effect, indicating that higher levels of C_trough,C1 were associated with improved PFS and OS. A summary of the E‐R models was presented in Table S3, and the visual predictive check (VPC) plots for evaluation were displayed in Figure S3. For each 10 μg/mL increase in C_trough,C1, there was an estimated reduction in the hazard of 9.2% for investigator‐assessed PFS, 12.4% for PFS by BICR, and 20.2% for OS, respectively. It is noteworthy that the OS model may be biased, as 45 patients from the placebo group who experienced disease progression subsequently crossed over to receive at least one dose of sugemalimab, in accordance with the study protocol [7].

No discernible relationship was observed between sugemalimab exposure metrics and overall response (OR). Furthermore, the ER analysis for safety in 842 patients across four studies demonstrated no relationship between sugemalimab exposure and increased risk of treatment‐emergent AEs of interest with the dosage ranging from 3 to 40 mg/kg Q3W, in addition to the dose of 1200 mg Q3W.

2.2.3. External Validation of popPK Model

The popPK model for sugemalimab was developed using data from subjects who predominantly received a dosage of 1200 mg Q3W. Therefore, it was crucial to evaluate the suitability of the model for higher doses, as higher doses may be required for patients with higher body weights in Europe. To this end, data from 13 Asian subjects in the 1800 mg Q4W group were used for external validation of the popPK model. Population and individual predicted concentrations were obtained from post hoc estimations using the $ESTIMATION routine with the MAXEVAL = 0 in NONMEM. Prediction error‐based metrics and VPC were employed to assess model performance [12, 13].

2.3. Simulation‐Based Evaluation for Heavier European NSCLC Patients

2.3.1. PK Simulation for High Weight Subjects Receiving 1200, 1500, and 1800 mg Q3W Dosages

The individual empirical Bayes estimates (EBE) were calculated from the popPK model for all 1002 patients included in the modeling. Simulated body weight values for these patients were randomly drawn from a uniform distribution for each 5 kg stratum in the range of 80–150 kg. Subsequently, individual PK model parameters (CL₀ and V _c ), which are dependent on body weight, were recalculated for the simulated body weight in accordance with the relationship derived from the final popPK model, while holding the effects of other covariates constant.

The re‐estimated PK parameters were then used to simulate concentration‐time profiles on a 1‐h sampling interval from 0 to 504 h for each virtual high‐weight patient under three distinct dosing scenarios (1200, 1500, and 1800 mg Q3W). Exposure metrics, including C_trough, AUC, and C_max for Cycle 1, were estimated for each individual and subsequently summarized for each 5 kg weight stratum. The results were then compared to the model‐predicted exposures from individuals in the pivotal study who received a dosage of 1200 mg Q3W. As η‐shrinkage was relatively high for K_des, CL_T, and V _p (39.5%–49.6%), EBEs for these parameters may attenuate the variability of simulated exposure metrics. To assess robustness, a sensitivity simulation was conducted in which individual η vectors were generated by sampling from the individual conditional distribution rather than using EBEs, while all other simulation settings were unchanged [14].

In addition, the 5th percentile of observed C_trough,C1 in study GEMSTONE‐302 (42.6 μg/mL) was designated as the clinical reference concentration. This served as an anchor for comparing exposures across weight groups. Consequently, the proportion of simulated C_trough,C1 above this reference within each weight stratum was also calculated.

2.3.2. Criteria for Dose Adjustment Based on PK

Refer to the guidelines from the FDA and EMA [15, 16], the adequacy and comparability of exposure for patients with higher body weights at the 1200 mg Q3W dose and the increased doses of 1500 and 1800 mg Q3W were evaluated using the following criteria:

• The geometric mean (90% CI) of AUC_C1, C_max,C1, and C_trough,C1 for patients with higher body weight should not be more than 20% lower at 1200 mg Q3W compared to the geometric mean for patients in the GEMSTONE‐302 study.

• The geometric mean (90% CI) of AUC_C1, C_max,C1 and C_trough,C1 should not exceed 25% higher at 1500 and 1800 mg Q3W doses than the geometric mean for patients in the GEMSTONE‐302 study.

In accordance with these criteria and the proportion of patients above the clinical reference C_trough, the body weight threshold for dose adjustment will be identified and a new dosing regimen for high‐weight patients will be proposed.

2.3.3. Simulation of Survival Probability for the Proposed Dose Adjustment

Following the proposal for dose adjustment based on body weight, 1000 virtual patients were generated for each dosage that fell below or above the identified body weight threshold. The body weights of the virtual patients were sampled from a uniform distribution with a minimum of 80 kg and a maximum equal to the threshold, or alternatively, a minimum equal to the threshold and a maximum of 150 kg. The remaining covariates in the popPK model and ER model were sampled with replacement from subjects in the sugemalimab arm of the GEMSTONE‐302 study. For comparison, 1000 virtual Asian subjects were also created using the same methodology, incorporating their actual body weights into the sampling with replacement.

For each virtual patient, the trough concentration of Cycle 1 was simulated based on individual EBEs from the final population PK model. For patients with simulated body weight, the updated body weight‐related individual PK parameters were used. The survival probabilities of PFS and OS for each virtual patient were simulated up to 24 months from the exposure‐response models that incorporated simulated exposure metrics and the related covariates. To account for parameter uncertainty, the values of parameters in the simulation were sampled from their typical estimates and the asymptotic variance–covariance matrix of the ER models. The results were summarized and displayed graphically for each weight stratum along with the referenced Asian group.

To further evaluate the extrapolative capability of the popPK model for overweight European patients, a reference‐corrected VPC will be applied by normalizing to the median body weight of the overweight patients who received the proposed high dosage [17]. No simulations were conducted for OR and safety events, as no significant relationship was identified.

2.4. Software

The external validation of popPK model was performed by using NONMEM version 7.5.0 (ICON, Hanover, MD), and Perl‐speaks‐NONMEM [18] (PsN) version 5.3.1. Data management, simulations, computation of summary statistics, and graphical analyses were performed using R [19] version 4.4. Simulation of pharmacokinetic profiles was conducted by using mrgsolve package [20] in R.

3. Results

3.1. PopPK External Validation

A total of 146 PK samples from 13 Asian patients receiving sugmalimab 1800 mg Q4W were used as an external data set to evaluate the established popPK model. The model demonstrated good capability to predict the concentration‐time profiles across a different dosing regimen. The precision of predictions was evaluated using median prediction error (MPE), median absolute prediction error (MAPE), and the percentages of prediction error within 20% (F20) and 30% (F30). The results fell within acceptable ranges [12] (within ±15% for MPE, 30% for MAPE, F20 ≥ 35%, and F30 ≥ 50%) as presented in Table 1. The absence of bias in the goodness‐of‐fit plots (Figure 1a.) and the observations falling within the 5th and 95th prediction intervals in the VPC plot (Figure 1b.) provided further support for the validity of the model.

TABLE 1.

Summary of prediction‐based diagnostics of the external validation for the popPK model.

	MPE (%)	MAPE (%)	F20 (%)	F30 (%)
Based on PRED	−7.84	17.14	80.14	86.3
Based on IPRED	−1.07	9.46	89.73	93.84

Abbreviations: F20 and F30, percentages of prediction error within 20% and 30%, respectively; MAPE, median absolute prediction error; MPE, median prediction error.

FIGURE 1.

External validation for the population pharmacokinetic model. (a) Goodness‐of‐fit plots. (b) Visual predictive check.

As no obvious bias was observed in the external validation, the popPK model of sugemalimab could serve as a reliable tool for dosing regimen optimization.

3.2. Exposure Simulation for High‐Weight Patients and Dose Adjustment

The exposure simulation was conducted in 42 scenarios, encompassing 14 body weight strata and 3 dose levels, with each scenario involving 1002 subjects. The distribution of simulated exposure for the three dose scenarios, categorized by body weight, was shown in Figure 2. As anticipated, exposure decreased with increasing body weight and increased with dose, in the range of 1200–1800 mg.

FIGURE 2.

Boxplots of simulated exposure (C_trough, AUC, and C_max of Cycle 1) for sugemalimab 1200, 1500, and 1800 mg Q3W dosing regimens from high body weight strata. Red indicates that the 90% confidence interval of geometric mean ratio for exposure between the high body weight stratum and GEMSTONE‐302 is not within the range of 0.8–1.25. Conversely, green means within 0.8–1.25.

Using the model‐predicted exposure metrics from the GEMSTONE‐302 study as a reference, acceptable exposure was demonstrated to be maintained up to 115, 90–150, and 135–150 kg for the 1200 mg Q3W, 1500 mg Q3W, and 1800 mg Q3W regimens, respectively (Table 2). And more than 89% of the simulated subjects exceeded the clinical reference concentration for C_trough in these scenarios. For the 1200 mg Q3W dose, the lower limit of the 90% confidence interval for geometric mean ratios (GMRs) compared to the exposure in the GEMSTONE‐302 study for C_trough,C1, C_max,C1 and AUC_C1 remained above 80% for body weights up to 115, 120, and 140 kg, respectively. In the case of the 1500 mg Q3W dosage, the upper limit of the 90% CI for GMRs exceeded 125% for C_trough,C1 and AUC_C1 when body weights were below 85 and 90 kg, respectively. For the 1800 mg Q3W dosage, the 90% CIs of GMRs fell within the range of 80%–125% for body weights between 135 and 150 kg, while increasing by over 25% for body weights below 135 kg. The sensitive analysis also confirmed that the exposure simulations based on η‐sampling were consistent with the EBE‐based approach across weight strata and dosing scenarios (Figure S4).

TABLE 2.

Comparison of exposure metrics by body weight and dose between simulation and GEMSTONE‐302.

Weight strata (kg)	% with Ctrough,C1 ≥ 42.6 μg/mL a	GMR (90% CI) of AUCC1	GMR (90% CI) of Ctrough,C1	GMR (90% CI) of Cmax,C1
1200 mg Q3W
80–85	96.0	1.022 (1.002–1.042)	0.990 (0.963–1.018)	0.960 (0.942–0.978)
85–90	95.6	0.996 (0.977–1.016)	0.958 (0.932–0.986)	0.934 (0.917–0.952)
90–95	94.7	0.974 (0.955–0.993)	0.930 (0.904–0.956)	0.912 (0.895–0.930)
95–100	93.3	0.952 (0.933–0.971)	0.902 (0.878–0.928)	0.891 (0.874–0.908)
100–105	91.9	0.932 (0.914–0.950)	0.878 (0.853–0.902)	0.871 (0.855–0.888)
105–110	90.4	0.912 (0.895–0.930)	0.854 (0.830–0.878)	0.853 (0.837–0.869)
110–115	89.2	0.894 (0.877–0.912)	0.831 (0.808–0.855)	0.836 (0.820–0.852)
115–120	86.8	0.877 (0.860–0.895)	0.810 (0.788–0.833)	0.820 (0.805–0.836)
120–125	84.6	0.861 (0.844–0.878)	0.790 (0.768–0.812)	0.805 (0.790–0.820)
125–130	82.4	0.846 (0.829–0.862)	0.771 (0.750–0.793)	0.791 (0.776–0.806)
130–135	79.9	0.831 (0.815–0.847)	0.752 (0.732–0.774)	0.777 (0.763–0.792)
135–140	77.4	0.817 (0.801–0.833)	0.735 (0.715–0.756)	0.764 (0.750–0.779)
140–145	74.6	0.803 (0.788–0.819)	0.719 (0.699–0.739)	0.752 (0.738–0.767)
145–150	72.0	0.790 (0.775–0.806)	0.703 (0.684–0.723)	0.741 (0.727–0.755)
1500 mg Q3W
80–85	98.8	1.277 (1.252–1.302)	1.238 (1.204–1.273)	1.199 (1.176–1.221)
85–90	98.4	1.246 (1.221–1.270)	1.199 (1.166–1.233)	1.168 (1.146–1.190)
90–95	98.1	1.217 (1.193–1.241)	1.162 (1.130–1.195)	1.139 (1.118–1.161)
95–100	98.1	1.190 (1.166–1.213)	1.129 (1.097–1.161)	1.112 (1.092–1.134)
100–105	97.8	1.165 (1.142–1.188)	1.098 (1.067–1.129)	1.088 (1.068–1.109)
105–110	97.6	1.141 (1.119–1.163)	1.068 (1.038–1.098)	1.065 (1.045–1.086)
110–115	97.5	1.118 (1.096–1.140)	1.040 (1.011–1.069)	1.044 (1.024–1.064)
115–120	96.7	1.096 (1.075–1.118)	1.012 (0.984–1.041)	1.024 (1.005–1.044)
120–125	96.1	1.076 (1.055–1.097)	0.988 (0.960–1.016)	1.005 (0.986–1.024)
125–130	95.4	1.056 (1.036–1.077)	0.964 (0.937–0.991)	0.987 (0.969–1.006)
130–135	94.4	1.038 (1.018–1.059)	0.941 (0.915–0.968)	0.971 (0.952–0.989)
135–140	93.5	1.021 (1.001–1.041)	0.919 (0.894–0.946)	0.955 (0.937–0.973)
140–145	92.7	1.004 (0.985–1.024)	0.899 (0.874–0.924)	0.940 (0.922–0.958)
145–150	91.3	0.988 (0.969–1.007)	0.879 (0.855–0.904)	0.925 (0.908–0.943)
1800 mg Q3W
80–85	99.5	1.532 (1.503–1.563)	1.487 (1.446–1.529)	1.438 (1.411–1.465)
85–90	99.2	1.494 (1.465–1.524)	1.439 (1.399–1.480)	1.399 (1.373–1.426)
90–95	99.2	1.461 (1.432–1.490)	1.396 (1.357–1.435)	1.366 (1.340–1.392)
95–100	99.1	1.427 (1.400–1.456)	1.355 (1.318–1.393)	1.333 (1.309–1.359)
100–105	99.0	1.397 (1.370–1.425)	1.318 (1.281–1.355)	1.304 (1.280–1.329)
105–110	98.8	1.368 (1.342–1.395)	1.282 (1.246–1.318)	1.277 (1.253–1.301)
110–115	98.5	1.341 (1.315–1.368)	1.248 (1.213–1.283)	1.252 (1.228–1.276)
115–120	98.5	1.316 (1.290–1.342)	1.216 (1.183–1.251)	1.228 (1.205–1.251)
120–125	98.4	1.291 (1.266–1.317)	1.186 (1.153–1.219)	1.205 (1.182–1.228)
125–130	98.4	1.268 (1.244–1.293)	1.157 (1.125–1.190)	1.184 (1.162–1.206)
130–135	97.9	1.246 (1.222–1.270)	1.13 (1.098–1.162)	1.164 (1.142–1.186)
135–140	97.7	1.225 (1.201–1.249)	1.104 (1.073–1.135)	1.144 (1.123–1.166)
140–145	97.4	1.205 (1.182–1.229)	1.079 (1.049–1.110)	1.127 (1.106–1.148)
145–150	97.3	1.185 (1.163–1.209)	1.056 (1.026–1.086)	1.109 (1.089–1.130)

Note: Shaded areas indicate 90% CI of GMR within the range of 80%–125%.

Abbreviations: CI, confidence interval; GMR, geometric mean ratio (weight strata vs. GEMSTONE‐302).

^a 42.6 μg/mL was the 5th percentile of observed C_trough,C1 in study GEMSTONE‐302.

Based on the findings from the simulated exposures and the criteria for dose adjustment, it is recommended that individuals with body weights exceeding 115 kg receive an increased dosage to ensure adequate exposure comparable to that of the GEMSTONE‐302 population. It is anticipated that raising the dosage of sugemalimab to 1500 mg Q3W for these patients will effectively address the under‐exposure in C_trough and AUC without elevating C_max by more than 25%, thereby achieving exposures comparable to those observed in patients from the pivotal study who were administered 1200 mg Q3W. In addition, the reference‐corrected VPC by setting the body weight of 130 kg for all patients as the reference (Figure S5) demonstrated that the popPK model performed well in extrapolation and accurately described the PK profiles in the target overweight patients.

3.3. Efficacy Simulation Based on Proposed Dose Adjustment

Following the recommended dose adjustment derived from the PK simulation, 3000 virtual subjects were simulated for the efficacy comparison. The representing 1000 Asian patients and 1000 patients with body weights ranging from 80 to 115 kg, who were administered sugemalimab at a dose of 1200 mg Q3W. The remaining 1000 patients, with body weights between 115 and 150 kg, received a dose of 1500 mg Q3W. As demonstrated in Figure S6, the density plot of C_trough,C1 exhibited that the distributions from the simulated 80–115 kg/1200 mg and 115–150 kg/1500 mg cohorts were predominantly within the range of those from Asian patients (1200 mg, mimic GEMSTONE‐302).

The simulations of the probability of PFS and OS, stratified by body weight according to the proposed dosing regimen, were presented in Figure 3. The median survival probabilities at 12 and 24 months together with the 90% predictive intervals (PIs), which incorporate both between‐subject exposure variability and ER parameter uncertainty, were summarized in Table 3. Notably, there was a notable degree of overlap in the 90% PIs among the three groups, with large variability in response. The results indicated that high‐weight patients (115–150 kg) receiving the escalated 1500 mg of sugmalimab achieved survival outcomes comparable to those of Asian patients treated with 1200 mg, who exhibited the most favorable PFS and OS. In contrast, patients weighing between 80 and 115 kg who received the same dose of 1200 mg demonstrated the least favorable outcomes. Nevertheless, the discrepancies in median survivals at 12 months between the two groups with the most and least favorable outcomes were modest, and were 4.2%, 5.8% and 8.6% for PFS (Investigator‐assessed), PFS (BICR) and OS, respectively.

FIGURE 3.

Survival probability of virtual patients from exposure‐response for progression free survival (investigator‐assessed), progression free survival (BICR), and overall survival. Shaded area represents the 90% predictive interval from 1000 virtual patients and line represents the median values within each group.

TABLE 3.

Probabilities of progression free survival (PFS) and overall survival (OS) at 12 and 24 months from exposure‐response simulation.

Endpoint/group	12 months			24 months
	q5	q50	q95	q5	q50	q95
Probability of PFS assessed by investigator
Asian (1200 mg Q3W)	0.239	0.386	0.538	0.093	0.205	0.357
80–115 kg (1200 mg Q3W)	0.204	0.344	0.453	0.071	0.170	0.268
115–150 kg (1500 mg Q3W)	0.209	0.353	0.466	0.074	0.178	0.281
Probability of PFS assessed by BICR
Asian (1200 mg Q3W)	0.239	0.411	0.606	0.109	0.251	0.459
80–115 kg (1200 mg Q3W)	0.196	0.353	0.495	0.080	0.199	0.336
115–150 kg (1500 mg Q3W)	0.203	0.366	0.512	0.085	0.210	0.354
Probability of OS
Asian (1200 mg Q3W)	0.576	0.722	0.859	0.332	0.522	0.738
80–115 kg (1200 mg Q3W)	0.509	0.636	0.778	0.259	0.405	0.605
115–150 kg (1500 mg Q3W)	0.515	0.657	0.795	0.265	0.432	0.631

Note: In GEMSTONE‐302 Study, 12‐month survival rate of PFS by investigator, PFS by BICR and OS were 0.364, 0.410, and 0.716, respectively; 24‐month survival rate of those were 0.199, 0.230, and 0.522, respectively.

Abbreviation: BICR, blinded independent central review.

4. Discussion

This comprehensive analysis commenced with the application of a different dosing regimen to externally validate the established population pharmacokinetic model. Subsequently, the PK exposures were simulated for different dosages in the high‐weight patient strata derived from the model. By comparing these exposures with those from the pivotal GEMSTONE‐302 study, the threshold body weight for dose adjustment was identified, leading to the proposal of a dosing regimen for patients whose weight exceeds this threshold. Finally, the efficacy of this proposed dosing regimen was further evaluated through simulations based on the exposure‐response models constructed from the pivotal study.

Monoclonal antibodies, such as sugemalimab, are eliminated through proteolytic catabolism, a non‐specific immunoglobulin elimination pathway [21]. Following their binding to targets, these antibodies are degraded intracellularly. This process is independent of the transporters and specific metabolic pathways that small molecules undergo. It is not expected that there will be an interaction between sugemalimab and chemotherapeutic drugs, nor that sugemalimab will influence the exposure of chemotherapeutic agents across different populations. Furthermore, sugemalimab exhibits linear pharmacokinetics (PK) within the dosage range of 3–40 mg/kg, characterized by a relatively flat exposure‐response profile of safety. The intravenous administration of sugemalimab ensures high bioavailability with no dietary absorption effects, a low likelihood of protein binding, and minimal potential for drug–drug interactions (DDI). These attributes are consistent with the ICH E5 guideline [22], indicating that sugemalimab is ethnically insensitive.

Like sugemalimab, many monoclonal antibodies show minimal ethnic sensitivity in their PK and benefit–risk profile, enabling a single global dose. When alternative dose regimens are required, the approach should be exposure‐matching, not based on ethnicity. For nivolumab, popPK analysis found no clinically relevant Chinese vs. non‐Asian difference after covariate adjustment [23]. Atezolizumab exhibited comparable PK between Japanese and non‐Japanese, with efficacy in Japanese subgroups consistent with global trials [24]. Early Japanese studies of cetuximab reported similar PK and safety to Western cohorts, likewise supporting unified dosing [25]. Collectively, these cases adhere to ICH‐E5 and E17 principles, which advocate for the acceptance of foreign clinical data and multi‐regional clinical trial (MRCT) strategies when ethnic sensitivity is low. Consequently, any future exposure‐matching alternative regimens will be maintained across regions.

Body weight is a key factor in determining the appropriate sugemalimab dosing regimen across different ethnic populations. Despite the fact that the evaluation of all covariates, including body weight, in the popPK model was found not to be clinically meaningful, this inference was drawn based on data from six clinical trials, with over 95% of patients being Asian. In particular, the pivotal study was conducted exclusively in Asian patients, whose body weights ranged from 41 to 96 kg, with a mean of 62 kg. In contrast, the average body weight for females in the European population is approximately 70 kg, while for males it is about 80 kg [26]. Therefore, it can be hypothesized that the fixed dose of 1200 mg Q3W in the pivotal study of Asian patients may not be the optimal dosage for European patients. To address this question, a conservative approach was employed. Heavier body weight patients were simulated from a uniform distribution for each 5 kg weight stratum, up to a maximum weight of 150 kg. Although NSCLC patients generally have a lower body weight than the general population, an upper limit of 150 kg was selected as this covers 99.9% of the European adult population. The exposure and efficacy results for these simulated patients were then compared with those from Asian patients whose body weight followed a normal distribution. This methodology effectively identified clinically low exposure among patients with extremely high body weights, despite such individuals being rare even within the European population.

Although the guidances from the FDA and EMA pertain to alternative dosing regimens for PD‐(L)1 drugs, the rationale behind these guidances was applied in our case to inform the dose adjustment based on body weight. The FDA requires only geometric means of exposures when comparing different dosing regimens. A rigorous standard was applied in our analysis, which included not only the comparison of the central tendency but also the dispersion factor by incorporating the 90% confidence interval of GMR in our evaluations. Stringent bioequivalence criteria were employed to ensure that the C_trough and AUC were similar to those observed in the pivotal study conducted with Asian patients, while also avoiding an increase in C_max of more than 25%. The operational acceptance bounds (≤ 20% lower for AUC/C_trough and ≤ 25% higher for C_max) were adopted primarily from the FDA PD‐(L)1 alternative dosing framework, rather than being formally derived from the EMA‐suggested ER slope, as the efficacy ER data were available from only a single dose level. To ensure clinical relevance and robustness in dose optimization, the proportion of simulated C_trough,C1 above the clinical reference concentration, defined as the 5th percentile of observed C_trough,C1 in GEMSTONE‐302, was also referred to determine the appropriate body weight threshold for dose adjustment.

Only pharmacokinetic exposures from the initial dose in cycle 1 were evaluated to determine the appropriate dose adjustment, whereas the FDA also mandates an evaluation of the C_max at steady state. Several considerations were taken into account. Firstly, sugemalimab exhibits time‐varying clearance, which partly reflects the response to treatment. Consequently, exposure at steady state may be confounded with the treatment response; however, cycle 1 exposure metrics avoid this bias and are recommended in the literature to mitigate response‐driven ER artifacts [9, 10]. Secondly, established ER models indicate that the exposures from cycle one have a significant influence on efficacy. Finally, the effect of time‐varying ADA, which typically arise later, could not be reasonably simulated from the popPK model. Importantly, omitting ADA is conservative for this analysis, as ADA tends to increase clearance and then lower exposure. Consequently, ADA inclusion would not contradict, and could further support, the argument for a higher dose in heavier patients.

Although η‐shrinkage was moderate to high for K_des, CL_T, and V _p (~40%–50%) in the popPK model, which could compress variability in simulated exposure distributions if individual EBEs were used exclusively, η‐shrinkage for the primary drivers of cycle‐1 exposure, CL₀ and V _c , was low. This indicates that EBEs for these key parameters remain informative for deriving cycle‐1 exposure metrics. Consistent with this, Cycle‐1 exposure summaries were stable when simulations were repeated using η‐sampling rather than EBEs, supporting that the recommended dosing regimen is robust and not materially influenced by shrinkage.

The efficacy of the proposed dose adjustment was further evaluated using established exposure‐response models for both PFS and OS. When compared with the observed 12‐ and 24‐month survival rates of the sugemalimab arm in GEMSTONE‐302, the median values from 1000 simulated Asian patients closely aligned, suggesting that the ER model accurately simulated efficacy. Despite considerable variability in response, the majority of survival curves overlapped with those from Asian patients, indicating no systematic difference in predicted efficacy between the proposed high dose regimen for high‐weighted patients and the 1200 mg Q3W regimen used in Asian patients. The maximum difference in median survival at 12 months was less than 10%, supporting the notion that the proposed weight‐based dose adjustment at 115 kg could preserve similar efficacy to that observed in Asian patients. Nevertheless, although the ER analyses were used to inform exposure‐metric selection (C_trough,C1) and to support non‐inferiority projections within the observed exposure range, these inferences should be interpreted cautiously as supportive rather than definitive evidence, given that the ER model was developed from a single dose level and may have limited ability to reliably quantify the exposure‐efficacy relationship.

The popPK model was constructed from a range of dose levels and was externally validated in this analysis. Consequently, the simulation of exposure was considered reliable and served as the foundation for weight‐based dose adjustment. As the exposures from the proposed weight‐based dose adjustment remained within the range established in the pivotal study, the simulation results from these ER models supported this dosage proposal. Given that no relationship was identified between sugemalimab exposure and safety, the proposed weight‐based dosage was evaluated in terms of pharmacokinetic exposure and efficacy. In accordance with rigorous bioequivalence criteria, no more than a 25% increase in exposure was deemed sufficient to ensure that the proposed dosing regimen has a comparable safety profile to that observed in the pivotal study. Across the PD‐(L)1 class, the adoption of higher flat doses/longer intervals (e.g., Q4W to Q6W regimens) has not been associated with consistent increases in immune‐related toxicity, with modest C_max/AUC shifts of no more than 25% [27, 28, 29, 30]. This supports the expectation that sugemalimab's benefit–risk is insensitive to the modest exposure increase at 1500 mg for heavier patients.

In summary, we employed a validated population pharmacokinetic model to simulate the exposures in high‐weight patients receiving different dose regimens. Subsequently, these exposures were compared with those from the pivotal study involving Asian patients. We proposed that an increased dose of 1500 mg for patients weighing over 115 kg would be optimal, considering the stringent bioequivalence standards and the regulator‐referenced exposure‐comparability framework. The simulation of PFS and OS based on exposure‐response models demonstrated that the proposed weight‐based dosing regimen for high‐weight patients exhibited efficacy comparable to that in Asian patients who received sugemalimab at 1200 mg Q3W. The modeling and simulation indicated that any potential negative impact of higher body weight on exposure and efficacy could be mitigated by the proposed weight‐based dose adjustment at 115 kg. Given the ethnically insensitive nature of sugemalimab, this proposed dose regimen would be optimal for European patients. Based on this model‐informed regimen, sugemalimab was approved by the European Commission [31] and the UK Medicines and Healthcare products Regulatory Agency [32] in July and October 2024, respectively, for the treatment of non‐small‐cell lung cancer. Our findings demonstrated the application of MIDD approaches to bridge clinical evidence between Asian and European patient populations. Utilizing MIDD may eliminate the need for additional clinical trials, which is particularly valuable in the context of a high unmet medical need.

Author Contributions

Y.S., Z.Y., and K.W. wrote the manuscript. Y.S., Q.S., J.W., and K.W. designed the research. Y.S., Z.Y., and F.X. performed the research. Y.S. and F.X. analyzed the data. All authors reviewed this manuscript.

Funding

This work was sponsored by Cstone Pharmaceuticals.

Conflicts of Interest

Yucheng Sheng, Zenglian Yue, Jingru Wang, and Qingmei Shi conducted this work as salaried employees and stockholders of Cstone pharmaceuticals (Suzhou) Co. Ltd. Fengyan Xu and Kun Wang are employees of Shanghai Qiangshi Information Technology Co. Ltd.

Supporting information

Data S1: psp470220‐sup‐0001‐DataS1.docx.

Data S2: psp470220‐sup‐0002‐DataS2.docx.