Population Pharmacokinetics and Exposure–Response Analysis of Serplulimab in Small Cell Lung Cancer Patients

ABSTRACT

While PD‐L1 antibodies have demonstrated efficacy in small cell lung cancer, the therapeutic benefits remain limited. To address this unmet medical need, serplulimab was developed as an innovative monoclonal antibody targeting PD‐1. This study evaluated the pharmacokinetic (PK) properties of serplulimab and their relationship with efficacy and safety in patients with extensive‐stage small cell lung cancer (ES‐SCLC), using population pharmacokinetics (PopPK) and exposure–response (E–R) analysis to inform dose selection. Data from 1144 patients across eight Phase I–III clinical trials supported a two‐compartment PopPK model with time‐dependent clearance. Cox proportional hazards models were employed to analyze the correlation between exposure (C _avg1 and C _min1) and overall survival (OS)/progression‐free survival (PFS), and adverse events (AEs) with exposure (C _avg1 and C _max1). Body weight, albumin (ALB), and gender significantly influenced the clearance and volume distribution of serplulimab; however, the observed differences in exposure ratio did not reach clinically relevant thresholds (0.8–1.25), thereby obviating the need for dose adjustments. Safety analysis revealed no monotonic increase in AE probability with increasing exposure (p > 0.05). Efficacy analysis indicated no significant correlation between exposure and OS (p > 0.05), whereas lactate dehydrogenase (LDH) and tumor burden emerged as significant predictors of OS (p < 0.05). Results confirm favorable PK and safety of serplulimab at the recommended dose, requiring no adjustment for above covariates. These findings suggest that the current dose is on the plateau of the E–R curve, and dose escalation is unlikely to improve clinical outcomes.

Trial Registration: ClinicalTrials.gov identifier: NCT03952403, NCT04818359, NCT05246164, NCT04747236, NCT03973112, NCT04297995, NCT04778904, NCT04063163

Study Highlights.

• What is the current knowledge on the topic?

  • ○
    Serplulimab is a fully humanized IgG4 anti‐PD‐1 monoclonal antibody with demonstrated clinical activity in cancer. Based on the early‐phase trials and model analyses, 4.5 mg/kg every 3 weeks (Q3W) was projected to achieve optimal target engagement with an acceptable tolerability profile. This dosing regimen was confirmed in the pivotal Phase III ASTRUM‐005 trial, which showed clinically meaningful efficacy and a favorable safety profile for first‐line treatment of patients with ES‐SCLC.
• What question did this study address?

  • ○
    This study aimed to determine which covariates significantly influence serplulimab PK and whether these require dose adjustments. It also sought to clarify if the selected dose achieves maximal efficacy or if higher doses might improve outcomes, as well as to evaluate the relationship between drug exposure and safety. To answer these questions, PopPK modeling and E–R analyses were conducted.
• What does this study add to our knowledge?

  • ○
    The final two‐compartment model with time‐varying clearance accurately described the PK profile. Body weight, albumin, gender, and tumor type significantly influenced PK parameters, but these factors were not clinically relevant, thereby negating the need for dose adjustments. Exposure–efficacy and exposure–safety analyses revealed no significant relationship between exposure and PFS, OS, or safety outcomes. Observed OS differences were driven by baseline LDH and tumor burden rather than exposure.
• How might this change clinical pharmacology or translational science?

  • ○
    This study highlights the value of integrating PopPK and E–R analyses to guide dose selection in oncology. The flat E–R relationship indicates that the selected dose reaches a pharmacodynamic plateau, beyond which higher exposure offers no additional benefit. These findings support more efficient drug development and inform regulatory decisions for immuno‐oncology therapies.

1. Introduction

Small cell lung cancer (SCLC) is a highly aggressive neuroendocrine tumor, accounting for approximately 15% of all lung cancer cases [1]. At the time of diagnosis, around 30%–35% of patients are found to have limited‐stage disease (LS‐SCLC), where the tumor is confined to one side of the chest, while 60%–70% are classified as having extensive‐stage disease (ES‐SCLC) due to metastases to the opposite chest or distant organs [2]. The traditional standard of care was platinum‐based doublet chemotherapy. However, its efficacy has plateaued since the 1980s; prognostic outcomes vary significantly across disease stages: patients with LS‐SCLC can achieve a 5‐year survival rate of 15%–25% following comprehensive treatment, while the 5‐year survival rate for ES‐SCLC patients is less than 3% [3]. Furthermore, most patients experience recurrence within 2 years, and long‐term survival is rare [4]. This therapeutic challenge stems largely from the distinct biological features of SCLC, particularly the lack of well‐characterized driver gene mutations, which have impeded progress in targeted therapy development [5]. Consequently, this limitation has led to a substantial treatment impasse, demanding immediate research breakthroughs.

Targeting the PD‐1/PD‐L1 pathway is a key strategy to overcome tumor immune evasion. By blocking this pathway, PD‐1/PD‐L1 inhibitors enhance T cell‐mediated anti‐tumor responses, and their combination with chemotherapy has shown promising outcomes for patients with treatment‐naïve ES‐SCLC. Atezolizumab, a PD‐L1 monoclonal antibody by Roche, showed a median OS of 12.3 months compared to 10.3 months in the control group in the IMpower133 trial [3], leading to its approval for ES‐SCLC in the United States and China in 2019 [4]. Similarly, AstraZeneca's durvalumab demonstrated a median OS of 13.0 months versus 10.3 months in the control arm in the CASPIAN trial [5], earning approval in 2020. More recently, three PD‐1/PD‐L1 inhibitors have expanded treatment options in China: adebrelimab (Hengrui, approved by China National Medical Products Administration [NMPA] in 2023) showed 15.3 months versus 12.8 months for the control group in the CAPSTONE‐1 trial [6]; toripalimab (Junshi, approved by NMPA in 2024) reported 14.6 months versus 13.3 months in the control arm in the EXTENTORCH study [7], and tislelizumab (BeiGene, approved by NMPA in 2024) achieved 15.5 months compared to 13.5 months for the control group in the RATIONALE‐312 trial [8].

Serplulimab is a fully humanized IgG4 anti‐PD‐1 monoclonal antibody developed by Shanghai Henlius Biotech Inc. Engineered through hybridoma technology, it targets the PD‐1 pathway to enhance anti‐tumor immunity. The pivotal Phase III ASTRUM‐005 trial (NCT04063163) demonstrated the efficacy of serplulimab combined with chemotherapy as a first‐line treatment option in patients with ES‐SCLC. Based on these results, serplulimab was approved by NMPA in December 2022 and by the European Medicines Agency (EMA) in February 2025 for this disease indication.

The 3 mg/kg Q2W regimen was initially selected as the RP2D based on FIH study results, demonstrating good safety, sufficient exposure to achieve PD‐1 receptor saturation, and favorable disease control compared to lower doses. Subsequent PopPK simulations showed that 3 mg/kg Q2W and 4.5 mg/kg Q3W provided comparable exposure levels, both exceeding the pharmacodynamic saturation threshold. Given the flat E–R relationships for efficacy and safety, and to align with the Q3W chemotherapy schedule for better treatment convenience and compliance, 4.5 mg/kg Q3W was chosen for the ASTRUM‐005 trial in ES‐SCLC patients. The ASTRUM‐005 demonstrated a median OS of 15.4 months for patients receiving serplulimab in conjunction with chemotherapy compared to 10.9 months for those receiving placebo plus chemotherapy (HR = 0.63), prolonging median overall survival by 4.7 months [9], thereby representing the greatest survival benefit reported to date among the PD‐1 inhibitors in the first‐line treatment of ES‐SCLC.

The safety analysis from the ASTRUM‐005 trial showed that serplulimab had comparable safety to controls, with grade ≥ 3 treatment‐emergent adverse events (TEAEs) occurring in 82.5% versus 80.1% of patients and treatment‐related adverse events (TRAEs) in 33.2% versus 27.6%. Immune‐related AEs (irAEs) were more frequent in the serplulimab group (37.0%) compared to the placebo group (18.4%), including grade ≥ 3 events (9.5% vs. 5.6%, respectively) [9]. Overall, the incidence and severity of TEAEs were similar between groups, suggesting that serplulimab did not significantly increase toxicity. Observed AEs were consistent with immunotherapy‐related toxicities in SCLC [9]. Serplulimab, administered at 4.5 mg/kg every 3 weeks (Q3W), maintained a manageable safety profile and provided significant survival benefits in ES‐SCLC patients.

This study aims to evaluate the dose selection of serplulimab for ES‐SCLC patients through PopPK and E–R analysis. Specifically, it investigated: (1) the impact of intrinsic factors (e.g., hepatic/renal function, tumor burden) and extrinsic factors (e.g., co‐medications) on the PK of serplulimab; (2) the relationship between drug exposure and AEs, particularly immune‐related toxicity; and (3) the dose–response plateau characteristic of PD‐1 inhibitors, where higher doses beyond a threshold do not yield additional efficacy benefits. The PopPK model was developed using data from 1144 patients with cancer from eight Phase I–III trials. Efficacy analysis focused on the ASTRUM‐005 Phase III trial, constructing an E–R model based on survival outcomes. Safety analyses included all subjects across a broad dose range (0.3–10 mg/kg) to explore E–R relationships for safety endpoints. By integrating PopPK and E–R findings, this study establishes the first quantitative pharmacology framework for serplulimab, providing critical support for personalized therapy, regulatory submissions, and evidence‐based clinical decision‐making.

2. Methods

2.1. Data and Study

The PopPK modeling in this study utilized serum drug concentration data from 1144 participants across eight clinical studies, which included two Phase I trials (HLX10‐001 [ClinicalTrials.gov identifier: NCT03952403] and HLX10HLX04‐001 [NCT04818359]), four Phase II trials (HLX10‐008‐HCC201 [NCT05246164], HLX10‐010‐MSI201 [NCT04747236], HLX10‐011‐CC201 [NCT03973112], and HLX10HLX07–001 [NCT04297995]), and two Phase III trials (HLX10‐004‐NSCLC303 [NCT04778904] and HLX10‐005‐SCLC301 [NCT04063163]). Details of the clinical trials included are presented in Table S1. All trial data were incorporated into the safety analysis, whereas the SCLC trial (HLX10‐005‐SCLC301) was specifically included in the efficacy analysis. All clinical studies adhered to the Declaration of Helsinki, the International Council for Harmonization (ICH) guidelines, and local regulations, with approval obtained from the Institutional Review Board (IRB) or Independent Ethics Committee (IEC) at each study site. Written informed consent was secured from all patients or their legal representatives prior to participation.

2.2. Population Pharmacokinetic Model

The PopPK structural model of serplulimab is a two‐compartment PK model with time‐dependent clearance. Since the basic model did not fit long‐term (post‐6 months) sampling data well, a time‐varying clearance PK model was adopted. Detailed information is provided in the Supporting Information. Covariate selection followed a stepwise approach, with significance levels of p = 0.01 (forward) and p = 0.001 (backward). Model fit was assessed using goodness‐of‐fit plots, pcVPC, and bootstrap methods [10]; each based on 1000 simulations (pcVPC and bootstrap).

2.3. Evaluation of the Influence of Covariates on PK Parameters

The influence of covariates on PK was assessed utilizing simulated exposure generated via Bayesian post hoc methods. A dosing regimen of 4.5 mg/kg of serplulimab, administered Q3W for eight cycles with a 1‐h infusion, was simulated, and PK samples were collected at 0.25‐h intervals. The area under the curve at steady‐state (AUC_ss) from the beginning of the eighth infusion to Day 21 post‐dose was calculated using R software. The steady‐state average concentration (C _avgss) was derived by dividing AUC_ss by the duration of the dosing interval. Additionally, the maximum concentration (C _maxss) and the minimum concentration (C _minss) at the last cycle were determined. Due to missing weight data for two subjects (0.18% of the study population), these two subjects were excluded from the model application. Predicted exposures were categorized according to the covariates under investigation, and the geometric mean of each category was compared to the geometric mean of the overall subject exposure. The results were presented using a forest plot. In this study, we considered the ratio of change to reflect no significant PK difference when it fell within the range of 0.8–1.25. However, when the ratio deviates beyond the 0.7–1.43 range, clinical significance may be indicated.

2.4. Exposure–Response Analysis

The average concentration (C _avg1), maximum concentration (C _max1), and minimum concentration (C _min1) following the initial dose were simulated as exposure metrics for the E–R analysis. C _avg1 and C _max1 were utilized for safety analysis, while C _avg1 and C _min1 were employed for the E–R efficacy analysis. The PopPK model analysis reveals changes in clearance over time. This phenomenon has also been observed in other approved PD‐1 monoclonal antibodies, such as pembrolizumab [11] and nivolumab [12]. PopPK results for both of these drugs show that their clearance rates decrease with prolonged drug administration, and the extent of this change relative to baseline clearance is associated with the treatment response in cancer patients [11, 13, 14]. In patients with partial response (PR) and complete response (CR), the reduction in steady‐state clearance relative to baseline was greater than in patients with stable disease (SD) or progressive disease (PD), suggesting that the efficacy of this class of drugs may affect steady‐state exposure, potentially confounding the E–R relationship. This may lead to an overestimation of the correlation between exposure and both efficacy and safety, especially with single‐dose regimens [15]. Furthermore, exposure levels at the first dose are more readily obtainable compared to the steady‐state exposures.

The safety analysis endpoints include grade 3 and above TEAEs, grade 3 and above adverse drug reaction (ADR), serious AEs, AEs of special interest, and immune‐related AEs. All these endpoints were binary; the incidence rates for patients across exposure quartiles were calculated to assess potential E–R trends.

The efficacy endpoints, PFS and OS, were stratified by PK exposure and relevant covariates, with Kaplan–Meier (K–M) survival curves used for plotting. The Cox proportional hazards models detailed in the Supporting Information were further established to describe the E–R relationship, considering PK exposure and other covariates as potential predictors.

3. Results

3.1. Data

The pharmacokinetic analysis dataset included 6677 serplulimab serum concentration measurements from 1144 subjects. Of these, 27 data points fell below the lower limit of quantification (LLOQ) and were excluded from the analysis. The E–R safety analysis was conducted using the same subject population as the PopPK analysis. The E–R efficacy analysis was performed using data from patients with SCLC in the HLX10‐005‐SCLC301 trial, which included a total of 389 subjects. The baseline demographic characteristics of these subjects are shown in Table 1. Matrix plots showing the correlation of covariates in the PK and E–R analysis datasets are presented in Figures S8 and S9, respectively.

TABLE 1.

Demographics and baseline covariates for PK modeling in patients.

Characteristics	PK dataset (n = 1144)	ER efficacy dataset (n = 389)
Baseline age (AGE, years)	61.0 (23.0, 83.0)	63.0 (28.0, 76.0)
Baseline bodyweight (WT, kg)	64.5 (33.0, 131)	67.0 (33.0, 120)
Baseline height (HT, cm)	168 (128, 191)	168 (142, 191)
Baseline body mass Index (BMI, kg/m2)	23.1 (13.0, 42.3)	23.8 (13.7, 42.3)
Baseline body surface area (BSA, m2)	1.73 (1.19, 2.55)	1.76 (1.19, 2.39)
Baseline albumin (ALB, g/L)	41.3 (23.9, 67.9)	41.1 (23.9, 67.9)
Baseline aspartate aminotransferase (AST, U/L)	22.0 (5.00, 179)	21.9 (7.00, 179)
Baseline alanine transaminase (ALT, U/L)	19.0 (4.00, 228)	20.0 (4.00, 181)
Baseline alkaline phosphatase (ALP, U/L)	95.0 (10.7, 863)	95.0 (10.7, 607)
Baseline lactate dehydrogenase (LDH, U/L)	219 (38.0, 2555)	248 (116, 2239)
Baseline total bilirubin (BILI, μmol/L)	10.2 (1.60, 42.9)	9.80 (1.60, 42.9)
Baseline creatinine (CREAT, μmol/L)	68.9 (23.0, 156)	69.2 (26.7, 156)
Baseline creatinine clearance (CRCL, mL/min)	90.9 (28.5, 291)	92.6 (28.5, 291)
Baseline tumor burden (TUMBUR, mm)	88.0 (10.0, 350)	117 (13.8, 323)
Sex
Male	917 (80.16%)	317 (81.49%)
Female	227 (19.84%)	72 (18.51%)
Race
Asian	897 (78.41%)	262 (67.35%)
Non‐Asian	247 (21.59%)	127 (32.65%)
Anti‐drug antibody (ADA)
Negative	1078 (94.23%)	376 (96.66%)
Positive	23 (2.01%)	5 (1.29%)
Missing	43 (3.76%)	8 (2.06%)
Baseline ECOG
0	306 (26.75%)	71 (18.25%)
1	835 (72.99%)	318 (81.75%)
2	3 (0.26%)	—
COMB
No	382 (33.39%)	—
Yes	762 (66.61%)	389 (100.00%)
COMBANTI
No	1003 (87.68%)	389 (100.00%)
Yes	141 (12.33%)	—
Tumor type (TUMTP)
Hepatic cancer	125 (10.93%)	—
Colorectal cancer	86 (7.52%)	—
Lung cancera	817 (71.42%)	389 (100.00%)
Other	116 (10.14%)	—

Note: Continuous variables are presented as median (min–max); categorical variables are presented as N (%). The missing data for continuous variables primarily included the following covariates: height (n = 28, 2.45%), body mass index (BMI) (n = 30, 2.62%), body surface area (BSA) (n = 30, 2.62%), and lactate dehydrogenase (LDH) (n = 30, 2.62%) among all patients (n = 1144). The missing rates for other covariates were all below 2%.

Abbreviations: COMB, combined with chemotherapy; COMBANTI, combination antibody‐based anti‐tumor therapy; ECOG, Eastern Cooperative Oncology Group performance status.

^aLung cancer includies non‐samll cell lung cancer and small cell lung cancer.

3.2. Population Pharmacokinetic Model

Weight was first tested as a covariate in the base model and subsequently retained in the base model due to a significant decrease in OFV. The stepwise covariate screening results show that weight and ALB significantly influenced CL, while weight, gender, and tumor type significantly influenced the V _c. The final model incorporated these relationships, which are described in detail below:

$${\mathrm{CL}}_i\left(\mathrm{L}/\mathrm{day}\right)=0.225\times \exp \left(E_{\max i}\times \frac{{\text{TIME}}^{2.05}}{{\text{TIME}}^{2.05}+{106}^{2.05}}+0.531\times \ln \left(\frac{\mathrm{WT}}{65}\right)-0.783\times \ln \left(\frac{\mathrm{ALB}}{41.3}\right)+{\eta }_{\mathrm{CL},i}\right)$$

$$V_{{\mathrm{c}}_i}\left(\mathrm{L}\right)=3.52\times \exp \left(0.450\times \ln \left(\frac{\mathrm{WT}}{65}\right)-0.0887\times \text{TUMTP}-0.121\times \mathrm{SEX}+{\eta }_{{\mathrm{V}}_{\mathrm{c},i}}\right)$$

$$E_{\max i}=-0.364+{\eta }_{E_{\max ,i}}$$

WT represents weight, ALB represents albumin, and SEX represents gender (SEX = 0 for male, SEX = 1 for female). TUMTP represents tumor type (TUMTP = 0 for lung cancer, TUMTP = 1 for hepatocellular carcinoma, colorectal cancer, and other tumor types). CL, V _c, and E _max correspond to clearance, central compartment volume of distribution, and the natural logarithm of the maximum relative change in clearance from baseline, respectively. η _CL,i , ${\eta }_{V_{\mathrm{c},i}}$, and ${\eta }_{E_{\max ,i}}$ represent the inter‐individual variability of CL, V _c, and E _max, respectively, for the ith subject.

The diagnostic plots of the final PopPK model for serplulimab are illustrated in Figure S1, demonstrating strong consistency between observed and predicted concentrations. No obvious bias was observed in the CWRES plots with respect to time and predicted concentration. The estimated parameters of the base model and the final model, along with the bootstrap results of the final model, are presented in Table 2. The median values of the bootstrap estimates closely matched the typical values from the final PopPK model estimates, and the 95% confidence intervals (CIs) of the final PopPK model estimates highly overlap with the 95% prediction intervals (PIs) (2.5th–97.5th percentiles) of the bootstrap estimates. These findings confirm the robustness and precision of the model. Additionally, the pcVPC plot of the final model and base model are depicted in Figures S2 and S3, illustrating that the final and base PopPK model effectively captures both the central tendency and variability of serplulimab concentrations across all included clinical studies.

TABLE 2.

Parameter estimates of the final model and bootstrap parameter estimates.

Parameter	Parameter description	Base model estimate (RSE%)	Final model estimate (RSE%, 95% CI)	Bootstrap estimate median (2.5th–97.5th percentiles)
CL0 (L/day)	Clearance at baseline	0.225 (2.11%)	0.225 (2.85%, 0.213; 0.238)	0.224 (0.209; 0.24)
V c (L)	Central volume	3.32 (0.757%)	3.52 (0.849%, 3.46; 3.58)	3.52 (3.46; 3.58)
Q (L/day)	Inter‐compartmental clearance	0.447 (4.87%)	0.463 (5.98%, 0.412; 0.52)	0.463 (0.38; 0.568)
V p (L)	Peripheral volume	2.22 (5.76%)	2.21 (7.41%, 1.91; 2.55)	2.22 (1.92; 2.58)
exp(E max)	Maximum magnitude change in CL	0.695 (3.1%)	0.695 (4.52%, 0.636; 0.759)	0.698 (0.63; 0.779)
T 50 (day)	Time when clearance maximal change equals 50%	108 (8.68%)	106 (4.52%, 83.2; 129)	108 (82.8; 139)
λ	Sigmoid factor of time‐varying CL	2.03 (5.99%)	2.05 (14.2%, 1.48; 2.62)	2.07 (1.55; 3.14)
CLwt	Influence of body weight on CL	—	0.531 (12.2%, 0.403; 0.658)	0.533 (0.452; 0.619)
Vwt	Influence of body weight on V c	—	0.450 (7.3%, 0.385; 0.514)	0.452 (0.383; 0.516)
CLalb	Influence of albumin on CL	—	−0.783 (13.4%, −0.893; −0.674)	−0.782 (−0.943; −0.626)
Vcsex	Influence of sex on V c	—	−0.121 (−0.152; −0.0891)	−0.119 (−0.15; −0.0866)
Vctumtp	Influence of tumor type on V c	—	−0.0887 (15.5%, −0.116; −0.0617)	−0.0881 (−0.116; −0.0604)
ωCL, V c	Covariance (CL, V c)	0.0346 (8.62%)	0.017 (12.6%, 0.0128; 0.0212)	0.0169 (0.0131; 0.0213)
IIVCL	Inter‐individual variability of CL (%)	29.4 (3.41%)	25.8 (4.37%, 23.5; 27.9)	25.7 (23.6; 27.9)
IIVV c	Inter‐individual variability of V c (%)	20.6 (4.14%)	15.4 (5.46%, 13.7; 17.0)	15.4 (13.7; 17.1)
IIVQ	Inter‐individual variability of Q (%)	56.6 (14.9%)	49.7 (22.5%, 17; 68.1)	50.0 (20.4; 66.4)
IIVV p	Inter‐individual variability of V p (%)	49.7 (8.67%)	51.5 (8.74%, 41.7; 59.7)	50.9 (41.1; 61.9)
IIVE max	Inter‐individual variability of E max (%)	27.3 (8.53%)	26.3 (9.08%, 21.1; 30.6)	26.3 (21.5; 32.2)
σ	Residual (%)	16.5 (3.02%)	16.6 (2.74%, 15.7; 17.4)	16.5 (15.5; 17.5)

Note: IIV for CL, V _c, Q, V _p, E _max, and residual are reported as approximate CV%.

Abbreviations: CI, confidence interval; IIV interindividual variability; RSE, relative standard error.

3.3. Influence of Covariates on PK Exposure

The PopPK analysis identified body weight, ALB, and gender as statistically significant covariates influencing the PK parameters of serplulimab. However, the resulting exposure differences did not reach clinically meaningful thresholds. As shown in Figure 1 (C _avgss) and Figures S4 (C _maxss) and S5 (C _minss), higher body weight was associated with lower exposure, with steady‐state exposure ratios across quartiles relative to the reference value ranging from 0.88 to 1.14. Additionally, ALB negatively correlated with CL, leading to higher exposure in individuals with higher ALB, with exposure ratios ranging from 0.818 to 1.17. Furthermore, gender influenced the V _c, resulting in slightly higher exposure in females compared to males, with exposure ratios ranging from 0.945 to 1.07. Although race was not identified as a statistically significant covariate, it was included in the forest plot due to its potential clinical interest. Non‐Asian subjects (21.6%, N = 247) showed slightly higher exposure compared to Asian subjects, with exposure ratios between 0.945 and 1.17; however, this observation may be confounded by differences in body weight between the groups. Overall, the exposure ratios influenced by these covariates ranged from 0.818 to 1.17, well within the 0.8 to 1.25 range and below the pre‐specified threshold of clinical relevance (0.7–1.43). These findings support the conclusion that no dose adjustments for serplulimab are necessary based on these covariates.

FIGURE 1.

Forest plots for serplulimab C _avgss by significant covariates in patients. The ratio of the geometric mean of individual C _avgss (average concentration at steady‐state) levels in the subgroup to the geometric mean of individual exposure levels in patients. A vertical line displays a geometric mean ratio of 1. Solid circles display the mean geometric mean ratios for patients in the subgroup, and error bars display the 95% confidence intervals of their geometric mean ratios. The gray and dark gray shaded areas represent the ranges of geometric mean ratios, 0.7–1.43 and 0.8–1.25, respectively.

3.4. Exposure–Response Analysis

The results of the E–R safety analysis are presented in Figure 2. No monotonic relationship was observed between increasing C _max1 of serplulimab and the probability of AEs. Similarly, no E–R trend was identified between C _avg1 and AEs, as presented in Figure S6.

FIGURE 2.

Probability of adverse events across serplulimab C _max1 quantiles in patients. The blue open circles represent the observed adverse events in patients. The black solid circles represent the observed probability of adverse events, and the error bars represent the standard errors in quantiles. The green vertical dotted lines represent the quantiles of exposures. The red lines are loess lines between C _max1 and the probability of adverse events. C _max1, peak concentration after first dose.

The results of the E–R efficacy analysis are depicted in Figure 3. The PFS curves stratified by the median of C _min1 (Figure 3A) and C _avg1 (Figure 3B) were comparable and show no E–R tendency. In contrast, the OS curves stratified by the median of C _min1 (Figure 3C) and C _avg1 (Figure 3D) suggest a tendency toward longer OS in the high‐exposure group compared to the low‐exposure group. The K–M plots for OS across quartiles of baseline LDH and tumor burden are shown in Figure 3E,F. To further explore the potential E–R relationship, a Cox proportional hazards analysis was conducted. The univariate Cox model results for OS and individual covariates are presented in Table S2. Covariates with p < 0.05, including ALB, AST, BILI, tumor burden, and LDH, were subsequently incorporated into multivariate Cox models with exposure as a predictor. The modeling process, using forward inclusion and backward elimination strategies, is detailed in Table S3. The final model equation using C _min1 (Model 17) and C _avg1 (Model 19) as predictors is as follows:

$$h_i\left(t\right)=h_0\left(t\right)·\exp \left({\beta }_1·\log \left(X_i\right)+{\beta }_2·\log \left(\mathrm{LDH}\right)+{\beta }_3·\log \left(\text{TUMBUR}\right)\right)$$

where h _i (t) is the risk function of subject i at time t, h ₀(t) is the baseline risk function, X _i represents the exposure of the subject (C _avg1 or C _min1), β ₁ represents the coefficient of the exposure effect, β ₂ represents the coefficient of the covariate LDH effect, and β ₃ represents the coefficient of the covariate tumor burden effect.

FIGURE 3.

PFS and OS assessed by quartiles of serplulimab exposure and significant covariates. Solid lines represent Kaplan–Meier curves, shaded areas represent 95% CI, and the p‐value is derived from a log‐rank test. C _min1, trough concentration after first dose; LDH, lactate dehydrogenase.

The results indicated that LDH and tumor burden were significant factors influencing OS (p < 0.05). Higher LDH levels and greater tumor burden were correlated with shorter OS. In contrast, no significant correlation was observed between exposure (C _avg1 and C _min1) and OS (p > 0.05) within the current exposure range. The Cox regression parameter estimates are shown in Table S4. Table S5 demonstrates that the distribution of LDH and tumor burden across different exposure quartiles is imbalanced; patients in the higher exposure groups (Q3 and Q4) had lower LDH and tumor burden values compared to those in the lower exposure groups (Q1 and Q2), indicating a better baseline disease status in the high exposure groups. When LDH and tumor burden were included in the model as confounding factors, their associated p‐values were statistically significant, while the “adjusted” p‐value for the effect of exposure on OS was no longer significant.

The hazard ratio of each covariate in the final model with C _min1 as the predictor is shown in Figure 4. The length of the bars reflects the magnitude of impact. LDH shows the strongest association, followed by tumor burden. The hazard ratios for baseline LDH and tumor burden at the 95th percentile were greater than 1, with 95% confidence intervals that do not include 1, indicating a significant association between higher values of these factors and increased mortality risk. As shown in Figure 3E,F, patients with lower baseline LDH levels and tumor burden exhibited longer OS compared to those with higher levels, reinforcing the strong relationship between OS and these prognostic biomarkers. The hazard ratios of the 5th and 95th percentiles of C _min1 were both close to 1, and their 95% confidence interval included 1, indicating no significant association with OS. Similar findings were obtained in the model using C _avg1 as the predictor, as shown in Figure S7. These results suggest that the apparent trend between exposure and OS observed in Figure 3C,D is likely due to a confounding effect, further supporting the conclusion that, within the exposure range of the studied dose of 4.5 mg/kg, exposure does not significantly impact OS.

FIGURE 4.

The effects of significant covariates and serplulimab C _min1 on the hazard ratio of OS in patients. The figure shows the impact of significant covariates and C _min1 on the hazard ratio (HR) for OS, relative to a subject with reference values. The x‐axis is on a logarithmic scale of HR, with a vertical line at 1.0 indicating no change in risk. The y‐axis lists the covariates and exposure metric included in the Cox model. Each box represents the HR relative to the reference value, with its left and right edges indicating the 5th (P05) and 95th (P95) percentiles, respectively. The blue hollow and red solid squares with error bars correspond to the HRs and 95% confidence intervals at P05 and P95 values, respectively, and align with the numerical values shown on the right. A blue color indicates the direction of increasing covariate value. Points to the left of 1.0 suggest a decreased risk (HR < 1), while those to the right indicate an increased risk (HR > 1). A 95% CI that crosses 1.0 suggests the result is not statistically significant. LDH, lactate dehydrogenase.

4. Discussion

This study developed a PopPK model of serplulimab in cancer patients using a large dataset. It estimated the PK parameters and inter‐individual variability among subjects and evaluated the effects of covariates on these parameters. The final structural model of serplulimab was a two‐compartment linear elimination PK model with clearance varying over time, which accurately described the PK characteristics of serplulimab within the studied dosage range. The final model developed based on current data demonstrated that incorporating a time‐dependent clearance (CL) significantly improved model performance. The diagnostic plots, individual fitting plots, and pcVPC demonstrated that the final PopPK model reliably represented the PK data following intravenous administration. The RSE% of key PK parameters CL₀, V _c, Q, V _p, E _max, T ₅₀, and λ in the final model were all less than 15%, indicating precise estimation. Weight, ALB, gender, and tumor type were identified as significant covariates on PK parameters, with their associated RSE% ranging from 7.12% to 15.5%, further supporting the accuracy of these estimates. The inter‐individual variability of PK parameters was consistent with the typical range observed for monoclonal antibody [16]. The final model estimated a time‐dependent CL for serplulimab, with a T ₅₀ of 106 days. Simulations based on individual parameter estimates indicated that the median half‐life of serplulimab was 19.0 days after the first dose and 24.4 days at steady state.

Body weight, ALB, gender, and tumor type were identified as significant covariates influencing the PK parameters of serplulimab. However, these findings do not warrant dose adjustments. To further evaluate the impacts of covariates on exposures (C _avgss, C _maxss, and C _minss), this study compared the geometric mean ratios (GMRs) of exposure between different covariate subgroups and the overall population. The results indicated that variations in exposure across body weight quartiles, ALB quartiles, genders, cancer types (lung cancer vs. other tumors), and racial groups (Asian vs. non‐Asian) all fell within the range of 0.8 to 1.25 relative to reference values. These differences were considered minor and not clinically significant. A comprehensive covariate analysis also showed that additional factors, including age, BMI, BSA, AST, ALT, tumor burden, ECOG, LDH, CREAT, BILI, CRCL, ADA, concomitant chemotherapy drugs, concomitant antibody‐based anti‐tumor therapy, and race, did not have a statistically significant effect on the PK parameters of serplulimab. It should be noted that among all 1144 subjects, the numbers of ADA‐negative, ADA‐positive, and subjects with missing ADA status were 1078, 23, and 43, respectively. The geometric mean clearance in ADA‐positive subjects was 13.3% higher than in ADA‐negative subjects. This difference is small, and the incidence of ADA was low—only 2.01% (23/1144)—which suggests that the impact of ADA is not clinically meaningful.

The E–R safety analysis involved eight clinical trials encompassing a broad dosing range, from 0.3 to 10 mg/kg. The incidence rates of AEs, including grade 3 and above TEAEs, grade 3 and above ADR, serious AEs, AEs of special interest, and immune‐related AEs, did not demonstrate a consistent, monotonic increase with rising drug exposure. In patients with SCLC, no E–R relationship was observed between PFS and exposures (C _avg1 and C _min1). Although an apparent E–R trend for OS was noted when patients were grouped by median exposure, further analysis using a Cox proportional hazards model revealed no significant association between exposure (C _avg1 and C _min1) and OS (p > 0.05). Instead, LDH levels and baseline tumor burden emerged as significant predictors of OS, with higher values correlating with shorter survival. LDH is a known marker of tumor burden and disease progression, and elevated levels are generally associated with poorer prognosis [17]. The longer OS observed in the higher exposure group may be attributed to imbalances in baseline LDH and tumor burden across exposure groups. In Phase III trials, only a single dose group is typically included, which often fails to display a complete dose–response curve or demonstrate a clear dose–response relationship. This is because the selected dose aims to ensure that the majority of patients achieve a high level of efficacy. It does not imply that very low exposure levels would still be effective, but rather that most subjects already reach a sufficiently efficacious exposure range.

The absence of exposure‐dependent efficacy beyond the plateau threshold further supports the pharmacodynamic saturation typical of PD‐1 blockade. Notably, this plateau effect suggests that increasing the dose would not overcome unfavorable prognostic biomarkers or lead to improved clinical outcomes. A similar observation was reported in the E–R analysis of nivolumab [18]. Therefore, distinguishing between statistical significance and clinical relevance of covariates is crucial when evaluating the dosing regimen of serplulimab. The E–R analysis data in this Phase III trial is based on validation results from earlier dose‐finding stages. The scientific focus has shifted from dose optimization to E–R validation, rather than directly verifying the impact of dose on efficacy. Based on the findings of this study, adjustments based on covariates such as weight, albumin, gender, or tumor type are not recommended.

One limitation of the study was that the established model estimated a time‐dependent CL for serplulimab, showing a T ₅₀ of 106 days—longer than the corresponding values of 67.4 days for pembrolizumab [11], 58 days for nivolumab [13] and 53.6 days for sugemalimab [14]. Given that a substantial proportion of subjects in the current dataset had relatively short clinical observation periods (with only 56.1% and 29% of subjects having a maximum observation time exceeding 100 days and 6 months, respectively), we anticipate that the inclusion of longer‐term follow‐up data in future analyses will further refine the precision of this parameter estimation.

In summary, this study not only provided a comprehensive and systematic characterization of the PK profile of serplulimab but also conducted an E–R analysis in patients with SCLC. These findings offer quantitative support for the development of prescribing information and establish a solid foundation for future pharmacokinetic and pharmacodynamic research.

Author Contributions

K.W. wrote the manuscript; L.Z., C.H., Q.W., and Y.S. designed the research; Q.W. and Y.S. performed the research; F.X., K.W., Y.G., Q.W., and Y.S. analyzed the data.

Conflicts of Interest

K.W., F.X., and Y.G. are employees of Shanghai Qiangshi Information Technology Co. Ltd. and serve as consultants for Shanghai Henlius Biotech Inc. Y.S., C.H., L.Z., and Q.W. are employees of Shanghai Henlius Biotech Inc.

Supporting information

Data S1: cts70322‐sup‐0001‐DataS1.docx.

Wang K., Shen Y., Hu C., et al., “Population Pharmacokinetics and Exposure–Response Analysis of Serplulimab in Small Cell Lung Cancer Patients,” Clinical and Translational Science 18, no. 9 (2025): e70322, 10.1111/cts.70322.

Data Availability Statement

The datasets used and/or analyzed in this study are available from the corresponding author (L.Z.) and the sponsor (Shanghai Henlius Biotech Inc.) on reasonable request.