Population pharmacokinetic modeling of adalimumab biosimilar SCT630 and Humira® in healthy participants and patients with moderate to severe plaque psoriasis to establish PK similarity

Abstract

Objective

SCT630 is a recombinant fully human anti-TNF-α IgG1 monoclonal antibody developed as a biosimilar to adalimumab reference product Humira® (AbbVie). This study aims to evaluate the pharmacokinetic (PK) similarity, characterize pharmacokinetics between healthy participants and patients, and identify potential sources of PK variability of SCT630 and Humira® using population pharmacokinetic modeling approach.

Methods

A population pharmacokinetic model was developed using PK data from a phase I study in healthy participants and a phase III study in patients with moderate to severe plaque psoriasis (PsO).

Results

A one-compartment PopPK model with first-order linear elimination adequately described SCT630 and Humira® concentration data. The covariates analysis identified anti-drug antibody, body weight, total bilirubin, alanine aminotransferase, participant type and study drug as significant covariates affecting clearance and volume of distribution. Model-predicted exposures, including C_max and AUC were obtained, and their geometric mean ratios (test versus the reference group) were calculated. The 90% confidence intervals (CI) for all these ratios fell entirely within the range of 80.00% to 125.00%.

Conclusion

This analysis integrated data from both healthy participants and patients with PsO, and provided a valuable assessment of adalimumab PK characteristics and significant covariates. This model-based approach further elucidates how the SCT630 PK profile compares to that of Humira® in healthy participants and patients with PsO, supporting the totality-of-evidence for the biosimilarity of SCT630.

Introduction

Adalimumab is a recombinant fully human monoclonal antibody against tumor necrosis factor-α (TNF-α). It binds to TNF-α with high specificity, thereby blocking the interaction between TNF-α and its receptor, and preventing TNF-α from activating downstream signaling pathways and exerting its proinflammatory effects. Adalimumab was initially approved by the US Food and Drug Administration (FDA) in 2002 and by the European Medicines Agency (EMA) in 2003 for the treatment of moderate to severe active rheumatoid arthritis, and subsequently had since been extended to several autoimmune diseases, including plaque psoriasis (PsO).^1,2

Biosimilar first appeared in a 2005 EMA guideline, and was subsequently adopted by US FDA.^3,4 Biosimilars are not innovative drugs; rather, they have been shown, based on evidence from clinical studies, to be highly similar to the original reference products.⁵ Multiple biosimilars of adalimumab have been approved since its patent expiration.⁶ SCT630, a recombinant fully human anti-TNF-α IgG1 monoclonal antibody developed by Sinocelltech, Ltd., is a biosimilar which has the same amino acid sequence as adalimumab. It was approved by China National Medical Products Administration (NMPA) in 2023 for treatment of 8 indications including rheumatoid arthritis, ankylosing spondylitis and psoriasis.⁷

During the development of adalimumab biosimilars, similarity in pharmacokinetics (PK) is usually confirmed in phase I studies in healthy participants, while phase III studies provide evidence of similarity in efficacy, safety and immunogenicity in patients.^8,9 PK comparison using observed data plays a pivotal role and meets regulatory expectations. However, dense sampling in patients is not always feasible due to the short intervals between administrations. Consequently, thorough PK comparisons in patients have been limited. Population pharmacokinetics (PopPK) analysis can integrate rich or sparse PK data from multiple clinical studies. Although not mandatory by regulators, it has become an essential tool in biosimilar development, providing additional supportive information on PK equivalence between biosimilar candidates and their reference products.¹⁰ The Technical Guideline for Clinical Pharmacology Studies of Biosimilars issued by the NMPA recommends performing PopPK assessment if the population or dose in clinical efficacy similarity study differs from that in clinical PK similarity study.¹¹ Several PopPK studies have explored the PK characteristics of biosimilar candidates against Humira® in the treatment of ankylosing spondylitis and rheumatoid arthritis, but few published biosimilar studies have been used a PopPK approach in patients with PsO.^{12, 13, 14}

The PK similarity of SCT630 has been established using non-compartmental analysis (NCA) based on observed data of healthy participants in a phase I clinical study SCT630PS01 (NCT03917628), subsequently a phase III study SCT630PS03 (NCT03927352) established its clinical similarity in efficacy and safety in treating moderate to severe PsO.^15,16 Here, we report the PopPK modeling analysis to integrate PK data from healthy participants and patients with moderate to severe PsO. This study aimed to leverage PopPK modeling to further consolidate the totality-of-evidence for SCT630. This model-based approach provided a comprehensive assessment of PK similarity across different populations and dosing conditions, quantifying potential covariate effects (e.g., immunogenicity, disease status) on drug exposure to support the primary NCA-based bioequivalence conclusion with additional mechanistic and integrative evidence. To our knowledge, this is the first published PopPK study of adalimumab biosimilar in Asian PsO patients.

Methods

Study design

The data for the PopPK analysis in this study were obtained from the phase I study SCT630PS01 (NCT03917628) and phase III study SCT630PS03 (NCT03927352). Both studies were conducted in compliance with the principles of the Declaration of Helsinki and Good Clinical Practice guidelines, and were reviewed and approved by the relevant institutional review boards and ethics committees. Written informed consents were obtained from all participants prior to enrollment in the studies.

SCT630PS01 was a randomized, double-blind, parallel controlled study. Participants were randomized in a 1:1 ratio to receive either SCT630 or Humira® with a single dose of 40 mg by subcutaneous (SC) injection. PK blood samples were intensively collected at pre-dose, 8 hours post-dose, and on days 2 to 9, 15, 22, 29, 36, 43, 57 and 71. Immunogenicity samples were collected at pre-dose, and on days 15, 29, 43, and 71.

SCT630PS03 was a randomized, double-blind clinical study consisting of 2 stages. In the first stage, patients with PsO were randomized in a 1:1 ratio to receive SCT630 or Humira® at doses of 80 mg and 40 mg on day 1 of week 1 and week 2, respectively, followed by repeated doses of 40 mg once every 2 weeks until week 16. In the second stage, all patients with clinical response in the SCT630 group continued the original treatment regimen until week 48. Patients in the Humira® group who achieved clinical response were re-randomized 1:1 at week 16 to receive either SCT630 or Humira® until week 48. Sparse trough concentration samples were collected at week 1 (pre-dose), 4, 16, 32, 48 and 50. Immunogenicity samples were collected at week 1 (pre-dose), 4, 16, 32, 48, 50, 52.

PK and immunogenicity sample analysis

Serum concentrations of SCT630 and Humira® were quantitatively determined using a single ELISA method with lower limit of quantitation of 15.6 ng/mL. The assay was fully validated and was able to measure SCT630 and Humira® equivalently, with accuracy ranging from -17.8% to 7.5% and precision below 19.4% (Supplementary Table S1). The use of a single assay eliminated assay-induced systematic bias and ensured the comparability of PK data. PopPK modeling approach was used to integrate the concentration data from both studies. Data post re-randomization in the phase III study were excluded from the dataset, as the concentrations of SCT630 and Humira® were indistinguishable.

Anti-drug antibody (ADA) was tested using a validated bridging electrochemiluminescence immunoassay (ECLIA) following a tiered approach consisting of screening, confirmation and titer determination. The neutralizing ability of the ADAs were evaluated using a validated cell-based assay. ADA and neutralizing antibody (Nab) status were incorporated into the model as binary covariates. A participant was identified as ADA-positive or Nab-positive if ADA or Nab was detected at any post-dose timepoints.

Softwares

The PopPK model was developed using NONMEM (Version 7.5). Perl-speak-NONMEM (PsN, Version 4.6.0) was used for model validation. R (Version 4.1.0) software was used for data processing, graphical analysis, and plotting. WinNonlin (Version 8.3.1) was used for non-compartmental analysis of PK parameters. SAS (Version 9.4) was used to arrange the data and analyze the simulation results.

Population model analysis and evaluation

The model development was performed using the first order conditional estimate with interaction (FOCE-I) algorithm. One-compartment model was widely used to describe the PK profiles of adalimumab.^{13,17, 18, 19} Furthermore, the logarithmic serum concentration-time curves indicated only 1 declining phase in healthy participants.¹⁵ Therefore, model development started with one-compartment model described by linear clearance (CL/F), volume of distribution (V/F) and absorption rate constant (Ka), implemented using NONMEM subroutine ADVAN2 TRANS2. Two-compartment model (ADVAN4 TRANS4) and different error models were subsequently tested and compared.

The inter-individual variability (IIV) of parameters was expressed by an exponential model:

$$P_i=P_{TV}\times exp\left({\eta }^{P_i}\right)$$

where P_i is the predicted value of the parameter for individual i. P_TV is the typical value of the parameter. ${\eta }^{P_i}$ is the interindividual variance, fitting normal distribution η∼N (0, ω²), defined by the covariance matrix Ω (OMEGA) in NONMEM.

For residual variance, additive, proportional, and mixed models were explored:

$$C_{ij}={\widehat{C}}_{ij}+\left({\epsilon }_{aij}\right)$$

$$C_{ij}={\widehat{C}}_{ij}\left(1+{\epsilon }_{P_{ij}}\right)$$

$$C_{ij}={\widehat{C}}_{ij}\left(1+{\epsilon }_{P_{ij}}\right)+{\epsilon }_{aij}$$

C_ij is the j-th observed plasma concentration for the i-th individual; Ĉ_ij is the model-predicted value of the j-th plasma concentration for the i-th individual; ε_Pij and ε_aij are proportional and additive residuals, respectively, fitting normal distribution ε∼NID (0, σ²), defined by Σ (SIGMA) in NONMEM.

The base model was selected primarily based on the reduction of objective function value (OFV), model simplicity, diagnostic plots and parameter stability.

Once a base model was selected, covariates were screened by the graphical method and the correlation between covariates was considered. Then the covariates those could potentially impact PK parameters were evaluated using a stepwise modeling approach, with a criteria of p < 0.05 for the forward inclusion and p < 0.01 for the backward elimination steps respectively. The effect of continuous variables on PK parameters was evaluated by the power function model:

$$P_i=P_{TV}\times {\left(\frac{Covariate}{Median}\right)}^{\theta }$$

where P_i is the parameter of i-th individual, P_TV is the typical value of the parameter, and θ is the exponential parameter that relate the continuous covariate to the population parameter.

The categorical variables were evaluated by the following models:

$$P_i=\{\begin{array}{c}{P}_{TV},ifCovariate=1\\ P_{TV}\times \left(1+\theta \right),ifCovariate=0\end{array}$$

where P_i is the parameter of i-th individual, P_TV is the typical value of the parameter and θ is the parameter that relate the categorical covariate to the population parameter.

Any significant covariates identified using the stepwise approach were incorporated into the model and this model was defined as the final model. The final model was evaluated using goodness-of-fit (GOF) plots, visual predictive check (VPC) and bootstrap analysis. GOF included plots of predicted concentrations against observed concentrations, as well as conditional weighted residuals (CWRES) against time after dose and against predicted concentrations. A good alignment between predicted and observed concentrations indicated that the model was able to accurately describe the observed data. CWRES were expected to be randomly distributed around the zero line. VPC was used to evaluate the predictive performance of the model. Using the parameters of the final model, 1000 simulations were performed to obtain the confidence intervals of the 5%, 50% and 95% quantiles of the predicted values. The distribution of predicted and observed values were compared. Bootstraps resampling with 1000 replicates were performed to assess the robustness of the final model. The parameters calculated from the original dataset were compared with those from the bootstrap analysis. The final model was considered robust if the final model estimates fell within the 95% CI of the bootstrap results.

Evaluation of biosimilarity

The individual predicted PK profiles were obtained using the empirical Bayes estimates (EBE) PK parameters from the final PopPK model. The PK parameters after single-dose (C_max, AUC_0-t and AUC_0-∞) and at steady state after multiple-dose (C_max,ss, AUC_tau) were calculated using non-compartmental analysis. The geometric least squares mean (GLSM) exposures were derived for the SCT630 and Humira® groups respectively. Finally, the ratios of GLSM (SCT630 to Humira®) and corresponding 90% confidence intervals were calculated. SCT630 was considered similar to Humira® if the confidence intervals fell entirely within the range of 80% to 125%.

Results

Dataset

A total of 513 participants from the phase I and III studies contributed 3528 concentration data. Among these, 700 values were below the quantification limit (BQL), including 511 pre-dose and 189 (5.4%) post-dose results. The proportion of post-dose BQL values was less than 10% and was evenly distributed between SCT630 and Humira® groups. Therefore all the BQL data were omitted (M1 method).^20,21 In the phase III study, 2 participants had drug concentration tested prior to the first dose, and 1 participant in SCT630 group incorrectly received Humira®. All 3 participants were excluded from the dataset. Finally, a total of 2822 serum concentration data from 510 participants were included in the PopPK analysis, including 2169 concentration data from 146 healthy participants (1075 concentration data from 73 participants treated with SCT630 and 1094 concentration data from 73 participants treated with Humira®) and 653 concentration data from 364 patients (319 concentration data from 182 patients treated with SCT630 and 334 concentration data from 182 patients treated with Humira®). The PK profiles of SCT630 and Humira® are presented in Fig. 1. The demographic characteristics of participants included in the PopPK model are shown in Table 1.

Fig. 1.

Serum concentration-time scatter plots of SCT630 and Humira®.

Table 1.

Summary of baseline characteristics.

	Healthy participants		Patients with PsO		All (n=510)
	SCT630 (n=73)	Humira® (n=73)	SCT630 (n=182)	Humira® (n=182)
Continuous covariates	Median (min, max)	Median (min, max)	Median (min, max)	Median (min, max)	Median (min, max)
Age (year)	29 (19, 44)	29 (19, 44)	38 (19, 69)	37 (20, 69)	34 (19, 69)
Body weight (kg)	64.4 (53.5, 77.0)	65.0 (50.8, 79.9)	74.8 (46.5, 128)	74.5 (49.0, 126)	69.9 (46.5, 128)
Height (cm)	169 (157, 181)	169 (154, 181)	173 (147, 184)	172 (148, 197)	170 (147, 197)
Albumin (g/L)	47.7 (43.1, 52.7)	48.0 (41.3, 52.1)	46.3 (39.3, 53.5)	46.2 (36.7, 54.4)	46.8 (36.7, 54.4)
Alkaline phosphatase (U/L)	73.0 (37.0, 110)	70.0 (35.0, 110)	82.0 (30.0, 183)	77.0 (30.7, 150)	76.0 (30.0, 183)
Alanine aminotransferase (U/L)	15.0 (8.00, 51.0)	17.0 (8.00, 40.0)	23.0 (7.00, 118)	22.0 (7.00, 107)	21.0 (7.00, 118)
Aspartate aminotransferase (U/L)	17.0 (12.0, 31.0)	18.0 (13.0, 34.0)	21.0 (8.90, 46.4)	21.4 (9.70, 83.9)	20.0 (8.90, 83.9)
Total protein (g/L)	73.7 (67.2, 81.2)	74.4 (65.4, 83.9)	74.5 (63.1, 86.0)	74.3 (62.8, 86.2)	74.1 (62.8, 86.2)
Creatinine clearance (mL/min)	119 (78.9, 161)	123 (76.7, 181)	122 (56.6, 301)	124 (62.1, 237)	122 (56.6, 301)
Hemoglobin (g/L)	156 (124, 173)	155 (112, 176)	154 (109, 184)	154 (118, 183)	154 (109, 184)
Lymphocyte count (109/L)	1.80 (1.00, 3.40)	1.90 (0.900, 3.20)	1.77 (0.610, 3.40)	1.78 (0.800, 4.31)	1.80 (0.610, 4.31)
Lymphocyte percentage (%)	30.9 (16.5, 46.0)	32.7 (19.3, 47.1)	25.1 (0.109, 44.0)	24.6 (0.167, 45.4)	27.2 (0.109, 47.1)
Monocyte count (109/L)	0.320 (0.170, 0.670)	0.300 (0.170, 0.650)	0.430 (0.170, 0.840)	0.420 (0.200, 1.00)	0.400 (0.170, 1.00)
Monocyte percentage (%)	5.50 (3.20, 9.70)	5.30 (3.20, 9.60)	6.05 (0.050, 12.2)	5.80 (0.048, 12.0)	5.70 (0.048, 12.2)
Platelet (109/L)	238 (130, 348)	249 (139, 365)	238 (101, 429)	241 (126, 436)	240 (101, 436)
Serum creatinine (μmol/L)	73.0 (58.0, 92.0)	73.0 (55.0, 94.0)	72.0 (40.0, 107)	72.0 (38.4, 127)	72.0 (38.4, 127)
Total bilirubin (μmol/L)	14.8 (5.90, 24.2)	15.6 (4.00, 23.6)	11.9 (5.30, 36.3)	12.3 (5.60, 41.5)	12.7 (4.00, 41.5)
Categorical covariates	n (%)	n (%)	n (%)	n (%)	n (%)
Sex
Female	0 (0.0%)	0 (0.0%)	35 (19.2%)	35 (19.2%)	70 (13.7%)
Male	73 (100.0%)	73 (100.0%)	147 (80.8%)	147 (80.8%)	440 (86.3%)
Nation
Han	69 (94.5%)	72 (98.6%)	175 (96.2%)	177 (97.3%)	493 (96.7%)
Other	4 (5.5%)	1 (1.4%)	7 (3.8%)	5 (2.7%)	17 (3.3%)
Type of participant
Healthy	73 (100.0%)	73 (100.0%)	0 (0.0%)	0 (0.0%)	146 (28.6%)
Patient	0 (0.0%)	0 (0.0%)	182 (100.0%)	182 (100.0%)	364 (71.4%)
Concomitant medication
Absent	65 (89.0%)	67 (91.8%)	24 (13.2%)	21 (11.5%)	177 (34.7%)
Present	8 (11.0%)	6 (8.2%)	158 (86.8%)	161 (88.5%)	333 (65.3%)
Underlying diseases
Absent	73 (100.0%)	73 (100.0%)	0 (0.0%)	0 (0.0%)	146 (28.6%)
Present	0 (0.0%)	0 (0.0%)	182 (100.0%)	182 (100.0%)	364 (71.4%)
Antidrug antibody
Negative	7 (9.6%)	4 (5.5%)	30 (16.5%)	23 (12.6%)	64 (12.5%)
Positive	66 (90.4%)	69 (94.5%)	152 (83.5%)	159 (87.4%)	446 (87.5%)
Neutralizing antibody
Negative	62 (84.9%)	57 (78.1%)	109 (59.9%)	100 (54.9%)	328 (64.3%)
Positive	11 (15.1%)	16 (21.9%)	73 (40.1%)	82 (45.1%)	182 (35.7%)

Note: A participant was identified as antidrug antibody positive or neutralizing antibody positive if ADA or Nab was tested at any post-dose timepoints respectively.

PopPK model analysis

Different model structures were compared, including one-compartment, two-compartment and different error models. Ultimately, one-compartment model with first order elimination and combined error resulted in the lowest OFV and was selected as the base model. The stepwise modeling approach identified ADA, body weight (BW), total bilirubin (TBIL) and participant type (TYPE) as covariates on clearance, and alanine aminotransferase (ALT), BW and study drug (DRUG) on volume of distribution. A summary of the model development process is presented in Supplementary Table S2. The final model parameters are summarized in Table 2.

Table 2.

Parameters of the final PopPK model and bootstrap estimates.

	Final model		Bootstrap
Parameter (Units)	Estimate	RSE(%)	Estimate	95%CI
Fixed effects
Ka (1/h)	0.0219	6.0%	0.0218	(0.0193, 0.0245)
CL/F (L/h)	0.0162	2.6%	0.0161	(0.0154, 0.0170)
ADA ON CL/F	-0.358	12.8%	-0.359	(-0.409, -0.298)
BW ON CL/F	1.12	10.4%	1.11	(0.910, 1.32)
TBIL ON CL/F	-0.134	35.8%	-0.135	(-0.235, -0.0320)
TYPE ON CL/F	0.317	21.1%	0.317	(0.231, 0.418)
V/F (L)	8.63	2.4%	8.62	(8.21, 9.02)
ALT ON V/F	-0.104	33.5%	-0.106	(-0.182, -0.0340)
BW ON V/F	1.04	10.6%	1.04	(0.780, 1.29)
DRUG ON V/F	0.185	20.1%	0.186	(0.111, 0.261)
Inter-individual variability
ω2CL/F	0.133	7.0%	0.131	(0.108, 0.155)
	Shrinkage	5.8%
ω2V/F	0.0503	16.0%	0.0511	(0.0308, 0.109)
	Shrinkage	34.3%
ω2Ka	0.358	14.6%	0.353	(0.238, 0.492)
	Shrinkage	47.5%
Residual variability
Proportional error	0.145	1.1%	0.144	(0.103, 0.173)
Additive error	182	2.9%	181	(142, 241)

Note: RSE: relative standard error; CI: confidence interval; V/F: volume of distribution; CL/F: clearance; Ka: absorption rate constant; ADA: anti-drug antibodies; BW: body weight; TBIL: total bilirubin; TYPE: participant type; ALT: alanine aminotransferase; DRUG: the study drug.

The diagnostic plots of the final PopPK model are shown in Fig. 2. Good correlation was observed between the individual or population predictions and the observed data. The trend lines coincided with the diagonal lines, indicating that the model provided a good fit to the observed data. The plots of CWRES against predicted concentrations and time after dose did not indicate any misspecification of the structural model or the residual error model.

Fig. 2.

Diagnostic plots of the final PopPK model.

Blue dots represent measured values, the black lines represent fitting lines, and the red lines represent smooth local regression curves.

VPC was used to assess the predictive performance of the final model. As shown in Fig. 3, the 5%, 50% and 95% quantile lines of the observed values fell within the 90% CI of the corresponding quantiles of the simulated concentration data, indicating similar distribution characteristics between the observed values and the predictions. Thus, the model demonstrated good predictability and adequately described the pharmacokinetic characteristics.

Fig. 3.

Visual predictive check plot of final model.

Blue circles are observed data. Red solid and dashed lines are the median and 5% and 95% percentiles of the observed data respectively. Black solid and dashed lines are the median and 5% and 95% percentiles of the predicted data. The shaded areas are the corresponding 90% confidence intervals of the predicted median and 5% and 95% percentiles.

A total of 995 out of 1000 bootstrap runs converged successfully. The bootstrap estimates are presented in Table 2 and are consistent with the final model parameters. Furthermore, the final model parameters all fell within the 95% confidence intervals of the bootstrap estimates, supporting the robustness of the final PopPK model.

Effect of significant covariates

In the final PopPK model, body weight, total bilirubin, participant type, ADA, ALT and study drug were identified as significant covariates. The effects of these covariates on the typical values of PK parameters were evaluated and are presented in Fig. 4. A typical individual was defined as a patient with moderate to severe PsO having a baseline body weight of 69.9 kg, TBIL of 12.7 μmol/L, and ALT of 21.0 U/L, who was treated with SCT630 and developed ADA during treatment. Compared with the typical individual, for participants whose body weight fell within 5% quantile (54 kg) to 95% quantile (99 kg) of the modeling population, the percent change in CL/F ranged from -29.3% to 59.99%, and the percent change in V/F ranged from -27.61% to 54.95%. The CL/F of ADA-negative participants was 44.78% to 26.82% lower than that of the ADA-positive patients. Humira® had slightly higher V/F than SCT630, with an increase of 11.21% to 25.79%. Healthy participants had higher CL/F than the patients, with an increase of 18.57% to 44.83%. The effects of TBIL on CL/F and ALT on V/F were minimal, with percent changes relative to the typical values within ±20%. The impact of TBIL and ALT may not hold substantial clinical relevance.

Fig. 4.

Impact of covariates on model parameters.

The ranges of BW, TBIL and ALT on the left side indicate the median values and 5% and 95% quantiles of each parameter of the model population; ADA-positive is taken as control for the status of ADA; patients are taken as control for participant type; SCT630 is taken as control for study drug.

Evaluation of PK similarity between SCT630 and Humira®

The model-predicted exposure parameters after single dose (C_max, AUC_0-t, AUC_0-∞) and at steady state following multiple doses (C_max, AUC_tau) were calculated. The GLSM ratios and corresponding 90% CIs are shown in Table 3.

Table 3.

Single- and multiple-dose biosimilarity.

Data type	Participant type and dosage	Parameter (Unit)	GLSM
			T	R	T/R Ratio (90% CI) (%)
Observed data	Healthy, SD	Cmax (ng/mL)	3736	3220	116.02 (108.66, 123.88)
		AUC0-t (h*ng/mL)	2106912	1924607	109.47 (99.80, 120.08)
		AUC0-∞ (h*ng/mL)	2210675	2023643	109.24 (98.80, 120.78)
Model predicted data	Healthy, SD	Cmax (ng/mL)	3633	3133	115.94 (109.03, 123.30)
		AUC0-t (h*ng/mL)	2101030	1952097	107.63 (98.98, 117.04)
		AUC0-∞ (h*ng/mL)	2208320	2082240	106.06 (96.49, 116.57)
	Healthy, MD	Cmax (ng/mL)	7750	7236	107.10 (98.20, 116.81)
		AUCtau (h*ng/mL)	2208234	2080825	106.12 (96.64, 116.53)
	Patient, SD	Cmax (ng/mL)	7228	6145	117.63 (113.29, 122.14)
		AUC0-t (h*ng/mL)	4727800	4287307	110.27 (103.07, 117.99)
		AUC0-∞ (h*ng/mL)	5192640	4722267	109.96 (101.81, 118.76)
	Patient, MD	Cmax.ss (ng/mL)	8908	7948	112.07 (105.09, 119.51)
		AUCtau (h*ng/mL)	2577006	2347555	109.77 (101.79, 118.38)

Note: T = test group SCT630; R = reference group Humira®. GLSM = geometric least-squares mean. SD=single dose. MD=multiple dose. CI = confidence interval.

In healthy participants, the GLSM ratios for SCT630 compared with Humira® after a single dose were 115.94% (90% CI: 109.03%, 123.30%) for C_max, 107.63% (90% CI: 98.98%, 117.04%) for AUC_0-t, and 106.06% (90% CI: 96.49%, 116.57%) for AUC_0-∞. These values were comparable to the results calculated from observed PK data in the phase I study.⁸ After repeated doses, the GLSM ratios for SCT630 compared with Humira® were 107.10% (90% CI: 98.20%, 116.81%) for C_max, and 106.12% (90% CI: 96.64%, 116.53%) for AUC_tau.

In patients with PsO, GLSM ratios for SCT630 compared with Humira® after a single dose were 117.63% (90% CI: 113.29%, 122.14%) for C_max, 110.27% (90% CI: 103.07%, 117.99%) for AUC_0-t, and 109.96% (90% CI: 101.81%, 118.76%) for AUC_0-∞. After repeated doses, the GLSM ratios for SCT630 compared with Humira® were 112.07% (90% CI: 105.09%, 119.51%) for C_max, and 109.77% (90% CI: 101.79%, 118.38%) for AUC_tau.

After single or multiple doses in healthy participants and patients with PsO, all the 90% CIs of GLSM ratios fell within the predefined range of 80.00% to 125.00%. Therefore, SCT630 is considered biosimilar to the reference product Humira® in the study population.

Discussion

PK similarity plays a vital role in demonstrating biosimilarity between an investigational drug and its reference product. To support the development of the adalimumab biosimilar SCT630, PK similarity between SCT630 and Humira® was initially demonstrated using an NCA approach based on observed data in healthy participants, and the clinical similarity was subsequently demonstrated in patients with moderate to severe PsO.^15,16 Phase I trials usually employ intensive PK sampling to perform NCA for each participant, while in phase III studies intensive sampling is not always feasible due to short intervals between repeated administrations. Instead, sparse sampling is commonly implemented. This PopPK analysis, as a model-based approach, is capable of generating comprehensive PK data, thereby providing further support to the totality-of-evidence in the assessment of PK similarity in patients. Similar population model-based approach has also been used for other biosimilars to adalimumab across various indications.^12,13,17,19

In this study, a one-compartment model with first order absorption and elimination adequately fitted the observed data. The robustness of the model was assessed through several approaches. The model-predicted PK exposure parameters were consistent with those calculated from the observed data in the phase I study (Table 3), suggesting good predictive capability of the model and supporting the reliability of the following biosimilarity analysis.¹⁵ Regulatory guidelines typically specify a bioequivalence margin of 80% to 125% for observed PK data. In PopPK model-based analysis, the employment of EBE estimates reduces variability in exposure parameters compared with observed data. Applying the same bioequivalence margin to less variable, model-based analysis should be conservative and robust to evaluate PK similarity. This acceptance criteria has been used in several adalimumab biosimilar studies.^13,17 The GLSM ratios and corresponding 90% CIs for all the key PK parameters fell entirely within the predefined equivalence range of 80.00% to 125.00% in this study. These results provided additional supportive evidence to establish the PK similarity between SCT630 and Humira®.

The immunogenicity characteristics of adalimumab vary widely across clinical studies and products. Several biosimilar phase III studies have reported ADA-positive rates ranging from 54.2% to 95.9%.^{22, 23, 24, 25, 26, 27} Notably, within each study, the positive rates were generally consistent between the reference product Humira® and the biosimilar products. In the phase I study of SCT630, the ADA-positive rate was 94.5% and 90.4% in the Humira® group and SCT630 group, respectively. In the phase III study, 83.6% and 76.1% participants developed ADA during treatment in the Humira® group and SCT630 group respectively. It should be noted that assay methods can grant substantial differences in immunogenicity results.²⁸ In clinical studies of Humira®, a higher ADA-positive rate was observed in PsO patients with drug concentrations below 2 μg/mL compared with the overall patient population (21% vs. 8%).²⁹ In addition, ADA incidence measured by sensitive ECL assay has been shown to be higher than that measured by ELISA method.²⁹ In our studies, a highly sensitive ECLIA method was employed which could detect ADA at levels as low as 5.42 ng/mL and tolerate drug concentrations as high as 100 μg/mL. The improvement in methodology may partly explain the higher ADA incidences reported in some biosimilar studies compared with those in earlier studies of Humira®.^18,29 Although the ADA positive rate was relatively high and the presence of ADA was identified as a covariate affecting CL/F in this study, the effect of ADA on efficacy was not statistically significant in either SCT630 or Humira® groups.¹⁶

Treatment (SCT630 vs. Humira®) was identified as a statistically significant covariate for V/F. However the SCT630/ Humira® ratio in exposures (97.91% to 117.63%) entirely fell within the biosimilar acceptance range of 80% to 125%, supporting the similarity in PK.

Body weight is a commonly identified covariate in PopPK models for monoclonal antibody drugs.^30,31 In this study, both CL/F and V/F increased with body weight, leading to lower drug exposures in participants with higher body weight. This finding is consistent with previously reported adalimumab PopPK analysis.^18,19 The reference product Humira® has a well-established broad therapeutic and safety window and has been approved for the treatment of PsO at a fixed dose regardless of body weight. This study demonstrated no efficacy or safety signals for patients with extreme body weight (46.5 to 128 kg), further supporting the appropriateness of the fixed dosing regimen in this population.

Additionally, participant type (healthy participants vs. patients with moderate to severe PsO) was identified as a significant covariate for CL/F. Healthy participants had 31.7% higher CL/F than patients. After multiple doses, the AUC_tau in patients is 1.17-fold (2577006 versus 2208234 h*ng/mL) and 1.13-fold (2347555 versus 2080825 h*ng/mL) of healthy participants for SCT630 and Humira® respectively (Table 3). The consistent higher exposure in patients for both SCT630 and Humira® suggests no significant difference between the two products in terms of drug exposures.

Furthermore, TBIL was identified as a significant covariate for CL/F, and ALT for V/F. Extreme maximum ALT and TBIL values were observed in patients (118 U/L and 41.5 μmol/L respectively) compared with healthy participants (51.0 U/L and 24.2 μmol/L respectively). Both covariates were retained in the final model to account any potential differences between the study populations, even though their impact on typical PK parameters was minimal (within ±20%).

Only Asian participants were enrolled in this study; therefore a head-to-head comparison with other ethnic groups was not feasible. However, one study reported C_max values for Humira®/biosimilar after multiple doses in White PsO patients ranging from 8657 to 8696 ng/mL.³² Another study reported C_max values for Humira®/biosimilar after a single dose in White healthy participants ranging from 3900 to 4100 ng/mL.³³ These results were consistent with those observed in this study (7948 to 8908 ng/mL in patients after repeated doses, and 3133 to 3633 ng/mL in healthy participants after a single dose), suggesting PK may be comparable across races and ethnicities.

This study has several limitations. External validation could not be performed due to lack of an independent external dataset, and sensitivity analysis (e.g. for covariate inclusion) was not conducted. Nevertheless, given that the primary objective of this study was to establish PK similarity rather than to support dose extrapolation or adjustment, the robustness and accuracy demonstrated by internal evaluations (GOF, VPC and bootstrap) were considered acceptable for this purpose. Another limitation was that only trough concentrations were collected in patients with PsO during the phase III study. The lack of observed C_max data may introduce bias to the estimate of absorption phase and overall exposure. In addition, ADA was incorporated into the model as a binary covariate rather than time-varying ADA status or titer, which might conceal the true impact of ADA on PK. However, the ADA-positive rates in SCT630 and Humira® groups were similar; therefore the effect on the PK similarity evaluation should be minimal.

In conclusion, the pharmacokinetic behavior of SCT630 and Humira® in healthy participants and patients with moderate to severe PsO was well described by the final PopPK model. The comparisons in exposures based on model-predicted data and observed data from the phase I study were consistent, further supporting the PK similarity of SCT630 to Humira®. The effects of significant covariates were quantified. Collectively, these findings add value to the totality-of-evidence for SCT630 biosimilarity and enhance the understanding of adalimumab PK characteristics in patients with PsO, particularly in the Asian population.

CRediT author contribution statement

Dongyang Li: Data curation, Formal analysis, Methodology, Writing – original draft. Yuhuan Jiao: Conceptualization, Investigation, Formal analysis, Project administration, Writing – review & editing. Kun Wu: Formal analysis, Writing – review & editing. Hongyun Ma: Formal analysis, Writing – review & editing. Qianqian Hong: Formal analysis, Writing – review & editing. Liangzhi Xie: Conceptualization, Supervision, Writing – review & editing.

Ethics statement

Both clinical studies involved in the PopPK analysis were reviewed and approved by institutional review boards and ethics committees. Written informed consent statements were obtained before the participants were enrolled in the studies.

Funding

This work was sponsored by Sinocelltech Ltd. and supported by National Science and Technology Major Projects of China [2018ZX09736002].

Declaration of competing interest

Sinocelltech Ltd. sponsored this work. Dongyang Li, Yuhuan Jiao, Kun Wu, Hongyun Ma, Qianqian Hong and Liangzhi Xie are employees of Sinocelltech Co., Ltd. and have stock ownership and/or potential stock option interests in the company. All authors declare no other conflicts of interests.