Pharmacokinetic/LDL‐C and exposure–response analysis of tafolecimab in Chinese hypercholesterolemia patients: Results from phase I, II, and III studies

Abstract Tafolecimab, a novel fully human monoclonal antibody targeting PCSK9, has been assessed in Chinese healthy volunteers and patients with hypercholesterolemia. This analysis is to develop and qualify a population pharmacokinetics (PopPKs)/LDL‐C model to characterize tafolecimab PK and LDL‐C profiles, evaluate the impact of potential covariates on tafolecimab, estimate individual predicted exposure, and LDL‐C decreasing, furthermore, explore exposure–response relationship to support clinical use. Data from six clinical trials in China were used to develop the PopPK/LDL‐C model. A Michaelis–Menten approximation of the target‐mediated drug disposition (TMDD) model was used to describe PK data and indirect response (IDR) model was developed to estimate the LDL‐C profile. A stochastic approximation expectation maximization algorithm was applied to estimate PopPK/LDL‐C parameters. The PK/LDL‐C time course data for tafolecimab were well described by TMDD/IDR model. Baseline covariates resulting in statistically significant changes in PK/LDL‐C parameters included: body weight and sex on absorption rate constant; body weight, sex, and unbound PCSK9 on central volume; body weight and sex on clearance; baseline LDL‐C on first‐order rate constants for the removal of an effect); and disease and sex on maximum effect. However, the magnitudes of changes associated with these covariates do not necessitate dose adjustment. Exposure–efficacy relationship indicated that the nadir of LDL‐C reduction achieved with the steady‐state trough plasma concentration (C trough) of tafolecimab at 5 μg/mL, and no further LDL‐C decreasing with the increasing C trough. There was no exposure dependency observed in exposure‐safety exploration. The PopPK/LDL‐C model was successfully developed, validated, and predicted tafolecimab/LDL‐C concentrations and individual exposures.


INTRODUCTION
Elevated serum cholesterol level is a major risk factor for cardiovascular diseases, and lowering low-density lipoprotein-cholesterol (LDL-C) is the primary target for patients with hypercholesterolemia. 1 Given the unsatisfactory LDL-C target attainment rate on statin therapy alone, especially for those on high or very-high cardiovascular risk, additional lipid-lowering therapeutics are highly desirable. 2,3DL receptor (LDLR) recycling plays an important role in regulating circulating LDL-C levels and maintaining cholesterol balance.4 Proprotein convertase subtilisin/kexin type 9 (PCSK9) binds to LDLR and facilitates its degradation after internalization, decreasing LDLR on the cell surface and increasing circulating LDL-C levels.5 Thus, PCSK9 inhibition is an effective strategy to lower circulating LDL-C levels.6 In recent years, two PCSK9 monoclonal antibodies, alirocumab and evolocumab, have been approved by the US Food and Drug Administration (FDA) to treat hypercholesterolemia as well as reduce cardiovascular events.[7][8][9] When administered once or twice monthly, these drugs reduced LDL-C levels by ~60% and provides long-term cardiovascular benefit to patients with hypercholesterolemia. 10,11 Tafolecimab is a fully human IgG2 PCSK9 monoclonal antibody produced through affinity maturation by chain shuffling and complementarity determining region mutagenesis on candidate antibodies discovered from a synthetic human antibody library. In addition, tafolecimab was approved in August 2023 in China as an adjunct to diet, in combination with a statin or statin with other LDL-Clowering therapies, for the treatment of adults with primary hyperlipidemia (including heterozygous familial

WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?
Tafolecimab is a novel long-acting LDL-C lowering monoclonal antibody.The new drug application submission of tafolecimab in China had been approved.Tafolecimab population pharmacokinetic (PopPK)/LDL-C model was successfully developed to guide the clinical development and clinical practice.The article first discloses the PopPK/LDL-C data of tafolecimab.This work will provide a reference for the clinical development of other PSCK9 antibodies.

WHAT QUESTION DID THIS STUDY ADDRESS?
This work presents a PopPK/LDL-C analysis characterizing tafolecimab PK and LDL-C profile, evaluating the impact of potential covariates on tafolecimab, and estimating individual predicted exposure and LDL-C decreasing of tafolecimab, also exploring exposure-response (E-R) relationship to support the dose regimen choice for clinical use.However, given the limitations regarding combination administration and disease phenotype, the conclusions from PopPK/LDL-C model would only be applicable for clinical practice in the population annotated in the label.

WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?
The PopPK/LDL-C modeling and E-R for tafolecimab has not yet been explored in Chinese patients with hypercholesterolemia.This study proposes a pragmatic but systematic approach to model building, coupled with E-R relationship exploration, to elucidate the impact of covariates on parameters and instruct target patients on medication.

WHAT MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?
A 2-cmt target-mediated drug disposition/indirect response model is well-developed and validated to describe PopPK/LDL-C time course data, after covariates being modeled, the impact of covariate on parameter is explored to determine dose adjustment among different population.The E-R relationship is also explored to support long-term dosing.Our analysis will provide a potential, useful tool to guide target patients' treatment in the future.

Clinical study design
Data are derived from six clinical studies of tafolecimab, including a phase I study (NCT03366688) in healthy volunteers (HVs), a phase IIa study (NCT03815812) in patients with hypercholesterolemia, and four phase III studies (NCT04031742, NCT04179669, NCT04289285, and NCT04709536) in patients with non-FH, homozygous familial hypercholesterolemia (HoFH) or HeFH (Table 1).All studies performed involving human participants were in accordance with local laws, the International Conference on Harmonization Goof Clinical Practice guidelines, and the ethical principles outlined in the Declaration of Helsinki.

Bioanalysis
Free tafolecimab concentrations in human serum samples were analyzed using a validated enzyme-linked immunosorbent assay (ELISA) method.A sandwich ELISA method utilizing 96-well ELISA plates was used with anti-tafolecimab neutralizing antibody-90 as a capture.Calibrators, controls, and samples were added to the plate.For detection, anti-tafolecimab neutralizing antibody-90 that was previously labeled with Biotin was used followed by addition of streptavidin-horseradish peroxidase conjugate and tetramethyl benzidine.The analytical format of tafolecimab was the unbound drug for both arms.The lower and upper limit of quantification were 0.4 μg/ mL and 12.8 μg/mL, respectively.Tafolecimab is stable for up to 397 days when stored in −20°C or 1267 days when stored in −70°C.
LDL-C was measured using a Beckman Coulter AU680 Biochemistry Analyzer.The assay is comprised of two distinct phases.In phase I, a protecting reagent is added and protect LDL.And all non-LDL-lipoprotein are broken down by reaction with cholesterol oxidase and cholesterol esterase.Hydrogen peroxide produced by this reaction is decomposed by catalase in the reagent.In phase II, protecting reagent is released from LDL.The LDL-cholesterol reacts with cholesterol esterase, cholesterol oxidase, and a chromogen system to yield a blue color complex which can be measured bichromatically at 540/660 nm.The resulting increase in absorbance is directly proportional to the LDL-C concentration in the sample.The reportable range of the LDL-C method is 0.078-1113.6mmol/L.
Unbound PSCK9 in human serum sample was measured by Ligand Binding Assay.Tafolecimab was coated on microplate as capture reagent, thus unbound PSCK9 in human serum sample was captured by the immobilized tafolecimab.The reaction system yielded a yellow color complex which can be measured at 450 nm with reference at 620 nm.The reportable range of unbound PCSK9 is 24.758-1390.652ng/mL.
An electrochemiluminescent immunoassay method on the Meso Scale Discovery (MSD) platform was used to assess the antibodies against tafolecimab in human serum samples.Samples and controls were pre-incubated with biotin-tafolecimab and ruthenium-tafolecimab to pre-form anti-drug antibody (ADA) immune-complexes, which were captured via biotin on streptavidin MSD plates.After removal of unbound materials, plate-bound ADA immune-complexes were detected by ruthenium-emitted electrochemiluminescence.ADA positive samples will further be conducted with a neutralizing assay for the detection of anti-tafolecimab neutralizing antibodies.

PK/LDL-C exclusion criteria
Participants receiving tafolecimab (including those crossed over to tafolecimab in studies 4 and 6) were included in the analysis.Data with missing sampling time were also excluded from analysis.Participants with missing baseline LDL-C value were excluded from LDL-C analysis.If fewer than 15% of all patients had missing values for a given continuous covariate, this covariate was imputed based on median value of all subjects, whereas   2), in which R was the concentration of LDL-C in serum; L was free serum tafolecimab concentration; K in and K out were the production and elimination rate of LDL-C in blood, respectively; E max (stimulation maximum) was the maximum effect that tafolecimab can achieve upon K out ; and EC 50 was the tafolecimab concentration inducing 50% of E max ; gamma (γ) was a Hill coefficient.The PK/LDL-C model structure was shown in Figure 1.

PK/LDL-C model development
In addition, between subject variability, also called interindividual variability (IIV) was modeled using exponential functions.Individual parameter (θ i ) (Equation 3) was calculated by the typical value and variance (η i ) which was a symmetrically distributed, random variable, with a mean of zero and variance ω 2 .Furthermore, based on the preliminary model testing, proportional and combined variability models were tested to estimate residual error, respectively.
Stochastic Approximation for Model Building Algorithm (SAMBA) 15 was used for preliminary covariate screening.Once potential covariates were selected, Conditional Sampling use for Stepwise Approach based on Correlation tests (COSSAC) 16 was performed to identify each potential covariate using a strict modeling criterion (p < 0.001).
(1) The continuous categorical covariates were modeled by the following Equations 4 and 5: Where θ i was the ith parameter value, θ TV was the typical value, cov i was the ith covariate value, cov median was the median value of cov i , θ x was the coefficient of influence covariate on parameter.

Model validation
Model selection was based on precision of parameter estimates, statistics, and GOF.Before simulation, model verification was performed for each model by examination of the GOF plots and visual predictive check (VPC).Nonparametric bootstraps were performed to evaluate the precision of the model parameters.The original dataset was replicated 100-times using random sampling with replacement and the model was run for each replicate dataset.The mean and 90% confidence interval of each parameter estimate were calculated and compared to the original model parameters.The bootstraps were performed with the R-based Rsmlx 5.0 package.

Simulated individual parameters
Once the final model was developed, dose regimens and tafolecimab/LDL-C parameters were calculated for target population using Simulx (version 2021R1; Lixoft).PK parameters were derived from final PopPK/LDL-C model, such as maximum concentration at steady-state (C max,ss ) and area under the curve at steady-state (AUC ss ) on week 12 and percent change in LDL-C from baseline to week 24/48.Descriptive statistics of PK parameters, simulated by Empirical Bayes Estimates (EBEs), were summarized for each dose in phase III studies.

Exposure-response exploration
Based on tafolecimab concentrations at 12-week and 48-week in non-FH/HeFH populations and the corresponding LDL-C ratio to baseline, plotted the relationship of trough plasma concentration (C trough ) and efficacy.
Meanwhile the 5th to 95th percentiles of LDL-C were estimated, the population distribution of C trough also was plotted.Finally, the LDL-C distribution graph and population information were graphically superimposed together.

Final PopPK/LDL-C model
For PopPK analysis, 0.03% (3/9530) of tafolecimab concentrations were excluded from the analyses based on the predefined outlined criteria.For LDL-C modeling process, 0.35% (4/1153) participants missing baseline LDL-C observations, and 0.5% (60/12067) of LDL-C measurements were excluded from the analyses based on the outlined criteria.
Finally, 9527 serum tafolecimab concentrations from 1153 participants and 12,007 LDL-C observations from 1149 participants were included in the PopPK/LDL-C analysis.A schematic of PopPK/LDL-C model is shown in Figure 1.

Final PopPK model
A two-compartment, Michaelis-Menten, TMDD model, as previously mentioned, was the best fit for serum tafolecimab concentrations.The PopPK model was parameterized with an absorption constant (K a , 1/h), the apparent distribution volume (V, L), the inter-compartmental clearance (Q, L/h), the peripheral compartment (V2, L), the first-order apparent linear clearance (CL, L/h), two Michaelis-Menten parameters (Vm, mg/h and Km, mg/L) and the bioavailability factor (F).The IIV was modeled through an exponential error model for all PK parameters except for F, K m , Q, and V2, for which no interindividual term could be provided.A proportional error model was used to model the residual variability.After examination of the GOF plots and VPC, this model was considered as the PopPK base model.
Potential co-linearity between continuous covariates was identified in Figure S1, for which correlation between two continuous covariates was above 0.75, 17 only one of them was modeled.
Three covariates were included in the final PopPK model (Table 2), two (BBWT and SEX) on K a , three (BBWT, BPCSK9, and SEX) on V and two (BBWT and SEX) on CL, through the following equations (x_pop = typical values, beta = coefficient for the covariate effect).Both structure model and residual variability parameters were estimated with good precision, as measured by a relative standard error (RSE) of less than 5%.Covariate effects also were precisely estimated (RSE < 20%).With the inclusion of covariates in the PopPK model, omega value (indicated IIV) on K a , V, and CL was reduced (from 63% to 55%, from 47% to 34% and from 44% to 33%).
All the final model parameter estimates along with the mean and 90% confidence intervals of the non-parametric bootstrap were reported in Table 2.The population parameter estimates were similar to the mean of the nonparametric bootstrap estimates.
The final model was evaluated by GOF plot and VPC (Figure S2), both showing the adequacy of the model and none major systematic bias.The PopPK final model was well-qualified and well-characterized among Chinese non-FH/FH patients.

Final PopPK/LDL-C model
A two-step procedure was used to build the PopPK/ LDL-C model.Once all PK parameters in the PopPK final model were fixed, an indirect response model was developed to link tafolecimab to LDL-C concentrations.The PopPK/LDL-C base model was IDR model (Figure 1) which was parameterized with a first-order rate constant for elimination of LDL-C (K out ), a maximum tafolecimabinduced effect (E max ), a tafolecimab concentration inducing 50% of E max (EC 50 ), and a Hill coefficient (gamma).An exponential error model was applied to IIV for all LDL-C parameters except for gamma and EC 50 .IIV for gamma and EC 50 would not be modeled.A proportional error model was used to capture the residual variability.Model validation by GOF and VPC could be explored.
Covariates, screened as previously described, were tested in the model to quantify IIV.The relationships between the LDL-C parameters and relevant covariates were described by the following equations: BLDLC showed a pronounced influence on R0 and K out .Additionally, R0 was BLDLC actually with no exploration further.K out increased by 83.3% in subjects with fifth BLDLC (76.6 mg/dL) and decreased by 61.7% in subjects with 95th BLDLC (202.7 mg/dL), compared with subjects with median BLDLC (112.13 mg/dL).DISEASE and sex had significant influence on E max .The E max was 2.38 and 2.04 for male and female non-FH patients, respectively.Meanwhile, E max was 2.19, 1.64 and 0.11 for HVs, and HeFH and HoFH male participants, respectively.The E max for HVs and non-FH participants were similar, albeit lower for HeFH and far lower for HoFH.
The final PopPK/LDL-C parameters were summarized in Table 2.The RSE of the LDL-C parameter in PopPK/LDL-C final model was small except for β HVx , the influence of HVs on E max (104%).No important systematic deviation or major bias was observed in any of the GOF plots (PRED and IPRED vs. OBS; IWRES and PWRES vs. time or individual/population predicted values).VPC plots stratified by DISEASE showed good fitting as well (Figure S3).

Dosage regimen
Based on the final PopPK/LDL-C model, different dose regimens (including 75, 150, or 300 mg every 2 weeks; 150, 300, or 450 mg every 4 weeks; 300, 450, or 600 mg every 6 weeks) were simulated by EBEs.The LDL-C reduction by 50% or greater from baseline was identified as the target for dose selection.Dose dependent response was observed at the same interval.150 mg q2w and 450 mg q4w dosing regimens would reduce LDL-C levels by 50% or greater in 85% of patients, and 600 mg q6w dosing regimen would reduce LDL-C levels by 50% or greater in 70% of patients (Figure 2).The observed percent of patients achieving 50% or greater reduction in LDL-C levels in clinical studies (NCT04709536, NCT04289285, and NCT04179669) were up to 63.6% at 150 mg q2w, 89.1% at 450 mg q4w, and 64.6% at 600 mg q6w, which were consistent with our prediction (Table 3).The accuracy of prediction for mg q2w was relatively low because this regimen was evaluated only in patients with HeFH (Table 3).Considering the covariate effect of DISEASE on LDL-C reduction, we performed simulation for difference dosing regimens in patients stratified by non-FH/HeFH.The predicted target LDL-C  reduction attainment rate for non-FH and HeFH was 85% and 70% at 150 mg q2w, 90% and 75% at 450 mg q4w, and 75% and 65% at 600 mg q6w (Figure S4).The prediction was essentially consistent with observation despite slightly higher predicted control rate at 150 mg q2w for HeFH and at 600 mg q6w for non-FH (Table 3).The prediction accuracy was all above 85%.The recommended dose regimens proved to be effective in reducing LDL-C by above 50% in at least 75% of non-FH and 65% of HeFH populations.
The time to maximum concentration at the first dose was 4.7-7.5 days.Bioavailability was 58%.The geometric mean (geocv) of CL was 0.162 (30.4%)L/day.The geocv of t 1/2 was 26.1 (24.5%) days.The geocv of V ss was 5.7 (21.1%)L. After 12 weeks' subcutaneous administration, AR for tafolecimab 150 mg q2w, 450 mg q4w, and 600 mg q6w were 2.4, 1.49, and 1.12, respectively.For TMDD PK characteristic, both target-mediated and typical IgG elimination mechanisms contribute to the overall clearance in the wide range of drug concentration.Typical IgG elimination pathway is major clearance mechanism after 450 mg q4w administration, thus tafolecimab PKs is adequately characterized for the proposed dosing regimen (Figure S6).

Effect of covariates on tafolecimab exposure and LDL-C reduction
The effect of the three significant covariates included in final PopPK model (SEX, BBWT, and BPCSK9) was evaluated on tafolecimab exposure at different dose regimens (150 mg q2w, 450 mg q4w, and 600 mg q6w) administrated in pivotal studies (NCT04709536, NCT04289285, and NCT04179669).Meanwhile, three covariates (BLDLC, SEX, and DISEASE) were included in the final PopPK/ LDL-C model to evaluate their effect on LDL-C reduction in pivotal studies.The effect of covariates on tafolecimab exposure (C max,ss , AUC ss , and C min,ss ) and on LDL-C reduction (LDLC_ max,ss and LDLC_ min,ss ) were consistent among the three dose regimens.Thus, distributions of PK exposures and LDL-C reduction were exhibited to explain covariate effects using 450 mg q4w as the representative (Figure S5).
Increasing body weight was associated with reducing tafolecimab exposure for all three dose regimens.In patients weighing above median weight (about 70 kg), tafolecimab median AUC ss , C max,ss , and C min,ss after administrated 450 mg q4w decreased by 30%, 17.2%, and 53.2%, respectively, compared with patients weighing less than median weight.Female patients had lower (13.4%-21.7%)tafolecimab exposures than male patients.However, no obvious change of tafolecimab exposures was observed in patients with levels higher than median (811.313ng/mL) and lower than median BPCSK9.
There was no obvious effect of covariates (BLDLC and SEX) on LDL-C reduction from baseline in pivotal studies.Patients with HeFH exhibited a bit lower LDL-C reduction (almost 9%) than non_FH patients, but still maintained about 60% LDL-C reduction.
Based on the above findings, covariates of interest, such as SEX, BBWT, BPCSK9, and BLDLC, were not identified as significant factors for the clinical practice (Figure S5).

Exposure-response
The exposure-efficacy relationship was analyzed at the primary end point in the pivotal clinical studies (NCT04709536, NCT04289285, and NCT04179669) by C trough,12/48 (C trough at week 12/48) and LDL-C ratio to the baseline.There was a clear exposure-efficacy relationship between tafolecimab trough concentrations and LDL-C response at week 12/48 (Figure 3).The shapes of the curves were similar, irrespective of the disease category or study duration, with the nadir in LDL-C reduction achieved with tafolecimab concentration of 5 μg/mL.The relationship suggested that increasing exposure might not further decrease LDL-C levels.
The exposures-safety relationship was explored between tafolecimab exposure (C max,ss and AUC ss by EBEs simulation) and adverse events (AEs) including treatment-emergent AE, all grade treatment-related AE (TRAE), moderate or higher TRAE summarized in Integrated Safety Summary.Table 4 shows the frequency of AEs by quartile of exposure.There was no clinically meaningful safety signal with increasing exposure.

DISCUSSION
A two-compartment Michaelis-Menten TMDD model, parameterized with a first-order absorption process and two elimination processes, was developed and qualified in a dataset of 1153 HVs and patients with hypercholesterolemia enrolled in phase I, II, and III studies.This analysis showed that the PKs of tafolecimab can be accurately predicted using this model.Because tafolecimab exhibited TMDD characteristics and showed different  disposal processes at different concentration ranges, the data in the low concentration range can better capture the characteristics of rapid clearance of tafolecimab in TMDD, wide concentration range would fully characterize different rates of clearance.Thus, dose escalation studies (NCT03366688 NCT03815812) were crucial to provide comprehensive information for PopPK modeling.
In addition, the full TMDD model was often over-parameterized and difficult to converge, while the Michaelis-Menten TMDD model had the same ability of curve fitting as the full TMDD model, but it converged efficiently. 18inally, the two-compartment Michaelis-Menten TMDD model, was developed and qualified to model free tafolecimab.Enough intensive PK sampling was conducted in two dose escalation studies.Both intravenous infusion and subcutaneous injection were adopted in the study of NCT03366688, so F was calculated based on the HVs' abounded absorption information.Simultaneously, the estimation of nonlinear clearance parameter, K m , highly relied on low dose level PK data.Thus, F and K m were fixed based on evaluation in a previous analysis with the phase I and phase II data only.Because of largely sparse PK sampling conducted in phase III studies, large variability was observed in final PopPK modeling, parameters of Q and V2 were fixed as the estimations of an interim model (NCT03366688, NCT03815812, and NCT04709536 data included).This model also enabled identification of covariates that explained part of the individual PK/LDL-C variability of the drug.All potential covariates were evaluated in PopPK/LDL-C model.Finally, SEX, BBWT, and BPCSK9 were significant covariates of tafolecimab PK parameters, and SEX, DISEASE, and BLDLC had significant impact on LDL-C related parameters.Commonly, STATIN would affect production of PCSK9, increase target-mediated clearance, leading to reduced tafolecimab exposure.However, among tafolecimab clinical trials, almost all of patients co-administrated with moderate or above intensity of statins.On the other hand, the impact of upregulated PCSK9 caused by co-administrated statin on PK parameters may be explained by baseline PCSK9.
With limited participants without co-administration with statins included in modeling and significant covariate of BPCSK9 being involved, STATIN was not included in the final model.
The impact of significant covariates on tafolecimab exposures or LDL-C reduction were evaluated.Weight was the greatest factor that influence tafolecimab exposure.Increased body weight would decrease tafolecimab exposure, but the influence on LDL-C reduction was controllable.Sex was a significant covariate on PK parameters, female subjects had lower tafolecimab exposure, but the impact of sex on LDL-C reduction was not obvious.BPCSK9 was a significant covariate, but the impact of this covariate on PK parameters was very limited, which was translated into a small change on tafolecimab exposure and little clinically meaningful difference in LDL-C reduction.Therefore, no dose adjustments were recommended based on body weight, sex, and BPCSK9.DISEASE was a significant covariate of LDL-C reduction but not of tafolecimab exposures.LDL-C reduction was maintained by greater than 70% in non-FH, compared with a slightly lower reduction of 60% in patients with HeFH.This discrepancy may result from the difference in pathological mechanism between population phenotypes, and exposure-response (E-R) relationship revealed that increased PK exposure would not assist in improving efficacy.Therefore, no dose adjustment was recommended based on disease phenotype.
Long-term medication was crucial for patients with hypercholesterolemia.The E-R relationship fully supported the long-term efficacy and safety of the current dose regimens.

CONCLUSION
The PK/LDL-C time course data for tafolecimab was welldescribed by a TMDD/IDR model.The PopPK/LDL-C model for tafolecimab was successfully developed, validated, and predicted tafolecimab/LDL-C concentrations and individual exposures.Baseline factors resulting in statistically significant changes in tafolecimab PK/LDL-C parameters included: body weight, sex, unbound PCSK9, baseline LDL-C, and disease.However, their impact on tafolecimab exposure did not translate into any clinically meaningful difference in efficacy or safety.Therefore, no dose adjustments are recommended according to patients' weight, sex, and baseline factors.
hypercholesterolemia[HeFH] and non-familial hypercholesterolemia [non-FH]) and mixed dyslipidemia who have failed to achieve LDL-C goals despite moderate or higher doses of statins, to reduce LDL-C, total cholesterol, and ApoB levels.One of objectives of this analysis is to develop a validated population pharmacokinetic (PopPK)/LDL-C model to characterize individual tafolecimab PK profile and predict LDL-C levels in both Chinese healthy volunteers and patients with hypercholesterolemia based on pooled data from phase I-III clinical studies.Potential covariates including participants' demography, renal/hepatic function, concomitant use of other lipid-lowering drugs, disease (healthy volunteers and different type of hypercholesterolemia), anti-drug antibody, baseline PCSK9, and LDL-C levels were assessed.The effect of these covariates was investigated by a two-compartment Michaelis-Menten approximate target-mediated drug disposition (TMDD) model joint indirect response (IDR) model for tafolecimab PKs and LDL-C.Ultimately, the individual predicted exposures and LDL-C levels were estimated by the final PopPK/LDL-C model.

A
stochastic approximation expectation maximization (SAEM) algorithm (MONOLIX, version 2021R1; Lixoft, Antony, France) was applied to approximate the likelihood of the nonlinear mixed-effect model.A simulated annealing version of SAEM was used to estimate population parameters.The conditional means and standard deviations for the individual parameters were estimated by Markov chain Monte Carlo simulation.The concentrations below the lowest quantification (BLQ) were handled according to M4 method,14 and the probability of BLQ at each sampling timepoint was incorporated into likelihood calculation.Depending on precision of parameter estimates, goodness-of-fit (GOF) and statistics, the model was developed.Several potential structure models were investigated, including two-compartment linear or Michaelis-Menten elimination, or a combination of both.A two-compartment TMDD model was the best fit for serum tafolecimab concentrations (Equation1).Estimated parameters were absorption rate constant (K a ), central volume (V), peripheral volume (V2), bioavailability (F), clearance (CL), and inter-compartmental clearance (Q).The PopPK/LDL-C model was built by two steps, first, the PopPK model to find appropriate typical values, and then the LDL-C model by IDR or turnover model after estimated PopPK parameters were fixed.The relationship between tafolecimab concentrations and changes in LDL-C were described using the IDR model (Equation

F I G U R E 1
Schematic of the PopPK/LDL-C model.Vm and Km were maximum elimination rate and Michaelis-Menten constant, respectively.CL, clearance; Dose, dose administration; EC50, the tafolecimab concentration inducing 50% of E max ; E max , a maximum drug-induced effect; F, bioavailability; IV, intravenous infusion; Ka, absorption constant; K in , production rate of LDL-C; K out , a first-order rate constant for loss of response; PopPK, population pharmacokinetic; Q, inter-compartmental clearance; R0, baseline LDL-C concentration; SC, subcutaneous; V, distribution volume of central compartment; V2, distribution volume of peripheral compartment.

F I G U R E 2
Projected LDL-C percentage of baseline with different dose intervals (y-axis LDL-C/Baseline LDL-C*100%, black line the median of LDL-C ratio, blue range 5th to 95th percentiles of all population LDL-C ratio, red line reference line of 70% and 50%, respectively.Top line: LDL-C results at 75, 150, and 300 mg, every 2 weeks dosing regimens; middle line: LDL-C results at 150, 300, and 450 mg, every 4 weeks dosing regimens; bottom line: LDL-C results at 300, 450, and 600 mg, every 6 weeks dosing regimens).T A B L E 3The LDL-C result at week 12 in pivotal studies.

F
I G U R E 3 Exposure-efficacy relationships for tafolecimab trough concentrations at week 12/48 and LDL-C ratio from baseline in pivotal studies.(a) Trial NCT04179669, HeFH, (b) Trial NCT04709536, non-FH/HeFH, (c) Trial NCT04289285, non-FH.Mean LDL-C and the range of 5th to 95th percentiles at the corresponding median C trough were shown for 2.5-3.5 bin with in three pivotal trials by the blue solid lines and shaded region.Orange lines (solid and broken lines) depict the distribution of tafolecimab C trough with each respective dosing regimens.C trough , trough plasma concentration; HeFH, heterozygous familial hypercholesterolemia; non-FH, non-familial hypercholesterolemia.

T A B L E 4
Exposure-safety relationship for tafolecimab exposures and safety based on ISS.
For LDL-C modeling, there was an in vivo balance between production and elimination of LDL-C, and tafolecimab could accelerate the elimination of LDL-C indirectly.Therefore, an IDR model was developed to characterize LDL-C parameters.All relevant parameters in the PopPK/ LDL-C model had been estimated with reasonable precisions, and the model had been validated by GOF and VPC.The validation showed that individual PK and LDL-C curves fitted well.The final PopPK/LDL-C model allowed characterization of tafolecimab PK/LDL-C properties in the target population as well as individual exposures and LDL-C levels.

T A B L E 1 Summary of clinical studies. Study number N a Route of administration Dose Dosing interval PK samples b Study population Background medication
a Number of subjects/patients.b Number of sample timepoints per patient/group.baseline LDL-C (BLDLC) and BPCSK9 based on median value of the same study population.No continuous covariate had more than 15% missing values.If categorical covariates were missing, a new category was set combined with other categories.