2021年2月5日

海洋之神优惠大厅主站云课堂 | 马格利单抗（Hu5F9-G4，5F9）在实体瘤和淋巴瘤患者中的群体药代动力学研究

海洋之神优惠大厅主站拥有一支具有平均十几年行业经验的临床药理团队，为中美新药研发公司提供计算机建模和模拟、SAS/R临床编程、动物和临床数据分析及报告服务。本次课堂将通过海洋之神优惠大厅主站临床药理团队带头人王冰博士参与的项目为您带来案例剖析。

01 试验剖析

马格利单抗的基本原理

马格利单抗（Magrolimab）是靶向CD47的人源化单克隆抗体，可作为免疫检查点抑制剂隔断肿瘤细胞与免疫抑制受体的接触，使巨噬细胞能够识别并吞噬肿瘤细胞，并激活抗肿瘤T细胞应答。马格利单抗被开发为用于实体瘤和血液系统恶性肿瘤的新型治疗候选药物，并已获得FDA突破性疗法的称号，用于治疗高危骨髓增生异常综合征（MDS）患者。

  • Magrolimab (aka Hu5F9-G4 or 5F9) is a 1st-in-class, investigational, humanized monoclonal antibody (mAb) targeting cluster of differentiation (CD)–47, which blocks the “don’t eat me” signal used by cancer cells to avoid immune surveillance
  • Magrolimab is being developed as a novel therapeutic candidate for solid tumors and hematologic malignancies, and has been granted the FDA breakthrough therapy designation for treatment of higher risk myelodysplastic syndrome (MDS) patients

本临床试验的目的

建立马格利单抗单独和联合疗法在多种肿瘤人群中的群体药代动力学（Population PK）模型；

分析疾病和患者特点对马格利单抗药代动力学的影响。

  • To develop a population-pharmacokinetics (Pop-PK) model of magrolimab as a monotherapy and under different combination treatments across single and multiple doses in a variety of tumor populations
  • To quantitate the effect of patient/disease characteristics on PK

试验方法

群体药代动力学模型使用的数据来自于三个临床一期和临床1b/2期测试，共有222名患者参与并接受静脉注射0.1-45 mg/kg的马格利单抗或与利妥昔单抗的联合疗法。SCI-CD47-001和SCI-CD47-002是两个临床一期试验，参与者分别为88名晚期实体恶性肿瘤患者和20名血液系统恶行肿瘤患者；5F9003是临床1b/2期试验，114名复发/难治性B细胞霍奇金淋巴瘤（NHL）接受了马格利单抗和利妥昔单抗联合疗法，或利妥昔单抗与化疗结合疗法。群体药代动力学模型使用了非线性混合效应模式 (NONMEM® 7.4, ICON plc, Dublin, Ireland) 。在全模型的基础上作协变量分析后得到最终模型，并通过模拟的方法核实最终模型对5F9003试验中的急性髓性白血病（AML）/MDS患者体内血药浓度的预测能力。

  • A preliminary PopPK model was previously developed based on preliminary data1
  • In all, 4245 PK observations from 222 patients in 3 studies provided data following magrolimab dosing ranging from 0.1 to 45 mg/kg IV for PopPK model development
    • SCI-CD47-001: 1st-in-human, Phase 1, dose-escalation trial of magrolimab in patients with advanced solid malignancies2
    • SCI-CD47-002: Phase 1, dose-escalation trial of magrolimab in patients with hematologic malignancies
    • 5F9003: Phase 1b/2 trial of magrolimab in combination with rituximab or rituximab + chemotherapy in patients with relapsed/refractory B-cellnon-Hodgkin lymphoma (NHL)
  • PopPK was performed using a nonlinear, mixed-effects modeling approach (NONMEM® 7.4, ICON plc, Dublin, Ireland)
  • Full-model approach was used for covariate analysis
  • Simulation based on final PopPK model was used to verify predictive ability of the model for PK in acute myeloid leukemia
  • (AML)/MDS patients from Study 5F9005 (NCT03248479)

数据结果

  • Only quantifiable concentrations were included in the PopPK analysis
    • 599 concentrations below limit of quantitation were excluded
    • 244 observations with missing concentration or missing actual elapsed times were excluded
  • 26 PK concentrations considered outliers were excluded (conditional weighted residuals [CWRES] >6) after modeling
  • Magrolimab PK was best characterized by a 2-compartment model with linear and nonlinear CL (Michaelis-Menten equation) in parallel
    • A time-varying Vmax was used to describe the PK difference between Weeks 1 and 2 likely due to the irreversible pruning of CD47 receptors from red blood cells (RBCs) in the 1st week
    • Vmax was reduced by ~95% between Weeks 1 (priming dose week) and 2
    • The model did not identify a significant CL change over time after initial reduction3
  • Goodness-of-fit plot, and population and individual predictions suggested the model well described the observed data
  • All PK parameters and variabilities were precisely estimated and physiologically plausible for a mA
    • Mean estimates (RSE) of linear CL and Vc were 0.29 L/d (6.4%) and 4.66 L (6.6%), respectively
    • Variability in PK parameters was moderate
  • Covariate analysis identified WT as the only statistically significant covariate on both CL and Vc; other demographic covariates, including antidrug antibody (ADA) and coadministration of rituximab were not significant
  • Simulations based on PopPK model using 5F9005 dosing regimens were consistent with observed PK data, suggesting no PK difference in AML/MDS patients

结论

群体PK结果

  • 拟合优度图，总体和个人的模型预测表明，并行的线性与非线性清除的二室模型很好地反映了马格利单抗的非线性体内过程特征。
  • 随时间变化的Vmax用于解释CD47受体在红细胞（RBC）上的不可逆修剪，并很好地描述第1周和第2周之间的PK差异。对比第一周，Vmax在第二周时降低了约95%。这表明了RBC是抗原库的来源，比其他CD47表达细胞（包括白细胞）约高20倍。
  • 从第二周后，PK特性并未显示出时间依赖性，因此并不对稳态药物浓度有影响。
  • 模型估量的消除率（CL）和中央室的分布量（VC）分别为0.29 L/d和4.66 L，在单克隆抗体PK的消除率和中央室分布量的预期范围内。在最初的下降后，消除率并未表现出显著的随时间的变化。
  • 协变量分析结果表明，体重是唯一对消除率和中央室分布量有显著影响的协变量。然而体重对PK的影响并不具备临床意义，因此马格利单抗可统一剂量。
  • 在单独的马格利单抗疗法，和马格利单抗与阿扎胞苷或利妥昔单抗的联合疗法下，PK模型预估的参数值都很相似，表明没有明显的药物间相互作用。  
  • Magrolimab exhibited nonlinear PK; the PopPK analysis adequately described magrolimab PK using a 2-compartment model, with parallel linear and nonlinear CL due to target-mediated drug disposition
    • Time-varying Vmax was implemented to account for the irreversible pruning of CD47 receptors from RBCs and well described the observed PK at Week 1, indicating that RBCs are a source of antigen sink ~20-fold higher than other CD47-expressing cells, including white blood cells
    • Time-dependent PK was not observed after Week 2 and thus had no impact on steady-state drug concentrations
    • The model-estimated linear CL (0.29 L/d) and Vc (4.66 L) were within the expected ranges for PK of mAbs; variability in PK parameters could be considered moderate4
  • Baseline demographic variables such as age, sex, race, hepatic/renal function, ADA, coadministered drugs (rituximab and azacitidine), and tumor type (solid tumor, CRC, NHL, and AML) were not statistically significant covariates of magrolimab PK
    • Adjustment doses for any of these subpopulations are not required
    • Magrolimab PK parameters were similar whether administered as monotherapy (SCI-CD47-001 and SCI-CD47-002) or in combination with azacitidine (5F9005) or rituximab (5F9003), indicating lack of drug-drug interactions with the coadministered compounds
    • ADAs were observed at low frequencies (<10% overall) and did not have an impact on magrolimab exposure
  • WT was identified as the only statistically significant covariate on both CL and Vc in the covariate analysis; the impact of WT on PK is considered not clinically significant and thus flat doses may be possible for magrolimab
  • An external predictive check of the model was performed with observed PK data in AML/MDS patients from Study 5F9005 and confirmed the predictive capability of the model
  • The model is being used to optimize the dosing schedule in future pivotal trials of magrolimab in AML/MDS and other patient populations

参考文献

  • Agoram B, et al. J Clin Oncol 2018;36(suppl):2525;
  • Sikic BI, et al. J Clin Oncol 2019;37:946-53;
  • Liu C, et al. Clin Pharmacol Ther 2017;101:657-66;
  • Dirks NL, Meibohm B. Clin Pharmacokinet 2010; 49:633-59.

Population-Pharmacokinetics of Magrolimab (Hu5F9-G4, 5F9) in Patients With Solid Tumors and Lymphomas

Denise Chao-Yu Jin,1 Bing Wang,2 Branimir Sikic,3 Ranjana Advani,3 Mark Chao,1 Chris Takimoto,1 Balaji Agoram1 — 1Gilead Sciences, Inc., Foster City, CA; 2Amador BioScience, Pleasanton, CA; 3Stanford Medicine School of Medicine, Stanford, CA