Pharmacokinetics of Odronextamab, A Bispecific T‐Cell‐Engaging Antibody, in Adult Patients With Relapsed or Refractory B‐Cell Non‐Hodgkin Lymphoma

ABSTRACT

Non‐Hodgkin lymphoma (NHL) is the fifth most common malignancy and accounts for 5% of all cancers in the US, with the largest proportion being B‐cell CD20 positive NHL. Odronextamab is a CD20xCD3 IgG4 bispecific T‐cell‐engaging monoclonal antibody under development for the treatment of relapsed or refractory (R/R) B‐NHL. The objectives of this analysis were to characterize the pharmacokinetics (PK) of odronextamab in adult patients, and elucidate sources and correlates of variability. PK data of 507 patients with R/R B‐NHL from ELM‐1 (NCT02290951, Phase I; n = 167) and ELM‐2 (NCT03888105, Phase II; n = 340) were analyzed. Odronextamab concentration–time profiles following intravenous administration of 0.03 mg to 320 mg doses were described by a bi‐exponential decline with parallel linear (first‐order) and non‐linear (Michaelis–Menten) elimination processes. The modified Michaelis–Menten or target‐mediated elimination was not only concentration–dependent but also time‐dependent. A reduction in target‐mediated clearance over time suggests a reduction in target abundance to a larger extent than associated with concentration alone, which is consistent with the treatment‐induced depletion of the B cells observed in patients who underwent assessment. Linear clearance (CL) and steady‐state volume of distribution were 0.189 L/day and 9.41 L, respectively. Target‐mediated clearance was ~5 L/day at baseline, with an asymptote of ~0.03 L/day at steady state. With the largest covariate effect on odronextamab exposure, baseline body weight was directly correlated with CL and volume of distribution, albumin was inversely correlated with CL and volume of distribution, and baseline interleukin‐10 was inversely correlated with CL.

Study Highlights

• What is the current knowledge on the topic?

  • ○
    Non‐Hodgkin lymphoma (NHL) is the fifth most common malignancy in the US. Odronextamab is a CD20xCD3 IgG4 bispecific T‐cell engager monoclonal antibody under development for the treatment of relapsed or refractory (R/R) B‐cell NHL.
• What question did this study address?

  • ○
    Objectives of this analysis were to characterize the pharmacokinetics (PK) of odronextamab in serum of adult patients with R/R B‐NHL, and elucidate sources and correlates of variability.
• What does this study add to our knowledge?

  • ○
    Odronextamab concentration–time profiles were described by a bi‐exponential decline with parallel linear and target‐mediated elimination processes. A decline in target‐mediated clearance over time suggests a temporal reduction in target abundance, consistent with the treatment‐induced depletion of B cells observed in patients with measurable B cells pre‐treatment. Body weight had the largest effect on odronextamab exposure, followed by albumin and IL‐10. Nevertheless, identified covariate effects on odronextamab exposure are not considered clinically relevant.
• How might this change drug discovery, development, and/or therapeutics?

  • ○
    Characterization of PK and predictors of variability helped inform the odronextamab dose regimen in late‐phase clinical trials.

1. Introduction

Non‐Hodgkin lymphoma (NHL) comprises a heterogeneous group of malignancies that arise from B cells or T cells and have lymphoid tissue characteristics. The largest proportion (90%) are of B‐lymphoid origin [1], the majority of which (> 95%) express CD20 on the cell surface [2]. Overall, NHL is the fifth most common malignancy and accounts for 5% of all cancers in the US [3].

Odronextamab (international nonproprietary name for REGN1979) is a hinge‐stabilized, fully human, immunoglobulin G (IgG) 4‐based, CD20xCD3 IgG4 bispecific T‐cell engager monoclonal antibody under clinical development by Regeneron Pharmaceuticals Inc. for the treatment of relapsed or refractory (R/R) B‐cell NHL (B‐NHL) [4]. It is designed to bind to CD20‐expressing target B cells and cross‐link them to CD3‐expressing T cells, generating a polyclonal antigen‐nonspecific cytotoxic T‐cell response. This mechanism of action differs from anti‐CD20 antibodies (e.g., rituximab), and is evident to provide therapeutic benefit to patients who are R/R from previous anti‐CD20 antibody treatment [4, 5, 6]. Odronextamab has been evaluated in a Phase I ELM‐1 study (NCT02290951) as a monotherapy in adult patients with CD20+ B‐cell malignancies who were previously treated with an anti‐CD20 antibody [4], as well as in a Phase II ELM‐2 study (NCT03888105) as a monotherapy in adult patients with R/R B‐NHL [5, 6].

The objectives of this analysis were to characterize the pharmacokinetics (PK) of odronextamab in adult patients with R/R B‐NHL from studies ELM‐1 and ELM‐2 to elucidate sources and correlates of variability in odronextamab concentrations, including the influence of intrinsic and extrinsic factors, and to generate post hoc metrics of odronextamab exposure.

2. Methods

2.1. Data

This population PK analysis used data from adult patients with R/R B‐NHL enrolled in the Phase I study (ELM‐1) or Phase II study (ELM‐2), who had received any amount of odronextamab and had at least one post‐baseline assessment of odronextamab concentration in serum. The studies were in accordance with the ethical standards of the institutional review board. Study designs of ELM‐1 and ELM‐2 have been reported previously [4, 5, 6, 7], and are summarized below.

ELM‐1 is a single‐arm, global, Phase I study including dose‐escalation and dose‐expansion phases in adult patients with R/R B‐NHL. It evaluated the safety, tolerability, PK, and anti‐tumor activity of odronextamab monotherapy. Patients received intravenous (IV) infusion of odronextamab, with doses ranging from 0.03 mg to 320 mg. A step‐up dosing schedule was implemented followed by treatment doses once weekly (QW) ranging from 0.1 mg to 320 mg in dose escalation. In expansion, maintenance treatment continued every 2 weeks (Q2W) after the QW treatment until disease progression or unacceptable toxicity was reached. If patients had a complete clinical response for 9 months in the post CAR T cohort, they had the option to receive odronextamab once every 4 weeks (Q4W) [8, 9]. An assessment of PK, efficacy, and safety data from the dose escalation informed the dosing regimens studied in the expansion phase of ELM‐1 and in ELM‐2. Three expansion cohorts of ELM‐1 were opened: (1) diffuse large B‐cell lymphoma (DLBCL) post‐CAR T therapy at 160 mg QW followed by 320 mg Q2W; (2) aggressive lymphoma (excluding patients with prior CAR T therapy) at 160 mg QW followed by 320 mg Q2W; (3) follicular lymphoma (FL) Grade 1–3a at 80 mg QW followed by 160 mg Q2W.

ELM‐2 is a multi‐cohort, global, Phase II, open‐label study in adult patients with R/R B‐NHL, to assess the anti‐tumor activity and safety of odronextamab monotherapy. There were five disease‐specific cohorts: (1) R/R FL Grade 1–3a after ≥ 2 prior lines of systemic therapy, including an anti‐CD20 antibody and an alkylating agent; (2) R/R DLBCL after ≥ 2 lines of systemic therapy, including an anti‐CD20 antibody and an alkylating agent; (3) R/R mantle cell lymphoma (MCL) following or with failure to tolerate Bruton's tyrosine kinase inhibitor therapy; (4) R/R marginal zone lymphoma (MZL) after ≥ 2 lines of systemic therapy; (5) other R/R B‐NHL subtypes, excluding Waldenström macroglobulinemia, after ≥ 2 lines of systemic therapy. Patients received IV infusion over 1–4 h. The regimens included step‐up doses prior to administration of QW treatment doses until study Week 12, followed by maintenance doses Q2W until disease progression or treatment discontinuation. For indolent B‐NHL (i.e., FL and MZL), treatment and maintenance doses were 80 mg QW and 160 mg Q2W, respectively; for aggressive B‐NHL (i.e., DLBCL, MCL, and other B‐NHL), treatment and maintenance doses were 160 mg QW and 320 mg Q2W, respectively. All patients with durable complete responses of 9 months could transition from Q2W to Q4W dosing [7].

2.2. Model Building

2.2.1. Base Model Structure

Development of a base structural model that described the concentration–time profile of odronextamab in serum leveraged prior knowledge on: (1) typically reported PK of monoclonal antibodies [10]; (2) odronextamab mechanism of action; (3) a preliminary odronextamab PK model, based upon data from 255 adult patients with B‐NHL enrolled in ELM‐1 or ELM‐2.

Odronextamab concentration–time profiles following IV administration of 0.03–320 mg doses were described by a bi‐exponential decline with parallel linear (first‐order) and non‐linear (Michaelis–Menten) elimination processes (Figure 1). The Michaelis–Menten or target‐mediated elimination—typically only concentration–dependent—was modified to also include a time‐dependent component. The model was formulated by a system of non‐linear ordinary differential equations:

$$\frac{\mathrm{d}{A}_{\text{central}}}{\mathrm{d}t}=-K_{\mathrm{MM}}·A_{\text{central}}-K·A_{\text{central}}-K_{\mathrm{cp}}·A_{\text{central}}+K_{\mathrm{pc}}·A_{\text{peripheral}}$$ (1)

$$\frac{\mathrm{d}{A}_{\text{peripheral}}}{\mathrm{d}t}=K_{\mathrm{cp}}·A_{\text{central}}-K_{\mathrm{pc}}·A_{\text{peripheral}}$$ (2)

$$\frac{\mathrm{d}{V}_{\max }}{\mathrm{d}t}=-K_v·V_{\max }·\left(1-\frac{V_{\max ,\text{asym}}}{V_{\max }}\right)$$ (3)

where $A_{\text{central}}$ is the amount of odronextamab in the central compartment with a volume of distribution $V_c$, $A_{\text{peripheral}}$ is the amount of odronextamab in the peripheral compartment with a volume of distribution $V_p$, $K$ is the first‐order odronextamab elimination rate constant, $K_{\mathrm{cp}}$ is the first‐order rate constant describing the transfer of odronextamab from the central to the peripheral compartment, $K_{\mathrm{pc}}$ is the first‐order rate constant describing the transfer of odronextamab from the peripheral to the central compartment, $V_{\max }$ is the maximum amount of odronextamab eliminated via target per unit of time and volume, $K_v$ is the rate constant describing the decline in $V_{\max }$ per unit of time, $V_{\max ,\text{asym}}$ is the minimum value of $V_{\max }$ achieved over time, $K_{\mathrm{MM}}=\frac{V_{\max }}{k_m+\frac{A_{\text{central}}}{V_c}}$ and describes the modified Michaelis–Menten odronextamab elimination, where $k_m$ is the odronextamab concentration in serum reaching 50% of $V_{\max }$. The structural model was parameterized in terms of:

• $V_c$
• $V_p$

• Intercompartmental clearance, $Q$, calculated as $K_{\mathrm{cp}}·V_c=K_{\mathrm{pc}}·V_p$

• Linear clearance, $\mathrm{CL}$, calculated as $K·V_c$

• Baseline $V_{\max }$ or $V_{\max ,0}$

• $K_v$

• $R_v$, calculated as $\frac{V_{\max ,\text{asym}}}{V_{\max ,0}}$

• $k_m$

FIGURE 1.

Schematic of the PK model for odronextamab. K _cp is the first‐order rate constant describing the transfer of odronextamab from the central compartment (plasma) to the peripheral compartment (tissue); K _pc is the first‐order rate constant describing the transfer of odronextamab back from the peripheral to the central compartment. Michaelis–Menten elimination, which is typically only concentration–dependent, was modified to include a time‐dependent component. PK, pharmacokinetics; V _c, volume of distribution in the central compartment; V _p, volume of distribution in the peripheral compartment.

The modified Michaelis–Menten clearance, $\mathrm{C}{\mathrm{L}}_{\mathrm{MM}}$, was defined as $K_{\mathrm{MM}}·V_c$, and odronextamab total clearance was defined as $\mathrm{C}{\mathrm{L}}_{\mathrm{tot}}=\mathrm{C}{\mathrm{L}}_{\mathrm{MM}}+\mathrm{CL}$.

Between‐subject variability (BSV) and residual variability were assumed to be log‐normally distributed.

2.3. Full Covariate Model

Taking the base model as a starting point, the full model included pre‐specified covariates based on pathophysiological plausibility or previously reported supportive evidence. Rationale for their selection is presented in Table S1. For evaluated continuous covariates, the proportion of missing values was less than 10%. For categorical covariates, evaluated subgroups had a proportion of at least 10%. Among pre‐specified covariates, CD20+ B and CD3+ T lymphocyte counts as well as CD3+ CD4+ and CD3+ CD8+ T‐cell counts were missing in approximately 50% of the patients, and hence not evaluated. Covariate effects used baseline covariate values except for the time‐varying effect of albumin on CL. For instance, IgG levels over time were not collected as often as albumin levels.

Evaluated pre‐specified covariate effects were:

• Albumin on CL, V _c , and V _p

• B‐NHL subtypes on parameters describing the time course of V _max (i.e., V _max,0, K _v , and R _v )

• Body weight on CL, V _c , and V _p

• IgG on CL

• Lymphocyte count on V _c , V _p , and V _max,0

• Prior CAR T therapy on V _max,0

• Tumor size on V _max,0.

2.4. Fit‐For‐Purpose Model

Taking the full model as a starting point, the fit‐for‐purpose model included statistically significant exploratory/ad hoc covariates, which were evaluated via forward inclusion followed by backward elimination using a significance level of 0.001 (i.e., equivalent to a reduction in the objective function value [OFV] of at least 10.8 points for 1 degree of freedom). All covariate effects used baseline covariate values.

Exploratory or ad hoc evaluated covariate effects were:

• Age on CL

• Alanine aminotransferase (ALT) on CL

• Ann Arbor stage on V _max,0

• Race on CL

• Bulky disease status on V _max,0

• C‐reactive protein on CL

• Creatinine clearance and renal function category on CL

• Eastern Cooperative Oncology Group (ECOG) score on V _max,0

• Sex on CL, V _c , and V _p

• IgG on V _c and V _p

• Interleukin (IL)‐2, IL‐6, IL‐8, IL‐10, interferon (IFN) gamma, tumor necrosis factor alpha on CL

• Lactate dehydrogenase on CL.

Finally, pre‐specified covariate effects were removed from the model if the 95% confidence interval (CI) of the effect on the PK parameter included null value and if the entire 95% CI of its marginal effect on odronextamab exposure was contained within a 80%–125% range relative to the reference.

2.5. Software

The NONMEM program version 7.5.0 was used to fit the models using first‐order conditional estimation with interaction. While visual predictive check (VPC) simulations were conducted in NONMEM, all others were conducted in mrgsolve (R package version 1.0.0) [11]. Data wrangling and visualization were performed using the R Statistical Software (version 4.1.2) [12]. Forest plots of marginal covariate effects were generated in coveffectsplot (R package version 1.0.0) [13].

2.6. Model Evaluation

Model evaluation included inspection of goodness‐of‐fit graphs and simulation‐based assessments. Parameter precision was obtained from the NONMEM $COV step which estimates the variance–covariance of the population parameters, with the square root of which being the standard errors [14]. Additionally, the model was fitted to 300 bootstrap samples drawn with replacement. Ill‐conditioning of the parameter space was assessed by inspecting the eigenvalues of the covariance matrix. VPCs were generated to evaluate the model performance.

2.7. Simulations

Simulations of different nature and objectives were performed using the fit‐for‐purpose model:

• Simulations without uncertainty in fixed‐effect parameters with BSV, identical patients' covariates and dose schedule to analysis dataset (500 simulations), to assess the model performance via VPC. Concentration–time profiles were generated and 5th, 50th, and 95th percentiles of the observed data were compared with the respective percentiles of the simulated data. The 95% CI of each simulated percentile was approximated with the 2.5th to 97.5th inter‐percentile range of the simulated concentrations at each time point.

• Simulations with uncertainty in fixed‐effect parameters without BSV [15] (500 simulations), to interpret the effect of covariates on odronextamab exposure after administration of the 0.7/4/20/160 mg QW/320 mg Q2W dose regimen. Forest plots illustrated the effect of covariates on steady‐state area under the concentration–time curve (AUC_ss), maximum concentration (C _max,ss), and minimum concentration (C _min,ss) as fold change relative to a typical patient (reference). One covariate was varied at a time keeping all the others at the reference values (marginal covariate effects). The distribution of the BSV given covariates at their reference values was also presented for comparison with the estimated covariate effects. Steady‐state metrics were calculated using drug exposure at Weeks 42–43.

• Simulations using post hoc estimates of individual PK parameter estimates to predict odronextamab concentration–time profiles and summarize exposure metrics stratified by patient characteristics after administration of the 0.7/4/20/80 mg QW/160 mg Q2W or 0.7/4/20/160 mg QW/320 mg Q2W dose regimen. Time to steady state, time to wash‐out, and ratio of odronextamab accumulation were determined for recommended dosing regimens for FL and DLBCL. Moreover, using the actual doses received over time by each patient, simulations were conducted to generate individual predictions of odronextamab concentration at granular time points (every 0.01 day or ~14 min), to allow accurate determination of exposure metrics.

3. Results

3.1. Data Summary

There was a total of 507 patients with R/R B‐NHL from the ELM‐1 (n = 167) or ELM‐2 study (n = 340). Most patients were diagnosed as having either R/R DLBCL or R/R FL (52% and 33%, respectively). The remaining patients had either R/R MCL, R/R MZL, or other type of R/R B‐NHL lymphoma with a prevalence of less than 10% each. Most of the patients were White (62%), male (62%), with a median age of 65 (interquartile range [IQR]: 55, 72) years, body weight of 72.6 (IQR: 61.4, 84) kg, ECOG score of 1 (60%), Ann Arbor stage of 4 (62%), and albumin in serum of 38.7 (IQR: 34, 42) g/L. Baseline patient characteristics by study are presented in Table 1 and by B‐NHL type are presented in Table S2. Graphical representation of the observed data stratified by dose regimen and grouped by B‐NHL type is presented in Figure S1.

TABLE 1.

Baseline patient characteristics by study.

	ELM‐1	ELM‐2	Overall
	(N = 167)	(N = 340)	(N = 507)
NHL type
DLBCL CAR T naive	49 (29.3%)	153 (45.0%)	202 (39.8%)
DLBCL CAR T failure	59 (35.3%)	0 (0%)	59 (11.6%)
FL	40 (24.0%)	126 (37.1%)	166 (32.7%)
MCL	11 (6.6%)	14 (4.1%)	25 (4.9%)
MZL	6 (3.6%)	19 (5.6%)	25 (4.9%)
Other B‐NHL	2 (1.2%)	28 (8.2%)	30 (5.9%)
Sex
Female	52 (31.1%)	142 (41.8%)	194 (38.3%)
Male	115 (68.9%)	198 (58.2%)	313 (61.7%)
Age (years)
Median [IQR]	66.0 [58.0, 71.5]	63.0 [53.0, 72.0]	65.0 [55.0, 72.0]
5th, 95th percentiles	39.0, 82.0	38.0, 80.0	38.3, 81.0
Body weight (kg)
Median [IQR]	78.8 [66.5, 89.7]	69.2 [59.8, 80.6]	72.6 [61.4, 84.0]
5th, 95th percentiles	54.1, 118	49.0, 105	49.8, 111
Eastern Cooperative Oncology Group score (points)
0	65 (38.9%)	135 (39.7%)	200 (39.4%)
1	101 (60.5%)	204 (60.0%)	305 (60.2%)
2	1 (0.6%)	1 (0.3%)	2 (0.4%)
Ann Arbor stage at study entry
I	8 (4.8%)	19 (5.6%)	27 (5.3%)
II	21 (12.6%)	35 (10.3%)	56 (11.0%)
III	34 (20.4%)	78 (22.9%)	112 (22.1%)
IV	104 (62.3%)	208 (61.2%)	312 (61.5%)
Tumor size as sum of products of perpendicular diameters (mm 2 )
Median [IQR]	3100 [1350, 5800]	2860 [1570, 5110]	2930 [1480, 5220]
5th, 95th percentiles	447, 17,100	453, 12,800	446, 14,600
Prior CAR T therapy
No	101 (60.5%)	340 (100%)	441 (87.0%)
Yes	66 (39.5%)	0 (0%)	66 (13.0%)
Anti‐drug antibody status
Missing	4 (2.4%)	103 (30.3%)	107 (21.1%)
Negative	162 (97.0%)	232 (68.2%)	394 (77.7%)
Positive	1 (0.6%)	5 (1.5%)	6 (1.2%)
Albumin (g/L)
Median [IQR]	38.0 [34.5, 42.0]	39.0 [34.0, 42.0]	38.7 [34.0, 42.0]
5th, 95th percentiles	28.0, 45.7	29.0, 46.0	28.3, 46.0
Immunoglobulin G (μmol/L)
Median [IQR]	32.5 [23.5, 44.6]	41.0 [27.9, 57.2]	38.2 [25.4, 53.8]
5th, 95th percentiles	11.4, 71.1	13.5, 94.8	12.2, 88.4
Lymphocyte (10 9 /L)
Median [IQR]	0.660 [0.360, 1.06]	0.890 [0.550, 1.40]	0.792 [0.490, 1.30]
5th, 95th percentiles	0.133, 3.40	0.200, 3.17	0.170, 3.25
Interleukin 10 (ng/L)
Median [IQR]	11.5 [8.10, 35.3]	1.65 [0.760, 4.75]	3.02 [0.930, 11.0]
5th, 95th percentiles	1.61, 130	0.600, 39.0	0.600, 79.0
Missing	27 (16.2%)	0 (0%)	27 (5.3%)
Interferon gamma (ng/L)
Median [IQR]	0.300 [0.138, 0.745]	7.46 [3.52, 17.5]	3.66 [1.68, 12.1]
5th, 95th percentiles	0.0408, 3.52	2.60, 91.6	0.120, 66.9
Missing	27 (16.2%)	0 (0%)	27 (5.3%)
Renal impairment status
Normal	78 (46.7%)	137 (40.3%)	215 (42.4%)
Mild	57 (34.1%)	121 (35.6%)	178 (35.1%)
Moderate	29 (17.4%)	81 (23.8%)	110 (21.7%)
Severe	3 (1.8%)	1 (0.3%)	4 (0.8%)
Hepatic function
Normal	150 (89.8%)	292 (85.9%)	442 (87.2%)
Mild	17 (10.2%)	37 (10.9%)	54 (10.7%)
Moderate	0 (0%)	4 (1.2%)	4 (0.8%)
Severe	0 (0%)	7 (2.1%)	7 (1.4%)
Dose regimen
0.03/0.1 mg	3 (1.8%)	0 (0%)	3 (0.6%)
0.1/0.3 mg	3 (1.8%)	0 (0%)	3 (0.6%)
0.3/1 mg	3 (1.8%)	0 (0%)	3 (0.6%)
1/2 mg	8 (4.8%)	0 (0%)	8 (1.6%)
1/3 mg	3 (1.8%)	0 (0%)	3 (0.6%)
1/4 mg	7 (4.2%)	0 (0%)	7 (1.4%)
1/5 mg	3 (1.8%)	0 (0%)	3 (0.6%)
1/6 mg	4 (2.4%)	0 (0%)	4 (0.8%)
1/7 mg	4 (2.4%)	0 (0%)	4 (0.8%)
1/8 mg	3 (1.8%)	0 (0%)	3 (0.6%)
1/6/12 mg	5 (3.0%)	0 (0%)	5 (1.0%)
1/9/18 mg	5 (3.0%)	0 (0%)	5 (1.0%)
1/13.5/27 mg	7 (4.2%)	0 (0%)	7 (1.4%)
1/20/40 mg	9 (5.4%)	0 (0%)	9 (1.8%)
1/20/80 mg	9 (5.4%)	0 (0%)	9 (1.8%)
1/20/160 mg	9 (5.4%)	0 (0%)	9 (1.8%)
1/20/320 mg	9 (5.4%)	0 (0%)	9 (1.8%)
1/20/80 mg QW/160 mg Q2W	6 (3.6%)	87 (25.6%)	93 (18.3%)
1/20/160 mg QW/320 mg Q2W	41 (24.6%)	95 (27.9%)	136 (26.8%)
1/20/320 mg QW/320 mg Q2W	0 (0%)	33 (9.7%)	33 (6.5%)
0.7/4/20/80 mg QW/160 mg Q2W	0 (0%)	58 (17.1%)	58 (11.4%)
0.7/4/20/160 mg QW/320 mg Q2W	26 (15.6%)	67 (19.7%)	93 (18.3%)

Abbreviations: B‐NHL, B‐cell non‐Hodgkin lymphoma; CAR T, chimeric antigen receptor T cell; DLBCL, diffuse large B‐cell lymphoma; FL, follicular lymphoma; IQR, interquartile range; MCL, mantle cell lymphoma; MZL, marginal zone lymphoma; Q2W, once every 2 weeks; QW, once weekly.

There was a total of 14,618 concentration records in the analysis dataset, with a median of 29 (IQR: 20, 38) observations per patient within the median follow‐up time of 5 [IQR: 1, 10] months. A total of 1342 observed records were excluded from this analysis, with most of them being pre‐first dose samples (n = 497, concentration = 0), followed by possibly mis‐matching of samples and pre‐specified collection tubes (e.g., pre‐infusion concentration greater than end of infusion concentration) (n = 177). Extreme observations (conditional weighted residual outside the [−5, 5] interval) led to a failed convergence due to rounding error and were also excluded (n = 76); yet no patient was entirely excluded. Post‐dose concentration records that were below the limit of quantification represented 3.4% (n = 515) were excluded from the analysis.

3.2. Fit‐For‐Purpose Model

A detailed interpretation of the fit‐for‐purpose model parameters, covariate effects, and their relationships is presented in Table 2. The fit‐for‐purpose model code and output are presented in Supporting Information. The model described the observed concentration–time profiles reasonably well and predicted the median, 5th and 95th percentiles of the observed data regardless of B‐NHL subtype (Figure 2). Figure S2 presents additional goodness‐of‐fit graphs.

TABLE 2.

Model parameter estimates, covariate relationships, ^a and their interpretation.

Parameter b	Estimate (95% CI) (base model)	Estimate (95% CI) (full model)	Estimate (95% CI) (fit‐for‐purpose model)	Bootstrap c median (quantile 0.025, 0.975)
Linear clearance, CL (L/day)	0.176 (0.169, 0.184)	0.191 (0.187, 0.196)	0.189 (0.185, 0.194)	0.1890 (0.170, 0.208)
Albumin effect on CL		−0.967 (−1.13, −0.8)	−0.933 (−1.1, −0.769)	−0.9730 (−1.76, −0.584)
Body weight effect on CL		0.807 (0.739, 0.876)	0.854 (0.783, 0.925)	0.8500 (0.637, 1.03)
Immunoglobulin G effect on CL		0.136 (0.109, 0.162)	0.118 (0.0912, 0.145)	0.1190 (0.015, 0.215)
Interferon gamma effect on CL			0.0274 (0.0173, 0.0375)	0.0288 (−0.0046, 0.0592)
Interleukin 10 effect on CL			−0.0717 (−0.0833, −0.0601)	−0.0712 (−0.1, −0.0362)
Intercompartmental clearance, Q (L/day)	1.31 (1.22, 1.41)	1.19 (1.11, 1.28)	1.21 (1.13, 1.29)	1.1800 (0.826, 1.44)
Volume of distribution of central compartment, V c (L)	4.73 (4.56, 4.91)	4.63 (4.48, 4.78)	4.99 (4.8, 5.2)	5.02 (4.81, 5.25)
Volume of distribution of peripheral compartment, V p (L)	4.11 (3.73, 4.53)	4.1 (3.73, 4.5)	4.42 (4.01, 4.87)	4.3900 (3.9, 5.03)
Albumin effect on V c and V p		−0.551 (−0.739, −0.364)	−0.561 (−0.742, −0.379)	−0.5510 (−0.788, −0.36)
Body weight effect on V c and V p		0.561 (0.444, 0.679)	0.372 (0.242, 0.501)	0.366 (0.241, 0.487)
Female sex effect on V c and V p			−0.186 (−0.24, −0.133)	−0.1890 (−0.25, −0.127)
Lymphocyte count effect on V c and V p		0.0887 (0.0556, 0.122)	0.0895 (0.0576, 0.121)	0.0885 (−0.0544, 0.13)
Baseline maximum velocity of Michaelis–Menten elimination, V max,0 (mg/L/day)	2.1 (1.91, 2.3)	2.4 (2.04, 2.82)	2.93 (2.56, 3.36)	2.84 (2.14, 3.79)
Ann Arbor stage < 4 effect on V max,0			−0.285 (−0.375, −0.195)	−0.2840 (−0.378, −0.179)
Lymphocyte count effect on V max,0		0.32 (0.247, 0.392)	0.313 (0.242, 0.384)	0.3140 (0.215, 0.448)
Follicular lymphoma effect on V max,0		−0.229 (−0.369, −0.0877)	−0.252 (−0.367, −0.137)	−0.2360 (−0.421, 0.187)
Prior CAR T therapy effect on V max,0		−0.229 (−0.385, −0.0737)	−0.201 (−0.357, −0.044)	−0.2030 (−0.34, −0.0453)
Tumor size effect on V max,0		0.0847 (0.0245, 0.145)	0.0601 (0.00114, 0.119)	0.0588 (−0.00296, 0.129)
Concentration reaching 50% of V max,0, k m (mg/L)	2.64 (2.47, 2.83)	1.89 (1.75, 2.05)	1.95 (1.81, 2.1)	1.9000 (1.38, 2.53)
Rate constant describing the decline in V max per unit of time, K v (1/day)	0.133 (0.121, 0.146)	0.994 (0.829, 1.19)	1.09 (0.926, 1.27)	1.0900 (0.774, 1.65)
Follicular lymphoma effect on K v		−0.88 (−0.902, −0.858)	−0.891 (−0.909, −0.873)	−0.8850 (−0.924, −0.399)
Other B‐NHL type effect on K v		−0.831 (−0.867, −0.795)	−0.844 (−0.872, −0.815)	−0.8400 (−0.901, −0.414)
Asymptotic V max (V max,asym) to V max,0 ratio, R v	0.274 (0.264, 0.286)	0.27 (0.242, 0.303)	0.262 (0.237, 0.289)	0.2620 (0.207, 0.325)
Follicular lymphoma effect on R v		−0.151 (−0.264, −0.0385)	−0.133 (−0.235, −0.0302)	−0.1650 (−0.414, 0.208)
Other B‐NHL type effect on R v		−0.292 (−0.401, −0.183)	−0.292 (−0.383, −0.201)	−0.2870 (−0.525, 0.00242)
Variance of between‐patient variability in V c	0.153 (0.13, 0.176)	0.107 (0.0904, 0.124)	0.0983 (0.0829, 0.114) a , b , d	0.0951 (0.0742, 0.13)
Variance of between‐patient variability in V max,0	0.539 (0.451, 0.626)	0.418 (0.349, 0.487)	0.403 (0.338, 0.467) a , b , d	0.4010 (0.305, 0.52)
Variance of between‐patient variability in V p	0.76 (0.621, 0.9)	0.721 (0.584, 0.857)	0.709 (0.576, 0.841) a , b , d	0.6890 (0.511, 0.94)
Variance of residual variability (ln mg/L)	0.149 (0.145, 0.153)	0.139 (0.136, 0.142)	0.136 (0.133, 0.14)	0.1350 (0.122, 0.148)

Note: Condition number is 69.1, minimum objective function value from NONMEM is −11043.

Abbreviations: B‐NHL, B‐cell non‐Hodgkin lymphoma; CAR T, chimeric antigen receptor T cell; CI, confidence interval; CL, clearance; V _max, the maximum amount of odronextamab eliminated via target per unit of time and volume.

^a Covariates effects used baseline values, except by the effect of albumin on CL.

^b Reference patient is a 72.6‐kg male with DLBCL, CAR T naive, Ann Arbor stage of 4, tumor size of 2930 mm², albumin of 38.7 g/L, immunoglobulin G of 38.2 μmol/L, interleukin 10 of 3.01 ng/L, interferon gamma of 3.66 ng/L, and lymphocyte count of 0.792 × 10⁹/L.

^c Bootstrap: summary results from 300 bootstrap samples drawn with replacement.

^d Eta‐shrinkage of 8.3%, 8.6%, and 20.1% for V _c , V _max, and V _p , respectively.

FIGURE 2.

Visual predictive checks. (A) All B‐NHL subpopulations, population prediction‐corrected; (B) All B‐NHL subpopulations; (C) DLBCL CAR T naïve, population prediction‐corrected; (D) DLBCL CAR T failure, population prediction‐corrected; (E) FL, population prediction‐corrected; (F) MCL, population prediction‐corrected; (G) MZL, population prediction‐corrected; (H) Other B‐NHL subtype, population prediction‐corrected. Open gray circles are individual observations. Lines are 95th, 50th, and 5th percentiles of the observed data, respectively, from top to bottom. Shaded areas are the 2.5th to 97.5th inter‐percentile range of the simulated data. B‐NHL, B‐cell non‐Hodgkin lymphoma; CAR T, chimeric antigen receptor T cell; DLBCL, diffuse large B‐cell lymphoma; FL, follicular lymphoma; MCL, mantle cell lymphoma; MZL, marginal zone lymphoma.

The typical (reference) patient was a 72.6‐kg male with DLBCL, CAR T naive, Ann Arbor stage of 4, tumor size of 2930 mm², albumin of 38.7 g/L, IgG of 38.2 μmol/L, IL‐10 of 3.01 ng/L, IFN gamma of 3.66 ng/L, and lymphocyte count of 0.792 × 10⁹/L. The population estimates of CL and Q were 0.189 L/day and 1.21 L/day, respectively. The sum of the estimates for V _c and V _p of 4.99 L and 4.42 L, respectively, yielded a volume of distribution at steady state of 9.41 L. Estimates of baseline V _max, k _m , K _v , and R _v were 2.93 mg/L/day, 1.95 mg/L, 1.09/day, and 0.262, respectively. Between‐subject variability was estimated for V _max, V _c , and V _p , with eta‐shrinkage of 8.3%, 8.6%, and 20.1%, respectively. An attempt to estimate BSV for CL biased the population predictions, suggesting that there was not enough information in the data to inform between‐subject random effects for both linear and Michaelis–Menten elimination; albumin levels, though, were incorporated as a time‐varying covariate of CL, explaining a significant portion of its variability between patients and within patients over time. Based on the criteria mentioned in the Methods, all pre‐specified covariate effects were retained in the fit‐for‐purpose model, except by the effect of other B‐NHL subtypes—grouped as MCL, MZL, or other B‐NHL—on V _max,0. Statistically significant exploratory/ad hoc covariates included in the fit‐for‐purpose model were: (1) IL‐10 and IFN gamma effect on CL; (2) female sex on V _c and V _p ; (3) Ann Arbor stage < 4 on V _max,0 (Table S3 presents forward inclusion and backward elimination procedure). While there was some degree of correlation among interleukins, IL‐10 was the most predictive of CL and other interleukins did not explain variability in CL above and beyond IL‐10.

Marginal covariate effects on odronextamab exposure were estimated for covariate subgroups (categorical covariates) and 5th and 95th percentile values of the covariate distribution (continuous covariates) for patients in the analysis dataset (Figure 3). Most 95% CI of covariate effects on odronextamab exposure relative to reference were contained within the 80%–125% range, except for body weight, albumin, and IL‐10. Body weight was directly correlated with CL and volume of distribution; albumin was inversely correlated with CL and volume of distribution; and IL‐10 was inversely correlated with CL. For instance, the estimated effects of these important covariates on C _min,ss relative to the reference patient were:

• A 49.8‐kg (5th percentile) or 111‐kg (95th percentile) patient was estimated to have 1.47‐fold (95% CI 1.42–1.51) and 0.632‐fold (95% CI 0.609–0.657) change, respectively.

• A patient with albumin level of 28.3 g/L (5th percentile) or 46 g/L (95th percentile) was estimated to have 0.689‐fold (95% CI 0.65–0.733) and 1.22‐fold (95% CI 1.18–1.25) change, respectively.

• A patient with an IL‐10 level of 0.6 ng/L (5th percentile) or 79 ng/L (95th percentile) was estimated to have 0.871‐fold (95% CI 0.852–0.892) and 1.3‐fold (95% CI 1.25–1.36) change, respectively.

FIGURE 3.

Marginal effects of covariates on odronextamab exposure in serum. (A) AUC at steady state; (B) C _max at steady state; (C) C _min at steady state. Reference patient is a 72.6‐kg male with DLBCL, CAR T naive, Ann Arbor stage of 4, tumor size of 2930 mm², albumin of 38.7 g/L, immunoglobulin G of 38.2 μmol/L, interleukin 10 of 3.01 ng/L, interferon gamma of 3.66 ng/L, and lymphocyte count of 0.792 × 10⁹/L. For continuous covariates, lower and higher values in the forest plot represent the 5th and 95th percentiles, respectively, of the covariate distribution at baseline across patients in the analysis dataset. Odronextamab 0.7/4/20/160 mg QW/320 mg Q2W dose regimen. AUC, area under the concentration–time curve; B‐NHL, B‐cell non‐Hodgkin lymphoma; BSV, between‐subject variability; CAR T, chimeric antigen receptor T cell; CI, confidence interval; C_max, maximum concentration; C_min, minimum concentration; DLBCL, diffuse large B‐cell lymphoma; FL, follicular lymphoma; QW, once weekly; Q2W, once every 2 weeks.

3.3. Simulations Using Post Hoc Estimates of Individual PK Parameters

3.3.1. Time to Steady State, Washout, and Accumulation

Simulations were performed following administration of the 0.7/4/20/80 mg QW/160 mg Q2W or 0.7/4/20/160 mg QW/320 mg Q2W dosing regimen to patients in the analysis dataset (Figures S3 and S4, Tables S4 and S5); 97% of steady state was reached 29 weeks after the first dose for the 0.7/4/20/80 mg QW/160 mg Q2W and 0.7/4/20/160 mg QW/320 mg Q2W dose regimens, respectively. Notwithstanding the limited amount of observed data to allow characterization of the wash‐out phase for odronextamab, simulations estimated that following the administration of the last steady‐state dose for 0.7/4/20/80 mg QW/160 mg Q2W and 0.7/4/20/160 mg QW/320 mg Q2W, the time to reach the assay lower limit of quantification (0.00313 mg/L) was 19 weeks and 24 weeks, respectively. For those dose regimens, the median ratio of odronextamab accumulation of full dose between the AUC at Week 12 and AUC at Week 4 for 80 mg and 160 mg QW was 3.44 (IQR: 3.18, 3.75) and 3.87 (IQR: 3.56, 4.25), respectively.

3.3.2. Covariate Effects

For covariates with marginal effects outside the 80–125% range, a comparison of steady‐state odronextamab exposure among patients grouped by baseline characteristics (Tables 3 and S6) revealed effects of body weight and IL‐10 that were in the same direction as the estimated marginal effects (Figure 3), although of smaller magnitude (i.e., 1.18‐fold vs. 1.30‐fold change) for patients with IL‐10 levels in the 0.95–1 quartile range (> 79 ng/L) (N = 25). Patients with albumin levels in the 0–0.05 quartile range (16, 28.3 g/L) had an estimated median fold change in C _min,ss of 1.09, which was within the 80%–125% range and in the opposite direction of the estimated marginal effect. A summary of model covariates by various albumin IQRs revealed an imbalance of other covariate values associated with decreased total clearance in these patients (Table S7). For instance, median body weight, IgG level, lymphocyte count, and IL‐10 level were 0.89‐fold, 0.68‐fold, 0.70‐fold, and 3.2‐fold, respectively, relative to the overall population.

TABLE 3.

Comparison of steady‐state odronextamab exposure among patients grouped by various baseline characteristics ^a in the analysis dataset after simulated administration of the 0.7/4/20/160 mg QW/320 mg Q2W dose regimen.

	C min (mg/L)	C max (mg/L)	AUC (mg*day/L)
	Median [IQR]	Median [IQR]	Median [IQR]
Anti‐drug antibody status
Missing (N = 107)	86.4 [69.1, 106]	153 [130, 191]	1490 [1250, 1780]
Negative (N = 394)	84.4 [67.9, 104]	157 [130, 187]	1460 [1240, 1800]
Positive (N = 6)	78.0 [43.3, 96.9]	156 [109, 192]	1390 [855, 1680]
Ann Arbor stage at study entry
I (N = 27)	83.1 [70.0, 104]	172 [138, 201]	1490 [1260, 1820]
II (N = 56)	81.2 [66.9, 101]	159 [134, 193]	1440 [1260, 1740]
III (N = 112)	86.7 [69.1, 101]	157 [130, 187]	1470 [1260, 1750]
IV (N = 312)	84.2 [68.0, 105]	154 [129, 185]	1460 [1230, 1800]
Eastern Cooperative Oncology Group score
0 (N = 200)	85.0 [66.2, 100]	157 [133, 188]	1460 [1230, 1750]
1 (N = 305)	84.5 [69.7, 106]	156 [125, 189]	1470 [1250, 1830]
2 (N = 2)	86.9 [52.1, 122]	128 [85.6, 170]	1420 [899, 1940]
Prior CAR T therapy
No (N = 441)	83.6 [66.7, 100]	154 [128, 186]	1460 [1230, 1760]
Yes (N = 66)	95.5 [72.1, 125]	169 [138, 212]	1650 [1270, 2040]
Renal impairment status b
Normal (N = 215)	76.0 [57.8, 90.9]	141 [114, 163]	1340 [1050, 1550]
Mild (N = 178)	85.4 [73.1, 103]	161 [136, 190]	1510 [1310, 1770]
Moderate (N = 110)	99.4 [82.5, 118]	179 [150, 209]	1720 [1440, 1950]
Severe (N = 4)	126 [114, 153]	230 [214, 251]	2120 [2000, 2440]
Hepatic function (NCI‐ODWG) [16]
Normal (N = 442)	84.5 [67.9, 104]	157 [130, 189]	1470 [1240, 1800]
Mild (N = 54)	85.6 [63.8, 105]	149 [124, 174]	1450 [1130, 1750]
Moderate c (N = 4)	42.4 [0.0308, 109]	89.7 [25.2, 177]	766 [56.3, 1810]
Severe (N = 7)	83.5 [73.3, 92.4]	197 [137, 205]	1420 [1330, 1630]
Sex
Female (N = 194)	99.8 [82.9, 117]	189 [159, 212]	1750 [1460, 1990]
Male (N = 313)	76.5 [61.3, 91.3]	141 [118, 165]	1350 [1110, 1560]
Age (years)
[22, 38.3] (N = 26)	84.5 [67.4, 108]	170 [126, 202]	1530 [1270, 1870]
(38.3, 81) (N = 454)	84.2 [66.7, 102]	154 [129, 187]	1460 [1230, 1770]
[81, 89] (N = 27)	85.3 [78.8, 108]	173 [139, 196]	1530 [1350, 1830]
Body weight (kg)
[31.8, 49.8] (N = 26)	142 [108, 159]	225 [204, 258]	2310 [1920, 2650]
(49.8, 111) (N = 454)	84.7 [70.2, 102]	157 [133, 186]	1470 [1260, 1770]
[111, 165] (N = 27)	53.2 [40.6, 61.1]	109 [101, 118]	985 [795, 1100]
Tumor size as sum of products of perpendicular diameters (mm 2 )
[75, 446] (N = 26)	102 [87.9, 118]	183 [162, 215]	1800 [1530, 1990]
(446, 14,600) (N = 455)	82.9 [66.6, 101]	153 [128, 185]	1440 [1230, 1760]
[1.46e+04, 4.61e+04] (N = 26)	96.9 [82.2, 116]	177 [136, 202]	1690 [1410, 1910]
Albumin (g/L)
[16, 28.3] (N = 26)	92.1 [78.3, 115]	157 [127, 201]	1560 [1290, 1910]
(28.3, 46) (N = 447)	84.6 [68.6, 103]	158 [131, 188]	1470 [1240, 1800]
[46, 56] (N = 34)	71.0 [59.1, 89.8]	141 [118, 171]	1310 [1070, 1530]
Immunoglobulin G (μmol/L)
[4.07, 12.2] (N = 26)	111 [85.3, 134]	179 [143, 213]	1810 [1430, 2210]
(12.2, 88.4) (N = 455)	84.6 [67.5, 103]	156 [129, 188]	1470 [1240, 1780]
[88.4, 4.54e+03] (N = 26)	69.2 [60.6, 80.1]	133 [118, 167]	1260 [1090, 1440]
Lymphocyte (10 9 /L)
[0.03, 0.17] (N = 28)	101 [87.0, 124]	187 [148, 213]	1760 [1490, 2020]
(0.17, 3.25) (N = 453)	83.8 [67.4, 103]	157 [130, 187]	1460 [1240, 1780]
[3.25, 44.3] (N = 26)	71.6 [19.0, 90.8]	122 [55.3, 151]	1230 [410, 1530]
Interferon gamma (ng/L)
[0.022, 0.12] (N = 24)	94.5 [78.3, 116]	169 [150, 188]	1610 [1410, 1940]
(0.12, 66.9) (N = 459)	83.8 [67.0, 103]	156 [128, 188]	1460 [1230, 1790]
[66.9, 2.04e+03] (N = 24)	89.4 [80.6, 101]	154 [135, 196]	1480 [1400, 1730]
Interleukin 10 (ng/L)
0.6 (N = 29)	79.1 [62.9, 101]	149 [125, 180]	1450 [1130, 1730]
(0.6, 79) (N = 426)	84.6 [67.4, 103]	156 [130, 189]	1470 [1240, 1790]
[79, 9.75e+03] (N = 25)	99.8 [83.5, 129]	159 [137, 186]	1730 [1440, 2000]
Overall (N = 507)	84.6 [67.5, 104]	156 [129, 188]	1470 [1240, 1800]

Abbreviations: AUC, area under the concentration–time curve; CAR T, chimeric antigen receptor T cell; C _max, maximum concentration; C _min, minimum concentration; FDA, US Food and Drug Administration; IQR, interquartile range; NCI‐ODWG, National Cancer Institute Organ Dysfunction Working Group; Q2W, once every 2 weeks; QW, once weekly; V _max, the maximum amount of odronextamab eliminated via target per unit of time and volume.

^a For continuous covariates, intervals correspond to [0, 0.05], (0.05, 0.95) and [0.95, 1] quantile ranges, respectively.

^b Classification based on Cockcroft‐Gault equation as per FDA Population Pharmacokinetics Guidance [17].

^c Two patients (ID 168 and ID 268) had short treatment duration with odronextamab doses that were lower than the target doses; the few collected samples showed concentrations in serum falling quickly below the assay lower limit of quantification. Therefore, estimated high steady‐state V _max values, and predicted low steady‐state concentrations, are not necessarily accurate given the level of extrapolation.

Higher steady‐state odronextamab exposure appeared to be in groups with poor renal function (Table S8); this is mostly confounded due to the fact that patients with impaired renal function also had lower body weight (Figure S5). Patients with severe renal impairment had a median C _min,ss that was approximately 1.49‐fold relative to the median C _min,ss in the overall population. In these patients, the median body weight was approximately 53 kg in contrast to a median body weight of approximately 73 kg in the overall population. As such, the higher exposure is consistent with the estimated marginal effect of body weight on exposure, in that patients with a body weight of approximately 50 kg were estimated to have a median change in C _min,ss of 1.47‐fold relative to the reference (Figure 3). Moreover, patients with severe renal impairment had IgG and IFN gamma levels, lymphocyte count, and tumor size that were below the median of the overall population, and associated with lower total clearance and higher exposure as per estimated marginal covariate effects (Figure 3).

3.3.3. DLBCL Vs. FL

Modest differences in predicted odronextamab concentration after simulated administration of the 0.7/4/20/160 mg QW/320 mg Q2W dose regimen to patients with DLBCL and FL in the analysis dataset were uncovered mainly within the initial weeks of the step‐up treatment at concentrations approximately 1000‐fold lower than those achieved with the target doses, and hence are unlikely to be of consequence. Median odronextamab exposure was higher in patients with DLBCL in comparison to FL during the first two to 3 weeks of treatment (Figure 4 Panels A–D, Table S9). This was consistent with the estimated faster V _max rate of decline over time in the former B‐NHL subpopulation (Figure 4 Panel E). V _max values for both subpopulations reached similar levels at around the third week of treatment approaching the asymptote at around the fourth week of treatment, as reflected by the modified $\mathrm{C}{\mathrm{L}}_{\mathrm{MM}}$ time profiles (Figure 4 Panels F and G). The estimated median $\mathrm{C}{\mathrm{L}}_{\mathrm{MM}}$ at baseline was 5.43 L/day and 4.67 L/day, and the estimated median $\mathrm{C}{\mathrm{L}}_{\mathrm{MM}}$ at asymptote was 0.0296 L/day and 0.025 L/day, for patients with DLBCL and FL, respectively.

FIGURE 4.

Predicted odronextamab concentration and clearance over time after simulated administration of the 0.7/4/20/160 mg QW/320 mg Q2W dose regimen to patients with DLBCL and FL in the analysis dataset. (A) Concentration–time profile stratified by B‐NHL subtype and overlayed by observed data during the first 3 weeks of treatment; (B) Concentration–time profile with overlay of B‐NHL subtypes during the first 3 weeks of treatment; (C) Concentration–time profile stratified by B‐NHL subtype with overlay of observed data at long‐term follow‐up; (D) Concentration–time profile with overlay of B‐NHL subtypes at long‐term follow‐up; (E) V _max time profile with overlay of B‐NHL subtypes; (F) Linear and modified MM clearance time profile stratified by B‐NHL subtype; (G) Modified MM clearance time profile with overlay of B‐NHL subtypes. B‐NHL, B‐cell non‐Hodgkin lymphoma; DLBCL, diffuse large B‐cell lymphoma; FL, follicular lymphoma; MM, Michaelis–Menten; QW, once weekly; Q2W, once every 2 weeks; V _max, the maximum amount of odronextamab eliminated via target per unit of time and volume.

4. Discussion

The developed PK model described the observed concentration–time profiles reasonably well, regardless of B‐NHL subtype, and was determined fit‐for‐purpose. The bi‐exponential decay of odronextamab concentrations in serum, with parallel first‐order (i.e., linear clearance or CL, consistent with a non‐saturable proteolytic process) and Michaelis–Menten elimination (i.e., non‐linear clearance or CL_MM, consistent with a target‐mediated elimination process), is commonly observed for an IV administered monoclonal antibody against an endogenous target on a cell membrane. An estimated linear clearance of 0.189 L/day is also consistent with typically reported values of linear clearance for monoclonal antibodies. The Michaelis–Menten elimination, typically only concentration–dependent, also included a time‐dependent component. An empirical function describing a decline in V _max over time improved fitting of the observed data, with baseline total lymphocyte count being a predictor of baseline V _max. A temporal reduction in V _max suggests a temporal reduction in target abundance, which is consistent with the treatment‐induced depletion of the B cells that was observed in patients who had measurable B‐cell counts in serum. An empirical time‐dependent, although concentration–independent, clearance mechanism has also been reported for rituximab (anti‐CD20 therapy) in patients with NHL [18] or chronic lymphocytic leukemia [19]. However, given the empirical nature of the model developed herein and the fact that odronextamab also binds to T cells, mechanistic‐related conclusions should be drawn with caution. Limitations in the data on odronextamab target abundance (i.e., CD20+ B cells and CD3+ T cells) made the development of a mechanistic model difficult. Baseline counts of total circulating B and T cells were not quantifiable in approximately half of the patient population, with an increased proportion of missing information at later time points. Moreover, circulating B cells do not account for target present in malignant cells accumulated in the form of nodal or extranodal tumor masses [20].

Body weight had the largest effect on odronextamab exposure, being directly correlated with linear clearance and volume of distribution via a power relationship with allometric exponents of 0.854 and 0.372, respectively. While the allometric exponent for clearance is consistent with that reported in the literature for monoclonal antibodies (i.e., 0.75–0.85), the relationship between body weight and volume of distribution was less steep than that typically reported (i.e., exponent of ~1) [21]. The body weight range of the analysis population, though, was not large enough to allow a reliable estimate of allometric exponents in that it included only adult patients.

Although not as pronounced as the effect of body weight, albumin and IL‐10 also had the 95% CI of their marginal effect on odronextamab exposure lying outside of the 80%–125% range. Albumin and IL‐10 were inversely correlated with linear clearance. Such effects align with what has been shown in the literature. Cancer has been associated with chronic inflammatory response, which, in turn, is associated with higher protein turnover rate. Hypoalbuminemia is a recognized marker of cachexia and elevated protein turnover secondary to chronic systemic inflammation. Indeed, serum albumin concentrations have frequently been reported as an inversely correlated covariate for monoclonal antibodies clearance, in which decreased albumin levels are indicative of increased IgG clearance [10]. In addition, interaction of proteins with circulating albumin in serum has been reported, with higher levels of albumin resulting in lower volume of distribution [22]. IL‐10 is a potent anti‐inflammatory cytokine playing an important role in preventing inflammatory pathologies [23]. As such, the association between higher IL‐10 levels and lower clearance is plausible.

Author Contributions

M.B.P., D.J.C., K.S., L.O.H. and J.D.D. wrote the manuscript. M.B.P., D.J.C., K.S., L.O.H., J.D.D. and M.Z. designed the research. M.B.P., D.J.C. and M.Z. performed the research. M.B.P. and D.J.C. analyzed the data.

Funding

This study is funded by Regeneron Pharmaceuticals Inc.

Disclosure

The opinions expressed in this article are those of the authors and should not be interpreted as the position of their employers.

Conflicts of Interest

M.B.P., K.S., L.O.H., J.D.D. and M.Z. are employees of and own stock of Regeneron Pharmaceuticals Inc. D.J.C. is an employee of and owns stock of Jazz Pharmaceuticals Inc., and is a former employee of Regeneron Pharmaceuticals Inc.

Supporting information

Appendix S1: psp470162‐sup‐0001‐Supinfo.pdf.

Bravo Padros M., Conrado D. J., Srinivasan K., Harnisch L. O., Davis J. D., and Zhu M., “Pharmacokinetics of Odronextamab, A Bispecific T‐Cell‐Engaging Antibody, in Adult Patients With Relapsed or Refractory B‐Cell Non‐Hodgkin Lymphoma,” CPT: Pharmacometrics & Systems Pharmacology 15, no. 1 (2026): e70162, 10.1002/psp4.70162.

Data Availability Statement

Qualified researchers may request access to study documents (including the clinical study report, study protocol with any amendments, blank case report form, statistical analysis plan) that support the methods and findings reported in this manuscript. Individual anonymized participant data will be considered for sharing (1) once the product and indication has been approved by major health authorities (e.g., FDA, EMA, PMDA, etc.) or development of the product has been discontinued globally for all indications on or after April 2020 and there are no plans for future development (2) if there is legal authority to share the data and (3) there is not a reasonable likelihood of participant re‐identification. Submit requests to https://vivli.org/.