Pharmacokinetic parameters of T-cell engagers
Fixed effects, random effects and covariates from the published population models
Three drugs are complete, three are partial, and nothing has been checked against a printed table. Teclistamab, mosunetuzumab and elranatamab were transcribed from nlmixr2lib model files, which carry their own provenance notes and table citations. Odronextamab was transcribed from the open-access full text. Epcoritamab, talquetamab, glofitamab and blinatumomab are summary-level only. Every entry is ⚠️ in References.
This is a reference page. It answers “what did the published model estimate” and nothing else. The argument the numbers were collected for is in the working specification.
1. What This Covers, and in What Order
Every approved T-cell engager (TCE) with a published population pharmacokinetic model, plus blinatumomab as the format that behaves differently. Ordered so that the first entries teach the ones after them: odronextamab because its model states target-mediated drug disposition (TMDD) explicitly, then the two whose time-dependent clearance is the same mechanism under another name, then the partial entries.
Each drug gets the same six slots, and a slot with nothing in it says so rather than being dropped.
- Study and data
- Structure
- Fixed effects
- Interindividual variability
- Residual error
- Covariates
Units are litres and days throughout, converted where a source reported hours. Volumes are absolute, not per kilogram.
2. Summary Across Drugs
| Drug | Target | Route | Structure | CL at start | CL asymptote | \(V_{ss}\) |
|---|---|---|---|---|---|---|
| Odronextamab | CD20 | IV | 2-cmt, linear + Michaelis-Menten, \(V_{\max}\) decaying | ~5.6 L/d | 0.22 L/d | 9.41 L |
| Mosunetuzumab | CD20 | IV | 2-cmt, \(CL_{\rm base}\to CL_{ss}\) | 1.08 L/d | 0.584 L/d | 11.7 L |
| Teclistamab | BCMA | SC | 2-cmt, first-order absorption, \(CL_1+CL_2e^{-K_{DES}t}\) | 0.996 L/d | 0.449 L/d | 5.47 L |
| Elranatamab | BCMA | SC | 2-cmt, first-order absorption, linear | 0.335 L/d | same | 12.05 L |
| Epcoritamab | CD20 | SC | 2-cmt quasi-steady-state TMDD, first-order absorption | CL/F 0.53 L/d | not transcribed | V/F 25.6 L |
| Talquetamab | GPRC5D | SC | 2-cmt, zero- then first-order absorption, linear + time-dependent linear | not transcribed | not transcribed | not transcribed |
| Glofitamab | CD20 | IV | 2-cmt, linear | not transcribed | same | not transcribed |
| Blinatumomab | CD19 | continuous IV | 1-cmt, linear | 70.1 L/d | same | 4.52 L (\(V_z\)) |
Reading down the clearance columns, the four drugs with a start-to-asymptote gap are the four whose models carry a time-varying or concentration-varying term. Elranatamab’s model is linear because it is a preliminary population analysis embedded in a systems model, not because elranatamab lacks the mechanism.
3. The Drugs
3.1 Odronextamab
Study and data. ELM-1 (NCT02290951, phase I, n = 167) and ELM-2 (NCT03888105, phase II, n = 340), 507 patients with relapsed or refractory B-cell non-Hodgkin lymphoma, 14,618 concentration records, median 29 per patient. Doses 0.03 mg to 320 mg intravenous. Kovalenko et al., CPT Pharmacometrics Syst Pharmacol 2026.
Structure. Two compartments, parallel first-order and Michaelis-Menten elimination, with the maximum velocity declining exponentially toward an asymptote:
\[ \frac{dV_{\max}}{dt}=-K_v\!\left(V_{\max}-V_{\max,\rm asym}\right), \qquad CL_{MM}=\frac{V_{\max}}{k_m+C}V_c, \qquad CL_{\rm tot}=CL+CL_{MM} \]
Fixed effects. Reference patient: 72.6 kg male, DLBCL, CAR-T naive, Ann Arbor stage 4, tumor 2930 mm², albumin 38.7 g/L.
| Parameter | Estimate | 95% CI |
|---|---|---|
| Linear clearance \(CL\) (L/day) | 0.189 | 0.185, 0.194 |
| Intercompartmental clearance \(Q\) (L/day) | 1.21 | 1.13, 1.29 |
| Central volume \(V_c\) (L) | 4.99 | 4.80, 5.20 |
| Peripheral volume \(V_p\) (L) | 4.42 | 4.01, 4.87 |
| Baseline maximum velocity \(V_{\max,0}\) (mg/L/day) | 2.93 | 2.56, 3.36 |
| Michaelis constant \(k_m\) (mg/L) | 1.95 | 1.81, 2.10 |
| Decline rate of \(V_{\max}\), \(K_v\) (1/day) | 1.09 | 0.926, 1.27 |
| Asymptote ratio \(R_v=V_{\max,\rm asym}/V_{\max,0}\) | 0.262 | 0.237, 0.289 |
Interindividual variability. Variances, with shrinkage. \(V_c\) 0.0983 (8.3%), \(V_{\max,0}\) 0.403 (8.6%), \(V_p\) 0.709 (20.1%). No random effect on \(CL\): attempting to estimate one biased the population predictions, so albumin was carried as a time-varying covariate on \(CL\) instead.
Residual error. Variance 0.136 on \(\ln\)(mg/L), so a coefficient of variation near 37%.
Covariates. On \(CL\): albumin −0.933, body weight 0.854, immunoglobulin G 0.118, interferon gamma 0.0288, interleukin-10 −0.0712. On \(V_c\) and \(V_p\): albumin −0.561, body weight 0.372, female sex −0.189, lymphocyte count 0.0895. On \(V_{\max,0}\): Ann Arbor stage < 4 −0.284, lymphocyte count 0.313, follicular lymphoma −0.252, prior CAR-T −0.201, tumor size 0.0601. On \(K_v\): follicular lymphoma −0.891, other B-NHL −0.844. On \(R_v\): follicular lymphoma −0.133, other B-NHL −0.292.
The covariates on \(V_{\max,0}\) read as a target-burden model. Lymphocyte count, tumor size, stage and prior CAR-T all move the size of the target sink, and all four land on the nonlinear clearance rather than the linear one.
3.2 Mosunetuzumab
Study and data. GO29781 phase I/II, 439 patients with relapsed or refractory B-cell non-Hodgkin lymphoma, 7250 observations. Doses 0.05 to 2.8 mg fixed and 1/2/60/30 mg step-up, intravenous every 3 weeks. Bender et al., Clin Transl Sci 2024;17(5):e13825. Via nlmixr2lib Bender_2024_mosunetuzumab.
Structure. Two compartments, clearance moving from a baseline value to a steady-state value with a transition half-life:
\[ CL(t)=CL_{\rm base}+\left(CL_{ss}-CL_{\rm base}\right) \left[1-e^{-\ln 2\,t/HL_{\rm trans}}\right] \]
The model also carries residual rituximab and obinutuzumab from prior therapy as states decaying at fixed literature half-lives of 24 and 28 days, and computes a competitive CD20 receptor occupancy from all three.
Fixed effects.
| Parameter | Estimate | %RSE |
|---|---|---|
| Baseline clearance \(CL_{\rm base}\) (L/day) | 1.08 | 5.6 |
| Steady-state clearance \(CL_{ss}\) (L/day) | 0.584 | 2.0 |
| Transition half-life \(HL_{\rm trans}\) (day) | 16.3 | 7.1 |
| Central volume \(V_1\) (L) | 5.49 | 2.5 |
| Peripheral volume \(V_2\) (L) | 6.17 | 3.6 |
| Intercompartmental clearance \(Q\) (L/day) | 1.46 | 3.7 |
| Mosunetuzumab CD20 \(K_D\) (µg/mL) | 10.2 | fixed |
| Rituximab CD20 \(K_D\) (µg/mL) | 0.675 | fixed |
| Obinutuzumab CD20 \(K_D\) (µg/mL) | 0.600 | fixed |
Interindividual variability. Two 2×2 blocks and one diagonal term. \((CL_{\rm base}, V_1)\): variances 0.426 and 0.0981, covariance 0.180, correlation 0.88, shrinkage 4.8% and 4.6%. \((CL_{ss}, HL_{\rm trans})\): variances 0.0343 and 0.739, covariance −0.0892, correlation −0.56, shrinkage 33.8% and 40.9%. \(V_2\): 0.0621, shrinkage 49.9%. No random effect on \(Q\).
Residual error. Log-normal, SD 0.259.
Covariates. Body weight (reference 78 kg) on \(CL_{ss}\) 0.549, \(V_1\) 0.433, \(V_2\) 0.737. Albumin (reference 39 g/L) on \(CL_{\rm base}\) −1.51, \(V_1\) −0.481. Composite baseline anti-CD20 concentration on \(CL_{\rm base}\) −0.573, entering as a ratio of logarithms rather than a concentration ratio. Square root of tumor sum of products (reference 54.5 mm) on \(CL_{ss}\) 0.0935. Female sex as a linear multiplier, \(CL_{ss}\) −0.128, \(V_1\) −0.126.
Albumin and residual anti-CD20 antibody land on the baseline clearance; body weight and tumor burden land on the steady-state clearance. The split assigns the decaying component to the covariates that describe the target and the inflammatory state.
3.3 Teclistamab
Study and data. MajesTEC-1 phase I/II (NCT03145181, NCT04557098), 338 patients with relapsed or refractory multiple myeloma, 4840 serum concentrations, 83 patients intravenous and 255 subcutaneous. Miao et al., Target Oncol 2023;18(5):667-684. Via nlmixr2lib Miao_2023_teclistamab.
Structure. Two compartments, first-order subcutaneous absorption, and total clearance \(CL(t)=CL_1+CL_2e^{-K_{DES}t}\). The authors tested a quasi-steady-state TMDD parameterization and a time-varying soluble BCMA driver, and rejected both; they describe the exponential decay as an approximation of TMDD combined with feedback from improving disease status.
Fixed effects.
| Parameter | Estimate | %RSE |
|---|---|---|
| Time-independent clearance \(CL_1\) (L/day) | 0.449 | 8.87 |
| Time-dependent clearance \(CL_2\) at \(t=0\) (L/day) | 0.547 | 15.6 |
| Decay rate \(K_{DES}\) (1/day) | 0.0292 | 13.0 |
| Central volume \(V_1\) (L) | 4.13 | 4.40 |
| Peripheral volume \(V_2\) (L) | 1.34 | 26.1 |
| Intercompartmental clearance \(Q\) (L/day) | 0.0390 | 55.5 |
| Absorption rate constant \(K_a\) (1/day) | 0.133 | 7.73 |
| Subcutaneous bioavailability \(F\) | 0.718 | 7.38 |
Interindividual variability. Reported as percent coefficient of variation. \(CL_1\) 53.6% (shrinkage 14.4%), \(CL_2\) 107% (33.8%), \(V_1\) 48.8% (29.5%), \(K_a\) 45.2% (44.3%). Random effects on \(Q\), \(K_{DES}\) and \(F\) were attempted and dropped. No off-diagonal covariances reported.
The 107% coefficient of variation on the decaying component is the largest random effect anywhere in this document, and it is on the term that represents the target sink. Patients differ in target burden by more than they differ in anything else.
Residual error. Additive on the log scale, 41.7% coefficient of variation.
Covariates. Body weight (reference 74 kg) on \(CL_1\) 0.704, \(V_1\) 0.358, \(V_2\) 1.40. International Staging System stage II 1.31 and stage III 1.67 on \(CL_1\). Non-IgG myeloma 0.689 on \(CL_1\) and 0.295 on \(CL_2\), the only covariate retained on both components. Screened and rejected: age, sex, creatinine clearance, albumin, soluble BCMA (both baseline and time-varying), anti-drug antibodies, ECOG performance status.
3.4 Elranatamab
Study and data. MagnetisMM-1 (NCT03269136) and MagnetisMM-3 (NCT04649359). The pharmacokinetic parameters below are a preliminary population analysis, fixed rather than estimated inside the quantitative systems pharmacology model of Poels et al., npj Syst Biol Appl 2025;11:102. Via nlmixr2lib Poels_2025_elranatamab_qsp. Patient counts and observation counts for the preliminary analysis are not reported.
Structure. Two compartments with first-order subcutaneous absorption and linear elimination, embedded in a three-compartment systems model that adds bone marrow, mass-action binding to membrane BCMA, soluble BCMA and CD3, trimer-driven myeloma killing and cytokine release.
Fixed effects. Converted from the source’s per-hour units.
| Parameter | Estimate | Source units |
|---|---|---|
| Clearance \(CL\) (L/day) | 0.335 | 13.96 mL/h |
| Intercompartmental clearance \(Q\) (L/day) | 0.202 | 0.008416 L/h |
| Central volume \(V_c\) (L) | 4.25 | 4.25 L |
| Peripheral volume \(V_p\) (L) | 7.8 | 7.8 L |
| Absorption rate constant \(K_a\) (1/day) | 0.148 | 6.167e-3 /h |
| Subcutaneous bioavailability \(F\) | 0.511 | 0.511 |
Interindividual variability. None. The parameters are fixed point values inside a systems model.
Residual error. Not reported.
Covariates. None carried.
Binding constants, which the other entries do not have. CD3: \(k_{on}\) 0.00198 /pM/h, \(k_{off}\) 82.17 /h, so \(K_D\) = 41.5 nM. BCMA: \(k_{on}\) 0.0054 /pM/h, \(k_{off}\) 0.162 /h, so \(K_D\) = 30 pM. Receptor densities: CD3 60,000 per T cell, membrane BCMA 12,590 per myeloma cell. Soluble BCMA degrades at 0.01925 /h and is carried as a drug sink.
A CD3 affinity of 41.5 nM puts elranatamab near the weak end of the range Pearce et al. simulate, where CD3-mediated clearance contributes little and the intrinsic IgG half-life survives. That is consistent with elranatamab having the longest reported half-life in the class, 22 days.
3.5 Epcoritamab
Study and data. EPCORE NHL-1 and related studies, relapsed or refractory B-cell non-Hodgkin lymphoma, subcutaneous only. Doses 1.5 to 60 mg full dose. Clin Pharmacokinet 2025, doi:10.1007/s40262-024-01464-2.
Structure. Quasi-steady-state approximation of a two-compartment TMDD model with first-order absorption. This is the only published TCE model that carries TMDD in its declared structure rather than as a time-dependent clearance.
Fixed effects. Not transcribed. Derived quantities reported at the 48 mg full dose: apparent total clearance 0.53 L/day (40% CV) after end of cycle 3, apparent volume of distribution 25.6 L (82%), terminal half-life 22 days (58%), median time to maximum concentration 4 days after the first full dose and 2.3 days at end of cycle 3. Exposure increased more than proportionally from 1.5 to 48 mg.
Interindividual variability. Not transcribed.
Residual error. Not transcribed.
Covariates. Age and body weight reported as significant; coefficients not transcribed.
The more-than-proportional exposure increase is the nonlinearity that the odronextamab and mosunetuzumab models express as time dependence. Epcoritamab is the entry that would let the two parameterizations be compared on the same mechanism, which is why its structural parameters are the highest-value gap in this document.
3.6 Talquetamab
Study and data. MonumenTAL-1 phase I/II, relapsed or refractory multiple myeloma, intravenous and subcutaneous. Clin Pharmacol Ther 2025, PMC12439005.
Structure. Two compartments, sequential zero- then first-order subcutaneous absorption, parallel time-independent linear clearance and time-dependent linear clearance.
Fixed effects. Not transcribed; the parameter table is in the supplement. Reported: median terminal half-life 7.56 days at initial treatment and 12.2 days at steady state, with steady state reached at week 16 for both recommended phase II doses.
Interindividual variability. Not transcribed.
Residual error. Not transcribed.
Covariates. Body weight on clearance at time zero and on central volume, carried a priori. Myeloma subtype (IgG versus non-IgG) and International Staging System stage on clearance at time zero. All other covariates tested were rejected, including baseline total T cells and anti-drug antibodies.
Baseline total T-cell count was tested as a covariate and rejected. If CD3-mediated clearance were the dominant route, T-cell count should carry signal, and it does not.
3.7 Glofitamab
Study and data. NP30179 and related, relapsed or refractory B-cell non-Hodgkin lymphoma, intravenous, with mandatory 1000 mg obinutuzumab pretreatment seven days before the first glofitamab dose. Djebli et al., Blood 2020;136(Suppl 1):1, conference abstract.
Structure. Two compartments with linear clearance. No time-varying and no concentration-varying term. The obinutuzumab pretreatment removes the CD20-bearing target before glofitamab arrives.
Fixed effects, interindividual variability, residual error. Not transcribed. The source is an abstract.
Covariates. Body weight on volumes and clearances.
3.8 Blinatumomab
Study and data. Pooled analyses supporting the label. Continuous intravenous infusion.
Structure. One compartment, linear.
Fixed effects. Clearance 2.92 L/h, so 70.1 L/day, with a standard deviation of 2.83 L/h. Terminal half-life 2.11 h (SD 1.42). Volume based on the terminal phase \(V_z\) 4.52 L (SD 2.89).
Interindividual variability, residual error, covariates. Not transcribed.
At 54 kDa blinatumomab is below the glomerular filtration threshold and carries no Fc domain, so neither neonatal Fc receptor recycling nor a molecular size that resists filtration applies. Its clearance is 370 times odronextamab’s. It is in this table as the format against which the others are read, not as a comparable entry.
4. What Is Missing from the nlmixr2 Model Library
nlmixr2lib (nlmixr2.github.io/nlmixr2lib) holds 2769 model files at version 0.3.2.9000, and was the default source for this page. It is maintained by Bill Denney, with Richard Hooijmaijers, Matthew Fidler and Kiranmayi Vedantham.
4.1 What Is Already There
| Model | Drug | What it carries |
|---|---|---|
Miao_2023_teclistamab |
Teclistamab | Population PK, complete, with covariates and the rejected-covariate list |
Bender_2024_mosunetuzumab |
Mosunetuzumab | Population PK, complete, plus competitive CD20 receptor occupancy |
Poels_2025_elranatamab_qsp |
Elranatamab | QSP, with preliminary PK fixed inside it |
ChandralayamAyyappaMenon_2026_teclistamab_qsp |
Teclistamab | QSP |
ChandralayamAyyappaMenon_2026_isb2001_qsp |
ISB 2001 | QSP |
Willemin_2024_interleukin6_cyp_talquetamab |
Talquetamab | IL-6 and CYP interaction, not a PK model |
Fiandaca_2025_mRNABiTE_scm, _lcm |
mRNA-encoded BiTE | PK of an mRNA-delivered engager |
Generic TMDD structures are present and usable: PK_1cmt_tmdd_full, PK_1cmt_tmdd_qss, PK_1cmt_tmdd_mm, PK_2cmt_tmdd_qss, PK_2cmt_tmdd_mm, Duffull_2025_mab_tmdd_qss and Duffull_2025_mab_tmdd_simplified.
4.2 What Is Not There
Searched by drug name across inst/, R/ and man/. No match for:
| Drug | Published model that would fill it | What it adds |
|---|---|---|
| Odronextamab | Kovalenko et al., CPT Pharmacometrics Syst Pharmacol 2026, PMC12823311, open access | The only published TCE model with an explicit Michaelis-Menten elimination and a fitted \(k_m\). Full parameter table, bootstrap, and covariate effects on \(V_{\max}\) are all in the open-access text. |
| Epcoritamab | Clin Pharmacokinet 2025, doi:10.1007/s40262-024-01464-2 | The only quasi-steady-state TMDD structure fitted to a TCE. Pairs with the generic PK_2cmt_tmdd_qss already in the library. |
| Talquetamab | Clin Pharmacol Ther 2025, PMC12439005 | Sequential zero- then first-order absorption with parallel time-independent and time-dependent clearance. No other model in the library combines those. |
| Glofitamab | Djebli et al., Blood 2020;136(Suppl 1):1 | Linear, and the one TCE given after target debulking. The negative control for every time-dependent clearance model in the library. Abstract-only, which may make it out of scope. |
| Blinatumomab | Zhu et al., Clin Pharmacokinet | The non-Fc BiTE format, 2-hour half-life, continuous infusion. |
| Tarlatamab | Clin Pharmacokinet 2025, PMID 40261494 | The half-life-extended solid-tumor format, and a DLL3 target with no published TMDD data. |
| Linvoseltamab | Not located | Reported half-life changes from 1.75 to 21 days between first dose and steady state, the largest such change in the class. |
4.3 The Note to Bill
Ready to send, and it asks two things.
First, four population pharmacokinetic models to add, in priority order: odronextamab (Kovalenko 2026, PMC12823311, open access, full parameter table with bootstrap), epcoritamab (Clin Pharmacokinet 2025, doi:10.1007/s40262-024-01464-2), talquetamab (Clin Pharmacol Ther 2025, PMC12439005, parameters in the supplement) and tarlatamab (Clin Pharmacokinet 2025, PMID 40261494). With
Miao_2023_teclistamab,Bender_2024_mosunetuzumaband the elranatamab parameters insidePoels_2025_elranatamab_qspalready present, those four would make the library complete for the approved T-cell engagers. The library would then hold, for a single mechanism, four different parameterizations of it: explicit Michaelis-Menten with a decaying \(V_{\max}\), quasi-steady-state TMDD, an exponentially decaying clearance component, and a parallel time-dependent linear clearance. That set is hard to assemble from anywhere else and is directly useful for teaching which parameterization to reach for.Second, a question about metadata. Teclistamab’s, mosunetuzumab’s and talquetamab’s models all describe a decaying clearance as an approximation of target-mediated drug disposition, and none of them can be found by searching for TMDD. Is there a place in the model metadata for the mechanism a model approximates, as distinct from the structure it uses? Something like a
mechanismfield alongsidedescriptionwould makemodellib()able to answer “show me every model of target-mediated clearance”, which currently returns the structural TMDD templates and misses every drug that actually exhibits it.
Not sent.