Epcoritamab Step‐Up Dosing Regimen Selection and Optimization Using Repeated Time‐to‐Event Modeling for Cytokine Release Syndrome Risk Mitigation

Abstract

Epcoritamab is a CD3 × CD20 T‐cell–engaging bispecific antibody approved for the treatment of various types of relapsed/refractory (R/R) large B‐cell lymphoma (LBCL) and R/R follicular lymphoma (FL), after at least two lines of systemic therapy. Here, we develop and calibrate repeated time‐to‐event models to assess the impact of optimized step‐up dosing (SUD) regimens and the effect of intravenous fluids and/or corticosteroids on cytokine release syndrome (CRS) risk. The analysis used pooled data from 600 patients with aggressive non‐Hodgkin lymphoma (aNHL) and indolent NHL (iNHL) who received subcutaneous epcoritamab in 28‐day cycles in the EPCORE® NHL‐1 and EPCORE® NHL‐3 studies (NCT03625037, NCT04542824). In the calibrated model, prior CAR T cell therapy (125/600 patients [20.8%]) was associated with a 69.7% reduction (95% confidence interval [CI], 38.2–85.2) in the maximum stimulatory effect of epcoritamab on the hazard of Grade ≥2 CRS. Intravenous fluids or dexamethasone during Cycle (C)1 were associated with a 2.89‐fold (95% CI, 1.57–5.30) increase in the half‐maximal effective plasma concentration of epcoritamab on stimulation (S50). Furthermore, prophylaxis with intravenous fluids and dexamethasone during C1 was associated with a 3.79‐fold (95% CI, 1.65–8.73) increase in S50. Simulations show that the use of dexamethasone further reduces Grade ≥2 CRS risk compared with prednisone in patients with aNHL/iNHL. Moreover, a 3‐SUD design further reduces CRS risk in patients with iNHL. Overall, our models showed that the approved 2‐SUD regimen for R/R LBCL and 3‐SUD regimen for R/R FL—together with intravenous fluids and dexamethasone prophylaxis—are adequate to reduce Grade ≥2 CRS risk.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Cytokine release syndrome (CRS) is a common adverse event occurring during treatment with T‐cell–engaging agents, such as epcoritamab. Due to an incomplete understanding of the pathophysiology of CRS, determining the most optimal epcoritamab dosing regimen and prophylaxis to reduce the risk of CRS events remains a challenge.

• WHAT QUESTION DID THIS STUDY ADDRESS?

How can we effectively determine optimized step‐up dosing regimens of subcutaneous epcoritamab and prophylaxis strategies in patients with aNHL/iNHL via modeling of CRS to mitigate the risk of occurrence of this adverse event?

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

This study shows the development, calibration, and application of repeated time‐to‐event models to characterize CRS events.

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

The outcomes of this study facilitate more precise model‐informed dosing regimens and prophylactic strategies for subcutaneous epcoritamab.

Epcoritamab is a subcutaneously administered CD3 × CD20 T‐cell–engaging bispecific antibody that induces T‐cell–mediated killing of malignant CD20‐positive B cells. ¹ Epcoritamab monotherapy is approved in the USA, the EU, and Japan for the treatment of various types of relapsed/refractory (R/R) large B‐cell lymphoma (LBCL) and R/R follicular lymphoma (FL), after ≥2 lines of systemic therapy, ² , ³ based on the results from the EPCORE® NHL‐1 (US/EU; NCT03625037) and NHL‐3 (Japan; NCT04542824) trials. ⁴ , ⁵ , ⁶

EPCORE NHL‐1, a multicenter, multicohort, single‐arm, phase 1/2 trial, demonstrated meaningful and durable clinical activity and manageable safety of epcoritamab in patients with R/R LBCL and R/R FL. ⁴ , ⁵ , ⁷ Cytokine release syndrome (CRS) was the most common treatment‐emergent adverse event (TEAE) in patients with LBCL (78/157 [49.7%]) and FL (85/128 [66.4%]). ⁴ , ⁵ Grade 1/2 CRS events were reported in 74 (47.1%) and 83 (64.8%) of patients with LBCL and FL, respectively, and Grade 3/4 CRS events were reported in 4 (2.5%) and 2 (1.6%) patients with LBCL and FL. ⁴ , ⁵ Most CRS events were low grade, predictable, and occurred consistently early, after the first full dose of epcoritamab with a median time to onset of 20 hours in patients with LBCL and 15.3 hours in patients with FL. ⁴ , ⁵

CRS is a frequent TEAE occurring in patients receiving T‐cell–engaging bispecific antibodies and CAR T cell therapies. ⁸ It is induced by the overactivation of immune effector cells and increased levels of various proinflammatory cytokines, including interleukin‐1, interleukin‐6, interferon‐γ, and granulocyte‐macrophage colony‐stimulating factor. ⁸ A common strategy to lower cytokine levels and CRS risk with T‐cell–engaging bispecifics is subcutaneous administration, which delays and reduces peak drug concentrations compared with intravenous (IV) administration. ⁹ The symptoms and severity of CRS are generally attenuated with subsequent dosing; early exposure to low levels of immune‐activating agents seems to mitigate the occurrence and severity of potential CRS. As a result, regimens involving step‐up doses (SUD) are commonly used. ⁹ , ¹⁰ , ¹¹ Furthermore, prophylaxis and management of CRS typically include IV fluid support and administration of corticosteroids. ¹²

Epcoritamab is administered in patients with R/R LBCL, including diffuse LBCL (DLBCL), in 28‐day cycles. The recommended dosing includes a 2‐SUD regimen in Cycle (C)1, consisting of priming and intermediate doses (0.16 mg on Day [D]1 of C1 and 0.8 mg on C1D8, respectively), followed by full 48 mg doses from C1D15 onward. ³ , ¹³ For the treatment of R/R FL, a 3‐SUD regimen, consisting of a priming (0.16 mg on C1D1) dose and two intermediate (0.8 mg C1D8 and 3 mg C1D15) doses, followed by full 48 mg doses from C1D22 onward, is approved. ³ , ¹³ To reduce the risk of CRS, the recommended prophylaxis is based on premedication with dexamethasone (15 mg oral or IV), as the preferred corticosteroid, or prednisolone (100 mg oral or IV) or equivalent, prior to each weekly administration of epcoritamab, and for three consecutive days following each weekly administration. Additionally, patients receive diphenhydramine (50 mg oral or IV) or equivalent and acetaminophen (650 to 1,000 mg oral), prior to each weekly administration of epcoritamab.

During early clinical development, it was unclear whether administration of corticosteroids or IV fluids during treatment with epcoritamab would be effective in preventing CRS events, and which SUD regimens best reduce the risk of CRS in patients with R/R aggressive non‐Hodgkin lymphoma (aNHL) or indolent non‐Hodgkin lymphoma (iNHL). Therefore, the objectives of these analyses were to use novel repeated time‐to‐event (RTTE) models of CRS in response to epcoritamab exposure to support the selection of optimized SUD regimens and different prophylaxis methods (IV fluids and/or dexamethasone), and to reduce the incidence of Grade ≥2 CRS in patients with R/R aNHL, including (D)LBCL, and in patients with R/R indolent iNHL, including FL.

METHODS

Study design and input data

Data were pooled from patients with aNHL or iNHL from the EPCORE NHL‐1 and EPCORE NHL‐3 studies. ¹ , ⁴ , ⁵ , ⁶ , ⁷

EPCORE NHL‐1 is a first‐in‐human, open‐label, dose‐escalation, expansion, and optimization trial of epcoritamab in patients with relapsed, progressive, or refractory B‐cell lymphomas, including (D)LBCL and FL. ¹ , ⁴ , ⁵ , ⁷ Patients in the dose‐escalation cohort received subcutaneous epcoritamab in 28‐day cycles, with priming, intermediate, and full doses ranging from 0.0128 to 60 mg. ¹ In the expansion cohort, patients with either R/R DLBCL or FL received subcutaneous epcoritamab in 28‐day cycles, with 2‐SUD consisting of a 0.16 mg priming dose on C1D1 and a 0.8 mg intermediate dose on C1D8, followed by full 48 mg doses on C1D15 and C1D22. Afterward, full 48 mg epcoritamab doses were administered once weekly (QW) in C2–3 (D1, D8, D15, D22); every 2 weeks (Q2W) in C4–9 (D1, D15); and every 4 weeks (Q4W) in C10+ (D1) until disease progression or unacceptable toxicity. ⁴ , ⁵ During dose escalation and expansion parts, prophylaxis for CRS included prednisolone 100 mg once daily (QD) orally or IV as the recommended corticosteroid (alternative corticosteroid equivalent, for example, dexamethasone, was permitted), prior to each epcoritamab dose in C1 and on D2–4, D9–11, D16–18, and D23–25 and diphenhydramine 50 mg QD orally or IV (or equivalent) and acetaminophen 650–1,000 mg QD orally administered on D1, D8, D15, and D22 of C1. If Grade ≥2 CRS occurred after the fourth epcoritamab administration on C1D22, corticosteroids were given for 4 days or until resolution of the CRS. ⁵ Hydration with IV fluids and prophylactic dexamethasone 15 mg, as the preferred corticosteroid, were evaluated during C1 of the dose optimization part to further reduce risk and severity of CRS. ⁴ , ⁵ In the optimization cohort, patients with R/R FL received a second intermediate dose (3 mg or 6 mg) on C1D15 in place of the first full 48 mg dose. ⁴

EPCORE NHL‐3 is an open‐label, dose‐escalation and expansion study of epcoritamab alone or in combination with other treatments in patients with R/R B‐cell NHL in Japan. ⁶ Only patients with aNHL or iNHL who were enrolled in the dose‐escalation and expansion parts who received epcoritamab monotherapy were included in the present study. Patients received subcutaneous epcoritamab until disease progression or unacceptable toxicity in 28‐day cycles as follows: in C1, epcoritamab was administered following a 2‐SUD regimen including a 0.16 mg dose on D1 and a 0.8 mg dose on D8, followed by full doses on D15 and D22 (24–48 mg in the dose‐escalation part, 48 mg as the recommended phase 2 dose in the expansion part). Epcoritamab was administered QW in C1–3, Q2W in C4–9, and Q4W C10+. Prophylaxis for CRS included prednisolone 100 mg oral or IV (or alternative corticosteroid equivalent), diphenhydramine 50 mg IV, and acetaminophen 650–1,000 mg administered orally prior to each epcoritamab dose in C1. Additionally, prednisolone 100 mg was given QD on D2–4, D9–11, D16–18, and D23–25 of C1. ⁶ , ¹⁴

A previously developed population pharmacokinetic (PPK) model indicated that the observed concentration–time data of epcoritamab were adequately described by a 2‐compartment target‐mediated drug disposition model with subcutaneous absorption. ¹⁵ The PPK model was employed together with individual empirical Bayes estimates (EBEs) for CRS RTTE models calibration. For subsequent simulations, epcoritamab plasma concentration profiles of virtual patients were generated with varying PPK model covariate values (body weight, age, and baseline tumor size) based on observed means and variances and typical values of PPK model parameters. The list of covariates and their values are presented in Table S1 .

The following covariates were considered potential modifiers for the Grade ≥2 CRS RTTE analysis: disease type (aNHL/iNHL); CRS prophylaxis administered with the first four epcoritamab doses in C1 (no administration of IV fluids nor dexamethasone; administration of IV fluids or dexamethasone; administration of both IV fluids and dexamethasone); and prior chimeric antigen receptor (CAR) T cell therapy (yes/no). Patients who received prednisolone and did not receive dexamethasone as prophylactic corticosteroid were included in either the “no IV fluids nor dexamethasone” or “administration of IV fluids or dexamethasone,” depending on whether they received IV fluids as CRS prophylaxis. Receipt of dexamethasone and IV fluids during treatment to manage CRS adverse events were not considered for patient categorization.

Model development

The base model for the development of the model for Grade ≥2 CRS events for the RTTE analysis is based on a hazard function that varies as a function of the longitudinal exposure. In this method, ¹⁶ the likelihood contribution was:

$$L\left(t\right)=f{\left(t\right)}^{\delta \left(t\right)}\times S{\left(t\right)}^{1-\delta \left(t\right)}$$

$$=h{\left(t\right)}^{\delta \left(t\right)}\times S\left(t\right)$$

$$=h{\left(t\right)}^{\delta \left(t\right)}\times \exp \left\{-{\int }_{-t_{\left(-1\right)}}^{t}h\left(t\right)\mathit{dt}\right\}$$

where t is the post‐randomization time of a CRS event or the censoring time. The indicator δ(t) = 1 when t corresponds to an observed CRS time, and δ(t) = 0 when t corresponds to an administrative censoring time. Integration of the hazard occurs from the time of a previous event or enrollment (t ₍₋₁₎) to the event at time t.

The hazard function h of an individual i varied along epcoritamab exposure (Cp _i (t)) and other covariates (X _i ), and it is parameterized by fixed effects (θ _p , where p indexes parameter) and random effects η _ik (for individual i and parameter k), such that:

$${\eta }_{\mathit{ik}}\sim \mathrm{MVN}\left(0,\mathrm{\Omega }\right)\text{and}{h}_i\left(t\right)=g\left(t,{\mathit{Cp}}_i\left(t\right),X_i,\theta ,{\eta }_i\right)$$

Clinical experience with epcoritamab showed that CRS events tend to occur early in a patient's dosing history, indicating a decreasing risk of CRS occurrence with accumulated exposure (accumulation of tolerance). To reflect this effect, the model was implemented with two components; the first component (STIM) allowed for the hazard to increase as a function of time and epcoritamab plasma concentration, while the second component (EFF_Inh) allowed for the hazard to decrease as a function of time and epcoritamab plasma concentration (Figure 1 ):

$$h\left(t\right)=\text{STIM}\left(t\right)\times {\mathrm{EFF}}_{\mathrm{Inh}}\left(t\right)$$

$$h_i\left(t\right)=g_{\text{STIM}}\left(t,{\mathit{Cp}}_i\left(t\right),X_i,{\eta }_i\right)\times g_{{\mathrm{EFF}}_{\mathrm{Inh}}}\left(t,{\mathit{Cp}}_i\left(t\right),X_i,{\eta }_i\right)$$

where STIM _i (t) ≥ 0 and 0 ≤ EFF_Inh,i (t) ≤ 1. A more detailed description of STIM and EFF_Inh components is included in the supporting information.

Figure 1.

RTTE model schematic. C _p, plasma concentration; CRS, cytokine release syndrome; EFF, component describing the inhibition of the hazard of CRS; h(t), hazard function; I50, half‐maximal effective C _p on inhibition; IHILL, Hill parameter for inhibition; IMAX, maximum drug effect on hazard inhibition; K _in, rate constant for production; K _out, rate constant for elimination; PK, pharmacokinetic; RTTE, repeated time‐to‐event; S50, half‐maximal effective C _p on stimulation; SHILL, Hill parameter for stimulation; SMAX, maximum drug effect on hazard stimulation; STIM, component addressing the onset of an adverse event risk.

Covariates were included as an exponentiated linear combination of parameters and covariates, adjusting typical‐value structural parameters in the model. Covariate inclusion was guided through simulation‐based model evaluation, diagnostics, and Akaike information criterion. Final models included only those covariates with significant effects, defined as whether the 95% confidence interval (CI) of the estimate excluded the null value.

Model evaluation

To assess the model's ability to replicate both the probability of CRS and the number of CRS events, visual predictive checks (VPCs) were used for model evaluation by comparing the simulated numbers of events per individual with the observed number of events per individual. To assess the model's ability to describe the exposure (and tolerance) effect, a time‐to‐first‐event VPC was created, and Kaplan–Meier (KM) curves were fitted to each simulation replicate and binned by one‐day increments. Parameter uncertainty was included in the VPCs, with parameter sets for each simulation replicate drawn from a multivariate normal distribution using mean parameter estimates and the estimated variance–covariance matrices among parameters.

Model‐based simulation

The final RTTE model was applied to predict the probability of experiencing a Grade ≥2 CRS event under different prophylactic approaches, SUD regimens, and other selected covariates that may influence CRS risk.

From the parameter uncertainty distribution, 200 samples were used to simulate CRS events for 1,000 individuals per sample over a period of 2 months under different SUD regimens. Pharmacokinetic profiles were generated using body weights, ages, and baseline tumor sizes from the observed means and variances of these variables in the studied pooled population.

Simulations were conducted for patients with aNHL receiving 2‐SUD epcoritamab regimens, including combinations of 20 priming doses (0.001–10 mg) and 20 intermediate doses (0.001–10 mg). Three additional 2‐SUD regimens (priming/intermediate/full‐dose levels: 0.16/0.8/48 mg; 0.32/1.6/48 mg; and 0.64/3/48 mg, respectively) were selected for further simulation, along with various CRS prophylaxis strategies, including administration or lack of IV fluids and/or dexamethasone during C1 epcoritamab dosing. Simulations of the rapid ramp‐up 2‐SUD regimen with priming/intermediate/first full doses on D1, D4, and D8 of C1, respectively, to shorten the time to first full‐dose administration were also carried out for patients with aNHL, with or without prior CAR T cell therapy receipt.

In patients with iNHL, simulations of 2‐SUD (priming/intermediate/full‐dose levels: 0.16/0.8/48 mg, respectively) and 3‐SUD regimens (priming/first intermediate/second intermediate/full‐dose levels: 0.16/0.8/3.0/48 mg and 0.16/0.8/6.0/48 mg) were compared. Administration/lack of IV fluids and/or dexamethasone for CRS prophylaxis during C1 epcoritamab dosing were also compared through simulation.

Software

Data manipulation, visualization, and simulations were conducted using R version 4.1 (R Core Team 2021) (R Foundation for Statistical Computing, Vienna, Austria). ¹⁷ RTTE modeling was performed using NONMEM® version 7.5.0 (NONMEM). ¹⁸

RESULTS

Patient population

The dataset included data from patients enrolled in the dose‐escalation, expansion and optimization parts of the NHL‐1 trial, and in the dose‐escalation and expansion parts of the NHL‐3 trial (Table 1 ). Data from a total of 600 patients with either aNHL or iNHL, distributed in near‐equal proportions (331 [55.2%] and 269 [44.8%], respectively), were used for this analysis. Most patients did not receive prior CAR T cell therapy (475 [79.2%]) and did not receive dexamethasone as the preferred corticosteroid or IV fluids for CRS prophylaxis (373 [62.2%]).

Table 1.

Baseline demographics and patient characteristics

Category	EPCORE NHL‐1 (N = 536)		EPCORE NHL‐3 Dose escalation/dose expansion (N = 64)	Summary (N = 600)
	Dose escalation/dose expansion (n = 364)	Dose optimization (n = 172)
	n (%)	n (%)	n (%)	n (%)
Indication
aNHL	198 (54.4)	91 (52.9)	42 (65.6)	331 (55.2)
iNHL	166 (45.6)	81 (47.1)	22 (34.4)	269 (44.8)
Prior CAR T cell therapy
No prior CAR T cell therapy	293 (80.5)	120 (69.8)	62 (96.9)	475 (79.2)
Prior CAR T cell therapy	71 (19.5)	52 (30.2)	2 (3.1)	125 (20.8)
CRS prophylaxis (dexamethasone/IV fluids)
No dexamethasone; no IV fluids	293 (80.5)	16 (9.3)	64 (100.0)	373 (62.2)
No dexamethasone; IV fluids	0	30 (17.4)	0	30 (5.0)
Dexamethasone; no IV fluids	71 (19.5)	20 (11.6)	0	91 (15.2)
Dexamethasone; IV fluids	0	106 (61.6)	0	106 (17.7)

aNHL, aggressive non‐Hodgkin lymphoma; CAR, chimeric antigen receptor; CRS, cytokine release syndrome; iNHL, indolent NHL; IV, intravenous(ly).

Model development and evaluation

Table 2 reports parameter estimates for the final model of Grade ≥2 CRS in the overall population of patients with aNHL or iNHL. The final model was fitted with the effect of prior CAR T cell therapy history on the maximum epcoritamab effect on CRS hazard stimulation (SMAX) and the effect of C1 CRS prophylaxis on the half‐maximal effective epcoritamab plasma concentration on stimulation (S50) as covariates. The selected model included three categories for C1 CRS prophylaxis (no administration of IV fluids nor dexamethasone; administration of IV fluids or dexamethasone; administration of both IV fluids and dexamethasone). Overall, VPCs for the final model showed no deficiencies (Figure 2 , Figures S1–S8 ). Baseline characteristics such as general demographic parameters, prior lines of therapy, common baseline laboratory values, and the study effect (NHL‐1 vs. NHL‐3) were explored as potential covariates but did not demonstrate statistical significance warranting inclusion in the final model.

Table 2.

Summary of parameter estimates for the Grade ≥2 CRS RTTE full covariate model with three categories for C1 prophylaxis

Parameter	Description	Estimate (% RSE)	Standard error	Shrinkage	Transformed estimate (95% CI)
θ1	Inhibition accumulation rate (K in, 1/day)	−0.832 (46.2)	0.385	–	0.435 (0.205–0.925)
θ2	Maximum drug effect on hazard stimulation (SMAX)	0.0539 (1400.0)	0.757	–	1.06 (0.239–4.65)
θ3	Half‐maximal effective C p, stimulation (S50, mg/L)	−2.36 (16.5)	0.389	–	0.0943 (0.0440–0.202)
θ4	Half‐maximal effective C p, inhibition (I50, mg/L)	−7.72 (5.29)	0.408	–	0.000446 (0.0002–0.000991)
θ5	Maximum drug effect on hazard inhibition (IMAX)	1.0 (−)	–	–	1.0 (NA)
θ6	Hill parameter for inhibition (IHILL)	1.0 (−)	–	–	1.0 (NA)
θ7	Hill parameter for stimulation (SHILL)	−0.124 (143.0)	0.177	–	0.884 (0.625–1.25)
θ8	Effect of prior CAR T cell therapy = yes on θ1	0 (−)	–	–	1.0 (NA)
θ9	Effect of prior CAR T cell therapy = yes on θ2	−0.964 (31.9)	0.308	–	0.382 (0.209–0.698)
θ10	Effect of prior CAR T cell therapy = yes on θ3	0 (−)	–	–	1.0 (NA)
θ11	Effect of study = NHL‐3 on θ3	0 (−)	–	–	1.0 (NA)
θ12	Effect of dexamethasone or fluids on θ3	1.19 (33.9)	0.403	–	3.28 (1.49–7.22)
θ13	Effect of dexamethasone and fluids on θ3	1.63 (35.8)	0.583	–	5.10 (1.63–16.0)
ω1,1	IIV on θ1	0 (−)	–	0	0 (NA)
ω2,2	IIV on θ2	0 (−)	–	0	0 (NA)
ω3,3	IIV on θ3	0 (−)	–	0	0 (NA)
ω4,4	IIV on θ4	0 (−)	–	0	0 (NA)

Model includes overall population of patients with aNHL or iNHL. Parameters fixed to 0 were not used in the model but are included for context on parameters considered during model building. Parameters fixed at 1 are used in the model at those fixed values.

aNHL, aggressive non‐Hodgkin lymphoma; C, cycle; CAR, chimeric antigen receptor; CI, confidence interval; C_p, plasma concentration; CRS, cytokine release syndrome; IIV, inter‐individual variability; iNHL, indolent NHL; K_in, rate constant for production; NA, not applicable; RSE, relative standard error; RTTE, repeated time‐to‐event.

Figure 2.

Visual predictive checks for the Grade ≥2 CRS RTTE model. The model includes the overall population of patients with aNHL or iNHL. The top panel shows observed (black line) and model‐simulated (colored line = mean across simulations; shaded region = 95% CI across simulations) KM curve for the time to first Grade ≥2 CRS event. The bottom panel shows the observed distribution of the number of events per patient (black points) and model‐simulated values (colored points = mean across simulations; error bars = 95% CIs across simulations). aNHL, aggressive non‐Hodgkin lymphoma; CI, confidence interval; CRS, cytokine release syndrome; iNHL, indolent NHL; IV, intravenous(ly); KM, Kaplan–Meier; RTTE, repeated time‐to‐event.

The final model showed that receipt of prior CAR T cell therapy was associated with a proportional decrease in SMAX by 61.8% (exp(θ₉) = 0.382; 95% CI, 0.209–0.698), which indicates a lower risk of CRS (Table 2 ). Prophylactic treatments were associated with reduced Grade ≥2 CRS risk, as determined by increased S50, indicating that more drug was necessary to reach the half‐maximal effect on SMAX. Administration of either IV fluids or dexamethasone during C1 was associated with a 3.28‐fold (exp(θ₁₂) = 3.28; 95% CI, 1.49–7.22) increase in S50. Furthermore, prophylaxis with IV fluids and dexamethasone during C1 was associated with a 5.10‐fold (exp(θ₁₄) = 5.10; 95% CI, 1.63–16.00) increase in S50 (Table 2 ).

Dosing regimen simulations

Simulations of multiple 2‐SUD regimens were carried out for patients with aNHL from the NHL‐1 study. A grid search across priming and intermediate dose levels indicated that the recommended phase 2 2‐SUD regimen (priming/intermediate/full‐dose levels: 0.16/0.8/48 mg, respectively) fell within the region of SUD combinations, with the lowest model‐predicted probability of at least one Grade ≥2 CRS event (Figure 3 ). Based on the simulation results, alternative 2‐SUD regimens (0.32/1.6/48 mg; 0.64/3/48 mg) were identified with a similar model‐predicted probability of at least one Grade ≥2 CRS event and were recommended for further investigation in the optimization part of the NHL‐1 study. Simulations of SUD combinations with higher priming and intermediate doses were predicted to increase the risk of at least one Grade ≥2 CRS event (Figure 3 ).

Figure 3.

2‐SUD combination simulation grid for patients with aNHL, assuming no IV fluids/dexamethasone. aNHL, aggressive non‐Hodgkin lymphoma; CRS, cytokine release syndrome; IV, intravenous(ly); SUD, step‐up dosing. (=excluding limit value;] = including limit value.

Simulations based on the final model—including data from the optimization arm of NHL‐1—showed that the administration of prophylactic IV fluids and dexamethasone significantly contributes to reducing the probability of having at least one Grade ≥2 CRS event in patients with aNHL, while the two alternative SUD regimens have a minimal effect on reducing the risk of Grade ≥2 CRS events over the recommended phase 2 2‐SUD regimen (Figure 4 a ).

Figure 4.

(a) 2‐SUD combination simulations for patients with aNHL under different CRS prophylaxis combinations; (b) 2‐/3‐SUD combination simulations for patients with iNHL under different CRS prophylaxis combinations. Error bars represent 90% confidence intervals for the probability of at least one Grade ≥2 event within 2 months of dosing; hollow black points represent observed fractions of patients with at least one Grade ≥2 CRS event. (a) No IV fluids or dexamethasone, 0.16/0.8/48/48 mg, n = 108; IV fluids or dexamethasone, 0.16/0.8/48/48 mg, n = 61; IV fluids and dexamethasone, 0.16/0.8/48 mg, n = 40. Only patient data from expansion and optimization cohorts of NHL‐1 and from the expansion cohort of NHL‐3 at the recommended phase 2 dosing regimen are presented; for remaining regimens and use of dexamethasone/IV fluids categories no observed probabilities are reported in the figure, due to either no patient data available or limited number of patients that may cause a bias. (b) IV fluids and dexamethasone, 0.16/0.8/3/48 mg, n = 52; IV fluids or dexamethasone, 0.16/0.8/48/48 mg, n = 32; No IV fluids or dexamethasone, 0.16/0.8/48/48 mg, n = 93; for remaining regimens and use of dexamethasone/IV fluids categories no observed probabilities are reported in the figure, due to either no patient data available or limited number of patients that may cause a bias. aNHL, aggressive non‐Hodgkin lymphoma; CAR, chimeric antigen receptor; CRS, cytokine release syndrome; iNHL, indolent non‐Hodgkin lymphoma; IV, intravenous(ly); SUD, step‐up dosing.

Finally, model‐predicted probabilities of at least one Grade ≥2 CRS event for patients with iNHL were directionally lower when IV fluids and/or dexamethasone were administered together with epcoritamab during C1 vs. no administration of IV fluids or dexamethasone, with 3‐SUD regimens further contributing to the reduced risk of at least one Grade ≥2 CRS event compared with the 2‐SUD regimen (Figure 4 b ).

Accelerated SUD regimen simulations

Simulations were performed to explore the feasibility of shortening the established 2‐week 2‐SUD regimen in patients with aNHL. Shortening the 2‐week 2‐SUD regimen generally increases the Grade ≥2 CRS risk (Figure 5 ), but in subpopulations with a lower CRS risk (e.g., patients with prior CAR T cell therapy), the increase in Grade ≥2 CRS risk is relatively small (Figure 5 b,d ). Patients with prior CAR T cell therapy are predicted to have a Grade ≥2 CRS risk of ≤15% with a 1‐week rapid ramp‐up SUD (e.g., with an intermediate dose at D4 and a full dose at D8) regardless of whether they received IV fluids and dexamethasone (Figure 5 b,d ). Patients with prior CAR T cell therapy with a 1‐week SUD who also received IV fluids and dexamethasone were predicted to have a Grade ≥2 CRS occurrence risk lessened to ≤10% (Figure 5 d ).

Figure 5.

Model‐predicted Grade ≥2 CRS rates under different rapid SUD ramp‐up times in patients with aNHL. aNHL, aggressive non‐Hodgkin lymphoma; CAR, chimeric antigen receptor; CRS, cytokine release syndrome; IV, intravenous(ly); SUD, step‐up dosing.

DISCUSSION

CRS is a major challenge in clinical practice with T‐cell–engaging bispecific antibodies and CAR T cell therapies. As a frequent and potentially serious AE, CRS can compromise patient safety, require intensive management, and disrupt treatment delivery. Its systemic inflammatory response can range from mild symptoms to severe, life‐threatening complications, making effective prevention and management strategies essential for optimal patient outcomes. ⁸

The scope of this work was the development and validation of a model that would enable informed decision‐making regarding epcoritamab SUD regimens. Particularly, results from the model‐based simulations were leveraged to help accelerate the optimization of the SUD regimens for patients with R/R aNHL or iNHL.

This manuscript presents an RTTE model in which the hazard of CRS event occurrence varied with longitudinal epcoritamab concentration. In this model, the hazard function was specified as the product of two components that empirically capture the onset and tolerance dynamics of epcoritamab exposure on the hazard of CRS events.

Two key categorical covariates for the risk of Grade ≥2 CRS were incorporated into the model: receipt of prior CAR T cell therapy and CRS prophylaxis. CAR T cell therapy is known to be associated with high rates and severity of CRS; however, in this model, prior CAR T cell therapy was found to be associated with a lower risk of Grade ≥2 CRS events compared with the risk in patients who had not previously received CAR T cell therapy, which is consistent with clinical observations. This suggests that personalized CRS risk mitigation strategies, such as accelerated SUD regimens, may be considered in appropriately selected patients.

The model also tested CRS prophylaxis as a three‐category covariate (no administration of IV fluids nor dexamethasone; administration of either IV fluids or dexamethasone; administration of both IV fluids and dexamethasone). The decision to test three categories for the prophylaxis covariate instead of four categories (no administration of IV fluids nor dexamethasone; administration of IV fluids only; administration of dexamethasone only; administration of both IV fluids and dexamethasone) was made a priori based on the observed sample sizes in each of the four categories and preliminary modeling results (not shown here), and guided by the Akaike information criterion model selection criterion.

Generally, the final model described the data well, as indicated by the VPCs, although some biases in the time‐to‐first‐event KM VPCs were evident and most pronounced at the intermediate SUD and at the end of C1 for all grade CRS events. Specifically, the models failed to capture the rapid onset of CRS events immediately following administration of full epcoritamab doses during C1, which implies that this approximation of the occurrence mechanism of CRS events misses some elements of exposure–cytokine dynamics. The models tend to predict a more gradual accumulation of Grade ≥2 CRS events during C1, compared with the clinical observations of sharp drops in the fractions of patients with no Grade ≥2 CRS events starting from D14, the day prior to the receipt of the first full dose (Figure 2 , Figures S1 , S3 , S5 , S7 ). Simulations that change the weekly interval should be interpreted with caution, and model performance should be validated if additional data can be generated. However, the VPCs strongly support the use of the model for simulation‐based inference when the time horizon of interest is longer than ~30 days.

Overall, the model estimates and predictions support the use of prophylaxis (fluids and/or dexamethasone) with the first four epcoritamab doses in C1 to significantly reduce the probability of patients experiencing at least one Grade ≥2 CRS event. The probability of Grade ≥2 CRS events with CRS prophylaxis (IV fluids and/or dexamethasone) administered with the first four epcoritamab doses in C1 is even lower in patients who received prior CAR T cell therapy.

Using our model and simulations, we were able to screen numerous combinations of 2‐ and 3‐SUD regimens that would not be feasible in a clinical setting. For aNHL, given the aggressive nature of the disease, simulations focused on evaluating alternative 2‐SUD regimens delivering maximum CRS risk reduction, without compromising treatment and risking disease progression. Accordingly, for slow‐growing iNHL, the simulations prioritized exploration of 3‐SUD regimens to achieve maximal reduction of CRS risk. Potential alternative 2‐ and 3‐SUD regimens were identified for clinical testing in the ongoing optimization part of the NHL‐1 study. Simulations also indicate that the development of tolerance plays a notable role in the risk of CRS, and SUD regimens play a larger role in modulating the development of tolerance. The choice of SUD can further modulate CRS risk, but to a lesser degree.

In addition, we incorporated the effects of different CRS prophylaxis strategies as covariates, based on the findings from the optimization cohort of NHL‐1, which allowed an independent assessment of the impact of CRS prophylaxis and alternative SUD regimens on CRS risk. Model simulation results suggest that the current SUD regimens, including the recommended phase 2 SUD regimens, are the most optimal in minimizing rates of Grade ≥2 CRS events in patients with aNHL (2‐SUD) and iNHL (3‐SUD). Furthermore, in patients with aNHL, simulations showed that administration of IV fluids and/or dexamethasone with the first four doses of epcoritamab in C1 and/or receipt of prior CAR T cell therapy had a larger impact on reducing Grade ≥2 CRS event rates than the SUD regimen design itself. In practice, this suggests that receiving IV fluids and dexamethasone may have a larger impact on reducing Grade ≥2 CRS risk than epcoritamab administration via SUD (regardless of regimen design).

Model‐informed optimization of epcoritamab SUD regimens together with standard prophylaxis measures, such as fluids and dexamethasone, represents a practical and clinically implementable strategy to attenuate the incidence of CRS events. The subcutaneous administration of epcoritamab offers operational advantages over IV delivery, including reduced chair time and reduced costs at treatment centers. ¹⁹ , ²⁰

The model does not include or consider the use of a prophylactic anti‐CD20 agent (eg, obinutuzumab) as a pre‐treatment to deplete peripheral and tissue‐based B cells, which is typically administered to prevent CRS in certain treatments with T‐cell–engaging bispecific antibodies. ²¹ , ²² However, the model showed that the hazard of CRS is reduced to an acceptable level with the use of the approved epcoritamab SUD regimens and prophylaxis (IV fluids and dexamethasone).

The model was also used for SUD strategy optimization by exploring the feasibility and safety of accelerated SUD regimens. The standard 2‐SUD approach requires ~2 weeks to reach the full therapeutic dose of epcoritamab. For patients with aggressive disease, this delay in achieving full‐dose exposure may increase the risk of disease progression. The RTTE model predicted a relatively lower risk of Grade ≥2 CRS in patients with aNHL with prior CAR T cell therapy than in those without prior CAR T cell therapy, even after shortening the approved 2‐SUD regimen (0.16 mg/0.8 mg/48 mg full dose) from 2 weeks to 1 week. However, in patients who did not receive CAR T cell therapy, the risk of Grade ≥2 CRS further increases under the 1‐week rapid SUD regimen compared with the 2‐week SUD regimen. Based on the results of these simulations, a 1‐week rapid ramp‐up SUD may feasibly allow for an expedited receipt of the full dose without undue risk for patients previously treated with CAR T cell therapy, although a confirmation in a clinical setting would be needed. In addition, the simulations showed that the increased risk with the 1‐week rapid SUD regimen for patients with low baseline CRS risk may be small enough to be acceptable, given the potential benefit of receiving the full dose of epcoritamab sooner than with the 2‐week SUD regimen. Assessment of the rapid SUD regimen in a clinical study is warranted to confirm the simulation results.

In conclusion, we successfully applied an RTTE model, based on data pooled from 600 patients, to optimize and validate CRS risk prediction in patients with R/R aNHL or iNHL receiving epcoritamab via SUD regimens. This analysis provides justification for the administration of prophylactic IV fluids and dexamethasone with the currently approved 2‐ and 3‐SUD regimens in order to mitigate CRS Grade ≥2 risk in patients with R/R LBCL and FL, respectively, treated with subcutaneous epcoritamab. Furthermore, simulations showed that a 1‐week rapid ramp‐up SUD may allow for an expedited receipt of the full epcoritamab dose in certain patient subgroups at a lower risk for CRS events, such as patients with DLBCL previously treated with CAR T cell therapy.

FUNDING

This study was funded by Genmab A/S and AbbVie.

CONFLICT OF INTEREST

T.L., C.J.T., and S.X. are employed by and own stock/stock options in Genmab. M.P. is employed by, owns stock/stock options in, and receives support for attending meetings and/or travel and other financial/non‐financial support from Genmab. M.G. owns stock/stock options in Genmab. B.E. is an inventor on related patents, is employed by, owns stock/stock options in, and receives support for attending meetings and/or travel from Genmab. A.P. is employed by and owns stock/stock options in AbbVie. B.N. is employed by, owns stock/stock options in, and receives other financial/non‐financial support from AbbVie. M.‐E.F.M. is employed by, owns stock/stock options in, and receives support for attending meetings and/or travel from AbbVie. A.T. and D.P. are employed by Metrum Research Group, contracted by Genmab. S.L., K.S., and C.L.G. declare no conflicts of interest.

AUTHOR CONTRIBUTIONS

T.L., A.T., D.P., M.P., S.L., K.S., C.T., A.P., B.N., M.E.M., C.LG., B.E., M.G., and S.X. wrote the manuscript; T.L., C.T., A.P., B.N., and ME.M. designed the research; T.L., A.T., D.P., M.P., S.L., and C.L.G. performed the research; T.L., D.P., M.P., S.L., and C.L.G. analyzed the data.

Supporting information

Table S1.

DATA AVAILABILITY STATEMENT

De‐identified individual participant data collected during the trial will not be available upon request for further analyses by external independent researchers. Aggregated clinical trial data from the trial is provided via publicly accessible study registries/databases as required by law. For more information, please contact clinicaltrials@genmab.com.