Main

Diffuse large B cell lymphoma (DLBCL) is an aggressive form of B cell non-Hodgkin lymphoma (B-NHL)1. Approximately 30% of people will relapse after first-line treatment with immunochemotherapy (for example, rituximab, cyclophosphamide, doxorubicin, vincristine and prednisolone) and poor outcomes are observed in this relapsed/refractory (R/R) setting; the median overall survival (OS) is approximately 6–7 months in people with primary refractory disease2,3,4.

T cell-engaging therapies, including chimeric antigen receptor (CAR) T cell therapies and bispecific antibodies, are important modalities in the management of R/R DLBCL. CAR T cell treatments were initially approved in people with two or more prior therapy lines5,6,7,8,9,10 and have since received approval in DLBCL refractory to or relapsed after first-line immunochemotherapy8,10,11,12. The shift to earlier CAR T cell therapy necessitates the development of effective options in the third-line setting.

Bispecific antibodies, which bind T cells to a tumor antigen on cancer cells, have shown encouraging activity in solid and hematologic malignancies and are manufactured to allow off-the-shelf administration13,14,15. Recently, glofitamab and epcoritamab received accelerated US approval for DLBCL treatment after at least two prior lines of systemic therapy16,17.

Odronextamab is an Fc-silenced, human CD20×CD3 bispecific antibody that simultaneously engages CD20 on malignant B cells and CD3 on cytotoxic T cells to induce T cell-mediated cytotoxicity of the former18. The phase 1 dose-escalation and expansion ELM-1 study demonstrated encouraging activity and generally manageable safety of odronextamab monotherapy in patients with heavily pretreated B-NHL, including those with R/R DLBCL after CAR T cell therapy (at the active dose for aggressive lymphoma (≥80 mg); the objective response rate (ORR) was 33% in those who had previous CAR T cell therapy)19. Following further safety, pharmacokinetic and pharmacodynamic evaluation, the recommended dose for DLBCL was established as 160 mg (ref. 19). Here, we report the long-term efficacy and safety results of odronextamab in patients with R/R DLBCL from the phase 2 ELM-2 study (NCT03888105).

Results

Patient disposition and characteristics

Between March 24, 2020 and May 18, 2022, 127 patients with DLBCL were enrolled, treated with odronextamab and evaluated for efficacy and safety. At data cutoff (August 18, 2023), the median duration of exposure was 18.0 weeks (range: 0.9–168.1) and 19 (15.0%) patients remained in the study. In total, 91.3% of patients completed cycle 1 (C1); of these, 67 (52.8%) received the C1 regimen with a step up of 1 to 20 mg, and 60 (47.2%) received the regimen with a step up of 0.7 to 4 to 20 mg. In total, 63.0% of patients completed four or more cycles of odronextamab treatment. The most common reasons for treatment discontinuation were progressive disease (PD; 47.2%), death (15.7%) and adverse events (AEs; 13.4%), occurring in a similar proportion of patients by step-up regimen (Extended Data Fig. 1).

Anti-infection prophylaxis was recommended as part of a protocol amendment; eight patients (6.3%) received prophylaxis for cytomegalovirus (CMV) infection, 84 patients (66.1%) received prophylaxis for Pneumocystis jirovecii pneumonia and nine patients (7.1%) received intravenous (IV) immunoglobulin prophylaxis.

At baseline, the median age of patients was 67 years (range: 24–88), with 23.6% of patients aged ≥75 years (Table 1). Most patients (81.1%) had advanced disease (Ann Arbor stages III–IV) and 55.9% had high–intermediate or high International Prognostic Index scores. A total of 31 (24.4%) patients had transformed DLBCL (24, non-Richter’s; 7, Richter’s) and 11 (8.7%) had double-hit or triple-hit cytogenetic rearrangements by local assessment. The median number of prior therapy lines was two (range: 2–8), with 20.5% of patients having received four or more prior lines, 55.1% of patients being primary refractory, 64.6% of patients being double refractory to an alkylator and anti-CD20 antibody and 17.3% of patients with prior autologous stem cell transplantation (ASCT). Baseline demographics were generally similar irrespective of step-up dosing regimen used (1 to 20 mg versus 0.7 to 4 to 20 mg) (Supplementary Table 1).

Table 1 Patient demographics and baseline characteristics
Full size table

Efficacy

The primary endpoint of ORR per independent central review (ICR) was 52.0% (66/127 (95% confidence interval (CI): 42.9–60.9); Table 2). The complete response (CR) rate was 31.5% (40/127 (95% CI: 23.5–40.3)). The median time to response was 2.6 months (range: 0.8–6.4) and 87.9% (58/66) of patients who responded did so by their first assessment at week 12. Response rates by ICR were similar in patients who received the step-up regimen of 1 to 20 mg and in those who received the step-up regimen of 0.7 to 4 to 20 mg, with overlapping CIs for ORR, CR and partial response (PR; Supplementary Table 2). Response rates as reported by local investigator assessment were similar to those reported by ICR, with an ORR of 49.6% (63/127 (95% CI: 40.6–58.6)) and a CR rate of 38.6% (49/127 (95% CI: 30.1–47.6)).

Table 2 Odronextamab response summary according to ICR
Full size table

Tumor size was reduced in 78.9% (71/90) of patients with postbaseline imaging (Fig. 1). With a median efficacy follow-up of 29.9 months (95% CI: 20.4–32.6), the median duration of response (DOR) per ICR was 10.2 months (95% CI: 5.0–17.9). Patients with a best response of CR had a median CR duration of 17.9 months (95% CI: 10.2–not evaluable (NE)). Among the 21 patients who sustained a CR for 9 months and were eligible to transition to dosing once every 4 weeks, 18 transitioned and the median DOR from the time of transition was 18.5 months (95% CI: 6.0–NE).

Fig. 1: Waterfall plot of best percentage change from baseline in tumor sum of the products of the diameters.
figure 1

Data for each evaluable patient are shown as a separate bar on the figure (n = 90 patients). SD, stable disease.

Source data

Full size image

Odronextamab demonstrated antitumor activity in patients across a range of key subgroups, including in patients aged 75 years and older (ORR = 50.0%), in those with more than two lines of prior therapy (ORR = 46.7%) and in those who were double refractory to an alkylator and anti-CD20 antibody (ORR = 40.2%) (Fig. 2).

Fig. 2: Subgroup analysis of ORR by ICR.
figure 2

ORR data are presented as the mean values ± 95% CIs. The vertical dashed line indicates the ORR for all patients (N = 127).

Source data

Full size image

The median progression-free survival (PFS) was 4.4 months (95% CI: 3.6–5.9) and median OS was 9.2 months (95% CI: 6.5–12.7). The median PFS in patients with CR (20.4 months) was longer than that in those with PR (5.8 months; hazard ratio (HR) = 0.29 (95% CI: 0.2–0.5)), as was the OS (not reached (NR) versus 17.0 months, respectively; HR = 0.48 (95% CI: 0.2–1.0)) (Fig. 3a,b).

Fig. 3: PFS and OS in patients treated with odronextamab, for all patients and by best overall response.
figure 3

a, PFS. b, OS. Data are presented as Kaplan–Meier curves, with tick marks indicating patients with censored data (N = 127 patients; n = 40 with CR, n = 26 with PR). Median values with 95% CIs are presented alongside the respective curves.

Source data

Full size image

Biomarker assessment

Among 63 patients evaluable for circulating tumor DNA (ctDNA) assessment who had a response assessment at C4 day 15 (C4D15), all were positive for minimal residual disease (MRD+) at baseline. At C4D15, 43 remained MRD+ and 20 were MRD−. PFS was longer in patients who were MRD− by C4D15 versus those who were MRD+ (HR = 0.27; 95% CI: 0.12–0.62) (Fig. 4). MRD negativity also predicted PFS benefit in patients who did not achieve CR by positron emission tomography–computed tomography at C4D15 (MRD− versus MRD+: HR = 0.11; 95% CI: 0.03–0.49) (Extended Data Fig. 2).

Fig. 4: PFS by C4D15 ctDNA MRD status.
figure 4

Data are presented as Kaplan–Meier curves, with tick marks indicating patients with censored data (n = 63 ctDNA-evaluable patients; n = 20 with cleared MRD at C4D15, n = 43 with detected MRD at C4D15). Median values with 95% CIs are presented alongside the respective curves. The HR for PFS in patients with MRD cleared versus MRD detected was calculated by univariate Cox regression.

Source data

Full size image

Safety

Treatment-emergent AEs (TEAEs) were reported in 126 (99.2%) patients, with 111 (87.4%) patients experiencing at least one treatment-related (per investigator assessment) TEAE (Table 3). Overall, the most common TEAEs were cytokine release syndrome (CRS; 55.1%), pyrexia (43.3%), anemia (38.6%) and neutropenia (30.7%). Grade ≥3 TEAEs occurred in 107 (84.3%) patients, the most common being neutropenia (26.0%), anemia (22.8%), thrombocytopenia (15.0%), and coronavirus disease 2019 (COVID-19; 10.2%). Serious TEAEs occurred in 82 (64.6%) patients and were considered treatment related in 62 (48.8%) patients.

Table 3 Summary of AEs with odronextamab treatment
Full size table

A total of 17 (13.4%) patients had TEAEs that led to treatment discontinuation. Treatment-related TEAEs leading to discontinuation were encephalopathy (n = 2), CRS, COVID-19, CMV reactivation, pulmonary tuberculosis, aphasia, supraventricular tachycardia, and cholangitis sclerosing (n = 1 each); CRS, tachycardia, pancreatitis, septic shock, pneumonia plus cough in one patient; and P. jirovecii plus neutrophil count decrease in one patient. TEAEs leading to death were reported in 20 (15.7%) patients (Supplementary Table 3); these were considered treatment related in five (3.9%) patients (COVID-19, pneumonia, P. jirovecii pneumonia, pseudomonal sepsis (n = 1 each), and CMV pneumonia plus CMV reactivation in one patient).

With the step-up regimen of 0.7 to 4 to 20 mg, CRS was reported in 32/60 (53.3%) patients and mostly occurred during C1 (Extended Data Fig. 3). With this regimen, the majority of events were of low grade (grade 1, 40%; grade 2, 11.7%), with one grade 3 case occurring in the setting of acute pancreatitis (pancreatic lymphoma mass causing obstruction of biliary drainage) at week 6; this was numerically lower than the rate of grade 3 CRS with the original step-up regimen of 1 to 20 mg (n = 5 (7.5%)) (Supplementary Table 4). There were no cases of grade 4 or grade 5 CRS. CRS was managed with tocilizumab in 15 (25.0%) patients and systemic steroids in 13 (21.7%) patients (Supplementary Table 5); no patients required mechanical ventilation or intensive care unit admission for CRS. The median time to onset of CRS was 18.0 h (range: −3.4 to 221.0) and CRS events resolved in a median of 7.7 h (range: 0.1–143.9). Infusion-related reactions occurred in five (8.3%) patients (all grade 1 or 2). Two cases of encephalopathy that led to discontinuation of odronextamab treatment occurred in the setting of CRS during step-up dosing: one grade 3 event with the step-up regimen of 1 to 20 mg in a 79-year-old patient, and one grade 2 event with the step-up regimen of 0.7 to 4 to 20 mg in an 85-year-old patient; both events resolved with steroids.

Neurologic AEs of any grade occurred in 54 (42.5%) patients (grade ≥3, five (3.9%) patients), including in 22 (36.7%) patients with the step-up regimen of 0.7 to 4 to 20 mg (all grade 1 or 2). Neurologic AEs reported in >5% of patients were insomnia (n = 20 (15.7%)), dizziness (n = 12 (9.4%)) and headache (n = 8 (6.3%)). No cases of immune effector cell-associated neurotoxicity syndrome (ICANS; preferred term) were reported with either regimen. Tumor lysis syndrome occurred in one patient (grade 3–4) with the step-up regimen of 1 to 20 mg (Supplementary Table 4). There was one case of low-grade tumor flare.

Infections occurred in 82/127 (64.6%) patients (grade ≥3, 38.6%). The most frequent type of infection was COVID-19, which was reported in 18.1% (grade ≥3, 12.6%) of patients. Febrile neutropenia was also observed in three (2.4%) patients. Overall, six (4.7%) patients discontinued treatment because of treatment-related infections. Grade 5 infection occurred in 15 patients, with five cases because of COVID-19. Other grade 5 infections included pneumonia and sepsis (n = 3 each), P. jirovecii pneumonia, CMV infection, and pseudomonal sepsis (n = 1 each), and CMV infection reactivation plus CMV pneumonia in one patient.

Discussion

In this phase 2 study, odronextamab monotherapy demonstrated substantial efficacy in heavily pretreated patients with R/R DLBCL. These results are consistent with those from the ELM-1 study in patients with R/R DLBCL who had received prior CAR T cell therapy and no new safety signals were observed19,20, indicating that odronextamab may have an important role in maintaining effective disease control in this aggressive lymphoma.

Overall, the baseline characteristics of enrolled patients were representative of a heavily pretreated, highly refractory population. High-risk factors included double-hit and triple-hit cytogenetic rearrangements (9% of patients), transformed disease (24%), age ≥ 75 years (24%), Ann Arbor stage III–IV (81%) and prior ASCT (17%). Despite the difficult-to-treat nature of this population, odronextamab demonstrated consistent ORRs across high-risk subgroups.

T cell-engaging therapies including CAR T cell therapies are now established as an important treatment option for people with R/R DLBCL2. Despite encouraging ORRs (52–82%), eligibility for CAR T cell therapy is low (6–22%) and uptake among eligible people is variable because of access barriers (for example, administration complexities, manufacturing timelines and costs), associated toxicities (including CRS, ICANS and prolonged cytopenias) and potential risk of secondary malignancy5,6,7,21,22,23,24,25. These challenges may underpin the less frequent use of CAR T cell therapies reported in elderly people26. Potential earlier use of CAR T cell therapy also highlights the need for therapies that are active after these agents, where outcomes are typically dismal27. Bispecific antibodies have, thus, assumed greater importance as an alternative treatment option, particularly in the third line following the recent approvals of axicabtagene ciloleucel and lisocabtagene maraleucel for people with DLBCL who are R/R within 12 months of first-line immunochemotherapy8,10. Although cross-study comparisons of different single-arm studies are challenging, in the third-line setting, odronextamab demonstrated ORR and CR rates of 52% (95% CI: 42.9–60.9) and 31% (95% CI: 23.5–40.3), respectively, similar to those across the field of bispecific antibodies in R/R DLBCL (glofitamab, 52% (95% CI: 43–60) and 39% (95% CI: 32–48); epcoritamab, 63% (95% CI: 55.0–70.6) and 39% (95% CI: 31.2–46.9), respectively)28,29. Although this study did not include people treated with prior CAR T cell therapy, the efficacy of odronextamab in this population is supported by data from a prospective cohort of patients with disease progression after CAR T cell therapy in ELM-1 (n = 44), where the ORR was 48% (CR rate, 30%) and median DOR was NR after a median efficacy follow-up of 4.9 months. These data are consistent with those reported in CAR T cell therapy-naive patients in the current study and there were no major differences in safety profile20.

Biomarker assessment revealed MRD clearance in 20/63 patients at C4D15 of odronextamab treatment, with improved PFS in patients with cleared versus detectable MRD. The observed association between MRD clearance and PFS benefit, even in patients without positron emission tomography–computed tomography CR, indicates the prognostic utility of MRD measurement at this early time point and supports further investigation at later stages of treatment and potentially even after treatment.

The odronextamab administration schedule involved step-up dosing during C1, which was optimized during the study to help mitigate the risk of CRS. The step-up regimens used differed by just 1 week before reaching full dose and pharmacokinetic data indicated that exposure levels were similar for both regimens after the first full dose was received30. Following step-up, patients received weekly odronextamab in C2–C4, before dosing once every 2 weeks (once every 4 weeks with durable CR) as maintenance treatment until disease progression or unacceptable toxicity. This treatment regimen provided compelling antitumor control while maintaining or improving patient-reported outcomes over 42 weeks31. Alternative treatment paradigms, including fixed duration, have been explored in this setting28,29 but the optimal treatment approach for R/R DLBCL is yet to be determined. The potential for growth of subclonal cell populations supports treatment to progression32,33,34 in highly refractory people with aggressive lymphoma and appeared feasible. In addition, administration frequency could be reduced to once every 4 weeks in patients with a durable CR, enabling continued antitumor control in the context of reduced treatment burden.

Among the key AEs associated with bispecific antibodies, severe CRS risk was generally mitigated with the optimized C1 step-up regimen, with a numerically lower rate of grade 3 CRS compared with the original step-up regimen of 1 to 20 mg. One case of grade 3 CRS occurred with the revised regimen, although this was confounded by concurrent acute pancreatitis. Grade ≥3 CRS was reported in 4% (6/154) and 2.5% (4/157) of patients treated with glofitamab and epcoritamab, respectively28,29. Tocilizumab and corticosteroids for CRS were given to 25% and 22% of patients, respectively, in the current study according to evolving institutional guidelines35,36,37,38, although no patients required ventilatory or intensive care unit support. Given the low incidence of severe CRS, ongoing studies are evaluating odronextamab dosing in the outpatient setting, an important consideration for promoting equitable access to effective treatment options in underserved communities. No ICANS was reported with odronextamab in contrast to glofitamab (8%) and epcoritamab (6%)28,29. Neurologic AEs occurred in 43% of patients treated with odronextamab, similar to rates reported with glofitamab in R/R B-NHL (40%) and epcoritamab in R/R DLBCL (35%)39,40. Neurologic AEs were mostly grade ≤2 with odronextamab and the events observed were generally consistent with those reported with other bispecific antibodies16,40. Two cases of grade 2–3 encephalopathy that led to treatment discontinuation were reported in elderly patients; however, both events occurred in the setting of CRS during step-up dosing and resolved with steroids.

Infections were observed in 65% of patients, which may be common in a population with impaired B cell functionality because of underlying malignancy, prior exposure to immunosuppressive agents and chemotherapy, and anticipated B cell depletion and hypogammaglobulinemia induced by odronextamab41,42. Anti-infection prophylaxis was added to the protocol during the study, although local practices for infection management and IV immunoglobulin supplementation may have differed between global sites. COVID-19 was the most frequent infection reported and the most frequent grade 5 infection, reflecting the course of ELM-2 enrollment. Enrollment began early in the pandemic when viral severity and mortality were high and no vaccines or anti-COVID-19 treatments were available. Later enrollment occurred when more transmissible variants were prevalent, vaccine availability had improved and social-distancing measures were relaxing. Randomized controlled trials are required to further investigate the risk of infections with odronextamab and better characterize the kinetics of B cell depletion following fixed durations of treatment.

TEAEs resulting in death were reported in 15.7% of patients treated with odronextamab, of which five (3.9%) were because of COVID-19 infection. With glofitamab and epcoritamab, TEAEs leading to death were reported in approximately 5% of patients, which were related to COVID-19 in five (3.2%) and two (1.3%) patients, respectively28,29. Most fatal TEAEs were caused by infections. The differences in fatal infection rates may be attributed to variations in populations, regional infection rates and local supportive care practices. In addition, the timing of enrollment during different phases of the COVID-19 pandemic, including the availability of treatments and vaccines, may have influenced outcomes.

In conclusion, odronextamab demonstrated highly encouraging clinical activity, including durable CRs, in heavily pretreated, highly refractory patients with R/R DLBCL. AEs were experienced by nearly all patients treated with odronextamab. However, these events were generally manageable with supportive care measures. Odronextamab is a potential treatment option for people with highly refractory R/R DLBCL. Phase 3 trials are currently enrolling in earlier lines of therapy and will inform the future management paradigm for aggressive lymphomas.

Methods

Study design and patients

ELM-2 is an ongoing phase 2, open-label, multicohort, multicenter, single-arm study of odronextamab monotherapy in R/R B-NHL (ClinicalTrials.gov identifier: NCT03888105). Here, we report long-term follow-up results of the primary analysis in the cohort of patients with R/R DLBCL. Patients were recruited from various centers across multiple countries, including the USA, Australia, Canada, China, France, Germany, Italy, Japan, the Republic of Korea, Poland, Singapore, Spain, Taiwan and the United Kingdom. Methods for the follicular lymphoma cohort of ELM-2, which used the same endpoints as the DLBCL cohort, have been published43. Eligible patients for the DLBCL cohort were aged ≥18 years with DLBCL (de novo or transformed) refractory to or relapsed after two or more prior lines of systemic therapy, including an anti-CD20 antibody and an alkylator. Other inclusion criteria were measurable disease on cross-sectional imaging, Eastern Cooperative Oncology Group (ECOG) performance status 0–1, and adequate bone marrow and hepatic functions. Patients with high-grade lymphoma (double-hit and triple-hit cytogenetic rearrangements) were accepted. People with primary central nervous system lymphoma or prior ASCT, CAR T cell therapy or CD20×CD3 bispecific antibody treatment were excluded.

Measures to ensure diverse and inclusive enrollment included diverse trial sites, translated consent forms for under-represented populations, extended screening windows for patients with access constraints, broad eligibility criteria to include patients with controlled human immunodeficiency virus, hepatitis B and hepatitis C infection, and lower thresholds for those with compromised organ function because of lymphoma.

Prophylaxis for P. jirovecii pneumonia was recommended for all patients. Other anti-infection prophylaxis measures included IV immunoglobulin supplementation and antivirals, in accordance with the protocol and local institutional standard, as well as the National Comprehensive Cancer Network44, American Society of Clinical Oncology45 or European Society for Medical Oncology guidelines46. In patients with severe hypogammaglobulinemia (<400 mg dl−1) or in patients with recurrent episodes of infection with immunoglobulin levels between 400 and 600 mg dl−1, supplementation with IV immunoglobulin was recommended. For patients with positive hepatitis B surface antigens, hepatitis B core antibodies and/or measurable viral load, an appropriate antiviral agent for hepatitis B virus was recommended. Appropriate antiviral prophylaxis was recommended for patients with prior herpes simplex virus or CMV infection.

Patients received IV odronextamab in 21-day cycles. The original step-up regimen (1 to 20 mg) during C1 comprised a dose of 1 mg split over day 1 (0.5 mg) and day 2 (0.5 mg) and 20 mg split over day 8 (10 mg) and day 9 (10 mg), followed by the full dose of 160 mg on day 15. The step-up dosing regimen was optimized during the study to further mitigate the risk of CRS by reducing the initial dose and adding an intermediary dose. The revised step-up regimen of 0.7 to 4 to 20 mg regimen consisted of 0.7 mg split over day 1 (0.2 mg) and day 2 (0.5 mg), 4 mg split over day 8 and day 9 and 20 mg split over day 15 and day 16 of C1. Following C1 step-up dosing, patients received odronextamab 160 mg on days 1, 8 and 15 of C2–C4 and then 320 mg once every 2 weeks as maintenance until disease progression or another protocol-defined reason for treatment discontinuation. Patients were admitted for inpatient monitoring for 24 h following each infusion up to and including C2D1.

In patients who had a CR that lasted for 9 months or longer by investigator evaluation, the frequency of dosing was reduced to 320 mg once every 4 weeks.

Premedication with dexamethasone, diphenhydramine and acetaminophen was given during C1 step-up dosing to help mitigate the risk of CRS. All patients received 20 mg of IV dexamethasone 1–3 h before each split or single infusion dose for both regimens. Patients on the step-up regimen of 0.7 to 4 to 20 mg also received 10 mg of oral dexamethasone 12–24 h before the first split infusion and 25 mg of IV diphenhydramine and 650 mg of oral acetaminophen 30–60 min before each split or single infusion. Patients then received 10 mg of oral dexamethasone 24 h after the second split infusion or first single infusion. Premedication was continued until the patient received the full weekly dose without experiencing infusion-related reactions or CRS. Patients who developed symptoms consistent with severe CRS were considered for treatment with tocilizumab, corticosteroid and other interventions according to the clinical judgment of the investigator.

The protocol and amendments were approved by the relevant institutional review boards and ethics committees (Supplementary Table 6). The study protocol is included in the Supplementary Information. The study was conducted according to applicable regulatory requirements, guidelines of Good Clinical Practice as specified by the International Conference on Harmonization and principles originating from the Declaration of Helsinki. All patients provided written informed consent before enrollment. Where possible, the present report was developed in accordance with CONSORT reporting guidelines47. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Endpoints and assessments

The primary endpoint was ORR, assessed by ICR and in accordance with Lugano criteria48. Secondary endpoints included ORR assessed by local investigator, CR rate, DOR, PFS, OS and patient-reported quality-of-life outcomes.

Disease assessments using computed tomography/magnetic resonance imaging and positron emission tomography were performed during screening, at week 12 and then every 8 weeks in year 1, every 12 weeks in year 2 and during follow-up as described in the protocol.

Exploratory endpoints included changes in select cytokine levels and MRD status using ctDNA, with samples taken at baseline, at week 12 and at every radiologic response assessment in patients with CR. A modified AVENIO ctDNA analysis workflow (Roche; research only) was used for next-generation sequencing according to the cancer personalized profiling obtained by deep sequencing49. Whole-blood-cell pellets were used to filter out germline allele variants and MRD negativity was reported when the P value for allele frequency was greater than 0.005 (ref. 50). The study was not powered for statistical testing of MRD analyses given their exploratory nature.

Safety and tolerability were assessed until 90 days after the last dose of study drug or initiation of another antilymphoma therapy, with AEs graded according to the National Cancer Institute Common Terminology Criteria for AEs (version 5). CRS grading was adapted from American Society for Transplantation and Cellular Therapy guidelines51. TEAE and treatment-related TEAE data are presented. TEAEs were defined as AEs that newly occurred or worsened during the on-treatment period and any treatment-related serious AEs that occurred during the post-treatment period. TEAEs were deemed treatment related by the investigator.

Statistics and reproducibility

In the ELM-2 study of B-NHL, approximately 512 patients were planned for enrollment into five disease-specific cohorts (DLBCL, follicular lymphoma, marginal zone lymphoma, mantle cell lymphoma and ‘other B-NHL’). This report included all patients in the DLBCL global cohort (160 mg once weekly or 320 mg once every 2 weeks), with no data excluded. Data distribution was assumed to be normal, but not formally tested.

An exact binomial design was adopted for the primary endpoint of ORR. The two-sided 95% CIs for the observed ORR were calculated on the basis of a sample size of 112. Assuming a clinically meaningful ORR as being greater than 35%, with 112 patients, an ORR of at least 45% would have a lower CI bound that excludes 35%. In addition, if the observed ORR was at least 50%, 55% or 60%, the lower bound of the 95% CI would exclude an ORR of 40%, 45% and 50%, respectively. With a sample size of 112 patients, if the true treatment effect of odronextamab was 50%, the probability of the observed lower bound of 95% CI excluding 35% was 89%. Enrollment was increased to include at least 60 patients treated with a step-up regimen of 0.7 to 4 to 20 mg, and up to 127 patients with 160 mg of weekly dosing. The step-up dosing regimen allocation for patients was nonrandomized and unblinded.

Patients NE for best overall response were considered nonresponders. The primary analysis for the primary endpoint was performed after all patients had completed 36 weeks of tumor assessment or withdrawn from study. Efficacy and safety analyses were performed in all patients who received odronextamab. DOR, PFS and OS were analyzed using Kaplan–Meier estimation. Data collection and analysis were not performed blind to the conditions in the experiments.

All analyses were performed using SAS (SAS Institute) version 9.4 or above. The statistical analysis plan is included in the Supplementary Information.

Subgroup analysis

Patient demographics were summarized for the DLBCL cohort, including age (<65 years, 65–75 years, and ≥65 years), sex (male or female; self-reported by patients), race (White, Black or African American, American Indian or Alaska native, native Hawaiian or other Pacific Islander, not reported, unknown, or other) and ethnicity (Hispanic or Latino, or not Hispanic or Latino).

ORR per ICR was analyzed in subgroups of patients with DLBCL defined by baseline characteristics, including age, cell of origin (germinal center B cell-like (GCB) DLBCL, activated B cell-like (ABC) DLBCL/non-GCB, or unclassified DLBCL) and cytogenetic status (triple hit or double hit). If a subgroup included fewer than ten patients, the analysis for the given subgroup was not performed or combined with another subgroup.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.