In this issue of Blood Advances, van Besien et al1 present a comprehensive systematic review and meta-analysis that compared the infection outcomes of patients with non-Hodgkin lymphoma (NHL) who were treated with CD3×CD20-bispecific antibodies (BsAb) with those of patients treated with CD19-directed chimeric antigen receptor T-cell (CAR-T) therapy by analyzing 3202 patients treated on trial. Their work responds to a key criticism in the field, namely that infection outcomes are often reported without adequate adjustment for follow-up duration.2 Furthermore, their comparative analysis yields several pertinent findings that raise important considerations for future evaluations of infections in cellular therapies.
van Besien et al demonstrated that, when controlling for duration of follow-up, both all-grade and grade ≥3 infections are more common among patients treated with BsAb than among those who received CAR-T therapy. In the main analysis, BsAb were assessed as a single group encompassing both fixed-duration regimens (eg, glofitamab) and indefinite-duration regimens (eg, epcoritimab and odronextamab).1 However, supplementary analyses offered additional insights into the impact of treatment duration on infection risk. Although the all-grade infection rates were similar between the fixed- and indefinite-duration regimens, the incidence of grade ≥3 infections was higher with indefinite-duration BsAb (29%; 95% confidence interval [CI], 21-38) than with fixed-duration therapies (16%; 95% CI, 14-20). The overall incidence of infection was also significantly greater with indefinite-duration therapies (0.0235 vs 0.0117; P = .045). These findings highlight the importance of further exploring how treatment modifications, such as augmented dosing schedules, planned interruptions for infection management, or cessation of therapy in patients who achieved complete metabolic response, might reduce the infection risk without compromising disease control.
A further important factor that needs to be considered when interpreting the heightened BsAb infection signal, particularly among those on indefinite-therapy regimens, is the need to define time-dependent periods of risk. A subanalysis of ELM-2, currently published in abstract form, evaluated fatal infections among patients who achieved complete remission following odronextamab.3 Fatal infections are more common earlier in treatment, particularly within the first 12 months, than at later time points despite continued BsAb exposure.3 Similar findings are reported in EPCORE NHL-1 trial with epcoritimab in which the rate of grade 3 or 4 infections (excluding COVID-19) was highest during the first 12 weeks (11%) and lower thereafter (2%-7%).4 These emerging patterns, however, are difficult to resolve through meta-analyses, including those that incorporate follow-up duration, because the onset of infections is not consistently reported across trials.
Although beyond the scope of a systematic review and meta-analysis, the approach of comparing BsAb and CAR-T treatment among patients with NHL has also been applied in real-world cohort studies to better delineate the infection risks. For example, a recent cohort study that compared B-cell maturation antigen–directed CAR-T and BsAb therapies in multiple myeloma found that treatment-emergent hypogammaglobulinemia disproportionately affected BsAb patients, correlating with both higher infection rates and increased severity.5 Extending the comparative framework used by van Besien et al, future analyses in lymphoma should explore infection risk factors, such as treatment-emergent neutropenia, cumulative corticosteroid exposure, and immune effector cell–associated hematotoxicity to help define periods of risk and inform treatment decisions.
Unfortunately, this analysis does not provide a comparative view of the types of infections seen with BsAb vs CAR-T. van Besien et al highlight a key theme in their discussion, namely that incomplete infection reporting precludes the meaningful characterization of infection types across trials. This limitation also lies outside the scope of a meta-analysis that is aimed at standardizing infection incidence. Nonetheless, the persistent infection signal in patients treated with cellular therapies underscores a broader gap in long-term infection data for this population. Registry cohort studies of CAR-T recipients have shown that bacterial infections decline after neutrophil recovery, whereas low-grade respiratory viruses become more prominent with time.6,7 Small but persistent risks for fungal and severe viral infections remain,6 however, infections such as cytomegalovirus and human herpesvirus 6B may not have been systematically sought or detected in these pivotal trials. Meanwhile, Clostridioides difficile has emerged as a relevant early bacterial complication as supportive strategies evolve.8 Comparable longitudinal data are not yet available for BsAb-treated patients. In the 2-year safety update for epcoritimab, 25% of patients experienced a grade 3 or 4 infection but limited details regarding the causative pathogens were provided.4 Similarly, odronextamab’s 2-year follow-up reported grade 3 infections in 38.6% of patients with COVID-19 listed as the most common cause, but no further information was provided.9 Prospective postpandemic studies are urgently needed to describe late infectious events and their clinical significance, particularly across BsAb therapies.
Turning to fatal infections, van Besien et al observed a higher proportion of fatal infections among patients who received CD28-based CAR-T therapies (axicabtagene and brexucabtagene) and those who received BsAb, although this difference was not statistically significant when expressed as incidence rates. Two recent meta-analyses of nonrelapse mortality (NRM) further underscore the complexity. In a pooled analysis of 7604 CAR-T recipients, axi-cel had a higher NRM than 4-1BB products with infections comprising >50% of the NRM cases, yet, no significant difference was found in infection-attributable NRM across the products (P = .28).10 In contrast, a second meta-analysis of 2535 patients (1479 with NHL) who were treated with BsAb reported that infections accounted for 70% of NRM cases.11 After excluding COVID-19 deaths, BsAb had a significantly lower NRM than CAR-T (3.1% vs 5.8%; P = .005).11 These somewhat discordant findings highlight the challenges in attributing infection-related mortality and reinforce the need to account for patient-level factors, disease characteristics, and timing of therapy when comparing immunotherapeutic platforms.
The systematic review and meta-analysis by van Besien et al represents a thoughtful and methodologically rigorous contribution to our understanding of infection risk across novel immunotherapies for NHL. By addressing key limitations in existing literature and enabling cross-platform comparisons between BsAb and CAR-T therapies, their work provides an important foundation for future inquiry. These findings underscore the need for harmonized infection reporting, time-aware risk models, and integrated analyses that consider treatment exposures, patient-level variables, and immune vulnerability. Although meta-analyses of trials with limited infection reporting provide important insights, they can only take us so far. We urgently call for contemporary, well-designed registries or cohort studies that more precisely characterize infection in cellular therapy recipients, to lay the groundwork for patient-centered risk stratification to inform trial design, prophylaxis strategies, and therapeutic decision-making.
Conflict-of-interest disclosure: G.K.R. reports a PhD scholarship from National Health and Medical Research Council (NHMRC; 2013970). M.A.S. reports research grants unrelated to the work from NHMRC, Merck, and F2G; consultation fees from Pfizer, Takeda, Gilead, and Merck; honoraria from Pfizer and F2G; serving on advisory boards for Roche and Pfizer; and serving in a leadership role for the Australian Society of Infectious Diseases Immunocompromised Host Special Interest Group.
