Just out todayDigital health tech: Ninety-nine nurses, seventy wearable features, and no control groupDigital health tech: Fifty-six trials, 6,154 patients, and headsets beat the threshold patients would noticeSynthetic biology & cell therapy: Roughly one in four patients on these therapies gets a severe infection

Synthetic biology & cell therapy/Italy, United Kingdom, Germany

Roughly one in four patients on these therapies gets a severe infection

Pooled across 2,097 patients, 24.2% on bispecific antibodies or CAR T-cells developed grade 3 to 4 infections.

What the paper found

An Italian, British and German team pooled 16 real-world retrospective studies covering 2,097 patients with relapsed refractory multiple myeloma treated with approved bispecific antibodies or CAR T-cells. Overall 24.2% developed grade 3 to 4 infections: 26% on bispecifics (27% BCMA-directed, 25% GPRC5D-directed) and 19% on CAR T-cells.

Grade 12 reading level6 min read

What happened

A team at the University of Bari Aldo Moro, the University of Hull and a German university hospital published in Experimental Hematology and Oncology a meta-analysis of infection risk in relapsed refractory multiple myeloma treated with the two newest therapy classes.

They pooled 16 real-world retrospective studies covering 2,097 patients treated with approved bispecific antibodies (BiTEs) or CAR T-cell therapies.

Overall, 24.2% developed grade 3 to 4 infections, a pooled event rate of 0.24 with a confidence interval from 0.21 to 0.28.

Split by class: bispecific antibodies, 1,602 patients, pooled rate 0.26; CAR T-cells, 495 patients, pooled rate 0.19.

Within bispecifics, BCMA-directed agents showed 0.27 and GPRC5D-directed 0.25.

The test

The choice to pool real-world retrospective studies rather than trials is deliberate and defensible. Trial populations in myeloma are fitter than the patients who receive these drugs in practice, and trial reporting of infection is inconsistent in both grading and ascertainment. Real-world series capture the patients who actually get treated.

The cost of that choice is that the two therapy classes are compared observationally. Nobody is randomised to a bispecific antibody versus CAR T-cell therapy. The two routes select different patients: CAR T-cell therapy requires a patient fit enough to survive apheresis, bridging therapy and a manufacturing wait, which selects for better performance status.

So the 0.26 against 0.19 difference is confounded by design, and the confounding runs in a predictable direction.

Infections were graded as grade 3 to 4, meaning severe or life-threatening.

The numbers

Group Patients Pooled grade 3-4 infection rate
All patients 2,097 0.24 (95% CI 0.21 to 0.28), 24.2%
Bispecific antibodies 1,602 0.26 (95% CI 0.23 to 0.30)
BCMA-directed bispecifics subset 0.27
GPRC5D-directed bispecifics subset 0.25
CAR T-cell therapies 495 0.19 (95% CI 0.12 to 0.27)
Studies pooled 16, real-world retrospective
Data statement no datasets were generated or analysed

The result

Roughly one patient in four on these therapies develops a severe or life-threatening infection. That is the number to carry, and it is a substantial toxicity burden for therapies often discussed primarily in terms of response rates.

The between-class difference, 0.26 against 0.19, has overlapping confidence intervals: the bispecific interval runs 0.23 to 0.30 and the CAR T-cell interval 0.12 to 0.27. The CAR T-cell interval is wide because there are only 495 patients, and it includes most of the bispecific range.

So the honest statement is that the point estimates differ in the direction patient selection would predict, and the data do not establish that the therapies themselves differ.

The subgroup difference between BCMA-directed and GPRC5D-directed bispecifics, 0.27 against 0.25, is small enough to be noise and should not drive a choice.

How this compares to what came before

This batch contains the complementary paper on the same therapy class. A multicentre eligibility study screened 1,844 lupus patients across seven certified centres and found 54 (2.9%) met severe disease criteria while three (0.16%) were treated with CAR T-cells.

Put the two together. In autoimmune disease, roughly one screened patient in six hundred currently reaches CAR T-cell therapy. In myeloma, roughly one in four who reach these therapies develops a severe infection.

Neither of those numbers appears in the remission reports that drive enthusiasm for cell therapy, and both belong in any serious account of it.

Among this batch’s five evidence syntheses, this one is methodologically mid-range: it pools 16 studies with 2,097 patients and reports confidence intervals and subgroups, and does not report a GRADE certainty rating. Two others in the batch, a network meta-analysis of perioperative digital health and a Bayesian network meta-analysis of telerehabilitation, applied GRADE throughout.

What it means

For haematologists choosing between these classes in a patient at high infectious risk, the authors’ own framing is reasonable: CAR T-cell therapy, or GPRC5D-directed bispecifics when CAR T-cell therapy is not feasible, within an individualised strategy. That is appropriately hedged given the confounding.

For service planning, 24.2% severe infection is the figure to resource against: antimicrobial prophylaxis, immunoglobulin replacement, monitoring and admission capacity.

For the field, the useful methodological point is that infection reporting in trials of these agents has been inconsistent enough that a real-world meta-analysis is the better source. That is an indictment of trial reporting standards rather than a virtue of retrospective data.

Business ideas from this paper

Infection surveillance and prophylaxis pathways specific to T-cell-engaging therapies, where a 24.2% severe infection rate is the business case.

Immunoglobulin replacement management, since B-cell-directed therapies cause hypogammaglobulinaemia and the monitoring is currently inconsistent.

Standardised toxicity reporting registries for cell and T-cell-engaging therapies, which would remove the need for retrospective meta-analysis to answer a basic safety question.

What could be wrong with it

Patients are not randomised between the two classes, and those reaching CAR T-cell therapy are selected for fitness in ways that independently lower infection risk.

Pooling 16 retrospective studies inherits all their differences in ascertainment, follow-up duration and infection grading.

The CAR T-cell confidence interval, 0.12 to 0.27, overlaps most of the bispecific range, so the between-class difference is not established.

The BCMA versus GPRC5D difference of 0.27 against 0.25 is within noise.

No GRADE certainty rating is reported.

The record states that no datasets were generated or analysed during the study, so the pooled data cannot be re-examined.

How sure can you be?

Confident that the pooled grade 3 to 4 infection rate across 2,097 patients was 24.2%, with a reasonably tight interval of 0.21 to 0.28.

Confident that the rate among the 1,602 bispecific-treated patients was 0.26.

Confident that the point estimate for CAR T-cell therapy was lower at 0.19.

Less confident that the therapies differ, because the intervals overlap substantially and patient selection confounds the comparison in the same direction as the observed difference.

Less confident about comparability across the 16 pooled studies, which differ in how infections were ascertained and graded.

Not confident about the subgroup difference between bispecific targets.

Our rubric places this in the upper part of the range for this batch: peer-reviewed, three institutions across three countries, a systematic review and meta-analysis of real-world data, with no certainty rating and no public data.

Do this today

Carry the 24.2% figure into consent conversations. Severe infection in roughly one patient in four is a material part of the risk profile of these therapies.

When comparing two therapies observationally, ask which one selects fitter patients. Here it is CAR T-cell therapy, and the difference in infection rate points the same way.

If you run a cell therapy service, resource the infection pathway against one in four rather than against trial-reported rates.

Source: Real-world evidence on infection risk in multiple myeloma treated with BiTEs and CAR-T cells: a meta-analysis, Experimental Hematology & Oncology, 2026-06-20. PMID 42323640 · doi.org ().

Just Out Tech explains new research in plain language. This article was drafted with AI assistance and checked by a human against the original source.

What to remember
  • Systematic review and meta-analysis of 16 real-world retrospective studies, 2,097 patients with relapsed refractory multiple myeloma.
  • Pooled grade 3 to 4 infection rate overall: 24.2% (pooled event rate 0.24, 95% CI 0.21 to 0.28).
  • Bispecific antibodies (n = 1,602): pooled rate 0.26 (95% CI 0.23 to 0.30), with BCMA-directed at 0.27 and GPRC5D-directed at 0.25.
  • CAR T-cell therapies (n = 495): pooled rate 0.19 (95% CI 0.12 to 0.27).
Where this really is

Systematic review and meta-analysis of real-world retrospective studies; observational comparison between non-randomised therapy groups.

Questions people ask

Why does the comparison need caution?

Because patients are not randomised to these therapies. Those who reach CAR T-cell therapy are selected for fitness, which lowers their infection risk independently of the treatment.

Is 24.2% high?

For a grade 3 to 4 infection rate it is substantial. Grade 3 to 4 means severe or life-threatening, requiring intervention.

Does the BCMA versus GPRC5D difference matter?

The difference is 0.27 against 0.25, which is small enough that it should not drive a treatment choice on its own.

Why real-world studies rather than trials?

Because trial reporting of infection is inconsistent and trial populations are healthier. Real-world retrospective data capture the patients who actually receive these therapies.

About the author

Mark Alex

Mark Alex is the founder and Managing Director of Real Biz Digital, a technology company operating out of Nairobi since 2018. He works in agentic AI and the Model Context Protocol, AI governance, enterprise software architecture and cybersecurity. He holds an MSc in Mechatronical Engineering from Obuda University in Budapest and a BSc in IT, Forensic Technology and Cybercrime, from USIU-Africa in Nairobi, and has published IEEE conference research on an AI-powered digital twin for greenhouse systems. He is the author of seven books. Between 2020 and 2024 he mentored more than 200 university students and interns in Nairobi. He writes every Just Out Tech article from the original research paper.