When CAR-T expansion becomes a toxicity signal: cilta-cel, flow counts and ALC

A 90-patient CAR-T flow cohort and 532-patient ALC analysis: what expansion metrics can and cannot predict.

Primary source: Hitomi Hosoya, Arash Velayati, Danai Dima et al. Rapid Peak Cilta-cel Expansion is Associated with Delayed Neurotoxicity in Multiple Myeloma. Blood, online 8 July 2026. DOI 10.1182/blood.2025032844. PubMed journal abstract.

What the investigators measured

A common flow-cytometry platform quantified CAR-T kinetics in 90 patients with relapsed or refractory myeloma: 54 receiving cilta-cel and 36 receiving ide-cel. Median peak CAR-T counts differed between products (106 vs 49 cells/µL). In the cilta-cel population, higher peak expansion was associated with delayed neurotoxicity (reported medians 1,009 vs 96 cells/µL in patients with vs without delayed neurotoxicity). In ide-cel, peak expansion tracked response, a relationship not observed in the same way for cilta-cel. These are distinct outcome associations, not a randomised head-to-head comparison.

The second analysis used a larger, multicentre cohort of 532 recipients (256 cilta-cel and 276 ide-cel). Absolute lymphocyte count (ALC) served as an accessible but non-specific surrogate. The reported risk criterion—peak ALC ≥3,000/µL or ≥2,500/µL following an increase of twofold or more—had sensitivity 81% and specificity 59% for delayed neurotoxicity. It is a signal worth investigating, not a validated intervention threshold.

Original CFCM comparison: 90-patient flow cohort median cilta-cel versus ide-cel peak CAR T cells 106 versus 49 per microliter; cilta-cel neurotoxicity comparison 1009 versus 96; multicentre ALC criterion sensitivity 81 percent specificity 59 percent. — **Figure 1 (CFCM original).** Three summaries of the published cohorts; these are not patient events or figures reproduced from Hosoya et al. Values are reported cohort statistics with different denominators.

Clinical flow interpretation

Quantitative CAR-T monitoring needs a defined analyte. The laboratory should specify the CAR-detection reagent, flow gating hierarchy, event thresholds, absolute counting method, acquisition timing relative to infusion, and how low-level counts are reported. Absolute CAR-T cells per µL and CAR-positive percentage of lymphocytes are not interchangeable. In longitudinal measurements, differences in specimen handling, leukocyte recovery, reference populations and analytical sensitivity can masquerade as biological changes.

Evidence boundary

This study is observational. ALC captures non-CAR lymphocytes and may change during infection, inflammation or immune recovery. Its predictive performance may differ across populations and clinical workflows. A related 2026 observational analysis did not demonstrate prevention of delayed-neurotoxicity events from prophylactic dexamethasone at high ALC (Blood Cancer Journal DOI). A risk marker is not proof of effective intervention.

Daily Learning Pill — a surrogate is not the analyte

A validated flow assay measures CAR-expressing T cells more specifically than ALC. ALC is not. The diagnostic properties of a surrogate should be judged with sensitivity, specificity, prevalence, and prospective clinical utility—not merely correlation with a more specific measurement. Academy reuse: CAR-T monitoring / absolute enumeration / surrogate biomarker validation.