A monthly roundup of key articles in other journals

AI-enabled forecasting of prehospital transfusion needs in patients with trauma: a multinational, registry-based, retrospective, machine learning development and validation study. Sigle M, Boos M, Weiss T, et al. Lancet Digit Health. 2026; 8(1): 100945. Machine-learning models can personalize clinical decision-making by determining early need for transfusion in patients with trauma. Uncontrolled hemorrhage is a leading cause of preventable death in patients with trauma. Prehospital transfusion with blood products is associated with reduced mortality, but past trials have reported conflicting outcomes, and there is thus a lack of consensus around prehospital transfusion triggers. Sigle et al. developed an algorithm to identify patients who would benefit from early, prehospital transfusion. The authors leveraged multinational trauma registries to develop a machine learning model for predicting need for prehospital transfusion. All available patient data obtained from the American National Trauma Data Bank (N = 364,350) that was considered potentially indicative of transfusion requirement (injury pattern, vital signs, etc.) was incorporated into models, which were trained to predict transfusion prior to internal validation (N = 91,087) and external validation on data from three additional registries (TraumaRegister DGU, National Office of Clinical Audit–Major Trauma Audit, and Alberta Trauma Registry; N = 54,210). The new injury severity score (NISS)—which considers the three most severe injury types, irrespective of body region—exerted the strongest influence in predicting need for transfusion, with injuries to the abdomen having higher predictive importance than other body regions. In internal validation, there was a near-perfect relationship between patients who were predicted to need a transfusion and those who received transfusion. While in external validation the model showed a tendency to overpredict the need for transfusion, performance remained strong. Furthermore, replacing the NISS with an alternative injury severity score (a four-point scale: none, minor, moderate, and severe across nine body regions) did not significantly impact performance. A second two-stage model developed to address transfusion volume showed more modest performance, with the NISS again exerting notable influence. Patients assigned the highest probability of needing transfusion showed the highest cumulative incidence of overall mortality over 10 weeks of follow-up as well as the highest rate of death from hemorrhage at 72 h. An evaluation of heterogeneity of transfusion prediction revealed that a positive predictive value (PPV) of 0.93 (95% CI 0.93–0.94) resulted in good precision and capacity to identify patients for whom early transfusion would improve outcomes. The largest differences in PPV were observed in penetrating injuries and injuries of the abdomen or pelvis. A detailed analysis of misclassification found that lacerations and fractures to the face were associated with underprediction whereas superficial injuries to lower extremities were associated with overprediction. While approaches such as those described here have the potential to personalize trauma care and improve patients' outcomes, prospective validation is required prior to clinical implementation.♦♦♦ CD4+ T cells mediate CAR-T cell-associated immune-related adverse events after BCMA CAR-T cell therapy. Ho M, Paruzzo L, Noll JH, et al. Nat Med. 2026; 32(2): 702–716. CD4+ T-cell mediated attack on normal tissues is the primary driver of chimeric antigen receptor (CAR) T cell-associated immune-related adverse events (CirAEs). B-cell maturation antigen (BCMA)-targeting CAR T cells—idecabtagene vicleucel (ide-cel) and ciltacabtagene autoleucel (cilta-cel)—have revolutionized care for patients with relapsed/refractory multiple myeloma, but concern around longer-term immune-related toxicity has arisen. CirAEs are highly heterogeneous and have not been well characterized, but the increasing use of BCMA CAR T-cells underscores the need for clarification of the risks and benefits of treatment. Here Ho et al. report the incidence, spectrum, risk factors and outcomes associated with CirAEs in a large cohort of patients treated with BCMA-targeting CAR T-cell therapy. The retrospective analysis identified 198 patients treated with BCMA targeting CAR T-cells (cilta-cel n = 125; ide-cel n = 73; median follow-up = 20.5 months; overall response rate = 85.4%; median progression-free survival = 21.1 months; 12-month overall survival = 83.3%). Cytokine response syndrome (CRS) and immune effector cell (IEC)-associated neurotoxicity syndrome (ICANS) were reported in 82.2% and 12.6% of the full cohort, respectively. Additionally, 20% of the patients treated with cilta-cel and 2.7% of those treated with idel-cel developed CirAEs (median delay to onset = 28 days). CirAEs were heterogeneous in their clinical presentation and could be sorted into four subtypes: rheumatologic (n = 2), neurologic (n = 19), gastrointestinal (n = 7), or pulmonary (n = 2). In patients treated with cilta-cel, CirAEs included cranial nerve palsies, enterocolitis, parkinsonism, arthritis, delayed ICANS, encephalopathy, and peripheral neuropathy. In contrast, the only reported CirAE in patients treated with idel-cel was pneumonitis. Notably, 1-year non-relapse mortality was 17% in patients who developed CirAEs, compared with 3% in patients without CirAEs. Incidence of CirAEs was strongly associated with ICANS, median Day 14 ALCPeak (highest lymphocyte count within 14 days of infusion), apheresis CD4:CD8 ratio >1, and higher circulating Day-14 CAR T-cell counts, but not pre-infusion myeloma burden or CRS. Infiltrating CAR T-cells detected in samples from CirAEs showed CD4+ skewing; and sera from 31 patients treated with cilta-cel who later developed CirAEs showed higher prelymphodepletion levels of EGF, CCL4 and CCL5. One heavily treated 74-year-old patient developed pronounced CD4+-skewed hyperleukocytosis at Day 13 concurrent with three CirAEs (facial palsy, delayed ICANS, and IEC-enterocolitis). Proteomic analysis of Day 7 serum sample showed elevated levels of lymphopoietic cytokines and chemokines. High levels of IL-15 preceded peak CAR T-cell expansion, and a sample of this patient's CAR T-cells showed elevated sensitivity to IL-15 in culture, resulting in CCR5 upregulation and heightened CCL5 secretion. Maraviroc—a CCR5 receptor antagonist—mitigated the IL-15-stimulated hypoproliferation without disrupting cytotoxicity. Together, the findings suggest that an early increase in cytokines may drive excessive CD4+ CAR T-cell expansion and subsequent CirAEs, paving the way for strategies to mitigate risks associated with BCMA-targeting CAR T cells.♦♦♦ Infant immunity after maternal nipocalimab in severe hemolytic disease of the fetus and newborn. de Winter DP, Moise KJ Jr, Ling LE, et al. NEJM Evid. 2026; 5(2): EVIDoa2500097. Following maternal administration of nipocalimab, infant serum concentrations fall below pharmacologically relevant levels by birth or soon thereafter, with negligible impact on postnatal immunity. Nipocalimab is a fully humanized antibody currently under investigation for preventing hemolytic disease of the fetus and newborn (HDFN) which acts by blocking the interaction between IgG and FcRn, preventing alloantibody transport across the placenta. However, blocking this interaction may also mitigate the transfer of beneficial IgG and, if transferred across the placenta, impact FcRn-mediated IgG recycling, thus impairing passive immunity. In this report, De Winter et al. explore the in vivo pharmacokinetics of nipocalimab and its impact on infant immunity across the first 96 weeks, postnatal. UNITY was a multicenter, open-label, single-group, Phase 2 study exploring safety, efficacy, pharmacokinetics, and pharmacodynamics of nipocalimab in pregnancies at high risk for anti-D or anti-K-mediated severe HDFN. Nipocalimab was administered weekly between gestational age 14–35 weeks. Serum nipocalimab levels were assessed in cordocentesis samples prior to intrauterine transfusion, cord blood at birth, venous samples taken 4 weeks postnatally, colostrum (Day 0–3 postnatal), and breast milk (Day 5–8). In total, 13 women delivered 12 live born infants (median gestational age = 36 weeks and 5 days; birthweight = 2825 g), one of whom was lost to follow-up after the 24-week appointment. Cordocentesis samples from the first intrauterine transfusion were available for four infants, of which three had undetectable levels of nipocalimab; the remaining sample had a low but detectable concentration of 0.04 μg/mL. Maternal serum concentrations evaluated within 1 week of these samples being drawn were in the pharmacologically active range (greater than 10 μg/mL). In cord blood samples (N = 10), only one had detectable levels (0.7 μg/mL). By 4 weeks after birth, nipocalimab levels were undetectable in the neonatal sera, and no receptor occupancy was observed. Only one colostrum sample had a pharmacologically active concentration of nipocalimab. All samples with detectable levels were collected from participants who had discontinued treatment less than 3 weeks prior to delivery, and none of the breastfed infants had detectable serum levels of nipocalimab at 4 weeks of life. Total serum IgG at the 48- and 96-week visits was within the normal range for six of seven infants, and IgA, IgM, and IgE were within the normal range in all available samples. Over the first 24 postnatal weeks, six infants developed eight infections, seven of which were mild or moderate in severity. One severe RSV event was recorded during a spike in the region, requiring hospitalization but not antiviral therapy. Protective anti-tetanus titers were evident in all available samples, and only one infant had a subthreshold anti-diptheria titer. While the results are preliminary, the analysis indicates that maternal nipocalimab treatment causes limited fetal and neonatal drug exposure and is not associated with longer-term negative impact on infant immune function.♦♦♦ Paper-based fluorescent assay for blood typing and antibody titer determination using long-term ambient-stored bioengineered RBCs. Li Y, Zhang Y, Liang S, et al. Nat Commun. 2026; 17(1): 2386. A novel, paper-based fluorescent assay can rapidly and accurately quantify anti-A and anti-B antibodies at the point-of-care. Forward and reverse blood typing are vital tools for detecting blood antigens and antibodies, respectively. Traditional tube testing can be time-consuming and is subject to inter-observer variability, while gel microcolumn agglutination (GMA) is more accurate but requires skilled operators and complex equipment. Several paper-based diagnostics using short-lived cell-based reagents have been developed, but these generally show limited ambient stability. Here, Li et al. report a rapid, paper-based blood typing test using bioengineered red blood cells (RBCs) and fluorescent labeling to quantify IgM. To begin, a mild “punching” solution was applied to RBCs to remove cellular contents while preserving the membrane, maintaining antigenicity. The resulting product was labeled with red fluorescent nanoparticles (RBCbm@Fluo) and freeze-dried into the reaction zone wells of a 3D-printed, flip-type device. Upon exposure to a test sample, RBCbm@Fluo rehydrate and IgM A-/B-antibodies activate an agglutination reaction. The device is then folded, directing the reaction mix through the filtration zone, and aggregates are retained while unaggregated RBCbm@Fluo are cleared. The test can be read after 5 min using a portable fluorescence analyzer to quantify the signal, in situ. Nanoflow cytometry confirmed that the freeze-drying process did not induce significant membrane protein damage, nor did it impact agglutination performance, and the freeze-dried reagent retained agglutination performance for up to 2 years. The RBCbm@Fluo device showed substantial comparability with existing commercial reverse typing tests and was capable of reliably typing a range of additional blood group systems including Rh, Kell, Duffy, Lewis, and MNS. A double-blind clinical head-to-head comparison with a GMA was conducted using residual blood from adult patients with unknown blood type (N = 641 blood samples; N = 388 plasma samples). Blood group typing showed 100% concordance with routinely used commercially available testing in this hospital. Antibody titers measured by PFA and GMA showed strong concordance, with discordances largely driven by the subjective thresholds used as cut offs for the GMA. The findings show that RBCbm@Fluo is a portable, objective reverse typing device. Moreover, the test can be adapted to detect IgG, broadening its clinical utility. Further work elucidating manufacturing and testing is required. The author declares no conflicts of interest. Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.