Abstract
Objective
Although several CAR-T products have been approved for hematologic malignancies, their post-marketing adverse-event reporting patterns may not be interchangeable. This study compared FAERS-derived positive signals across approved CAR-T products to distinguish recurrent class-related toxicities from product-skewed reporting patterns.
Methods
FAERS reports from Q1 2014 to Q2 2025 were reviewed for CAR-T products. Demographic characteristics were summarized for seven products, whereas PT- and SOC-level signal analyses were restricted to six products with sufficient signal data: Breyanzi, Kymriah, Tecartus, Yescarta, Abecma, and Carvykti. Positive signals were detected using ROR, PRR, and BCPNN algorithms. After predefined exclusion of nonclinical or non-AE terms, indication- or disease-progression-related terms, procedure- or investigation-related terms, efficacy- or persistence-related terms, and product-quality-related terms, eligible signals were compared at the PT and SOC levels. Nonspecific MedDRA-coded PTs were retained if they met the predefined signal criteria, but they were interpreted descriptively and were not used to define product-level profile labels.
Results
CRS was consistently among the leading PT-level signals across all six products. Yescarta and Tecartus showed higher reporting of ICANS and neurotoxicity-related PTs. Kymriah was more strongly represented by infection-, immune-, and hematologic-related signals, whereas Carvykti showed several neurologic PTs including Parkinsonism, cranial nerve paralysis, facial paralysis, and Bell’s palsy. Abecma showed relatively high-ranking systemic symptom-related PTs, including fatigue, headache, and decreased appetite. At the SOC level, nervous system disorders and infections and infestations were among the most frequently represented categories, but their proportions differed across products.
Conclusion
FAERS-based signal detection showed a common CAR-T reporting backbone centered on CRS, ICANS/neurotoxicity, hypotension, pyrexia, infection-related events, and hematologic events, together with product-skewed signal patterns. These findings represent disproportionality signals rather than comparative safety rankings, and FAERS data cannot establish causal relationships between CAR-T products and adverse events.
Keywords: adverse events, CAR-T, FAERS, pharmacovigilance, real-world study
1. Introduction
Chimeric antigen receptor T-cell (CAR-T) therapy has become an important therapeutic strategy for patients with relapsed or refractory hematologic malignancies and has achieved durable remission in a subset of patients, as demonstrated in pivotal studies of several approved CAR-T products (Abramson et al., 2020; Schuster et al., 2019; Wang et al., 2020; Neelapu et al., 2017; Munshi et al., 2021; Berdeja et al., 2021). However, treatment-related toxicities remain a major challenge limiting its broader clinical application (Morris et al., 2022; Schubert et al., 2021). In addition to typical class-effect toxicities such as cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS), differences among products may also exist in terms of infection, myelosuppression, and neurologic injury (Morris et al., 2022; Schubert et al., 2021; Lee et al., 2019; Guo et al., 2024).
Currently approved CAR-T products differ in target design, co-stimulatory domains, cellular expansion kinetics, and in vivo persistence, which may contribute to heterogeneity in their toxicity profiles (Abramson et al., 2020; Schuster et al., 2019; Wang et al., 2020; Neelapu et al., 2017; Munshi et al., 2021; Berdeja et al., 2021; Morris et al., 2022; Schubert et al., 2021). Previous studies have mainly focused on individual products or specific categories of toxicities, whereas systematic comparisons of the real-world safety spectra across different CAR-T products remain relatively limited (Guo et al., 2024). The FDA Adverse Event Reporting System (FAERS) database provides large-scale real-world data for post-marketing drug safety evaluation and is particularly suitable for generating hypotheses regarding rare or product-specific safety signals (Cutroneo et al., 2024). Previous FAERS-based pharmacovigilance research has shown that CAR-T-cell therapy is associated with a broad spectrum of reported toxicities across multiple organ systems, including CRS, ICANS, infection-related events, hematologic toxicities, and neurologic toxicities (Guo et al., 2024; Liu et al., 2026).
To address this gap, we conducted a comparative FAERS-based pharmacovigilance analysis of six CAR-T products using an integrated framework combining PT-level signal ranking, SOC-level signal distribution, and cross-product overlap patterns. Unlike previous studies focusing primarily on individual products or selected toxicity categories, the present analysis was designed to characterize both recurrent class-related signals and product-skewed reporting patterns across multiple CAR-T therapies. To improve clinical interpretability, predefined PT-exclusion rules were applied to remove non-AE terms and terms related to underlying indications or disease progression, lack of efficacy or treatment response, procedures or investigations, pharmacodynamic persistence, and product-quality issues before product-level comparisons were interpreted.
2. Materials and methods
2.1. Data source
This study used AE reports extracted from FAERS for CAR-T products. FAERS is a spontaneous reporting database and was used here for signal detection and descriptive comparison only. Demographic analyses included seven products: Aucatzyl, Breyanzi, Kymriah, Tecartus, Yescarta, Abecma, and Carvykti. Preferred term (PT)- and system organ class (SOC)-level analyses were restricted to six products with sufficient positive-signal data: Breyanzi, Kymriah, Tecartus, Yescarta, Abecma, and Carvykti. The overall workflow and analytic cohorts are shown in Figure 1.
FIGURE 1.
Study flow and analytic overview of CAR-T product cohorts and signal profiles. The bottom row shows descriptive product-level profile labels derived from clinically interpretable eligible PT-level signals, SOC-level distributions, forest plots, and cross-product overlap patterns. These labels should be interpreted as shorthand descriptions of reporting-signal patterns rather than as incidence-based or causality-based toxicity phenotypes.
2.2. Signal detection and data processing
FAERS reports from the first quarter of 2014 through the second quarter of 2025 were included in the analysis. Duplicate and follow-up reports were removed before signal detection. For records sharing the same CASEID, the report with the most recent FDA_DT was retained; when both CASEID and FDA_DT were identical, the record with the largest PRIMARYID was retained. Only reports in which the CAR-T product was designated as the primary suspect (PS) were included. Reports with missing demographic information were retained, and missing values were not imputed. Missing age or sex was categorized separately in the corresponding descriptive and subgroup analyses. Adverse events were standardized at the Preferred Term (PT) and System Organ Class (SOC) levels according to the Medical Dictionary for Regulatory Activities (MedDRA), version 25.0.
For each CAR-T product, PT-level disproportionality analyses were performed using three algorithms: the reporting odds ratio (ROR), proportional reporting ratio (PRR), and Bayesian confidence propagation neural network (BCPNN) (Cutroneo et al., 2024; Evans et al., 2001; Rothman et al., 2004; Bate et al., 1998). The fourfold contingency table was constructed at the drug–PT pair level. In the primary analysis, ROR was used as the principal criterion for signal detection, and a drug–PT pair was considered a positive signal when it met the predefined ROR criterion listed in Table 1. PRR and BCPNN were calculated as complementary disproportionality measures.
TABLE 1.
Formulas and signal detection criteria for reporting odds ratio (ROR), proportional reporting ratio (PRR), and Bayesian confidence propagation neural network (BCPNN).
| Algorithms | Equation | Criteria | |
|---|---|---|---|
| ROR | ROR = ad/bc 95% CI = exp[ln(ROR) ± 1.96√(1/a +1/b + 1/c + 1/d)] |
Lower limit of 95% CI > 1, a ≥3 | |
| PRR | PRR = a(c + d)/c(a + b) χ 2 = [(ad - bc)^2](a + b + c + d)/[(a + b) (c + d) (a + c) (b + d)] |
PRR >2, χ 2 > 4, a ≥3 | |
| BCPNN | IC = log2 [a(a + b + c + d)/((a + b) (a + c))] 95% CI = E(IC) ± 2√V(IC) |
IC025 > 0 | |
| Fourfold table of disproportionality measures | Target AE | Other AE | Total |
| Target drugs | a | b | a + b |
| Other drugs | c | d | c + d |
| Total | a + c | b + d | a + b + c + d |
After initial signal detection, PTs were further screened to remove terms unlikely to represent interpretable adverse-event signals, including nonclinical or non-AE terms, laboratory investigation terms, surgical or medical procedure terms, underlying indications or disease progression, lack of efficacy or treatment response, CAR-T persistence or other pharmacodynamic durability indicators, and product quality issues. Incomplete case records that could not support PT-level interpretation were also excluded. PT screening was performed by three investigators, and borderline or uncertain PTs were discussed until consensus was reached before the final inclusion or exclusion decision. Nonspecific but formally coded MedDRA PTs were not removed solely because of limited clinical specificity if they met the primary ROR-based signal criterion; these terms were retained for descriptive reporting and tabulation but were not considered when characterizing product-level reporting patterns. Eligible PT signals were then ranked by reporting frequency, and the top 30 PTs for each product were used for the main comparative visualizations. For the subsequent subgroup analyses, a more stringent three-algorithm concordance criterion was applied, whereby a drug–PT pair was considered a positive signal only when the predefined thresholds for ROR, PRR, and BCPNN listed in Table 1 were simultaneously satisfied.
2.3. Statistical analysis and visualization
Descriptive analyses were performed for sex, age, reporting year, and geographic distribution of each product. The frequencies and RORs (95% confidence intervals [95% CI]) of the top 30 PT-level signals were summarized, and visualization analyses were conducted using bubble plots, forest plots, and a network diagram. In addition, eligible PT-level signal distribution patterns among different products were compared at the SOC level.
3. Results
3.1. Demographic characteristics of AE reports
The demographic characteristics of adverse event (AE) reports associated with seven CAR-T products in the FAERS database are presented in Figure 2 and Table 2. The United States was the primary source of AE reports for all products. Differences were observed among products with respect to sex, age, and reporting-year distributions. Among the included products, Yescarta accounted for the largest number of AE reports, followed by Carvykti, whereas Aucatzyl had the fewest reports.
FIGURE 2.
Demographic characteristics of adverse event reports in the FAERS database. (A) Annual number of reports for seven CAR-T products; (B) Geographic distribution of reports for the seven products; (C) Sex distribution of reports for the seven products.
TABLE 2.
Demographic characteristics of AE reports associated with seven CAR-T products in the FAERS database.
| Characteristic | Aucatzyl | Breyanzi | Kymriah | Tecartus | Yescarta | Abecma | Carvykti | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Category | n (%) | Category | n (%) | Category | n (%) | Category | n (%) | Category | n (%) | Category | n (%) | Category | n (%) | |
| Sex | Male | 5 (62.5) | Male | 297(55.5) | Male | 1409 (47.2) | Male | 885 (61.0) | Male | 2592 (48.5) | Male | 425 (49.8) | Male | 1053 (33.4) |
| Female | 3 (37.5) | Female | 189 (35.3) | Female | 908 (30.4) | Female | 325 (22.4) | Female | 1674 (31.3) | Female | 298 (34.9) | Female | 808 (25.6) | |
| Unknown | 0 (0) | Unknown | 49 (9.2) | Unknown | 669 (22.4) | Unknown | 240 (16.6) | Unknown | 1075 (20.1) | Unknown | 131 (15.3) | Unknown | 1291 (41.0) | |
| Age (years) | <18 | 0(0) | <18 | 1(0.2) | <18 | 637 (21.3) | <18 | 8 (0.6) | <18 | 11 (0.2) | <18 | 0(0) | <18 | 0(0) |
| 18-<65 | 6(75.0) | 18-<65 | 147(27.5) | 18-<65 | 731 (24.5) | 18-<65 | 485 (33.4) | 18-<65 | 2134(40.0) | 18-<65 | 247(28.9) | 18-<65 | 350 (11.1) | |
| 65–85 | 1(12.5) | 65–85 | 316(59.1) | 65–85 | 488 (16.3) | 65–85 | 480(33.1) | 65–85 | 1414(26.5) | 65–85 | 388(45.4) | 65–85 | 519 (16.5) | |
| >85 | 0(0) | >85 | 5(0.9) | >85 | 5 (0.2) | >85 | 2 (0.1) | >85 | 9 (0.2) | >85 | 5 (0.6) | >85 | 2 (0.1) | |
| Missing | 1(12.5) | Missing | 66(12.3) | Missing | 1125(37.7) | Missing | 475(32.8) | Missing | 1773(33.2) | Missing | 214(25.1) | Missing | 2281(72.4) | |
| Reporting year | 2014 | 0(0) | 2014 | 0(0) | 2014 | 1 (0.0) | 2014 | 0(0) | 2014 | 0(0) | 2014 | 0(0) | 2014 | 0(0) |
| 2015 | 0(0) | 2015 | 0(0) | 2015 | 0(0) | 2015 | 0(0) | 2015 | 0(0) | 2015 | 0(0) | 2015 | 0(0) | |
| 2016 | 0(0) | 2016 | 0(0) | 2016 | 0(0) | 2016 | 0(0) | 2016 | 0(0) | 2016 | 0(0) | 2016 | 0(0) | |
| 2017 | 0(0) | 2017 | 0(0) | 2017 | 10 (0.3) | 2017 | 0(0) | 2017 | 1 (0.0) | 2017 | 0(0) | 2017 | 0(0) | |
| 2018 | 0(0) | 2018 | 1(0.2) | 2018 | 152 (5.1) | 2018 | 0(0) | 2018 | 438 (8.2) | 2018 | 0(0) | 2018 | 0(0) | |
| 2019 | 0(0) | 2019 | 0(0) | 2019 | 318 (10.6) | 2019 | 0(0) | 2019 | 616 (11.5) | 2019 | 4 (0.5) | 2019 | 0(0) | |
| 2020 | 0(0) | 2020 | 1(0.2) | 2020 | 640 (21.4) | 2020 | 25 (1.7) | 2020 | 530 (9.9) | 2020 | 0(0) | 2020 | 0(0) | |
| 2021 | 0(0) | 2021 | 86(16.1) | 2021 | 527 (17.6) | 2021 | 206(14.2) | 2021 | 504 (9.4) | 2021 | 104(12.2) | 2021 | 0(0) | |
| 2022 | 0(0) | 2022 | 120(22.4) | 2022 | 505 (16.9) | 2022 | 285(19.7) | 2022 | 641 (12.0) | 2022 | 264(30.9) | 2022 | 159 (5.0) | |
| 2023 | 0(0) | 2023 | 99(18.5) | 2023 | 353 (11.8) | 2023 | 362(25.0) | 2023 | 1036(19.4) | 2023 | 207(24.2) | 2023 | 606 (19.2) | |
| 2024 | 0(0) | 2024 | 135(25.2) | 2024 | 344 (11.5) | 2024 | 411(28.3) | 2024 | 1126(21.1) | 2024 | 183(21.4) | 2024 | 1308(41.5) | |
| 2025 | 8(100) | 2025 | 93(17.4) | 2025 | 136 (4.6) | 2025 | 161(11.1) | 2025 | 449 (8.4) | 2025 | 92 (10.8) | 2025 | 1079(34.2) | |
| Geographical distribution | US | 8(100) | US | 462(86.4) | US | 1952(65.4) | US | 1034(71.3) | US | 3440(64.4) | US | 702(82.2) | US | 2997(95.1) |
| Spain | 0(0) | Spain | 1(0.2) | Spain | 57(1.9) | Spain | 34(2.3) | Spain | 394(7.4) | Spain | 2(0.2) | Spain | 22(0.7) | |
| France | 0(0) | France | 4(0.7) | France | 65(2.2) | France | 69(4.8) | France | 183(3.4) | France | 51(6.0) | France | 12(0.4) | |
| Germany | 0(0) | Germany | 3(0.6) | Germany | 65(2.2) | Germany | 59(4.1) | Germany | 181(3.4) | Germany | 19(2.2) | Germany | 35(1.1) | |
| Japan | 0(0) | Japan | 49(9.2) | Japan | 163(5.5) | Japan | 0(0) | Japan | 91(1.7) | Japan | 52(6.1) | Japan | 0(0) | |
| Italy | 0(0) | Italy | 2(0.4) | Italy | 38(1.3) | Italy | 65(4.5) | Italy | 217(4.1) | Italy | 1(0.1) | Italy | 5(0.2) | |
| | UK | 0(0) | UK | 1(0.2) | UK | 37(1.2) | UK | 52(3.6) | UK | 168(3.1) | UK | 1(0.1) | UK | 8(0.3) |
| | Australia | 0(0) | Australia | 0(0) | Australia | 127(4.3) | Australia | 10(0.7) | Australia | 118(2.2) | Australia | 0(0) | Australia | 10(0.3) |
| Canada | 0(0) | Canada | 2(0.4) | Canada | 46(1.5) | Canada | 27(1.9) | Canada | 62(1.2) | Canada | 0(0) | Canada | 1(0.0) | |
| Poland | 0(0) | Poland | 0(0) | Poland | 42(1.4) | Poland | 14(1.0) | Poland | 71(1.3) | Poland | 0(0) | Poland | 4(0.1) | |
| Switzerland | 0(0) | Switzerland | 3(0.6) | Switzerland | 12(0.4) | Switzerland | 16(1.1) | Switzerland | 45(0.8) | Switzerland | 23(2.7) | Switzerland | 1(0) | |
| Unknown | 0(0) | Unknown | 0(0) | Unknown | 95(3.2) | Unknown | 0(0) | Unknown | 1(0.0) | Unknown | 1(0.1) | Unknown | 0(0) | |
| Austria | 0(0) | Austria | 4(0.7) | Austria | 12(0.4) | Austria | 14(1.0) | Austria | 57(1.1) | Austria | 1(0.1) | Austria | 5(0.2) | |
| CZ | 0(0) | CZ | 0(0) | CZ | 57(1.9) | CZ | 4(0.3) | CZ | 25(0.5) | CZ | 0(0) | CZ | 3(0.1) | |
| Korea | 0(0) | Korea | 0(0) | Korea | 75(2.5) | Korea | 0(0) | Korea | 0(0) | Korea | 0(0) | Korea | 1(0) | |
| Israel | 0(0) | Israel | 0(0) | Israel | 14(0.5) | Israel | 5(0.3) | Israel | 45(0.8) | Israel | 0(0) | Israel | 8(0.3) | |
| Portugal | 0(0) | Portugal | 0(0) | Portugal | 0(0) | Portugal | 7(0.5) | Portugal | 60(1.1) | Portugal | 0(0) | Portugal | 0(0) | |
| China | 0(0) | China | 0(0) | China | 0(0) | China | 3(0.2) | China | 50(0.9) | China | 0(0) | China | 12(0.4) | |
| Netherlands | 0(0) | Netherlands | 0(0) | Netherlands | 9(0.3) | Netherlands | 15(1.0) | Netherlands | 24(0.4) | Netherlands | 0(0) | Netherlands | 7(0.2) | |
| Greece | 0(0) | Greece | 0(0) | Greece | 1(0) | Greece | 12(0.8) | Greece | 36(0.7) | Greece | 0(0) | Greece | 0(0) | |
| Others | 0(0) | Others | 4(0.7) | Others | 119(4.0) | Others | 10(0.7) | Others | 73(1.4) | Others | 1(0.1) | Others | 21(0.7) | |
US, UK, and CZ, denote the United States, the United Kingdom, and the Czech Republic, respectively.
The extent of missing demographic information varied substantially across products. Missing age information accounted for 12.5%, 12.3%, 37.7%, 32.8%, 33.2%, 25.1%, and 72.4% of reports for Aucatzyl, Breyanzi, Kymriah, Tecartus, Yescarta, Abecma, and Carvykti, respectively. Unknown sex accounted for 0%, 9.2%, 22.4%, 16.6%, 20.1%, 15.3%, and 41.0% of reports, respectively. Missing demographic values were not imputed and were retained as separate missing/unknown categories in the descriptive analyses.
3.1.1. Demographic characteristics table
Table 2 provides detailed information on the sex, age, reporting year, and geographic distribution of AE reports associated with the seven CAR-T products.
3.2. Identification and analysis of positive AE signals at the PT level
After application of the predefined PT-exclusion rules, the numbers of eligible PT-level positive AE signals were 39 for Breyanzi, 141 for Kymriah, 81 for Tecartus, 149 for Yescarta, 57 for Abecma, and 106 for Carvykti. The complete eligible signal lists are provided in Supplementary Tables S1–S6, and the top 30 PT-level signals for each product are summarized in Figure 3 and Table 3.
FIGURE 3.
Top 30 PT-level positive AE signals and corresponding SOC classifications for six CAR-T products in the FAERS database. (A) Breyanzi; (B) Kymriah; (C) Tecartus; (D) Yescarta; (E) Abecma; and (F) Carvykti. CRS, cytokine release syndrome; ICANS, immune effector cell-associated neurotoxicity syndrome; AKI, acute kidney injury; HLH, haemophagocytic lymphohistiocytosis; IRR, infusion-related reaction; FN, febrile neutropenia; AF, atrial fibrillation; TLS, tumour lysis syndrome; MODS, multiple organ dysfunction syndrome; GBS, Guillain-Barré syndrome; CDI, Clostridium difficile infection; IEC-HS, immune effector cell-associated HLH-like syndrome.
TABLE 3.
Top 30 positive AE signals associated with six CAR-T products in the FAERS database.
| Breyanzi | Kymriah | Tecartus | Yescarta | Abecma | Carvykti | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| PT | n (%) | ROR (95%CI) | PT | n (%) | ROR (95% CI) | PT | n (%) | ROR (95% CI) | PT | n (%) | ROR (95% CI) | PT | n (%) | ROR (95% CI) | PT | n (%) | ROR (95% CI) |
| Cytokine release syndrome | 215 (15.60) | 741.51 (640.44–858.53) | Cytokine release syndrome | 1084 (7.52) | 350.05 (328.2–373.36) | Cytokine release syndrome | 641 (15.49) | 362.82 (332.8–395.55) | Cytokine release syndrome | 2474 (17.13) | 1007.23 (960.68–1056.03) | Cytokine release syndrome | 494 (20.20) | 1037.73 (938.73–1147.17) | Cytokine release syndrome | 423 (7.65) | 337.63 (305.29–373.4) |
| Immune effector cell-associated neurotoxicity syndrome | 106 (7.69) | 1,142.69 (934.88–1396.69) | Pyrexia | 756 (5.25) | 9.89 (9.19–10.65) | Immune effector cell-associated neurotoxicity syndrome | 439 (10.61) | 743.51 (669.63–825.54) | Immune effector cell-associated neurotoxicity syndrome | 1207 (8.36) | 1772.44 (1652.26–1901.35) | Fatigue | 151 (6.17) | 5.23 (4.44–6.17) | Pyrexia | 179 (3.24) | 5.96 (5.14–6.92) |
| Neurotoxicity | 84 (6.10) | 245.12 (196.46–305.85) | Neurotoxicity | 309 (2.14) | 84.13 (75.07–94.29) | Neurotoxicity | 163 (3.94) | 124.96 (106.65–146.41) | Neurotoxicity | 875 (6.06) | 258.6 (240.99–277.5) | Immune effector cell-associated neurotoxicity syndrome | 147 (6.01) | 886.48 (748.02–1050.57) | Immune effector cell-associated neurotoxicity syndrome | 162 (2.93) | 420 (358.04–492.68) |
| Hypotension | 32 (2.32) | 7.33 (5.16–10.4) | Hypotension | 303 (2.10) | 6.63 (5.92–7.43) | Death | 119 (2.88) | 2.06 (1.72–2.48) | Death | 514 (3.56) | 2.62 (2.4–2.86) | Pyrexia | 54 (2.21) | 4.02 (3.07–5.27) | Parkinsonism | 152 (2.75) | 182.07 (154.73–214.24) |
| Pyrexia | 31 (2.25) | 4.1 (2.87–5.86) | Hypogammaglobulinaemia | 275 (1.91) | 221.34 (195.75–250.27) | Pyrexia | 86 (2.08) | 4.12 (3.32–5.1) | Pyrexia | 421 (2.92) | 5.36 (4.86–5.91) | Headache | 52 (2.13) | 2.13 (1.62–2.81) | Neurotoxicity | 104 (1.88) | 72.46 (59.64–88.04) |
| Confusional state | 18 (1.31) | 5.01 (3.15–7.97) | Immune effector cell-associated neurotoxicity syndrome | 249 (1.73) | 250.67 (220.22–285.33) | Hypotension | 81 (1.96) | 6.83 (5.48–8.51) | Hypotension | 206 (1.43) | 4.46 (3.89–5.12) | Decreased appetite | 50 (2.04) | 5.3 (4.01–7.02) | Malaise | 94 (1.70) | 2.38 (1.94–2.92) |
| Aphasia | 16 (1.16) | 23.31 (14.24–38.16) | Hypoxia | 174 (1.21) | 22.01 (18.95–25.58) | Confusional state | 51 (1.23) | 5.76 (4.37–7.6) | Neutropenia | 191 (1.32) | 6.18 (5.35–7.12) | Neurotoxicity | 48 (1.96) | 75.38 (56.62–100.36) | Cranial nerve paralysis | 90 (1.63) | 3890.16 (3038.74–4980.14) |
| Decreased appetite | 15 (1.09) | 2.8 (1.68–4.65) | Cytopenia | 164 (1.14) | 66.55 (56.98–77.73) | Encephalopathy | 48 (1.16) | 31.04 (23.33–41.29) | Encephalopathy | 174 (1.21) | 31.03 (26.71–36.06) | Neutropenia | 44 (1.80) | 8.43 (6.26–11.36) | Facial paralysis | 74 (1.34) | 58.98 (46.85–74.23) |
| Mental status changes | 12 (0.87) | 19.3 (10.93–34.08) | Thrombocytopenia | 119 (0.83) | 4.67 (3.9–5.59) | Hypoxia | 41 (0.99) | 17.68 (12.99–24.06) | Confusional state | 164 (1.14) | 4.35 (3.73–5.08) | Anaemia | 41 (1.68) | 5.48 (4.02–7.46) | Bell’s palsy | 66 (1.19) | 578.66 (450.66–743.01) |
| Tremor | 12 (0.87) | 3.22 (1.83–5.69) | Neutropenia | 111 (0.77) | 3.57 (2.96–4.31) | Pancytopenia | 39 (0.94) | 11.54 (8.42–15.83) | Pancytopenia | 150 (1.04) | 11.79 (10.04–13.86) | Hypogammaglobulinaemia | 29 (1.19) | 129.45 (89.66–186.89) | Haemophagocytic lymphohistiocytosis | 64 (1.16) | 75.51 (58.96–96.71) |
| Encephalopathy | 11 (0.80) | 20.31 (11.22–36.77) | Tachycardia | 98 (0.68) | 4.79 (3.93–5.84) | Tremor | 39 (0.94) | 4.53 (3.3–6.21) | Aphasia | 130 (0.90) | 18.1 (15.22–21.52) | Acute kidney injury | 28 (1.14) | 3.65 (2.52–5.3) | COVID-19 | 58 (1.05) | 3.82 (2.95–4.94) |
| Hypogammaglobulinaemia | 9 (0.65) | 70.63 (36.65–136.12) | Confusional state | 86 (0.60) | 2.27 (1.84–2.81) | Tachycardia | 38 (0.92) | 7.33 (5.33–10.1) | Tremor | 126 (0.87) | 3.23 (2.71–3.85) | Hypotension | 28 (1.14) | 3.57 (2.46–5.18) | Illness | 57 (1.03) | 7.33 (5.64–9.51) |
| Memory impairment | 9 (0.65) | 2.93 (1.52–5.65) | Depressed level of consciousness | 69 (0.48) | 7.73 (6.1–9.8) | Acute kidney injury | 37 (0.89) | 2.98 (2.15–4.11) | Thrombocytopenia | 104 (0.72) | 4.07 (3.35–4.93) | Tremor | 27 (1.10) | 4.09 (2.8–5.98) | Infection | 52 (0.94) | 4.17 (3.17–5.48) |
| Haemophagocytic lymphohistiocytosis | 8 (0.58) | 37.41 (18.66–74.98) | Haemophagocytic lymphohistiocytosis | 67 (0.47) | 30.14 (23.69–38.36) | Aphasia | 32 (0.77) | 19.61 (13.84–27.79) | Hypoxia | 99 (0.69) | 12.4 (10.17–15.12) | Tachycardia | 24 (0.98) | 6.92 (4.63–10.35) | Infusion related reaction | 44 (0.80) | 7.77 (5.78–10.46) |
| Pleural effusion | 8 (0.58) | 5.84 (2.92–11.7) | Multiple organ dysfunction syndrome | 57 (0.40) | 5.44 (4.19–7.06) | Neutropenia | 29 (0.70) | 2.66 (1.84–3.83) | Tachycardia | 94 (0.65) | 4.58 (3.74–5.61) | Haemophagocytic lymphohistiocytosis | 21 (0.86) | 55.56 (36.14–85.42) | Cytopenia | 43 (0.78) | 44.69 (33.08–60.37) |
| Infusion related reaction | 8 (0.58) | 5.66 (2.82–11.34) | Aphasia | 54 (0.37) | 7.47 (5.72–9.76) | Seizure | 27 (0.65) | 2.95 (2.02–4.31) | Cytopenia | 93 (0.64) | 37.17 (30.28–45.63) | Somnolence | 21 (0.86) | 2.65 (1.73–4.08) | Colitis | 42 (0.76) | 12.84 (9.48–17.4) |
| Chills | 8 (0.58) | 3.06 (1.53–6.13) | Febrile neutropenia | 54 (0.37) | 3.55 (2.72–4.64) | Infection | 27 (0.65) | 2.67 (1.83–3.9) | Febrile neutropenia | 85 (0.59) | 5.59 (4.52–6.92) | Myalgia | 20 (0.82) | 2.98 (1.92–4.63) | Neutropenia | 38 (0.69) | 3.18 (2.31–4.38) |
| Febrile neutropenia | 7 (0.51) | 4.82 (2.29–10.12) | Encephalopathy | 53 (0.37) | 9.34 (7.13–12.23) | Haemophagocytic lymphohistiocytosis | 26 (0.63) | 29.59 (20.1–43.56) | Infection | 84 (0.58) | 2.57 (2.08–3.19) | Hypoalbuminaemia | 19 (0.78) | 64.84 (41.26–101.9) | Hypotension | 36 (0.65) | 2.02 (1.45–2.8) |
| Hypoxia | 6 (0.44) | 7.83 (3.51–17.46) | Respiratory failure | 53 (0.37) | 3.09 (2.36–4.05) | Febrile neutropenia | 25 (0.60) | 5.13 (3.46–7.6) | Haemophagocytic lymphohistiocytosis | 75 (0.52) | 33.72 (26.85–42.36) | Cytopenia | 19 (0.78) | 44.53 (28.34–69.97) | Sepsis | 35 (0.63) | 3.51 (2.52–4.89) |
| Depressed level of consciousness | 6 (0.44) | 7.02 (3.15–15.65) | Staphylococcal infection | 46 (0.32) | 5.95 (4.46–7.95) | Sepsis | 25 (0.60) | 3.85 (2.6–5.71) | Sepsis | 72 (0.50) | 2.76 (2.19–3.48) | Pancytopenia | 19 (0.78) | 8.77 (5.59–13.78) | Thrombocytopenia | 30 (0.54) | 3.06 (2.13–4.37) |
| Septic shock | 6 (0.44) | 6.33 (2.84–14.11) | Coagulopathy | 45 (0.31) | 11.6 (8.65–15.55) | Infusion related reaction | 21 (0.51) | 3.65 (2.38–5.61) | Septic shock | 69 (0.48) | 6.96 (5.49–8.82) | Chills | 18 (0.74) | 3.88 (2.44–6.18) | Immune-mediated enterocolitis | 27 (0.49) | 106.03 (72.49–155.08) |
| Cardiac arrest | 6 (0.44) | 3.2 (1.44–7.14) | Pancytopenia | 40 (0.28) | 3.12 (2.29–4.26) | Brain oedema | 20(0.48) | 28.73 (18.49–44.63) | Disorientation | 57 (0.39) | 6.06 (4.67–7.87) | Infusion related reaction | 15 (0.61) | 5.98 (3.6–9.93) | Respiratory failure | 27 (0.49) | 4.1 (2.81–5.99) |
| Atrial fibrillation | 6 (0.44) | 2.76 (1.24–6.14) | Enterococcal infection | 34 (0.24) | 30.76 (21.94–43.12) | Mental status changes | 20(0.48) | 15.13 (9.74–23.49) | Atrial fibrillation | 57 (0.39) | 2.5 (1.93–3.24) | Thrombocytopenia | 14 (0.57) | 3.22 (1.91–5.45) | Bacterial infection | 26 (0.47) | 16.2 (11.02–23.83) |
| Hypoalbuminaemia | 5 (0.36) | 30.09 (12.5–72.44) | Rhinovirus infection | 33 (0.23) | 42.9 (30.42–60.49) | Depressed level of consciousness | 20(0.48) | 8.88 (5.72–13.79) | Multiple organ dysfunction syndrome | 52 (0.36) | 4.95 (3.77–6.5) | Confusional state | 14 (0.57) | 2.18 (1.29–3.68) | Febrile neutropenia | 23 (0.42) | 3.94 (2.62–5.94) |
| Tumour lysis syndrome | 5 (0.36) | 27.07 (11.25–65.16) | Tachypnoea | 32 (0.22) | 10.4 (7.35–14.72) | Thrombocytopenia | 20(0.48) | 2.86 (1.84–4.44) | Mental status changes | 51 (0.35) | 7.8 (5.92–10.27) | Haematotoxicity | 13 (0.53) | 38.78 (22.47–66.91) | Pancytopenia | 22 (0.40) | 4.48 (2.94–6.8) |
| Incontinence | 5 (0.36) | 21.31 (8.85–51.3) | Nervous system disorder | 32 (0.22) | 4.19 (2.96–5.92) | Multiple organ dysfunction syndrome | 19(0.46) | 7.64 (4.86–11.99) | Hypogammaglobulinaemia | 49 (0.34) | 36.88 (27.82–48.9) | Febrile bone marrow aplasia | 12 (0.49) | 75.14 (42.57–132.63) | Unevaluable event | 21 (0.38) | 2.99 (1.95–4.6) |
| COVID-19 pneumonia | 5 (0.36) | 18.86 (7.84–45.4) | Bone marrow failure | 31 (0.22) | 6.24 (4.39–8.89) | Agitation | 19(0.46) | 5.76 (3.67–9.04) | Pleural effusion | 49 (0.34) | 3.41 (2.57–4.51) | Sepsis | 12 (0.49) | 2.71 (1.54–4.79) | Guillain-barre syndrome | 20 (0.36) | 48.4 (31.17–75.17) |
| Acute respiratory failure | 5 (0.36) | 11.93 (4.96–28.72) | Clostridium difficile infection | 30 (0.21) | 6.16 (4.31–8.82) | Respiratory failure | 18(0.43) | 4.53 (2.85–7.2) | Agitation | 48 (0.33) | 2.74 (2.06–3.64) | Immunodeficiency | 10 (0.41) | 18.27 (9.81–34.01) | Influenza | 20 (0.36) | 2.07 (1.33–3.21) |
| Pneumonia aspiration | 5 (0.36) | 9.05 (3.76–21.79) | Disorientation | 30 (0.21) | 3.19 (2.23–4.56) | Delirium | 17(0.41 ) |
7.85 (4.88–12.65) | Dysgraphia | 44 (0.30) | 28.25 (20.99–38.02) | Hypoxia | 10 (0.41) | 7.35 (3.95–13.68) | Immune effector cell-associated hlh-like syndrome | 18 (0.33) | 5772.29 (3203.84–10399.8) |
| Hyperkalaemia | 5 (0.36) | 6.54 (2.72–15.73) | Septic shock | 30 (0.21) | 3.02 (2.11–4.32) | Cytopenia | 16(0.39 ) |
13.49 (8.25–22.06) | Respiratory failure | 43 (0.30) | 2.5 (1.85–3.37) | Febrile neutropenia | 9 (0.37) | 3.48 (1.81–6.71) | Septic shock | 17 (0.31) | 4.46 (2.77–7.18) |
Unevaluable event was retained as a nonspecific but formally coded MedDRA PT, that met the predefined signal criteria; it was interpreted descriptively and was not used to define product-level profile labels.
Across the six products, cytokine release syndrome was consistently among the leading PT-level signals, indicating a recurrent CAR-T class-related reporting pattern. Other repeatedly observed signals included immune effector cell-associated neurotoxicity syndrome, neurotoxicity, hypotension, pyrexia, and several infection- and hematologic-related PTs.
Product-level differences were also observed. Yescarta and Tecartus showed higher reporting proportions of immune effector cell-associated neurotoxicity syndrome, with 1207 reports (8.36%) for Yescarta and 439 reports (10.61%) for Tecartus. Neurotoxicity was also more frequently reported for Yescarta [875 reports (6.06%)] than for Kymriah [309 reports (2.14%)] or Carvykti [104 reports (1.88%)]. Kymriah was more strongly represented by infection-, immune-, and hematologic-related PTs, including hypogammaglobulinaemia [275 reports (1.91%)], hypoxia [174 reports (1.21%)], and neutropenia [111 reports (0.77%)]. Carvykti showed several neurologic PTs, including Parkinsonism [152 reports (2.75%)], cranial nerve paralysis [90 reports (1.63%)], facial paralysis [74 reports (1.34%)], and Bell’s palsy [66 reports (1.19%)]. In Abecma, systemic symptom-related PTs ranked relatively high, including fatigue [151 reports (6.17%)], headache [52 reports (2.13%)], and decreased appetite [50 reports (2.04%)].
The concise product-level profile labels shown in Figure 1 were derived from the integrated interpretation of clinically interpretable eligible PT-level signals, top-ranked clinically interpretable PT signals, SOC-level distributions, and cross-product overlap patterns. These labels were used as descriptive summaries of reporting-signal patterns rather than as incidence-based or causality-based toxicity phenotypes.
3.3. SOC-level distribution of eligible PT signals
Table 4 summarizes the SOC-level distribution of eligible PT-level positive AE signals for the six CAR-T products. The percentages in Table 4 represent the proportion of eligible PT signals assigned to each SOC within a given product, rather than the incidence or reporting frequency of adverse events. Overall, nervous system disorders and infections and infestations were among the most frequently represented SOC categories, but their relative contributions differed across products.
TABLE 4.
SOC-level distribution of eligible PT-level positive AE signals for six CAR-T products.
| SOC | Breyanzi n (%) | Kymriah n (%) | Tecartus n (%) | Yescarta n (%) | Abecma n (%) | Carvykti n (%) |
|---|---|---|---|---|---|---|
| Nervous system disorders | 7 (17.9) | 22 (15.6) | 16 (19.8) | 25 (16.8) | 10 (17.5) | 31 (29.2) |
| Infections and infestations | 6 (15.4) | 47 (33.3) | 15 (18.5) | 45 (30.2) | 9 (15.8) | 29 (27.4) |
| Metabolism and nutrition disorders | 6(15.4) | 7(5.0) | 9(11.1) | 7(4.7) | 5(8.8) | 4(3.8) |
| Blood and lymphatic system disorders | 4(10.3) | 14(9.9) | 9(11.1) | 13(8.7) | 11(19.3) | 7(6.6) |
| Respiratory, thoracic and mediastinal disorders | 4(10.3) | 9(6.4) | 7(8.6) | 9(6.0) | 4(7.0) | 5(4.7) |
| Immune system disorders | 3(7.7) | 8(5.7) | 5(6.2) | 6(4.0) | 4(7.0) | 8(7.5) |
| General disorders and administration site conditions | 2(5.1) | 5(3.5) | 4(4.9) | 5(3.4) | 3(5.3) | 5 (4.7) |
| Psychiatric disorders | 2(5.1) | 4(2.8) | 5(6.2) | 6(4.0) | 2(3.5) | 3(2.8) |
| Cardiac disorders | 2(5.1) | 5(3.5) | 4(4.9) | 9(6.0) | 2(3.5) | 0(0) |
| Vascular disorders | 1(2.6) | 6(4.3) | 2(2.5) | 6(4.0) | 1(1.8) | 1(0.9) |
| Renal and urinary disorders | 1(2.6) | 3(2.1) | 3(3.7) | 2(1.3) | 2(3.5) | 0(0) |
| Injury, poisoning and procedural complications | 1(2.6) | 1(0.7) | 1(1.2) | 3(2.0) | 1(1.8) | 2(1.9) |
| Gastrointestinal disorders | 0(0) | 2(1.4) | 0(0) | 7(4.7) | 0(0) | 9(8.5) |
| Hepatobiliary disorders | 0(0) | 3(2.1) | 1(1.2) | 2(1.3) | 2(3.5) | 1(0.9) |
| Eye disorders | 0(0) | 3(2.1) | 0(0) | 2(1.3) | 0(0) | 1(0.9) |
| Congenital, familial and genetic disorders | 0(0) | 1(0.7) | 0(0) | 1(0.7) | 0(0) | 0(0) |
| Skin and subcutaneous tissue disorders | 0(0) | 1(0.7) | 0(0) | 0(0) | 0(0) | 0(0) |
| Musculoskeletal and connective tissue disorders | 0(0) | 0(0) | 0(0) | 0(0) | 1(1.8) | 0(0) |
| Endocrine disorders | 0(0) | 0(0) | 0(0) | 1(0.7) | 0(0) | 0(0) |
Percentages were calculated as the number of eligible PT-level signals assigned to a given SOC, divided by the total number of eligible PT-level signals for that product.SOC, percentages reflect the distribution of eligible positive PT, signals across SOCs, and should not be interpreted as event frequencies or incidence rates.
Carvykti showed the highest proportion of signals assigned to nervous system disorders (29.2%), followed by Tecartus (19.8%), Breyanzi (17.9%), Abecma (17.5%), Yescarta (16.8%), and Kymriah (15.6%), indicating a relatively stronger neurologic signal concentration for Carvykti at the SOC level. For infections and infestations, Kymriah, Yescarta, and Carvykti showed higher proportions, accounting for 33.3%, 30.2%, and 27.4% of eligible PT signals, respectively, compared with Tecartus (18.5%), Abecma (15.8%), and Breyanzi (15.4%).
Other SOC categories also contributed to product-skewed patterns. Abecma had the highest proportion of blood and lymphatic system disorder signals (19.3%), compared with Tecartus (11.1%), Breyanzi (10.3%), Kymriah (9.9%), Yescarta (8.7%), and Carvykti (6.6%). Breyanzi showed comparable proportions of infections and infestations and metabolism and nutrition disorders (both 15.4%), whereas Carvykti also showed a relatively higher proportion of gastrointestinal disorder signals (8.5%) than the other products. These findings indicate that the SOC-level signal distribution was not uniform across CAR-T products and supported the product-skewed reporting patterns observed in the PT-level analyses.
3.4. Comparison of positive PT signals among different products
Figure 4 presents the ROR forest plot for the top 30 PT-level positive AE signals associated with Carvykti. Carvykti was selected for main-text presentation because it showed the highest proportion of nervous-system SOC signals and several distinctive neurologic PT-level signals. Forest plots for Breyanzi, Kymriah, Tecartus, Yescarta, and Abecma are provided in Supplementary Figures S1–S5.
FIGURE 4.
Forest plot of RORs (95% CI) for the top 30 PT-level positive AE signals associated with Carvykti. Unevaluable event was retained as a nonspecific but formally coded MedDRA PT and was interpreted descriptively.
Figure 5 shows the overlap and product-skewed distribution of the top 30 PT-level positive AE signals across the six CAR-T products. In this network, shared PTs were linked to multiple products, whereas product-skewed PTs were mainly connected to one product or a limited number of products. CRS, immune effector cell-associated neurotoxicity syndrome, neurotoxicity, hypotension, pyrexia, and several infection- and hematologic-related PTs formed the recurrent CAR-T class-related signal pattern.
FIGURE 5.
Network diagram of the top 30 PT-level positive AE signals associated with six CAR-T products in the FAERS database. The diagram illustrates shared and product-skewed PT-level signals among different CAR-T products.
Product-skewed patterns were also evident. Breyanzi showed a relatively balanced reporting profile without a single dominant product-skewed category. Kymriah was more strongly represented by infection- and immunosuppression-related PTs, including hypogammaglobulinaemia, hypoxia, and neutropenia. Tecartus showed an ICANS-focused profile, whereas Yescarta showed an acute neurotoxicity-skewed reporting profile. Abecma was characterized by cytopenia/fatigue-related reporting, supported by blood and lymphatic system signals and high-ranking fatigue-related systemic symptoms. Carvykti was mainly characterized by a delayed/atypical neurologic reporting profile, including Parkinsonism, cranial nerve paralysis, facial paralysis, and Bell’s palsy.
3.5. Subgroup analyses
3.5.1. Age-subgroup analysis
Age-stratified analyses showed heterogeneous reporting patterns across the six CAR-T products (Supplementary Figure S6). Breyanzi showed a positive signal in patients aged 65–85 years (ROR 1.42, 95% CI 1.13–1.78), whereas Kymriah showed a positive signal in patients aged <18 years (ROR 1.40, 95% CI 1.29–1.52). No consistent age-related pattern was observed for the other products. Missing-age reports were retained as a separate subgroup, and estimates based on small subgroup counts should be interpreted cautiously.
3.5.2. Sex-subgroup analysis
Sex-stratified analyses showed generally modest differences between male and female subgroups across the six CAR-T products (Supplementary Figure S7). ROR estimates were broadly comparable between sexes for Breyanzi, Kymriah, Abecma, and Carvykti, whereas lower RORs were observed for Tecartus in males and for Yescarta in both sexes. Missing-sex reports were retained as a separate subgroup and showed higher ROR estimates for Tecartus and Yescarta. Overall, no consistent sex-specific reporting pattern was observed across products.
3.5.3. Outcome-subgroup analysis
Outcome-subgroup analysis showed heterogeneous reporting patterns across CAR-T products (Supplementary Figure S8). Positive signals were observed for hospitalization with Breyanzi (ROR 1.72, 95% CI 1.37–2.17), and for the “Other” outcome category with Kymriah (ROR 1.38, 95% CI 1.28–1.50), Tecartus (ROR 1.34, 95% CI 1.15–1.56), and Abecma (ROR 2.17, 95% CI 1.80–2.61). Carvykti showed positive signals for serious outcomes (ROR 1.22, 95% CI 1.09–1.37) and hospitalization (ROR 1.46, 95% CI 1.29–1.66). No corresponding positive outcome signal was observed for Yescarta.
Collectively, subgroup analyses showed variation in disproportionality patterns across age, sex, and clinical outcome categories. Age-related and outcome-specific variations were observed for selected products, whereas sex-related differences were generally limited.
4. Discussion
In this FAERS-based pharmacovigilance analysis, six CAR-T products showed a recurrent adverse-event reporting pattern centered on CRS, ICANS/neurotoxicity, hypotension, pyrexia, infection-related events, and hematologic events. This pattern is consistent with the established safety profile of CAR-T therapy, in which excessive immune activation, neurologic involvement, infection susceptibility, and myelosuppression represent major clinical concerns (Morris et al., 2022; Schubert et al., 2021; Lee et al., 2019; Shahid et al., 2024; Hill and Seo, 2020). Because FAERS is a spontaneous reporting system, these findings should be interpreted as reporting signals rather than incidence estimates or evidence of causality (Cutroneo et al., 2024).
Despite this shared reporting backbone, product-level differences were observed. Yescarta and Tecartus showed a higher concentration of neurotoxicity-related PTs, including ICANS and neurotoxicity. This finding is consistent with the recognized importance of neurologic toxicity in CAR-T therapy and with previous pharmacovigilance evidence suggesting that AE reporting patterns may differ among CAR-T products (Morris et al., 2022; Schubert et al., 2021; Lee et al., 2019; Guo et al., 2024; Liu et al., 2026). However, the present FAERS-based analysis cannot determine whether these differences reflect intrinsic product biology, differences in indications, patient characteristics, treatment settings, or reporting behavior.
Kymriah showed a reporting pattern more strongly represented by infection-, immune-, and hematologic-related PTs, including hypogammaglobulinaemia, hypoxia, and neutropenia. These findings are clinically plausible because infection risk after CAR-T therapy may be influenced by immune reconstitution, B-cell aplasia, hypogammaglobulinaemia, cytopenia, prior therapies, and disease-related immunosuppression (Schubert et al., 2021; Shahid et al., 2024; Hill and Seo, 2020). Therefore, the Kymriah-associated pattern observed in this study should be interpreted as a signal of relatively prominent infection- and immunosuppression-related reporting, rather than as a direct comparison of true infection incidence across products.
Carvykti showed several neurologic PTs that were less typical of conventional acute ICANS, including Parkinsonism, cranial nerve paralysis, facial paralysis, and Bell’s palsy. These findings are consistent with published concerns regarding atypical neurologic events after BCMA-directed CAR-T therapy (Berdeja et al., 2021; Van Oekelen et al., 2021; Karschnia et al., 2023). This pattern is also consistent with recent real-world evidence showing more frequent delayed neurotoxicity with cilta-cel than with ide-cel, while conventional ICANS did not show a similar difference (Hansen et al., 2025). The present FAERS findings support an atypical neurologic reporting pattern for Carvykti that is broadly consistent with the delayed neurologic toxicities described in previous clinical studies. Accordingly, Figure 1 label of delayed neurotoxicity was used as a descriptive shorthand for this delayed/atypical neurologic reporting profile, supported by the observed signal pattern and prior clinical literature. However, because FAERS reports do not provide uniform time-to-onset information, this label should not be interpreted as direct evidence that delayed onset was established in the present dataset. Further studies with clinical timing, exposure denominators, and standardized neurologic assessment are needed to clarify whether these signals represent delayed neurologic toxicity and to define their clinical risk factors.
The observed product-skewed reporting patterns may be related to differences in target antigen, co-stimulatory domain, expansion kinetics, in vivo persistence, disease indication, prior treatment exposure, and clinical monitoring practices (Abramson et al., 2020; Schuster et al., 2019; Wang et al., 2020; Neelapu et al., 2017; Munshi et al., 2021; Berdeja et al., 2021; Morris et al., 2022; Schubert et al., 2021). Patient composition may further contribute to this heterogeneity, as CAR-T products are used across pediatric and adult populations and across different underlying malignancies, particularly lymphoma and multiple myeloma, with differences in disease characteristics, prior treatment exposure, and treatment line. For example, differences between CD28−and 4-1BB-based CAR constructs may affect expansion and persistence profiles, which could contribute to heterogeneity in toxicity reporting. Nevertheless, FAERS data alone cannot separate biological mechanisms from confounding by indication, patient population, or reporting bias. Therefore, these product-level differences should be regarded as pharmacovigilance hypotheses that require validation in clinical cohorts.
From a clinical perspective, these findings suggest that CAR-T safety management should combine standardized prevention of class-related toxicities with product-aware monitoring. CRS and ICANS remain core toxicities requiring early recognition and standardized grading (Schubert et al., 2021; Lee et al., 2019). For products with prominent neurotoxicity-related reporting, dynamic neurologic assessment and ICANS-oriented monitoring should be emphasized (Lee et al., 2019; Van Oekelen et al., 2021; Karschnia et al., 2023). For products with higher infection- and immunosuppression-related reporting, infection surveillance, antimicrobial prophylaxis when appropriate, and immunoglobulin monitoring or replacement should be considered according to clinical guidelines and patient-specific risk factors (Shahid et al., 2024; Hill and Seo, 2020). These recommendations should be interpreted as signal-informed considerations for clinical surveillance rather than evidence of comparative safety or causal differences among CAR-T products. Product-aware monitoring should complement, rather than replace, established guideline-based management and individualized patient assessment.
Several limitations should be considered when interpreting these findings. As a spontaneous reporting system, FAERS is subject to under-reporting, stimulated reporting, potential residual duplicate reporting, and other reporting biases, which may also vary across countries and regions. Cross-product comparisons may also be confounded by differences in approved indications, patient populations, disease stage, prior and concomitant therapies, manufacturing platforms, year of approval, and duration of post-marketing exposure, while incomplete clinical information limits adjustment for these factors. Moreover, the lack of reliable exposure denominators precludes estimation of true incidence, and disproportionality signals reflect reporting associations rather than causal relationships or direct measures of clinical risk. Therefore, observed differences across CAR-T products should be interpreted as differences in post-marketing reporting patterns rather than as evidence of differences in intrinsic product toxicity. Despite predefined PT-exclusion rules, residual misclassification may remain, and the limited post-marketing data for Aucatzyl precluded its inclusion in the PT- and SOC-level analyses. More broadly, evidence across different phases of clinical trials underscores the complementary role of post-marketing pharmacovigilance in identifying uncommon, delayed, or incompletely characterized adverse events that may not be fully captured during pre-approval clinical development (Upadhye et al., 2025).
Future studies should validate FAERS-derived signals using independent pharmacovigilance databases such as VigiBase and EudraVigilance, together with longitudinal and machine-learning approaches. Integration with active surveillance systems and clinically characterized real-world registries may further strengthen the clinical interpretation and translational relevance of CAR-T safety signals.
5. Conclusion
In this FAERS-based pharmacovigilance analysis, six CAR-T products shared a recurrent reporting pattern centered on CRS, ICANS/neurotoxicity, hypotension, pyrexia, infection-related events, and hematologic events. At the same time, the distribution of positive signals varied across products. Breyanzi showed a relatively mixed reporting pattern, whereas infection- and immunosuppression-related signals were relatively prominent for Kymriah, ICANS and other neurologic signals for Tecartus and Yescarta, cytopenia- and fatigue-related signals for Abecma, and several neurologic signals, including Parkinsonism and cranial nerve-related events, for Carvykti. These findings represent product-associated disproportionality patterns rather than differences in clinical incidence or intrinsic toxicity and do not establish causal relationships. Further validation using prospective real-world studies with reliable exposure denominators and clinical time-to-event information is warranted.
Funding Statement
The author(s) declared that financial support was not received for this work and/or its publication.
Footnotes
Edited by: Mohamed Hamed, Al Azhar University, Egypt
Reviewed by: Yuan Liu, Indiana University Bloomington, United States
Zeinab Bakr, Assiut University, Egypt
Krishna S. Upadhye, Krescent Medical Research Pvt Ltd., India
Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.
Ethics statement
Ethical approval was not required for the study involving humans in accordance with the local legislation and institutional requirements. Written informed consent to participate in this study was not required from the participants or the participants’ legal guardians/next of kin in accordance with the national legislation and the institutional requirements.
Author contributions
SX: Conceptualization, Validation, Methodology, Investigation, Formal Analysis, Software, Visualization, Data curation, Writing – original draft. YL: Conceptualization, Validation, Methodology, Visualization, Data curation, Writing – original draft. HW: Validation, Methodology, Formal Analysis, Writing – review and editing, Data curation, Software, Investigation. LT: Writing – review and editing, Methodology, Software, Validation, Data curation. RL: Conceptualization, Validation, Resources, Project administration, Writing – review and editing, Supervision, Methodology.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar.2026.1913235/full#supplementary-material
References
- Abramson J. S., Palomba M. L., Gordon L. I., Lunning M. A., Wang M., Arnason J., et al. (2020). Lisocabtagene maraleucel for patients with relapsed or refractory large B-cell lymphomas (TRANSCEND NHL 001): a multicentre seamless design study. Lancet 396 (10254), 839–852. 10.1016/S0140-6736(20)31366-0 [DOI] [PubMed] [Google Scholar]
- Bate A., Lindquist M., Edwards I. R., Olsson S., Orre R., Lansner A., et al. (1998). A bayesian neural network method for adverse drug reaction signal generation. Eur. J. Clin. Pharmacol. 54 (4), 315–321. 10.1007/s002280050466 [DOI] [PubMed] [Google Scholar]
- Berdeja J. G., Madduri D., Usmani S. Z., Jakubowiak A., Agha M., Cohen A. D., et al. (2021). Ciltacabtagene autoleucel, a B-cell maturation antigen-directed chimeric antigen receptor T-cell therapy in patients with relapsed or refractory multiple myeloma (CARTITUDE-1): a phase 1b/2 open-label study. Lancet 398 (10297), 314–324. 10.1016/S0140-6736(21)00933-8 [DOI] [PubMed] [Google Scholar]
- Cutroneo P. M., Sartori D., Tuccori M., Crisafulli S., Battini V., Carnovale C., et al. (2024). Conducting and interpreting disproportionality analyses derived from spontaneous reporting systems. Front. Drug Saf. Regul. 3, 1323057. 10.3389/fdsfr.2023.1323057 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Evans S. J. W., Waller P. C., Davis S. (2001). Use of proportional reporting ratios (PRRs) for signal generation from spontaneous adverse drug reaction reports. Pharmacoepidemiol. Drug Saf. 10 (6), 483–486. 10.1002/pds.677 [DOI] [PubMed] [Google Scholar]
- Guo Z., Ding Y., Wang M., Zhai Q., Liu J., Du Q. (2024). Comparing the differences in adverse events among chimeric antigen receptor T-cell therapies: a real-world pharmacovigilance study. Pharmaceuticals 17 (8), 1025. 10.3390/ph17081025 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hansen D. K., Peres L. C., Dima D., Richards A., Shune L., Afrough A., et al. (2025). Comparison of standard-of-care idecabtagene vicleucel and ciltacabtagene autoleucel in relapsed/refractory multiple myeloma. J. Clin. Oncol. 43 (13), 1597–1609. 10.1200/JCO-24-01730 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hill J. A., Seo S. K. (2020). How I prevent infections in patients receiving CD19-targeted chimeric antigen receptor T cells for B-cell malignancies. Blood 136 (8), 925–935. 10.1182/blood.2019004000 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Karschnia P., Miller K. C., Yee A. J., Rejeski K., Johnson P. C., Raje N., et al. (2023). Neurologic toxicities following adoptive immunotherapy with BCMA-directed CAR T cells. Blood 142 (14), 1243–1248. 10.1182/blood.2023020571 [DOI] [PubMed] [Google Scholar]
- Lee D. W., Santomasso B. D., Locke F. L., Ghobadi A., Turtle C. J., Brudno J. N., et al. (2019). ASTCT consensus grading for cytokine release syndrome and neurologic toxicity associated with immune effector cells. Biol. Blood Marrow Transplant. 25 (4), 625–638. 10.1016/j.bbmt.2018.12.758 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Liu Y., Yang J., Iqbal M., Uyanik M., Luo M., Yang L., et al. (2026). A retrospective pharmacovigilance analysis based on the FAERS database reveals sex-associated differences in toxicities of CAR T-cell therapy. Nat. Commun. 17 (1), 6025. 10.1038/s41467-026-72816-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Morris E. C., Neelapu S. S., Giavridis T., Sadelain M. (2022). Cytokine release syndrome and associated neurotoxicity in cancer immunotherapy. Nat. Rev. Immunol. 22 (2), 85–96. 10.1038/s41577-021-00547-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Munshi N. C., Anderson L. D., Shah N., Madduri D., Berdeja J., Lonial S., et al. (2021). Idecabtagene vicleucel in relapsed and refractory multiple myeloma. N. Engl. J. Med. 384 (8), 705–716. 10.1056/NEJMoa2024850 [DOI] [PubMed] [Google Scholar]
- Neelapu S. S., Locke F. L., Bartlett N. L., Lekakis L. J., Miklos D. B., Jacobson C. A., et al. (2017). Axicabtagene ciloleucel CAR T-cell therapy in refractory large B-cell lymphoma. N. Engl. J. Med. 377 (26), 2531–2544. 10.1056/NEJMoa1707447 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Rothman K. J., Lanes S., Sacks S. T. (2004). The reporting odds ratio and its advantages over the proportional reporting ratio. Pharmacoepidemiol. Drug Saf. 13 (8), 519–523. 10.1002/pds.1001 [DOI] [PubMed] [Google Scholar]
- Schubert M. L., Schmitt M., Wang L., Ramos C. A., Jordan K., Müller-Tidow C., et al. (2021). Side-effect management of chimeric antigen receptor (CAR) T-cell therapy. Ann. Oncol. 32 (1), 34–48. 10.1016/j.annonc.2020.10.478 [DOI] [PubMed] [Google Scholar]
- Schuster S. J., Bishop M. R., Tam C. S., Waller E. K., Borchmann P., McGuirk J. P., et al. (2019). Tisagenlecleucel in adult relapsed or refractory diffuse large B-cell lymphoma. N. Engl. J. Med. 380 (1), 45–56. 10.1056/NEJMoa1804980 [DOI] [PubMed] [Google Scholar]
- Shahid Z., Jain T., Dioverti V., Pennisi M., Mikkilineni L., Thiruvengadam S. K., et al. (2024). Best practice considerations by the American society of transplant and cellular therapy: infection prevention and management after chimeric antigen receptor T cell therapy for hematological malignancies. Transplant. Cell. Ther. 30 (10), 955–969. 10.1016/j.jtct.2024.07.018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Upadhye K. S., Tayade P., Patidar H., Yadav N. (2025). A meta-analysis of serious adverse events across different phases of clinical trials. Natl. J. Pharmacol. Ther. 3 (2), 95–104. 10.4103/njpt.njpt_3_25 [DOI] [Google Scholar]
- Van Oekelen O., Aleman A., Upadhyaya B., Schnakenberg S., Madduri D., Gavane S., et al. (2021). Neurocognitive and hypokinetic movement disorder with features of parkinsonism after BCMA-targeting CAR-T cell therapy. Nat. Med. 27 (12), 2099–2103. 10.1038/s41591-021-01564-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wang M., Munoz J., Goy A., Locke F. L., Jacobson C. A., Hill B. T., et al. (2020). KTE-X19 CAR T-cell therapy in relapsed or refractory mantle-cell lymphoma. N. Engl. J. Med. 382 (14), 1331–1342. 10.1056/NEJMoa1914347 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.