A retrospective pharmacovigilance analysis based on the FAERS database reveals sex-associated differences in toxicities of CAR T-cell therapy Authors: Yuan Liu, Jingwen Yang, Madiha Iqbal, Mehmet Uyanik, Mei Luo, Liqi Yang, Yanyan Lou, Lixia Diao, Olalekan O. Oluwole, Javid J. Moslehi, Douglas B. Johnson, Leng Han Published: 2026 ABSTRACT Chimeric antigen receptor (CAR) T-cell therapy has significantly advanced treatment outcomes for hematological malignancies; however, therapy-associated toxicities are often severe and sex-related differences in toxicity remain under-investigated. Here, we extract data from FDA Adverse Event Reporting System (FAERS) to evaluate sex-associated differences in CAR T-cell therapy toxicities. Among 7700 cases, females show higher reporting odds of cytokine release syndrome (CRS), with a reporting odds ratio (ROR) of 1.10 and a confidence interval (CI) of [1.00, 1.20], as well as leukemias (ROR = 1.34; CI [1.05, 1.70]). Elevated reporting odds in females are observed across multiple organ systems. Comparisons across cancer treatments indicate that sex-associated reporting patterns in CAR T-cell therapy cannot be attributed to baseline sex differences across cancer therapies. Stratified analyses further identify heterogeneity by cancer type and CAR T product. These results highlight distinct sex-associated toxicity patterns and may support incorporating sex into toxicity management. Introduction Chimeric antigen receptor (CAR) T-cell therapy has transformed the therapeutic landscape for various hematological malignancies and is now being explored for solid tumors1,2. Despite its great potential, CAR T-cell therapy is associated with a broad spectrum of toxicities2–4, including critical toxicities, such as cytokine release syndrome (CRS), immune effector cell-associated neurotoxicity syndrome (ICANS), and, more recently, concerns regarding secondary malignancies5,6. Nevertheless, comprehensive studies investigating sex-associated differences in CAR T-cell therapy-related toxicity remain limited. A primary barrier is the small sample size in clinical trials, typically fewer than 100 participants, which significantly reduces the ability to detect sex-associated differences. Statistical power analyses indicate that detecting a 10% absolute risk difference (e.g., comparing a 20% risk in one sex with a 30% risk in the other) requires enrolling at least 294 patients per sex (see Methods). Additionally, clinical trials of CAR T-cell therapy often do not report toxicity outcomes stratified by sex, complicating meta-analyses7. Therefore, we turn to real-world data, which typically provides larger sample sizes to enable such analyses8. Here, we conduct intra-CAR T and inter-treatment analyses to address two complementary questions: (1) whether sex-differential reporting patterns are observed among CAR T recipients, and (2) whether these patterns differ from those observed in other cancer treatments with comparable disease contexts. We identify higher reporting odds in females across multiple toxicity categories and organ systems, with additional heterogeneity observed across cancer types and CAR T products. Comparative analyses indicate that these patterns are not explained by baseline sex differences in cancer therapies, suggesting CAR T–specific sex-associated reporting profiles. These findings highlight the importance of considering sex as a variable in CAR T-cell therapy and provide a framework for future studies to refine risk stratification and toxicity management. Results Cohort identification and curation of CAR T–associated toxicity terms To characterize sex-differential reporting patterns associated with CAR T-cell therapy, we analyzed FAERS reports using a curated set of guideline-informed CAR T–associated toxicities (Fig. 1a) and a comparator cohort of non–CAR T cancer treatments matched by disease context (Fig. 1b). We evaluated enrichment of toxicity reporting in CAR T and assessed sex-associated differences across organ systems (Fig. 1c), with parallel analyses in the comparator cohort to distinguish CAR T–associated patterns from background sex differences in oncology treatments (Fig. 1d).Fig. 1Overview of the analytical workflow for identifying sex-associated differences in CAR T–associated toxicities using FDA Adverse Event Reporting System (FAERS).a FAERS reports were used to identify CAR T cases, curate clinically relevant toxicity terms, and aggregate them at the MedDRA High Level Group Term (HLGT) level. b Enrichment of curated toxicities was evaluated in CAR T versus non–CAR T cancer therapies. c Sex differences in toxicity reporting were assessed within the CAR T cohort. d Specificity of CAR T–associated sex differences was evaluated by comparison with corresponding sex differences in non–CAR T therapies. The figure was created in BioRender (https://BioRender.com/pbls5hw). Overview of the analytical workflow for identifying sex-associated differences in CAR T–associated toxicities using FDA Adverse Event Reporting System (FAERS). a FAERS reports were used to identify CAR T cases, curate clinically relevant toxicity terms, and aggregate them at the MedDRA High Level Group Term (HLGT) level. b Enrichment of curated toxicities was evaluated in CAR T versus non–CAR T cancer therapies. c Sex differences in toxicity reporting were assessed within the CAR T cohort. d Specificity of CAR T–associated sex differences was evaluated by comparison with corresponding sex differences in non–CAR T therapies. The figure was created in BioRender (https://BioRender.com/pbls5hw). A total of 13,597 CAR T-cell–related safety reports were identified, of which 7,700 were retained after applying predefined data-quality filters and eligibility criteria (Fig. 1a and Supplementary Table 1). We extracted 2,788 unique reaction terms reported in CAR T cases. Because FAERS reaction terms may capture laboratory procedures, administrative entries, or clinical events unlikely to be attributable to CAR T-cell therapy, we excluded terms that did not plausibly represent treatment-related toxicities (e.g., lab test items) or did not conform to standard Medical Dictionary for Regulatory Activities (MedDRA) reporting9, resulting in 1170 candidate terms. Clinicians with CAR T expertise subsequently conducted manual curation to generate a guideline-informed toxicity term set. Specifically, each candidate term was systematically mapped to established CAR T toxicity management guidelines and reviews2,10–13. This process yielded 914 terms that were either explicitly referenced or conceptually aligned with recognized CAR T-associated toxicities, including cytokine release syndrome, immune effector cell–associated neurotoxicity syndrome, and other treatment-emergent adverse events (Supplementary Table 2). To assess whether the curated toxicity terms were preferentially reported in CAR T-cell recipients, we assembled a non-CAR T cancer-treatment cohort from FAERS using the National Cancer Institute Cancer Medications Enquiry Database (NCI CanMED) drug list, encompassing conventional chemotherapy through immunotherapy (Fig. 1b). To improve contextual comparability, we restricted this cohort to the same cancer types for which CAR T-cell therapy is indicated, yielding approximately 240,000 non-CAR T reports. Following prior FAERS studies, adverse event terms were collapsed into High Level Group Terms (HLGT) using MedDRA9 to reduce redundancy and enhance interpretability6. Using a consistent analytical workflow across cohorts, 48 HLGTs were evaluated. Of these, 45 showed statistically significant differences between CAR T and non–CAR T cohorts, and 39 exhibited higher reporting odds in CAR T reports (Fig. 2). A binomial test confirmed significant enrichment of these HLGTs among CAR T-cell reports (P < 0.001), supporting the relevance of the curated toxicity term set.Fig. 2Comparison of adverse event reporting between CAR T and non-CAR T cancer therapies.This circular plot summarizes reporting odds ratios (ROR) for MedDRA High Level Group Terms (HLGT) comparing CAR T with non-CAR T cancer treatments. Each sector represents an individual HLGT, grouped by its corresponding System Organ Class (SOC). The outer rings display disproportionality results, including both the ROR and the proportional reporting ratio (PRR), for CAR T versus non-CAR T. Color encoding denotes the direction of enrichment (red: higher reporting in CAR T; cyan: higher reporting in non-CAR T), with color intensity corresponding to effect size or statistical significance as shown in the legend. Only HLGTs meeting the predefined statistical significance criteria (false discovery rate, FDR < 0.1) are shown. Comparison of adverse event reporting between CAR T and non-CAR T cancer therapies. This circular plot summarizes reporting odds ratios (ROR) for MedDRA High Level Group Terms (HLGT) comparing CAR T with non-CAR T cancer treatments. Each sector represents an individual HLGT, grouped by its corresponding System Organ Class (SOC). The outer rings display disproportionality results, including both the ROR and the proportional reporting ratio (PRR), for CAR T versus non-CAR T. Color encoding denotes the direction of enrichment (red: higher reporting in CAR T; cyan: higher reporting in non-CAR T), with color intensity corresponding to effect size or statistical significance as shown in the legend. Only HLGTs meeting the predefined statistical significance criteria (false discovery rate, FDR < 0.1) are shown. Sex-associated differences in CAR T toxicities We next assessed sex-associated differences of critical CAR T-cell therapy toxicities, including CRS, ICANS, and secondary malignancies (Figs. 1c and 3). Females had significantly higher reporting odds of CRS (Reporting odds ratio [ROR] = 1.10, CI [1.00, 1.20], N = 2503/4775 males and 1603/2925 females) compared to males (Fig. 3a and Supplementary Fig. 1). Females also had slightly higher reporting odds of ICANS (ROR = 1.06, N = 943/4775 males and 602/2925 females), although it did not reach statistical significance (Fig. 3a and Supplementary Fig. 1).Fig. 3Multivariable logistic regression analysis on real-world pharmacovigilance data for toxicity difference between male and female among CAR T patients.a presents a forest plot of cytokine release syndrome and immune effector cell-associated neurotoxicity syndrome. b shows significant high level group terms in neoplasms by sex, and c presents significant high level group terms across all system organ classes by sex. In panels (b) and (c), bold and italic subtitles indicate the System Organ Classes (SOC) to which the High Level Group Terms (HLGT) belong. Points represent reporting odds ratios (ROR), and horizontal lines indicate 95% confidence intervals. Statistical significance was assessed using multivariable logistic regression, with P values derived from Wald tests and adjusted for multiple comparisons in panels b and c using the false discovery rate (FDR) method. Multivariable logistic regression analysis on real-world pharmacovigilance data for toxicity difference between male and female among CAR T patients. a presents a forest plot of cytokine release syndrome and immune effector cell-associated neurotoxicity syndrome. b shows significant high level group terms in neoplasms by sex, and c presents significant high level group terms across all system organ classes by sex. In panels (b) and (c), bold and italic subtitles indicate the System Organ Classes (SOC) to which the High Level Group Terms (HLGT) belong. Points represent reporting odds ratios (ROR), and horizontal lines indicate 95% confidence intervals. Statistical significance was assessed using multivariable logistic regression, with P values derived from Wald tests and adjusted for multiple comparisons in panels b and c using the false discovery rate (FDR) method. To investigate the potential for reported secondary malignancy, we evaluated all malignancy-related terms (i.e., neoplasms) in FAERS. Under the neoplasm system organ class (SOC), three HLGTs showed significant sex-associated differences: miscellaneous and unspecified malignant neoplasms (ROR = 1.50, CI [1.07, 2.10], N = 74/4775 males and 66/2,925 females) and leukemias (ROR = 1.34, CI [1.05, 1.70], N = 157/4775 males and 125/2925 females) were more frequent in females, whereas skin malignancies (ROR = 0.49, CI [0.29, 0.83], N = 62/4775 males and 18/2,925 females) were more common in males (Fig. 3b and Supplementary Fig. 2). In light of growing concerns regarding secondary T-cell malignancies following CAR T-cell therapy14, we also inspected the sex difference in ‘Lymphomas non−Hodgkin’s T−cell’. Higher reporting odds were observed in females (ROR = 1.49), although due to limited case numbers (17 cases total: 9 males, 8 females), the test did not reach statistical significance. CAR T-cell therapy exhibits toxicity across multiple organ systems, including the neurologic and gastrointestinal systems10. We therefore expanded our analysis to all reported toxicities in FAERS, and our analysis identified 10 HLGTs with statistically significant sex-associated differences in addition to neoplasm, spanning 7 system organ classes, (Fig. 3c and Supplementary Fig. 3). Females exhibited higher reporting odds than males in all of these categories, with the most pronounced difference observed in bone, calcium, magnesium, and phosphorus metabolism disorders (ROR = 2.38, CI [1.56, 3.62], N = 37/4775 males and 54/2925 females), and seizures (incl subtypes) (ROR = 1.69, CI [1.23, 2.31], N = 78/4775 males and 82/2925 females). We also found that females had 1.28-fold higher odds of reporting white blood cell disorders than males (ROR = 1.28; CI [1.10, 1.50]; N = 407/4775 males and 317/2925 females). This finding is consistent with a prior study using the Nationwide Readmissions Database, which reported higher odds of leukopenia, a component of white blood cell reduction, among females (OR = 1.26, CI [1.06, 1.50])15. Assessment of CAR T–specific sex-associated differences To determine whether the observed sex-associated reporting patterns were specific to CAR T-cell therapy rather than reflecting background sex differences common across cancer treatments, we performed an internal comparator analysis using the matched non–CAR T cancer cohort (Fig. 1d). Sex-stratified reporting patterns differed between CAR T and non–CAR T cohorts across multiple HLGTs, and statistical testing confirmed that 7 of the 13 HLGTs with significant sex-associated differences in the CAR T cohort exhibited sex differences that were statistically distinguishable from those observed in the non–CAR T cohort (Fig. 4). For example, bone, calcium, magnesium, and phosphorus metabolism disorders showed markedly higher reporting odds in female CAR T recipients (ROR = 2.38, CI [1.56, 3.62], N = 37/4775 males and 54/2925 females), whereas the corresponding sex difference in the non–CAR T cohort was modest (ROR = 1.12, CI [1.00, 1.26], N = 571/131,315 males and 582/116,131 females). These findings suggest that several sex-associated reporting patterns observed among CAR T-cell recipients are not solely attributable to baseline sex differences shared across cancer therapies.Fig. 4Sex-differential reporting patterns in CAR T and non-CAR T cancer therapies.Reporting odds ratios (ROR) compare female versus male reporting within each treatment group. Total numbers of cases in each group are as follows: non–CAR T, male (N = 131,315); non–CAR T, female (N = 116,131); CAR T, male (N = 4775); CAR T, female (N = 2925). Colors indicate the direction and magnitude of reporting disproportionality. Asterisks denote HLGTs for which the sex-differential reporting pattern in CAR T recipients differs significantly from that observed in the non–CAR T cancer-treatment cohort. Points represent reporting odds ratios (ROR), and horizontal lines indicate 95% confidence intervals. Within each cohort, statistical significance was assessed using multivariable logistic regression, with P values derived from Wald tests and adjusted for multiple comparisons using the false discovery rate (FDR) method. Differences in sex-associated effects between cohorts were evaluated by comparing coefficients from separate multivariable logistic regression models using Wald-type Z tests. Sex-differential reporting patterns in CAR T and non-CAR T cancer therapies. Reporting odds ratios (ROR) compare female versus male reporting within each treatment group. Total numbers of cases in each group are as follows: non–CAR T, male (N = 131,315); non–CAR T, female (N = 116,131); CAR T, male (N = 4775); CAR T, female (N = 2925). Colors indicate the direction and magnitude of reporting disproportionality. Asterisks denote HLGTs for which the sex-differential reporting pattern in CAR T recipients differs significantly from that observed in the non–CAR T cancer-treatment cohort. Points represent reporting odds ratios (ROR), and horizontal lines indicate 95% confidence intervals. Within each cohort, statistical significance was assessed using multivariable logistic regression, with P values derived from Wald tests and adjusted for multiple comparisons using the false discovery rate (FDR) method. Differences in sex-associated effects between cohorts were evaluated by comparing coefficients from separate multivariable logistic regression models using Wald-type Z tests. Stratified analyses by cancer type and product CAR T-cell therapy is used across several hematological malignancies, prompting us to perform a stratified analysis by cancer type. Among 5557 patients diagnosed with B-cell lymphoma, significant sex-associated differences were identified in 11 HLGTs, with females exhibiting higher reporting odds in 11 (Fig. 5a and Supplementary Fig. 4), spanning 8 system organ classes, including blood and lymphatic system disorders and Infections and infestations. Among patients with plasma cell myeloma, sex-associated differences emerged in headaches.Fig. 5Forest plot of significant HLGTs across cancer types or products by sex.a Cancer types are indicated in bold and italic, system organ classes (SOC) in bold, and High Level Group Terms (HLGT) in regular text. b Products are indicated in bold and italic, SOCs in bold, and HLGTs in regular text. Points represent reporting odds ratios (ROR), and horizontal lines indicate 95% confidence intervals. Statistical significance was assessed using multivariable logistic regression, with P values derived from Wald tests and adjusted for multiple comparisons using the false discovery rate (FDR) method. Forest plot of significant HLGTs across cancer types or products by sex. a Cancer types are indicated in bold and italic, system organ classes (SOC) in bold, and High Level Group Terms (HLGT) in regular text. b Products are indicated in bold and italic, SOCs in bold, and HLGTs in regular text. Points represent reporting odds ratios (ROR), and horizontal lines indicate 95% confidence intervals. Statistical significance was assessed using multivariable logistic regression, with P values derived from Wald tests and adjusted for multiple comparisons using the false discovery rate (FDR) method. We further investigated sex-associated differences across different CAR T-cell therapy products. Among the products, sex-associated differences in toxicities were observed in multiple products, such as Tisagenlecleucel and Axicabtagene Ciloleucel (Fig. 5b and Supplementary Fig. 5). Notably, products displayed product-specific sex-associated differential reporting odds. For example, females had significantly higher reporting odds of headaches (ROR = 2.72, CI [1.50, 4.94], N = 18/305 males and 35/239 females) when treated with Idecabtagene Vicleucel, and bone, calcium, magnesium, and phosphorus metabolism disorders (ROR = 2.09, CI [1.31, 3.34], N = 31/2492 males and 43/1628 females) when treated with Axicabtagene Ciloleucel. Discussion In conclusion, our study highlights previously unrecognized sex-associated differences in CAR T-cell therapy toxicities potentially due to limited sample size. Females exhibited significantly higher reporting odds of toxicities after CAR T-cell infusion, including CRS, and exhibited a consistent pattern of elevated adverse event reporting odds across diverse cancer types and therapy products. Our analysis also identifies sex-associated differences across multiple toxicity categories. For example, females show increased reporting odds of bacterial infectious disorders (ROR = 1.33; 95% CI [1.11, 1.61]), as well as leukemias (ROR = 1.34; 95% CI [1.07, 1.70]) and miscellaneous malignant neoplasms (ROR = 1.50; 95% CI [1.07, 2.09]), along with elevated reporting across multiple organ systems. Although FAERS data do not permit causal inference, these findings suggest the potential need for increased vigilance in monitoring infectious complications and long-term malignancy adverse events in female patients receiving CAR T-cell therapy. These findings underscore the need for further research on clinical management strategies to ensure precision medicine approaches in cancer therapy16. The observed sex-associated differences in reported toxicities may reflect a complex interplay of pharmacokinetics, pharmacodynamics, immune regulation, and tumor–host interactions. Prior work in immuno-oncology has highlighted systematic sex biases driven by differences in sex hormones, X-chromosome gene dosage, immune cell composition and activation, cytokine signaling, and drug metabolism, all of which can influence both therapeutic efficacy and toxicity profiles17,18. Accordingly, the associations identified for individual toxicity categories in our study should be viewed as testable, treatment-context–specific hypotheses that warrant mechanistic investigation and validation in prospective and biologically annotated cohorts. Our study has several limitations inherent to FAERS, a retrospective pharmacovigilance database. Certain cancer types, CAR T-cell products, and rare toxicities were sparsely represented in FAERS, limiting statistical power in some subgroup analyses. For example, the small number of patients with acute leukemia or reports of secondary T-cell malignancies constrained our ability to detect sex-associated differences in these contexts. In addition, FAERS does not capture key individual-level factors such as race, menopausal status, prior treatment history, or time to toxicity onset, precluding adjustment for these potential confounders. More broadly, FAERS does not establish causal relationships between products and adverse events, lacks a defined exposed population denominator, and may contain incomplete, duplicate, or selectively reported cases, precluding estimation of incidence or absolute risk. Although we implemented data filtering and internal comparator analyses to mitigate some of these limitations, they cannot be fully eliminated; accordingly, our findings should be interpreted as reflecting patterns of reported events within FAERS and viewed as hypothesis-generating rather than definitive evidence of sex-specific toxicity risk. Method Power analysis A statistical power analysis was conducted to determine the minimal detectable difference in toxicity event rates between male and female participants. Assuming a baseline adverse event rate of 20% in males and 30% in females, calculations used a two-sided significance level of 0.05 and 80% power. Analyses were performed in R using the function ‘power.prop.test’ standard methods for two-sample comparisons of proportions19. Data collection This study was a retrospective analysis utilizing the FDA Adverse Event Reporting System (FAERS), a public pharmacovigilance database with reports submitted by healthcare professionals, manufacturers, and consumers. Complete FAERS data were downloaded through May 2024. Cases with treatment including ‘Idecabtagene icleucel’, ‘Abecma’, ‘Idecabtagene vicleucel’, ‘Brexucabtagene autoleucel’, ‘Tecartus’, ‘Axicabtagene Ciloleucel’, ‘Yescarta’, ‘Tisagenlecleucel’, ‘Kymriah’, ‘Lisocabtagene maraleucel’, ‘Breyanzi’, ‘Ciltacabtagene autoleucel’, or ‘CARVYKTI’ were selected for analysis. Cases lacking sex or age information were excluded. Duplicate entries identified by identical report IDs were removed, retaining only a single entry per case. Only cases involving adults were retained, defined as patients aged 20 years or older. Reported adverse events were manually curated and screened against professional guidelines and review articles on CAR T toxicity management2,10–13. Given the emerging nature of secondary malignancies following CAR T therapy, all reported symptoms classified under neoplasm-related terms were retained for analysis. Adverse reactions were required to match Preferred Terms (PT) according to the Medical Dictionary for Regulatory Activities (MedDRA) classification system. Reports not matching any PT were excluded due to ambiguous or inconsistent reporting. PTs were aggregated into HLGTs to reduce redundancy and improve interpretability. Reactions overlapping with drug indications were excluded to avoid misclassification errors. Cancer types were aggregated into three major groups. For cancer type identification, the drug indication field was analyzed and categorized as follows: Acute Leukemia: ‘B-cell type acute leukemia’, ‘Acute lymphocytic leukemia’, ‘B precursor type acute leukemia’, ‘Lymphocytic leukemia (not specified)’; B-Cell Lymphoma: ‘Diffuse large B-cell lymphoma’, ‘Mantle cell lymphoma’, ‘Follicular lymphoma’, ‘Mediastinal large B-cell lymphoma’, ‘Non-Hodgkin’s lymphoma (not specified)’, ‘B-Cell lymphoma (not specified)’, ‘Lymphoma (not specified)’; Plasma Cell Myeloma: classified into a single category; Unspecified Indication: all other uncategorized indications. A non–CAR T cancer treatment comparator cohort was constructed by identifying cancer therapies listed in the NCI CanMED resource (https://seer.cancer.gov/oncologytoolbox/canmed/hcpcs/) within FAERS. Comparator cases were restricted to adult patients ( ≥ 20 years) with cancer types overlapping those of CAR T therapy. Reports with missing sex or age information were excluded. Informed consent was waived because only deidentified data were used in this study. Statistical analysis Multiple logistic regression analysis was conducted using the Python package ‘statsmodels’. To evaluate enrichment of curated toxicity terms in CAR T versus non–CAR T reports, multivariable logistic regression was used. The model (1) was specified as:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mbox{logit}}}\left(p\right)=\log \left(\frac{p}{1-p}\right)={\beta }_{0}+{\beta }_{1}\times {\mbox{Age}}+{\beta }_{2}\times {\mbox{Sex}}+{\beta }_{3}\times {\mbox{CAR \,T}}.$$\end{document}logitp=logp1−p=β0+β1×Age+β2×Sex+β3×CAR T.Where:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p$$\end{document}p is the probability of experiencing a toxicity event.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{0}$$\end{document}β0 is the intercept.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{1},{\beta }_{2},\,{\beta }_{3}$$\end{document}β1,β2,β3 are the coefficients for age, sex, and treatment type, respectively.Age is a continuous variable.Sex is coded as 0 for male and 1 for femaleCAR T is coded as 0 for non-CAR T and 1 for CAR T \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p$$\end{document}p is the probability of experiencing a toxicity event. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p$$\end{document}p is the probability of experiencing a toxicity event. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{0}$$\end{document}β0 is the intercept. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{0}$$\end{document}β0 is the intercept. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{1},{\beta }_{2},\,{\beta }_{3}$$\end{document}β1,β2,β3 are the coefficients for age, sex, and treatment type, respectively. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{1},{\beta }_{2},\,{\beta }_{3}$$\end{document}β1,β2,β3 are the coefficients for age, sex, and treatment type, respectively. Age is a continuous variable. Age is a continuous variable. Sex is coded as 0 for male and 1 for female Sex is coded as 0 for male and 1 for female CAR T is coded as 0 for non-CAR T and 1 for CAR T CAR T is coded as 0 for non-CAR T and 1 for CAR T To assess sex-associated differences within the cohort, the following model (2) was used:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mbox{logit}}}\left(p\right)=\log \left(\frac{p}{1-p}\right)={\beta }_{0}+{\beta }_{1}\times {\mbox{Age}}+{\beta }_{2}\times {\mbox{Sex}}.$$\end{document}logitp=logp1−p=β0+β1×Age+β2×Sex.Where:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p$$\end{document}p is the probability of experiencing a toxicity event.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{0}$$\end{document}β0 is the intercept.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{1},{\beta }_{2}$$\end{document}β1,β2 are the coefficients for age and sex, respectively.Age is a continuous variable.Sex is coded as 0 for male and 1 for female \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p$$\end{document}p is the probability of experiencing a toxicity event. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p$$\end{document}p is the probability of experiencing a toxicity event. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{0}$$\end{document}β0 is the intercept. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{0}$$\end{document}β0 is the intercept. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{1},{\beta }_{2}$$\end{document}β1,β2 are the coefficients for age and sex, respectively. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{1},{\beta }_{2}$$\end{document}β1,β2 are the coefficients for age and sex, respectively. Age is a continuous variable. Age is a continuous variable. Sex is coded as 0 for male and 1 for female Sex is coded as 0 for male and 1 for female Comparisons with a two-sided P-value < 0.05 were considered statistically significant for CRS and ICANS20. For all other analyses, multiple testing corrections were applied using the false discovery rate (FDR), and results with an FDR < 0.1 were considered statistically significant. The logistic regression model was used to estimate curated term enrichment and sex effect and adjust age effect. ROR comparing sexes were calculated from the logistic regression model. Analyses required more than 10 patients of each sex per toxicity event; comparisons not meeting this criterion were excluded. Differences in sex effects between CAR T and non–CAR T cohorts were tested using Wald tests. For stratification analysis, cases with specific cancer types or products were analyzed separately using the same multivariable logistic regression described above. For CRS analysis, cases reported with ‘Cytokine release syndrome’ were classified as CRS cases. For ICANS analysis, cases reported with ‘Immune effector cell-associated neurotoxicity syndrome’ were classified as ICANS cases. Analysis was performed in Jetstream2 server21,22. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Transparent Peer Review file Supplementary Information Reporting Summary Transparent Peer Review file Source data Source Data Source Data