Cross-study safety analysis of risk factors in CAR T cell clinical trials: An FDA database pilot project.

Introduction

Despite the approval of six chimeric antigen receptor (CAR) T cell products, cytokine release syndrome (CRS), a cytokine-mediated systemic inflammatory response, and neurological toxicities (NTX) continue to be life-threatening adverse events associated with CAR T cell administration. 1 , 2 , 3 , 4 , 5 , 6 , 7 The clinical presentation of CRS varies from mild flu-like symptoms to high fevers, sinus tachycardia, hypotension, hypoxia, depressed cardiac function, and other signs of organ dysfunction, while NTX presents as a cluster of neurological symptoms ranging from mild confusion, headaches, and hallucinations to aphasia, seizures, and somnolence. 8 CRS pathophysiology is postulated to be related to CAR T cell activation, expansion, and recruitment of other immune mediator cells, resulting in cascading cytokine production. Although the pathophysiology of NTX remains unknown, the leading hypothesis is a cytokine-mediated disruption of the blood-brain barrier. 9 Symptoms of CRS and NTX often occur concurrently, and groups such as the American Society for Transplantation and Cellular Therapy (ASTCT) have published guidelines to better harmonize the definitions and grading systems for these adverse events. 10 Treatment of CRS consists primarily of symptomatic treatment at lower toxicity grades but may involve the use of supplemental oxygen, high-dose vasopressor therapy, and tocilizumab (a humanized anti-human IL-6 receptor monoclonal antibody) with or without concomitant corticosteroid treatment for worsening CRS. 7 , 11 Treatment strategies for NTX vary among different centers but consist primarily of corticosteroid therapy with or without cytokine-directed therapies and supportive care.

Many groups have attempted to identify risk factors associated with the development of CRS and NTX after CAR T cell product administration. Teachey et al. 12 measured cytokines and clinical biomarkers in 51 subjects and found that peak levels of 24 cytokines (including IFN-ɣ, IL-6, sgp130, and sIL-6R) in the first month after infusion were highly associated with severe CRS. In addition, they developed multiple predictive models, including a logistic regression model, which used a combination of IFN-ɣ and CCL3 to predict which subjects would develop severe CRS (sensitivity 82%, specificity 93%). Hay et al. 13 performed multivariate analysis of baseline characteristics in 133 adult subjects who received CD19 CAR T cells and identified high marrow tumor burden, lymphodepletion using fludarabine and cyclophosphamide, higher CAR T cell dose, thrombocytopenia before lymphodepletion, and manufacturing of CAR T cells without selection of CD8 + central memory T cells as independent predictors of CRS. Brudno and Kochenderfer 14 performed a review of factors contributing to CRS and NTX and found that higher peak in vivo proliferation of CAR T cells, higher cell doses, conditioning chemotherapy containing fludarabine, acute lymphocytic leukemia (ALL) rather than non-Hodgkin’s lymphoma (NHL), higher burden of disease, baseline thrombocytopenia, and baseline elevated markers of endothelial activation (e.g., angiopoietin-2 and von Willebrand factor) were all risk factors for both CRS and NTX. In addition, they found that the CAR structure may contribute to patterns of toxicity. Tedesco and Mohan 15 performed a systematic review of 33 CAR T cell clinical trials to identify biomarkers predictive of post-treatment CRS and NTX and found that circulating IL-6, IFN-ɣ, IL-10, and IL-15 appear to be associated with the severity of CAR T cell therapy toxicities in both leukemia and lymphoma subjects. Greenbaum et al. 16 found that endothelial activation and stress index, a clinical surrogate for endothelial dysfunction, combined with ferritin and C-reactive protein was associated with both the incidence and severity of CRS and immune effector cell-associated neurotoxicity syndrome (ICANS) in subjects treated with CAR T cell products.

Although academic researchers have made important contributions to the current understanding of risk factors associated with CAR T cell-induced CRS and NTX, many groups are limited by either relatively small sample sizes or limited access to patient-level data. In order to address these limitations and increase the statistical power to detect risk factors associated with CAR T cell product administration, the FDA Center for Biologics Evaluation and Research (CBER), Office of Tissues and Advanced Therapies (OTAT), initiated a project to (1) assess the feasibility of integrating cross-study CAR T cell product safety data into a central CAR T cell safety database and (2) perform exploratory analyses to validate the use of cross-study data for risk factor identification and predictive modeling. 17 In this project, clinical safety and chemistry manufacturing and control (CMC) data from 17 studies were voluntarily provided by sponsors and incorporated into a CAR T cell safety database. Maximum CRS and NTX grades within 28 days of CAR T cell product administration were calculated for each subject, and rates of severe (toxicity grade ≥ 3) CRS and NTX were compared among various demographic, clinical, and manufacturing groups to identify risk factors associated with severe CRS and NTX after CAR T cell product administration.

Results

The CAR T cell safety database contains data for 1,926 subjects from 17 studies received prior to September 1, 2020. Six studies were phase 1, five were phase 2, and six were phases 1 and 2. Six studies contained primarily ALL subjects, five studies contained NHL subjects, and six studies contained multiple myeloma (MM) subjects. Four studies contained primarily pediatric subjects, while 13 studies contained primarily adult subjects.

A total of 1,277 subjects received at least one administration of a CAR T cell product, while the remaining 649 subjects did not receive treatment because of failure to meet the sponsor’s study requirements. Of the 1,277 subjects, 963 (75.4%) received CD19-targeting products, while 314 (24.6%) received B cell maturation antigen (BCMA)-targeting products. Median treated patient count per study was 61 (interquartile range [IQR]: 29–93) across all studies, 49 (IQR: 29–70) for ALL studies, 92 (IQR: 19–182) for NHL studies, and 47 (IQR: 18–79) for MM studies.

A summary table of group sizes, rates of severe CRS and NTX, and odds ratios comparing select variables from pooled analysis are shown in Figure 1 . Further results for indication, CRS management protocol, vector design, age group, cytokines, expansion in vivo , dosing parameters, and CMC characteristics are reported below. Figure 1 Pooled analysis for select clinical and manufacturing variables The first column lists select variables, while the second column lists the group size. For both severe cytokine release syndrome (sCRS) and severe neurological toxicities (sNTX), the “Count (%)” column gives the number of subjects who experienced sCRS/sNTX as well as the percentage relative to the group size. The “Odds Ratio (95% CI)” columns contain odds ratios and 95% confidence intervals calculated using the first variable in each group as a reference point (reference points: adult for age group, LGI for CRS intervention, NHL for indication, yes for prior transplantation, female for sex, BCMA for target antigen, and non-gCD28vec for vector design). Forest plots are displayed next to the “Odds Ratio (95% CI)” column to visually represent relative risk. The “P Value” column lists p values from chi-square tests comparing severe toxicity rates in each group. ALL, acute lymphocytic leukemia; NHL, non-Hodgkin’s lymphoma; MM, multiple myeloma; BCMA, B cell maturation antigen; gCD28vec, products produced with gammaretrovirus vectors with CD28 sequences in the transgene; LGI, low-grade intervention; HGI, high-grade intervention.

CRS

Within the first 28 days after CAR T cell administration, 807 subjects (63.2%) experienced CRS and 174 subjects (13.6%) experienced severe CRS (sCRS) (toxicity grade ≥ 3). Median onset time for the first sCRS was 3.9 days post-infusion.

Indication

Of 1,277 treated subjects, 304 (23.8%) had ALL, 659 (51.6%) had NHL, and 314 (24.6%) had multiple myeloma. Subtypes of NHL were not analyzed. ALL subjects had higher rates of sCRS than NHL subjects, while there was no statistically significant difference between NHL and MM subjects in the frequency of sCRS ( Figure 1 , “Indication”). ALL subjects had higher rates of sCRS than NHL subjects in all subgroups except in subjects with no prior transplantation ( Figure 2 , “Prior Transplant: No”). An inadequate pediatric NHL sample size (n = 4) did not allow comparisons among pediatrics subjects ( Figure 2 , “Age Group: Pediatric: NHL”). For products administered to both adult ALL and adult NHL subjects, ALL subjects had higher rates of sCRS ( Figure S1 A). There was no statistically significant difference in sCRS rates between NHL and MM subjects ( Figure 2 ), except in subjects with no prior transplantation ( Figure 2 , “Prior Transplant: No”). Figure 2 Analysis of indication within select subgroups Within each subgroup (as defined in the first column), rates of severe cytokine release syndrome (sCRS) and severe neurological toxicities (sNTX) were calculated for various indications (non-Hodgkin’s lymphoma [NHL], acute lymphocytic leukemia [ALL], and multiple myeloma [MM]) and presented in the “Count (%)” column. The “Odds Ratio (95% CI)” columns contain odds ratios and 95% confidence intervals calculated using NHL as a reference point within each subgroup. Forest plots are displayed next to the “Odds Ratio (95% CI)” column to visually represent relative risk. The “P Value” column lists p values from chi-square tests comparing severe toxicity rates in each group. gCD28vec, products produced with gammaretrovirus vectors with CD28 sequences in the transgene; LGI, low- grade intervention; HGI, high-grade intervention.

CRS management protocol

CRS management protocols could be divided into two groups on the basis of the timing of intervention with tocilizumab. In the high-grade intervention (HGI) group, intervention with tocilizumab was generally reserved for sCRS while in the low-grade intervention (LGI) group, tocilizumab intervention could also be used as needed for grade 1 or 2 CRS. Only the rate of sCRS (using maximum CRS grade per subject) was considered in our analysis, not progression of CRS from lower to higher grades. HGI subjects experienced greater rates of sCRS than LGI subjects ( Figure 1 , “CRS Intervention”). Subgroup analyses supported this finding in all groups except pediatrics ( Figure 3 , “Age: Pediatric”). Because almost all pediatric subjects had ALL (n = 177 of 181), results from this subgroup analysis may not be applicable in other indications. Figure 3 Analysis of CRS management protocol within select subgroups Within each subgroup (as defined in the first column), rates of severe cytokine release syndrome (sCRS) and severe neurological toxicities (sNTX) were calculated for the low-grade intervention (LGI) protocol and the high-grade intervention (HGI) protocol and presented in the “Count (%)” column. The “Odds Ratio (95% CI)” column contains odds ratios and 95% confidence intervals calculated using the LGI group as a reference point within each subgroup. Forest plots are displayed next to the “Odds Ratio (95% CI)” column to visually represent relative risk. The “P Value” column lists p values from chi-square tests comparing severe toxicity rates in each group. ALL, acute lymphocytic leukemia; NHL, non-Hodgkin’s lymphoma; MM, multiple myeloma; gCD28vec, products produced with gammaretrovirus vectors with CD28 sequences in the transgene.

Vector design

Four different combinations of vector type (gammaretrovirus or lentivirus) and costimulatory domain (4-1BB or CD28) were used across 17 studies to manufacture CAR T cell products ( Table S1 ). Although subjects receiving CAR T cells with gammaretrovirus vectors containing CD28 sequences (gCD28vec) had higher rates of sCRS ( Figure 1 , “Vector Design”), this result was inconclusive upon subgroup analysis ( Figure 4 ). Figure 4 Analysis of vector design within select subgroups Within each subgroup (as defined in the first column), rates of severe cytokine release syndrome (sCRS) and severe neurological toxicities (sNTX) were calculated for products produced with gammaretrovirus vectors with CD28 sequences in the transgene (gCD28vec) and products not produced with this vector design (non-gCD28vec) and presented in the “Count (%)” column. The “Odds Ratio (95% CI)” column contains odds ratios and 95% confidence intervals calculated using the non-gCD28vec group as a reference point within each subgroup. Forest plots are displayed next to the “Odds Ratio (95% CI)” column to visually represent relative risk. The “P Value” column lists p values from chi-square tests comparing severe toxicity rates in each group. ALL, acute lymphocytic leukemia; NHL, non-Hodgkin’s lymphoma; MM, multiple myeloma; LGI, low-grade intervention; HGI, high-grade intervention.

Age group

Although pediatric subjects experienced higher rates of sCRS compared with adults ( Figure 1 , “Age Group”), this difference did not persist upon indication and CRS management protocol subgroup analyses ( Figure S2 A). As there were few pediatric subjects with NHL or MM (n = 4 of 181), subgroup analysis of age by indication could not be performed for these indications. Among ALL subjects, there was no statistically significant difference in sCRS rates between adults (26.0% [n = 33 of 127]) and pediatrics (36.7% [n = 65 of 177]) (p = 0.06). Further subgroup analysis of ALL subjects by CRS management protocol found no statistically significant difference in sCRS rates between pediatrics and adults in either the LGI (28.6% [n = 10 of 35] versus 23.5% [n = 27 of 115], p = 0.70) or the HGI (38.7% [n = 55 of 142] versus 50.0% [n = 6 of 12], p = 0.54) groups.

Cytokines

Of 116 cytokines/biomarkers in the CAR T cell safety database, there were data for 17 (IFN-ɣ, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12, IL-13, IL-15, CCL2, CCL3, CCL4, GMCSF, and TNF-α) of these for at least 60% of treated subjects. Differences in cytokine levels between subjects who experienced sCRS and those who did not are reported in Tables 1 and S2 and Figure S3 . Table 1 Differences in cytokine levels for subjects with sCRS and sNTX Cytokine sCRS (36 h) sCRS (before CRS) sNTX (36 h) CCL2 higher (p < 0.001, n = 582) higher (p = 0.01, n = 787) higher (p < 0.001, n = 582) CCL3 NSD NSD lower (p < 0.001, n = 623) CCL4 NSD NSD higher (p = 0.040, n = 627) GMCSF NSD NSD NSD IFN-ɣ higher (p = 0.01, n = 925) higher (p = 0.03, n = 1,133) NSD IL-1β NSD NSD NSD IL-2 higher (p = 0.04, n = 913) NSD NSD IL-4 lower (p < 0.001, n = 764) lower (p < 0.001, n = 1,014) higher (p = 0.014, n = 764) IL-5 NSD NSD NSD IL-6 NSD NSD NSD IL-7 NSD NSD higher (p = 0.048, n = 600) IL-8 higher (p = 0.001, n = 925) NSD higher (p = 0.023, n = 925) IL-10 NSD NSD NSD IL-12 NSD lower (p < 0.001, n = 1,004) NSD IL-13 NSD NSD NSD IL-15 NSD NSD NSD TNF-α NSD NSD NSD Average maximum cytokines levels within 36 h after CAR T cell product administration in subjects with severe cytokine release syndrome (sCRS) and severe neurological toxicities (sNTX) compared with subjects with non-sCRS/non-sNTX. For subjects who experienced sCRS, average maximum cytokine concentrations were also calculated using all cytokine concentrations before CRS. NSD, no significant difference at p = 0.05.

Expansion in vivo

Higher peak CAR T cell expansion has been shown to be associated with CRS. 18 , 19 Expansion in vivo was evaluated as the peak level (Cmax) of CAR T cell transgene copies in blood within 28 days after CAR T cell product administration. Data were available for 8 of 17 studies and 661 of 1,277 treated subjects. Median maximum transgene concentration was greater in subjects who experienced sCRS (57,988 versus 19,260 transgene copies/μg DNA, p < 0.001). Median maximum transgene concentration was greatest for ALL subjects (50,679 transgene copies/μg DNA, n = 243), followed by NHL (15,870 transgene copies/μg DNA, n = 410) and MM (243 transgene copies/μg DNA, n = 8). Because of differences in measurement units, not all MM expansion data could be included in the analysis. Among ALL subjects, median maximum transgene concentration was higher in adult (62,532 transgene copies/μg DNA, n = 78) compared with pediatric (44,652 transgene copies/μg DNA, n = 243) subjects (p=0.004). However, in subgroup analysis, differences between adult and pediatric ALL subjects receiving products with the same costimulatory domain were not significant. Transgene concentration measurements during the first 36 h after CAR T cell administration were not available for most subjects and were therefore not examined.

Dosing parameters

Three dosing parameters (transduced cell count, transduction frequency, and total cell count) were analyzed for correlation with sCRS. Transduced cell count is the quantity of T cells in the final CAR T cell product that have been transduced with the CAR, transduction frequency is the ratio of transduced cells to total cells in the final CAR T cell product, and total cell count is the quantity of all cells in the final CAR T cell product. Although products within a study are manufactured to meet the sponsors’ specified transduced cell dose, transduction frequency may vary widely among products because of the transduction rate of the patient-specific lot.

There was no statistically significant difference in transduced cell counts between subjects who experienced sCRS versus non-sCRS ( Figure 5 Ai). Subjects who experienced sCRS received products with lower transduction frequencies ( Figure 5 Aii). Subgroup analysis by indication found that for subjects with ALL or NHL, but not MM, transduction frequencies were significantly lower among subjects who experienced sCRS ( Figure S4 ). Subgroup analysis indicated that for LGI subjects, transduction frequencies were lower for those who experienced sCRS, while there was no statistically significant difference among HGI subjects (results not shown). Although subjects who experienced sCRS received products with greater total cell counts ( Figure 5 Aiii), this may be driven by subjects in the HGI group (which reported higher rates of sCRS) who received products with greater total cell counts than LGI subjects (7.90 × 10 8 versus 1.92 × 10 8 , p < 0.001, n = 1,254). Subgroup analysis within CRS management protocols (LGI versus HGI) revealed no statistically significant difference in median total cell count between subject who experienced sCRS versus non-sCRS. Figure 5 Relationship of dose variables with sCRS and sNTX (Ai) There was no statistically significant difference in median transduced cell count for products administered to subjects with and without severe cytokine release syndrome (sCRS) (1.2 × 10 8 versus 1.3 × 10 8 , n = 1,165). (Aii) Subjects with sCRS received products with smaller median transduction frequency (37.9% versus 57.8%, n = 1,150). (Aiii) Subjects with sCRS received products with greater median total cell counts (3.7 × 10 8 versus 2.4 × 10 8 , n = 1,254). (Biv) Subjects with severe neurological toxicities (sNTX) received products with smaller median transduced cell counts (1.1 × 10 8 versus 1.3 × 10 8 , n = 1,165). (Bv) There was no statistically significant difference in median transduction frequency for products administered to subjects with versus without sNTX (56.1% versus 48.9%, n = 1,150). (Bvi) Subjects with sNTX received products with smaller median total cell counts (2.5 × 10 8 versus 2.6 × 10 8 , n = 1,254). Results were considered significant at p ≤ 0.05.

Percentage T cells

Percentage T cells was not a differentiating factor for sCRS ( Figure 6 Ai). Figure 6 Relationship of cell manufacturing and control variables with sCRS and sNTX (Ai) Percentage T cells was not a differentiating factor for severe cytokine release syndrome (sCRS) (99.0% versus 99.4%, n = 1,047). (Aii) Subjects with sCRS received products with greater percentage viable cells (92.3% versus 90.4%, n = 1,149). (Aiii) Rank-normalized vector copy number (VCN) was not a differentiating factor for sCRS (0.28 versus 0.33, n = 1,061). (Aiv) Rank-normalized interferon gamma (IFN-ɣ) was not a differentiating factor for sCRS (0.25 versus 0.22, n = 745). (Bv) Percentage T cells was not a differentiating factor for severe neurological toxicities (sNTX) (99.3% versus 99.3%, n = 1,047). (Bvi) Percentage viable cells was not a differentiating factor for sNTX (91.0% versus 90.9%, n = 1,149). (Bvii) Subjects with sNTX received products produced with lower rank-normalized VCN (0.29 versus 0.32, n = 1,061). (Bviii) Subjects with sNTX received products with greater rank-normalized IFN-ɣ (0.30 versus 0.20, n = 745). Results were considered significant at p ≤ 0.05.

Percentage viable cells

Although cell viability was >90% for both groups, subjects who experienced sCRS received products with higher percentages of viable cells ( Figure 6 Aii). Subgroup analysis by indication revealed that among NHL subjects, percentages of viable cells were higher among subjects who experienced sCRS (94.5% [n = 52 of 493] versus 89.3% [n = 493 of 545], p < 0.001), whereas there was no statistically significant difference among subjects with ALL (92.2% [n = 97 of 298] versus 90.0% [n = 201 of 298], p = 0.16) or MM (92.8% [n = 18 of 292] versus 91.3% [n = 274 of 292], p = 0.97). Subgroup analysis by CRS management protocol found that among LGI subjects, subjects who experienced sCRS received products with a higher percentage of viable cells compared with subjects who did not experience sCRS (92.5% [n = 84 of 886] versus 89.8% [n = 802 of 886], p = 0.002). Figure S5 provides additional analysis of the relationship between cell viability and sCRS for both indication and CRS management protocol.

Vector copy number

Because of different underlying assay methods and measurement units (e.g., per cell, per transduced cell, per amount of DNA), vector copy number (VCN) data were rank-normalized; VCN levels in each study were assigned a value between 0 and 1. VCN rank was not a differentiating factor for sCRS ( Figure 6 Aiii).

Potency

Because of different underlying assay methods and measurement units, product potency results (as measured by IFN-ɣ secretion assay) were rank-normalized in each study by assigning a value between 0 and 1. Product potency rank was not a differentiating factor for sCRS ( Figure 6 Aiv).

NTX

Within the first 28 days after CAR T cell administration, 851 subjects (66.6%) experienced NTX, and 237 subjects (18.6%) experienced severe NTX (sNTX) (toxicity grade ≥ 3). Median onset time for the first sNTX was 5.6 days post-infusion.

Indication

Compared with subjects with NHL, subjects with ALL had higher rates of sNTX, while subjects with MM had lower rates of sNTX ( Figure 1 , “Indication”). Subgroup analyses indicated that ALL subjects had increased risk of sNTX compared with NHL subjects in certain subgroups (adults, LGI CRS management protocol, no prior transplantation, males, and subjects receiving gCD28vec; Figure 2 ). There was no statistically significant difference in sNTX rate between adult ALL and NHL subjects administered the same CD19-targetting CAR T cell product ( Figure S1 B). MM subjects had lower rates of sNTX than NHL subjects ( Figure 2 ), except in subjects with no prior transplantation ( Figure 2 , “Prior Transplant: No”).

CRS management protocol

There was no statistically significant difference in rates of sNTX between HGI and LGI subjects ( Figure 1 , “CRS Intervention”). Subgroup analysis supported this finding in all groups except ALL ( Figure 3 , “Indication: ALL”) and non-gCD28vec ( Figure 3 , “Vector Design: non-gCD28vec”). In the ALL subgroup, this was an artifact of HGI subjects’ consisting mostly of pediatrics who had lower rates of sNTX than adults ( Figure S2 B). Lower sNTX rates in the non-gCD28vec subgroup may be an artifact of LGI subjects’ receiving a disproportionate number of BCMA products, which had lower rates of sNTX than CD19 products ( Figure 1 ). After controlling for age group and target antigen in the ALL group and non-CD28vec group, respectively, there were no statistically significant differences in sNTX rate between HGI and LGI subjects ( Figures S6 and S7 ).

Vector design

Subjects in the gCD28vec group had higher rates of sNTX ( Figure 1 , “Vector Design”). Higher rates of sNTX were maintained regardless of sex, prior transplant status, CRS management protocol, or indication ( Figure 4 ). Among adults, the gCD28vec group had higher rates of sNTX; while among pediatric subjects, there was no difference in sNTX rate between the gCD28vec and non-gCD28vec group ( Figure 4 , “Age Group”). As noted above, most pediatric subjects (n = 177 of 181) had underlying ALL, and the pediatric gCD28vec group was relatively small (n = 35), therefore we cannot rule out product- or indication-specific factors.

Age group

Although pooled analysis revealed no statistically significant difference in sNTX rates between adults and pediatrics ( Figure 1 , “Age Group”), subgroup analysis of ALL subjects indicated that pediatrics had lower rates of sNTX than adults ( Figure S2 B, “Indication: ALL”). Although some subgroups show pediatrics with higher rates of sNTX ( Figure S2 B, “Vector Design: non-gCD28vec”), this was secondary to indication as pediatrics primarily had ALL, which was associated with higher rates of sNTX ( Figures 1 and 2 ). Controlling for indication in the non-gCD28vec group revealed pediatrics associated with lower rates of sNTX (19.7% [n = 28 of 142] versus 41.7% [n = 5 of 12]).

Cytokines

Differences in cytokine levels between subjects who experienced sNTX and those who did not are reported in Tables 1 and S2 and Figure S8 .

Expansion in vivo

Median maximum transgene concentration was greater in subjects who experienced sNTX compared with those who did not (71,273 versus 19,997 transgene copies/μg DNA, p = 0.002).

Dosing parameters

Although subjects who experienced sNTX received products with lower transduced cell counts ( Figure 5 Biv), this finding did not persist in subgroup analysis by either indication or vector design. Transduction frequency was not a differentiating factor for sNTX ( Figure 5 Bv). Although subjects who experienced sNTX received products with lower total cell counts ( Figure 5 Bvi), this may be due to subjects in the gCD28vec group (which reported higher rates of sNTX) who received products with lower total cell counts than non-CD28vec subjects (2.36 × 10 8 versus 3.00 × 10 8 , p < 0.001, n = 1,254). Subgroup analysis by vector design revealed no statistically significant difference in median total cell count between sNTX and non-sNTX subjects in either the gCD28vec or non-gCD28vec groups.

Percentage T cells

Percentage T cells was not a differentiating factor for sNTX ( Figure 6 Bv).

Percentage viable cells

Percentage of viable cells was not a differentiating factor for sNTX ( Figure 6 Bvi).

Vector copy number

Subjects who experienced sNTX received products manufactured with a lower rank VCN compared with subjects who did not experience sNTX ( Figure 6 Bvii). In subgroup analysis by indication, ALL subjects with sNTX received products manufactured with a lower rank VCN (0.21 [n = 84 of 288] versus 0.29 [n = 204 of 288], p = 0.002), whereas there was no statistically significant difference in VCN rank for subjects with NHL (0.33 [n = 103 of 508] versus 0.34 [n = 405 of 508], p = 0.61) or MM (0.31 [n = 12 of 265] versus 0.33 [n = 253 of 265], p = 0.34).

Potency

Although subjects who experienced sNTX received higher rank potency products ( Figure 6 Bviii), this finding did not persist in subgroup analysis by indication for subjects with ALL (0.3 [n = 46 of 175] versus 0.26 [n = 129 of 175], p = 0.32), NHL (0.3 [n = 88 of 329] versus 0.24 [n = 241 of 329], p = 0.09), or MM (0.17 [n = 11 of 241] versus 0.14 [n = 230 of 214], p = 0.45).

Multivariate classification models

Multivariate models using data from all studies and all domains (clinical and CMC) were developed to identify which combination of factors show a strong association with the occurrence of sCRS or sNTX. Significant parameters for sCRS and sNTX are listed in Table 2 . Forward-selected logistic regression models typically outperformed other classification methods on validation datasets. The best performing logistic regression model predicted the occurrence of sCRS with an accuracy of 73% and sensitivity of 80%, while the best model for predicting sNTX had an accuracy of 68% and sensitivity of 62%. Table 2 Significant parameters for multivariate logistic regression models Predictors Odds ratio 95% CI p value CRS (Grade 3–5) Maximum temperature (36 h) 1.44 1.26–1.64 <0.001 Maximum IL-4 concentration (36 h) 0.93 0.88–0.97 0.004 ALL indication 3.01 1.69–5.49 <0.001 gCD28vec products 3.96 1.55–12.21 0.008 HGI CRS management protocol 5.35 1.95–17.27 0.002 NTX (Grade 3–5) Maximum temperature (36 h) 1.43 1.27–1.61 <0.001 MM indication 0.17 0.08–0.35 <0.001 gCD28vec products 2.93 1.90–4.56 <0.001 Multivariate models using data from all studies and all domains were developed to identify which combination of factors show a strong association with the occurrence of severe cytokine release syndrome (CRS) or neurological toxicities (NTX). Model parameters were selected using forward selection using cross validation scores from a logistic regression estimator and selected parameters were fit using logistic regression with L1 regularization. Odds ratios and associated p values for significant variables were obtained from a logistic regression estimator. For CRS, indication, CRS intervention, vector design, temperature, IL-4, IL-8, and TNF-α were selected during parameter selection step. Temperature, IL-4 concentration, acute lymphocytic leukemia (ALL) indication, products produced with gammaretrovirus vectors with CD28 sequences in the transgene domain (gCD28vec), and high-grade intervention (HGI) CRS management protocol had significant effects on risk for severe CRS. For NTX, indication, vector design, temperature, CCL2, IL-8, IL-1β, and potency (IFN-ɣ) were selected during parameter selection step. Temperature, multiple myeloma (MM) indication, and gCD28vec vector design had significant effects on risk for severe NTX.

Discussion

The first goal of this project was to assess the feasibility of integrating CAR T cell product data from multiple studies to enable cross-study data analysis. Although data formats varied among sponsors, the creation of a CAR T cell product-specific standard format and the development of data extraction, transformation, and loading tools allowed successful integration of cross-study data into a single database.

The second goal of this project was to perform exploratory analyses to validate the use of the integrated database for risk factor identification and predictive modeling. Although subgroup analysis and normalization methods were used to adjust for differences between studies, not all sources of confounding were able to be controlled for. As a result, risk factors identified in this exploratory analysis may not be directly related to sCRS/sNTX but instead may reflect fundamental differences in the included phase 1 and 2 trials. 20 , 21 , 22 , 23 In addition, as analyses were exploratory in nature, p values were not adjusted for multiple comparisons, further limiting the interpretation of results. 24 Therefore, findings should be interpreted as hypothesis generating and are not intended to support clinical or regulatory decisions.

Despite these limitations, several findings will be of interest to the CAR T cell clinical development community and may be considered when designing early phase dose-escalation cohorts. First, adults with ALL had a higher risk for developing both sCRS and sNTX compared with adults with either NHL or MM. Although evidence for the association of ALL with sNTX was not as strong as that for sCRS, it was still evident in multiple subgroup analyses. Although the association of ALL with sCRS and sNTX may be related to CAR T cell expansion in vivo (ALL subjects had substantially higher maximum transgene levels after administration), exploration of this hypothesis was beyond the scope of this analysis. A limitation of this analysis is that although disease burden was captured in our database, there was no meaningful way to standardize disease burden across different indications. As disease burden has been associated with both sCRS and sNTX, differences in underlying disease burden may have confounded this result. 13 , 14 , 25 Despite this limitation, our findings support work by others who have found that rates of sCRS are higher in patients with ALL than in those with NHL. 14 , 26 , 27 , 28 Thus, when considering the design of phase 1 CAR T cell therapy studies (particularly for CD19-targeted products), separate dose-escalation cohorts based on indication may be considered for determining maximum tolerable dose, which may be different among different indications with the same target.

Second, among subjects with ALL, pediatric subjects were less likely to experience sNTX and equally likely to experience sCRS as adult subjects. This finding supports work by others who have compared rates of sCRS and sNTX in adult versus pediatric ALL subjects. 29 , 30 Because of the limited number of pediatric subjects with NHL or MM in our database, it is unclear if these findings are applicable outside of ALL, although others have found age related associations with ICANS in adults with ALL and NHL. 31 , 32 , 33 , 34

Third, studies using LGI with tocilizumab had lower rates of sCRS, with no difference in rates of sNTX. This finding supports the paradigm shift in CRS management that has been observed when comparing earlier studies with more recent ones, with the use of cytokine-directed therapies and corticosteroids at lower grades of toxicity leading to a significant reduction in sCRS with no evident impact on ICANS. 30 , 35 , 36 , 37 , 38 , 39 These findings suggest different risk mitigation strategies may be needed for CRS and ICANS.

This study reports several CAR T cell CMC findings. First, CAR T cells produced with gammaretrovirus vectors that included CD28 sequences in the CAR design appeared to increase the risk for sNTX. This finding corroborates other who have noted that subjects receiving CAR T cell products with CD28 domains have experienced higher levels of sNTX. 14 , 30 , 40 Because of the lack of structural variety in products produced using gammaretrovirus vectors, it was not possible to determine if the increased risk for sNTX was a result of vector type or CAR structure ( Table S1 ). As vector design was typically uniform within a clinical study, the possibility of confounding factors unique to studies using this vector design cannot be ruled out.

Second, lower transduction frequency in the CAR T cell product was significantly associated with higher rates of sCRS and this relationship was observed in subjects grouped by both indication and study. This correlation could relate to the relationship between higher total cell counts and increased sCRS incidence, as lower transduction frequency would necessitate administration of a higher total number of cells to meet dose. This finding suggests that untransduced cells or cellular traits that contribute to lower transduction frequency could be a sCRS risk factor. Although others have found associations between T cell subset composition and toxicity, because neither the cellular composition of untransduced cells nor the T cell subsets were evaluated by most sponsors, we could not determine if particular cell types were associated with higher rates of sCRS. 13 , 41 , 42 , 43 , 44 , 45

There were multiple challenges associated with interpreting results from this cross-study data analysis. The primary challenge was how to properly assess the trade-off between achieving greater statistical power for risk factor detection (by increasing the number of eligible subjects in the analysis) with the increase in potential confounders when combining datasets with different clinical and CMC profiles. Differences in the severity and frequency of CRS and NTX among subjects from different clinical and CMC subgroups (such as subjects with different cancers or different product profiles) may also have caused the analysis results to be confounded by factors unique to these subgroups. In addition, pertinent baseline characteristics such as tumor burden and inclusion/exclusion criteria across studies was not captured in our analysis and limits interpretation of results.

Although we tried to account for some sources of confounding by using subgroup analyses, differences in prognostic factors and limitations in the quantity of reported data between studies made it impossible to perform comparisons in a completely standardized way. Although disease burden data were sometimes available for leukemias, lymphomas, which require specifications for imaging techniques in addition to the standardized PET/CT-based imaging routinely used to evaluate response to therapy and disease status at study entry, were not available for most studies. Even if included, it would be difficult to interpret the results of an analysis evaluating the impact of disease burden on safety outcomes within and across indications, as differences in imaging parameters used to determine disease burden may also confound these results. Despite the limitations from confounding and missing data, this is the largest analysis of pooled data across CAR T cell products and results from this exploratory analysis can help researchers identify areas to test in more rigorously controlled populations.

A second major challenge was incorporating differences in grading and management strategies between studies. For CRS, differences in the Penn, Lee, and ASTCT criteria not only influence CRS toxicity grade but also the timing of tocilizumab administration, further complicating cross-study comparison. 10 , 46 , 47 , 48 , 49 , 50 Multiple steps were taken to control for and investigate how differences in grading systems affected our findings. Subgroup analyses were performed within studies using the same grading systems to test the robustness of pooled results. We also repeated our analysis with sCRS defined as toxicity grade ≥ 4 (the grade at which different grading criteria converge in assessment) and found the primary findings of our analysis remained consistent (results not shown). We also investigated the algorithmic conversion between grading criteria and again found the primary findings of our analysis remained consistent (results not shown). These complementary approaches lend validity to our exploratory findings, despite variations in the CRS grading systems.

For NTX, outcomes were standardized using Medical Dictionary for Regulatory Activities (MedDRA) system organ classes. Despite not taking a more selective approach for classifying neurological events (e.g., using the ICANS definition), almost all studies used the same version of the Common Terminology Criteria for Adverse Events (CTCAE), and rates of sNTX were consistent with listed rates in the package insert of marketed products. 51 , 52 , 53 , 54 , 55 Therefore this standardization strategy allowed accurate comparisons between studies in this exploratory analysis, although we do not recommend this approach for use in clinical trial design or decision making.

This study showed that although it is feasible to combine cross-study safety data for analysis, data standardization and differences in study design complicate analysis and interpretation of results. Although standardization efforts in the field may facilitate future efforts to assess safety risks across products, CAR T cell products are still an evolving therapeutic class and adherence to specified standards may not be feasible or desirable at this time. Although this study focuses specifically on risk factors associated with CRS and NTX, future analyses could consider additional safety factors and emerging toxicities (e.g., prolonged cytopenia, macrophage activation syndrome, or hemophagocytic lymphohistiocytosis). 56 , 57 Findings from these exploratory analyses offer researchers and stakeholders in the CAR T cell clinical development community information on broader safety trends, considerations for early phase trial design, and avenues for future research.

Materials and methods

Proprietary and confidential clinical safety and manufacturing data from 17 early to late phase studies received prior to September 1, 2020, were voluntarily provided by investigational new drug (IND) application sponsors for incorporation into the CAR T cell safety database. 17 Studies were selected by asking sponsors with relatively large numbers of patients in CAR T cell trials to voluntarily contribute data for cross-study safety analysis. Such data, unlike submissions for licensing applications, were not cleaned or locked. Demographics, adverse events, concomitant medications, serum cytokine profiles, dosing regimen, cellular kinetic profiles, and CMC information were included in the CAR T cell safety database. Most data were provided in SAS transport files and followed the Clinical Data Interchange Standards Consortium (CDISC) Study Data Tabulation Model (SDTM). 58 All data voluntarily submitted for this project were included in the analysis.

To integrate cross-study data into a standardized database, a custom data schema was designed based on the CDISC SDTM format. Software tools were developed to transform sponsor-submitted data to meet the requirements of the integrated database. As data were received, the initial data schema was updated to incorporate additional CAR T cell-specific information (e.g., cell population metrics). Reference tables were also developed to standardize terminology and reduce inconsistencies in the database. Additional information on data standardization ( Notes S1–S5 ), as well as the final schema for the CAR T cell safety database ( Appendix I ), can be found in the supplemental information .

Occurrence of sCRS or sNTX within 28 days after CAR T cell product administration were the primary safety outcomes of interest. A time window of 28 days after product administration was selected to focus our analysis on short-term toxicities likely due to product administration and not related to disease progression. Subjects were classified as experiencing sCRS if they experienced at least one occurrence of CRS at toxicity grade ≥ 3 within 28 days after CAR T cell product administration. CRS toxicity grades were not reviewed for accuracy or adherence to the sponsor’s stated grading criteria and were accepted as reported. NTX was broadly defined as the occurrence of one or more adverse events under the neurologic disorders or psychiatric disorders Medical Dictionary for Regulatory Activities 20.1 system organ class. 59 Subjects were classified as experiencing sNTX if they experienced at least one occurrence of NTX at toxicity grade ≥ 3 within 28 days after CAR T cell product administration. For subjects who had multiple episodes of CRS or NTX, only the maximum CRS or NTX toxicity grade was used for analysis. The most common adverse events used to define both NTX and sNTX are listed in Tables S3 and S4 .

After integration of cross-study safety data, risk factors analysis was performed separately for sCRS and sNTX. Risk factors analyzed included indication (ALL versus NHL versus MM), CRS management protocol (HGI versus LGI; Note S6 ), vector design (gCD28vec versus non-gCD28vec), age group (adult versus pediatric; Note S7 ), cytokines (maximum concentration in blood within 36 h post-CAR T cell product infusion), expansion in vivo (maximum concentration of CAR T cell transgene copies in blood within 28 days post-CAR T cell product infusion), dosing parameters (transduced cell count, transduction frequency, and total cell count), and CMC characteristics (percentage T cells, percentage viable cells, vector copy number, and potency).

Bivariate analysis of continuous data was performed by splitting subjects into two groups on the basis of adverse event severity (i.e., severe versus non-severe) and using t tests to determine significant differences between groups (p ≤ 0.05). Bivariate analysis of categorical data used the chi-square test to test for differences in the percentages of subjects with sCRS or sNTX in different subgroups (p ≤ 0.05). For select variables, odds ratios and 95% confidence intervals were calculated to evaluate comparative risk. Initial bivariate analysis was followed by subgroup analyses to rule out potential confounding factors. For cases in which integration of numerical data from across studies was not possible, because of differences in units or underlying assay methods (e.g., if two studies used different target cells for IFN-ɣ secretion assays), data were either rank-normalized to values between 0 and 1 within each study or dropped from analysis. Despite the potential for different underlying assay methods, cytokine data were not rank normalized, as comparisons of cytokine distributions between studies were comparable after converting all measurements to picograms per milliliter.

Multivariate models such as logistic regression, decision trees, and random forests were developed to identify which combination of factors show a strong association with the occurrence of sCRS or sNTX. Inputs to these models included demographics data (including indication), CMC data (IFN-ɣ, VCN, cellular composition data, and vector design data), dose data (total cell count, transduced cell count, percentage transduction, total cell count per kilogram, and transduced cell count per kg), CRS management protocol, maximum cytokine concentrations within 36 h after CAR T cell administration, and maximum temperature within 36 h after CAR T cell administration ( Note S8 ). All multivariate models were developed using training data and validated on data not used during model training. As cytokine data were typically submitted in a timeseries format with sparse measurements at irregular time points, cytokine data were interpolated to be compatible with multivariate analysis algorithms ( Note S9 ).

Data availability

The data that support the findings of this study are not openly available. Confidentiality agreements with sponsors who voluntarily provided clinical safety and manufacturing for this project prohibit data sharing.

References

1 U.S. Food & Drug Administration FDA Approves Ciltacabtagene Autoleucel for Relapsed or Refractory Multiple Myeloma 2022 https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-ciltacabtagene-autoleucel-relapsed-or-refractory-multiple-myeloma

2 U.S. Food & Drug Administration FDA Approves Brexucabtagene Autoleucel for Relapsed or Refractory Mantle Cell Lymphoma 2020 https://www.fda.gov/drugs/fda-approves-brexucabtagene-autoleucel-relapsed-or-refractory-mantle-cell-lymphoma

3 U.S. Food & Drug Administration FDA Approves Lisocabtagene Maraleucel for Relapsed or Refractory Large B-Cell Lymphoma 2021 https://www.fda.gov/drugs/drug-approvals-and-databases/fda-approves-lisocabtagene-maraleucel-relapsed-or-refractory-large-b-cell-lymphoma

4 U.S. Food & Drug Administration FDA Approves Idecabtagene Vicleucel for Multiple Myeloma 2021 https://www.fda.gov/drugs/drug-approvals-and-databases/fda-approves-idecabtagene-vicleucel-multiple-myeloma

5 U.S. Food & Drug Administration FDA Approval Brings First Gene Therapy to the United States 2017 https://www.fda.gov/news-events/press-announcements/fda-approval-brings-first-gene-therapy-united-states

6 U.S. Food & Drug Administration FDA Approves Axicabtagene Ciloleucel for Large B-Cell Lymphoma 2017 https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-axicabtagene-ciloleucel-large-b-cell-lymphoma

7 Murthy H. Iqbal M. Chavez J.C. Kharfan-Dabaja M.A. Cytokine release syndrome: current perspectives Immunotargets Ther. 8 2019 43 52 10.2147/ITT.S202015 31754614 PMC6825470

8 Shimabukuro-Vornhagen A. Gödel P. Subklewe M. Stemmler H.J. Schlößer H.A. Schlaak M. Kochanek M. Böll B. von Bergwelt-Baildon M.S. Cytokine release syndrome J. Immunother. Cancer 6 2018 56 10.1186/s40425-018-0343-9 29907163 PMC6003181

9 Gust J. Hay K.A. Hanafi L.A. Li D. Myerson D. Gonzalez-Cuyar L.F. Yeung C. Liles W.C. Wurfel M. Lopez J.A. Endothelial activation and blood-brain barrier disruption in neurotoxicity after adoptive immunotherapy with CD19 CAR-T cells Cancer Discov. 7 2017 1404 1419 10.1158/2159-8290.Cd-17-0698 29025771 PMC5718945

10 Lee D.W. Santomasso B.D. Locke F.L. Ghobadi A. Turtle C.J. Brudno J.N. Maus M.V. Park J.H. Mead E. Pavletic S. ASTCT consensus grading for cytokine release syndrome and neurologic toxicity associated with immune effector cells Biol. Blood Marrow Transpl. 25 2019 625 638 10.1016/j.bbmt.2018.12.758 PMC12180426 30592986

11 Chou C.K. Turtle C.J. Assessment and management of cytokine release syndrome and neurotoxicity following CD19 CAR-T cell therapy Expert Opin. Biol. Ther. 20 2020 653 664 10.1080/14712598.2020.1729735 32067497 PMC7393694

12 Teachey D.T. Lacey S.F. Shaw P.A. Melenhorst J.J. Maude S.L. Frey N. Pequignot E. Gonzalez V.E. Chen F. Finklestein J. Identification of predictive biomarkers for cytokine release syndrome after chimeric antigen receptor T-cell therapy for acute lymphoblastic leukemia Cancer Discov. 6 2016 664 679 10.1158/2159-8290.Cd-16-0040 27076371 PMC5448406

13 Hay K.A. Hanafi L.A. Li D. Gust J. Liles W.C. Wurfel M.M. López J.A. Chen J. Chung D. Harju-Baker S. Kinetics and biomarkers of severe cytokine release syndrome after CD19 chimeric antigen receptor-modified T-cell therapy Blood 130 2017 2295 2306 10.1182/blood-2017-06-793141 28924019 PMC5701525

14 Brudno J.N. Kochenderfer J.N. Recent advances in CAR T-cell toxicity: mechanisms, manifestations and management Blood Rev. 34 2019 45 55 10.1016/j.blre.2018.11.002 30528964 PMC6628697

15 Tedesco V.E. 5th Mohan C. Biomarkers for predicting cytokine release syndrome following CD19-targeted CAR T cell therapy J. Immunol. 206 2021 1561 1568 10.4049/jimmunol.2001249 33692146

16 Greenbaum U. Strati P. Saliba R.M. Torres J. Rondon G. Nieto Y. Hosing C. Srour S.A. Westin J. Fayad L.E. CRP and ferritin in addition to the EASIX score predict CAR-T–related toxicity Blood Adv. 5 2021 2799 2806 10.1182/bloodadvances.2021004575 34264268 PMC8341350

17 Thompson Reuters. FDA Proposes New Databases to Monitor CAR T-Cell Safety across INDs (US). Practical Law Life Sciences W-001-6501. https://uk.practicallaw.thomsonreuters.com/w-001-6501 .

18 Wei J. Liu Y. Wang C. Zhang Y. Tong C. Dai G. Wang W. Rasko J.E.J. Melenhorst J.J. Qian W. The model of cytokine release syndrome in CAR T-cell treatment for B-cell non-Hodgkin lymphoma Signal Transduct. Target. Ther. 5 2020 134 10.1038/s41392-020-00256-x 32728035 PMC7388484

19 Wang Z. Han W. Biomarkers of cytokine release syndrome and neurotoxicity related to CAR-T cell therapy Biomark. Res. 6 2018 4 10.1186/s40364-018-0116-0 29387417 PMC5778792

20 Kim H. Gurrin L. Ademi Z. Liew D. Overview of methods for comparing the efficacies of drugs in the absence of head-to-head clinical trial data Br. J. Clin. Pharmacol. 77 2014 116 121 10.1111/bcp.12150 23617453 PMC3895352

21 Greenland S. Quantitative methods in the review of epidemiologic literature Epidemiol. Rev. 9 1987 1 30 10.1093/oxfordjournals.epirev.a036298 3678409

22 Haidich A.B. Meta-analysis in medical research Hippokratia 14 Suppl 1 2010 29 37 21487488 PMC3049418

23 Tierney J.F. Stewart L.A. Clarke M. on behalf of the Cochrane Individual Participant Data Meta-analysis Methods Group Individual participant data Cochrane Handbook Syst. Rev. Interventions 2019 643 658

24 Jafari M. Ansari-Pour N. Why, when and how to adjust your P values? Cell J. 20 2019 604 607 10.22074/cellj.2019.5992 30124010 PMC6099145

25 Chavez J.C. Bachmeier C. Kharfan-Dabaja M.A. CAR T-cell therapy for B-cell lymphomas: clinical trial results of available products Ther. Adv. Hematol. 10 2019 10.1177/2040620719841581 2040620719841581 PMC6466472 31019670

26 Teachey D.T. Bishop M.R. Maloney D.G. Grupp S.A. Toxicity management after chimeric antigen receptor T cell therapy: one size does not fit 'ALL Nat. Rev. Clin. Oncol. 15 2018 218 10.1038/nrclinonc.2018.19 29434335

27 Maude S.L. Laetsch T.W. Buechner J. Rives S. Boyer M. Bittencourt H. Bader P. Verneris M.R. Stefanski H.E. Myers G.D. Tisagenlecleucel in children and young adults with B-cell lymphoblastic leukemia N. Engl. J. Med. 378 2018 439 448 10.1056/NEJMoa1709866 29385370 PMC5996391

28 Schuster S.J. Bishop M.R. Tam C.S. Waller E.K. Borchmann P. McGuirk J.P. Jäger U. Jaglowski S. Andreadis C. Westin J.R. Tisagenlecleucel in adult relapsed or refractory diffuse large B-cell lymphoma N. Engl. J. Med. 380 2019 45 56 30501490 10.1056/NEJMoa1804980

29 Anagnostou T. Riaz I.B. Hashmi S.K. Murad M.H. Kenderian S.S. CD19 directed chimeric antigen receptor T cell therapy in acute lymphoblastic leukemia: a systematic review and meta-analysis Biol. Blood Marrow Transplant. 25 2019 S169 S170 10.1016/j.bbmt.2018.12.308

30 Gauthier J. Turtle C.J. Chimeric antigen receptor T-cell therapy for B-cell acute lymphoblastic leukemia: current landscape in 2021 Cancer J. 27 2021 98 106 10.1097/ppo.0000000000000508 33750068

31 Grant S.J. Grimshaw A.A. Silberstein J. Murdaugh D. Wildes T.M. Rosko A.E. Giri S. Clinical presentation, risk factors, and outcomes of immune effector cell-associated neurotoxicity syndrome following chimeric antigen receptor T cell therapy: a systematic review Transpl. Cell. Ther. 28 2022 294 302 10.1016/j.jtct.2022.03.006 PMC9197870 35288347

32 Santomasso B.D. Park J.H. Salloum D. Riviere I. Flynn J. Mead E. Halton E. Wang X. Senechal B. Purdon T. Clinical and biological correlates of neurotoxicity associated with CAR T-cell therapy in patients with B-cell acute lymphoblastic leukemia Cancer Discov. 8 2018 958 971 10.1158/2159-8290.Cd-17-1319 29880584 PMC6385599

33 Rubin D.B. Al Jarrah A. Li K. LaRose S. Monk A.D. Ali A.B. Spendley L.N. Nikiforow S. Jacobson C. Vaitkevicius H. Clinical predictors of neurotoxicity after chimeric antigen receptor T-cell therapy JAMA Neurol. 77 2020 1536 1542 10.1001/jamaneurol.2020.2703 32777012 PMC7418044

34 Gauthier J. Cearley A. Perkins P. Kirk A. Shadman M. Williamson S. Myers J. Chen A.I. Nagle S. Hayes-Lattin B.M. CD19 CAR T-cell product type independently impacts CRS and ICANS severity in patients with aggressive NHL J. Clin. Oncol. 39 2021 7532 10.1200/JCO.2021.39.15_suppl.7532

35 Caimi P.F. Pacheco Sanchez G. Sharma A. Otegbeye F. Ahmed N. Rojas P. Patel S. Kleinsorge Block S. Schiavone J. Zamborsky K. Prophylactic tocilizumab prior to anti-CD19 CAR-T cell therapy for non-hodgkin lymphoma Front. Immunol. 12 2021 745320 10.3389/fimmu.2021.745320 34712233 PMC8546323

36 Oluwole O.O. Bouabdallah K. Muñoz J. De Guibert S. Vose J.M. Bartlett N.L. Lin Y. Deol A. McSweeney P.A. Goy A.H. Prophylactic corticosteroid use in patients receiving axicabtagene ciloleucel for large B-cell lymphoma Br. J. Haematol. 194 2021 690 700 10.1111/bjh.17527 34296427 PMC8457222

37 Gardner R. Leger K.J. Annesley C.E. Summers C. Rivers J. Gust J. Tarlock K. Cooper T.M. Pinto N.R. Finney O. Decreased rates of severe CRS seen with early intervention strategies for CD19 CAR-T cell toxicity management Blood 128 2016 586 10.1182/blood.V128.22.586.586

38 Gardner R.A. Ceppi F. Rivers J. Annesley C. Summers C. Taraseviciute A. Gust J. Leger K.J. Tarlock K. Cooper T.M. Preemptive mitigation of CD19 CAR T-cell cytokine release syndrome without attenuation of antileukemic efficacy Blood 134 2019 2149 2158 10.1182/blood.2019001463 31697826 PMC6908832

39 Kadauke S. Maude S. Gladney W. Motley L. Shenoy V. Callahan C. Baniewicz D. Teachey D.T. Grupp S. DiNofia A. Early administration of tocilizumab (Toci) for the prevention of grade 4 cytokine release syndrome (CRS) after CD19-directed CAR T-cell therapy (CTL019) Cytotherapy 21 2019 e2 e3 10.1016/j.jcyt.2019.04.009

40 Cappell K.M. Kochenderfer J.N. A comparison of chimeric antigen receptors containing CD28 versus 4-1BB costimulatory domains Nat. Rev. Clin. Oncol. 18 2021 715 727 10.1038/s41571-021-00530-z 34230645

41 Sommermeyer D. Hudecek M. Kosasih P.L. Gogishvili T. Maloney D.G. Turtle C.J. Riddell S.R. Chimeric antigen receptor-modified T cells derived from defined CD8+ and CD4+ subsets confer superior antitumor reactivity in vivo Leukemia 30 2016 492 500 10.1038/leu.2015.247 26369987 PMC4746098

42 Singh N. Perazzelli J. Grupp S.A. Barrett D.M. Early memory phenotypes drive T cell proliferation in patients with pediatric malignancies Sci. Transl. Med. 8 2016 320ra3 10.1126/scitranslmed.aad5222 26738796

43 Turtle C.J. Hanafi L.A. Berger C. Hudecek M. Pender B. Robinson E. Hawkins R. Chaney C. Cherian S. Chen X. Immunotherapy of non-Hodgkin's lymphoma with a defined ratio of CD8+ and CD4+ CD19-specific chimeric antigen receptor-modified T cells Sci. Transl. Med. 8 2016 355ra116 10.1126/scitranslmed.aaf8621 PMC5045301 27605551

44 Shah N.N. Highfill S.L. Shalabi H. Yates B. Jin J. Wolters P.L. Ombrello A. Steinberg S.M. Martin S. Delbrook C. CD4/CD8 T-Cell selection affects chimeric antigen receptor (CAR) T-cell potency and toxicity: updated results from a phase I anti-CD22 CAR T-cell trial J. Clin. Oncol. 38 2020 1938 1950 10.1200/jco.19.03279 32286905 PMC7280047

45 Arcangeli S. Bove C. Mezzanotte C. Camisa B. Falcone L. Manfredi F. Bezzecchi E. El Khoury R. Norata R. Sanvito F. CAR T-cell manufacturing from naive/stem memory T-lymphocytes enhances antitumor responses while curtailing cytokine release syndrome J. Clin. Invest. 132 2022 e150807 10.1172/jci150807 35503659 PMC9197529

46 Lee D.W. Gardner R. Porter D.L. Louis C.U. Ahmed N. Jensen M. Grupp S.A. Mackall C.L. Current concepts in the diagnosis and management of cytokine release syndrome Blood 124 2014 188 195 10.1182/blood-2014-05-552729 24876563 PMC4093680

47 Porter D. Frey N. Wood P.A. Weng Y. Grupp S.A. Grading of cytokine release syndrome associated with the CAR T cell therapy tisagenlecleucel J. Hematol. Oncol. 11 2018 35 10.1186/s13045-018-0571-y 29895316 PMC5998449

48 Locke F.L. Ghobadi A. Jacobson C.A. Miklos D.B. Lekakis L.J. Oluwole O.O. Lin Y. Braunschweig I. Hill B.T. Timmerman J.M. Long-term safety and activity of axicabtagene ciloleucel in refractory large B-cell lymphoma (ZUMA-1): a single-arm, multicentre, phase 1-2 trial Lancet Oncol. 20 2019 31 42 10.1016/s1470-2045(18)30864-7 30518502 PMC6733402

49 Schuster S.J. Maziarz R.T. Rusch E.S. Li J. Signorovitch J.E. Romanov V.V. Locke F.L. Maloney D.G. Grading and management of cytokine release syndrome in patients treated with tisagenlecleucel in the JULIET trial Blood Adv. 4 2020 1432 1439 10.1182/bloodadvances.2019001304 32271899 PMC7160283

50 Schuster S.J. Bishop M.R. Tam C.S. Waller E.K. Borchmann P. McGuirk J.P. Jäger U. Jaglowski S. Andreadis C. Westin J.R. Tisagenlecleucel in adult relapsed or refractory diffuse large B-cell lymphoma N. Engl. J. Med. 380 2019 45 56 10.1056/NEJMoa1804980 30501490

51 Kite Pharma Incorporated YESCARTA®(axicabtagene Ciloleucel) https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=9b70606e-b99c-4272-a0f1-b5523cce0c59

52 Novartis Pharmaceuticals Corporation KYMRIAH (Tisagenlecleucel) https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=aad3ba54-dfd3-4cb3-9e2b-c5ef89559189

53 Corporation C. ABECMA®(idecabtagene Vicleucel) https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=b90c1fe7-f5cc-464e-958a-af36e9c26d7c

54 Juno Therapeutics Incorporated BREYANZI® (lisocabtagene maraleucel) https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=594bb413-af3b-4b97-afb3-bfe2b174f2ed

55 Kite Pharma Incorporated TECARTUS™(brexucabtagene Autoleuce) https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a16108c2-7ca7-45af-965e-54bda4713022

56 Martín-Rojas R.M. Gómez-Centurión I. Bailén R. Bastos M. Diaz-Crespo F. Carbonell D. Correa-Rocha R. Pion M. Muñoz C. Sancho M. Hemophagocytic lymphohistiocytosis/macrophage activation syndrome (HLH/MAS) following treatment with tisagenlecleucel Clin. Case Rep. 10 2022 e05209 10.1002/ccr3.5209 35028140 PMC8741874

57 Sandler R.D. Tattersall R.S. Schoemans H. Greco R. Badoglio M. Labopin M. Alexander T. Kirgizov K. Rovira M. Saif M. Diagnosis and management of secondary HLH/MAS following HSCT and CAR-T cell therapy in adults; A review of the literature and a survey of practice within EBMT centres on behalf of the autoimmune diseases working party (ADWP) and transplant complications working party (TCWP) Front. Immunol. 11 2020 524 10.3389/fimmu.2020.00524 32296434 PMC7137396

58 National Cancer Institute CDISC terminology https://datascience.cancer.gov/resources/cancer-vocabulary/cdisc-terminology

59 MedDRA MedDRA Hierarchy https://www.meddra.org/how-to-use/basics/hierarchy

Supplemental information

Document S1. Notes S1–S9, Tables S1–S4, Figures S1–S8, and Appendix I Document S2. Article plus supplemental information

Acknowledgments

We would like to thank all the sponsors and patients who contributed data for this project, this project would not have been possible without their contributions. We would also like to thank Nirjal Bhattarai, Arya Eskandarian, Richard Forshee, Chao Liu, Donna Monk, Graeme Price, John Scott, Mark Walderhaug, Wei Wang, and Yaning Wang, who provided feedback and expertise at various stages throughout the project. CBER/FDA provided funding for these studies as part of its intramural research program. The graphical abstract was created using BioRender.com .

Author contributions

Conceptualization, B.G., K.S., X.W., and W.W.B.; Methodology, Y.N. and M.F.; Software, Y.N. and M.F.; Validation; Y.N. and M.F.; Formal Analysis, Y.N.; Data Curation, Y.N. and M.F.; Writing – Original Draft, M.F.; Writing – Review and Editing, M.F., Y.N., L.E., W.W.B., X.W., Y.X., K.S., and B.G.; Visualization, Y.N. and Y.X.; Supervision, B.G., K.S., X.W., and W.W.B.; Project Administration, L.E.; Funding Acquisition, B.G. and W.W.B.

Declaration of interests

The authors declare no competing interests.

Supplemental information can be found online at https://doi.org/10.1016/j.omto.2022.10.006 .