Early neurologic response predicts outcomes after anakinra for immune effector cell–associated neurotoxicity syndrome

Key Points

• •Outcomes after anakinra treatment for ICANS are suboptimal: 72-hour SNI rate of 43%, median resolution of 8 days, and 28-day TRM of 13%.
• •Achievement of 72-hour SNI after anakinra initiation predicted faster ICANS resolution, decreased corticosteroid exposure, and lower TRM.

Visual Abstract

Abstract

Anakinra is increasingly used for corticosteroid-refractory and/or severe immune effector cell–associated neurotoxicity syndrome (ICANS) following chimeric antigen receptor (CAR) T-cell therapy; however, robust data on its efficacy and predictors of response are lacking. We evaluated the outcomes of 101 patients treated with anakinra for ICANS (corticosteroid-refractory, n = 90) and investigated factors associated with anakinra efficacy. The median time to ICANS resolution from anakinra initiation was 8 days, and the 28-day cumulative incidences of ICANS resolution and treatment-related mortality (TRM) were 86% and 13%, respectively. Anakinra treatment failure occurred in 28%. The day +28 antitumor response rate was 91% (complete response, 47%). Significant neurologic improvement (SNI; ≥2-grade improvement in ICANS) occurred in 43% of patients within 72 hours after anakinra initiation. Achieving 72-hour SNI was associated with faster time to ICANS resolution (median, 3 vs 9 days; P < .001) and hospital discharge (28-day cumulative incidence, 93% vs 53%; P < .001), lower 2-week cumulative exposure to dexamethasone (120 vs 190 mg; P = .029) and anakinra (3650 vs 6600 mg; P = .059), lower TRM (28-day cumulative incidence, 0% vs 21%; P = .011), and a trend toward superior overall survival (28 days, 98% vs 74%; P = .094). In multivariable analysis, older age, CAR T-cell product type, and higher day 0 C-reactive protein were independently associated with lower odds of 72-hour SNI. Our study benchmarks key clinical outcomes after anakinra treatment for ICANS; 72-hour SNI may provide a practical clinical decision point to identify patients who may require additional treatment strategies.

Introduction

Immune effector cell–associated neurotoxicity syndrome (ICANS) is a potentially life-threatening complication observed in up to 60% of patients receiving chimeric antigen receptor (CAR) T-cell therapy.1, 2, 3, 4, 5 As the indications for CAR T-cell therapy expand across hematologic malignancies and solid tumors, ICANS remains a barrier to optimal outcomes and broader access to CAR T-cell therapy. Dexamethasone remains the cornerstone of ICANS management; however, ∼30% to 50% of patients with grade ≥2 ICANS require additional therapies.^6,7

The recombinant interleukin-1 receptor antagonist anakinra has emerged as a promising second-line option to treat corticosteroid-refractory and/or severe ICANS. Although anakinra may be effective in this setting8, 9, 10, 11, 12, 13, 14 and expert recommendations support its use,¹⁵ the outcomes of patients receiving anakinra to treat ICANS have not been well characterized to date. Moreover, in the absence of validated early response criteria, clinicians lack a framework to determine whether anakinra is effective, futile, or requires escalation to alternative therapies. To address this important knowledge gap, we conducted a retrospective multicenter study of 101 patients treated with anakinra for ICANS after CAR T-cell therapy. We comprehensively evaluated the efficacy of anakinra across multiple clinical end points and sought to identify the key predictors of outcomes. In addition, we evaluated whether early neurologic improvement could serve as a clinically actionable decision point during ICANS management.

Methods

Study design

We included all consecutive patients treated with anakinra for ICANS after CAR T-cell therapy for hematologic malignancies between 2017 and 2025 at Fred Hutchinson Cancer Center (FHCC: July 2017-December 2024) and Oregon Health & Science University (OHSU: December 2020-September 2024). Patient, disease, and treatment characteristics were extracted from the electronic medical record. This study was approved by the institutional review boards at FHCC and OHSU and conducted in accordance with the Declaration of Helsinki.

ICANS management at FHCC and OHSU

ICANS was treated according to FHCC and OHSU standard practice guidelines during the study period. At both institutions, frontline management of grade ≥2 ICANS consisted of dexamethasone administered IV at 10 mg every 6 to 12 hours, which could be escalated to methylprednisolone IV at 1 g/d for up to 3 days in the absence of neurologic improvement after dexamethasone initiation. Anakinra was recommended to be administered IV at 8 to 10 mg/kg per day in 3 divided doses (FHCC) or 200 mg every 8 hours (OHSU) for corticosteroid-refractory ICANS and/or for severe (grade ≥3) ICANS.^13,16,17 For patients at high risk, anakinra could be initiated earlier or for grade 1 to 2 ICANS at the treating physician’s discretion.

Grading and end point definitions

In the absence of available immune effector cell-associated encephalopathy scores before their implementation for patients who received infusions from 2017 to 2019 (n = 2), we graded ICANS per the National Cancer Institute Common Terminology Criteria for Adverse Events version 5.0. We applied the American Society for Transplantation and Cellular Therapy criteria¹⁸ for the remaining patients who received infusions between 2019 and 2025. We defined baseline ICANS grade as the grade at the time of anakinra initiation. The day of ICANS resolution was defined as the first day of grade 0 ICANS, stable for at least 3 consecutive days, as described previously.¹³

To assess the efficacy of anakinra at early time points after treatment initiation, we defined significant neurologic improvement (SNI) as a ≥2-grade improvement in ICANS grade from baseline in the absence of subsequent ICANS-directed therapy and death from any cause, thus reflecting a clinically meaningful improvement beyond interobserver variability and minor fluctuations in neurologic status. Anakinra treatment failure was defined by the initiation of a subsequent ICANS-directed therapy after anakinra initiation. Concurrent initiation of anakinra and escalation from dexamethasone to methylprednisolone was not considered treatment failure. We assessed CAR T-cell antitumor efficacy using the Lugano classification, International Myeloma Working Group criteria, and National Comprehensive Cancer Network criteria for patients with non-Hodgkin lymphoma, multiple myeloma, and acute lymphoblastic leukemia, respectively.

Statistical analyses

Median follow-up was estimated using the reverse Kaplan-Meier method, and overall survival (OS) using the Kaplan-Meier method. We compared changes between the baseline ICANS grade and the ICANS grades at 24, 48, and 72 hours after anakinra initiation using paired Wilcoxon signed-rank tests with continuity correction. We estimated the cumulative incidences of ICANS resolution and hospital discharge from anakinra initiation, with death from any cause as the competing risk; patients who received subsequent ICANS-directed therapies after anakinra initiation were not censored. We also estimated the cumulative incidence of treatment-related mortality (TRM) from the time of anakinra initiation, with death from progressive disease as the competing risk.

The 72-hour time point for SNI was selected to reflect a clinically relevant interval at which treatment escalation or alternative interventions are often considered in clinical practice. Comparisons of time-to-event outcomes by 72-hour SNI status were landmarked at 72 hours after anakinra initiation and performed using the cumulative incidence functions, accounting for the respective competing risks, and assessed using the Gray test. Comparisons of cumulative dexamethasone and anakinra exposure by 72-hour SNI status were performed using the Wilcoxon rank-sum test.

Across 30 patient, disease, treatment, and laboratory variables, we modeled associations with 72-hour SNI as a binary outcome using Firth penalized logistic regression¹⁹ to address small-sample bias, separation, and convergence limitations; and modeled associations with time to ICANS resolution using cause-specific Cox regression, landmarked at the time of anakinra initiation and censoring for death.²⁰ For each end point, a modified random forest model based on the approach of Altmann et al²¹ was applied. In this approach, empirical P values for variable importance were calculated by permuting the outcome 1000 times to generate a null distribution of importance for each variable. The 3 variables with the smallest permutation-based P value (corresponding to the highest variable importance in our data) were selected for inclusion in multivariable analysis. We performed all analyses using R software version 4.4.1 (R Core Team, Vienna, Austria).

Results

Patient, disease, and treatment characteristics

A total of 101 patients (FHCC, n = 64; OHSU, n = 37) received anakinra for ICANS (Table 1). The median age was 65 years (interquartile range [IQR], 57-71). The most common disease types were large B-cell lymphoma (n = 49 [49%]), mantle cell lymphoma (n = 22 [22%]), and acute lymphoblastic leukemia (n = 12 [12%]). The most common CAR T-cell products were axicabtagene ciloleucel (n = 28 [28%]), brexucabtagene autoleucel (n = 21 [21%]), and tisagenlecleucel (n = 20 [20%]). The median lactate dehydrogenase (LDH) and ferritin on day 0 before CAR T-cell infusion were 231.5 U/L (range, 108-3447) and 903.5 ng/mL (23-17 473), respectively. Baseline characteristics were similar across patients at FHCC and OHSU.

Table 1.

Patient, disease, and treatment characteristics

Characteristic	Overall, N = 101	FHCC, n = 64	OHSU, n = 37	P value∗
Age				.39
Median (IQR)	65.0 (57.4-71.0)	65.1 (55.9-71.1)	65.0 (58.8-70.6)
Range	23.8-83.0	28.0-78.1	23.8-83.0
Sex, n (%)				.33
Male	62 (61)	37 (58)	25 (68)
Female	39 (39)	27 (42)	12 (32)
Disease type, n (%)				.89
Large B-cell lymphoma	49 (49)	31 (48)	18 (49)
Mantle cell lymphoma	22 (22)	12 (19)	10 (27)
Acute lymphoblastic leukemia	12 (12)	7 (11)	5 (14)
Multiple myeloma	11 (11)	8 (13)	3 (8)
Follicular lymphoma	4 (4)	3 (5)	1 (3)
Chronic lymphocytic leukemia	2 (2)	2 (3)	0 (0)
Primary CNS lymphoma	1 (1)	1 (2)	0 (0)
CAR T-cell product, n (%)
Axi-cel	28 (28)	18 (28)	10 (27)
Brexu-cel	21 (21)	16 (25)	5 (14)
Tisa-cel	20 (20)	2 (3)	18 (49)
Investigational CAR T-cell product	14 (14)	14 (22)†	0 (0)
Liso-cel	9 (9)	8 (13)	1 (3)
Ide-cel	6 (6)	3 (5)	3 (8)
Cilta-cel	3 (3)	3 (5)	0 (0)
Costimulatory domain, n (%)				.15
4-1BB	49 (49)	27 (42)	22 (59)
CD28	49 (49)	34 (53)	15 (41)
CD28 and 4-1BB	3 (3)	3 (5)	0 (0)
Day 0 LDH (U/L)				.091
Median (IQR)	231.5 (165.0-391.0)	216.5 (152.0-358.0)	239.0 (199.5-392.5)
Range	108.0-3 447.0	112.0-2 300.0	108.0-3 447.0
Missing	3	2	1
Day 0 CRP (mg/L)				.17
Median (IQR)	33.7 (12.0-82.8)	30.2 (10.2-71.0)	36.0 (18.0-114.0)
Range	1.5-362.0	2.2-354.9	1.5-362.0
Missing	3	3	0
Day 0 ferritin (ng/mL)				.72
Median (IQR)	903.5 (265.0-2 340.0)	1 065.0 (270.0-2 340.0)	736.0 (253.0-2 039.0)
Range	23.0-17 473.0	77.0-17 473.0	23.0-14 547.0
Missing	3	3	0
Day +28 response, n (%)‡				.21
CR	37 (42)	22 (40)	15 (45)
PR	35 (40)	21 (38)	14 (42)
Not evaluable	9 (10)	6 (11)	3 (9)
PD	6 (7)	6 (11)	0 (0)
SD	1 (1)	0 (0)	1 (3)
Not performed	13	9	4

Axi-cel, axicabtagene ciloleucel; BCMA, B-cell maturation antigen; brexu-cel, brexucabtagene autoleucel; cilta-cel, ciltacabtagene autoleucel; CNS, central nervous system; CR, complete response; ide-cel, idecabtagene vicleucel; liso-cel, lisocabtagene maraleucel; PD, progressive disease; PR, partial response; SD, stable disease; tisa-cel, tisagenlecleucel.

^∗Continuous variables were compared using the Wilcoxon rank-sum test. Proportions were compared using the Pearson χ² test or the Fisher exact test. P values were not calculated for CAR T-cell product due to sparse counts and 0 values across multiple categories.

^†Targets: CD19, n = 9; CD20, n = 3; BCMA, n = 2.

^‡Percentages for response categories are among patients with response assessment performed (n = 88); ORR and CR rate in the text are among evaluable patients (n = 79).

In 79 patients evaluable for disease assessment, the day +28 overall antitumor response rate was 91% (complete response, 47%). With a median follow-up of 20.7 months (IQR, 15.1-35.4), the median OS was 19.5 months (95% confidence interval [CI], 9.63 to not reached).

Indications for anakinra and ICANS treatment characteristics

ICANS occurred at a median of 5 days (IQR, 4-7; range, 0-21) after CAR T-cell infusion, and the median initial ICANS grade at onset was 1 (IQR, 1-2; range, 1-4). The median peak ICANS grade at any time was 3 (IQR, 3-4; range, 1-5), occurring at a median of 7 days (IQR, 5-11; range, 2-45) after CAR T-cell infusion. First ICANS-directed treatment was initiated at a median of 0 days (IQR, 0-0; range, 0-4) after ICANS onset. The median grade at the time of first ICANS-directed treatment initiation was 2 (IQR, 1-3; range, 1-4).

Anakinra was initiated at a median of 1 day (IQR, 1-3; range, 0-35) after ICANS onset. The median ICANS grade at anakinra initiation was 3 (IQR, 2-3; range, 1-4). Peak ICANS occurred before, or at the time of, anakinra initiation in 84 patients (83%), and after anakinra initiation in 17 patients (17%). Detailed treatment patterns of the cohort are shown in supplemental Figure 1. Initial anakinra dosages and routes of administration are shown in supplemental Table 1.

The initial route of anakinra administration was IV in 96 patients (95%) and subcutaneous in 5 (5%). Overall, 90 patients (89%) received anakinra for persistent or worsening ICANS after frontline corticosteroid therapy (corticosteroid-refractory group): dexamethasone-refractory ICANS, n = 72 (72%); dexamethasone- and methylprednisolone-refractory ICANS, n = 16 (16%); and methylprednisolone-refractory ICANS, n = 2 (2%). The remaining 11 patients (11%) received frontline anakinra with or without corticosteroids for grade ≥2 ICANS.

The median number of tocilizumab administrations was 2 (IQR, 1-3). The total cumulative doses of dexamethasone and methylprednisolone were 240 mg (IQR, 169-341) and 3000 mg (IQR, 3000-3000), respectively. In 90 patients with corticosteroid-refractory ICANS, we observed 4 patterns of treatment escalation: anakinra added to dexamethasone (n = 53), anakinra added to methylprednisolone (n = 16), anakinra and methylprednisolone initiated concurrently after dexamethasone failure (n = 19), and initiation of anakinra with discontinuation of corticosteroids (n = 2).

Efficacy end points and short-term outcomes

The median time to ICANS resolution from anakinra initiation was 8 days. The 28-day cumulative incidence of ICANS resolution from anakinra initiation was 86% (95% CI, 78-92; Figure 1A). The 28-day cumulative incidence of TRM from anakinra initiation was 13% (95% CI, 7-20; Figure 1B) and the 28-day OS from anakinra initiation was 86% (95% CI, 80-93). In 13 patients who died of TRM within 28 days of anakinra initiation, 10 (77%) had ongoing ICANS at the time of death.

Figure 1.

Outcomes after anakinra initiation for ICANS. Cumulative incidence of ICANS resolution (A) and TRM (B) in all patients. Cumulative incidence of ICANS resolution (C) and TRM (D) in patients with corticosteroid-refractory ICANS. Shaded bands represent 95% CIs.

The median change in ICANS grade at 24, 48, and 72 hours after anakinra initiation was 0 (IQR, −1 to 0; range, −3 to 3), −1 (IQR, −2 to 0; range, −4 to 3), and −1.5 (IQR, −2 to −1; range, −4 to 2), respectively (Supplemental Figure 2). ICANS grades at each of these intervals after anakinra initiation were significantly lower than at baseline (24 hours: V = 491 [P < .001]; 48 hours: V = 530 [P < .001]; 72 hours: V = 338 [P < .001]).

Overall, 28 patients (28%) experienced anakinra treatment failure. Subsequent therapies administered after anakinra failure included methylprednisolone (n = 23), intrathecal therapy (n = 9), siltuximab (n = 2), and cetuximab (n = 1); specific regimens are listed in supplemental Table 2. A total of 12 patients died with ongoing ICANS; causes of death in these patients were infection (n = 4, all of whom had received methylprednisolone for ICANS), ICANS (n = 4), cytokine release syndrome (CRS; n = 1), CRS and ICANS (n = 1), progressive disease (n = 1), and acute heart and kidney failure (n = 1).

Despite standard practice guidelines in place at our institutions, residual heterogeneity in practice patterns may still have affected outcomes; therefore, we performed a subgroup analysis restricted to the patients treated with anakinra for corticosteroid-refractory ICANS (n = 90). In this subgroup, the median time to ICANS resolution was 8 days, and the 28-day cumulative incidence of ICANS resolution was 86% (95% CI, 76-91; Figure 1C). Anakinra treatment failure occurred in 25 patients (27%). The 28-day cumulative incidence of TRM was 12% (95% CI, 7-20; Figure 1D).

SNI is an early surrogate end point of ICANS outcomes

To date, early clinical end points to assess treatment efficacy on ICANS are lacking. To address this gap, we evaluated whether an improvement in ICANS of at least 2 grades (SNI) was associated with clinical outcomes. SNI is an easily measurable binary end point that can be assessed at early time points after anakinra initiation. Among 95 total patients evaluable for SNI assessment (nonevaluable: n = 5, grade 1 ICANS at anakinra initiation; n = 1, death due to grade 5 CRS within 24 hours of anakinra initiation), the rates of SNI at 24, 48, and 72 hours after anakinra initiation were 16% (n = 15), 31% (n = 29), and 43% (n = 41), respectively. Among 85 evaluable patients who received anakinra for corticosteroid-refractory ICANS, the rates of SNI at 24, 48, and 72 hours after anakinra initiation were 16% (n = 14), 29% (n = 25), and 44% (n = 37), respectively.

The 28-day cumulative incidence of ICANS resolution after anakinra initiation was significantly higher in patients with vs without 72-hour SNI (97% [95% CI, 67-100] vs 81% [95% CI, 67-90], Gray P < .001; median time to resolution, 3 vs 9 days; Figure 2A). SNI at 72 hours was associated with a higher 28-day cumulative incidence of hospital discharge after anakinra initiation (93% [95% CI, 77-98] vs 53% [95% CI, 38-65]; Gray P < .001; Figure 2B). At 28 days after anakinra initiation, no patient with 72-hour SNI experienced TRM, whereas the cumulative incidence of TRM in patients without 72-hour SNI was 21% (95% CI, 11-33; Gray P = .011; Figure 2C). We observed superior OS at 28 days after anakinra initiation in patients with 72-hour SNI than those without 72-hour SNI (98% [95% CI, 93-100] vs 74% [95% CI, 63-86]), although this was not statistically significant (P = .094; supplemental Figure 3).

Figure 2.

Outcomes by 72-hour SNI status. Cumulative incidence of ICANS resolution (A), hospital discharge (B), and TRM (C) by 72-hour SNI status. Cause-specific Gray test P values are shown. Analyses were landmarked at 72 hours after anakinra initiation. Shaded bands represent 95% CIs. (D) Cumulative dexamethasone exposure at 2 weeks after anakinra initiation by 72-hour SNI status. Boxes indicate IQRs, center lines indicate medians, whiskers indicate values within 1.5 times the IQR, and points indicate individual patients. Dexamethasone data were available for patients at FHCC only. PD, progressive disease.

At 2 weeks after anakinra initiation, patients without 72-hour SNI had significantly higher median cumulative exposure to dexamethasone (190 vs 120 mg; P = .029; Figure 2D) and numerically higher median cumulative exposure to anakinra (6600 vs 3650 mg; P = .059). Thus, the absence of 72-hour SNI identified a subgroup with persistently poor outcomes despite continued corticosteroid and anakinra exposure, suggesting that lack of early neurologic improvement may represent a clinically meaningful signal of treatment resistance. A proposed schema illustrating 72-hour SNI as a clinical decision point is shown in Figure 3.

Figure 3.

Proposed schema highlighting 72-hour SNI as a potential clinical decision point after anakinra initiation.

Predictors of ICANS outcomes

To further risk-stratify patients and inform future treatment practices, we applied univariate regression to identify predictors of response to anakinra as measured by 72-hour SNI (Table 2) and time to ICANS resolution (Table 3). First, in analyses of disease and treatment characteristics, multiple myeloma and treatment with a commercial B-cell maturation antigen product or an investigational CAR T-cell product were significantly associated with lower odds of 72-hour SNI.

Table 2.

Univariate Firth penalized logistic regression models for 72-hour SNI

Characteristic∗	N†	Event N	OR	95% CI	P value
Center	95	41
FHCC			—	—
OHSU			1.30	0.57-2.98	.53
Age	95	41	0.97	0.94-1.00	.10
Sex	95	41
Female			—	—
Male			0.92	0.40-2.11	.84
Disease type	95	41
Non-Hodgkin lymphoma			—	—
Acute lymphoblastic leukemia			0.92	0.26-3.15	.89
Multiple myeloma			0.15	0.02-0.71	.014
CAR T-cell product	95	41
Axi-cel/brexu-cel			—	—
Cilta-cel/ide-cel			0.17	0.02-0.84	.029
Investigational products			0.11	0.01-0.54	.004
Liso-cel/tisa-cel			1.26	0.50-3.26	.62
Costimulatory domain	95	41
4-1BB			—	—
CD28			1.72	0.76-3.94	.19
CD28 and 4-1BB			0.33	<0.01-4.35	.44
Day 0 LDH (log10 U/L)	92	40	0.20	0.04-0.79	.020
Day 0 platelet count (log10 × 103/μL)	59	24	2.75	0.67-13.5	.17
Platelet count at ICANS onset (log10 × 103/μL)	53	20	1.96	0.56-7.51	.30
Platelet count at anakinra initiation (log10 × 103/μL)	43	16	3.73	0.85-20.5	.081
Day 0 CRP (log10 mg/L)	92	40	0.63	0.29-1.33	.22
CRP at ICANS onset (log10 mg/L)	51	19	0.73	0.24-2.11	.55
CRP at anakinra initiation (log10 mg/L)	41	14	0.86	0.30-2.43	.78
Day 0 ferritin (log10 ng/mL)	92	40	0.42	0.19-0.87	.018
Ferritin at ICANS onset (log10 ng/mL)	51	19	0.60	0.21-1.61	.31
Ferritin at anakinra initiation (log10 ng/mL)	40	14	0.39	0.12-1.03	.058
Day 0 fibrinogen (log10 mg/dL)	59	24	3.43	0.18-84.9	.42
Fibrinogen at ICANS onset (log10 mg/dL)	53	20	1.11	0.09-15.3	.93
Fibrinogen at anakinra initiation (log10 mg/dL)	43	15	4.86	0.46-67.1	.19
Day 0 D-dimer (log10 ng/mL FEU)	59	24	0.58	0.19-1.64	.31
D-dimer at ICANS onset (log10 ng/mL FEU)	51	20	0.81	0.22-2.88	.75
D-dimer at anakinra initiation (log10 ng/mL FEU)	37	15	0.44	0.09-1.73	.24
ICANS grade at first ICANS-directed treatment initiation	95	41	0.82	0.50-1.34	.44
ICANS grade at anakinra initiation	95	41	1.02	0.54-1.92	.95
Days from ICANS onset to anakinra initiation	95	41	0.93	0.79-1.02	.13
Total dexamethasone dose from ICANS onset to anakinra initiation (mg)	59	24	0.98	0.96-1.00	.075
Receipt of methylprednisolone between ICANS onset and anakinra initiation	95	41	1.46	0.59-3.58	.41
Total number of tocilizumab doses from ICANS onset to anakinra initiation	32	11	0.79	0.23-2.02	.63
Treatment group‡	94	41
Anakinra added to dexamethasone			—	—
Anakinra added to methylprednisolone			1.31	0.42-4.10	.64
Concurrent anakinra + methylprednisolone after dexamethasone failure			1.64	0.58-4.74	.35
Up-front anakinra ± steroids for severe ICANS			1.03	0.26-3.85	.97
Anakinra dose§	78	35
200 mg IV every 8 h			—	—
300-400 mg IV every 8 h			0.27	0.06-0.90	.032

Bold indicates P value < 0.05.

Axi-cel, axicabtagene ciloleucel; brexu-cel, brexucabtagene autoleucel; cilta-cel, ciltacabtagene autoleucel; FEU, fibrinogen equivalent units; ide-cel, idecabtagene vicleucel; liso-cel, lisocabtagene maraleucel; tisa-cel, tisagenlecleucel.

^∗Daily dexamethasone, anakinra, and tocilizumab administrations; day 0 platelet count, fibrinogen, and D-dimer; and laboratory values at the time of ICANS onset and at the time of anakinra initiation were available for patients at FHCC only.

^†Excluding 5 patients with grade 1 ICANS at anakinra initiation and 1 patient who died within 24 hours of anakinra initiation.

^‡Excluding 2 patients who had corticosteroids discontinued at the time of anakinra initiation due to low number of events.

^§Limited to patients who received anakinra 200 to 400 mg IV every 8 hours.

Table 3.

Univariate cause-specific Cox models of time to ICANS resolution from anakinra initiation

Characteristic∗	N	Event N	HR	95% CI	P value
Center	101	89
FHCC			—	—
OHSU			0.92	0.59-1.43	.71
Age	101	89	0.99	0.97-1.00	.13
Sex	101	89
Female			—	—
Male			1.20	0.78-1.87	.41
Disease type	101	89
Non-Hodgkin lymphoma			—	—
Acute lymphoblastic leukemia			1.03	0.53-2.02	.92
Multiple myeloma			0.64	0.31-1.34	.24
CAR T-cell product	101	89
Axi-cel/brexu-cel			—	—
Cilta-cel/ide-cel			0.63	0.28-1.39	.25
Investigational products			0.61	0.30-1.25	.18
Liso-cel/tisa-cel			0.96	0.59-1.56	.87
Costimulatory domain	101	89
4-1BB			—	—
CD28			1.24	0.81-1.90	.32
CD28 and 4-1BB			0.72	0.17-3.00	.65
Day 0 LDH (log10 U/L)	98	86	0.39	0.18-0.84	.016
Day 0 platelet count (log10 × 103/μL)	64	56	3.02	1.39-6.57	.005
Platelet count at ICANS onset (log10 × 103/μL)	57	49	1.78	0.90-3.52	.10
Platelet count at anakinra initiation (log10 × 103/μL)	47	41	2.94	1.42-6.08	.004
Day 0 CRP (log10 mg/L)	98	87	0.64	0.42-0.97	.037
CRP at ICANS onset (log10 mg/L)	55	48	0.71	0.41-1.23	.22
CRP at anakinra initiation (log10 mg/L)	45	39	0.80	0.48-1.32	.39
Day 0 ferritin (log10 ng/mL)	98	87	0.66	0.49-0.89	.007
Ferritin at ICANS onset (log10 ng/mL)	55	48	0.54	0.32-0.92	.024
Ferritin at anakinra initiation (log10 ng/mL)	44	38	0.49	0.30-0.79	.003
Day 0 fibrinogen (log10 mg/dL)	64	56	2.71	0.50-14.5	.25
Fibrinogen at ICANS onset (log10 mg/dL)	57	49	1.04	0.25-4.29	.96
Fibrinogen at anakinra initiation (log10 mg/dL)	48	41	1.32	0.43-4.08	.63
Day 0 D-dimer (log10 ng/mL FEU)	64	56	0.59	0.32-1.08	.089
D-dimer at ICANS onset (log10 ng/mL FEU)	54	47	0.90	0.45-1.81	.77
D-dimer at anakinra initiation (log10 ng/mL FEU)	40	35	0.80	0.39-1.64	.54
ICANS grade at first ICANS-directed treatment initiation	101	89	0.81	0.63-1.05	.11
ICANS grade at anakinra initiation	101	89	0.88	0.66-1.16	.36
Days from ICANS onset to anakinra initiation	101	89	0.95	0.89-1.02	.19
Total dexamethasone dose from ICANS onset to anakinra initiation (mg)	63	55	1.00	0.99-1.00	.20
Receipt of methylprednisolone between ICANS onset and anakinra initiation	101	89	1.26	0.77-2.06	.36
Total number of tocilizumab doses from ICANS onset to anakinra initiation	35	30	1.09	0.59-1.99	.79
Treatment group†	99	87
Anakinra added to dexamethasone			—	—
Anakinra added to methylprednisolone			1.16	0.62-2.18	.63
Concurrent anakinra + methylprednisolone after dexamethasone failure			1.40	0.80-2.45	.24
Up-front anakinra ± steroids for severe ICANS			1.09	0.55-2.17	.81
Anakinra dose‡	82	75
200 mg IV every 8 h			—	—
300-400 mg IV every 8 h			0.75	0.42-1.32	.31

Bold indicates P value < 0.05.

Axi-cel, axicabtagene ciloleucel; brexu-cel, brexucabtagene autoleucel; cilta-cel, ciltacabtagene autoleucel; FEU, fibrinogen equivalent units; ide-cel, idecabtagene vicleucel; liso-cel, lisocabtagene maraleucel; tisa-cel, tisagenlecleucel.

^∗Daily dexamethasone, anakinra, and tocilizumab administrations; day 0 platelet count, fibrinogen, and D-dimer; and laboratory values at the time of ICANS onset and at the time of anakinra initiation were available for patients at FHCC only.

^†Excluding 2 patients who had corticosteroids discontinued at the time of anakinra initiation due to low number of events.

^‡Limited to patients who received anakinra 200 to 400 mg IV every 8 hours.

Because disease burden, systemic inflammation, and coagulopathy are known to be associated with severe ICANS,22, 23, 24, 25, 26, 27 we evaluated associations between response to anakinra and such biomarkers at baseline, at ICANS onset, and at anakinra initiation. We found that higher day 0 LDH and higher day 0 ferritin were significantly associated with lower odds of 72-hour SNI. Higher day 0 LDH; higher day 0 C-reactive protein (CRP); lower platelet count at day 0 and anakinra initiation; and higher ferritin at day 0, ICANS onset, and anakinra initiation were significantly associated with slower ICANS resolution. Lower platelet count at ICANS onset was also associated with slower ICANS resolution, although this did not reach statistical significance (P = .10).

In random forest models, the top 3 variables most frequently ranked highest by variable importance and thus included in multivariable models were age, day 0 CRP, and CAR T-cell product for the odds of 72-hour SNI (supplemental Figure 4A); and day 0 ferritin, day 0 LDH, and time from ICANS onset to anakinra initiation for the hazard of ICANS resolution (supplemental Figure 4B). In the multivariable model of 72-hour SNI, older age (odds ratio [OR], 0.96 per year increase; 95% CI, 0.92-0.99; P = .023), receipt of a commercial B-cell maturation antigen-targeted CAR T-cell product (OR, 0.06; 95% CI, <0.01-0.59; P = .011) or investigational CAR T-cell product (OR, 0.06; 95% CI, 0.01-0.32; P < .001), and higher day 0 CRP (OR, 0.32 per log₁₀ mg/L increase; 95% CI, 0.11-0.81; P = .016) remained independently associated with lower odds of 72-hour SNI. In the multivariable model of ICANS resolution, none of the 3 covariates remained independently associated with time to ICANS resolution (day 0 LDH: hazard ratio [HR], 0.63 per log₁₀ U/L increase; 95% CI, 0.26-1.56; P = .32; day 0 ferritin: HR, 0.75 per log₁₀ng/mL increase; 95% CI, 0.52-1.09; P = .14; and time from ICANS onset to anakinra initiation: HR, 0.97 per day increase; 95% CI, 0.90-1.04; P = .35).

Impact of ICANS-directed treatment patterns on outcomes

Next, we evaluated the impact of anakinra dose on ICANS outcomes in patients who received 200 to 400 mg IV every 8 hours (n = 82 overall; n = 78 evaluable for 72-hour SNI). We observed that patients who received the higher dose of 300 to 400 mg every 8 hours had a significantly lower odds of achieving 72-hour SNI (Table 2). In contrast, we could not confirm an association between anakinra dose and time to ICANS resolution (Table 3).

We investigated whether differences in preceding or concurrent corticosteroid treatment (ie, type and dosing) could have confounded the observed effects of anakinra on 72-hour SNI (Table 2) and time to ICANS resolution (Table 3). Given poor model convergence with cumulative methylprednisolone dose from ICANS onset to anakinra initiation, receipt of methylprednisolone between ICANS onset and anakinra initiation was modeled as a binary variable. Univariate regression models could not confirm an association between outcomes and receipt of methylprednisolone for ICANS before anakinra initiation. Higher cumulative dexamethasone dose from ICANS onset to anakinra initiation was associated with reduced odds of 72-hour SNI (OR, 0.98 per mg increase; 95% CI, 0.96-1.00; P = .075), although this did not reach statistical significance, and was not associated with ICANS resolution. Notably, the simultaneous escalation of dexamethasone to methylprednisolone and addition of anakinra did not appear to affect 72-hour SNI or ICANS resolution compared with the addition of anakinra to dexamethasone.

An association between tocilizumab treatment and higher ICANS risk has been suggested based on results from the ZUMA-1 cohort 3, which investigated the use of prophylactic tocilizumab in patients receiving axicabtagene ciloleucel.²⁸ Thus, we sought to test the hypothesis that variations in tocilizumab usage could have confounded our findings. Univariate analyses could not confirm associations between the number of tocilizumab doses from ICANS onset to anakinra initiation and odds of 72-hour SNI or time to ICANS resolution.

Discussion

Anakinra has emerged as a safe and promising treatment for corticosteroid-refractory and/or severe ICANS, yet robust data on treatment patterns and clinical outcomes in this patient population are lacking. In this multicenter retrospective study, we comprehensively benchmarked treatment characteristics and clinical outcomes in a large cohort of patients who received anakinra primarily as second-line treatment for corticosteroid-refractory, moderate to severe (ie, grade ≥2) ICANS. Despite the early initiation of anakinra after ICANS onset (median time to initiation, 1 day [IQR, 1-3]) at high doses, ICANS remained prolonged, with a 72-hour SNI rate of only 43% and a median time to ICANS resolution of 8 days after anakinra initiation. As expected, we observed complete resolution of ICANS in most patients (28-day cumulative incidence of ICANS resolution after anakinra initiation, 86%) with nonetheless significant TRM (28-day cumulative incidence of TRM after anakinra initiation, 13%), most of whom had ongoing ICANS at the time of death. These findings suggest that outcomes remain suboptimal despite the addition of anakinra to standard-of-care corticosteroid management. We also observed comparable outcomes between patients in whom dexamethasone was escalated to methylprednisolone compared with those who continued on dexamethasone. This important finding suggests that corticosteroid intensification beyond dexamethasone may not improve ICANS outcomes, while potentially increasing susceptibility to corticosteroid-induced side effects, such as myopathy and infections.

Given the lack of a validated ICANS-specific efficacy end point, we evaluated a novel and clinically relevant end point: 72-hour SNI. SNI at 72 hours provided early risk stratification, distinguishing patients with rapidly improving ICANS vs those with persistently poor outcomes despite continued treatment. From a clinical standpoint, the absence of SNI at 72 hours after anakinra initiation may represent a practical decision point during ICANS management. Patients without 72-hour SNI experienced significantly longer ICANS, higher cumulative corticosteroid exposure, and higher TRM. These findings suggest that continued escalation of corticosteroids and anakinra in the absence of early neurologic response may offer limited benefit and expose patients to additional toxicity. Instead, such patients may be appropriate candidates for alternative interventions, including intrathecal therapy, additional cytokine blockade (eg, siltuximab²⁹), or enrollment in clinical trials.

We next sought to identify factors associated with clinical outcomes and potentially predictive of anakinra efficacy. Successful identification of such factors would enable risk-adapted strategies (eg, treatment de-escalation in low-risk patients vs escalation in high-risk patients). Older age was associated with lower odds of 72-hour SNI, possibly reflecting age-related differences in susceptibility to neuroinflammation. This is in line with previous studies demonstrating an association between older age and greater severity of ICANS.30, 31, 32 In patients with multiple myeloma and in those receiving investigational CAR T-cell products, anakinra treatment yielded lower odds of 72-hour SNI, which may reflect differences in CAR T-cell biology (eg, constructs, manufacturing, dosing, expansion kinetics), concurrent targeted therapies (eg, γ-secretase inhibitors,³³ tyrosine kinase inhibitors), lymphodepletion intensity, and disease-specific factors. However, these factors did not appear to affect overall time to ICANS resolution.

In addition, baseline biomarkers of disease burden (LDH); biomarkers of systemic inflammation and coagulopathy (CRP, ferritin, platelet count) at baseline, ICANS onset, and anakinra initiation correlated with clinical outcomes. Although previous studies have linked biomarkers of disease burden, systemic inflammation, and coagulopathy to ICANS severity,22, 23, 24, 25, 26, 27 our findings extend these observations by suggesting a potential association with response to anakinra. These pretreatment factors may help identify patients who are less likely to achieve early neurologic improvement with anakinra, although these findings are exploratory and require prospective validation. Importantly, ICANS grade at the time of first ICANS-directed treatment initiation or at the time of anakinra initiation did not appear to affect outcomes. Although time to anakinra initiation was among the top 3 variables associated with time to ICANS resolution, it did not remain independently associated in multivariable modeling. Given these results, the relatively long median time to ICANS resolution from anakinra initiation, and the absence of a comparator group to define the natural course of ICANS without anakinra, the impact of anakinra on the overall trajectory of ICANS remains uncertain. Randomized prospective studies will be critical to assess the efficacy of anakinra on ICANS and to identify predictors of response.

In identifying key biomarker associations with response to anakinra, our work extends previous clinical experience and aligns with emerging reports of anakinra efficacy. In a recent conference abstract, Shah et al reported outcomes of anakinra to treat ICANS occurring in 42 patients with large B-cell lymphoma after CD19 CAR T-cell therapy.¹⁴ They defined response to anakinra as ICANS resolution after anakinra initiation. Consistent with their findings, we also confirmed elevated baseline serum LDH and ferritin as predictors of poor response, and identified greater dexamethasone exposure among patients with suboptimal responses. Their study was limited by a relatively small sample size and by the absence of analyses examining time to ICANS resolution and detailed treatment patterns. In contrast, our study provided an in-depth characterization of ICANS treatment patterns, specifically different escalation strategies and their associations with outcomes.

To our knowledge, this analysis represents the largest study to date investigating the use of anakinra in CAR T-cell therapy recipients developing ICANS. We comprehensively benchmarked clinical outcomes and identified key factors associated with anakinra efficacy, enabling potential risk-adapted strategies. However, we acknowledge several limitations. Due to its retrospective nature, treatment allocation bias could have confounded outcomes, including the inverse association between anakinra dose and odds of 72-hour SNI, which likely reflects greater ICANS severity in patients receiving higher doses. To address this, we modeled associations between a broad range of clinical and biological variables and outcomes before applying multivariable analyses. Importantly, the efficacy of anakinra in combination with corticosteroids over corticosteroids alone remains undetermined, emphasizing the urgent need for prospective randomized clinical trials to investigate the efficacy of anakinra on ICANS in the era of modern toxicity management.

In conclusion, our findings provide critical benchmarks that will inform future studies aimed at improving ICANS management. Anakinra was used primarily as a second-line treatment of ICANS after a lack of response to frontline treatment with dexamethasone, with or without concurrent escalation to high-dose methylprednisolone. Anakinra treatment was associated with a median time to ICANS resolution of 8 days. The 28-day cumulative incidence estimates of ICANS resolution and TRM were 86% and 13%, respectively. In addition, we identified potential predictive biomarkers of anakinra efficacy and identified a novel, clinically actionable end point for prospective validation. Lack of 72-hour SNI identified a high-risk subgroup with prolonged ICANS, higher corticosteroid exposure, and increased TRM. Early response assessment may therefore provide a practical framework for risk-adapted ICANS management and for identifying patients who may be considered for alternative or investigational therapies.

Conflict-of-interest disclosure: E.C.L. reports consultancy with Glass Health; honoraria from Eisai, Inc; and stock options in Bristol Myers Squibb (BMS). J.J.H. reports honoraria from Eisai, Inc; research funding to the institution from BMS; and consultancy with AbbVie. A.V.H. reports research funding from Juno Therapeutics (a BMS company) and Nektar Therapeutics; a licensing agreement with Cell Signaling Technology; and advisory board role with BMS. A.I.C. reports research funding from Novartis, Kite Pharma, BMS, Fate Therapeutics, Atara Biotherapeutics, Pierre Fabre, March Bio, Estrella Immunopharma, and Poseida Therapeutics; and consultancy with ADC Therapeutics, Pierre Fabre, and Elsevier. J.L. reports consultancy with Autolus, Kite, Takeda, Amgen, Adaptive Biotechnologies, Novartis, and AstraZeneca. S.E.S. reports consulting/advisory board roles with AstraZeneca, AbbVie, Genmab, Genentech, and Merck; research funding from Merck, BMS/Celgene, Genentech, AbbVie, AstraZeneca, Incyte, Schrödinger, Janssen, and Nurix; and expert testimony services for AbbVie. T.S. reports research funding from AstraZeneca; and reports spouse’s employment with Gilead Sciences. E.M. reports serving as an ad hoc consultant for, and receiving honoraria from, Prothena Biosciences; and reports research funding from Janssen, AstraZeneca, AbbVie, and BMS. R.T.M. reports advisory/consulting roles with ADC Therapeutics, Artiva, CRISPR Therapeutics, Incyte Corporation, Oricell Therapeutics, Orca Biotech, Poseida Biosciences, and Novartis; research funding from BMS, Kite, Orca Biotech, and Novartis; and serving on the data and safety monitoring board for Novartis, March Bio, Vor Pharmaceuticals, and Century Therapeutics. J.G. reports serving as an ad hoc consultant for, and receiving honoraria from, BMS, Kite Pharma, and Cartesian Therapeutics, Inc; research funding from Sobi, Juno Therapeutics (a BMS company), Celgene (a BMS company), Angiocrine Bioscience, Faron Pharmaceuticals, CARGO Therapeutics, Miltenyi Biotec, and CytoAgents; and serving on an independent data review committee for Century Therapeutics and University of Pennsylvania. The remaining authors declare no competing financial interests.