Table 5 Factors associated with CAR-T therapy toxicity

Lymphodepletion regimen (chemotherapy given before CAR-T infusion)

Anti-transgene response was observed in the absence of lymphodepletion^116,117,119

Combined lymphodepletion (Cy&Flu) resulted in better CAR T-cells expansion^11,120, a higher serum CAR-T peak^61,75 and higher toxicity (CRS and CRES)^28,61

Combined lymphodepletion (Cy&Flu) was mentioned⁹⁶ as the risk factor of fatal cerebral edema (ROCKET trial, see Table 1), however, after reverting back to monoCy lymphodepletion two more deaths were observed

Antigen type/epitope/scFv

Some of the tumor-specific CARs and TCRs are known to cross-react with normal tissue antigens (on-target off-tumor toxicities: B-cell aplasia in anti-CD19-therapies, cardiopulmonary toxicity in HER2¹²¹, and MAGE-A3-directed therapy¹²²)

Anaphylaxis and anti-CAR-T immune response are associated with murine epitopes in CAR¹¹

Toxicity profiles may theoretically differ between the scFv domains due to their different affinities for specific epitopes of the target antigen

JCAR015 (ROCKET trial, see Table 1) bore the recognition module derived from SJ25C1 in contrast to FMC63-based scFv used in other products by Novartis, Kite, and Juno Therapeutics. Toxicity impact unclear

CAR generation

Early trials with the first-generation CAR-Ts showed lack of both toxicity and efficacy^116,117,119

Across second-generation CAR-Ts with different co-stimulatory domains the toxicity profile is very similar

CD28-based CAR-Ts proliferate more actively and their peak expansion level is higher than that of the 4-1BB-containing CAR- Ts⁹⁶

In turn, 4-1BB module ameliorates CAR-T exhaustion¹²³

Little toxicity (low-grade CRS, no evidence of CRES) for the 4th generation 4SCAR19 bearing three co-stimulatory domains⁶. No comments on the toxicity profile are reported yet

T-cell subpopulation composition

Bulk CD8+ subset was an independent risk factor for CRS (JCAR014)²⁸, as well as for severe CRES (JCAR015)⁷⁴

JCAR014 and JCAR017 with defined CD4+:CD8+ composition^11,25 are being developed by Juno Therapeutics

JCAR017 demonstrated low rate of side effects (CRS and CRES)¹²⁴, however, extended data are expected

Disease type

NHL appears to show less frequent CRS in comparison to ALL (30–57%^24,25 vs 74–100%^6,10,26 in the largest trials), however in ZUMA-1 (NHL) CRS incidence was 94% (39%—grade 1)²⁷

For JCAR014, the type of disease impacted neither the severity of CRS nor CRES frequency^28,61

CAR T-cell dose and expansion peak

Infusion of 5*10⁸ CAR T-cells resulted in unacceptable toxicity (all 6 patients developed CRS and 3 died). Splitting this dose over 3 days with flexible administration schedule resulted in 86% response rate and 66% CRS rate. 5*10⁷ cells dose resulted in efficacy decrease and comparable toxicity (n = 27, ALL)¹²⁵

CAR T-cells dose was found to be a significant factor associated either with CRS and CRES^28,61

For CRS, the interplay between CAR T-cell dose and Cy&Flu lymphodepletion was found, i.e., at any given CAR T-cell dose addition of fludarabine increased the probability of CRS²⁸ onset

Only weak association between severity of CRS and the peak of CAR T-cell expansion was shown (n = 51)⁴⁷, but other data (n = 133)²⁸ demonstrate the correlation of peak CAR T cell serum levels with both efficacy and toxicity of the therapy

Logistic regression modeling performed to detect the therapeutic window²⁸ balancing between toxicity and efficacy 

Serum IL15 levels are associated with higher CAR T-cells level^74,126, efficacy of the therapy¹²⁶ and ≥3 CRES risk^74,126

Tumor burden

Borderline positive predictive value for sCRS (10 of 23 patients with >25% of marrow blasts developed sCRS)⁴⁷, but strong negative predictive value (1 out of 15 patients with <5% bone marrow blasts experienced sCRS)⁴⁷

In other studies, bone marrow blasts were included into predictive models^47,75 for CRS and CRES

The tumor burden-adapted treatment protocol was developed (JCAR014): the dose of 2*10⁵ CAR T-cells/kg for B-ALL with >20% marrow blasts; 2*10⁶ CAR T-cells/kg for B-ALL with ≤20% marrow blasts and for patients with NHL or CLL^11,78,79,95