Risk factors for inferior outcomes after CD19 CAR T cells in pediatric B-ALL
| Risk factor . | Summary . | Specific data . |
|---|---|---|
| Baseline disease characteristics | ||
| Leukemia cytogenetics | Overall, there is not a clear impact of leukemia cytogenetic lesions on CD19 CAR outcomes. However, several rare, specific lesions may deserve additional consideration: 1. KMT2Ar: similar relapse risk to other groups, but higher incidence of myeloid lineage switch with dismal survival after lineage switch leading to worse OS. 2. TP53: associated with substantially worse LFS and OS in a small sample. 3. Hypodiploid: poor outcomes reported in several small studies. | CHOP study21 (N = 231): similar RFS for patients with high-, intermediate-, and low-risk cytogenetics. Patients with KMT2Ar infant B-ALL (n = 13) had similar RFS to other groups, but worse OS (HR 3.6). Of 8 patients with hypodiploid B-ALL, 5 relapsed. Robert Debre and Saint Louis University Hospitals study22 (N = 51): high-risk genetic lesions were not associated with CIR, EFS, or OS. CAR-MA studies20,23 (N = 420): KMT2Ar (n = 38) associated with higher risk of lineage switch (HR 32.35) and lower median OS (25.3 vs 51.9 mo). EFS for patients with hypodiploid B-ALL (n = 14) was approximately 25% by 12 mo after CAR. Lu Daopei Hospital phase 1/2 study24 (N = 110): patients with TP53 mutations had lower 6-mo LFS (42% vs 83%, P = .0002) and OS (52% vs 89%, P < .0001) than those without TP53 alterations. |
| Molecular targets | As above, CD19 CAR T cells are effective across multiple molecular subtypes of B-ALL. For patients with Philadelphia-positive or certain subtypes of Philadelphia-like B-ALL, the question has been raised of combining CAR T cells with the multi–tyrosine kinase inhibitor dasatinib. This is not currently recommended, as high-dose dasatinib is toxic to T cells. | Preclinical dasatinib experiments:25 in a xenograft model, dasatinib rapidly and reversibly inhibited CAR T-cell activation and proliferation, cytokine production, and CAR T-cell killing. The authors suggested that dasatinib could potentially be repurposed as an “on” and “off” switch for CAR T cells. Further, the authors posited that intermittent or low-dose dasatinib may provide a synergistic therapeutic effect with CAR T cells. No human data exists, so this approach is not recommended. |
| Previous therapy | ||
| Blinatumomab | Two single/dual-center analyses suggested that prior blinatumomab may affect response rates and relapse risk. A large, multicenter analysis showed that nonresponse to prior blinatumomab, rather than blinatumomab exposure alone, was associated with inferior response rates and EFS. | CHOP study (N = 166): composite outcome of NR, CD19− MRD/relapse was more frequent in blinatumomab-exposed patients.26, Robert Debre and Saint Louis University Hospitals study22 (N = 51): prior blinatumomab was associated with early CAR failure (P = .01), increased CIR (HR 2.6), and shorter EFS (HR 3.0) and OS (HR 5.5). CAR-MA study (N = 420): blinatumomab nonresponders had lower CR rates to CD19 CAR T cells and worse 6-mo EFS (CR, 65%; EFS, 27%) than blinatumomab responders (CR, 93%; EFS, 67%) or blinatumomab-naïve patients (CR, 94%; EFS, 73%).20 |
| Quantity of previous therapy | Two multicenter analyses showed that a greater quantity of prior therapy, possibly indicating more refractory disease, was associated with worse outcomes. | PRWCC study (N = 185): a greater number of prior lines of therapy was associated with worse OS by multivariable analysis (HR 1.4, P = .022).27 CAR-MA study23 (N = 420): at least 2 prior CRs were associated with a higher cumulative incidence of CD19-positive relapse by multivariable analysis (HR 1.3, P = .008). |
| Factors at time of infusion | ||
| Bone marrow disease burden | Multiple clinical trials and multicenter analyses clearly demonstrate that high disease burden (HD burden), generally defined as ≥5% bone marrow blasts, is associated with increased risk of relapse, compared with low disease burden (LD burden). | CHOP clinical trials: in a trial of tisagenlecleucel (N = 70),28 patients with HD burden (>40% blasts) had inferior 24-mo EFS (34% vs 78%) and OS (60% vs 92%) compared with LD burden. In a trial of humanized CD19 CAR (N = 74),29 HD burden was associated with inferior RFS. CAR-MA studies20,23 (N = 420): HD burden (≥5% bone marrow blasts) was associated with inferior EFS, RFS, and OS. HD burden was independently associated with worse EFS (HR 2.5, P < .001) by multivariable analysis, and specifically associated with a higher cumulative incidence of CD19− relapse (HR 5.2, P < .001). PRWCC study27 (N = 185): patients with HD burden (≥5% bone marrow blasts) had lower 12-mo EFS (31% vs 70%, P < .0001) and OS (58% vs 85%, P < .0001) compared with LD burden. HD burden was independently associated with OS by multivariable analysis (HR 5.1, P = .002). St Jude and JHU study30 (N = 30): HD burden (≥5% bone marrow blasts) was independently associated with inferior EFS (HR 6.0, P = .038) and OS (HR 4.2, P = .015). Robert Debre and Saint Louis University Hospitals study22 (N = 51): HD burden (≥1% bone marrow blasts) was associated with a higher cumulative incidence of CD19− relapse (SHR 10.4, P = .03) in a competing risks analysis. |
| EMD | Active non-CNS EMD at time of CD19 CAR infusion has been associated with worse EFS. Prior extramedullary or CNS disease, however, has not been associated with worse outcomes. | CAR-MA study20 (N = 420): active non-CNS EMD at infusion was independently associated with worse EFS (HR 1.9, P = .01) PRWCC study31 (N = 184): CNS and non-CNS EMD were not associated with RFS or OS differences. |
| CD19+ antigen load | Bone marrow CD19+ antigen load has been associated with CAR persistence, which, in turn, would be expected to affect relapse risk. | Seattle PLAT-02 trial (N = 45): CD19+ antigen load measured as a combination of normal B cells and blasts <15% before lymphodepletion was independently associated with early loss of BCA (HR 3.0, P = .005).32 |
| Lymphodepletion | Although most patients receive standard dosing of cyclophosphamide and fludarabine based on body surface area for lymphodepletion, this may not be the optimal regimen for all patients. In 2 recent retrospective studies, for example, suboptimal fludarabine exposure was associated with increased relapse risk and shorter persistence. In addition, bendamustine has been explored as an alternative lymphodepleting agent, especially in light of recent fludarabine shortages; however, there is no published data on its use in pediatric B-ALL. | PRWCC study33 (N = 152): suboptimal fludarabine exposure, defined as AUC <13.8 mg × h/L and estimated by a validated population pharmacokinetic model, was associated with a higher CIR (HR 2.5, P = .005) and higher risk of a composite end point of relapse or loss of BCA (HR 2.0, P = .01) compared with optimal fludarabine exposure. Princess Maxima study34 (N = 26): a cumulative fludarabine AUC <14 mg × h/L was associated with a higher frequency of CD19+ relapse within 1 y (100% vs 27%, P = .0001) and probability of losing BCA within 6 mo (77% vs 37%, P = .009) than AUC >14 mg × h/L. Multicenter, adult large B-cell lymphoma study35 (N = 132): compared with fludarabine/cyclophosphamide lymphodepletion before tisagenlecleucel, bendamustine lymphodepletion demonstrated similar efficacy and lower rates of CRS, neurotoxicity, and hematologic toxicities. |
| Postinfusion factors | ||
| Short CAR persistence (loss of BCA) | Short CAR persistence, as evidenced by loss of peripheral BCA, has been highly associated with increased risk of relapse. However, the duration of BCA needed to mitigate relapse risk has not been fully elucidated. | Seattle PLAT-02 trial32 (N = 45): loss of BCA was associated with increase relapse risk (HR 3.5, P = .04). Pooled ELIANA/ENSIGN analysis36 (N = 143): loss of BCA within 1 y was associated with increased relapse risk (HR 4.5, P < .001). Patients with loss of BCA within 6 mo had a 24-mo EFS of 14%. CHOP humanized CD19 CAR T-cell trial29 (N = 74): when treated as a time-varying covariate, B-cell recovery was associated with worse RFS (P = .011). |
| Detectable NGS- MRD or IG/TCR RQ-PCR MRD | In a large, retrospective analysis, any detectable NGS-MRD at different time points after CD19 CAR T cells was highly predictive of subsequent relapse. In another retrospective analysis, detectable IG/TCR MRD (>10−4) at day 28 was highly predictive of subsequent relapse. | Pooled ELIANA/ENSIGN analysis36 (N = 143): NGS-MRD >0 at day 28 (HR 4.9, P < .001) and month 3 (HR 12, P < .001) after CD19 CAR T cells were highly and independently associated with relapse. Robert Debre and Saint Louis University Hospitals study22 (N=51): detectable IG/TCR RQ-PCR-MRD (>10−4) at day 28 after CD19 CAR T cells was highly and independently associated with relapse (HR 6.0, P = .006). |
| Risk factor . | Summary . | Specific data . |
|---|---|---|
| Baseline disease characteristics | ||
| Leukemia cytogenetics | Overall, there is not a clear impact of leukemia cytogenetic lesions on CD19 CAR outcomes. However, several rare, specific lesions may deserve additional consideration: 1. KMT2Ar: similar relapse risk to other groups, but higher incidence of myeloid lineage switch with dismal survival after lineage switch leading to worse OS. 2. TP53: associated with substantially worse LFS and OS in a small sample. 3. Hypodiploid: poor outcomes reported in several small studies. | CHOP study21 (N = 231): similar RFS for patients with high-, intermediate-, and low-risk cytogenetics. Patients with KMT2Ar infant B-ALL (n = 13) had similar RFS to other groups, but worse OS (HR 3.6). Of 8 patients with hypodiploid B-ALL, 5 relapsed. Robert Debre and Saint Louis University Hospitals study22 (N = 51): high-risk genetic lesions were not associated with CIR, EFS, or OS. CAR-MA studies20,23 (N = 420): KMT2Ar (n = 38) associated with higher risk of lineage switch (HR 32.35) and lower median OS (25.3 vs 51.9 mo). EFS for patients with hypodiploid B-ALL (n = 14) was approximately 25% by 12 mo after CAR. Lu Daopei Hospital phase 1/2 study24 (N = 110): patients with TP53 mutations had lower 6-mo LFS (42% vs 83%, P = .0002) and OS (52% vs 89%, P < .0001) than those without TP53 alterations. |
| Molecular targets | As above, CD19 CAR T cells are effective across multiple molecular subtypes of B-ALL. For patients with Philadelphia-positive or certain subtypes of Philadelphia-like B-ALL, the question has been raised of combining CAR T cells with the multi–tyrosine kinase inhibitor dasatinib. This is not currently recommended, as high-dose dasatinib is toxic to T cells. | Preclinical dasatinib experiments:25 in a xenograft model, dasatinib rapidly and reversibly inhibited CAR T-cell activation and proliferation, cytokine production, and CAR T-cell killing. The authors suggested that dasatinib could potentially be repurposed as an “on” and “off” switch for CAR T cells. Further, the authors posited that intermittent or low-dose dasatinib may provide a synergistic therapeutic effect with CAR T cells. No human data exists, so this approach is not recommended. |
| Previous therapy | ||
| Blinatumomab | Two single/dual-center analyses suggested that prior blinatumomab may affect response rates and relapse risk. A large, multicenter analysis showed that nonresponse to prior blinatumomab, rather than blinatumomab exposure alone, was associated with inferior response rates and EFS. | CHOP study (N = 166): composite outcome of NR, CD19− MRD/relapse was more frequent in blinatumomab-exposed patients.26, Robert Debre and Saint Louis University Hospitals study22 (N = 51): prior blinatumomab was associated with early CAR failure (P = .01), increased CIR (HR 2.6), and shorter EFS (HR 3.0) and OS (HR 5.5). CAR-MA study (N = 420): blinatumomab nonresponders had lower CR rates to CD19 CAR T cells and worse 6-mo EFS (CR, 65%; EFS, 27%) than blinatumomab responders (CR, 93%; EFS, 67%) or blinatumomab-naïve patients (CR, 94%; EFS, 73%).20 |
| Quantity of previous therapy | Two multicenter analyses showed that a greater quantity of prior therapy, possibly indicating more refractory disease, was associated with worse outcomes. | PRWCC study (N = 185): a greater number of prior lines of therapy was associated with worse OS by multivariable analysis (HR 1.4, P = .022).27 CAR-MA study23 (N = 420): at least 2 prior CRs were associated with a higher cumulative incidence of CD19-positive relapse by multivariable analysis (HR 1.3, P = .008). |
| Factors at time of infusion | ||
| Bone marrow disease burden | Multiple clinical trials and multicenter analyses clearly demonstrate that high disease burden (HD burden), generally defined as ≥5% bone marrow blasts, is associated with increased risk of relapse, compared with low disease burden (LD burden). | CHOP clinical trials: in a trial of tisagenlecleucel (N = 70),28 patients with HD burden (>40% blasts) had inferior 24-mo EFS (34% vs 78%) and OS (60% vs 92%) compared with LD burden. In a trial of humanized CD19 CAR (N = 74),29 HD burden was associated with inferior RFS. CAR-MA studies20,23 (N = 420): HD burden (≥5% bone marrow blasts) was associated with inferior EFS, RFS, and OS. HD burden was independently associated with worse EFS (HR 2.5, P < .001) by multivariable analysis, and specifically associated with a higher cumulative incidence of CD19− relapse (HR 5.2, P < .001). PRWCC study27 (N = 185): patients with HD burden (≥5% bone marrow blasts) had lower 12-mo EFS (31% vs 70%, P < .0001) and OS (58% vs 85%, P < .0001) compared with LD burden. HD burden was independently associated with OS by multivariable analysis (HR 5.1, P = .002). St Jude and JHU study30 (N = 30): HD burden (≥5% bone marrow blasts) was independently associated with inferior EFS (HR 6.0, P = .038) and OS (HR 4.2, P = .015). Robert Debre and Saint Louis University Hospitals study22 (N = 51): HD burden (≥1% bone marrow blasts) was associated with a higher cumulative incidence of CD19− relapse (SHR 10.4, P = .03) in a competing risks analysis. |
| EMD | Active non-CNS EMD at time of CD19 CAR infusion has been associated with worse EFS. Prior extramedullary or CNS disease, however, has not been associated with worse outcomes. | CAR-MA study20 (N = 420): active non-CNS EMD at infusion was independently associated with worse EFS (HR 1.9, P = .01) PRWCC study31 (N = 184): CNS and non-CNS EMD were not associated with RFS or OS differences. |
| CD19+ antigen load | Bone marrow CD19+ antigen load has been associated with CAR persistence, which, in turn, would be expected to affect relapse risk. | Seattle PLAT-02 trial (N = 45): CD19+ antigen load measured as a combination of normal B cells and blasts <15% before lymphodepletion was independently associated with early loss of BCA (HR 3.0, P = .005).32 |
| Lymphodepletion | Although most patients receive standard dosing of cyclophosphamide and fludarabine based on body surface area for lymphodepletion, this may not be the optimal regimen for all patients. In 2 recent retrospective studies, for example, suboptimal fludarabine exposure was associated with increased relapse risk and shorter persistence. In addition, bendamustine has been explored as an alternative lymphodepleting agent, especially in light of recent fludarabine shortages; however, there is no published data on its use in pediatric B-ALL. | PRWCC study33 (N = 152): suboptimal fludarabine exposure, defined as AUC <13.8 mg × h/L and estimated by a validated population pharmacokinetic model, was associated with a higher CIR (HR 2.5, P = .005) and higher risk of a composite end point of relapse or loss of BCA (HR 2.0, P = .01) compared with optimal fludarabine exposure. Princess Maxima study34 (N = 26): a cumulative fludarabine AUC <14 mg × h/L was associated with a higher frequency of CD19+ relapse within 1 y (100% vs 27%, P = .0001) and probability of losing BCA within 6 mo (77% vs 37%, P = .009) than AUC >14 mg × h/L. Multicenter, adult large B-cell lymphoma study35 (N = 132): compared with fludarabine/cyclophosphamide lymphodepletion before tisagenlecleucel, bendamustine lymphodepletion demonstrated similar efficacy and lower rates of CRS, neurotoxicity, and hematologic toxicities. |
| Postinfusion factors | ||
| Short CAR persistence (loss of BCA) | Short CAR persistence, as evidenced by loss of peripheral BCA, has been highly associated with increased risk of relapse. However, the duration of BCA needed to mitigate relapse risk has not been fully elucidated. | Seattle PLAT-02 trial32 (N = 45): loss of BCA was associated with increase relapse risk (HR 3.5, P = .04). Pooled ELIANA/ENSIGN analysis36 (N = 143): loss of BCA within 1 y was associated with increased relapse risk (HR 4.5, P < .001). Patients with loss of BCA within 6 mo had a 24-mo EFS of 14%. CHOP humanized CD19 CAR T-cell trial29 (N = 74): when treated as a time-varying covariate, B-cell recovery was associated with worse RFS (P = .011). |
| Detectable NGS- MRD or IG/TCR RQ-PCR MRD | In a large, retrospective analysis, any detectable NGS-MRD at different time points after CD19 CAR T cells was highly predictive of subsequent relapse. In another retrospective analysis, detectable IG/TCR MRD (>10−4) at day 28 was highly predictive of subsequent relapse. | Pooled ELIANA/ENSIGN analysis36 (N = 143): NGS-MRD >0 at day 28 (HR 4.9, P < .001) and month 3 (HR 12, P < .001) after CD19 CAR T cells were highly and independently associated with relapse. Robert Debre and Saint Louis University Hospitals study22 (N=51): detectable IG/TCR RQ-PCR-MRD (>10−4) at day 28 after CD19 CAR T cells was highly and independently associated with relapse (HR 6.0, P = .006). |
AUC, area under the curve; CHOP, Children’s Hospital of Philadelphia; CIR, cumulative incidence of relapse; JHU, Johns Hopkins University; NGS-MRD, detectable minimal residual disease by next-generation sequencing; IG/TCR RQ-PCR MRD, allele-specific real-time quantitative PCR of leukemia-specific immune-receptor gene rearrangements (either IG or TCR sequences); SHR, subdistribution hazard ratio; St Jude, St Jude Children’s Research Hospital; TCR, T-cell receptor.