Cytogenetic Subgroups Drive Risk Stratification and Response to Chemotherapy and Blinatumomab in Children and Young Adults with Relapsed B-ALL: A Children’s Oncology Group Study

Background

At initial diagnosis, sentinel cytogenetic (CG) abnormalities differ based on demographic/clinical features and are predictive of relapse risk in B-lymphoblastic leukemia (B-ALL). The relevance of CG at 1^st relapse is not well defined. Children’s Oncology Group (COG) AALL1331 enrolled non-BCR-ABL1+, non-Down syndrome 1^st relapse B-ALL patients (pts) 1-30 years (yr) old. After a common chemotherapy induction (Block 1), pts were risk assigned as: High risk (HR): Early bone marrow (BM) relapse <36 months from diagnosis or isolated extramedullary (IEM) relapse <18 months; Intermediate risk (IR): Late BM relapse (>36 months) with minimal residual disease (MRD) ≥0.1% at end Block-1 (EB-1); Low risk (LR): Late BM relapse with MRD <0.1% at EB-1; Treatment failure (TF): M3 BM or failure to clear central nervous system at EB-1. The primary objective of the trial for HR/IR pts was to randomly compare post-reinduction therapy with 2 blocks of chemotherapy vs. blinatumomab, followed by allogeneic transplant. We previously reported the superiority of blinatumomab for HR/IR pts (Brown, ASH 2019). We hypothesized that CG would correlate with clinical features at relapse and response to therapy, enhancing relapse risk stratification.

Methods

CG data (karyotype and FISH) at relapse were centrally reviewed. CG groups were defined as: favorable (fav) (double trisomies (DT) or ETV6-RUNX1), unfavorable (unfav) (hypodiploid, KMT2A-rearranged (KMT2A-R) or iAMP21), and neutral (all others). Of 662 eligible pts, 130 IEM relapses were excluded, since CG data were not available due to lack of BM involvement. Of 532 BM relapses, CG data were available for 394 (74%); this group was used for associations of CG with clinical features and response to Block 1. Of 188 BM relapses risk assigned HR/IR and randomized, CG data were available for 147 (78%; 72 Arm A chemotherapy, 75 Arm B blinatumomab); this group was used for associations of CG with response to randomized post-reinduction therapy. CG distributions at initial diagnosis were based on data from over 16,000 pts enrolled on COG upfront trials from 1996-2014. Statistical comparisons used Pearson’s chi-squared, Fisher’s exact, Wilcoxon rank sum, or logrank test as appropriate. All p values are two-sided.

Results

As expected, unfav CG were more common at relapse compared to diagnosis (17% vs 7% p<0.001), and the opposite for fav CG (22% vs 42% p<0.001). Also expected was the younger age at relapse for the fav (median 8 yr) and KMT2A-R (median 2 yr) CG vs. others (median 12 yr, p=0.002). WBC at relapse was higher in the KMT2A-R (median 11k/uL), but similar in all others (median 5k/uL, p=0.041). Time from diagnosis to relapse differed by CG group, with late BM relapses comprising 79% of fav vs. 61% of neutral vs. 48% of unfav (p<0.001). EB-1 MRD-negative rates were highest for fav (53%), intermediate for neutral (38%) and lowest for unfav (25%, p=0.002). Interestingly, in fav CG group, 62% of ETV6-RUNX1 vs. 40% of DT were EB-1 MRD-negative (p=0.043), suggesting that as for newly diagnosed pts, MRD clearance during relapse reinduction may be slower for DT than for ETV6-RUNX1. EB-1 risk assignments differed significantly by CG group, with rates for fav/neutral/unfav CG of 57%/35%/27% for LR, 16%/31%/34% for HR, and 3%/7%/7% for TF (p<0.01). After 1^st course of randomized therapy for HR/IR pts, the rate of EB-2 MRD negativity was (as previously reported) much higher for arm B (75%) vs arm A (25%, p<0.001). Interestingly, EB-2 MRD clearance rates did not vary significantly by CG group for either randomized arm. There was, however, an influence of CG group on survival for arm B, but not for arm A. The 2 yr DFS/OS for fav CG on arm B was 77%/93%, vs. 49%/65% for neutral and 49%/49% for unfav (p=0.069 for DFS, p=0.035 for OS, univariate analysis). For arm A, DFS/OS was similar and consistently poor in all three groups (44%/52% fav vs. 49%/61% neutral and 39%/61% unfav, p=0.92 for DFS, p=0.62 for OS) (Table 1).

Conclusions

Pts with fav CG B-ALL relapse later and are more likely to achieve MRD negativity after reinduction chemotherapy, resulting in higher rates of LR risk assignment. Interestingly, while the superior MRD clearance with blinatumomab vs. chemotherapy is seen in all CG groups, this only translated into higher rates of DFS/OS in the fav CG group. Thus, the influence of CG on outcomes after 1^st relapse may differ based on whether chemotherapy or immunotherapy is used as post-reinduction consolidation.