Risk factors and outcomes in children with high-risk B-cell precursor and T-cell relapsed acute lymphoblastic leukaemia: combined analysis of ALLR3 and ALL-REZ BFM 2002 clinical trials

Aim: Outcomes of children with high-risk (HR) relapsed acute lymphoblastic leukaemia (ALL) (N Z 393), recruited to ALLR3 and ALL-REZ BFM 2002 trials, were analysed. Minimal residual disease (MRD) was assessed after induction and at predetermined time points until haematopoietic stem cell transplantation (SCT). Methods: Genetic analyses included karyotype, copy-number alterations and mutation analyses. Ten-year survivals were analysed using Kaplan-Meier and Cox models for multivariable analyses. Results: Outcomes of patients were comparable in ALLR3 and ALL-REZ BFM 2002. The event-free survival of B-cell precursor (BCP) and T-cell ALL (T-ALL) was 22.6% and 26.2% ( P Z 0 . 94), respectively, and the overall survival (OS) was 32.6% and 28.2% ( P Z 0.11), respectively. Induction failures (38%) were associated with deletions of NR3C1 ( P Z 0.002) and BTG1 ( P Z 0 . 03) in BCP-ALL. The disease-free survival (DFS) and OS in patients with good vs poor MRD responses were 57.4% vs 22.6% ( P < 0.0001) and 57.8% vs 32.0% ( P Z 0 . 0004), respectively. For BCP- and T-ALL, the post-SCT DFS and OS were 42.1% and 56.8% ( P Z 0 . 26) and 51.6% and 55.4% ( P Z 0 . 67), respectively. The cumulative incidences of post-SCT relapse for BCP- and T-ALL were 36.9% and 17.8% ( P Z 0 . 012) and of death were 10.7% and 25.5% ( P Z 0 . 013), respectively. Determinants of outcomes after SCT were acute graft versus host disease, pre-SCT MRD ( (cid:1) 10 (cid:3) 3 ), HR cytogenetics and TP53 alterations in BCP-ALL. Conclusion: Improvements in outcomes for HR ALL relapses require novel compounds in induction therapy to improve remission rates and immune targeted therapy after induction to maintain remission after SCT.


Introduction
Duration of first complete remission and immunophenotypes predict outcomes in patients with relapsed acute lymphoblastic leukaemia (ALL) [1e3]. B-cell precursor (BCP) ALL with bone marrow relapses occurring later than 6 months after stopping therapy has survival rates of more than 80% [4,5]. In contrast, patients with highrisk (HR) relapses defined as BCP-ALL relapses within 18 months of first diagnosis or with isolated medullary relapses occurring within 6 months of stopping therapy, and T-cell isolated or combined medullary relapses at any time, have survival rates between 15% and 30% [3,6,7] even after allogeneic haematopoietic stem cell transplantation (SCT) [2,3,8e10]. Second complete remission (CR2) rates in HR patients range between 68% and 88% [3,9,10]. More recently, HR relapsed patients are receiving immune-directed therapies. For patients with HR BCP-ALL, CD19-and CD22-targeted therapies with the CD19-directed bispecific T-cell engager blinatumomab, the CD22-directed toxin-conjugated monoclonal antibody inotuzumab ozogamicin [11e16] or chimeric antigen receptor T-cells (CAR-Ts) [17,18] offer superior remission rates, although the long-term outcomes of these novel approaches are awaited.
In this context, we have analysed the long-term outcomes of HR patients treated in the ALLR3 and ALL-REZ BFM 2002 clinical trials for relapsed ALL. Both ALLR3 and ALL-REZ BFM 2002 used a common risk stratification method and assessed minimal residual disease (MRD) after induction and before SCT, and all HR patients were eligible for SCT after 12 weeks of chemotherapy. This permitted combining the trial data sets for patients with HR bone marrow ALL relapses treated in both trials to evaluate different chemotherapy approaches for induction remission in HR relapsed ALL and the relationship of MRD at different time points with survival, along with additional genetic analyses.

Study population
This analysis reports on patients with HR (Table 1) relapsed ALL without a previous SCT, treated in ALLR3 (NCT00967057, N Z 136) or ALL-REZ BFM 2002 (NCT00114348, N Z 257) trials. Early isolated extramedullary relapses were excluded. Patients at the age between 1 and 18 years at relapse diagnosis were included in the analyses. Trials were approved by relevant institutional ethical committees, and patients were recruited after written consent between 28th December 2001 and 18th June 2011 (ALL- REZ BFM 2002) and between 31st January 2003 and 31stOctober 2013 (ALLR3).

Therapy
Chemotherapy for ALL-REZ BFM 2002 [8] and ALLR3 [19] has been previously reported and is briefly described in the supplemental section (Supplemental Table S1). ALLR3 randomised patients to idarubicin or mitoxantrone in induction, and early closure of the randomisation in favour of mitoxantrone [19] allowed the evaluation of clofarabine, cyclophosphamide and etoposide (CCE) in induction.

Assessment of response
CR2 was defined as <5% blasts in the marrow and no blasts in the cerebrospinal fluid (CSF) at the end of induction (EOI)dotherwise defined as induction failure. MRD was measured after induction and in some patients after each block of therapy before SCT, with the validated and standardised real-time quantitative polymerase chain reaction assays for clonal gene rearrangements in immunoglobulin/T-cell receptor loci [8,19]. Two postinduction MRD response groups were defined. The good response (GR) group was that with an MRD of <10 À4 at the EOI or between !10 À4 and <10 À3 at the EOI, with a subsequent MRD value of <10 À4 after induction. All others were defined as poor response (PR).

Statistical analysis
Survival analysis considered five main end-points. Event-free survival (EFS) was calculated as the time from first relapse, using the date of relapse, until the first event (induction death, induction failure, second relapse, death in CR2 and second malignancy) or last follow-up. Disease-free survival (DFS) included only patients who achieved CR2 and was calculated as the time from first relapse until the first event (second relapse, death, second malignancy) or last follow-up. For outcomes after SCT, the date of SCT was used as the starting point for DFS. Overall survival (OS) was defined as time from first relapse to death censoring at the last contact. Cumulative incidence of subsequent relapses (CIR) or cumulative incidence of death (CID) was calculated as the time from first relapse until second relapse or death in CR2, taking other DFS events into account as competing events. EFS, DFS and OS probabilities and cumulative incidence functions for competing events are given at 10 years in the following Results and Discussion sections. Further details are in Supplemental section.

Comparison of patients treated uisng the different regimens
Twenty-five, 60 and 51 HR ALLR3 patients received idarubicin, mitoxantrone, CCE arms; 257 patients ALL-REZ BFM 2002 respectively (Supplemental Results). CR2 rates and EFS were comparable within the different treatment protocols for both BCP-ALL and T-cell ALL (T-ALL) (Fig. 1, Table 2). As the EFS of the different therapeutic groups was comparable, data from all 393 patients were combined to further investigate the determinants of outcomes in HR relapsed ALL.

Overall outcomes of the combined cohort
Two hundred seventy-eight (71%) patients had BCP-ALL, and 115 (29%) had T-ALL. Progression through treatment for the patients with BCP-and T-ALL is shown in Fig. 1 and for the whole cohort is shown in Supplementary Fig. S1. The proportion of patients attaining CR2 and reaching SCT was comparable between the two groups. The EFS and OS were 22.6% (95% confidence interval [CI]: 17e29) and 26.2% (18e35) as well as 32.6% (27e39) and 28.2% (20e37), for BCP-and T-ALL, respectively (Fig. 2). Although outcomes of patients with BCP-ALL and T-ALL relapse were comparable, T-ALL relapses at any time point are classified as HR. In this cohort, 25% of T-ALL were late relapses, and these patients were more likely to achieve CR2 (Table 3).

Determinants of induction failure
Overall, for both BCP-and T-ALL, 4% of patients died in induction and 34% of patients failed induction (Fig. 1). Induction failures were more frequent in those relapsing within 18 months from first diagnosis in both BCP-and T-ALL (P Z 0.020and P Z 0.017, respectively), in older children with BCP-ALL (P Z 0.012) and in BCP-ALL with isolated medullary relapse (P Z 0.040). None of these factors appeared to influence EFS in those who achieved CR2 (Table 3). Induction failures were more frequent in patients with cytogenetic HR BCP-ALL (48%) than in cytogenetic SR (24%) and B-other patients (31%, P Z 0.032), especially those with hypodiploidy (64%, P Z 0.017) and associated with deletions of BTG1 and NR3C1 (P Z 0.031, P < 0.001; Table 4). Eighteen (15%) of 123 patients with induction failures subsequently achieved remission, 9 with continuing protocol treatment and 9 with alternative treatment strategies.

Postinduction MRD response and correlation to outcomes
One hundred seventy-eight patients with BCP-ALL and 74 patients with T-ALL achieved CR2 (Fig. 1). A second relapse before SCT occurred in 18 of 140 patients (13%)  with BCP-ALL and 15 of 55 patients (27%) with T-ALL (P Z 0.0157, Fig. 1). The EOI MRD thresholds for the ALL-REZ BFM 2002 [8] and ALLR3 [19] trials were 10 À3 and 10 À4 , respectively. DFS and OS were significantly better in those with an EOI MRD of <10 À3 in both trials, but in ALLR3 only, these were better in those with an EOI MRD of <10 À4 (Table 5)     Discrepancies of patient numbers between the 10 À4 and 10 À3 cut-off groups are due to sensitivity of MRD markers (only 10 À3 or 5 Â 10 À4 , but not 10 À4 ) or the non-available quantitative MRD values to assign the MRD results to one of the positive MRD groups, either !10 À3 or <10 À3 e!10 À4 . P values in bold indicate those <0.05. Table 6 Outcomes of patients with high-risk relapsed ALL based on MRD response. a Total numbers between the three MRD categories (EOI </!10 À4 , EOI </!10 À3 , GR/PR) vary because of the sensitivity of MRD markers or missing MRD time points after induction. b p values calculated using the log-rank test. c p values calculated using the Gray test. d MRD good response was defined as either <10 À4 at the end of induction or <10 À3 at the end of induction and subsequent MRD values during consolidation/before SCT <10 À4 . The MRD poor response group included all other responses, !10 À4 at the end of induction and one or more subsequent MRD values during consolidation/before SCT !10 À4 . anthracycline after induction. As EFS in both trials was comparable, we analysed EOI and post-EOI MRD values as MRD-GR and MRD-PR. Grouped as MRD-GR and MRD-PR, DFS, CIR and OS were significantly better for patients with MRD-GR for both BCP-and T-ALL (Table 6 and Fig. 3). In BCP-or T-ALL, clinical risk parameters (Supplemental Table S3) or cytogenetic/ genetic parameters (Supplemental Table S4) in BCP-ALL did not distinguish between MRD-GR and MRD-PR, although numbers in each group were small. The EOI MRD results were available in 18 of the 33 patients relapsing before SCT. Sixteen (89%) had an EOI MRD of !10 À3 , 6 had an EOI MRD of !10 À2 and one had an EOI MRD of <10 À4 . MRD was not available for 89 (35%) of 252 patients in CR2 (Table 6) and is detailed in Supplemental Table S5.

Determinants of outcomes after SCT
Of the 207 patients who reached the SCT time point, 183 (88%) received an SCT, 18 did not receive an SCT and SCT status was unknown for 6 patients (Fig. 1). The DFS and OS of patients who received an SCT and who did not  Fig. 3. Different disease-free survival, overall survival and cumulative incidence of subsequent relapses in patients with MRD good vs MRD poor response after induction treatment. Kaplan-Meier analyses of disease-free (A) and overall (B) survival and cumulative incidence of subsequent relapses and death (C) in patients who were in CR and MRD until SCT was measured. CR, complete remission; CI, confidence interval; CIR, cumulative incidence of relapses; CID, cumulative incidence of death; MRD, minimal residual disease; BCP, B-cell precursor; ALL, acute lymphoblastic leukaemia; GR, good response; PR poor response.

Discussion
The outcomes of HR relapsed ALL treated in the two trials were similar, allowing observations to be made on a combined analysis of prospectively treated and uniformly defined HR relapsed ALL. Nevertheless, the data need to be interpreted cautiously as a number of patients were withdrawn because of toxicity with idarubicin and CCE, the study population was heterogeneous, numbers in subgroup analyses were small and the median follow-up time is variable in the different groups, although most events occurred within 36 months. The trials also accrued patients between 2001 and 2013. Although the data are mature, during the current era, there have been improvements in transplant outcomes and the availability of immunotherapies. The results show comparable outcomes for both BCP-and T-ALL HR relapse with a benefit for SCT. The results of this paper confirm previous observations of poor outcomes in older patients [24], relapses within 18 months of diagnosis, with isolated medullary relapse [2] and with HR cytogenetics [22]. Neither the donor nor the stem cell source influenced outcomes [25].  Although the trials were associated with differences in EOI MRD responses, the survival rates were similar. Those with an EOI MRD of <10 À4 or those with an EOI MRD of !10 À4 e<10 À3 with sequential decrease in MRD had the best outcomes, possibly reflecting sensitivity of residual cells to concurrent therapy. Thus, in HR relapses, serial assessment of MRD may better guide subsequent therapeutic interventions. Pre-SCT MRD levels of !10 À3 [26,27] and !10 À4 [5,28] have been reported to be associated with poor outcomes after SCT. In this cohort, a pre-SCT MRD of <10 À4 and !10 À4 e<10 À3 was associated with comparable outcomes, and only those with an MRD of !10 À3 showed a significantly poorer outcome. Methodologically newer MRD assays (next-generation sequencing) may further increase the sensitivity of detecting lower levels of MRD [29]. Nevertheless, patients with low or absence of MRD before SCT relapsed after SCT in our cohort. MRD kinetics vary with different ALL subtypes, with MRD being less predictive of relapses in HR cytogenetic subtypes and T-ALL [30]. We speculate that in T-ALL, there may also be extramedullary reservoirs of disease [31] not readily assessed using marrow-based MRD assays only.
aGVHD appeared to be associated with a graftversus-leukaemia effect in BCP-ALL but not in T-ALL in this cohort. Patients with T-ALL also had a higher post-SCT therapyerelated mortality. A 3-year therapyrelated mortality of 30% with an OS of 48% has been reported for 229 patients with T-ALL who underwent transplantation in CR2, with no impact of acute or chronic GVHD on the outcome [32]. Our study does not have the required data to provide a satisfactory explanation for the apparent benefit of aGVHD in BCP-ALL or the increased transplant-related toxicity in T-ALL, and this requires further prospective evaluation.
Achieving a second remission remains a major problem. Addition of the proteasomal inhibitor bortezomib to induction therapy in relapsed ALL has been reported to achieve CR2 rates of 63e72%, benefitting both BCP-and T-ALL, and was well tolerated [33,34]. Of those reaching SCT, other than high MRD, recurrence was seen more frequently in those with a TCF3 rearrangement, TP53 del/mut or hypodiploidy [35]. Preclinical data suggest that TCF3-rearranged and hypodiploid ALL may be susceptible to the BCL-2 inhibitor venetoclax [36,37] and that TP53-rearranged ALL may be susceptible to a combination of APR-246 and doxorubicin [38].
Newer-generation CAR-Ts with enhanced expansion and long-term persistence may prevent post-SCT relapse or even replace the need for SCT. Nevertheless, disease recurrence in ALL is associated with a long latency, and careful long-term follow-up will be required. This study provides the background comparative data required to evaluate the benefit of the new drugs. A caveat is the high cost of immunotherapy and its non-availability to patients in low-and middle-income countries where most cases of ALL occur. For these newer therapies to make a significant impact to global outcomes of childhood ALL, this is a gap that needs to be bridged.