Primary Induction Failure Acute Myeloid Leukemia: Immediate Use of Allogeneic Stem Cell Transplantation Vs. Further Induction Chemotherapy

Several studies estimate that primary induction failure (PIF), meaning persistent leukemia following intensive induction chemotherapy (1,2) occurs in approximately 10-40% of patients with acute myeloid leukemia (AML). The lack of response to treatment is considered a major predictor for poor outcomes. Curative treatment is unlikely to be achieved without use of allogeneic stem cell transplantation (ASCT) (3). There has been debate about the importance of achieving complete remission (CR) prior to ASCT at the cost of toxicity related to further treatment.

Thus, we performed a retrospective single center analysis of PIF AML patients who underwent ASCT, comparing those who proceeded immediately to ASCT or those who received further remission induction to achieve CR beforehand. The primary outcome was progression-free survival (PFS).

PIF was defined as the lack of CR following either two cycles of 7+3 induction, one cycle of FLAG-Ida or CLIA. Based on timing and disease status at ASCT, we generated four cohorts to be compared. Cohorts 1 and 2 included patients who proceeded directly to ASCT, however, cohort 1 included patients with bone marrow (BM) <5% blasts (n=78) whereas cohort 2 included patients with BM blasts ≥5% (n=38). Cohorts 3 and 4 continued to receive additional lines of treatment prior to ASCT, however, cohort 3 had BM blasts <5% (n=49) and cohort 4 had BM blasts ≥5% (n=31). Patients without count recovery but <5% BM blasts were included.

The four comparison cohorts had a similar median age at transplant: 61, 54, 57 and 56 years for cohorts 1, 2, 3 and 4 respectively. In cohorts 1 and 3, 39 (50%) and 30 (61%) patients were MRD positive by multiparameter flow cytometry (MFC). Donor sources for ASCT were fairly consistent across the groups with matched sibling donor (MSD), matched unrelated (MUD) and haploidentical donors accounting for 26 (33%), 39 (50%), 12 (15%) in cohort 1; 12 (32%), 18 (47%), 7 (18%) in cohort 2; 11 (22%), 23 (47%), 12 (24%) in cohort 3; and 9 (29%), 14 (45%), 5 (16%) in cohort 4. Median time from diagnosis to transplant differed as expected with 5.6, 5.9, 8.8 and 9.1 months for cohorts 1-4, respectively. Myeloablative conditioning (MAC) was used in 53%, 63%, 57%, 55% for cohorts 1-4, respectively.

The best PFS was observed in cohort 1 who proceeded directly to ASCT with BM blast<5% with 5- year estimate of 40%. Cohort 2 (proceeded directly to ASCT with BM blast ≥5%) had 25%; while cohorts 3 and 4 who received additional chemotherapy before ASCT had a 5-year PFS of 32% and 24% respectively. Similarly, the risk of relapse at 5-years was lowest for cohort 1 with cumulative incidence of 37%, cohort 2 who had BM ≥ 5% at ASCT had an incidence of 65% while cohorts 3 and 4 had incidence of 52% and 55% respectively. The NRM at day 100, at 1-year and 2-years were similar for the cohorts.

PFS estimates were stratified by ELN 2017 risk classification by adverse risk and non-adverse (intermediate and favorable) risk. In the adverse risk group, best outcomes were observed in cohorts 1 and 2 who immediately proceeded with ASCT independent of BM blast count: 5- year PFS was 32% and 27% (HR-1.1 [0.6-2.1], p=0.7). Cohorts 3 and 4 who received more chemotherapy had 5-year estimates of 17% (HR-1.8 [1.0-3.1], p=0.04) and 21% (HR-1.8 [0.9-3.4], p=0.08) compared with cohort 1, respectively. This suggests that in the adverse risk AML PIF patient, immediate ASCT might be preferable rather than proceeding with more treatment to achieve better response before ASCT.

In the non-adverse risk group, the best outcomes were in cohorts 1 and 3 with BM blasts <5%: cohort 1 had 5-year PFS of 49% and cohort 3 had an estimate of 52% (HR=0.9 [0.4-2.2], p=0.9). Cohorts 2 and 4 with BM blasts ≥5% at ASCT had 5-year PFS of 15% (HR-2.1 [0.9-4.8], p=0.07) and 25% (HR-2.6 [1.1-5.9], p=0.02) compared with cohort 1, suggesting that in non-adverse risk PIF AML patients may benefit from more intensive therapy with an aim to achieve BM blasts <5% prior to ASCT.

Timing of transplantation in PIF AML patients is dependent on treatment response as well as ELN risk classification. Our data suggests that in adverse-risk AML patients with PIF, immediate ASCT is preferable independent of blast count. Transplant decisions need to be individualized per patient.