Intensive chemotherapy for acute myeloid leukemia relapse after allogeneic hematopoietic cell transplantation

Abstract

Outcomes for acute myeloid leukemia (AML) relapsing after allogeneic hematopoietic cell transplantation (HCT) are generally poor.1 Achieving remission is considered essential for long-term survival after post-HCT relapse. “Intensive” chemotherapy is typically used to achieve remission, but there is no standard regimen in this setting.1 While “high-dose” cytarabine (HDAC) plus purine analogue (PA) chemotherapy produces higher response rates than 7 + 3 in de novo and relapsed AML,2 there is much less evidence they improve overall survival (OS) or event-free survival (EFS) and may produce excessive toxicity without commensurate benefit for patients in post-HCT relapse. Also, it is unknown whether PAs carry added mortality risks after HCT (perhaps increasing vulnerability to infection through profound lymphodepletion or increased risk of graft-vs-host disease [GvHD], as has been observed in patients receiving PAs prior to donor lymphocyte infusion [DLI]3) or whether they may confer added anti-leukemic benefits (perhaps through rebound lymphocyte recovery with improved repertoire against leukemia-associated antigens4 and possibly less lymphocyte exhaustion1, 5). We sought to estimate the remission rates (primary endpoint), early mortality, EFS, and OS associated with HDAC, HDAC + PA, and non-HDAC/non-PA regimens for post-HCT AML relapse. We also examined how immunosuppression management affected the outcome of chemotherapy. Patients transplanted at Fred Hutchinson Cancer Research Center (FHCRC) were eligible for this retrospective analysis if they first presented with post-HCT MRD (detected by flow cytometry, cytogenetics or PCR) or overt relapse between January 1, 2005, and March 31, 2018 and received “intensive” relapse-directed chemotherapy. They were assigned to one of the three groups according to the initial chemotherapy administered for post-HCT relapse. HDAC was defined as individual doses ≥1 g/m2. Methods are detailed in the online supplement. One hundred and seventy-five patients were included, of whom 123 (70%) received HDAC (78 with and 44 without PA) and 53 did not. (Characteristics are in Table S1.) The most common treatments in the latter group were 7 + 3 and mitoxantrone/etoposide (Figure S1). Most (79%) were treated outside our center. Overall, the use of HDAC+PA increased in recent years, although there was an overrepresentation of non-HDAC and non-HDAC/non-PA regimens among those treated at our institution. Otherwise, there were no statistically significant intergroup differences pre-transplant, at transplant, at first post-HCT AML detection, or at initiation of the first course of intensive chemotherapy (Table S1). The remission rates with and without measurable residual disease (MRD) after a first course were 10% and 26%, respectively, while 54% had persistent morphologic disease and 10% died of infection or multiorgan failure before restaging. Remission rates did not differ significantly among chemotherapy categories (Figure S2). Multivariable analysis found that non-myeloablative conditioning was associated with higher odds of achieving remission (OR 3.92, 95% CI 1.52–10.50, p = .005), but found no significant effect of HDAC/non-PA (OR 1.23, 95% CI 0.46–3.33, p = .680) or HDAC/PA (OR 1.68, 95% CI 0.71–4.11, p = .246) versus non-HDAC/non-PA (Table S2). The supplement discusses concurrent (Supporting Section 3/Figure S3) and subsequent (Supporting Section 4/Tables S3 and S4) treatments. Forty-two patients received a second intensive chemotherapy course within 90 days of the initial course, n = 37 for morphologic disease and n = 5 for MRD. Of the 37 non-responding patients, after the second course, 21 (57%) remained in relapse, 4 (11%) attained MRD+ remission, 5 (14%) MRD-negative remission, and 7 (19%) died of infection or multiorgan failure before AML restaging. Switching regimens did not affect the chance of achieving remission or treatment-related mortality (TRM) (p = .591). Early mortality (death ≤28 days of chemotherapy initiation) occurred in 22 patients (12%; 95% CI, 8%–19%). It was no different among chemotherapy categories (p = .473) or among patients treated at FHCRC (4/36; 11%) versus elsewhere (18/139; 12%) (p = .988). The median EFS, defined as the time to morphologic relapse/documented disease persistence (N = 132, 75%) or death (N = 41, 23%), occurred at 35 days (95% CI, 30–45 days) after starting chemotherapy, again with no difference detected among chemotherapy groups (p = .316, Figure 1A). Figure 1B displays EFS landmark analysis starting 30 days post-chemo. The median EFS among patients with MRD-negative remission was an additional 150 days (95% CI, 91–308 days), and for MRD+ remission, an additional 104 days (95% CI, 61–398 days). Although there was no clear advantage associated with MRD negativity (HR 1.10, 95% CI 0.63–1.93, p = .735), consider that most (82%) MRD+ patients received further MRD-directed treatment and most (82%) patients achieving MRD-negative remission subsequently received consolidation. In landmark analyses accounting for remission depth and initiation of subsequent therapy, the initial chemotherapy approach had no significant effect on EFS (Table S5). Overall survival for the whole cohort was 32% (95% CI, 26%–40%) at 1 year and 18% (95% CI, 13%–25%) at 2 years post chemotherapy. The median OS was 188 days (95% CI, 153–248 days). Only 13 patients (8%) remained alive at last follow-up after a median of 57 months (range, 7.8–161 months). Median OS from chemotherapy initiation was 344 days (95% CI, 144–539 days) for those treated with HDAC/non-PA, 168 days (95% CI, 129–230 days) for HDAC+PA, 216 days (95% CI, 132–325 days) for those who received neither (p = .231; Figure 1C). In landmark analysis starting at day 30 post-chemotherapy, among patients with MRD-negative remission, median OS was an additional 308 days (95% CI, 212–609 days). For those with MRD+ remission, it was 325 days (95% CI, 158–718 days), and for those who did not respond to chemotherapy but were alive with morphologic disease, it was an additional 156 days (95% 123–240 days) (overall p = .223; Figure 1D). To better elucidate the “pure” effect of the initial chemotherapy regimen on OS and compensate for the fact that frail patients or those with severe comorbidities might have been disproportionately represented among those who died early post-chemo, we conducted a landmark analysis among 30-day survivors, adjusted for disease status after one chemotherapy course and initiation of subsequent AML-directed treatment or consolidation. There was no statistically significant difference between chemotherapy regimens (Table S6). Patients who received chemotherapy at FHCRC tended to be much earlier post-HCT than those treated elsewhere (median 88 vs. 444 days, p < .001). Yet, there was no substantive difference in EFS or OS according to treatment site (Supporting Section 7; cf. landmark analyses adjusted for unbalanced factors in Tables S7–S9). One-hundred and twenty-seven patients (72%) were receiving systemic immunosuppression when AML was first detected after HCT. All but 29 (16% of the cohort) tapered off immunosuppression before death or last follow-up. The immunosuppression-free interval (IFI) was calculated for those who discontinued immunosuppression before or up to 30 days after onset of chemotherapy. There was no statistically significant relationship between the IFI and the probability of post-chemo remission (p = .870 by permutation testing). Among 84 patients who discontinued systemic immunosuppression to treat relapse no earlier than 42 days before chemotherapy (IQR, −8 to +10 days peri-chemotherapy initiation), 1-year cumulative incidence of grade 2–4 acute GvHD was 20% (95% CI, 13%–31%), grade 3–4 acute GvHD 3.6% (95% CI, 1.2%–11%), and chronic GvHD requiring systemic immunosuppression 7.1% (95% CI, 3.3%–15%). The 1-year cumulative incidence of restarting immunosuppression was 26% (95% CI, 18%–38%) (Figure S4). In summary, we did not observe a material difference in outcomes between HDAC, HDAC+PA, and non-HDAC/non-PA chemotherapy regimens for patients suffering from post-HCT AML relapse. To our knowledge, this is the largest reported cohort to attempt to distinguish the effect of HDAC, PA, and alternative intensive chemotherapy regimens, and the only study of chemotherapy selection exclusively in the post-HCT setting. Our analysis benefits from granular detail on MRD, pharmacologic immunosuppression, and GvHD. The risk of a type 2 error (concluding there is no effect of chemotherapy selection despite an actual effect) persists, but the data presented here suggest that effect size is small. The natural limitations of retrospective studies include unmeasured confounders. For example, performance status and frailty metrics were not routinely available. Our work suggests that administering one cycle of intensive chemotherapy is reasonable, but fewer than half of patients will attain remission, which is generally considered a prerequisite for DLI or second transplant – the modalities most strongly associated with prolonged post-relapse survival.1 The risk of TRM from one cycle of chemotherapy is acceptable but not negligible. TRM may increase with a second intensive chemotherapy cycle, while the chance of remission is only about a quarter, so patients who do not respond to one cycle should consider alternative and investigational options that might have a more favorable risk:benefit ratio. These may include venetoclax-based combinations, novel cellular or antibody-based immunotherapies, tumor vaccines, or treatment based on functional or genomic drug selection assays. The data presented here provide useful real-world information about TRM and MRD-negative remission rates for HCT patients facing the choice of whether to proceed with intensive salvage chemotherapy. These data also provide historical benchmarks for researchers who are developing new strategies for post-HCT AML relapse. Dr. Eli Estey contributed extensively to this manuscript and passed away prior to submission. The authors have no conflicts of interest. All participants gave informed consent on Fred Hutch IRB-approved protocols for the use of their medical data and protected health identifiers in research. Appendix S1. Supporting information. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Topics & Concepts