Association between Prior Cytotoxic Therapy, Antecedent Hematologic Disorder, and Outcome after Allogeneic Hematopoietic Cell Transplantation in Adult Acute Myeloid Leukemia

Simple Summary To assess the association between clinical history and AML outcomes in the context of allogeneic hematopoietic cell transplantation (HCT), we retrospectively analyzed 739 patients with de novo AML, 125 with antecedent hematologic disorder (AHD)/AML, and 115 with therapy-related AML who received first allografts while in first or second remission. Relative to patients with de novo AML, relapse rates were similar for patients with AHD and therapy-related AML after multivariable adjustment, as were relapse-free survival and overall survival. Non-relapse mortality was, however, higher for AHD AML. These data suggest that the clinical history by itself contains limited prognostic value for adults with AML undergoing allografting, supporting the most recent approach to use this information as a diagnostic qualifier rather than a disease entity. Abstract (1) Background: Secondary acute myeloid leukemia (AML), i.e., AML arising from prior therapy (therapy-related) and/or an antecedent hematologic disorder (AHD) is generally associated with worse outcomes compared to de novo AML. However, recognizing the prognostic importance of genetic characteristics rather than clinical history, secondary AML is now considered a diagnostic qualifier rather than a separate disease entity. (2) Methods: To assess the association between clinical history and AML outcomes in the context of allogeneic hematopoietic cell transplantation (HCT), we retrospectively analyzed 759 patients with de novo AML, 115 with AHD AML, and 105 with therapy-related AML who received first allografts while in first or second remission. (3) Results: At the time of HCT, these three cohorts differed significantly regarding many patient and disease-specific characteristics, including age (p < 0.001), gender (p < 0.001), disease risk (p = 0.005), HCT-CI score (p < 0.001), blood count recovery (p = 0.003), first vs. second remission (p < 0.001), remission duration (p < 0.001), measurable residual disease (MRD; p < 0.001), and conditioning intensity (p < 0.001). Relative to patients with de novo AML, relapse rates were similar for patients with AHD (hazard ratio [HR] = 1.07, p = 0.7) and therapy-related AML (HR = 0.86, p = 0.4) after multivariable adjustment, as were relapse-free survival (HR = 1.20, p = 0.2, and HR = 0.89, p = 0.5) and overall survival (HR = 1.19, p = 0.2, and HR = 0.93, p = 0.6). Non-relapse mortality was higher for AHD AML (HR = 1.59, p = 0.047). (4) Conclusions: These data suggest that the clinical history by itself contains limited prognostic value for adults with AML undergoing allografting, supporting the most recent approach to use this information as a diagnostic qualifier rather than a disease entity.

Given these mixed data, and the notion that post-HCT outcomes within the subset of patients with secondary AML might differ depending on the details of clinical history (e.g., whether therapy-related or after AHD, or the type of underlying AHD) [17,22], we examined a large cohort of adults with therapy-related AML and after AHD who underwent allogeneic HCT while in first or second remission at our institution between May 2006 and May 2021. In this analysis, we also assessed how the most recent changes in the classification for secondary AML might impact study findings.

Study Cohort
We included all adults with AML (2016 WHO criteria) [23] who proceeded to first allogeneic HCT while in first or second remission (i.e., had <5% marrow blasts) with or without peripheral blood count recovery between 5/2006 and 5/2021 at a single institution (Fred Hutchinson Cancer Center/University of Washington/Seattle Cancer Care Alliance, Seattle, WA, USA). High-dose fractionated total body irradiation (TBI; ≥12 Gy) with or without cyclophosphamide (CY) or fludarabine (FLU), high-dose TBI/thiotepa/FLU, busulfan (4 days) with CY or FLU, treosulfan/FLU with or without low-dose TBI, or any regimen containing a radiolabeled antibody were considered myeloablative conditioning (MAC) regimens [24]. Reduced-intensity conditioning (RIC) and non-myeloablative conditioning (i.e., 2-3 Gy TBI with or without fludarabine) were grouped together as non-MAC regimens. Data on post-HCT outcomes were obtained via the Long-Term Follow-Up Program from our outpatient clinic and local clinics that provided care for post-HCT patients; additionally, information was collected on patients in research studies. All treatments were on Institutional Review Board-approved research protocols (all registered within ClinicalTrials.gov) or standard protocols; patients gave informed consent in accordance with the Declaration of Helsinki. The cut-off date for follow-up was 10 February 2022.

Classification of Disease Risk and Treatment Response
Secondary AML was defined as disease following an AHD (i.e., myelodysplastic syndrome [MDS], myeloproliferative neoplasm [NPM], and MDS/MPN such as chronic myelomonocytic leukemia [CMML]) or treatment with systemic chemotherapy and/or radiotherapy for a different disorder. All previous cytotoxic regimens were considered for the main analysis, including methotrexate, mercaptopurine, and cyclophosphamide for auto-immune disease. Immunosuppressive treatment using nonchemotherapeutic agents was not considered cytotoxic. Patients developing MDS, MPN, or MDS/MPN between the chemotherapy or radiation treatment for their primary disease and the diagnosis of AML were classified as therapy-related AML unless stated otherwise. In sensitivity analyses, we explored the impact of new classification criteria for therapy-related AML (not including previous exposure to methotrexate, and, by analogy, mercaptopurine, and cyclophosphamide) and AHD (requirement for AHD diagnosis ≥3 months before AML diagnosis) [10,11]. The 2010 refined MRC/NCRI criteria were used to assess the cytogenetic risk at diagnosis [25]. The karyotype analysis was routinely based on 20 metaphases (if available); FISH studies were performed according to standard procedures in a subset of patients.

Statistical Analysis
Categorical variables were reported as numbers (with proportions) and compared using the Chi 2 test or Fisher's exact test, as appropriate. Continuous variables were reported as medians (with interquartile range [IQR]) and compared using the Mann and Whitney test. We estimated unadjusted probabilities of relapse-free survival (RFS; events: relapse or death) and overall survival (OS; event: death) with the Kaplan-Meier method and compared them with the Log-Rank test. Associations with RFS and OS were assessed using Cox regression models. Probabilities of relapse (with NRM as a competing event) and NRM (death without relapse with relapse as a competing event) were reported using cumulative incidence estimates. Associations with relapse and NRM were assessed using causespecific regression models [24]. All tests were two-sided with a significant level of p < 0.05. Statistical analyses were performed with R (R Foundation for Statistical Computing, Vienna, Austria; http://www.r-project.org).
The characteristics of the 979 patients included in the analysis are summarized in Table 1. Patients with AHD were older than patients with either de novo or therapy-related AML (63 vs. 53 vs. 57 years, p < 0.001). Due to the high number of therapy-related AML after treatment for breast cancers, there were more females in the therapy-related group (60% vs. 46% vs. 35% for patients with de novo and AHD, respectively, p < 0.001) and, as prior cancer is used to calculate the HCT-CI, patients in this group were more likely considered as high-risk according to this score (68% vs. 25% vs. 27%, respectively, p < 0.001). On the other hand, patients with AHD were less likely to be classified as favorable risk (0 vs. 8% vs. 11% for patients with de novo and therapy-related AML, respectively, p = 0.005) and they were more likely to have incomplete blood count recovery (43% vs. 28% vs. 34%, respectively, p = 0.003) and pre-HCT MRD (38% vs. 16% vs. 24%, respectively, p < 0.001) whereas the time from the last remission to HCT was shorter (72 vs. 104 vs. 96 days, respectively, p < 0.001). Most patients with secondary AML, either therapy-related or AHD, were transplanted in first remission (91% and 87% vs. 73% for patients with de novo AML, p < 0.001) and they were more likely to receive non-MAC regimens (53% and 56% vs. 36%, respectively, p < 0.001). As the definitions of sAML were recently changed [10,11], we also compared the three groups after reclassifying five patients with previous auto-immune conditions treated with either methotrexate (n = 3), and, by analogy, mercaptopurine (n = 1) or cyclophosphamide (n = 1), and 34 patients with AHD diagnosed ≤3 months before AML diagnosis (MDS, n = 28; CMML, n = 3; MPN, n = 3) as de novo AML. As summarized in Supplementary Table S2, the characteristics of the three patient subsets did not significantly change with this reclassification.

Relationship between Secondary AML Status and Post-HCT Outcome
After a median follow-up of 5.13 years after HCT among survivors (IQR: 2.30-9.01), there were 308 relapses, 460 deaths, and 191 NRM events. Relapse was non significantly higher in patients with AHD (35% [26-45%] vs. 23% [20-26%] vs. 28% [20-36%] at three years for those with de novo and therapy-related AML, respectively, Log-Rank test: p = 0.099; Figure 1) as was NRM (24% [15-32%] vs. 16% [13-19%] Figure S1). As the 34 patients not considered as having AHD in this last analysis were more likely to have prior MDS or not have been previously treated for the AHD, we explored whether these characteristics were associated with post-HCT outcomes. After considering all patients with AHD, including those with prior exposure to cytotoxic therapy, neither type of AHD nor previous therapy for AHD were associated with post-HCT outcome (Supplementary Figures S2 and S3).

AML after AHD as an Independent Prognostic Factor for Post-HCT Outcome
To study the relationship between AML after AHD and post-HCT outcomes in more detail, we evaluated univariable and multivariable regression models for the endpoints of NRM, relapse, RFS, and OS. In univariable analysis, AML after AHD was associated with NRM (hazard ratio [ (Table 2). Similar results were observed when limiting patients with AHD to those with a three-month interval between AHD and AML diagno-      Table 3). As AML after AHD was not independently associated with post-HCT outcomes other than NRM, we sought to identify variables that were associated with the outcome by performing univariable regression models in patients with AHD (n = 126). In these patients, adverse cytogenetic risk (HR = 1.  Table 4). As patients with AHD were significantly older, we specifically explored the association between age and post-HCT outcomes, dichotomizing patients into those younger vs. older than 60 years old. While the NRM for younger and older patients was similar (at three years: 22% [

Discussion
Recently introduced classification systems consider secondary AML as a diagnostic qualifier, in line with the increasing understanding that underlying genetic characteristics rather than clinical history largely account for worse outcomes seen in this patient subset [9][10][11]. Consistent with this notion, in our large retrospective analyses, we found very little evidence of an independent prognostic role for clinical history in multivariable analyses, ex-

Discussion
Recently introduced classification systems consider secondary AML as a diagnostic qualifier, in line with the increasing understanding that underlying genetic characteristics rather than clinical history largely account for worse outcomes seen in this patient subset [9][10][11]. Consistent with this notion, in our large retrospective analyses, we found very little evidence of an independent prognostic role for clinical history in multivariable analyses, except for a significant interaction between conditioning intensity and AML arising from AHD. Specifically, we found that compared to patients with de novo or therapy-related AML, those with AML arising from an AHD had a significantly higher risk of NRM if treated with MAC (but this did not translate into differences in post-transplant relapse rates or OS).
In the 2022 disease classifications (Internal Consensus Classification, WHO), some of the parameters for when exactly an AML should be considered secondary have changed. Specifically, it is now proposed that patients who developed AML after receiving cytotoxic therapy such as methotrexate for auto-immune conditions, or patients for whom the time interval between AHD and AML diagnosis was less than three months, should no longer be considered as secondary AML. Nevertheless, for our main analysis, we chose to consider these patients as secondary AML for comparability with prior studies [1,2,4,12,[16][17][18][19]22,40,41]. Complicating comparability, some of the older studies already used the 3-month time interval [5,15], but none classified AML cases after prior exposure to cytotoxic therapy such as methotrexate as de novo. However, we then performed sensitivity analyses considering these classification changes, which revealed a limited impact on the characteristics of each patient subgroup and the observed post-HCT outcomes. In these sensitivity analyses, our main findings were largely unchanged. As multilineage dysplasia is no longer recognized as a prognostic feature [42], these patients were considered as having de novo AML unless they were previously diagnosed with an AHD or had prior exposure to cytotoxic therapy.
Without adjustments, we observed that AML after AHD but not therapy-related AML was associated with post-HCT outcomes in our cohort. This finding is consistent with a recent report on a cohort of patients developing myeloid neoplasms after prior therapy for solid tumors [43]. The finding contrasts, however, with one previous report showing that both patients with AHD and therapy-related AML had inferior outcomes compared to patients with de novo AML [17]. A large part of these outcome differences could be attributed to differences in patient/disease characteristics. Patients with AHD were, on average, older and less likely to have favorable-risk cytogenetics but more likely to have pre-HCT MRD by MFC than the other patients. These differences explained why AHD was no longer associated with relapse or RFS after multivariable analyses [16,19], although one study previously observed an independent association between secondary AML and relapse [17].
As in previous reports [6,16,17], we observed in our study that secondary AML, and more specifically, AHD, was independently associated with NRM, suggesting that patients with AHD experienced more complications after transplant. Although there is no clear explanation for this observation with potential patient differences that are not fully captured in our multivariable adjustments, it is supported by the fact that the association between AHD and NRM was mostly observed in patients receiving MAC. As patients previously treated for AHD had similar outcomes to those who were not treated, increased exposure to pre-HCT therapies does not appear to have a role in increased NRM. Despite being older, patients with AHD had a similar number of comorbidities, as assessed by the HCT-CI. This may represent a bias toward selecting fitter subsets of patients for allografting. As stated, the association between AHD and NRM was attenuated after multivariable adjustment for other variables associated with increased NRM such as age.
Of note, conditioning intensity was not equally associated with post-HCT outcomes in patients with AHD. As in one previous report [16], but not another from the EBMT registry [44], we observed that patients with secondary AML receiving MAC were more likely to have increased NRM even after multivariable adjustment. In contrast to that previous report, but similar to a study from the EBMT registry [44], patients who received non-MAC had increased relapse and decreased RFS in our cohort, albeit not statistically significantly so after multivariable adjustment. Although we usually recommend that patients who are fit to tolerate MAC should receive a high-intensity conditioning regimen, more caution should be applied in patients with AHD. The intensity of conditioning was not associated with post-HCT outcomes in patients with therapy-related AML, which was observed in one study [45], but not in two others [46,47], in which relapse risk was higher in patients receiving non-MAC regimens.
The retrospective nature of our study analyzing post-HCT outcomes of patients for whom conditioning intensity was nonrandomly assigned limits our ability to draw definitive conclusions regarding the management of patients with AHD AML. Our general management has been to use MAC whenever it was felt that it could be safely administered based on patient age and comorbidities. Another limitation is the fact that mutational profiles were available only for a small subset of patients.

Conclusions
The fact that AML developed post cytotoxic therapy or after AHD provides little prognostic value (except for an increased risk of NRM in patients with AHD) supports the most recent proposal to use this information largely as a diagnostic qualifier rather than a prognostic variable.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/cancers15020352/s1, Figure S1: Post-HCT outcomes for 990 adults with AML undergoing allogeneic HCT while in first or second morphologic remission, stratified by disease status at diagnosis (de novo vs. therapy-related vs. antecedent hematologic disorder) after considering as de novo AML, five patients with auto-immune disorders treated with methotrexate, mercaptopurine, and cyclophosphamide and 34 patients with AHD diagnosed less than three months before AML; Figure S2 Figure S3: Post-HCT outcomes for 125 adults with AHD AML undergoing allogeneic HCT while in first or second morphologic remission, stratified by prior treatment for AHD (yes vs. no); Figure S4: Post-HCT outcomes for 143 adults with AML undergoing allogeneic HCT following reduced-intensity conditioning while in first or second morphologic remission, stratified by disease status at diagnosis (de novo vs. therapy-related vs. antecedent hematologic disorder); Figure S5: Post-HCT outcomes for 253 adults with AML undergoing allogeneic HCT following non-myeloablative conditioning while in first or second morphologic remission, stratified by disease status at diagnosis (de novo vs. therapy-related vs. antecedent hematologic disorder); Table S1: Primary disease and treatment-related characteristics in patients with therapy-related acute myeloid leukemia (n = 115) and after antecedent hematologic disorder (n = 125); Table S2: Pre-HCT demographic and clinical characteristics of study cohort (n = 979), stratified according to disease status at diagnosis (de novo vs. post cytotoxic therapy vs. after antecedent hematologic disorder [AHD]), after considering as de novo AML, five patients with auto-immune disorders treated with methotrexate, mercaptopurine, and cyclophosphamide and 34 patients with AHD diagnosed less than three months before AML.  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The datasets analyzed during the current study are available from the corresponding author upon reasonable request.