Therapy-Related AML: Clinical/Biological Features and Long-Term Outcome in a Cohort of 1133 Adult AML Patients

Introduction: Therapy-related acute myeloid leukemia (t-AML) affects approximately 7% of AML patients and exhibits adverse genetic features that are assumably caused by increased DNA damage after antecedent radiation/chemotherapy. There is an ongoing medical need for a better understanding of differences between patients with t-AML and de novo AML. Here, we assess clinical and biological characteristics of patients with t-AML and compare long-term survival with de novo AML in a large cohort of AML patients.

Methods: 225 t-AML patients who were newly diagnosed between 1995-2018 were retrospectively analyzed and compared to a cohort of 908 patients with de novo AML. Mann-Whitney U- and Chi square test were used to assess differences in baseline characteristics between subgroups. Kaplan Meier procedure was applied to calculate overall survival (OS). Cox regression was used to identify hazard ratios (HR) of independent prognostic risk factors. Non-relapse mortality (NRM) and risk of relapse (RR) were calculated by the cause-specific hazard method.

Results: As compared to de novo AML, t-AML patients were older (61 vs. 55 years, p<0.001), had increased comorbidity and a higher frequency of adverse genetic alterations (e.g. monosomal or complex karyotypes, -5/-5q, -7, t(v;11)(v;23)). Prior to t-AML, solid cancer was diagnosed in 61% of patients (138/225), whereas 31% had hematologic malignancies (69/225) and 8% had autoimmune diseases (18/225). Previous malignancies had been treated with combined radiochemotherapy in 33% (75/225), chemotherapy in 27% (60/225), radiation therapy in 22% (50/225), immunosuppressive therapy (Cyclophosphamide, Mitoxantrone, Methotrexate, Azathioprine) in 7% (16/225), radio-iodine therapy in 4% (9/225) or multiple preceding cytotoxic therapies in 7% (15/225). Median latency period between cytotoxic therapy and t-AML was 5.8 years. In intensively treated t-AML patients (172/225), median OS was 13.7 months as compared to 39.4 months in de novo AML (823/908, p<0.001). The substantial difference in median OS particularly applied to patients <60 years with a 5-year OS of 31% in t-AML (96/225) as compared to 51% in de novo AML (562/908, p<0.001). In patients ≥60 years, there was a tendency towards better OS in de novo AML, however, this did not reach statistical significance (p=0.066). Interestingly, OS in patients receiving intensive therapy for t-AML was comparable with de novo AML within the favorable genetic risk groups (APL p=0.927, ELN-favorable p=0.714). In contrast, a significant difference in favor of de novo AML was observed within the ELN-intermediate (p=0.009) and ELN-adverse (p=0.016) risk groups, which was most likely caused by age differences between these groups. However, there was no age difference within the favorable ELN risk group (p=0.878). Considering only patients who received non-intensive therapy, OS (p=0.394), age (p=0.087), ECOG status (p=0.123) and ELN risk group (p=0.098) did not differ significantly between t-AML (53/225) and de novo AML (85/908). Multivariate analysis (including several covariates such as age, ELN risk, ECOG and CCI) did not confirm t-AML per se as an independent risk factor for OS (p=0.081), RR (p=0.982) or NRM (p=0.320) in intensively treated patients. Interestingly, OS was particularly favorable after radio-iodine therapy as compared to other treatment modalities (p=0.042) and highly unfavorable after pretreatment with multiple cytotoxic therapies (p=0.009). Multivariate analysis revealed ECOG score >1 (HR 2.3, p=0.002), adverse ELN risk (HR 1.7, p=0.003), BMI <18,5 kg/m² (HR 2.7, p=0.029), diabetes (HR 2.0, p=0.029) and intensive treatment (HR 0.3, p<0.001) as independent prognostic factors in t-AML.

Conclusion: Our data indicate that inferior OS in t-AML as compared to de novo AML rather results from an accumulation of adverse genetic alterations and patient-associated risk factors than from t-AML itself. Thus, our study strongly supports the current view that risk stratification in t-AML (including indication for allo-HSCT) should align with strategies used in de novo AML patients (based on genetic features), both in the case of intensive and non-intensive therapy. Regarding the new International Consensus Classification of AML, our data are in accordance with the classification of t-AML as a “diagnostic qualifier” within the different AML subgroups rather than a separate subcategory.