Therapy-related acute myeloid leukemia (t-AML) is a complex disease entity. It results from molecular abnormalities induced by chemotherapy, radiation and immunosuppressive therapies. As a group of diseases, t-AML may represent cases that progressed from therapy-related myelodysplastic syndromes (t-MDS) and "de novo" t-AML. The classification t-AML also includes patients (pts) whose AML is a second primary cancer e.g., due to a genetic predisposition to develop multiple, distinct cancers (those cases would be indistinguishable from the non-therapy-related AML). Origins of the disease may also vary: t-AML may evolve from clonal hematopoiesis of indeterminate potential (CHIP) that preceded the first cancer, as a de novo disease or as a disease which progressed from "de novo" CHIP. Comprehensive genomic analyses involving clonal hierarchy may reveal genetic patterns pointing towards a potential molecular pathogenesis.

We applied targeted gene sequencing to analyze a large cohort of pts with AML (n=2696) for the presence of somatic mutations: comparator subtypes include primary AML (pAML, n=2133) and secondary AML (evolving from an antecendent MDS; sAML, n=446) to be compared to t-AML (n=117). These pts have had a history of other primary malignancies for which they received cytotoxic treatments including chemotherapy and/or radiation.

t-AML pts were younger than pts with other AML types (median age: 60 years for t-AML vs. 65 and 69 for pAML and sAML). t-AML pts were more likely to have leukopenia compared to pAML (23% vs. 21%, P=0.7) but significantly less likely than pts with sAML (23% vs. 38% P=0.002). Normal cytogenetics were significantly less present in t-AML when compared to pAML (39% vs. 62% P<0.0001), but sAML had similar prevalence as t-AML (both 39%). Focusing on abnormal cytogenetics, t-AML pts had a significantly higher percentage of -5/del(5q) compared to pAML (14% vs. 7% P=0.004) but slightly less than sAML (19%). Similarly, -7/del(7q) and complex karyotyping were significantly more prevalent in t-AML compared to pAML (19 vs. 8% P<.0001) and (25% vs. 11% P<.0001) but similar to sAML (-7/del(7q): 19% vs. 16%; complex karyotype: 25% vs. 25%). In sum, cytogenetic analyses indicated similarities between t-AML and sAML, with both carrying more complex genetic events compared to pAML. While abnormal karyotyping was often more frequent in t-AML, molecular mutations were less frequently noted in t-AML vs. pAML (except for TP53 mutations). Strong predictors of pAML phenotype (e.g., CEPBA, NPM1) were less represented in t-AML vs. pAML (CEBPA, 3% vs. 9% P=0.01; NPM1, 18% vs. 34% P=0.0002). A typical signature of genes (DNMT3A, ASXL1, TET2, TP53) representing a phenotype of CHIP has been associated with t-AML. We observed a lower percentage of mutations in ASXL1 (4% vs. 10% P=0.01), DNMT3A (20% vs. 30% P=0.02), and TET2 (10% vs. 17% P=0.04) in the t-AML population compared to pAML, likely due to the younger age of t-AML pts. TP53 mutations were the only CHIP mutations often enriched in t-AML when compared to pAML (16% vs. 8% P=0.001). On the other hand, when compared to sAML, t-AML had significantly more mutations affecting genes like DNMT3A (21% vs. 13% P= 0.03), FLT3 (30% vs. 9% P<0.0001), NPM1 (18% vs. 5% P<0.0001) and NRAS (16% vs. 9% P=0.04) but significantly fewer lesions in ASXL1 (4% vs. 22% P<0.0001), BCOR (2% vs. 7% P=0.02), RUNX1 (12% vs. 22% P=0.01) and SRSF2 (7% vs. 19% P=0.002). In general, mutation types were most commonly missense and predominantly transversions within t-AML pts. Clonal hierarchy analyses demonstrated that TP53 mutations were acquired before the acquisition of other molecular mutations, as they were often ancestral in t-AML compared to pAML (15% vs. 8% P=0.007). As a consequence, they were the only molecular events carrying a dismal prognosis, as they significantly impacted the median survival of pts with t-AML compared to pAML (6 mo. vs. 17 mo.; P=0.004). On the contrary, when compared to sAML, CBL (5% vs. 1% P=0.03), FLT3 (5% vs. 1% P=0.03) and NRAS (10% vs. 3% P=0.007) were significantly more represented as ancestral lesions in t-AML but did not impact survival outcomes.

In sum, cytogenetic and molecular features of t-AML were more similar to those of sAML rather than pAML. TP53 was the most commonly mutated gene and carried a dismal prognosis, possibly representing a selective growth advantage after receiving chemotherapy and/or radiation therapies for primary cancers.


Nazha:Novartis: Speakers Bureau; Abbvie: Consultancy; Daiichi Sankyo: Consultancy; Incyte: Speakers Bureau; Jazz Pharmacutical: Research Funding; MEI: Other: Data monitoring Committee; Tolero, Karyopharma: Honoraria. Meggendorfer:MLL Munich Leukemia Laboratory: Employment. Haferlach:MLL Munich Leukemia Laboratory: Employment, Equity Ownership. Sekeres:Millenium: Membership on an entity's Board of Directors or advisory committees; Celgene: Membership on an entity's Board of Directors or advisory committees; Syros: Membership on an entity's Board of Directors or advisory committees. Maciejewski:Alexion: Consultancy; Novartis: Consultancy.

Author notes


Asterisk with author names denotes non-ASH members.