Clinical outcome of therapy‐related acute myeloid leukemia patients. Real‐life experience in a University Hospital and a Cancer Center in France

Abstract Background t‐AML occurs after a primary malignancy treatment and retains a poor prognosis. Aims To determine the impact of primary malignancies, therapeutic strategies, and prognostic factors on clinical outcomes of t‐AML. Results A total of 112 adult patients were included in this study. Fifty‐Five patients received intensive chemotherapy (IC), 33 non‐IC, and 24 best supportive care. At t‐AML diagnosis, 42% and 44% of patients presented an unfavorable karyotype and unfavorable 2010 ELN risk profile, respectively. Among treated patients (n = 88), 43 (49%) achieved complete remission: four out of 33 (12%) and 39 out of 55 (71%) in non‐IC and IC groups, respectively. With a median follow‐up of 5.5 months, the median overall survival (OS) and disease‐free survival (DFS) for the whole population were 9 months and 6.3 months, respectively, and for the 88 treated patients 13.5 months and 8.2 months, respectively. Univariate analysis on OS and DFS showed a significant impact of high white blood cells (WBC) and blast counts at diagnosis, unfavorable karyotype and ELN classification. Multivariate analysis showed a negative impact of WBC count at diagnosis and a positive impact of chemotherapy on OS and DFS in the whole population. It also showed a negative impact of previous auto‐HCT and high WBC count on OS and DFS and of IC on OS in treated patients which disappeared when we considered only confounding variables (age, previous cancers, marrow blasts, and 2010 ELN classification). In a pair‐matched analysis comparing IC treated t‐AML with de novo AML, there was no difference of OS and DFS between the two populations. Conclusion We showed, in this study that t‐AML patients with unfavorable features represented almost half of the population. Best outcomes obtained in patients receiving IC must be balanced by known confounding variables and should be improved by using new innovative agents and therapeutic strategies.


| INTRODUCTION
Therapy-related myeloid neoplasms (t-MNs) including therapy-related acute myeloid leukemias (t-AML) and therapy-related myelodysplastic syndromes (t-MDS) are considered as a distinct category in the 2016 revised World Health Organization (WHO) classification. 1 This entity is characterized by myeloid neoplasms emerging after treatment for a primary malignancy and associated with a poor prognosis. 2,37][8] The most common subtype occurs 5-7 years after first exposure to alkylating agents or radiation, is often preceded by MDS, and frequently accompanied by chromosomes 5 and/or 7 abnormalities, complex karyotype, and TP53 mutations. 9,10Otherwise, t-AML after topoisomerase II inhibitors occurs 1-3 years after first exposure, rarely preceded by MDS and usually associated with multiple gene rearrangements. 11,12Results comparing t-AML/MDS and de novo AML/MDS, have been conflicting so far. 13Because of similarity regarding cytogenetic features of de novo MN and t-MN, patients have been treated for a long time in the same way. 14Nowadays, according to knowledge of molecular markers and innovative therapies, new strategies were proposed regarding de novo and t-MN.
Because over time, the incidence of t-MN is increasing as more patients survive their primary cancers, a better description of the biologic and prognostic factors associated with this entity is highly warranted. 15,16Survival of patients with t-AML remains poor and some retrospective studies have identified many factors that contribute to poor clinical outcomes. 17,18In this subset of patients, there is no clear guidelines regarding frontline treatment strategies which are usually based on IC 13 or non-IC as hypomethylating agents (HMA) associated to Bcl2 inhibitors 19 and allogeneic hematopoietic cell transplantation (Allo-HCT). 20,21Therapy-related AML patients have often been excluded from frontline clinical trials until availability of liposomal daunorubicin and cytarabine association (Vyxeos®) which demonstrated encouraging results for this subset of patients. 22he aim of this retrospective study is to describe general characteristics and treatment modalities of primary malignancies of 112 consecutive t-AML patients, from two French centers (one University Hospital and one Cancer Center), who received chemotherapy and/or radiotherapy for previous neoplastic and non-neoplastic diseases and to analyze their clinical outcomes according to treatment modalities and prognostic factors.

| General and biological characteristics of patients
Between 01/01/2006 and 01/06/2019, we analyzed retrospectively, in two French centers, 112 adult patients diagnosed with t-AML according to 2016 revised WHO criteria 1 and occurring after treatment of solid tumors (ST), hematologic malignancies (HM), or autoimmune diseases (AID).At leukemia diagnosis, we performed peripheral blood count and bone marrow morphology with blast percentage.Samples of bone marrow and blood were also examined for cytogenetic abnormalities.Cytogenetic and molecular risk profiles were recorded according to 2010 European Leukemia Net (ELN) classification. 23Next generation sequencing (NGS) 24 was not available in our two centers until 2016 and was done subsequently for only 13 patients which explained that 2017 and 2022 ELN risk stratifications 25,26 by genetics were not used in our study.Mixed-lineage leukemia (MLL) gene fusion and MECOM1 positivity were detected by fluorescence in situ hybridization (FISH).Since 2006, available leukemia samples were analyzed for mutations in the FLT3 protein domains (TKD and ITD), NPM1, IDH1/2, and Evi1 overexpression, using real-time quantitative reverse transcriptase-polymerase chain reaction (RT-PCR). 27

Conclusion:
We showed, in this study that t-AML patients with unfavorable features represented almost half of the population.Best outcomes obtained in patients receiving IC must be balanced by known confounding variables and should be improved by using new innovative agents and therapeutic strategies.

K E Y W O R D S
clinical outcome, real-life management, therapy-related AML

| Treatment of previous malignancies and t-AML
Patients received chemotherapy and/or radiotherapy as treatment for their primary malignancy and for some of them, high dose chemotherapy followed by autologous hematopoietic cell transplantation (Auto-HCT).All patients eligible for IC were treated by an anthracycline and cytarabine-based induction chemotherapy regimen (N = 55 including CPX-351 in seven patients).Patients achieving complete remission (CR) after one or two courses of induction received consolidation chemotherapy.Patients unfit for intensive chemotherapy received frontline non-IC: nine low dose cytarabine (LDC) and 24 hypomethylating agents (HMA).All HMA patients received azacytidine.Policy regarding transfusions, antibiotics, and antifungal prophylaxis were determined according to established local practice.Twentyfour patients received best supportive care (BSC) with for some patient's administration of hydroxyurea or 6-mercaptopurine.

| Therapy-related AML response criteria
CR was defined according to 2017 ELN criteria. 25Patients in partial response or failure after treatment were considered as nonresponders.Measurable residual disease (MRD) response by flow cytometry (FCM) or RT-qPCR, when possible, has not been included in this study.

| Statistical analysis
We performed a global descriptive analysis of the baseline characteristics of the cohort and a specific analysis of each treatment group.Continuous variables were summarized as median (min; max) and [interquartile range] and compared between groups with a Kruskal-Wallis test.Qualitative variables were synthetized as counts and percentages and compared between groups by a Pearson's chi-square test (with Monte Carlo simulations if one count is below 5).
Overall survival (OS) was defined as the time from diagnosis to death at last follow-up and disease-free survival (DFS) as the time since diagnosis to death or relapse date.Patients with no events or allografted before whichever event were censored.Survival curves were illustrated with the Kaplan-Meier method and compared with a log-rank test.Univariate and multivariate analyses on survival were done using Cox regression.Relapse was illustrated thanks to the cumulative incidence (CI) function.Univariate and multivariate analyses on CI of relapse was done using the Fine and Gray regression.
The bilateral level of significance was set at 5%.All the statistical analyses and graphics were made under the R program (v3.5.1) and its "survival," "cmprsk," and "gg-plot2" packages.
A pair-matched control analysis of t-AML patients treated by IC with de novo AML also treated by IC was matched for age, cytogenetic features, and 2010 ELN classification.
Therapy-related AML after AID had similar characteristics to ST and HM populations in terms of delay of t-AML occurrence, demographic and genetics features, response to therapy and outcome.At t-AML diagnosis, cytogenetics was unfavorable in 44 patients (42%) with 28 (64%) complex karyotypes and 16 no complex karyotypes (three patients with 5q deletion, five with monosomy 7, three with 7q deletion, two with both chromosome abnormalities, and three patients with other abnormalities).Nine patients presented chromosome 17 abnormalities (two deletions and seven monosomies) and all belonged to complex karyotypes.Considering t-AML karyotype, among intermediate risk patients, 70% were observed after ST, 17% after HM, and 13% after AID.In unfavorable risk patients, 59% were observed after ST and 41% after HM. (p = 0.03).

| Clinical outcome
With a median follow-up of 5.5 months (0-144), the median OS for the whole population and DFS for the treated population were 9 months (5.9-13.5)and 6.3 months (5.3-10.3),respectively (Figure 1A,B).There was a significant difference between the three groups (entire population: BSC vs. non-IC vs. IC) in terms of OS (p < 0.001) as well as between IC (median of 14.8 months) and non-IC (median of 11 months) groups (p = 0.02) (Figure 1C).There was no difference between IC (median of 10.3 months) and non-IC (median of 8 months) groups in terms of DFS (Figure 1D).Univariate analysis showed a significant impact of the act of receiving chemotherapy, high WBC and high blast counts, unfavorable karyotype, and 2010 ELN (Table 3).Overall survival (p = 0.035) and DFS (p = 0.326) according to 2010 ELN classification were shown in Figure 2B.
In multivariate analysis of the entire cohort (Table 3A), we observed a significant negative impact of WBC count >4 Giga/L on OS (p = 0.02) and a significant positive impact of the act of receiving chemotherapy on OS (p < 0.001).
In multivariate analysis including only treated patients (Table 3B), previous Auto-HCT had a significant negative impact on OS (p = 0.028) and DFS (p = 0.045), WBC count >4 G/L had a significant negative impact on OS (p = 0.03) and DFS (p < 0.001) and type of chemotherapy (IC vs. non-IC) had a significant negative impact on OS (p = 0.026).In order to have an unbiased estimate of the effect of chemotherapy intensity, we performed univariate and multivariate analyses on OS concerning only confounding variables (age, number of previous cancers, marrow blast count, and 2010 ELN classification).We observed a persistent significant impact of chemotherapy intensity in univariate which disappeared in multivariate analysis (Table S1).

| Results of pair-matched analysis
In order to compare clinical outcome of de novo AML (n = 117) and t-AML (n = 55) treated by IC, we performed a pair-matched analysis.Characteristics of the two matched cohorts were not significantly different although a difference of age between the two cohorts was observed but which did not reach a significant level (Table S2).Median OS was 16.5 months (12.7-24.9)and 14.8 months (11-32.6) in de novo and t-AML, respectively (p = 0.345).Median DFS was 9.45 months (6.8-13.7)and 9.9 months (6.6-15.8)for de novo and t-AML, respectively (p = 0.708) (Figure 4,B).In addition, there was no difference concerning 2010 ELN classification between t-AML and de novo AML when we analyzed the different prognostic groups (Figure S2).CI of relapse were 47.6% (36.7-58) and 33% (19.5-46.8)at 2 years and 51% (40-62.3)and 38.5% (24-53) at 5 years for de novo and t-AML patients, respectively, with a lower CI of relapse for t-AML (Figure 5).

| Allogeneic HCT results
According to international recommendations for HCT in AML patients, 28 15 patients [five males and 10 females with a median age of 53 years (21-65)] underwent Allo-HCT in first CR.Before transplantation, 14 patients have received IC, one patient HMA and no patient received CPX-351.As conditioning regimen, patients received either myeloablative (MAC) (n = 2) or reduced intensity conditioning (RIC) (n = 13) including antithymocyte globulins (ATG) in nine patients.Hematopoietic cell sources were two bone marrow, one cord blood unit, and 11 peripheral hematopoietic cells, from six geno-identical donors, four unrelated HLA matched donors, and four unrelated HLA mismatched donors.After transplantation, two patients presented a graft failure and among 13 evaluable patients, 12 achieved a full donor chimerism and one a mixed chimerism.Six out of 13 patients developed acute graft-versus-host disease (GVHD) (one grade I, three grade II, and two grade III) and six developed a chronic GVHD (one extensive and five limited).The best response after transplantation was CR in 10 patients.At the last follow-up, four patients are alive (three in CR and one in relapse) and 11 died (five from relapse, five from transplant related mortality causes, and one from another subsequent malignancy).

| DISCUSSION
In our study and as described in several studies, 17,18,29 50%-60% of patients had a ST as primary malignancy with a majority of breast cancer and 30%-40% HM with a majority of lymphoproliferative diseases before occurrence of t-AML.All patients, as in our analyzed population, received chemotherapy combinations with alkylating agents, topoisomerase II inhibitors, platinum compounds, and/or radiotherapy with variable median interval between primary malignancy and t-AML 15,17,29,30 Recent study from AML Swedish registry have showed, among patients treated for ST, lymphoma, and multiple myeloma (MM), 4.5% yearly overtime increase of incidence of t-AML. 313][34][35][36] Singh et al have evaluated risk of t-MN in a very large series of ST and MM patients during three periods of time (from 2000 to 2016), except in melanoma, authors observed in the most recent period, an increase rate of t-MDS and decrease rate of t-AML probably due to introduction of nonchemotherapy innovative agents. 37A study of t-MN after radiotherapy alone done by Patel et al has shown that the most common cytogenetic abnormality was a clonal abnormality in chromosome 5 and/or 7 with a worse outcome compared with other cytogenetic groups. 38Concerning t-MN after innovative agents, patients treated for ST, with different peptide receptor radionuclide therapy (PRRT) 39 and more recently, with Poly (ADP-ribose) polymerase (PARP) inhibitors, developed also high risk t-MN. 7,40Several studies highlighted the role of genetic mutations which belong to CHIP, found at ST diagnosis, growing within malignancy history and exposure to risk factors with a proven impact on t-MN development. 6,7,8,10,41,42,43,44Concerning general characteristics and genetic features of t-AML patients, as others, we reported more frequent abnormal karyotypes compared to de novo AML 2,15,17,18,45,46 and no difference in the incidence and distribution of mutated NPM1 and FLT3-ITD between t-AML and de novo AML. 47TP53 mutations or 17p abnormalities are frequently observed in t-MNs and associated with complex karyotype, a reduced probability of achieving response after induction chemotherapy and a poor outcome, as shown by Lindsley et al. 48Nowadays, prognosis of TP53 mutated t-AML, according to new AML classifications 26 must consider variant allelic frequency, single or multihit characteristic and presence of comutations. 49Due to NGS nonavailability at the beginning of our study, we reported only few results of genomic assessment in 13 patients with 2 or more mutations and TP53 mutations in 46% co-occurring, in half of them with chromosome 17 abnormalities.Concerning treatment of t-AML, a study about 931 older patients with secondary AML using IC-based regimens compared to epigenetic and low dose cytarabine regimens did not demonstrate a statistically significant difference in OS except for patients who underwent Allo-HCT. 50In our study, we observed positive impact of treatment with a better median OS than in Chen and Chicago group reports. 15,45Nowadays, patients with t-AML and AML-MRC could receive CPX-351 which allowed promising results 22 and specially for patients who underwent Allo-HCT.In our study, patients receiving IC had a significant higher OS and DFS in univariate but not in multivariate analysis, compared to those treated by non-IC after adjustment on confounding variables.Nevertheless, we have to mention that, at that period of time, patients in non-IC group, have received only HMA.Actually, AML including t-AML patients who are not eligible for IC are treated by combinations of HMA or cytarabine with Bcl2 inhibitors with or without targeted therapies (FLT3 inhibitors, IDH1 and 2 inhibitors, and anti-CD47 antibody). 19,51,52In addition, magrolimab and APR-246 showed interesting results in TP53 mutant t-MN. 51,53In this real-life study, we integrated BSC patients, frequent in this category of AML, who should be included in future AML studies.In Allo-HCT for t-AML setting, a positive impact of CPX-351 on OS and TRM was observed. 21,54Considering Allo-HCT, we showed in our study a usual GVHD incidence but a high balanced mortality rate related to either toxicity or relapse.A comparative study between sAML/t-AML and de novo AML undergoing Allo-HCT showed after adjustment for ELN risk and pre-HCT MRD status that disease subtype did not impact outcome. 55I G U R E 3 Cumulative incidence of relapse in treated population.
More recently, Baranwal et al had identified a very high-risk score (TP53/BCOR/IDH1/ GATA2/BCORL1) for Allo-HCT outcome with survival of 0% versus 64.6% (p = 0.001). 56ur results do not suggest t-AML as an independent poor prognostic factor, because of the similarity of OS between t-AML and de novo AML.This is related to a matching of IC-treated patients on age, cytogenetics, and ELN2010 in the pair-matched analysis of t-AML and de novo AML.These results are similar to those obtained by Chen et al. who also performed a pair-matched analysis between de novo AML and t-AML after breast cancer.
This OS similarity is controversial in other studies as German-Austrian AMLSG registry study 18 and PETHEMA registry study 57 who did not perform a pairmatched analysis between t-AML and de novo AML and included all consecutive newly diagnosed AML.These two studies showed difference of characteristics between the two populations with older patients and more frequent abnormal cytogenetics in t-AML than in de novo AML.The only similarities between all these studies is that t-AML patients had less frequently NPM1 mutations and FLT3-ITD.Due to pair-matched analysis results observed, we maintain that t-AML seem not an independent risk factor of poor outcome in younger patients who received IC.
Our study has some limitations related to its retrospective design, relatively small number of patients studied, and NGS nonavailability.Focusing on pair-matched analysis, our results were controversial with some literature data, suggesting that t-AML is not an independent risk factor for poor prognosis but rather an accumulation of patients whose probability of harboring other risk factors is higher.The recent ELN2022 update of AML providing an interesting hierarchical classification should allow to better characterize t-AML and prove if this entity is an independent risk factor for poor prognosis or not.
In conclusion, the best therapeutic strategy for patients with t-AML needs to take into consideration age, performance status, prognostic factors while integrating newly available treatment.For older t-AML patients and young patients with comorbidities not eligible for IC and/or Allo-HCT, HMA or cytarabine associated to venetoclax with or without targeted therapies (FLT3, IDH inhibitors, or anti-CD47 antibody especially for TP53 mutated patients) seems to be the best therapeutic option.Considering patients eligible for Allo-HCT, to decrease toxicity and improve results after transplantation, prior CPX-351 based chemotherapy should be privileged.Finally, the very poor results of the BSC strategy in t-AML and availability of new therapeutic approaches, lead us to ensure an exhaustive analysis of noneligibility criteria of t-AML patients before allocating them to BSC strategy.

F I G U R E 1
Kaplan Meier curves evaluatiing OS and DFS (A) OS of whole population, (B) DFS of treated population, (C) OS of whole population according to treatment groups, (D) DFS of treated population according to treatment intensity.

F 3
Multivariate analysis on OS (A) for total population (n = 112 patients) and OS and DFS (B) for treated population (IC/Non-IC) (n = 88 patients).F I G U R E 2 Kaplan-Meier curves evaluatiing OS and DFS according to 2010 ELN classification (A) OS of whole population, (B) DFS of treated population.

F I G U R E 4
Kaplan-Meier curves evaluating OS and DFS of pair-matched populations (IC treated AML vs de novo AML), (A) OS of t-AML and matched de novo AML, (B) DFS of t-AML and matched de novo AML.

F I G U R E 5
Cumulative incidence of relapse of pair-matched populations.
Descriptive analysis of total population according to therapeutic strategy.
T A B L E 1
T A B L E 2