Introduction

Acute myeloid leukemia (AML) is a clinically and biologically heterogeneous hematologic malignancy characterized by uncontrolled proliferation of hematopoietic precursors and loss of the ability to differentiate. Although the clinical outcome improves steadily in younger patients in the past 40 years, the survival for older patients remains very poor.1, 2

In addition to patient-specific factors, such as concomitant comorbidity, poor performance status and intolerance to intensive chemotherapy,3, 4 a number of leukemia-associated factors are related to the poor outcome in older AML patients.5, 6 Traditionally, cytogenetic findings establish the backbone for prognostic and therapeutical strategies in AML and are long used to risk-stratify AML patients and guide the treatment plan.7, 8 Appelbaum et al.5 demonstrated that older patients had less frequently favorable-risk but more commonly unfavorable-risk cytogenetics, particularly abnormalities in chromosomes 5, 7 and 17.

Many acquired gene mutations have been detected in AML patients, especially those with intermediate-risk cytogenetics, and some of them, such as mutations in NPM1, CEBPA, RUNX1, WT1, DNMT3A, ASXL1, IDH2 and FLT3 genes, have been shown to have prognostic significance.9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22 However, less is known about the clinical implications of gene mutations in older patients with AML. In this study, we aimed to comprehensively investigate the clinico-biological features and molecular genetic alterations and their clinical relevance in older AML patients. The findings from this study may pave ways for future targeted therapies in this group of patients with poor clinical outcome under the current treatment modality.

Materials and methods

Subjects

Totally, 462 adult patients who were newly diagnosed as having de novo non-M3 AML according to the FAB Cooperative Group Criteria23 at the National Taiwan University Hospital (NTUH) had cryopreserved cells for mutational analyses, and had complete clinical, cytogenetic and laboratory data were recruited for this study. Among them, 177 patients were 60 years or older. Patients with antecedent hematological diseases, history of cytopenia, family history of myeloid neoplasms or therapy-related AML were all excluded. Diagnosis and classification of AML were made according to the FAB Cooperative Group Criteria.23 This study was approved by the Research Ethics Committee of the NTUH and written informed consents were obtained from all participants in accordance with the Declaration of Helsinki. Among these patients, 329 (71.2%) received standard induction chemotherapy (Idarubicin 12 mg/m2 per day on days 1–3 and Cytarabine 100 mg/m2 per day on days 1–7) and then consolidation chemotherapy with two to four courses of high-dose Cytarabine (2000 mg/m2 q12h, total eight doses), with or without an anthracycline (Idarubicin or Mitoxantrone), after achieving complete remission (CR).20, 22 Because hypomethylating agents have not been reimbursed for the treatment of AML by the Taiwan government, only few patients received hypomethylating agents in this cohort; analysis of prognostic impact of hypomethylating agents was not carried out. The remaining patients received palliative therapy with supportive care and/or low-dose chemotherapy owing to underlying comorbidities or based on the decision of the physicians and patients.

Cytogenetics

Bone marrow cells were harvested directly or after 1–3 days of unstimulated culture as described previously.24 Metaphase chromosomes were banded by trypsin-Giemsa technique and karyotyped according to the International System for Human Cytogenetic Nomenclature.

Mutation analysis

Analyses of relevant mutations in 21 genes, including Class I mutations, such as FLT3/ITD,25 FLT3/TKD,26 NRAS,27 KRAS,27 JAK2,27 KIT28 and PTPN11(ref. 28) mutations, and Class II mutations, such as CEBPA29 and RUNX1(ref. 17) mutations, as well as mutations in NPM1,12 WT1,30 TP53,31 Cohesin complex genes (including STAG1/2, SMC1A, SMC3, and RAD21),32 and those genes related to epigenetic modification, such as MLL/PTD,33 ASXL1,34 IDH1,35 IDH2,36 TET2(ref. 37) and DNMT3A20 were performed as previously described. Abnormal sequencing results were confirmed by at least two repeated analyses.

Statistical analysis

The discrete variables of patients with and without specific molecular alteration were compared using the Fisher exact test. If the continuous data were not normally distributed, Mann–Whitney U tests were used to compare continuous variables and medians of distributions. To evaluate the impact of age, molecular alterations and other variables on clinical outcome, only the patients who received conventional standard chemotherapy were included in analyses.20, 22 Overall survival (OS) was measured from the date of first diagnosis to the date of last follow-up or death from any cause, whereas relapse was defined as a reappearance of at least 5% leukemic blasts in bone marrow aspiration smears or new extramedullary leukemia in patients with a previously documented CR.38 Disease-free survival (DFS) was applied to patients receiving standard intensive chemotherapy and was measured from the date of CR until relapse from CR or death from any cause, whichever occurred first. Multivariate Cox proportional hazards regression analysis was used to investigate independent prognostic factors for OS and DFS. The proportional hazards assumption (constant hazards assumption) was examined by using time-dependent covariate Cox regression before conducting multivariate Cox proportional hazards regression. The variables including age, white blood cell counts at diagnosis, karyotype, NPM1/FLT3-ITD, WT1, CEBPA, RUNX1, MLL/PTD, ASXL1, TET2, IDH2, DNMT3A and TP53 mutations that showed prognostic implication with P value less than 0.1 in univariate analysis were used as covariates. Those patients who received hematopoietic stem cell transplantation (HSCT) were censored at the time of HSCT in survival analysis to ameliorate the influence of the treatment. A P value <0.05 was considered statistically significant. All statistical analyses were performed with the SPSS 20 (SPSS Inc., Chicago, IL, USA) and Statsdirect (Cheshire, England, UK).

Results

Comparison of clinical and laboratory features between older and younger patients

Among the 462 AML patients recruited, 261 were males and 201 were females (Table 1). One hundred and seventy-seven patients were 60 years or older with a median age of 71 years (range 60–90 years). There was no difference in gender, hemogram and lactate dehydrogenase level between younger patients and older patients. Older age was negatively associated with the expression of CD19 (P=0.022), CD15 (P=0.007) and CD34 (P=0.002) on the leukemic cells (Supplementary Table 1). There was no difference in the expression of other antigens.

Table 1 Clinical manifestations of AML patients stratified by age
Full size table

Comparison of cytogenetic abnormalities and molecular gene mutations between older and younger patients

Chromosome data were available in 444 patients at diagnosis, including 166 older and 278 younger patients (Table 2). Compared with younger patients, the elderly had more frequently unfavorable-risk cytogenetic changes (21.1 vs 10.8%, P=0.004), but less commonly favorable-risk cytogenetics (4.2 vs 19.4%, P<0.001), such as t(8;21) (2.4 vs 13.7%, P<0.001) based on the Medical Research Council (MRC) classification.39 Specifically, the older patients had higher frequencies of complex chromosomal abnormalities (16.3 vs 8.6%, P=0.032), monosomy 5/5q deletion (8.4 vs 1.8%, P=0.001) and monosomy 7/7q deletion (8.4 vs 2.9%, P=0.012) but a lower incidence of t(7;11) (0 vs 3.6%, P=0.016). The distribution of simple chromosomal abnormalities with two or less changes involving chromosomes 8, 11, 13 and 21 was not different between the two groups.

Table 2 Association of age with cytogenetic abnormalitiesa
Full size table

To investigate the difference of gene mutations in the pathogenesis of leukemia between older and younger AML patients, a complete mutational screening of 21 genes was performed. The most common molecular event in total cohort was FLT3/ITD (22.5%), followed by NPM1 (22.3%), DNMT3A (15.2%), TET2 (14.3%) and CEBPA mutations (14.3%). Among the elderly, the most prevalent molecular event was NPM1 (28.2%), followed by TET2 (24.3%), FLT3/ITD (22.6%), DNMT3A (20.9%) and RUNX1 mutations (19.8%) (Table 3). The median number of molecular gene mutations at diagnosis was higher in the older patients than the younger ones (2.0, range 0–5 vs 1.0, range 0–5, P<0.001). Older patients had significantly higher incidences of PTPN11, NPM1, RUNX1, ASXL1, TET2, DNMT3A and TP53 mutations than younger patients (6.2% vs 2.5%, P=0.050; 28.2% vs 18.6%, P=0.021; 19.8% vs 9.5%, P=0.002; 17.6% vs 6.7%, P<0.001; 24.3% vs 8.1%, P<0.001; 20.9% vs 11.6%; P=0.008; and 13.0% vs 4.2%, P=0.001, respectively). On the contrary, WT1 mutations were rarely seen in patients aged 60 years or older (3.4 vs 9.1%, P=0.023). Other genetic alterations were not significantly different between these two age groups. The distributions of molecular gene mutations in these two groups are distinct (Figure 1 and Supplementary Figure 1). Older patients had a higher frequency to harbor one or more adverse genetic alterations (including FLT3/ITD, WT1, RUNX1, ASXL1, DNMT3A and TP53 mutations)22, 31 than younger ones (69.5 vs 49.5%, P<0.001), and the difference remained similar between the two groups when two or more such gene mutations were counted (26.6 vs 13.3%, P=0.001). We further showed that 85 pairwise associations were significant with P<0.1 in the elderly cohort (Supplementary Figure 2).

Table 3 Distribution of molecular genetic alterations by age
Full size table
Figure 1
figure 1

The Circos plots depicted the relative frequency and pairwise co-occurrence of genetic alterations in the older (a) and younger AML patients (b). The length of the arc corresponds to the frequency of the first gene mutation, and the width of the ribbon corresponds to the proportion of the second gene mutation.

Full size image

Prognostic impact of gene mutations in older patients

Fewer older patients received standard chemotherapy than younger patients (69/177, 40.0% vs 260/285, 91.2%, P<0.001); however, standard chemotherapy lead to a longer OS than palliative care only in this group of patients (median, 10.0 vs 3.0 months, P<0.001). Among the total cohort of 329 AML patients undergoing conventional intensive induction chemotherapy, 252 (76.6%) patients achieved CR. Older patients had a lower CR rate than younger population (56.5 vs 81.9%, P<0.001, Table 1). With a median follow-up of 69 months (ranges, 0.1–160), the elderly had significantly poorer OS and DFS than those aged below 60 years (median, 10.0 vs 61.0 months, P<0.001, Figure 2a, and median, 3.0 vs 9.0 months, P=0.001, Figure 2b, respectively). In multivariate Cox proportional hazards regression analysis for total cohort (Table 4), the independent poor risk factors for OS and DFS were older age, high white blood cell count >50 000/μl, and WT1, DNMT3A and TP53 mutations. On the other hand, NPM1+/FLT3-ITD− and CEBPAdouble-mutation were independent favorable prognostic factors. We also found that unfavorable-risk cytogenetics and RUNX1 mutations independently conferred poorer DFS and IDH2 predicted better OS.

Figure 2
figure 2

The Kaplan–Meier survival curves for OS (a) and DFS (b) in 329 AML patients who received standard intensive chemotherapy. The older patients have significantly poorer OS and DFS than those aged below 60 years (median, 10.0 vs 61.0 months, P<0.001, and median, 3.0 vs 9.0 months, P=0.001, respectively).

Full size image
Table 4 Multivariate analysis (Cox regression) on the disease-free survival and overall survival
Full size table

In the multivariate Cox proportional hazards regression analysis for OS in the elderly, DNMT3A and TP53 mutations were independent poor prognostic factors, while NPM1+/FLT3-ITD− remained to have good prognostic impact. Intriguingly, the older patients harboring any unfavorable genetic alteration, including FLT3/ITD, DNMT3A or TP53 mutation, had more dismal survival compared with those not (median OS, 7.0 vs 14.0 months, P=0.042, Figure 3).

Figure 3
figure 3

The Kaplan–Meier survival curve for OS in older AML patients stratified by having adverse genetic alterations or not. The survival of older patients who harbor any unfavorable genetic alterations is more dismal compared with those who do not (OS, 7.0 vs 14.0 months, P=0.042).

Full size image

The poor prognostic impacts of some mutations, such as FLT3/ITD, RUNX1 and DNMT3A mutations were lost when the patients receiving HSCT were not censored on the date of transplantation, implying allogeneic HSCT might overcome the poor risk of the patients with these mutations. Unfortunately, none of the 69 elder patients underwent allogeneic HSCT.

Further, the older patients with intermediate-risk cytogenetics could be further separated into three risk groups according to the molecular genotype:22 mutations of NPM1 or IDH2 or CEBPAdouble-mutation in the absence of FLT3/ITD as a favorable genotype,9 mutations of RUNX1, WT1, ASXL1, DNMT3A or TP53 as an unfavorable genotype,9, 27 and the remaining mutation patterns as an intermediate-risk genotype. Among the older patients with intermediate-risk cytogenetics, those with a favorable genotype had a higher CR rate and a trend of lower relapse rate than those with intermediate-risk and unfavorable genotypes (CR, 81.8 vs 64.7 vs 30.3%, P=0.002 and relapse, 44.4 vs 72.7 vs 87.5%, P=0.150). These three groups also had distinct OS (median, 26.0 vs 15.0 vs 8.0 months, P<0.001, Figure 4a) and DFS (median, 12.0 vs 7.0 vs 0 months, P=0.002, Figure 4b). Patients with intermediate-risk cytogenetics but favorable genotype had OS and DFS similar to those with favorable-risk cytogenetics (P=0.349), while patients with intermediate-risk cytogenetics but unfavorable genotype had OS and DFS similar to those with unfavorable-risk cytogenetics (P=0.420).

Figure 4
figure 4

The Kaplan–Meier survival curve for OS (a) and DFS (b) in older AML patients stratified by molecular genotypes. Older patients could be risk-stratified into groups with distinct outcomes.

Full size image

Genetic ontogeny was first proposed by Lindsley et al.40 Three types of mutations were defined: secondary-type mutations, TP53 mutations and de novo/pan AML mutations. Presence of secondary-type mutations predicted characteristic phenotype and poor outcome. We validated the impact of genetic ontogeny in our cohort. Five secondary-type genes, including three splicing factor genes (SRSF2, SF3B1 and U2AF1),41 ASXL1 and STAG2, were analyzed. In the 177 de novo AML patients aged 60 years or older, 34.5% had secondary-type mutations, 13.0% had TP53 mutation, 49.2% had de novo/pan AML mutations and 3.3% were undetermined. Among the 69 patients receiving standard chemotherapy, patients with secondary-type or TP53 mutations had a lower CR rate and a poor OS than those with de novo/pan AML mutations (43.5% vs 25.0% vs 66.7%, P=0.006, and median, 10.0 vs 3.0 vs 14.0 months, P=0.004, respectively).

Discussion

Most studies concerning prognostic factors in AML patients were focused on younger patients with less comorbidities and better performance status, and thus might not be representative of the general AML population in the real world.42 In this study, we recruited consecutively all de novo AML patients who had adequate samples for mutation analyses without restriction of age, so that we could compare the genetic alterations between older and younger patients and explored their clinical implications. We found that older AML patients had distinct clinico-biological features and genetic alterations from younger patients, and the status of mutations could predict the prognosis in this group of patients.

Traditionally, karyotype is one of the strongest prognostic factors in AML patients.43, 44 We showed that older patients had more frequently poor risk cytogenetics, such as complex chromosomal abnormalities, or aberrations involving chromosomes 5 and 7, but less frequently the core binding factor abnormalities. To better stratify AML patients into different risk groups, the European LeukemiaNet (ELN) panel first proposed a standardized classification according to both cytogenetics and molecular mutations in three genes, including FLT3/ITD, NPM1 and CEBPA mutations.7 Recently, several other genetic alterations were also found to have prognostic significance and were incorporated into risk stratification of AML patients.22, 31, 45, 46 However, there were only few reports in literature regarding the clinical impact of molecular alterations on older AML patients. In the studies from Cancer and Leukemia Group B (CALGB), RUNX1 and ASXL1 mutations were found to be more prevalent in the elder population with cytogenetically normal AML and were poor prognostic factors.14, 15 Older patients with WT1(ref. 47) or TP53(ref. 48) mutations had a shorter OS, while those with NPM149 mutations had a better CR rate and OS. Ostronoff et al.50 further depicted that NPM1 mutations in the absence of FLT3/ITD had a survival benefit among patients aged 55–65 years, but not in those older than 65 years. To our knowledge, this study is the first to comprehensively investigate the molecular genetic alterations of 21 genes among older patients with non-M3 AML. First, we showed that the distribution of genetic alterations and the burdens of gene mutations differed across the age groups. Second, older patients had higher incidences of PTPN11, NPM1, RUNX1, ASXL1, TET2, DNMT3A and TP53 mutations but less WT1 mutations. With the exception of NPM1 mutation, most of the other mutations that were more prevalent in older patients had unfavorable prognostic impact. Furthermore, older patients had a higher frequency to harbor one or more adverse genetic alterations (including FLT3/ITD, WT1, RUNX1, ASXL1, DNMT3A and TP53) than younger ones. Taken together, in addition to a higher incidence of adverse cytogenetics, the higher frequency and burdens of molecular mutations that are associated with poor prognosis in the elderly might explain the dismal outcome in this group of patients. Another possible cause to explain the dismal clinical outcome in older patients is that they are more vulnerable to the toxicity of chemotherapy agents, so may have higher treatment-related mortality.5, 51 However, our study demonstrated that the early mortality rate were comparable between the two age groups, similar to the national registration data of the United States52 and Swedish Acute Leukemia Registry.53, 54

Cytogenetic changes could well separate older AML patients into three risk groups in this study, similar to previous reports.55, 56, 57, 58 However, about 60–70% of the patients were in the intermediate-risk cytogenetic group which would hinder risk stratification of these patients for proper treatment. With the incorporation of nine gene mutations, including FLT3/ITD and mutations of CEBPA, NPM1, RUNX1, WT1, IDH2, ASXL1, DNMT3A and TP53, that are associated with prognosis,9, 27 we showed that older AML patients with intermediate-risk cytogenetics could be further stratified into three groups with different outcomes. The patients with favorable genotype (NPM1, IDH2 and CEBPAdouble-mutation in the absence of FLT3/ITD) had the longest survival, whereas those with unfavorable genotype (DNMT3A, ASXL1, WT1, RUNX1 or TP53) had the poorest outcome. The incorporation of the mutation status of these genes is helpful to stratify this highly heterogeneous population with intermediate-risk cytogenetics into distinct risk groups.

Genetic ontogeny, first proposed by Lindsley et al.,40 can help risk-stratify AML patients irrespective of their clinical assignment. In the 42 de novo AML patients aged 60 years or older in their study, 33.3% had secondary-type mutations, 21.4% had TP53 mutations and 45.2% had de novo/pan AML mutations. The frequencies of these three types of mutations in our cohort were not much different from those reported by Lindsley et al. The poor prognostic implication of secondary-type mutations in elderly patients with de novo AML was shown in this study as that of Lindsley et al.40 Similar to the high incidence of secondary-type mutations, elderly AML patients aged 60 years or older had higher incidence of dysplastic morphological features than those younger than 60 years (23.7 vs 14.4%; P=0.013); even none of them had a history of myelodysplastic syndrome, myeloproliferative neoplasm or other hematologic diseases.

In summary, we showed that older AML patients had distinct clinico-biological features, more frequently high-risk cytogenetics and gene mutations, and poorer prognosis. Integration of both cytogenetics and molecular alterations can better stratify older patients into different risk groups with distinct outcomes. It is warranted to develop novel therapies to improve the outcome of older patients with poor prognosis under current treatment modalities.