- Email: [email protected]
DNMT3A mutation is a poor prognosis biomarker in AML: Results of a meta-analysis of 4500 AML patients
Leukemia Research 37 (2013) 1445–1450
Contents lists available at ScienceDirect
Leukemia Research journal homepage: www.elsevier.com/locate/leukres
DNMT3A mutation is a poor prognosis biomarker in AML: Results of a meta-analysis of 4500 AML patients Velizar Shivarov a,∗ , Ralitza Gueorguieva b , Angel Stoimenov c , Ramon Tiu d a
Laboratory of Hematopathology and Immunology, National Hematology Hospital, Sofia, Bulgaria Department of Biostatistics, Schools of Public Health and Medicine, Yale University, New Haven, CT, USA Laboratory of Cytogenetics and Molecular Biology, National Hematology Hospital, Sofia, Bulgaria d Department of Translational Hematology and Oncology Research, Cleveland Clinic, Cleveland, OH, USA b c
a r t i c l e
i n f o
Article history: Received 18 March 2013 Accepted 27 July 2013 Available online 5 August 2013 Keywords: DNMT3A mutations AML Prognosis Meta-analysis
a b s t r a c t Somatic DNA methyl transferase 3A (DNMT3A) mutations have been recognized recently as recurrent molecular aberrations in acute myeloid leukemia (AML). The precise role of these mutations in leukemogenesis remains elusive but a number of studies have already been conducted to study their potential prognostic value in AML patients with variable results. We performed a meta-analysis on published data from over 4500 AML patients to provide robust evidence supporting DNMT3A mutation testing in clinical setting for AML patients. Our meta-analysis showed that DNMT3A mutations were associated with M4 and M5 AML subtypes. Those mutations conferred significantly worse prognosis with both shorter OS (p = 0.0004) and shorter RFS (p = 0.002). Notably, DNMT3A mutations appeared to be an independent adverse prognostic factor also in younger patients with normal cytogenetics AML (OS (p = 0.01) and RFS (p = 0.0005)) and also in the subgroup of patients with high risk genotypes defined according to the criteria of the European Leukemia Net (ELN) (OS (p = 0.002)). Therefore, DNMT3A mutational status can improve the risk stratification of AML patients in the setting of integrated mutational profiling. © 2013 Elsevier Ltd. All rights reserved.
1. Introduction Advances in molecular genetics have led to improvement in our understanding of the complex biological and clinical heterogeneity of patients with acute myeloid leukemia (AML) [1]. This was made possible through the identification of numerous recurrent somatic mutations in key genes involved in processes such as intracellular signaling, epigenetic modifications and transcriptional regulation of gene expression, apoptosis and cell cycle regulators ultimately causing enhanced proliferation and perturbed differentiation of hematopoietic precursors [2]. From a clinical standpoint, most of the newly identified molecular mutations immediately gain recognition as important outcome predictors and factors for initial risk stratification of patients with AML. For instance, a general consensus has already been reached regarding the implication and practical application of CEBPA, NPM1 and FLT3 mutational status [3]. Other more recently identified mutations (such as IDH1/2, DNMT3A) have also been investigated as predictive factors in various studies and will probably be an important part of the integrated mutational profiling in AML [4].
∗ Corresponding author at: National Hematology Hospital, 6 Plovdivsko pole Street, 1756 Sofia, Bulgaria. Tel.: +359 29701 218. E-mail address: [email protected] (V. Shivarov). 0145-2126/$ – see front matter © 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.leukres.2013.07.032
One of those potential candidate genes is DNA methyl transferase 3A (DNMT3A). Initially, mutations in DNMT3A were identified through targeted resequencing [5] and a whole genome and exome sequencing approach in AML [6,7]. DNMT3A enzyme is known to be essential for de novo DNA methylation [8], but there is still a lack of consensus on whether DNMT3A mutations cause changes in the global DNA methylation in AML samples and its precise role in leukemogenesis remains unclear [6,7,9]. A number of studies have been conducted to study the potential prognostic value of DNMT3A mutations in AML patients with variable results. Here, we performed a meta-analysis on published data from large studies to provide robust evidence supporting DNMT3A mutation testing in clinical setting for AML patients. 2. Methods 2.1. Studies selection A literature search for potential relevant publications was performed using the PubMed and Scopus databases. The search was performed using the following combination of terms: DNMT3A AND mutation AND leukemia, and was limited to publications in English listed in the databases between 2010 and the end of 2012. Titles and abstracts of all initially selected papers were reviewed manually and a total of 14 papers were left for full text review before final inclusion in the metaanalysis. Only papers that dealt with adult AML patients and provided adequate data on patients’ demographics (age and gender distribution) and clinical features (such as WBC, bone marrow blast count, FAB subtype, cytogenetic and molecular profile)
1446
V. Shivarov et al. / Leukemia Research 37 (2013) 1445–1450
as well as detailed data at least on overall survival (OS) analysis were included in the final selection. During revision two additional papers were reviewed and one of them included in the final meta-analysis of a total of nine reports.
2.2. Data extraction Key data of interest from the selected studies were summarized on a spreadsheet. The data were extracted by one investigator (VS) and independently cross-checked by another (AS). These data included country of origin, year of publication, number of patients, age and gender distribution of patients, mean white blood cell counts (WBC), FAB subtypes, mutational analysis, distribution of patients by cytogenetic risk, hazard ratios (HRs) and the 95% confidence intervals (CIs) for overall survival (OS) and relapse-free survival (RFS). For the studies that did not provide data on mean WBC, mean values were estimated based on the median and the range of WBC as described previously [10]. For studies that did not report HRs for OS and RFS, these values were estimated based on the reported data on the number of analyzed patients and events or directly extracted from the published Kaplan–Meier curves using the method developed by Parmar et al. [11,12]. For several studies assistance from the authors was requested [13,14].
2.3. Analysis Meta-analysis of the data was performed with RevMan version 5.2 software following the recommendation of the Cochrane Collaboration (http://www.cochrane.org/). Difference in mean WBC and percentage of bone marrow blasts between DNMT3A mutated and unmutated cases was assessed using the inverse variance method for continuous variables. Association of the DNMT3A mutation status with gender, FAB subtype and other mutations was tested with the Mantel-Haenszel method for dichotomous data. Prognostic role of DNMT3A mutations on OS and RFS was assessed by estimation of the pooled hazard ratios and their respective 95% confidence intervals with the inverse variance method. For all types of data fixed effect model analysis was performed first and if the heterogeneity was moderate with inconsistency (I2 ) of more than 20%, random effect model analysis was also performed. Presented pooled data are from fixed effect model analysis unless explicitly stated otherwise.
3. Results We identified nine studies [4,6,9,13–18] meeting the abovementioned criteria and included them in the current meta-analysis (Table 1). The geographic regions of origin of those studies were as follows: Eastern Asia (1), Western Europe (4) and North America (4), i.e. they originated from developed countries with high standards of care for AML patients. The studies included a total of 4582 AML patients and 988 DNMT3A mutated cases. The frequencies of DNMT3A mutated cases varied between 14.0% and 34.2%. Notably, DNMT3A mutation carrier status was not associated with sex (pooled OR: 0.83 (95% CI: 0.66, 1.06) p = 0.14, I2 = 51% in the random effect model analysis). Most studies except for two reported sufficient data on the white blood cells counts (WBC) in mutated and unmutated cases. Random effect model analysis did not reveal a significant mean difference (MD: 0.23; (95%CI: −16.86,17.31); p = 0.98) in WBC between the two groups but the heterogeneity was substantial (I2 = 91%). Sufficient data for analysis of the bone marrow blast percentage were extracted from four studies and did not show any significant mean difference (MD: 3.01; (95%CI: −0.61,6.62); p = 0.10; I2 = 79% in the random effect model).
3.1. Association of DNMT3A mutations with FAB subtype Initial reports on DNMT3A mutations in AML suggested that these mutations were associated with M4 and M5 subtypes according to the French–American–British (FAB) classification. We performed meta-analysis on the data from the selected studies. Indeed, as shown in Table 2, DNMT3A mutations were associated with non-M2 subtypes, which was due to the strong association with M4 and M5 subtypes.
3.2. Association of DNMT3A mutations with other recurrent mutations in AML We further tested the hypothesis that DNMT3A mutations were associated with other prognostically relevant mutations in AML. As shown in Table 3 DNMT3A mutations showed significant association with IDH1/2 mutations, FLT3-ITD and TKD mutations and NPM1 mutations and did not show association with CEBPA and NRAS mutational status. However, association with IDH1/2 and NPM1 mutations had a significant level of heterogeneity among the selected studies of 58% and 80%, respectively. 3.3. Prognostic power of DNMT3A mutations in AML For the entire cohort of patients the meta-analysis revealed significantly worse prognosis in terms of both OS (OR: 1.27; 95%CI: 1.11–1.45; p = 0.0004, I2 = 24% in the random effect model) and RFS (OR: 1.24; 95%CI: 1.08–1.42; p = 0.002) of DNMT3A mutation carrier. The data are presented in Fig. 1. Several studies provided data on mutation status and prognosis of younger adult patients with normal cytogenetics AML. Notably, DNMT3A mutations appeared to be an independent adverse prognostic factor (OS (OR: 1.48; 95%CI: 1.09–2.01; p = 0.01; I2 = 54% in the random effect model) and RFS (OR: 1.32; 95%CI: 1.13–1.55; p = 0.0005)) in this cohort as well (Fig. 2). Because of the apparent association of DNMT3A mutations with NPM1 and FLT3 mutations we investigated the prognostic impact of DNMT3A mutations in patients with low- and high-risk genotypes defined according to European Leukemia Net criteria [3] (Fig. 3), which implemented NPM1 and FLT3 mutational status. Notably, DNMT3A mutations conferred adverse prognosis only in patients with high-risk genotypes (OR: 1.38; 95%CI: 1.13–1.69; p = 0.002). 4. Discussion The era of personalized medicine particularly in the management of neoplastic diseases is now within reach. In the advent of novel Next Generation Sequencing (NGS) approaches and the constant drop of the cost of DNA sequencing in general will make it possible to perform a large scale mutational profiling of various patients cancer samples. This will have a particular implication in the field of hematology as it has already been shown that integrated mutational profiling can improve initial risk stratification of AML patients. However, the decision to include a specific gene of interest in a hematologic cancer panel for integrated mutational testing must be supported by robust data supporting its independent prognostic value in patients with that specific disease. Here we assessed the independent prognostic value of DNMT3A mutations in adult AML. DNMT3A is a key gene for the epigenetic regulation of human cells as the encoded enzyme is known to be a major player in the de novo DNA methylation at CpG sites [8]. Much remains to be learned on the precise leukemogenic effect of the epigenetic changes caused by DNMT3A mutations but we were able to identify several studies that evaluated its prognostic role in AML patients. The pooled meta-analysis of these studies showed that DNTM3A mutations confer a worse prognosis for OS and RFS. This worse prognosis held true even in the subgroup of younger adult AML patients (<60 years) with normal cytogenetics and in the subgroup of patients with ELN defined high-risk genotypes. This may be considered not surprising as DNMT3A mutations were found significantly more frequently in M4 and M5 subtypes of AML and were associated with FLT3/NPM1 and IDH1/2 mutations. These associations may even be a result of a direct pathogenic effect of mutations affecting genes involved in epigenetic regulation causing a general genomic instability and promoting mutations in
V. Shivarov et al. / Leukemia Research 37 (2013) 1445–1450
1447
Table 1 Summary of the data extracted from the eight studies included in the meta-analysis. Demographic and clinical characteristics of the patients included in the study. First author’s name
Thol
Ribeiro
Marcucci
Ostronoff
Renneville
Hou
Ley
Patel
Gaidzik
Journal Year Published Population Patients (n) Type of DNMT3A mutations Number of mutated cases Gender (M/F) Age (median, range; years)
JCO 2011 Germany 489 Any 87 49/38 52 (30–60)
Blood 2012 Netherlands 415 Any 96 47/49 50.5(18–60)
JCO 2012 USA 415 Any 142 70/72 61(22–82)
Leukemia 2012 USA 191 Any 37 20/17 68(57–81)
Leukemia 2012 France 123 Any 36 19/17 47(23–58)
Blood 2012 Taiwan 500 Any 70 41/29 61(16–87)
NEJM 2010 USA 281 Any 62 28/34 53.1(39–67)
NEJM 2012 USA 398 Any 88 207/191 46.5(18–60)
Blood 2013 Germany 1770 Any 370 907/863 NR(18–60)
FAB subtype n mutated (n total in the subgroup) M0 3(10) 1(16) M1 15(89) 15(87) 8(121) 23(104) M2 – – M3 37(167) 15(79) M4 22(66) 36(100) M5 1(12) 1(6) M6 0(3) – M7 – 3(16) RAEB – 2(7) ND Missing 1(21) –
1(8) 29(83) 18(88) – 33(74) 26(47) 0(0) – – – –
2(5) 9(44) 6(61) 0(1) 10(48) 7(20) 0(2) 0(5) – – –
0(5) 6(30) 10(37) – 11(22) 9(16) 0(7) – – 0(6) –
2(12) 14(126) 13(184) 0(38) 28(152) 12(36) 0(12) – – 1(10) –
– – – 1(48) 20(61) 12(21) – – – – –
NR NR NR NR NR NR NR NR NR NR NR
NR NR NR NR NR NR NR NR NR NR NR
Cytogenetics risk group n mutated (n total in the subgroup) Favorable 1(79)a 0(57)e Intermediate 82(344)a 85(276)e Adverse 4(61)a 6(70)e Missing data 0(5)a 5(12)e Unclassifiable – – NR 72(194) Normal karyotype
NR NR NR NR NR 142(415)
0(12)d 24(78)d 1(33)d 9(55)d 3(13)d NR
NR NR NR NR NR 36(123)
0(99)f 62(318)f 4(65)f 4(18)f – 51(223)
0(79)c 56(166)c 4(30)c 2(6)c – 44/120
NR NR NR NR NR NR
4(256)b 309(1060)b 33(320)b NR NR 268(794)b
Mutations n mutated (n total in the subgroup) 22(70) IDH1/IDH2 34(129) FLT3-ITD NR FLT3-TKD NPM1 56(163) 8(57) NRAS NR CEBPA NR TET2
46(79) 62(147) 10(31) 107(253) NR 7(65) 31(98)
6(45) 13(49) NR 17(51) NR 0(3) NR
13(36) 7(25) 3(5) 24(52) NR 0(12) NR
20(82) 30(113) 9(38) 38(104) 8(61) 3(66) 6(65)
20(45) 15(54) 9(21) 37(64) NR NR NR
19(NR) 53(NR) NR 57(NR) 10(NR) 7(NR) 6(NR)
96(247) 126(387) 28(125) 234(475) NR 18(123) NR
35(69) 39(116) 15(41) 73(133) 8(42) 3(31) NR
NR, data not reported; –, Data not included; a Medical Research Council (MRC) Cytogenetic risk subgroups. b European LeukemiaNet (ELN) guidelines. c Cancer and Leukemia Group B (CALGB) Cytogenetic risk subgroups. d Southwest Oncology Group (SWOG) Cytogenetic risk subgroups. e Cytogenetic risk group favorable includes t(8;21), inv(16) or t(15;17); adverse, inv(3)/t(3;3), t(6;9), 11q23 abnormalities other than t(9;11), del5, del5(q), del7, del7(q), t(9;22) or monosomal karyotypes (MK); and intermediate, the remaining AML cases. f Favorable, t(15;17), t(8;21), inv(16); unfavorable, −7, del(7q), −5, del(5q), 3q abnormality, complex abnormalities; and intermediate, normal karyotype and other abnormalities. Table 2 Pooled analysis of association of DNMT3A mutations with specific FAB subtypes of AML. FAB subtype
Random effect model
Fixed effect model
Overall effect OR M0 M1 M2 M4 M5 M6
Heterogeneity CI
0.85 0.83 0.48 1.45 2.65 0.4
0.33–2.21 0.64–1.08 0.31–0.73 1.06–1.99 2.02–3.48 0.13–1.23
2
Overall effect
p-Value
I (%)
OR
0.74 0.16 0.0008 0.02 <0.00001 0.11
36 0 58 38 0 0
0.7 0.82 0.47 1.43 2.63 0.36
Heterogeneity CI 0.35–1.41 0.63–1.07 0.36–0.62 1.13–1.82 2.00–3.45 0.12–1.08
p-Value
I2 (%)
0.32 0.14 <0.00001 0.003 <0.00001 0.07
36 0 58 38 0 0
Table 3 Pooled analysis of association of DNMT3A mutations with other recurrent somatic mutations in AML. Mutation status subtype
Random effect model
Fixed effect model
Overall effect OR IDH1/2 FLT3-ITD FLT3-TKD NPM1 NRAS CEBPA
2.62 2.06 1.61 5.92 0.8 0.35
CI 1.92–3.58 1.74–2.43 1.08–2.40 3.97–8.83 0.51–1.26 0.25–0.60
Heterogeneity 2
Overall effect
p-Value
I (%)
OR
<0.00001 <0.00001 0.02 <0.00001 0.34 0.0002
58 4 39 80 0 31
2.75 2.05 1.84 6.31 0.8 0.37
CI 2.30–3.30 1.74–2.41 1.08–1.91 5.35–7.44 0.50–1.26 0.25–0.54
Heterogeneity p-Value
I2 (%)
<0.00001 <0.00001 0.01 <0.00001 0.33 <0.00001
58 4 39 80 0 31
1448
V. Shivarov et al. / Leukemia Research 37 (2013) 1445–1450
Fig. 1. Forest plots of pooled HRs and the corresponding 95% CIs for the prognostic significance of DNMT3A mutations in the entire cohort of AML patients: (A) OS endpoint and (B) RFS endpoint.
other genes (such as FLT3) [19]. Actually, several studies reported multivariate analysis of the prognostic effect of DNMT3A including FLT3/NPM1 mutational status. Thol et al., [9] showed that DNMT3A mutations remained an independent negative prognostic marker for OS. Marcucci et al. [17] did not find adverse impact on OS in multivariate analysis based on the entire cohort they analyzed but did so in the subgroup analysis based on the age stratification.
Renneville et al. [14] included NPM1/FLT3-ITD/CEBPA low- vs. highrisk mutational status as covariate in their multivariate analysis and DNMT3A did not appear to be a significant factor for adverse prognosis. Ribeiro et al. [15] also included FLT3/NPM1 mutational status in their multivariate analysis and found that DNMT3A mutations were independent factor for worse OS. Gaidzik et al. [18], however, failed to demonstrate that DNMT3A mutations worsen
Fig. 2. Forest plots of pooled HRs and the corresponding 95% CIs for the prognostic significance of DNMT3A mutations in the subgroup of AML patients with normal cytogenetics aged under 60 years: (A) OS endpoint and (B) RFS endpoint.
V. Shivarov et al. / Leukemia Research 37 (2013) 1445–1450
1449
Fig. 3. Forest plots of pooled HRs and the corresponding 95% CIs for OS for the prognostic significance of DNMT3A mutations in the ELN defined subgroups of AML patients: (A) Low-risk genotypes subgroup and (B) high-risk genotypes subgroups.
OS. Ideally, this issue could be resolved by performing a large scale meta-analysis on individual patients’ data, which was beyond the scope of the current work. To partially address this question we investigated the prognostic role of DNMT3A mutations within the ELN defined risk subgroups and found that DNMT3A mutations conferred adverse prognosis only in patients with high-risk genotypes (OR: 1.38; 95%CI: 1.13–1.69; p = 0.002). It remains unclear whether this effect is identical for all genotypes within the highrisk group. For instance, Ribeiro et al. [15] reported adverse effect of DNMT3A within the subgroup with FLT3wt/NPM1wt genotype but not in the subgroup with FLT3-ITD/NPM1wt genotype, whereas Thol et al. [9] reported adverse impact of DNMT3A mutations in both subgroups. Furthermore, Ley et al. [6] and Markova et al. [20] (study not included in this meta-analysis) reported adverse effect of DNMT3A mutations in FLT3 mutated cases. This is consistent with the proposed inclusion of DNMT3A mutations along with other mutations as an additional enhancer of the poor prognosis conferred by the presence of FLT3-ITD mutations as proposed Patel et al. [4] based on integrated mutational profiling. The reason for the differences in the reported OS between various AML studies with relatively similar patient profile is not clear. However, several factors not reported in most AML studies that investigates the prognostic value of these molecular mutations may be responsible including the differences in number of patients who subsequently underwent allogeneic stem cell transplantation, the presence of other molecular mutations that may confer differences in survival outcomes in AML including mutations in spliceosome genes (SF3B1, U2AF1, SRSF2), EZH2, SETBP1 and others and differences in other disease related pathogenetic factors like differences in microRNA profile, number and types of additional lesions specifically acquired somatic uniparental discomy detected by single nucleotide polymorphism array analysis (SNP-A) karyotyping. Another level of complexity adds the observation that DNMT3A mutations are scattered along the full length of the gene but tend to cluster at codon 882. The prognostic effect of R882 mutations compared to missense DNMT3A mutations affecting other codons was not assessed in this study due to lack of available published studies directly comparing the effect of the two types of mutations. However, it is likely that there might not be a difference in the prognostic power compared to that of other mutations as suggested by Ley et al. [6] who did not find any difference in the OS between the two groups. Furthermore, Ribeiro et al. [15] showed that R882 only mutations are associated with shorter OS (p = 0.018)
and RFS (p = 0.029). However, Marcucci et al. [17] showed that R882 mutations did not modify the risk of younger adults with NC-AML but did so in the age group over 60 years. In the cohort analyzed by Gaidzik et al. [18] there was strong evidence that R882-mutated vs. non-R882-mutated cases had opposing effects – unfavorable for R882 DNMT3Amut on RFS, and favorable for non-R882 DNMT3Amut on OS. These effects were consistent for both entire cohort and CN-AML. In addition to the prognostic value of DNMT3A mutations it is interesting to address the potential predictive role of these mutations. Although this issue was not consistently covered by the studies included in the current analysis, there were some hints that this might indeed be the case. Patel et al. [4] showed that doseintensified chemotherapy improved the outcome of the DNMT3A mutated cases. Notably, however, Ribeiro et al. [15] did not find any change in the risk conferred by DNMT3A mutation in AML patients who underwent allogeneic transplantation. Gaidzik et al. [18] also showed that DNMT3A mutations did not have any effect on RFS and OS of patients who have undergone allogeneic hematopoietic stem cell transplantation. Another study not included in the analysis reported a better response of DNMT3A mutated cases to idarubicin in comparison to daunorubicin [21]. Finally, it has already been proposed that DNMT3A mutations might be predictive for a better response to decitabine treatment [22]. In conclusion, our meta-analysis of eight studies with over 4500 AML patients and almost 1000 DNMT3A mutated cases showed conclusively an independent adverse prognostic effect of the mutations on the OS and RFS in the entire cohort. This was true also for the intermediate risk patients with NC-AML under the age of 60 years and patients with ELN defined high-risk genotypes. Therefore, DNMT3A mutational status can improve the risk stratification of AML patients in the settings of integrated mutational profiling as it has already been shown by some individual studies [4]. Conflict of interest All authors have no conflict of interest to declare. Funding source This work was partially supported through Grant ID 09 157 (Contract 5/16.12.2009) of the National Science Fund, Bulgaria.
1450
V. Shivarov et al. / Leukemia Research 37 (2013) 1445–1450
Acknowledgements None. Contributions. VS proposed research, gathered and analyzed data and wrote the manuscript, RG analyzed data and wrote the manuscript, AS gathered data and RT analyzed data and wrote the manuscript. References [1] Marcucci G, Haferlach T, Dohner H. Molecular genetics of adult acute myeloid leukemia: prognostic and therapeutic implications. J Clin Oncol 2011;29(5):475–86. [2] Sanders MA, Valk PJ. The evolving molecular genetic landscape in acute myeloid leukaemia. Curr Opin Hematol 2013;20(2):79–85. [3] Dohner H, Estey EH, Amadori S, et al. Diagnosis and management of acute myeloid leukemia in adults: recommendations from an international expert panel, on behalf of the European LeukemiaNet. Blood 2010;115(3):453–74. [4] Patel JP, Gonen M, Figueroa ME, et al. Prognostic relevance of integrated genetic profiling in acute myeloid leukemia. N Engl J Med 2012;366(12):1079–89. [5] Yamashita Y, Yuan J, Suetake I, et al. Array-based genomic resequencing of human leukemia. Oncogene 2010;29(25):3723–31. [6] Ley TJ, Ding L, Walter MJ, et al. DNMT3A mutations in acute myeloid leukemia. N Engl J Med 2010;363(25):2424–33. [7] Yan XJ, Xu J, Gu ZH, et al. Exome sequencing identifies somatic mutations of DNA methyltransferase gene DNMT3A in acute monocytic leukemia. Nat Genet 2011;43(4):309–15. [8] Okano M, Bell DW, Haber DA, Li E. DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development. Cell 1999;99(3):247–57. [9] Thol F, Damm F, Ludeking A, et al. Incidence and prognostic influence of DNMT3A mutations in acute myeloid leukemia. J Clin Oncol 2011;29(21):2889–96. [10] Hozo SP, Djulbegovic B, Hozo I. Estimating the mean and variance from the median, range, and the size of a sample. BMC Med Res Methodol 2005;5:13.
[11] Tierney JF, Stewart LA, Ghersi D, Burdett S, Sydes MR. Practical methods for incorporating summary time-to-event data into meta-analysis. Trials 2007;8:16. [12] Parmar MK, Torri V, Stewart L. Extracting summary statistics to perform meta-analyses of the published literature for survival endpoints. Stat Med 1998;17(24):2815–34. [13] Hou HA, Kuo YY, Liu CY, et al. DNMT3A mutations in acute myeloid leukemia: stability during disease evolution and clinical implications. Blood 2012;119(2):559–68. [14] Renneville A, Boissel N, Nibourel O, et al. Prognostic significance of DNA methyltransferase 3A mutations in cytogenetically normal acute myeloid leukemia: a study by the Acute Leukemia French Association. Leukemia 2012;26(6):1247–54. [15] Ribeiro AF, Pratcorona M, Erpelinck-Verschueren C, et al. Mutant DNMT3A: a marker of poor prognosis in acute myeloid leukemia. Blood 2012;119(24):5824–31. [16] Ostronoff F, Othus M, Ho PA, et al. Mutations in the DNMT3A exon 23 independently predict poor outcome in older patients with acute myeloid leukemia: a SWOG report. Leukemia 2013;27(1):238–41. [17] Marcucci G, Metzeler KH, Schwind S, et al. Age-related prognostic impact of different types of DNMT3A mutations in adults with primary cytogenetically normal acute myeloid leukemia. J Clin Oncol 2012;30(7):742–50. [18] Gaidzik VI, Schlenk RF, Paschka P, et al. Clinical impact of DNMT3A mutations in younger adult patients with acute myeloid leukemia: results of the AML Study Group (AMLSG). Blood 2013;121(23):4769–77. [19] Wakita S, Yamaguchi H, Omori I, et al. Mutations of the epigenetics-modifying gene (DNMT3a, TET2, IDH1/2) at diagnosis may induce FLT3-ITD at relapse in de novo acute myeloid leukemia. Leukemia 2013;27(5):1044–52. [20] Markova J, Michkova P, Burckova K, et al. Prognostic impact of DNMT3A mutations in patients with intermediate cytogenetic risk profile acute myeloid leukemia. Eur J Haematol 2012;88(2):128–35. [21] LaRochelle O, Bertoli S, Vergez F, et al. Do AML patients with DNMT3A exon 23 mutations benefit from idarubicin as compared to daunorubicin? A single center experience. Oncotarget 2011;2(11):850–61. [22] Metzeler KH, Walker A, Geyer S, et al. DNMT3A mutations and response to the hypomethylating agent decitabine in acute myeloid leukemia. Leukemia 2012;26(5):1106–7.
Contents lists available at ScienceDirect
Leukemia Research journal homepage: www.elsevier.com/locate/leukres
DNMT3A mutation is a poor prognosis biomarker in AML: Results of a meta-analysis of 4500 AML patients Velizar Shivarov a,∗ , Ralitza Gueorguieva b , Angel Stoimenov c , Ramon Tiu d a
Laboratory of Hematopathology and Immunology, National Hematology Hospital, Sofia, Bulgaria Department of Biostatistics, Schools of Public Health and Medicine, Yale University, New Haven, CT, USA Laboratory of Cytogenetics and Molecular Biology, National Hematology Hospital, Sofia, Bulgaria d Department of Translational Hematology and Oncology Research, Cleveland Clinic, Cleveland, OH, USA b c
a r t i c l e
i n f o
Article history: Received 18 March 2013 Accepted 27 July 2013 Available online 5 August 2013 Keywords: DNMT3A mutations AML Prognosis Meta-analysis
a b s t r a c t Somatic DNA methyl transferase 3A (DNMT3A) mutations have been recognized recently as recurrent molecular aberrations in acute myeloid leukemia (AML). The precise role of these mutations in leukemogenesis remains elusive but a number of studies have already been conducted to study their potential prognostic value in AML patients with variable results. We performed a meta-analysis on published data from over 4500 AML patients to provide robust evidence supporting DNMT3A mutation testing in clinical setting for AML patients. Our meta-analysis showed that DNMT3A mutations were associated with M4 and M5 AML subtypes. Those mutations conferred significantly worse prognosis with both shorter OS (p = 0.0004) and shorter RFS (p = 0.002). Notably, DNMT3A mutations appeared to be an independent adverse prognostic factor also in younger patients with normal cytogenetics AML (OS (p = 0.01) and RFS (p = 0.0005)) and also in the subgroup of patients with high risk genotypes defined according to the criteria of the European Leukemia Net (ELN) (OS (p = 0.002)). Therefore, DNMT3A mutational status can improve the risk stratification of AML patients in the setting of integrated mutational profiling. © 2013 Elsevier Ltd. All rights reserved.
1. Introduction Advances in molecular genetics have led to improvement in our understanding of the complex biological and clinical heterogeneity of patients with acute myeloid leukemia (AML) [1]. This was made possible through the identification of numerous recurrent somatic mutations in key genes involved in processes such as intracellular signaling, epigenetic modifications and transcriptional regulation of gene expression, apoptosis and cell cycle regulators ultimately causing enhanced proliferation and perturbed differentiation of hematopoietic precursors [2]. From a clinical standpoint, most of the newly identified molecular mutations immediately gain recognition as important outcome predictors and factors for initial risk stratification of patients with AML. For instance, a general consensus has already been reached regarding the implication and practical application of CEBPA, NPM1 and FLT3 mutational status [3]. Other more recently identified mutations (such as IDH1/2, DNMT3A) have also been investigated as predictive factors in various studies and will probably be an important part of the integrated mutational profiling in AML [4].
∗ Corresponding author at: National Hematology Hospital, 6 Plovdivsko pole Street, 1756 Sofia, Bulgaria. Tel.: +359 29701 218. E-mail address: [email protected] (V. Shivarov). 0145-2126/$ – see front matter © 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.leukres.2013.07.032
One of those potential candidate genes is DNA methyl transferase 3A (DNMT3A). Initially, mutations in DNMT3A were identified through targeted resequencing [5] and a whole genome and exome sequencing approach in AML [6,7]. DNMT3A enzyme is known to be essential for de novo DNA methylation [8], but there is still a lack of consensus on whether DNMT3A mutations cause changes in the global DNA methylation in AML samples and its precise role in leukemogenesis remains unclear [6,7,9]. A number of studies have been conducted to study the potential prognostic value of DNMT3A mutations in AML patients with variable results. Here, we performed a meta-analysis on published data from large studies to provide robust evidence supporting DNMT3A mutation testing in clinical setting for AML patients. 2. Methods 2.1. Studies selection A literature search for potential relevant publications was performed using the PubMed and Scopus databases. The search was performed using the following combination of terms: DNMT3A AND mutation AND leukemia, and was limited to publications in English listed in the databases between 2010 and the end of 2012. Titles and abstracts of all initially selected papers were reviewed manually and a total of 14 papers were left for full text review before final inclusion in the metaanalysis. Only papers that dealt with adult AML patients and provided adequate data on patients’ demographics (age and gender distribution) and clinical features (such as WBC, bone marrow blast count, FAB subtype, cytogenetic and molecular profile)
1446
V. Shivarov et al. / Leukemia Research 37 (2013) 1445–1450
as well as detailed data at least on overall survival (OS) analysis were included in the final selection. During revision two additional papers were reviewed and one of them included in the final meta-analysis of a total of nine reports.
2.2. Data extraction Key data of interest from the selected studies were summarized on a spreadsheet. The data were extracted by one investigator (VS) and independently cross-checked by another (AS). These data included country of origin, year of publication, number of patients, age and gender distribution of patients, mean white blood cell counts (WBC), FAB subtypes, mutational analysis, distribution of patients by cytogenetic risk, hazard ratios (HRs) and the 95% confidence intervals (CIs) for overall survival (OS) and relapse-free survival (RFS). For the studies that did not provide data on mean WBC, mean values were estimated based on the median and the range of WBC as described previously [10]. For studies that did not report HRs for OS and RFS, these values were estimated based on the reported data on the number of analyzed patients and events or directly extracted from the published Kaplan–Meier curves using the method developed by Parmar et al. [11,12]. For several studies assistance from the authors was requested [13,14].
2.3. Analysis Meta-analysis of the data was performed with RevMan version 5.2 software following the recommendation of the Cochrane Collaboration (http://www.cochrane.org/). Difference in mean WBC and percentage of bone marrow blasts between DNMT3A mutated and unmutated cases was assessed using the inverse variance method for continuous variables. Association of the DNMT3A mutation status with gender, FAB subtype and other mutations was tested with the Mantel-Haenszel method for dichotomous data. Prognostic role of DNMT3A mutations on OS and RFS was assessed by estimation of the pooled hazard ratios and their respective 95% confidence intervals with the inverse variance method. For all types of data fixed effect model analysis was performed first and if the heterogeneity was moderate with inconsistency (I2 ) of more than 20%, random effect model analysis was also performed. Presented pooled data are from fixed effect model analysis unless explicitly stated otherwise.
3. Results We identified nine studies [4,6,9,13–18] meeting the abovementioned criteria and included them in the current meta-analysis (Table 1). The geographic regions of origin of those studies were as follows: Eastern Asia (1), Western Europe (4) and North America (4), i.e. they originated from developed countries with high standards of care for AML patients. The studies included a total of 4582 AML patients and 988 DNMT3A mutated cases. The frequencies of DNMT3A mutated cases varied between 14.0% and 34.2%. Notably, DNMT3A mutation carrier status was not associated with sex (pooled OR: 0.83 (95% CI: 0.66, 1.06) p = 0.14, I2 = 51% in the random effect model analysis). Most studies except for two reported sufficient data on the white blood cells counts (WBC) in mutated and unmutated cases. Random effect model analysis did not reveal a significant mean difference (MD: 0.23; (95%CI: −16.86,17.31); p = 0.98) in WBC between the two groups but the heterogeneity was substantial (I2 = 91%). Sufficient data for analysis of the bone marrow blast percentage were extracted from four studies and did not show any significant mean difference (MD: 3.01; (95%CI: −0.61,6.62); p = 0.10; I2 = 79% in the random effect model).
3.1. Association of DNMT3A mutations with FAB subtype Initial reports on DNMT3A mutations in AML suggested that these mutations were associated with M4 and M5 subtypes according to the French–American–British (FAB) classification. We performed meta-analysis on the data from the selected studies. Indeed, as shown in Table 2, DNMT3A mutations were associated with non-M2 subtypes, which was due to the strong association with M4 and M5 subtypes.
3.2. Association of DNMT3A mutations with other recurrent mutations in AML We further tested the hypothesis that DNMT3A mutations were associated with other prognostically relevant mutations in AML. As shown in Table 3 DNMT3A mutations showed significant association with IDH1/2 mutations, FLT3-ITD and TKD mutations and NPM1 mutations and did not show association with CEBPA and NRAS mutational status. However, association with IDH1/2 and NPM1 mutations had a significant level of heterogeneity among the selected studies of 58% and 80%, respectively. 3.3. Prognostic power of DNMT3A mutations in AML For the entire cohort of patients the meta-analysis revealed significantly worse prognosis in terms of both OS (OR: 1.27; 95%CI: 1.11–1.45; p = 0.0004, I2 = 24% in the random effect model) and RFS (OR: 1.24; 95%CI: 1.08–1.42; p = 0.002) of DNMT3A mutation carrier. The data are presented in Fig. 1. Several studies provided data on mutation status and prognosis of younger adult patients with normal cytogenetics AML. Notably, DNMT3A mutations appeared to be an independent adverse prognostic factor (OS (OR: 1.48; 95%CI: 1.09–2.01; p = 0.01; I2 = 54% in the random effect model) and RFS (OR: 1.32; 95%CI: 1.13–1.55; p = 0.0005)) in this cohort as well (Fig. 2). Because of the apparent association of DNMT3A mutations with NPM1 and FLT3 mutations we investigated the prognostic impact of DNMT3A mutations in patients with low- and high-risk genotypes defined according to European Leukemia Net criteria [3] (Fig. 3), which implemented NPM1 and FLT3 mutational status. Notably, DNMT3A mutations conferred adverse prognosis only in patients with high-risk genotypes (OR: 1.38; 95%CI: 1.13–1.69; p = 0.002). 4. Discussion The era of personalized medicine particularly in the management of neoplastic diseases is now within reach. In the advent of novel Next Generation Sequencing (NGS) approaches and the constant drop of the cost of DNA sequencing in general will make it possible to perform a large scale mutational profiling of various patients cancer samples. This will have a particular implication in the field of hematology as it has already been shown that integrated mutational profiling can improve initial risk stratification of AML patients. However, the decision to include a specific gene of interest in a hematologic cancer panel for integrated mutational testing must be supported by robust data supporting its independent prognostic value in patients with that specific disease. Here we assessed the independent prognostic value of DNMT3A mutations in adult AML. DNMT3A is a key gene for the epigenetic regulation of human cells as the encoded enzyme is known to be a major player in the de novo DNA methylation at CpG sites [8]. Much remains to be learned on the precise leukemogenic effect of the epigenetic changes caused by DNMT3A mutations but we were able to identify several studies that evaluated its prognostic role in AML patients. The pooled meta-analysis of these studies showed that DNTM3A mutations confer a worse prognosis for OS and RFS. This worse prognosis held true even in the subgroup of younger adult AML patients (<60 years) with normal cytogenetics and in the subgroup of patients with ELN defined high-risk genotypes. This may be considered not surprising as DNMT3A mutations were found significantly more frequently in M4 and M5 subtypes of AML and were associated with FLT3/NPM1 and IDH1/2 mutations. These associations may even be a result of a direct pathogenic effect of mutations affecting genes involved in epigenetic regulation causing a general genomic instability and promoting mutations in
V. Shivarov et al. / Leukemia Research 37 (2013) 1445–1450
1447
Table 1 Summary of the data extracted from the eight studies included in the meta-analysis. Demographic and clinical characteristics of the patients included in the study. First author’s name
Thol
Ribeiro
Marcucci
Ostronoff
Renneville
Hou
Ley
Patel
Gaidzik
Journal Year Published Population Patients (n) Type of DNMT3A mutations Number of mutated cases Gender (M/F) Age (median, range; years)
JCO 2011 Germany 489 Any 87 49/38 52 (30–60)
Blood 2012 Netherlands 415 Any 96 47/49 50.5(18–60)
JCO 2012 USA 415 Any 142 70/72 61(22–82)
Leukemia 2012 USA 191 Any 37 20/17 68(57–81)
Leukemia 2012 France 123 Any 36 19/17 47(23–58)
Blood 2012 Taiwan 500 Any 70 41/29 61(16–87)
NEJM 2010 USA 281 Any 62 28/34 53.1(39–67)
NEJM 2012 USA 398 Any 88 207/191 46.5(18–60)
Blood 2013 Germany 1770 Any 370 907/863 NR(18–60)
FAB subtype n mutated (n total in the subgroup) M0 3(10) 1(16) M1 15(89) 15(87) 8(121) 23(104) M2 – – M3 37(167) 15(79) M4 22(66) 36(100) M5 1(12) 1(6) M6 0(3) – M7 – 3(16) RAEB – 2(7) ND Missing 1(21) –
1(8) 29(83) 18(88) – 33(74) 26(47) 0(0) – – – –
2(5) 9(44) 6(61) 0(1) 10(48) 7(20) 0(2) 0(5) – – –
0(5) 6(30) 10(37) – 11(22) 9(16) 0(7) – – 0(6) –
2(12) 14(126) 13(184) 0(38) 28(152) 12(36) 0(12) – – 1(10) –
– – – 1(48) 20(61) 12(21) – – – – –
NR NR NR NR NR NR NR NR NR NR NR
NR NR NR NR NR NR NR NR NR NR NR
Cytogenetics risk group n mutated (n total in the subgroup) Favorable 1(79)a 0(57)e Intermediate 82(344)a 85(276)e Adverse 4(61)a 6(70)e Missing data 0(5)a 5(12)e Unclassifiable – – NR 72(194) Normal karyotype
NR NR NR NR NR 142(415)
0(12)d 24(78)d 1(33)d 9(55)d 3(13)d NR
NR NR NR NR NR 36(123)
0(99)f 62(318)f 4(65)f 4(18)f – 51(223)
0(79)c 56(166)c 4(30)c 2(6)c – 44/120
NR NR NR NR NR NR
4(256)b 309(1060)b 33(320)b NR NR 268(794)b
Mutations n mutated (n total in the subgroup) 22(70) IDH1/IDH2 34(129) FLT3-ITD NR FLT3-TKD NPM1 56(163) 8(57) NRAS NR CEBPA NR TET2
46(79) 62(147) 10(31) 107(253) NR 7(65) 31(98)
6(45) 13(49) NR 17(51) NR 0(3) NR
13(36) 7(25) 3(5) 24(52) NR 0(12) NR
20(82) 30(113) 9(38) 38(104) 8(61) 3(66) 6(65)
20(45) 15(54) 9(21) 37(64) NR NR NR
19(NR) 53(NR) NR 57(NR) 10(NR) 7(NR) 6(NR)
96(247) 126(387) 28(125) 234(475) NR 18(123) NR
35(69) 39(116) 15(41) 73(133) 8(42) 3(31) NR
NR, data not reported; –, Data not included; a Medical Research Council (MRC) Cytogenetic risk subgroups. b European LeukemiaNet (ELN) guidelines. c Cancer and Leukemia Group B (CALGB) Cytogenetic risk subgroups. d Southwest Oncology Group (SWOG) Cytogenetic risk subgroups. e Cytogenetic risk group favorable includes t(8;21), inv(16) or t(15;17); adverse, inv(3)/t(3;3), t(6;9), 11q23 abnormalities other than t(9;11), del5, del5(q), del7, del7(q), t(9;22) or monosomal karyotypes (MK); and intermediate, the remaining AML cases. f Favorable, t(15;17), t(8;21), inv(16); unfavorable, −7, del(7q), −5, del(5q), 3q abnormality, complex abnormalities; and intermediate, normal karyotype and other abnormalities. Table 2 Pooled analysis of association of DNMT3A mutations with specific FAB subtypes of AML. FAB subtype
Random effect model
Fixed effect model
Overall effect OR M0 M1 M2 M4 M5 M6
Heterogeneity CI
0.85 0.83 0.48 1.45 2.65 0.4
0.33–2.21 0.64–1.08 0.31–0.73 1.06–1.99 2.02–3.48 0.13–1.23
2
Overall effect
p-Value
I (%)
OR
0.74 0.16 0.0008 0.02 <0.00001 0.11
36 0 58 38 0 0
0.7 0.82 0.47 1.43 2.63 0.36
Heterogeneity CI 0.35–1.41 0.63–1.07 0.36–0.62 1.13–1.82 2.00–3.45 0.12–1.08
p-Value
I2 (%)
0.32 0.14 <0.00001 0.003 <0.00001 0.07
36 0 58 38 0 0
Table 3 Pooled analysis of association of DNMT3A mutations with other recurrent somatic mutations in AML. Mutation status subtype
Random effect model
Fixed effect model
Overall effect OR IDH1/2 FLT3-ITD FLT3-TKD NPM1 NRAS CEBPA
2.62 2.06 1.61 5.92 0.8 0.35
CI 1.92–3.58 1.74–2.43 1.08–2.40 3.97–8.83 0.51–1.26 0.25–0.60
Heterogeneity 2
Overall effect
p-Value
I (%)
OR
<0.00001 <0.00001 0.02 <0.00001 0.34 0.0002
58 4 39 80 0 31
2.75 2.05 1.84 6.31 0.8 0.37
CI 2.30–3.30 1.74–2.41 1.08–1.91 5.35–7.44 0.50–1.26 0.25–0.54
Heterogeneity p-Value
I2 (%)
<0.00001 <0.00001 0.01 <0.00001 0.33 <0.00001
58 4 39 80 0 31
1448
V. Shivarov et al. / Leukemia Research 37 (2013) 1445–1450
Fig. 1. Forest plots of pooled HRs and the corresponding 95% CIs for the prognostic significance of DNMT3A mutations in the entire cohort of AML patients: (A) OS endpoint and (B) RFS endpoint.
other genes (such as FLT3) [19]. Actually, several studies reported multivariate analysis of the prognostic effect of DNMT3A including FLT3/NPM1 mutational status. Thol et al., [9] showed that DNMT3A mutations remained an independent negative prognostic marker for OS. Marcucci et al. [17] did not find adverse impact on OS in multivariate analysis based on the entire cohort they analyzed but did so in the subgroup analysis based on the age stratification.
Renneville et al. [14] included NPM1/FLT3-ITD/CEBPA low- vs. highrisk mutational status as covariate in their multivariate analysis and DNMT3A did not appear to be a significant factor for adverse prognosis. Ribeiro et al. [15] also included FLT3/NPM1 mutational status in their multivariate analysis and found that DNMT3A mutations were independent factor for worse OS. Gaidzik et al. [18], however, failed to demonstrate that DNMT3A mutations worsen
Fig. 2. Forest plots of pooled HRs and the corresponding 95% CIs for the prognostic significance of DNMT3A mutations in the subgroup of AML patients with normal cytogenetics aged under 60 years: (A) OS endpoint and (B) RFS endpoint.
V. Shivarov et al. / Leukemia Research 37 (2013) 1445–1450
1449
Fig. 3. Forest plots of pooled HRs and the corresponding 95% CIs for OS for the prognostic significance of DNMT3A mutations in the ELN defined subgroups of AML patients: (A) Low-risk genotypes subgroup and (B) high-risk genotypes subgroups.
OS. Ideally, this issue could be resolved by performing a large scale meta-analysis on individual patients’ data, which was beyond the scope of the current work. To partially address this question we investigated the prognostic role of DNMT3A mutations within the ELN defined risk subgroups and found that DNMT3A mutations conferred adverse prognosis only in patients with high-risk genotypes (OR: 1.38; 95%CI: 1.13–1.69; p = 0.002). It remains unclear whether this effect is identical for all genotypes within the highrisk group. For instance, Ribeiro et al. [15] reported adverse effect of DNMT3A within the subgroup with FLT3wt/NPM1wt genotype but not in the subgroup with FLT3-ITD/NPM1wt genotype, whereas Thol et al. [9] reported adverse impact of DNMT3A mutations in both subgroups. Furthermore, Ley et al. [6] and Markova et al. [20] (study not included in this meta-analysis) reported adverse effect of DNMT3A mutations in FLT3 mutated cases. This is consistent with the proposed inclusion of DNMT3A mutations along with other mutations as an additional enhancer of the poor prognosis conferred by the presence of FLT3-ITD mutations as proposed Patel et al. [4] based on integrated mutational profiling. The reason for the differences in the reported OS between various AML studies with relatively similar patient profile is not clear. However, several factors not reported in most AML studies that investigates the prognostic value of these molecular mutations may be responsible including the differences in number of patients who subsequently underwent allogeneic stem cell transplantation, the presence of other molecular mutations that may confer differences in survival outcomes in AML including mutations in spliceosome genes (SF3B1, U2AF1, SRSF2), EZH2, SETBP1 and others and differences in other disease related pathogenetic factors like differences in microRNA profile, number and types of additional lesions specifically acquired somatic uniparental discomy detected by single nucleotide polymorphism array analysis (SNP-A) karyotyping. Another level of complexity adds the observation that DNMT3A mutations are scattered along the full length of the gene but tend to cluster at codon 882. The prognostic effect of R882 mutations compared to missense DNMT3A mutations affecting other codons was not assessed in this study due to lack of available published studies directly comparing the effect of the two types of mutations. However, it is likely that there might not be a difference in the prognostic power compared to that of other mutations as suggested by Ley et al. [6] who did not find any difference in the OS between the two groups. Furthermore, Ribeiro et al. [15] showed that R882 only mutations are associated with shorter OS (p = 0.018)
and RFS (p = 0.029). However, Marcucci et al. [17] showed that R882 mutations did not modify the risk of younger adults with NC-AML but did so in the age group over 60 years. In the cohort analyzed by Gaidzik et al. [18] there was strong evidence that R882-mutated vs. non-R882-mutated cases had opposing effects – unfavorable for R882 DNMT3Amut on RFS, and favorable for non-R882 DNMT3Amut on OS. These effects were consistent for both entire cohort and CN-AML. In addition to the prognostic value of DNMT3A mutations it is interesting to address the potential predictive role of these mutations. Although this issue was not consistently covered by the studies included in the current analysis, there were some hints that this might indeed be the case. Patel et al. [4] showed that doseintensified chemotherapy improved the outcome of the DNMT3A mutated cases. Notably, however, Ribeiro et al. [15] did not find any change in the risk conferred by DNMT3A mutation in AML patients who underwent allogeneic transplantation. Gaidzik et al. [18] also showed that DNMT3A mutations did not have any effect on RFS and OS of patients who have undergone allogeneic hematopoietic stem cell transplantation. Another study not included in the analysis reported a better response of DNMT3A mutated cases to idarubicin in comparison to daunorubicin [21]. Finally, it has already been proposed that DNMT3A mutations might be predictive for a better response to decitabine treatment [22]. In conclusion, our meta-analysis of eight studies with over 4500 AML patients and almost 1000 DNMT3A mutated cases showed conclusively an independent adverse prognostic effect of the mutations on the OS and RFS in the entire cohort. This was true also for the intermediate risk patients with NC-AML under the age of 60 years and patients with ELN defined high-risk genotypes. Therefore, DNMT3A mutational status can improve the risk stratification of AML patients in the settings of integrated mutational profiling as it has already been shown by some individual studies [4]. Conflict of interest All authors have no conflict of interest to declare. Funding source This work was partially supported through Grant ID 09 157 (Contract 5/16.12.2009) of the National Science Fund, Bulgaria.
1450
V. Shivarov et al. / Leukemia Research 37 (2013) 1445–1450
Acknowledgements None. Contributions. VS proposed research, gathered and analyzed data and wrote the manuscript, RG analyzed data and wrote the manuscript, AS gathered data and RT analyzed data and wrote the manuscript. References [1] Marcucci G, Haferlach T, Dohner H. Molecular genetics of adult acute myeloid leukemia: prognostic and therapeutic implications. J Clin Oncol 2011;29(5):475–86. [2] Sanders MA, Valk PJ. The evolving molecular genetic landscape in acute myeloid leukaemia. Curr Opin Hematol 2013;20(2):79–85. [3] Dohner H, Estey EH, Amadori S, et al. Diagnosis and management of acute myeloid leukemia in adults: recommendations from an international expert panel, on behalf of the European LeukemiaNet. Blood 2010;115(3):453–74. [4] Patel JP, Gonen M, Figueroa ME, et al. Prognostic relevance of integrated genetic profiling in acute myeloid leukemia. N Engl J Med 2012;366(12):1079–89. [5] Yamashita Y, Yuan J, Suetake I, et al. Array-based genomic resequencing of human leukemia. Oncogene 2010;29(25):3723–31. [6] Ley TJ, Ding L, Walter MJ, et al. DNMT3A mutations in acute myeloid leukemia. N Engl J Med 2010;363(25):2424–33. [7] Yan XJ, Xu J, Gu ZH, et al. Exome sequencing identifies somatic mutations of DNA methyltransferase gene DNMT3A in acute monocytic leukemia. Nat Genet 2011;43(4):309–15. [8] Okano M, Bell DW, Haber DA, Li E. DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development. Cell 1999;99(3):247–57. [9] Thol F, Damm F, Ludeking A, et al. Incidence and prognostic influence of DNMT3A mutations in acute myeloid leukemia. J Clin Oncol 2011;29(21):2889–96. [10] Hozo SP, Djulbegovic B, Hozo I. Estimating the mean and variance from the median, range, and the size of a sample. BMC Med Res Methodol 2005;5:13.
[11] Tierney JF, Stewart LA, Ghersi D, Burdett S, Sydes MR. Practical methods for incorporating summary time-to-event data into meta-analysis. Trials 2007;8:16. [12] Parmar MK, Torri V, Stewart L. Extracting summary statistics to perform meta-analyses of the published literature for survival endpoints. Stat Med 1998;17(24):2815–34. [13] Hou HA, Kuo YY, Liu CY, et al. DNMT3A mutations in acute myeloid leukemia: stability during disease evolution and clinical implications. Blood 2012;119(2):559–68. [14] Renneville A, Boissel N, Nibourel O, et al. Prognostic significance of DNA methyltransferase 3A mutations in cytogenetically normal acute myeloid leukemia: a study by the Acute Leukemia French Association. Leukemia 2012;26(6):1247–54. [15] Ribeiro AF, Pratcorona M, Erpelinck-Verschueren C, et al. Mutant DNMT3A: a marker of poor prognosis in acute myeloid leukemia. Blood 2012;119(24):5824–31. [16] Ostronoff F, Othus M, Ho PA, et al. Mutations in the DNMT3A exon 23 independently predict poor outcome in older patients with acute myeloid leukemia: a SWOG report. Leukemia 2013;27(1):238–41. [17] Marcucci G, Metzeler KH, Schwind S, et al. Age-related prognostic impact of different types of DNMT3A mutations in adults with primary cytogenetically normal acute myeloid leukemia. J Clin Oncol 2012;30(7):742–50. [18] Gaidzik VI, Schlenk RF, Paschka P, et al. Clinical impact of DNMT3A mutations in younger adult patients with acute myeloid leukemia: results of the AML Study Group (AMLSG). Blood 2013;121(23):4769–77. [19] Wakita S, Yamaguchi H, Omori I, et al. Mutations of the epigenetics-modifying gene (DNMT3a, TET2, IDH1/2) at diagnosis may induce FLT3-ITD at relapse in de novo acute myeloid leukemia. Leukemia 2013;27(5):1044–52. [20] Markova J, Michkova P, Burckova K, et al. Prognostic impact of DNMT3A mutations in patients with intermediate cytogenetic risk profile acute myeloid leukemia. Eur J Haematol 2012;88(2):128–35. [21] LaRochelle O, Bertoli S, Vergez F, et al. Do AML patients with DNMT3A exon 23 mutations benefit from idarubicin as compared to daunorubicin? A single center experience. Oncotarget 2011;2(11):850–61. [22] Metzeler KH, Walker A, Geyer S, et al. DNMT3A mutations and response to the hypomethylating agent decitabine in acute myeloid leukemia. Leukemia 2012;26(5):1106–7.