E-26 Transformation-specific Related Gene Expression and Outcomes in Cytogenetically Normal Acute Myeloid Leukemia: A Meta-analysis.

BACKGROUND: The E-26 transformation-specific related gene (ERG) is frequently expressed in cytogenetically normal acute myeloid leukemia (CN-AML). Herein, we performed a meta-analysis to investigate the relationship between the prognostic significance of ERG expression and CN-AML.
METHODS: A systematic review of PubMed database and other search engines were used to identify the studies between January 2005 and November 2016. A total of 667 CN-AML patients were collected from seven published studies. Of the 667 patients underwent intensive chemotherapy, 429 had low expression of ERG and 238 had high expression of ERG. Summary odds ratio (OR) and the 95% confidence interval (CI) for the ERG expression and CN-AML were calculated using fixed- or random-effects models. Heterogeneity was assessed using Chi-squared-based Q- statistic test and I2 statistics. All statistical analyses were performed using R.3.3.1 software packages (R Foundation for Statistical Computing, Vienna, Austria) and RevMan5.3 (Cochrane Collaboration, Copenhagen, Denmark).
RESULTS: Overall, patients with high ERG expression had a worse relapse (OR = 2.5127, 95% CI: 1.5177-4.1601, P = 0.0003) and lower complete remission (OR = 0. 3495, 95% CI: 0.2418-0.5051, P< 0.0001). With regard to the known molecular markers, both internal tandem duplications of the fms-related tyrosine kinase 3 gene (OR = 3.8634, 95% CI: 1.8285-8.1626, P = 0.004) and brain and acute leukemia, cytoplasmic (OR = 3.1538, 95% CI: 2.0537-4.8432, P< 0.0001) were associated with the ERG expression. In addition, the results showed a statistical significance between French-American-British (FAB) classification subtype (minimally differentiated AML and AML without maturation, OR = 4.7902, 95% CI: 2.7772-8.2624, P< 0.0001; acute monocytic leukemia, OR = 0.2324, 95% CI: 0.0899-0.6006, P = 0.0026) and ERG expression.
CONCLUSION: High ERG expression might be used as a strong adverse prognostic factor in CN-AML.

Introduction

Acute myeloid leukemia (AML) is a heterogeneous disease with diverse genomic aberrations in the subtypes.[1] While a great number of cytogenetic abnormalities has been identified in AML,[2] approximately 45% of de novo adult AML patients and 20% of pediatric AML patients are diagnosed with cytogenetically normal AML (CN-AML).[12] It is important to study the predictive molecular markers so that patients can get better treatment. The molecular aberrations that have been previously studied[3] in CN-AML patients include internal tandem duplications of the fms-related tyrosine kinase 3 gene (FLT3-ITD),[456] the nucleophosmin gene (NPM1) mutations,[7] MLL partial tandem duplication (MLL-PTD),[58] E-26 transformation-specificrelated gene (ERG)[910] and brain and acute leukemia, cytoplasmic (BAALC) expression levels.[11]

The ERG, a member of the E-26 transformation-specific (ETS) family of transcription factors, plays a major role in multiple cancers, such as Ewing's sarcoma,[1213] prostate cancer,[14] and leukemia.[1015] The ERG, which is located on chromosome band 21q22, was involved in cell proliferation, apoptosis, and differentiation.[16] Moreover, ERG over-expression was demonstrated with complex karyotypes in AML patients and found in patients with CN-AML.[17] In addition, Marcucci et al.[18] also was the first time to study and report the prognostic significance of ERG in CN-AML. In accordance with the prior report, Marcucci et al.[19] sought to validate and demonstrated that the ERG overexpression was associated with worse outcome. Therefore, ERG overexpression might be used as important markers for CN-AML patients.

Until now, however, a comprehensive analysis of all reported ERG-related CN-AML is lacking. Therefore, in the current study, we assess the prognostic values of the expression of ERG in the patients with de novo CN-AML by a meta-analysis on all published studies. The well-established genetic markers were also investigated.

Methods

Search strategy

We searched the PubMed database and other search engines for all articles on the association between ERG and leukemia (last search update Nov 3, 2016). The following terms were used in the search: “ERG or ETS-related” and leukemia and “CN-AML or cytogenetically normal acute myeloid leukemia”. All eligible studies on the topic were identified by a manual search for references of retrieved articles. Finally, 239 studies were identified.

Selection criteria

The association studies between ERG overexpression and CN-AML were included if all the following conditions were met:[1917181920] ERG gene expression was analyzed and grouped into “high” and “low” in studies; the cancer type is CN-AML; the study provides the total number of ERG high and low patients; the study is published in English or Chinese; the publication year range from January 2005 to November 2016.

The major exclusion criteria were as follows: Duplicate data; abstract, comment, review or editorial; poor study quality; the incomplete data.

We have no contact with authors. Ethical approval and informed patient consent are not required as this study is a literature review and have no direct patient contact or influence on patient care.

Data extraction and quality assessment

Two investigators independently extracted data from each study by following the predefined selection criteria and discrepancies were resolved by consensus of all investigators. All study personnel was blinded throughout the meta-analysis.

The following information was recorded for each study: The surname of the first author; year of publish; cancer type; ethnicity; number of cases and controls; risk factors.

Two investigators conducted the risk of bias assessments independently using the Cochrane Collaboration tool.[21] To assess the quality of each eligible study, two investigators worked independently to determine the adequacy of the studies, and discrepancies were resolved by discussion with all investigators. All assessors were blinded throughout the meta-analysis.

Statistical analysis

The Chi-squared-based Q-statistic test and I2 statistics were used to assess the heterogeneity. When the result of the heterogeneity test was P < 0.005, the random-effects model was used.[22] Otherwise, the fixed-effects model was selected. Funnel plots were used to diagnose a potential publication bias.[23] For the possible publication bias, trim and fill method were used to evaluate the influence to the result.[24] Sensitivity analyses were performed to assess the stability of the results by excluding one study at a time. All analyses were performed using R.3.3.1 software package (R Foundation for Statistical Computing, Vienna, Austria). All the P values were two-sided. The value of P < 0.05 was defined as statistical significance. The risk of bias assessment was performed in RevMan5.3 (Cochrane Collaboration, Copenhagen, Denmark).

Results

Characteristics of studies

Of the 239 studies identified initially, seven studies met criteria and were included in the analysis [Figure 1]. Overall, these studies contained a total of 667 patients with high/low expression of ERG. Characteristics of the included studies are summarized in Tables 1–4. The association analyses were performed between the ERG expression and the risk factor are indicated in Table 1.

Figure 1

Flow chart of the study selection process.

Table 1

General information of the patients reported in the seven references

References	Groups	Cases (n)	Age (years)	Risk factor
Marcucci et al.[18]	Low ERG	63	18–59	Sex; race; CR; relapse; FAB; M
	High ERG	21	26–59	Sex; race; CR; relapse; FAB; M
Marcucci et al.[19]	Low ERG	38	19–59	Sex; race; CR; relapse; M
	High ERG	38	19–59	Sex; race; CR; relapse; M
Metzeler et al.[9]	Low ERG	157	17–83	Sex; CR; relapse; M; FAB
	High ERG	53	18–78	Sex; CR; relapse; M; FAB
Eid et al.[20]	Low ERG	20	18–42	Sex; CR
	High ERG	10	16–40	Sex; CR
Schwind et al.[17]	Low ERG	79	60–81	Sex; CR; race; M
	High ERG	79	60–83	Sex; CR; race; M
Aref et al.[1]	Low ERG	25	3–14	Sex; CR; M
	High ERG	22	2–15	Sex; CR; M
Zheng YC[25]	Low ERG	47	24–77	Sex; CR; FAB
	High ERG	15	20–76	Sex; CR; FAB

ERG: E-26 transformation-specific related gene; M: Relative gene mutations; CR: Complete remission; FAB: French-American-British.

Table 4

ERG expression levels and FAB classification reports in the three references

References	Groups	M0/M1, n/N	M2, n/N	M4, n/N	M5, n/N	M6, n/N
Marcucci et al.[18]	Low ERG	9/63	21/63	17/63	11/63	1/63
	High ERG	10/21	6/21	5/21	0/21	0/21
Metzeler et al.[9]	Low ERG	35/157	53/157	35/157	25/157	7/157
	High ERG	28/53	12/53	11/53	0/53	21/53
Zheng YC[25]	Low ERG	0/47	14/47	10/47	19/47	4/47
	High ERG	4/15	5/15	11/15	5/15	0/15

ERG: E-26 transformation-specific related gene; FAB: French-American-British; M0: Minimally differentiated acute myeloid leukemia; M1: Acute myeloid leukemia without maturation; M2: Acute myeloid leukemia with maturation; M4: Acute myelomonocytic leukemia; M5: Acute monocytic leukemia; M6: Erythroleukemia.

Association between the E-26 transformation-specific related gene expression level and gender risk

Gender information was included in the seven studies [Table 2]. The result showed a statistical significance of heterogeneity between studies (τ2 = 0.2732; I2 = 55.6%; P = 0.0354; Figure 2); thus, random effects model was used for this analysis. Compared with the low expression ERG, there was no statistically significant difference among gender (odds ratio [OR] = 0.9639, 95% confidence interval [CI]: 0.5640–1.6476, P = 0.8932).

Table 2

ERG expression levels and clinic-pathologic report in the seven references

References	Groups	Sex (female), n/N	Race (white), n/N	CR, n/N	Relapse, n/N
Marcucci et al.[18]	Low ERG	28/63	54/63	52/63	17/63
	High ERG	11/21	19/21	16/21	13/21
Marcucci et al.[19]	Low ERG	23/38	35/38	37/38	19/38
	High ERG	24/38	35/38	30/38	27/38
Metzeler et al.[9]	Low ERG	100/157	–	108/157	99/157
	High ERG	22/53	–	25/53	41/53
Eid et al.[20]	Low ERG	7/20	–	19/20	–
	High ERG	5/10	–	3/10	–
Schwind et al.[17]	Low ERG	41/79	70/79	60/79	–
	High ERG	30/79	71/79	51/79	–
Aref et al.[1]	Low ERG	12/25	–	21/25	–
	High ERG	11/22	–	9/22	–
Zheng YC[25]	Low ERG	21/47	–	31/42	–
	High ERG	11/15	–	6/15	–

ERG: E-26 transformation-specific related gene; CR: Complete remission; –: No data.

Figure 2

Forest plot of gender risk associated with ERG overexpression and ERG low expression level reported in the seven references. ERG: E-26 transformation-specific related gene.

Association between the E-26 transformation-specific related gene expression level and race risk

It has been shown previously that the frequency of ERG overexpression varied significantly between white and nonwhite population.[171819] Here, three studies were included to assess the effect of race [Table 2]. There was no evidence of heterogeneity between studies (τ2 = 0; I2 = 0; P = 0.9193; Figure 3), thus fixed effects model was used for the analysis. Compared with the low expression ERG, there was no statistically significant difference among race (OR = 1.2012, 95% CI:0.5645–2.5564, P = 0.6342).

Figure 3

Forest plot of race risk of high ERG expression level and low ERG expression level reported in the three references. ERG: E-26 transformation-specific related gene.

Association between the E-26 transformation-specific related gene expression level and complete remission

Complete remission was reported in all seven studies,[191718192025] and the results indicated that there was no heterogeneity between studies (τ2 = 0.3021; I2 = 49.7%; P = 0.0634; Figure 4), thus fixed effects model was employed in the merging analysis. Compared with the low expression ERG, there was statistically significant difference among complete remission (OR = 0.3495, 95% CI:0.2418–0.5051, P < 0.0001).

Figure 4

Forest plot of complete remission of high ERG expression level and low ERG expression level reported in the studies. ERG: E-26 transformation-specific related gene.

Association between theE-26 transformation-specific related gene expression level and relapse

The relapse information was reported in three studies.[91819] Since there was no heterogeneity between studies (τ2 = 0; I2 = 0; P = 0.4739; Figure 5), the fixed effects model was used for this analysis. Compared with the low expression ERG, there was statistically significant difference in relapse (OR = 2.5127, 95% CI:1.5177–4.1601, P = 0.0003).

Figure 5

Forest plot of relapse of high ERG expression level and low ERG expression level reported in the studies. ERG: E-26 transformation-specific related gene.

Association between the E-26 transformation-specific related gene expression level and internal tandem duplication of the fms-related tyrosine kinase 3 gene

A total of five studies reported data for FLT3-ITD.[19171819] There was evidence of statistically significant heterogeneity between studies (τ2 = 0.4047; I2 = 62.0%; P = 0.0325; Figure 6) and a random effects model was used for merging analysis. Compared with the low expression ERG, there was statistically significant difference in FLT3-ITD (OR = 3.8634, 95% CI:1.8285–8.1626, P = 0.004).

Figure 6

Forest plot of FLT3-ITD present of ERG overexpression and ERG low expression reported in the studies. FLT3-ITD: Internal tandem duplication of the FLT3 gene; ERG: E-26 transformation-specific related gene.

Association between the E-26 transformation-specific related gene expression level and MLL partial tandem duplication

A total of four studies contained data for MLL-PTD.[9171819] There was no evidence of statistically significant heterogeneity between studies (τ2 = 0; I2 = 0; P = 0.4169; Figure 7). Compared with the low expression ERG, there was no statistically significant difference in MLL-PTD (OR = 0.7817,95% CI:0.3915–0.4078, P = 0.4851).

Figure 7

Forest plot of MLL-PTD present of ERG overexpression and ERG low expression reported in the studies. MLL-PTD: Partial tandem duplication of the MLL gene; ERG: E-26 transformation-specific related gene.

Association between the E-26 transformation-specific related gene expression level and brain and acute leukemia, cytoplasmic

A total of five studies reported data for BAALC.[117181920] There was no evidence of statistically significant heterogeneity between studies (τ2 = 0.0777; I2 = 21.8%; P = 0.2755; Figure 8). Compared with low expression ERG, there was statistically significant difference in BAALC (OR = 3.1538, 95% CI:2.0537–4.8432, P < 0.0001).

Figure 8

Forest plot of BAALC expression of ERG overexpression and ERG low expression reported in the studies. BAALC: Brain and acute leukemia, cytoplasmic; ERG: E-26 transformation-specific related gene.

Association between the E-26 transformation-specific related gene expression level and the nucleophosmin gene mutations

A total of four studies reported data for NPM1 mutations.[191719] There was no evidence of statistically significant heterogeneity between studies (τ2 = 0.046; I2 = 19.0%; P = 0.2953; Figure 9) and merging analysis is using fixed effects model. Compared with low expression ERG, there was no statistically significant difference in NPM1 (OR = 1.2471, 95% CI: 0.8378–1.8563, P = 0.2766).

Figure 9

Forest plot of NPM1 mutation of ERG overexpression and ERG low expression reported in the studies. NPM1: Nucleophosmin gene; ERG: E-26 transformation-specific related gene.

Association between theE-26 transformation-specific related gene expression level and French-American-British classification

A total of three studies reported data for French-American-British subtype.[91825] Compared with low ERG expression, there was statistically significant difference in minimally differentiated AML (M0)/AML without maturation (M1) and acute monocytic leukemia (M5) [Table 5].

Table 5

Analyses of FAB classification reports in the three references

FAB classification	Q	I2 (%)	P	OR	95% CI	P
M0/M1	2.23	10.2	0.3285	4.7902	2.7772–8.2624	<0.0001
M2	1.02	0.0	0.6016	0.7095	0.4143–1.2151	0.2112
M4	10.37	80.7	0.0056	1.8455	0.4528–7.5214	0.3927
M5	4.90	59.2	0.0863	0.2324	0.0899–0.6006	0.0026
M6	8.07	75.2	0.0177	2.1898	0.1426–33.6154	0.5738

CI: Confidence interval; OR: Odds ratio; FAB: French-American-British; M0: Minimally differentiated acute myeloid leukemia; M1: Acute myeloid leukemia without maturation; M2: Acute myeloid leukemia with maturation; M4: Acute myelomonocytic leukemia; M5: Acute monocytic leukemia; M6: Erythroleukemia.

Sensitivity analysis

Sensitivity analyses were performed by sequential removal of each eligible study to assess the influence of each study on the pooled OR in each comparison in complete remission. The results showed the reliability of the prognostic impact [Figure 10]. Due to the limited studies included in this analysis, we did not carry out the sensitivity analysis for the relapse risk factor.

Figure 10

Sensitivity analysis of fixed effect model in the risk factor of complete remission.

Subgroup analysis

Six studies[91718192025] conducted in adult-only population for complete remission were selected for subgroup analysis. Compared with low expression ERG, there was statistically significant difference among complete remission (OR = 0.38, 95% CI: 0.2586–0.5584, P < 0.0001). There was no heterogeneity between studies (τ2 = 0.2663; I2 = 47.8%; P = 0.0880; Figure 11); thus; fixed effects model was employed in the merging analysis.

Figure 11

Forest plot graph: High ERG level versus low ERG level for complete remission in adult-only participants. ERG: E-26 transformation-specific related gene.

Four studies[191719] with median as criteria for ERG high and low expression were selected for subgroup analysis. Compared with low expression ERG, there was statistically significant difference among complete remission (OR = 0.3779, 95% CI:0.2472–0.5778, P < 0.0001). We selected four studies[191719] and the results indicated that there was no heterogeneity (τ2 = 0.1764; I2 = 42.6%; P = 0.1561; Figure 12); thus, fixed effects model was employed in the merging analysis.

Figure 12

Forest plot graph: Median as cut-off ERG level for complete remission in participants. ERG: E-26 transformation-specific related gene.

Risk of bias assessment

Assessments using the Cochrane Risk of Bias tool[21] are presented in Figure 13. Detailed are provided in [Supplementary Information].

Figure 13

Risk of Bias assessment results. Green circle for low risk of bias, red circle for high risk of bias, yellow circle for unclear risk of bias.

S

Table for Risk of Bias

Items	Authors’ judgement/support for judgement
	Zheng YC[25]	Aref et al.[1]	Schwind et al.[17]	Eid et al.[20]	Metzeler et al.[9]	Marcucci et al.[19]	Marcucci et al.[18]
Random sequence generation (selection bias)	Unclear/no information on sequence generation was reported	Unclear/no information on sequence generation was reported	Unclear/no information on sequence generation was reported	Unclear/no information on sequence generation was reported	Unclear/no information on sequence generation was reported	Unclear/no information on sequence generation was reported	Unclear/no information on sequence generation was reported
Allocation concealment (selection bias)	Unclear/this information was not reported	Unclear/this information was not reported	Unclear/this information was not reported	Unclear/this information was not reported	Unclear/this information was not reported	Unclear/this information was not reported	Unclear/this information was not reported
Blinding of participants and personnel (performance bias)	High risk/the patients were treated	High risk/the patients were treated	High risk/the patients were treated	High risk/the patients were treated	High risk/the patients were treated	High risk/the patients were treated	High risk/the patients were treated
Blinding of outcome assessment (detection bias)	Unclear/this information was not provided	Unclear/this information was not provided	Unclear/this information was not provided	Unclear/this information was not provided	Unclear/this information was not provided	Unclear/this information was not provided	Unclear/this information was not provided
Incomplete outcome data (attrition bias)	Low risk/all participants completed the trial	Low risk/all participants completed the trial	Low risk/all participants completed the trial	Low risk/all participants completed the trial	Low risk/all participants completed the trial	Low risk/all participants completed the trial	Low risk/all participants completed the trial
Selective reporting (reporting bias)	Low risk/the outcome measures were reported	Low risk/the outcome measures were reported	Low risk/the outcome measures were reported	Low risk/the outcome measures were reported	Low risk/the outcome measures were reported	Low risk/the outcome measures were reported	Low risk/the outcome measures were reported
Other bias	Low risk/there was no exist inappropriate influence	Low risk/there was no exist inappropriate influence	Low risk/there was no exist inappropriate influence	Low risk/there was no exist inappropriate influence	Low risk/there was no exist inappropriate influence	Low risk/there was no exist inappropriate influence	Low risk/there was no exist inappropriate influence

Publication bias

The funnel plot found the evidence for publication bias in complete remission. The trim and fill method showed that the funnel plot needed three studies to be symmetrical [Figure 14]. Since there was heterogeneity between studies (τ2 = 0.6399; I2 = 62.9%; P = 0.0039), the merging analysis was performed using random effects model. Compared to the low expression ERG, there was statistically significant difference (OR = 0.4527, 95% CI: 0.2301–0.8905, P = 0.0217). We did not test the publication bias in relapse due to the limited number of relevant studies.

Figure 14

Funnel plots of meta-analysis before (a) and after (b) trim and fill adjustment for publication in complete remission. Each point represents a separate study for the indicated association. The red points represent the filled studies.

Discussion

ERG, located on chromosome 21q22,[26] is widely overexpressed in AML patients with complex karyotypes. The product of ERG is involved in many important pathways, such as cell proliferation, differentiation, and apoptosis.[162728] In this study, we performed a systematic study between ERG and cancer risk based on seven studies. Although the results suggested that there was no association with race (OR = 1.2012, 95% CI:0.5645–2.5564, P = 0.6342) or gender (OR = 0.9639, 95% CI: 0.5640–1.6476, P = 0.8932), the analysis showed high ERG expression level was significantly associated with high relapse (OR = 2.5127, 95% CI: 1.5177–4.1601, P = 0.0003) and inferior complete remission (OR = 0.3495, 95% CI: 0.2418–0.5051, P < 0.0001). In accordance with previous studies, ERG overexpression predicted the increased relapse risk and fewer complete remission.[1819]

FMS-like tyrosine kinase-3 gene (FLT3), a receptor tyrosine kinase, is important for the development of the hematopoietic and immune systems. Activating mutations of FLT3 are now recognized as the most common molecular abnormality in AML. FLT3/ITD occurs in 20–30% of young adults with AML and is associated with poor prognosis.[293031] In the meta-analysis, we detect an association between FLT3-ITD positive and ERG expression (OR = 3.8634, 95% CI:1.8285–8.1626, P = 0.004), in line with the previous report by Marcucci et al.[919]

BAALC, located on chromosome 8q22.3, is widely expressed in CN-AML.[32] While no significant association was observed between the mutations in NPM1 and ERG expression, we found a correlation between ERG and BAALC expression (OR = 3.1538, 95% CI: 2.0537–4.8432, P < 0.0001). The association between high ERG expression and high BAALC was consistent with the previous studies.[191933]

Although the number of the study was small in this meta-analysis, there was no heterogeneity among studies in complete remission (τ2 = 0.3021; I2 = 49.7%; P = 0.0634) and relapse (τ2 = 0; I2 = 0; P = 0.4739), indicating the results were reliable. Publication bias was found in the complete remission, however, the trim and fill analysis proved that the combined effect was statistically significant (OR = 0.4527, 95% CI: 0.2301–0.8905, P = 0.0217) in random effect model. Moreover, sensitivity analysis results showed the results of this meta-analysis were statistically reliable. Considering the age factor in CN-AML, we selected six adult-only studies to conduct the subgroup analysis. The results showed that high ERG expression was also associated with complete remission significantly in adult group (OR = 0.38, 95% CI: 0.2586–0.5584, P < 0.0001). Due to the criterion of ERG high/low group was not consistent in seven studies, we conducted the subgroup analysis with four studies using median as criteria, and there was also statistically significant difference (OR = 0.3779, 95% CI: 0.2472–0.5778, P < 0.0001) in the subgroup analysis. The results demonstrated the different criteria might not affect the analysis results, further studies will be needed in the future for validation.

There are some innovative points in this analysis. First, to the best of our knowledge, this is a rare meta-analysis study to explore the association between ERG expression and CN-AML. Second, the results found a correlation between BAALC expression and the ERG expression. Third, this meta-analysis was rarely to explore the relationship between the ERG expression and molecular markers.

However, there are some limitations for this meta-analysis. First, the selected studies are completely blind and lack of the detail clinical information. Second, the number of included studies is relatively small in this study which might be the major reason to create bias or heterogeneity. Due to the limited studies or the limited sample sizes, more studies with the large sample are required to confirm the results in the future.

In conclusion, the meta-analysis provides the prognostic value of the ERG expression in CN-AML. ERG expression appears to be associated with complete remission and relapse.

Supplementary information is linked to the online version of the paper on the Chinese Medical Journal website.

Financial support and sponsorship

The study was supported by grants from the Collaborative Innovation Projects of Science and Technology Department of Zhejiang (No. 2014F50014), and Natural Science Foundation of Zhejiang Province (No. LY13H310004).

Conflicts of interest

There are no conflicts of interest.

Table 3

ERG expression levels and other molecular markers in the six references

References	Groups	FLT3-ITD (present), n/N	MLL-PTD (yes), n/N	BAALC (high), n/N	NPM1 (mutation), n/N
Marcucci et al.[18]	Low ERG	10/63	5/63	27/63	–
	High ERG	4/21	3/21	15/21	–
Marcucci et al.[19]	Low ERG	10/38	4/38	18/38	25/38
	High ERG	25/38	1/38	23/38	29/38
Metzeler et al.[9]	Low ERG	57/157	20/157	–	87/157
	High ERG	30/53	6/53	–	27/53
Eid et al.[20]	Low ERG	–	–	13/20	–
	High ERG	–	–	8/10	–
Schwind et al.[17]	Low ERG	12/79	4/79	28/79	45/79
	High ERG	46/79	2/79	51/79	52/79
Aref et al.[1]	Low ERG	0/25	–	6/25	1/25
	High ERG	6/22	–	17/22	4/22

ERG: E-26 transformation-specific related gene; MLL-PTD: Partial tandem duplication of the MLL gene; FLT3-ITD: Internal tandem duplication of the FLT3 gene; BAALC: Brain and acute leukemia, cytoplasmic; NPM1: Nucleophosmin gene; –: No data.