Prognostic role of EZH2 in gliomas: a meta-analysis

Abstract

Background:

The aim of this study was to elucidate the clinicopathological significance of enhancer of zeste homolog 2 (EZH2) expression in gliomas, including its incidence and prognostic role. In addition, we investigated the concordance between immunohistochemistry and polymerase chain reaction for determining the presence or absence of EZH2 in these tumors.

Methods:

The current meta-analysis included 1,049 gliomas with various WHO tumor grades from 12 eligible studies, which were analyzed for positivity of EZH2 and correlation between EZH2 expression and prognosis. Subgroup analyses were performed based on detection methods and WHO tumor grades.

Results:

We found the estimated positive rate of EZH2 in gliomas to be 0.663 (95% confidence interval [95% CI], 0.549-0.761). There was no difference between immunohistochemistry and polymerase chain reaction in determination of EZH2 positivity (0.706, 95% CI, 0.539-0.831 vs. 0.673, 95% CI, 0.472-0.825). The positive rate of EZH2 increased by increasing WHO tumor grade. EZH2 expression was significantly correlated with worse overall and progression-free survival (hazard ratio [HR] = 2.436, 95% CI, 1.350-4.393, and HR = 4.071, 95% CI, 1.325-12.508, respectively). The overall concordance rate between immunohistochemistry and polymerase chain reaction was 0.885 (95% CI, 0.300-0.993).

Conclusions:

EZH2 positivity was significantly correlated with WHO tumor grade and worse prognosis in gliomas.

Keywords

Concordance analysis EZH2 Glioma Meta-analysis Prognosis

Introduction

Up-regulation of enhancer of zeste homology 2 (EZH2) has been reported in several malignant tumors, such as tumors of the head and neck, breast, stomach, liver, lung and prostate (1 -6). EZH2, which is a component of polycomb repressive complexes 2, is involved in proliferation, invasion and metastasis of cancer cells (7). Additionally, EZH2 leads to methylation and phosphorylation of STAT3 in carcinogenesis of glioblastoma stem-like cells (8). Recent studies have reported that EZH2 is significantly correlated with worse survival in various cancers, including non-small cell lung cancer and female reproductive system cancer (4, 9). The prognostic roles of EZH2 in gliomas are not fully understood (10 -16). EZH2 is overexpressed in isocitrate dehydrogenase (IDH)-wildtype glioblastoma (17); however, the incidence of EZH2 has not been fully elucidated in gliomas of various grades. Methods such as immunohistochemistry (IHC) and polymerase chain reaction (PCR) can be used in the detection of EZH2 protein or mRNA expression. Although PCR-based assays are more sensitive, they are more expensive and time-consuming when compared with IHC. PCR-based assays are not available in every pathology laboratory, whereas IHC methods are widely available. In addition, in daily practice, the IHC method is preferentially used as a screening test. However, the diagnostic accuracy of IHC for determining EZH2 positivity in glioma has not yet been fully investigated.

Detailed information such as the positive rate of EZH2 and the concordance between EZH2 IHC and PCR is not fully available for gliomas (11, 12, 14, 16, 18 -24). In this study, we performed a meta-analysis to confirm the incidence and prognostic potential of EZH2 in gliomas. In addition, a concordance analysis between EZH2 IHC and PCR was conducted to examine the relative utility of these methods.

Materials and methods

Published study search and selection criteria

Relevant articles were found by searching the PubMed and MEDLINE databases through December 31, 2016, using the following key words: “glioma” and “enhancer of zeste homolog 2 or EZH2.” Titles and abstracts of all articles found were screened for exclusion, and review articles were screened to find additional eligible studies. Search results were then reviewed and included if the study had been performed in human glioma and if there was information about the positive rate of EZH2, the correlations between EZH2 expression and prognosis, and the concordance between EZH2 IHC and PCR. Articles were excluded if they were case reports or not original articles, or non-English-language publications.

Data extraction

Data from all eligible studies were extracted by 2 independent authors. The extracted data from each of the eligible studies included the following (11, 12, 14, 16, 18 -24): the first author’s name, year of publication, study location, antibody clone and manufacturer, antibody dilution ratio, evaluation criteria, number of patients analyzed and WHO tumor grade. For concordance analysis, we extracted all data associated with results of IHC and PCR. Some patients in 2 reports (21, 23) were duplicated with those of other studies (14, 18). Two reports were excluded in estimation of overall EZH2 positive rate (21, 23). For quantitative aggregation of survival results, the correlation between EZH2 expression and overall or disease-free survival rate was analyzed according to the hazard ratio (HR) using one of the following 3 methods: In studies not quoting the HR or its confidence interval (CI), these variables were calculated from the presented data using the HR point estimate, log-rank statistic or its p value, and the O-E statistic (difference between the number of observed and expected events) or its variance. If those data were unavailable, HR was estimated using the total number of events, number of patients at risk in each group, and the log-rank statistic or its p value. Finally, if the only useful data were in the form of graphical representations of survival distributions, survival rates were extracted at specified times to reconstruct the HR estimate and its variance, under the assumption that patients were censored at a constant rate during the time intervals (25). The published survival curves were read independently by 2 authors to reduce reader variability. The HRs were then combined into an overall HR using Peto’s method (26).

Statistical analysis

For meta-analysis, all data were analyzed using the Comprehensive Meta-Analysis software package (Biostat, Englewood, NJ, USA). We investigated the positive rate of EZH2 in gliomas, regardless of detection method. In addition, the correlations between EZH2 expression and worse survival were evaluated. The concordance rates were measured by agreement rates between EZH2 IHC and PCR in 2 eligible studies (14, 18). Because eligible studies used various IHC and PCR methods and evaluation criteria for various populations, a random-effects model was more suitable than a fixed-effects model. Heterogeneity between the studies was checked using the Q and I² statistics, and presented using p values. Additionally, sensitivity analysis was used to assess the heterogeneity of eligible studies and the impact of each study on the combined effect. Publication bias was assessed using Begg’s funnel plot and Egger’s test. If significant publication bias was found, the fail-safe N and trim and fill tests were performed to confirm the degree of publication bias. The results were considered statistically significant at a p value of <0.05.

Results

Selection and characteristics

One hundred twenty-four reports were identified in the database search. Thirty-nine reports were duplicated articles and were excluded. Eighteen reports were excluded due to insufficiency or lack of information on the positive rate of EZH2 or the correlation between EZH2 and survival rate. In addition, 45 reports were excluded because they reported the results using animals or cell lines (n = 38) or of other diseases (n = 7), and 10 were excluded because they were not original articles (n = 9) or were not in English (n = 1). Finally, 12 reports were included in this meta-analysis (Fig. 1; Tab. I). We ultimately included 1,049 patients from 12 eligible studies. The number of patients from each eligible study ranged from 24 to 128.

Fig. 1

Flow chart of study search and selection methods.

Table I

Main characteristics of the eligible studies

Study lead author, date (ref.)	Country	Method	No. of patients	No. of positive cases	WHO tumor grade				Histotype	No. of cases
					I	II	III	IV
Ahmed, 2016 (11)	Egypt	IHC	40	22	5	9	11	15	PA	5
									DA	9
									AA	11
									GBM	15
Ailon, 2015 (12)	Canada	IHC	201	ND					Intracranial ependymoma	201
Lin, 2015 (18)	China	PCR	128	70	18	14	38	58	ND
Orzan, 2011 (19)	UK	IHC	57	54				57	GBM	57
Pathak, 2015 (20)	India	IHC	24	12				24	Pediatric GBM	24
Purkait, 2016 (14)	India	IHC	93	78		10	12	71	DA	10
		PCR	93	93					AA	12
									GBM	71
Purkait, 2015 (21)*	India	IHC	108	76		26	27	55	PA	10
									DA	15
									AA	17
									GBM	55
									OG	11
									AOG	10
Sharma, 2016 (22)	India	IHC	55	39		14		41	DA	14
									GBM	41
Wiese, 2016 (15)	Germany	PCR	72	34				72	GBM	72
Wu, 2013 (23)^†	China	IHC	128	80	18	14	38	58	PA	18
									DA	14
									AA	38
									GBM	58
						15^‡			OA/OG	15
Zhang, 2015 (16)	China	IHC	83	43				83	GBM	83
Zheng, 2011 (24)	USA	PCR	60	45		15	25	30	DA	15
									GBM	30
									OA/OG	15

AA = anaplastic astrocytoma; AOG = anaplastic oligodendroglioma; DA = diffuse astrocytoma; GBM = glioblastoma; IHC = immunohistochemistry; ND = no description; OA = oligoastrocytoma; OG = oligodendroglioma; PA = pilocytic astrocytoma; PCR = polymerase chain reaction.

Some patients were duplicated in the report by Purkait et al (14).

†

Some patients were duplicated in the report by Lin et al (18).

‡

Included both grade II and III tumors.

Meta-analysis

The estimated positive rate of EZH2 in cases overall was 0.663 (95% CI, 0.549-0.761; Tab. II). EZH2 positive rates ranged from 47.2% to 94.7%. Subgroup analyses were performed according to the WHO tumor grades and detection methods. The incidence of EZH2 significantly increased by increasing WHO tumor grade. The EZH2 positive rates of WHO grade I, II, III and IV subgroups were 0.100 (95% CI, 0.018-0.399), 0.381 (95% CI, 0.258-0.520), 0.531 (95% CI, 0.358-0.697) and 0.772 (95% CI, 0.613-0.879), respectively (Tab. II). The subgroup of studies utilizing the IHC method showed slightly higher EZH2 positivity than the PCR subgroup (0.706, 95% CI, 0.539-0.831 vs. 0.673, 95% CI, 0.472-0.825). However, there was no statistically significant difference between the IHC and PCR subgroups in the metaregression test (p = 0.848).

Table II

Positive rates of EZH2 in gliomas

Stratified analysis	Number of studies	Fixed effect (95% CI)	Heterogeneity, p value	I²	Random effect (95% CI)	Egger’s test p value
Overall	9	0.616 (0.574-0.657)	<0.001	85.3%	0.663 (0.549-0.761)	0.097
WHO grade
Grade I	2	0.126 (0.044-0.310)	0.201	38.7%	0.100 (0.018-0.399)	-
Grade II/III	7	0.482 (0.389-0.575)	0.156	35.7%	0.482 (0.362-0.603)	0.949
Grade II	4	0.381 (0.258-0.520)	0.518	0.0%	0.381 (0.258-0.520)	0.238
Grade III	2	0.518 (0.379-0.653)	0.249	24.6%	0.531 (0.358-0.697)	-
Grade IV	8	0.658 (0.603-0.708)	<0.001	89.4%	0.772 (0.613-0.879)	0.007
Method
IHC	6	0.662 (0.605-0.715)	<0.001	87.1%	0.706 (0.539-0.831)	0.314
PCR	4	0.576 (0.515-0.636)	<0.001	86.8%	0.673 (0.472-0.825)	0.160

CI = confidence interval; IHC = immunohistochemistry; PCR = polymerase chain reaction.

The WHO grade IV subgroup exhibited a publication bias as determined by the Egger’s test (p = 0.007). To define the degree of publication bias, we conducted the fail-safe N and the trim and fill tests as secondary assessment. The publication bias of the WHO grade IV subgroup was insignificant in both the fail-safe N test and the trim and fill test. In other subgroups, there emerged no significant publication biases.

Next, we investigated the correlation between EZH2 and prognosis and found that EZH2 positivity was significantly correlated with worse overall and progression-free survival rates (HR = 2.436, 95% CI, 1.350-4.393 and HR = 4.071, 95% CI, 1.325-12.508, respectively) (Tab. III). Regarding overall survival, the estimated HRs of IHC and PCR methods were 3.018 (95% CI, 1.616-5.638) and 1.802 (95% CI, 0.592-5.482), respectively. However, a significant correlation was not found between EZH2 and worse prognosis in studies that utilized the PCR method. No publication bias emerged from analysis of the correlation between EZH2 and prognosis.

Table III

Correlation between EZH2 and survival in gliomas

Stratified analysis	Number of studies	Fixed effect (95% CI)	Heterogeneity, p value	I²	Random effect (95% CI)	Egger’s test p value
Overall survival	6	2.018 (1.589-2.563)	<0.001	81.7%	2.436 (1.350-4.393)	0.230
Method
IHC	5	2.587 (1.962-3.411)	0.001	77.5%	3.018 (1.616-5.638)	0.375
PCR	2	1.832 (1.310-2.561)	0.001	90.9%	1.802 (0.592-5.482)	-
WHO grade IV	1	1.277 (0.869-0.877)	0.128	56.9%	1.342 (0.732-2.459)	-
Progression-free survival	3	2.529 (1.741-3.672)	0.010	78.3%	4.071 (1.325-12.508)	0.349

CI = confidence interval; IHC = immunohistochemistry; PCR = polymerase chain reaction.

Concordance analysis

To evaluate the diagnostic role of EZH2 IHC, concordance analysis between EZH2 IHC and PCR was performed. Overall concordance rate between IHC and PCR was 0.885 (95% CI, 0.300-0.993) (Tab. IV). Concordance rates between IHC and PCR of EZH2 positive and negative cases were 0.961 (95% CI, 0.552-0.998, p = 0.036 in the metaregression test) and 0.639 (95% CI, 0.001-1.000, p = 0.886 in the metaregression test), respectively; however, concordance rates between EZH2 positive and negative cases in the metaregression test were not significantly different (p = 0.456).

Table IV

Concordance analysis of EZH2 immunohistochemistry and polymerase chain reaction in gliomas

Stratified analysis	Number of studies	Fixed effect (95% CI)	Heterogeneity, p value	I²	Random effect (95% CI)	Egger’s test p value
Overall	4	0.878 (0.795-0.930)	<0.001	86.5%	0.885 (0.300-0.993)	0.969
Positive cases	2	0.892 (0.813-0.940)	0.033	78.0%	0.961 (0.552-0.998)	-
Negative cases	2	0.649 (0.203-0.930)	<0.001	93.6%	0.639 (0.001-1.000)	-

CI = confidence interval.

Discussion

This study investigated EZH2 positivity in human gliomas of WHO grades I-IV from eligible studies. The incidence of EZH2 ranged from 47.2% to 94.7% among eligible studies. The inclusion of tumors with variable grades or detection methods could be responsible for the variability in EZH2 incidence. In a subgroup analysis based on the WHO tumor grade, the positive rates of EZH2 significantly increased by increasing WHO tumor grade (p<0.001 in the metaregression test). The EZH2 positive rates of WHO grade I, II, III and IV gliomas were 0.100, 0.381, 0.531 and 0.772, respectively. Given that many of the cases we included were glioblastoma (WHO grade IV) cases, the overall positive rate of EZH2 is understandably high (0.663, 95% CI, 0.549-0.761). In previous studies, loss-of-function mutations of EZH2 appeared rare in IDH-wildtype pediatric glioblastoma (27, 28). However, in our meta-analysis, the positive rate of EZH2 in pediatric glioblastoma was 0.500 (95% CI, 0.310-0.690), albeit this was lower than the overall positive rate of EZH2 in glioblastoma. More investigations are needed to further delineate EZH2 expression in specific types of gliomas.

Each eligible study used particular detection methods and evaluation criteria for EZH2 positivity. We performed a subgroup analysis based on EZH2 antibody manufacturers and found no significant difference between manufacturers (p = 0.714 in the metaregression test; data not shown). However, subgroup analysis of evaluation criteria of EZH2 IHC could not be performed because of insufficient information. Therefore, the impact of different cutoffs on the EZH2 IHC positivity was unclear.

In our subgroup analysis based on detection methods, the difference of EZH2 positivity between IHC and PCR methods was insignificant (p = 0.848 in the metaregression test). Molecular tests, such as PCR, possess disadvantages such as high costs, the need for multiple steps, skill and time compared with IHC. Therefore, we performed a concordance analysis to confirm the diagnostic accuracy of EZH2 IHC and determined an overall concordance rate between IHC and PCR of 0.885 (95% CI, 0.300-0.993). To evaluate the cause of discordance, an additional subgroup analysis was performed to examine EZH2 IHC positivity versus negativity. The concordance rate of EZH2 positive cases was higher than that of EZH2 negative cases (0.961, 95% CI, 0.552-0.998 vs. 0.639, 95% CI, 0.001-1.000). In other words, in EZH2 IHC, false negative cases were frequently found. Discordance between EZH2 IHC and PCR may result from the discrepancy between EZH2 mRNA expression and EZH2 protein expression after translation. In addition, individual evaluation criteria for EZH2 IHC and PCR may have influenced concordance rates between detection methods. Prospective studies with standardized evaluation criteria for EZH2 IHC and PCR would more precisely elucidate the difference in results obtained with these 2 methods.

Our result that EZH2 was correlated with worse survival in glioma is consistent with a previously published meta-analysis (10). However, EZH2 can also exist in lower grade gliomas (WHO I to III); therefore, the diagnostic role of EZH2 may be limited in gliomas. Interestingly, in glioblastomas, there was no significant correlation between EZH2 positivity and overall survival rate (HR = 1.342, 95% CI, 0.732-2.459). A higher EZH2 positive rate may affect the correlation between EZH2 positivity and survival. Therefore, assessment of EZH2 positivity may be more useful for predicting prognosis in lower grade gliomas. However, in the lower grade gliomas included in this study, subgroup analysis for the correlation between EZH2 positivity and survival could not be performed due to insufficient information.

In the current meta-analysis, cases with EZH2 positivity by IHC endured a worse overall survival rate (HR = 3.018, 95% CI, 1.616-5.638). However, in the 2 eligible studies employing the PCR method, we did not find a significant correlation between EZH2 presence and worse overall survival (HR = 1.802, 95% CI, 0.592-5.482), as in a previous meta-analysis (10). There was a discrepancy between 2 eligible studies regarding the correlation with survival (15, 18). The HRs of each of the eligible studies were 1.020 (95% CI, 0.630-1.651) and 3.174 (95% CI, 1.991-5.061), respectively. The criteria for EZH2 positivity in PCR may influence the correlation between EZH2 expression and survival. Further cumulative studies will be required to elucidate the optimal method for predicting prognosis using EZH2.

There were some limitations to the current meta-analysis. First, due to the nature of the eligible studies, we included more glioblastoma cases (WHO grade IV) than lower grade glial tumors. The impact of this composition of glial tumors on EZH2 positivity and the subsequent correlation between EZH2 and survival rate is unclear. Currently, an insufficient number of studies exist in the literature to address this issue. Second, while the present study was the first meta-analysis of the concordance rate between detection methods in gliomas, a thorough diagnostic test accuracy review could not be conducted due to the small number of studies included. Third, the optimal cutoff value for EZH2 IHC or PCR positivity has not been elucidated in gliomas. The positive rates of EZH2 and concordance rates between IHC and PCR could differ by cutoff value. Unfortunately, a subgroup analysis to clarify this could not be performed due to insufficient information in the eligible studies. Fourth, to evaluate the impact of ependymomas on overall estimated HR, sensitivity analysis was performed. The HRs of overall and progression-free survival were slightly increased in sensitivity analysis (HR = 2.813, 95% CI, 1.366-5.791 and HR = 6.404, 95% CI, 3.073-13.345, respectively). However, to confirm the impact of EZH2 on the prognostic role of ependymomas, further cumulative evaluations will be needed. Fifth, although our results were largely consistent with those of a previous meta-analysis, the detailed results, such as the value of HR, differed slightly (overall survival: HR = 2.18, 95% CI, 1.25-3.80 vs. HR = 2.436, 95% CI, 1.350-4.393; and progression-free survival: HR = 2.23, 95% CI, 1.56-3.19 vs. HR = 4.071, 95% CI, 1.325-12.508) (10). To avoid selection bias introduced by different follow-up periods, extraction after a 60-month follow-up period was conducted in the present study. However, information on follow-up periods was not shown in the previous meta-analysis.

In conclusion, EZH2 expression was significantly correlated with worse overall and disease-free survival in gliomas. In addition, EZH2 positivity was higher in advanced WHO grades of glioma. However, further studies will require detailed evaluation criteria to determine the optimal EZH2 detection method.

Footnotes

Disclosures

Financial support: None.

Conflict of interest: The authors declare that they have no conflict of interest.

References

Cao

Feng

Cui

et al . Up-regulation of enhancer of zeste homolog 2 is associated positively with cyclin D1 overexpression and poor clinical outcome in head and neck squamous cell carcinoma. Cancer. 2012;118(11):2858-2871.

Chang

Yang

Xia

et al . EZH2 promotes expansion of breast tumor initiating cells through activation of RAF1-β-catenin signaling. Cancer Cell. 2011;19(1):86-100.

Cai

et al . Prognostic significance of overexpression of EZH2 and H3k27me3 proteins in gastric cancer. Asian Pac J Cancer Prev. 2012;13(7):3173-3178.

Huqun

Ishikawa

Zhang

et al . Enhancer of zeste homolog 2 is a novel prognostic biomarker in nonsmall cell lung cancer. Cancer. 2012;118(6):1599-1606.

Groner

et al . EZH2 oncogenic activity in castration-resistant prostate cancer cells is Polycomb-independent. Science. 2012;338(6113):1465-1469.

Zhuang

Wang

Huang

et al . A double-negative feedback loop between EZH2 and miR-26a regulates tumor cell growth in hepatocellular carcinoma. Int J Oncol. 2016;48(3):1195-1204.

Chang

Hung

. The role of EZH2 in tumour progression. Br J Cancer. 2012;106(2):243-247.

Kim

Woo

et al . Phosphorylation of EZH2 activates STAT3 signaling via STAT3 methylation and promotes tumorigenicity of glioblastoma stem-like cells. Cancer Cell. 2013;23(6):839-852.

Chen

Huang

Sun

et al . Enhancer of zeste homolog 2 as an independent prognostic marker for cancer: a meta-analysis. PLoS ONE. 2015;10(5):e0125480.

10.

Zhang

Chen

Zhang

Feng

. EZH2 overexpression is associated with poor prognosis in patients with glioma. Oncotarget. 2017;8(1):565-573.

11.

Ahmed

Rashed

Hegazy

Mohamed

Elmesallamy

. Prognostic value of ALDH1, EZH2 and Ki-67 in astrocytic gliomas. Turk Patoloji Derg. 2016;32(2):70-81.

12.

Ailon

Dunham

Carret

et al . The role of resection alone in select children with intracranial ependymoma: the Canadian Pediatric Brain Tumour Consortium experience. Childs Nerv Syst. 2015;31(1):57-65.

13.

Wang

et al . Combined aberrant expression of Bmi1 and EZH2 is predictive of poor prognosis in glioma patients. J Neurol Sci. 2013;335(1-2):191-196.

14.

Purkait

Sharma

Kumar

et al . Expression of DNA methyltransferases 1 and 3B correlates with EZH2 and this 3-marker epigenetic signature predicts outcome in glioblastomas. Exp Mol Pathol. 2016;100(2):312-320.

15.

Wiese

Schill

Sturm

et al . No significant cytotoxic effect of the EZH2 inhibitor tazemetostat (EPZ-6438) on pediatric glioma cells with wildtype histone 3 or mutated histone 3.3. Klin Padiatr. 2016;228(3):113-117.

16.

Zhang

Chen

Han

et al . EZH2 is a negative prognostic factor and exhibits pro-oncogenic activity in glioblastoma. Cancer Lett. 2015;356(2)(2 Pt B):929-936.

17.

Louis

Ohgaki

Wiestler

et al . WHO classification of tumours of the central nervous system. 4th ed. Lyon, France: International Agency for Research on Cancer; 2016.

18.

Lin

Zheng

et al . MicroRNA-144 suppresses tumorigenesis and tumor progression of astrocytoma by targeting EZH2. Hum Pathol. 2015;46(7):971-980.

19.

Orzan

Pellegatta

Poliani

et al . Enhancer of Zeste 2 (EZH2) is up-regulated in malignant gliomas and in glioma stem-like cells. Neuropathol Appl Neurobiol. 2011;37(4):381-394.

20.

Pathak

Jha

Purkait

et al . Altered global histone-trimethylation code and H3F3A-ATRX mutation in pediatric GBM. J Neurooncol. 2015;121(3):489-497.

21.

Purkait

Sharma

Jha

et al . EZH2 expression in gliomas: correlation with CDKN2A gene deletion/p16 loss and MIB-1 proliferation index. Neuropathology. 2015;35(5):421-431.

22.

Sharma

Purkait

Takkar

et al . Analysis of EZH2: micro-RNA network in low and high grade astrocytic tumors. Brain Tumor Pathol. 2016;33(2):117-128.

23.

Wang

et al . Combined aberrant expression of Bmi1 and EZH2 is predictive of poor prognosis in glioma patients. J Neurol Sci. 2013;335(1-2):191-196.

24.

Zheng

Houseman

Morrison

et al . DNA hypermethylation profiles associated with glioma subtypes and EZH2 and IGFBP2 mRNA expression. Neuro-oncol. 2011;13(3): 280-289.

25.

Parmar

Torri

Stewart

. Extracting summary statistics to perform meta-analyses of the published literature for survival endpoints. Stat Med. 1998;17(24):2815-2834.

26.

Yusuf

Peto

Lewis

Collins

Sleight

. Beta blockade during and after myocardial infarction: an overview of the randomized trials. Prog Cardiovasc Dis. 1985;27(5):335-371.

27.

Schwartzentruber

Korshunov

Liu

et al . Driver mutations in histone H3.3 and chromatin remodelling genes in paediatric glioblastoma. Nature. 2012;482(7384):226-231.

28.

Broniscer

McEachron

et al . St. Jude Childrens Research Hospital–Washington University Pediatric Cancer Genome Project. Somatic histone H3 alterations in pediatric diffuse intrinsic pontine gliomas and non-brainstem glioblastomas. Nat Genet. 2012;44(3):251-253.