The Correlation of Mouse Double Minute 4 (MDM4) Polymorphisms (rs4245739, rs1563828, rs11801299, rs10900598, and rs1380576) with Cancer Susceptibility: A Meta-Analysis.

BACKGROUND Mouse double minute 4 (MDM4) has been extensively investigated as a negative regulator of P53, its negative feedback loop, and the effect of its genetic polymorphisms on cancers. However, many studies showed varying and even conflicting results. Therefore, we employed meta-analysis to further assess the intensity of the connection between MDM4 polymorphisms and malignancies. MATERIAL AND METHODS We searched eligible articles in 5 databases (Cochrane Library, PubMed, Web of Science, Wan Fang Database, and China National Knowledge Infrastructure) up to August 2021. Odds ratios (ORs) and 95% confidence intervals (CIs) were utilized to probe the correlation of 5 MDM4 polymorphisms (rs4245739, rs1563828, rs11801299, rs10900598, and rs1380576) with carcinomas. We employed meta-regression and subgroup analysis to probe for sources of heterogeneity; Funnel plots, Begg's test, and Egger's test were used to evaluate publication bias. Sensitivity analysis was applied to assess the stability of the study. RESULTS Twenty-two studies, comprising 77 reports with 29 853 cases and 72 045 controls, were included in our meta-analysis. We found that rs4245739 polymorphism was a factor in reducing overall cancer susceptibility (dominant model, OR=0.85, 95% CI=0.76-0.95; heterozygous model, OR=0.86, 95% CI=0.78-0.96; additive model, OR=0.87, 95% CI=0.79-0.95), especially in Asian populations, and it also reduces the risk for esophageal squamous cell carcinoma (ESCC). The remaining 4 SNPs were not associated with cancers. CONCLUSIONS The rs4245739 polymorphism might reduce the risk of malignancies, especially in Asian populations, and it is a risk-reducing factor for ESCC incidence. However, rs1563828, rs11801299, rs10900598, and rs1380576 are not relevant to cancer susceptibility.

Background

According to the World Health Organization (WHO), cancer causes a tremendous burden worldwide, in both men and women, second in severity only to ischemic heart disease and stroke. Over the next 40 years, cancer will become the primary cause of death by replacing ischemic heart disease, according to WHO prediction model [1]. Studies on the oncogenesis and therapy of malignancies are urgently required. The occurrence of cancer is due to multiple factors, including alcohol consumption, smoking, obesity, and working and living environment. In addition, hereditary elements, such as single-nucleotide polymorphisms (SNPs), may contribute to the pathogenesis of cancer, as suggested by molecular epidemiological studies [2,3].

Mutation of anti-oncogene has been recognized as an essential factor in cancer genesis and progression. Mutations in the TP53 gene, an anti-oncogene found on chromosome 17p13 and coding for the p53 protein, commonly occur in human carcinomas, leading to p53 loss of activity, which results in tumor initiation and progression [4,5]. Inactivation of p53 tumor suppressor is essential for oncogenesis, and almost all cancers have p53 abnormalities. MDM2/MDM4 functions as a negative regulator for p53 and tightly regulates the activity of p53 [6]. MDM4 is named for its structural similarity to MDM2 and forms a family with it [7]. MDM2 and MDMX (also named HDMX and MDM4) proteins show deregulation in many human cancers and perform carcinogenic functions, primarily by the inhibition of tumor suppressor p53 [8]. Moreover, single-nucleotide polymorphisms in MDM4 may increase or decrease the activity of MDM4 proteins, thus affecting tumorigenesis and progression.

Since its discovery, MDM4 has been widely studied due to its structural and functional similarity to MDM2, especially focusing on its single-nucleotide polymorphisms in tumors. The associations of rs4245739, rs1563828, rs11801299, rs10900598, and rs1380576 polymorphisms with cancers have been studied more frequently, including breast cancer, gastric cancer, nasopharyngeal cancer (NPC), colorectal cancer, squamous cell carcinoma of the head and neck (SCCHN), lung cancer, esophageal squamous cell carcinoma (ESCC), prostate cancer, endometrial cancer, hereditary melanoma, ovarian cancer, non-Hodgkin lymphoma (NHL), thyroid cancer, retinoblastoma, and acute myeloid leukemia(AML) [9-31]. The association of MDM4 polymorphisms with cancers has been discussed in several published meta-analyses [32-36], but they focused on a single SNP of MDM4 or included a small number of studies, and there has been no comprehensive evaluation of the correlation of MDM4 polymorphisms with cancers. Therefore, we assessed 5 SNPs (rs4245739, rs1563828, rs11801299, rs10900598, and rs1380576), and included a greater number of studies for a more reliable investigation of the relationship between MDM4 polymorphisms and cancers.

Material and Methods

This study derived all data from published literature and did not directly involve patients. Therefore, the approval of the ethics committee and informed consent was not a requirement.

Registration Information

Platform: INPLASY

Registration number.: INPLASY2021110083

Register name: The Correlation of MDM4(rs4245739, rs1563828, rs11801299, rs10900598 and rs1380576) Polymorphisms with Cancer Susceptibility: A meta-analysis.

View website: https://doi.org/10.37766/inplasy2021.11.0083.

Literature Search

Included in this meta-analysis were studies obtained from 5 databases (Cochrane Library, PubMed, Web of Science, Wan Fang Database, and CNKI) up to August 2021. The search terms on PubMed are shown in Table 1; the terms in the other 4 databases are roughly the same. No language restrictions were used in this search. We also accessed other relevant articles based on the references of the included studies.

Table 1

Search strategy for PubMed database.

(((((((((((((“Neoplasms”[Mesh]) OR (Neoplasia[Title/Abstract])) OR (Neoplasias[Title/Abstract])) OR (Neoplasm[Title/Abstract])) OR (Tumors[Title/Abstract])) OR (Tumor[Title/Abstract])) OR (Cancer[Title/Abstract])) OR (Cancers[Title/Abstract])) OR (Malignancy[Title/Abstract])) OR (Malignancies[Title/Abstract])) OR (Malignant Neoplasms[Title/Abstract])) OR (Malignant Neoplasm[Title/Abstract])) OR (Neoplasm, Malignant[Title/Abstract])) OR (Neoplasms, Malignant[Title/Abstract])

AND

((((((((((((“MDM4 protein, human” [Supplementary Concept]) OR (MDMX protein, human[Title/Abstract])) OR (Mdm2-like p53-binding protein, human[Title/Abstract])) OR (hMDMX protein, human[Title/Abstract])) OR (Double minute 4 protein, human[Title/Abstract])) OR (Mdm4, transformed 3T3 cell double minute 4, p53 binding protein (mouse) protein, human[Title/Abstract])) OR (Hdmx protein, human[Title/Abstract])) OR (MDM4[Title/Abstract])) OR (rs4245739[Title/Abstract])) OR (rs1563828[Title/Abstract])) OR (rs11801299[Title/Abstract])) OR (rs10900598[Title/Abstract])) OR (rs1380576[Title/Abstract])

AND

((((((((((“polymorphism, single nucleotide”[MeSH]) OR (“Mutation”[Mesh])) OR (“Genetic Variation”[Mesh])) OR (“Alleles”[Mesh])) OR (nucleotide polymorphism single[Title/Abstract])) OR (nucleotide polymorphisms single[Title/Abstract])) OR (polymorphisms single nucleotide[Title/Abstract])) OR (single nucleotide polymorphisms[Title/Abstract])) OR (SNPs[Title/Abstract])) OR (single nucleotide polymorphism[Title/Abstract])) OR (Polymorphism[Title/Abstract])

AND

(((“Case-Control Studies”[Mesh]) OR (Case-Control Study[Publication Type])) OR (Studies, Case-Control[Publication Type])) OR (Study, Case-Control[Publication Type])

Inclusion criteria: (a) Studies for MDM4 gene polymorphisms and cancers (at least 1 of the 5 SNPs). (b) Case-control studies, with the case group having confirmed pathological malignancies. (c) Ability to access complete available data.

Exclusion criteria: (a) Duplicate literature. (b) Non-human experiments. We selected only the most recently published articles or studies with the largest sample sizes when multiple studies had overlapping data.

Two researchers separately performed the search and screening according to the same strategy. In case inconsistencies were found, discussions were held until the results were consistent between the 2 individuals.

Data Extraction

Both investigators separately extracted the following data from the available literature: first author, year of publication, cancer type, country region, ethnicity, control group source, genotyping method, gene frequencies in the case and control groups, and the P value of Hardy-Weinberg equilibrium (HWE). If a study did not mention the P value of HWE, the goodness-of-fit test was used to identify whether the gene frequency distribution of the control groups conformed to HWE; P>0.05 was regarded as consistent with the HWE. Reports that did not conform to HWE were excluded from this meta-analysis. The 2 researchers cross-checked the extracted data to avoid any discrepancies.

Quality Assessment

Two researchers evaluated the quality of the studies included based on the Newcastle-Ottawa Scale (NOS). The case-control trials were scored on 3 dimensions: selection, comparability, and exposure, with a total score of 9. A score of 5~9 was considered high quality, while a score of 0~4 was considered low quality [37]. In case of disagreement between 2 investigators, discussion with a third party was required until agreement was reached.

Statistical Analysis

Statistical software: STATA 16.0

Quantitative synthesis: The combined ORs and 95% CI were employed to evaluate the association of 5 MDM4 polymorphisms (Y>X) with malignancy in the dominant (XY+YY vs XX), recessive (YY vs XY+XX), heterozygous (XY vs XX), homozygous (YY vs XX), and additive (Y vs X) models. X represents the major allele and Y represents the minor allele. Gene frequencies for all genotypes were derived from included case-control studies, and related studies lacking gene frequencies were eliminated.

Heterogeneity analysis: We used Cochran’s Q test and the I2 statistic to assess the level of heterogeneity for the original studies. P values of Cochran’s Q test less than 0.05 or I2 more significant than 50% were considered significant heterogeneity. In cases of insignificant heterogeneity, fixed-effects models were employed to merge ORs to assess the association of individual models with tumor risk; otherwise, random-effects models were applied. We performed a meta-regression analysis based on publication year, ethnicity, cancer type, genotyping methods, and source of controls to explore the sources of heterogeneity (P value less than 0.05 shows heterogeneity).

Subgroup analysis: We performed subgroup analyses of the included studies for ethnicity, cancer type, and control source to further probe sources of heterogeneity and investigate the association of MDM4 polymorphisms with cancer risk.

Publication bias: We utilized the contour-enhanced funnel plot, Begg’s test, and Egger’s test to evaluate the publication bias risks. Publication bias was suggested if the funnel plot was asymmetric and/or the P value for Begg’s and Egger’s test was less than 0.05. When publication bias existed, the trim and fill method would be employed to estimate how the bias affected the study outcomes.

Sensitivity analysis: When included studies exceeded 10, the leave-one-out method was implemented to evaluate the stability of the results by re-merging the ORs with 95% CIs after excluding individual studies in turn. The results of this meta-analysis were stable if the ORs and CIs were not reversed.

Results

After systematically searching the 5 significant databases, we initially obtained 324 records. After excluding 40 duplicate records, 284 articles remained. We then closely read titles and abstracts to exclude 255 articles, including 253 irrelevant publications and 2 meta-analyses. We performed a full-text reading of the 29 papers which were left and eliminated 6 papers, 5 of which lacked gene frequencies, and 1 for overlapping data with another study. We also included 1 study that matched the criteria by hand searching. A total of 23 studies were enrolled, including 78 reports, of which 61 were dedicated to studying the link between the rs4245739 polymorphism and cancer risk [10-21,29,30]. The above process can be more intuitively understood from Figure 1.

Figure 1

Flow diagram of the search and selection of literature. (This figure was created and processed using Photoshop, CS6, Adobe Systems Software Ireland, Ltd.)

We obtained 23 studies and extracted relevant data to fill in Table 2. Thirteen articles concentrated on White populations and another 10 on Asian populations; there were no studies on African populations. The article on 3 genome-wide association studies (GWASs) by Garcia-Closaset et al contained 40 case-control studies investigating rs4245739 polymorphism and breast cancer risk for White populations. After recalculating the P values of HWE, we noticed that the gene frequencies of the control group of Florin Tripon’s study on rs4245739 and acute myeloid leukemia were not matched with HWE [31], as well as Mohammad Hashemi’s and Guo-cong Wu’s studies on rs1380576 [15,24]. We did not include these 3 reports in the meta-analysis (P values of HWE can be seen in Table 2). The quality assessment results are shown in Supplementary Table 1, where all studies had scores greater than 4, suggesting that the studies included for this meta-analysis were of relatively high quality. Finally, 21 studies were included in this meta-analysis, comprising 77 reports with 29 853 cases and 72 045 controls.

Table 2

Characteristics of the included studies.

Author	Year	Cancer-type	Country	Ethnicity	Control source	Genotype method	Case			Control			HWE (Control)
rs4245739							AA	AC	CC	AA	AC	CC
Garcia-Closas	2013	Breast cancer	Multi-center	Caucasian	Mixed	Illumina array	3318	2637	557	22825	15798	2828	0.412
JB Liu	2013	Breast cancer	China	Asian	PB	PCR-RFLP	733	67	0	686	111	3	0.801
JB Liu	2013	Breast cancer	China	Asian	PB	PCR-RFLP	278	22	0	501	96	3	0.782
LQ Zhou	2013	ESCC	China	Asian	PB	PCR-RFLP	501	37	2	478	70	2	0.946
LQ Zhou	2013	ESCC	China	Asian	PB	PCR-RFLP	529	56	3	510	88	2	0.679
CB Fan	2014	NHL	China	Asian	PB	PCR-RFLP	187	13	0	346	53	1	0.785
JB Feng	2014	Gastric cancer	China	Asian	HB	PCR-RFLP	208	209	51	210	219	64	0.845
Gansmo	2015	Breast cancer	Norway	Caucasian	PB	LightSNiP assay	966	643	108	1021	703	146	0.271
Gansmo	2015	Colon cancer	Norway	Caucasian	PB	LightSNiP assay	823	600	108	2042	1439	266	0.848
Gansmo	2015	Lung cancer	Norway	Caucasian	PB	LightSNiP assay	715	515	101	2042	1439	266	0.848
Gansmo	2015	Prostate cancer	Norway	Caucasian	PB	LightSNiP assay	1412	927	161	1021	736	120	0.271
F Gao	2015	Lung cancer	China	Asian	PB	PCR-RFLP	297	22	1	548	90	2	0.701
F Gao	2015	Lung cancer	China	Asian	PB	PCR-RFLP	183	17	0	321	77	2	0.514
Pedram	2016	Breast cancer	Iran	Caucasian	HB	T-ARMS-PCR assay	123	87	10	165	81	14	0.061
Gansmo	2016	Ovarian cancer	Norway	Caucasian	PB	LightSNiP assay	716	564	105	1021	703	146	0.271
Gansmo	2016	Endometrial cancer	Norway	Caucasian	PB	LightSNiP assay	757	541	106	1021	703	146	0.271
Khanlou	2017	Thyroid cancer	Iran	Caucasian	HB	T-ARMS-PCR assay	63	34	5	144	76	12	0.893
Hashemi	2018	Breast cancer	Iran	Caucasian	HB	T-ARMS-PCR assay	175	83	7	142	70	9	0.995
Pedram	2020	Breast cancer	Iran	Caucasian	PB	T-ARMS-PCR assay	114	82	10	120	68	11	0.946
DM Zhao	2020	Colorectal cancer	China	Asian	HB	MassARRAY	304	128	11	323	180	25	1.000
Kotarac	2020	Prostate cancer	Serbia	Caucasian	HB	TaqMan	198	131	23	182	144	31	0.890
Tripon	2020	AML	Romania	Caucasian	HB	T-ARMS-PCR assay	202	144	57	209	114	83	<0.001
rs1563828							CC	CT	TT	CC	CT	TT
CG Song	2012	Breast cancer	China	Asian	HB	MassArray	53	57	14	44	43	14	0.802
YW Zhang	2012	NPC	China	Asian	PB	RT-PCR	98	91	21	90	88	22	0.998
Thunell	2014	Hereditary melanoma	Sweden	Caucasian	PB	RT-PCR	27	21	2	389	340	70	0.940
rs1380576							CC	CG	GG	CC	CG	GG
HP Yu	2011	SCCHN	America	Caucasian	HB	TaqMan	487	477	111	518	455	106	0.917
HP Yu	2012	SCCHN	America	Caucasian	HB	TaqMan	170	158	43	150	135	36	0.798
GC Wu	2015	Gastric cancer	China	Asian	HB	TaqMan	188	281	173	212	290	218	<0.001
MY Wang	2017	Gastric cancer	China	Asian	HB	TaqMan	487	493	97	552	485	136	0.180
Hashemi	2018	Breast cancer	Iran	Caucasian	HB	T-ARMS-PCR assay	86	151	26	44	142	27	<0.001
FH Yu	2019	Retinobla-stoma	China	Asian	HB	TaqMan	77	39	10	71	59	18	0.426
rs11801299							GG	AG	AA	GG	AG	AA
HP Yu	2011	SCCHN	America	Caucasian	HB	TaqMan	684	351	40	665	376	38	0.229
HP Yu	2012	SCCHN	America	Caucasian	HB	TaqMan	118	179	74	202	109	10	0.589
MY Wang	2017	Gastric cancer	China	Asian	HB	TaqMan	380	539	158	449	532	192	0.271
Hashemi	2018	Breast cancer	Iran	Caucasian	HB	T-ARMS-PCR assay	183	75	6	164	50	4	0.997
FH Yu	2019	Retinobla-stoma	China	Asian	HB	TaqMan	39	49	38	57	64	27	0.491
rs10900598							GG	GT	TT	GG	GT	TT
HP Yu	2011	SCCHN	America	Caucasian	HB	TaqMan	307	545	223	296	552	231	0.677
HP Yu	2012	SCCHN	America	Caucasian	HB	TaqMan	233	126	12	93	156	72	0.913
MY Wang	2017	Gastric cancer	China	Asian	HB	TaqMan	547	447	83	604	462	107	0.393

ESCC – esophageal squamous cell carcinoma; NHL – non-Hodgkin lymphoma; NPC – nasopharyngeal cancer; SCCHN – squamous cell carcinoma of the head and neck; AML – acute myeloid leukemia.

Meta-Analysis Findings

Table 3 shows the outcomes of meta-analysis for all SNPs.

Table 3

Summary of the association between MDM4 polymorphisms (X>Y#) and cancers.

SNPs	N	Dominant model (XY+YY vs XX)		Recessive model (YY vs XY+XX)		Heterozygous model (XY vs XX)		Homozygous model (YY vs XX)		Additive model (Y vs X)
		OR (95% CI)	P/I2 (%)	OR (95% CI)	P/I2 (%)	OR (95% CI)	P/I2 (%)	OR (95% CI)	P/I2 (%)	OR (95% CI)	P/I2 (%)
rs4245739 (A>C)	60	0.85 (0.76, 0.95)	<0.001/ 83.9%	0.96 (0.85, 1.08)	0.038/ 38.5%	0.86 (0.78, 0.96)	<0.001/ 81.5%	0.95 (0.82, 1.10)	0.003/ 51.7%	0.87 (0.79, 0.95)	<0.001/ 84.6%
Cancer type
BC	46	0.90 (0.72, 1.13)	<0.001/ 87.9%	0.90 (0.64,1.28)	0.008/ 65.2%	0.92 (0.74, 1.15)	<0.001/ 85.3%	0.92 (0.62, 1.36)	0.002/ 70.8%	0.88 (0.72, 1.08)	<0.001/ 89.0%
ESCC	2	0.58 (0.44, 0.76)	0.464/ 0.0%	1.28 (0.34, 4.76)	0.763/ 0.0%	0.56 (0.43, 0.74)	0.484/ 0.0%	1.20 (0.32, 4.48)	0.759/ 0.0%	0.61 (0.48,0.79)	0.438/ 0.0%
GCC	3	0.91 (0.74, 1.11)	0.053/ 66.0%	0.90 (0.74, 1.09)	0.196/ 38.6%	0.94 (0.78, 1.12)	0.125/ 51.9%	0.82 (0.58, 1.16)	0.118/ 53.2%	0.90 (0.75, 1.08)	0.029/ 71.9%
LC	3	0.59 (0.29, 1.20)	<0.001/ 90.5%	1.07 (0.84, 1.35)	0.814/ 0.0%	0.58 (0.29, 1.19)	<0.001/ 90.0%	1.07 (0.84, 1.37)	0.762/ 0.0%	0.61 (0.31, 1.19)	<0.001/ 90.4%
PC	2	0.90 (0.81, 1.01)	0.433/ 0.0%	0.96 (0.77, 1.20)	0.312/ 2.1%	0.90 (0.80, 1.01)	0.618/ 0.0%	0.92 (0.73, 1.15)	0.271/ 17.5%	0.93 (0.85, 1.02)	0.299/ 7.4%
Other*	4	0.99 (0.81, 1.21)	0.044/ 63.1%	0.96 (0.80, 1.16)	0.997/ 0.0%	1.00 (0.81, 1.23)	0.045/ 62.8%	1.00 (0.83, 1.20)	0.985/ 0.0%	0.99 (0.85, 1.15)	0.062/ 59.0%
Ethnicity
Asian	9	0.57 (0.46,0.70)	0.008/ 61.7%	0.72 (0.52,0.99)	0.872/ 0.0%	0.58 (0.46, 0.71)	0.006/ 62.8%	0.68 (0.49, 0.95)	0.833/ 0.0%	0.59 (0.48, 0.72)	0.002/ 67.9%
Caucasian	51	1.04 (0.96, 1.12)	0.001/ 64.1%	0.99 (0.88, 1.13)	0.019/ 51.7%	1.04 (0.98, 1.12)	0.024/ 50.1%	1.00 (0.86, 1.16)	0.002/ 62.4%	1.02 (0.95, 1.09)	<0.001/ 70.5%
Source of controls
HB	7	0.90 (0.79,1.02)	0.130/ 41.4%	0.75 (0.58,0.97)	0.899/ 0.0%	0.93 (0.82,1.07)	0.162/ 36.7%	0.73 (0.56,0.95)	0.788/ 0.0%	0.89 (0.80,0.99)	0.181/ 34.0%
PB	13	0.79 (0.69,0.91)	<0.001/ 82.5%	0.96 (0.87,1.06)	0.941/ 0.0%	0.79 (0.69,0.91)	<0.001/ 82.3%	0.97 (0.87,1.07)	0.911/ 0.0%	0.82 (0.73,0.92)	<0.001/ 81.7%
Mixed	40	1.18 (1.12,1.24)	−/−	1.28 (1.16,1.40)	−/−	1.15 (1.09,1.21)	−/−	1.35 (1.23,1.49)	−/−	1.16 (1.11, 1.21)	−/−
rs1380576 (C>G)	4	1.02 (0.86, 1.21)	0.104/ 51.4%	0.89 (0.75, 1.06)	0.264/ 24.5%	1.09 (0.97, 1.22)	0.144/ 44.5%	0.93 (0.77, 1.12)	0.207/ 34.1%	0.98 (0.86, 1.12)	0.092/ 53.4%
Ethnicity
Asian	2	0.83 (0.46, 1.50)	0.020/ 81.7%	0.74 (0.57, 0.96)	0.660/ 0.0%	0.88 (0.47, 1.63)	0.022/ 80.9%	0.77 (0.59, 1.01)	0.313/ 1.9%	0.83 (0.55, 1.24)	0.040/ 76.2%
Caucasian	2	1.10 (0.95, 1.27)	0.681/ 0.0%	1.05 (0.83, 1.34)	0.948/ 0.0%	1.09 (0.94, 1.28)	0.680/ 0.0%	1.10 (0.85, 1.41)	0.850/ 0.0%	1.07 (0.95, 1.19)	0.735/ 0.0%
rs11801299 (G>A)	5	1.47 (0.94, 2.30)	<0.001/ 93.1%	1.75 (0.85, 3.60)	<0.001/ 90.0%	1.35 (0.93, 1.96)	<0.001/ 88.7%	2.01 (0.86, 4.71)	<0.001/ 92.0%	1.41 (0.94, 2.12)	<0.001/ 94.9%
Ethnicity
Asian	2	1.16 (0.99, 1.37)	0.448/ 0.0%	1.25 (0.58, 2.69)	0.011/ 84.5%	1.19 (1.00, 1.41)	0.820/ 0.0%	1.33 (0.65, 2.75)	0.033/ 78.1%	1.19 (0.84, 1.70)	0.045/ 75.1%
Caucasian	3	1.64 (0.68, 3.97)	<0.001/ 96.4%	2.21 (0.51, 9.55)	<0.001/ 91.7%	1.50 (0.73, 3.09)	<0.001/ 94.3%	2.63 (0.43,16.1)	<0.001/ 94.5%	1.56 (0.70, 3.48)	<0.001/ 97.1%
rs10900598 (G>T)	3	0.63 (0.31,1.26)	<0.001/ 96.9%	0.48 (0.21, 1.13)	<0.001/ 95.0%	0.70 (0.40, 1.25)	<0.001/ 95.0%	0.40 (0.13, 1.18)	<0.001/ 96.5%	0.66 (0.37, 1.17)	<0.001/ 97.7%
Ethnicity
Asian	1	1.03 (0.87, 1.21)	−/−	0.83 (0.62, 1.12)	−/−	1.07 (0.40, 1.27)	−/−	0.86 (0.63, 1.17)	−/−	0.98 (0.86, 1.12)	−/−
Caucasian	2	0.48 (0.13, 1.83)	<0.001/ 98.1%	0.34 (0.04, 2.77)	<0.001/ 97.5%	0.56 (0.19, 1.61)	<0.001/ 96.6%	0.25 (0.02, 3.44)	<0.001/ 98.2%	0.53 (0.16, 1.73)	<0.001/ 98.7%
rs1563828 (C>T)	3	0.93 (0.71, 1.22)	0.825/ 0.0%	0.78 (0.49,1.23)	0.657/ 0.0%	0.97 (0.73, 1.30)	0.867/ 0.0%	0.77 (0.47, 1.24)	0.642/ 0.0%	0.91 (0.74, 1.12)	0.750/ 0.0%
Ethnicity
Asian	2	0.97 (0.71, 1.33)	0.764/ 0.0%	0.86 (0.52, 1.40)	0.804/ 0.0%	1.00 (0.72, 1.39)	0.678/ 0.0%	0.86 (0.51, 1.45)	0.921/ 0.0%	0.95 (0.75, 1.20)	0.928/ 0.0%
Caucasian	1	0.81 (0.46, 1.43)	−/−	0.43 (0.10, 1.82)	−/−	0.89 (0.49, 1.60)	−/−	0.41 (0.10, 1.77)	−/−	0.78 (0.49, 1.24)	−/−

Bold text indicates meaningful results;

X: major allele, Y: minor allele.

BC – breast cancer; ESCC – esophageal squamous cell carcinoma; GCC – gastric cancer and colorectal cancer; LC – lung cancer; PC – prostate cancer, Other* – ovarian cancer, endometrial cancer, thyroid cancer, and non-Hodgkin lymphoma; HB – hospital-based; PB – population-based.

rs4245739 (A>C)

This polymorphism was shown to reduce the risk of overall cancers in dominant, heterozygous, and additive models by our meta-analysis (dominant model, OR=0.85, 95% CI=0.76–0.95; heterozygous model, OR=0.86, 95% CI=0.78–0.96; additive model, OR=0.87, 95% CI=0.79–0.95), as shown in Figure 2A–2C.

Figure 2

(A) Forest plot related to rs4245739 polymorphism and cancer in dominant model (CC+AC vs AA). (B) Forest plot related to rs4245739 polymorphism and cancer in the heterozygous model (AC vs AA). (C) Forest plot related to rs4245739 polymorphism and cancer in the additive model (C vs A). CI – Confidence interval; OR – odds ratio. (Figures were created using Stata.16.0 and processed with Photoshop. Stata, 16.0, StataCorp. Photoshop, CS6, Adobe Systems Software Ireland, Ltd.)

After performing subgroup analysis of cancer types, we found that this SNP was only related to ESCC risk independent of other cancers (dominant model, OR=0.58, 95% CI=0.44–0.76; heterozygous model, OR=0.56, 95% CI=0.43–0.74; additive model, OR=0.61, 95% CI=0.48–0.79). This SNP was discovered to reduce the risk of tumor development in Asian populations by performing subgroup analysis of ethnicity (dominant model, OR=0.57, 95% CI=0.46–0.70; recessive model, OR=0.72, 95% CI=0.52–0.99; heterozygous model, OR=0.58, 95% CI=0.46–0.71; homozygous model, OR=0.68, 95% CI=0.49–0.95; additive model, OR=0.59, 95% CI=0.48–0.72). For subgroup analyses of control group sources, we discovered that the SNP was associated with neoplasm risk in hospital-based, population-based, and mixed groups (HB: recessive model, OR=0.75, 95% CI=0.58–0.97; homozygous model, OR=0.73, 95% CI=0.56–0.95; additive model, OR=0.89, 95% CI=0.80–0.99. PB: dominant model, OR=0.79, 95% CI=0.69–0.91; heterozygous model, OR=0.79, 95% CI=0.69–0.91; additive model, OR=0.82, 95% CI=0.73–0.92).

rs1380576 (C>G), rs11801299 (G>A), rs10900598 (G>T) and rs1563828 (C>T)

Meta-analysis revealed that rs1380576 (C>G), rs11801299 (G>A), rs10900598 (G>T), and rs1563828 (C>T) polymorphisms were unrelated to the risk of carcinomas. Through subgroup analysis, the rs1380576 polymorphism was found to reduce the risk of tumors in Asian populations (recessive model, OR=0.74, 95% CI=0.57–0.96).

Heterogeneity Analysis

High heterogeneity was found among the studies of rs4245739, rs11801299, and rs10900598, for which we used random-effects models to combine ORs and performed subgroup analysis.

Meta-regression was performed to explore the origin of the heterogeneity for rs4245739 among publication year, ethnicity, cancer type, genotyping methods, and source of controls, the P value was less than 0.001 for ethnicity only, indicating that heterogeneity was derived from ethnicity (Supplementary Table 2), and we further discovered by subgroup analysis that heterogeneity existed mainly in studies on Asian populations.

Publication Bias

Since only the number of studies with rs4245739 was greater than 10, we performed an evaluation of publication bias for rs4245739.

The publication bias was evaluated for the 5 models of rs4245739 with contour-enhanced funnel plots, and a significant asymmetry was observed in the funnel plots for the dominant, heterozygous, and additive models (Figure 3A–3C). However, the asymmetry resulted from the distribution of studies in areas with statistical significance outside the funnel plot’s white color. This suggests that it is incredibly likely that the asymmetry in funnel plots is caused by factors other than publication bias and, most likely, by heterogeneity, since the 3 models share a significant amount of heterogeneity.

Figure 3

(A) Contour-enhanced funnel plot on the dominant model (CC+AC vs AA) of the relationship between rs4245739 and cancer susceptibility. (B) Contour-enhanced funnel plot on the heterozygous model (AC vs AA) of the relationship between rs4245739 and cancer susceptibility. (C) Contour-enhanced funnel plot on the additive model (C vs A) of the relationship between rs4245739 and cancer susceptibility. (Figures were created using Stata.16.0 and processed with Photoshop. Stata, 16.0, StataCorp. Photoshop, CS6, Adobe Systems Software Ireland, Ltd.)

To perform a quantitative assessment of publication bias, we performed Begg’s test and Egger’s test. Results of Begg’s test indicated dominant, heterozygous, and additive models with publication bias. However, Egger’s test showed publication bias for all 5 models (dominant model: Pbegg=0.032, Pegger<0.001; recessive model: Pbegg=0.526, Pegger=0.001; heterozygous model: Pbegg=0.032, Pegger<0.001; homozygous mode: Pbegg=0.487, Pegger=0.001; additive model: Pbegg=0.012, Pegger<0.001). Therefore, the publication bias of the 5 models was further evaluated with the trim and fill method, and the findings showed that the outcomes of the 5 models were not reversed after trim and fill, with no significant change in the ORs and confidence intervals. This indicates that the publication bias is within the acceptable range, and the outcomes are robust.

Sensitivity Analysis

We conducted a sensitivity analysis of the pooled results to assess the individual impact of each study. Overall findings did not change significantly after sequentially eliminating each included study, showing that the outcomes of this meta-analysis are robust and stable (Supplementary Figures 1–5).

Discussion

As a negative regulator of the P53 protein, MDM4 can downregulate P53 activity and lead to cancer. Meanwhile, MDM4 has been investigated extensively as a target spot for targeted malignancy therapy [38]. Single-nucleotide polymorphisms in the MDM4 gene can impact the MDM4 activity and hence tumor susceptibility and prognosis. The relationship between MDM4 polymorphisms and various neoplasms was investigated; however, the results were inconsistent. For this reason, it is essential to comprehensively evaluate the relationship between MDM4 polymorphisms and cancers.

This is the most comprehensive meta-analysis to date to study the association between MDM4 polymorphisms and cancers. We analyzed 5 MDM4 polymorphisms and found that rs4245739 polymorphism was a factor in reducing cancer susceptibility, while the remaining 4 SNPs (rs4245739, rs1563828, rs11801299, rs10900598, and rs1380576) were not associated with malignancies. The findings of meta-analyses by Ming-Jie Wang et al, Chaoyi Xu et al, and Yajing Zhai et al suggest that rs4245739 polymorphism decreases carcinoma susceptibility, and the conclusions were consistent with ours [32-34]; For rs4245739, rs1563828, rs11801299, rs10900598, and rs1380576, Ming-Jie Wang et al and Yajing Zhai et al concluded, in agreement with the present study, that these 4 SNPs were not associated with cancers [32,34]. Xin Jin et al investigated rs4245739 only, and found that this SNP could reduce the risk of cancer in dominant, heterozygous, and additive models, but the opposite was found in the recessive model [35]. Although Yaxuan Wang et al worked on all 5 MDM4 polymorphisms, they only evaluated 3 gene models, including alleles, dominant, and recessive; they found that rs4245739 reduced cancer risk, while the remaining 4 SNP were not associated with cancer [36]. Our meta-analysis was updated with 7 articles, including 48 case-control studies, compared to a similar meta-analysis published recently [36].

Polymorphism rs4245739 (A>C) is the most studied MDM4 gene polymorphism. Thus far, 60 case-control studies were performed to evaluate its relationship with cancer. The most studies (46) have been done on breast cancer [12,14,15,18,20,21]. Two studies from Negar Pedram concluded that the rs4245739 polymorphism has no association with breast cancer [20,21]. Three studies suggested that SNP may reduce the risk of breast cancer [12,15,18]. Garcia-Closaset et al found that rs4245739 polymorphism enhanced the chances of estrogen receptor (ER)-negative but not ER-positive breast cancer in their GWAS studies (including 40 case-control studies) [14]. Two studies from China showed the rs4245739 polymorphism can reduce ESCC risk in the Chinese population [30]. A Chinese study suggested that the rs4245739 polymorphism might be a risk factor for gastric cancer [16]. Studies on colorectal cancer and the SNP reached different conclusions; Zhao et al found that the SNP reduced the risk of colorectal cancer[29], and Gansmo et al concluded that there was no association between these [12]. Gao et al discovered that the rs4245739 polymorphism decreased lung cancer incidence [13], while Gansmo et al concluded that it was not associated with lung cancer [12]. Kotarac et al and Gansmo et al both discovered that this SNP was not associated with prostate cancer [12,17]. Ovarian cancer, endometrial cancer, thyroid cancer, and non-Hodgkin lymphoma have been less studied [10,11,19]. The rs4245739 polymorphism has been reported to be unrelated to endometrial cancer and thyroid cancer, and was reported to be associated with ovarian cancer and non-Hodgkin lymphoma. Our findings suggest that the rs4245739 polymorphism can reduce cancer risk, especially in subgroups of the Asian populations. In addition, only the risk of ESCC was associated with rs4245739 polymorphism in the subset of cancer types. All 3 teams in the subgroup of the source of controls showed statistical significance, and the mixed group contained only 1 study that showed the opposite result. We investigated the reasons for heterogeneity with meta-regression and subgroup analysis. Through meta-regression, we discovered that heterogeneity originated from ethnicity, and we further discovered that heterogeneity existed mainly in studies on Asian populations, which may also be related to factors such as living environment and diet. The asymmetry of contour-enhanced funnel plots for the dominant, recessive, and additive models may be associated with heterogeneity. However, the results of the Begg’s and Egger’s tests indicated publication bias for all 5 models, so we further evaluated them by the trim and fill method and found the publication bias to be within acceptable limits.

Relatively few studies were conducted for rs1380576 (C>G), rs11801299 (G>A), rs10900598 (G>T) and rs1563828 (C>T) [9,22-28]. Our meta-analysis demonstrated that these 4 SNPs were not associated with tumor sensitivity. However, in ethnic subgroups, we found that the rs1380576 polymorphism could reduce the risk of tumors in Asian populations; therefore, future studies on the rs1380576 polymorphism could regard this as a breakthrough. The analysis of rs11801299, rs10900598, and rs1380576 polymorphisms had a high level of heterogeneity; we speculate that this might be related to variations in tumor type and ethnicity of the included individuals.

MDM2/MDM4 is an endogenous negative regulatory factor of p53, and its expression can be activated by p53, forming a p53-MDM2/MDM4 negative feedback loop [39]. Almost all malignancies present abnormalities in the p53-MDM2/MDM4 loop, mainly involving p53 mutations or MDM2/MDM4 overexpression [40]; this abnormal pathway leads to loss of p53 tumor suppressor activity. Due to its prevalence in a variety of tumors, p53 has long been a potential target for tumor therapy. However, because of its complex mechanism, only a few drugs have entered preliminary clinical trials until recent years, in particular, drugs targeting the MDM2/MDM4 family [41-44]. Therefore, future studies need to define the specific mechanisms by which the p53-MDM2/MDM4 loop plays a role in cancers, as well as feasible interventions. In addition, the screening of populations suitable for p53-related therapy and the search for biomarkers with the predictive value of efficacy will be the focus of future research.

Our study has some potential limitations. First, this meta-analysis suffers from publication bias and heterogeneity. Second, the remaining 4 SNP datasets (rs1563828, rs11801299, rs10900598, and rs1380576) were small, making it difficult to draw reliable conclusions. Third, selective bias exists as only studies from Asian populations versus White populations were included. Fourth, the literature search only included English and Chinese articles.

Conclusions

In conclusion, we demonstrated that the rs4245739 polymorphism reduces the risk of cancers, especially in Asian populations, and it is a risk-reducing factor for ESCC incidence. More studies are needed to further explore the relationship between MDM4 polymorphisms and cancer susceptibility in the future.

Sensitivity analysis of MDM4 rs4245739 and cancer risk in dominant model. (The figure was created using Stata.16.0 and processed with Photoshop. Stata, 16.0, StataCorp. Photoshop, CS6, Adobe Systems Software Ireland, Ltd.)

Sensitivity analysis of MDM4 rs4245739 and cancer risk in the recessive model. (The figure was created using Stata.16.0 and processed with Photoshop. Stata, 16.0, StataCorp. Photoshop, CS6, Adobe Systems Software Ireland Ltd.)

Sensitivity analysis of MDM4 rs4245739 and cancer risk in heterozygous model. (The figure was created using Stata.16.0 and processed with Photoshop. Stata, 16.0, StataCorp. Photoshop, CS6, Adobe Systems Software Ireland, Ltd.)

Sensitivity analysis of MDM4 rs4245739 and cancer risk in homozygous model. (The figure is made by Stata.16.0 and processed by Photoshop. Stata, 16.0, StataCorp. Photoshop, CS6, Adobe Systems Software Ireland, Ltd.)

Sensitivity analysis of MDM4 rs4245739 and cancer risk in the additive model. (The figure was created using Stata.16.0 and processed using Photoshop. Stata, 16.0, StataCorp. Photoshop, CS6, Adobe Systems Software Ireland, Ltd.)

Supplementary Table 1

Assessment of the quality of the included studies by the Newcastle-Ottawa Scale (NOS).

Author	Year	Cancer-type	Selection				Comparability	Exposure			Score
			Criteria 1	Criteria 2	Criteria 3	Criteria 4	Criteria 5	Criteria 6	Criteria 7	Criteria 8
rs4245739
Garcia-Closas	2013	Breast cancer	*	*	–	*	*	*	–	–	5
JB Liu	2013	Breast cancer	*	*	*	*	**	–	*	–	7
JB Liu	2013	Breast cancer	*	*	*	*	**	–	*	–	7
LQ Zhou	2013	ESCC	*	*	*	*	**	–	*	–	7
LQ Zhou	2013	ESCC	*	*	*	*	**	–	*	–	7
CB Fan	2014	NHL	*	*	*	*	**	–	*	–	7
JB Feng	2014	Gastric cancer	*	*	–	*	**	–	*	–	6
Gansmo	2015	Breast cancer	*	*	*	*	**	–	*	–	7
Gansmo	2015	Colon cancer	*	*	*	*	**	–	*	–	7
Gansmo	2015	Lung cancer	*	*	*	*	**	–	*	–	7
Gansmo	2015	Prostate cancer	*	*	*	*	**	–	*	–	7
F Gao	2015	Lung cancer	*	*	*	*	**	–	*	–	7
F Gao	2015	Lung cancer	*	*	*	*	**	–	*	–	7
Pedram	2016	Breast cancer	*	*	–	*	**	–	*	–	6
Gansmo	2016	Ovarian cancer	*	*	*	*	*	–	*	–	6
Gansmo	2016	Endometrial cancer	*	*	*	*	*	–	*	–	6
Khanlou	2017	Thyroid cancer	*	*	–	*	**	–	*	–	6
Hashemi	2018	Breast cancer	*	*	–	*	**	–	*	–	6
Pedram	2020	Breast cancer	*	*	*	*	**	–	*	–	7
DM Zhao	2020	Colorectal cancer	*	*	–	*	**	–	*	–	6
Kotarac	2020	Prostate cancer	*	*	–	*	**	–	*	–	6
Tripon	2020	AML	*	*	–	*	*	*	*	–	6
rs1563828
CG Song	2012	Breast cancer	*	*	–	*	**	–	*	–	6
YW Zhang	2012	NPC	*	*	*	*	**	–	*	–	7
Thunell	2014	Hereditary melanoma	*	*	*	*	**	–	*	–	7
rs1380576
HP Yu	2011	SCCHN	*	*	–	*	**	*	*	–	7
HP Yu	2012	SCCHN	*	*	–	*	**	*	*	–	7
GC Wu	2015	Gastric cancer	*	*	–	*	**	–	*	–	6
MY Wang	2017	Gastric cancer	*	*	–	*	**	–	*	–	6
Hashemi	2018	Breast cancer	*	*	–	*	**	–	*	–	6
FH Yu	2019	Retino-blastoma	*	*	–	*	**	–	*	–	6
rs11801299
HP Yu	2011	SCCHN	*	*	–	*	**	*	*	–	7
HP Yu	2012	SCCHN	*	*	–	*	**	*	*	–	7
MY Wang	2017	Gastric cancer	*	*	–	*	**	–	*	–	6
Hashemi	2018	Breast cancer	*	*	–	*	**	–	*	–	6
FH Yu	2019	Retino-blastoma	*	*	–	*	**	–	*	–	6
rs10900598
HP Yu	2011	SCCHN	*	*	–	*	**	*	*	–	7
HP Yu	2012	SCCHN	*	*	–	*	**	*	*	–	7
MY Wang	2017	Gastric cancer	*	*	–	*	**	–	*	–	6

Criteria 1 – adequate definition of case; Criteria 2 – representativeness of the case; Criteria 3 – selection of controls; Criteria 4 – definition of controls; Criteria 5 – control for important factor; Criteria 6 – assessment of exposure; Criteria 7 – same method of ascertainment for cases and controls; Criteria 8 – non-response rate.

Supplementary Table 2

Results for meta-regression of rs4245739.

Covariates	Number of dummy variables	Dominant model	Recessive model	Heterozygous model	Homozygous model	Additive model
rs4245739
Publication year	–	0.580	0.363	0.653	0.361	0.486
Ethnicity	2	<0.001	0.548	<0.001	0.534	<0.001
Cancer type	6	0.292	0.258	0.211	0.445	0.493
Genotyping methods	4	0.517	0.679	0.487	0.744	0.496
Source of controls	3	0.275	0.412	0.203	0.358	0.291