The association between the migration inhibitory factor -173G/C polymorphism and cancer risk: a meta-analysis.

Previous studies have suggested that macrophage migration inhibitory factor (MIF) -173G/C polymorphism may be associated with cancer risk. However, previous research has demonstrated conflicting results. Therefore, we followed the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines and the meta-analysis on genetic association studies checklist, and performed a meta-analysis to investigate the association between MIF -173G/C polymorphisms and the risk of cancer. Odds ratios (ORs) and corresponding 95% confidence intervals (CIs) were combined to measure the association between MIF promoter polymorphisms and cancer risk. The pooled ORs were performed for the dominant model, recessive model, allelic model, homozygote comparison, and heterozygote comparison. The publication bias was examined by Begg's funnel plots and Egger's test. A total of ten studies enrolling 2,203 cases and 2,805 controls met the inclusion criteria. MIF (-173G/C) polymorphism was significantly associated with increased cancer risk under the dominant model (OR=1.32, 95%, CI=1.00-1.74, P=0.01) and the heterozygote comparison (OR=1.38, CI=1.01-1.87, P=0.04). In subgroup analysis, MIF polymorphism and prostate were related to increased risk of prostate and non-solid cancer. In conclusion, MIF polymorphism was significantly associated with cancer risk in heterozygote comparison. The MIF -173G/C polymorphism may be associated with increased cancer risk.

Introduction

Macrophage migration inhibitory factor (MIF) was first identified nearly 50 years ago and has been used as a cytokine and an enzyme.1,2 MIF is a member of the transferring growth factor-β (TGF-β) super family, which is expressed by a broad variety of cells, including B- and T-lymphocytes as well as endocrine, endothelial, and epithelial cells of diverse histogenetic origin.3 Presently, MIF is considered to play an important role in the pro- and anti-inflammatory response to infection since it is constitutively expressed and acts as an upstream regulator of many other inflammatory cytokines.4,5

Recently, several studies have shown that MIF can promote tumor growth and viability by modulating immune responses and supporting tumor-associated angiogenesis.6 A few experiments suggested that MIF mRNA and MIF protein are overexpressed in a number of cancers.7 Tan et al reported that MIF is upregulated in patients with pancreatic cancer and causes dysfunction of insulin secretion in β-cells.8 Krockenberger et al reported that MIF is clearly overexpressed on the protein level in invasive cervical cancer compared to cervical dysplasia.9 Two polymorphisms in the promoter region of MIF have been reported in the past. One is a single nucleotide polymorphism (SNP) at the nucleotide position −173 (G to C)10 and the other is a tetranucleotide CATT repeat beginning at position −794.11 The association between these two polymorphisms and diseases has been extended to several inflammatory conditions including Graves’ disease,12 idiopathic thrombocytopenic purpura,13 and Vogt-Koyanagi-Harada (VKH) syndrome.14 These studies indicate that these two polymorphisms of MIF are associated with inflammatory diseases. Similarly, some studies have reported that the polymorphism of MIF resulted in an increased risk of cancer. With new studies about the polymorphism of MIF and the risk of cancer emerging, there has been no meta-analysis conducted regarding the association between MIF promoter polymorphism and the risk of cancer in recent times. The aim of this study is to perform a meta-analysis of all available studies that analyze the association between the polymorphism of MIF promoter and the risk of cancer.

Materials and methods

Literature search

The preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement (Figure S1) and the meta-analysis on genetic association studies checklist (Figure S2) were followed in our meta-analysis. A comprehensive search of EMBASE, PubMed, Web of Science, OVID, Cochrane Library, and China National Knowledge Infrastructure (CNKI) was done from database inception to July 22, 2014 without language restriction. The search strategy was “macrophage migration inhibitory factor or MIF” and “polymorphism or variant or mutation or genotype.” To complete our research, we also studied the review articles and references of retrieved articles manually. The literature review was performed independently by X Zhang and J Wang and the disagreements were resolved through consensus by all the authors.15,16

Selection criteria

Studies were included in the meta-analysis if the following inclusion criteria were satisfied: 1) case-control studies focused on association between the MIF promoter polymorphism and cancer risk, 2) studies enrolled more than 30 patients, 3) studies provided sufficient data to estimate the odds ratio (OR) and 95% confidence intervals (CIs) according to MIF promoter polymorphism, and 4) when study patients overlapped with patients in other included studies, we selected the first study published. The two researchers (J Wang and X Zhang) independently read the titles and abstracts and excluded the uncorrelated studies; then the full-texts were examined by our review team. The studies were selected according to the inclusion criteria.15,16

Data abstraction

Two independent reviewers (X Zhang and J Wang) extracted the following information: authors, year of publication, country, tumor type, number of cases and controls analyzed, mean value of age, source of controls (hospital-based controls or population-based controls), and genotyping method. If both univariate and multivariate analyses were reported, we utilized the multivariate analysis because it involves observation and analysis of more than one statistical outcome variable at a time thus is more accurate. If articles provided insufficient data (missing data, inconsistencies, or any other uncertainties), we attempted to contact the first and corresponding authors for necessary information via telephone or email.15,16

Statistical analysis

ORs and corresponding 95% CIs were combined to measure the association between MIF promoter polymorphisms and cancer risk. Hardy–Weinberg equilibrium (HWE) for each study was determined by the chi-square test. The pooled ORs were calculated for the allelic model (mutation [M] allele versus [vs] wild [W] allele), dominant model (WM + MM vs WW), recessive model (MM vs WM + WW), homozygote comparison (MM vs WW), and heterozygote comparison (WM vs WW) respectively, and P<0.05 denoted statistical significance. Statistical heterogeneity among the studies was evaluated using the Q-test and I2-test. When heterogeneity among the studies was observed, the pooled OR was calculated by random-effect models. Sensitivity analyses were performed to identify the potential influence of the individual data set to the pooled ORs. Subgroup analyses were conducted with respect to cancer type and source of controls. The statistical significance was analyzed by Student’s t-test. These analyses were performed by Review Manager Version 5.1 software (http://ims.cochrane.org/revman). Both Begg’s and Egger’s tests was performed using R (http://cran.r-project.org/bin/windows/base).15,16

Results

Characteristics of identified studies

Following an initial search, 166 studies were retrieved from PubMed; 233 studies from EMBASE; 313 studies from OVID; 266 studies from Web of Science; 50 studies from Cochrane Library; 532 studies from CNKI; and five additional review articles were added to make our search comprehensive. After duplicated records were removed, 878 published studies were identified. We excluded 780 unrelated studies by reading the titles and abstracts. Next, we downloaded the full-text of the remaining 98 studies and excluded 65 unrelated studies. Of the remaining 33 studies considered for performing the meta-analysis, some studies were found to report incomplete data or report other associations between MIF and cancer. We tried our best to communicate with the first and corresponding authors to get the necessary data. Some authors were able to provide the necessary data for our study, while others did not. Ultimately, after further reviewing in detail, ten studies were included in our meta-analysis.17–26 Figure 1 shows in detail the selection process. These ten studies were published between 2005 and 2014. There were 2,203 cases and 2,805 controls included in our meta-analysis. Studies were carried out in People’s Republic of China, Taiwan, Japan, Iran, Italy, and USA. Polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) was used in seven studies.17,18,20,21,23,25,26 One study used polymerase chain reaction-single strand conformation polymorphism (PCR-SSCP).24 The other two studies employed denaturing high-performance liquid chromatography (DHLPC) wave analysis19 and a Genetic Analyzer,22 respectively. Three studies assessed prostate cancer;20,22,26 three studies assessed leukemia17,19,25 and one each for gastric cancer,24 cervical cancer,18 colorectal cancer,21 and bladder cancer.23 The genotype distribution in one study deviated from HWE.26 The main characteristics of all the included studies are listed in Table 1.

Figure 1

Flow diagram summarizing the selection of eligible studies.

Table 1

Baseline characteristics of studies included in the meta-analysis

Study	Year	Country	Tumor Type	Cases	Controls	Age	Source of controls	Genotyping method	HWE
Ramireddy et al17Leukemia	2014	Taiwan	Acute myeloid leukemia	256	256	Mean age: cases: 53.4 controls: 55.8	HB	PCR-RFLP	0.06
Wu et al18	2011	People’s Republic of China	Cervical cancer	250	147	Mean age: cases: 49.08±9.405 controls: 47.99±10.750	PB	PCR-RFLP	0.28
Ziino et al19	2005	Italy	Acute lymphoblastic leukemia	151	355	NR	PB	PCR and DHLPC Wave analysis	0.05
Razzaghi et al20	2012	Iran	Prostate cancer	61	71	NR	PB	PCR-RFLP	0.88
Ramireddy et al21CRC	2014	Taiwan	Colorectal cancer	192	256	Mean age: cases: 62.1 controls: 55.8	PB	PCR-RFLP	0.13
Meyer-Siegler et al22	2007	USA	Prostate cancer	131	128	Mean age: cases: 70.16±0.89 controls: 64.39±1.09	PB	PCR and ABI 310 Genetic analyzer	–
Yuan et al23	2012	People’s Republic of China	Bladder cancer	325	345	Cases: ≤55 years: 66 persons, >55 years: 259 persons; controls: ≤55 years: 83 persons, >55 years: 262 persons	PB	PCR-RFLP	0.94
Arisawa et al24	2007	Japan	Gastric cancer	232	430	Mean age: cases: 62.99±10.73 controls: 54.72±18.84	HB	PCR-SSCP	0.81
Xue et al25	2010	People’s Republic of China	Acute lymphoblastic leukemia	346	516	Cases: <6 years: 156 persons, ≥6 years: 190 persons; controls: <6 years: 251 persons, ≥6 years: 265 persons	PB	PCR-RFLP	0.8
Ding et al26	2009	People’s Republic of China	Prostate cancer	259	301	Cases: ≤70 years: 123 persons, >70 years: 136 persons; controls: ≤70 years: 153 persons, >70 years: 148 persons	HB	PCR-RFLP	0.01

Abbreviations: HB, hospital-based; PB, population-based; HWE, Hardy–Weinberg equilibrium; PCR-RFLP, polymerase chain reaction-restriction fragment length polymorphism; DHLPC, denaturing high-performance liquid chromatography; PCR-SSCP, polymerase chain reaction-single strand conformation polymorphism; NR, no report.

Meta-analysis

Overall, ten prospective studies enrolling 2,203 cases and 2,805 controls were included in our meta-analysis. A statistically significant association between MIF (−173G/C) polymorphism and cancer risk was found under the dominant model (OR=1.32, CI=1.00–1.74, P=0.01) (Figure 2) and the heterozygote comparison (OR=1.38, CI=1.01–1.87, P=0.04) (Figure S3). There was no statistical significant association under the recessive model (OR=0.98, 95% CI 0.67–1.45, P=0.93) (Figure S4), homozygote comparison (OR=1.02, 95% CI 0.64–1.63, P=0.93) (Figure S5), and allelic model (OR=1.32, 95% CI 1.00–1.74, P=0.05) (Figure S6). Furthermore, in our subgroup analysis, a significant association was found in the prostate group under the dominant model (OR=3.34, 95% CI 2.24–4.97, P<0.001), allelic model (OR=2.94, 95% CI 1.91–4.54, P<0.001), and heterozygote comparison (OR=2.39, 95% CI 1.65–3.47, P<0.001). MIF (−173G/C) polymorphism was also significantly associated with non-solid cancer risk under the dominant model (OR=1.27, 95% CI 1.03–1.56, P=0.03) and heterozygote comparison (OR=1.32, 95% CI 1.06–1.63, P=0.01). Table S1 presents the results of overall and subgroup analyses.

Figure 2

Forest plot of MIF –173G/C polymorphism and cancer risk in dominant model.

Abbreviation: CI, confidence interval.

Sensitivity analysis

We performed sensitivity analysis by omitting one study at a time and calculating the pooled ORs again. However, the results did not show any significant statistical differences when studies were omitted. Therefore, the stability of the study was not influenced by any individual study. Table S2 presents the sensitivity analysis in the dominant model.

Publication bias

Both Begg’s funnel plot and Egger’s test were carried out to evaluate the publication bias of the studies. The results are presented in Figure 3 and Table 2. Publication bias was found under the dominant model (P=0.0286) according to Begg’s funnel plot. When Egger’s test was performed, publication bias was found under the recessive model (P=0.0075) and homozygote comparison (P=0.03). Results indicate that there may be publication bias existing in our meta-analysis. Table 2 presents the results of Begg’s funnel plot and Egger’s test under the five genetic models.

Figure 3

Publication bias in this meta-analysis.

Notes: (A) Begg’s funnel plots of MIF −173G/C polymorphism in dominant model. (B) Egger’s test of MIF −173G/C polymorphism in dominant model.

Abbreviation: MIF, migration inhibitory factor.

Table 2

A summary of P-values for Begg’s funnel plot and Egger’s test in five genetic models

	Begg’s funnel plot	Egger’s test
Dominant model	0.0286	0.1128
Recessive model	0.1361	0.0075
Homozygote comparison	0.1361	0.03
Heterozygote comparison	0.4767	0.2992
Allelic model	0.7614	0.2373

Discussion

In our meta-analysis, ten studies enrolling 2,203 cases and 2,805 controls were included. The results indicated that MIF −173G/C polymorphism was significantly associated with cancer risk.

MIF is known as a major regulator of inflammation and a central upstream mediator of innate immune response, and functions as a key mediator to counter-regulate the inhibitory effects of glucocorticoids within the immune system.27 There are numerous studies suggesting that MIF polymorphism might be associated with the risk of immune disease. Liu et al reported that MIF polymorphism is associated with new-onset Graves’ disease in a Taiwanese Chinese population.12 Hao et al carried out a meta-analysis to investigate the association between MIF polymorphism and the risk of inflammatory bowel disease (IBD).28 They found that MIF −173G/C polymorphism contributed to the susceptibility of IBD.

MIF is also involved in cancer growth and progression. The elevated MIF and mRNA levels have been observed in many tumor cells and pre-tumor states. Krockenberger et al found that MIF was significantly overexpressed on both the protein level and the mRNA level in invasive cervical cancer and MIF protein was overexpressed in SiHA and CaSki cervical cancer cell lines.9 Huang et al reported that MIF expression levels in hepatocellular carcinoma tissues and cell lines were significantly up-regulated compared with adjacent normal tissues or a normal liver cell line.29 Moreover, several studies suggested that MIF polymorphism might be associated with the risk of cancer. Only one study reported that MIF −173G/C polymorphism is associated with a decreased risk of cancer.23 All the other studies reported the opposite conclusion. We also found a meta-analysis that investigated the association between the MIF −173G/C polymorphism and cancer risk.30 However, there were only five studies included in that meta-analysis, and the result was only under the dominant model. In recent times, some new studies have been emerging; for instance, Yuan et al reported that MIF −173G/C polymorphism is associated with decreased cancer risk.23 This conclusion contradicted with the conclusion in the previous meta-analysis. Therefore, we added new studies in our meta-analysis and calculated ORs in the dominant model, recessive model, homozygote comparison, heterozygote comparison, and allelic model. In our meta-analysis, we found that MIF −173G/C polymorphism is significantly associated with cancer risk in the dominant model (OR=1.32, 95% CI 1.00–1.74, P=0.01) and heterozygote comparison (OR=1.38, 95% CI 1.01–1.87, P=0.04). There were no significant associations between MIF −173G/C polymorphism and cancer risk in the recessive model (OR=0.98, 95% CI 0.67–1.45, P=0.93), homozygote comparison (OR=1.02, 95% CI 0.64–1.63, P=0.93), and allelic model (OR=1.32, 95% CI 1.00–1.74, P=0.05). Drawing from these results, we conclude from our meta-analysis that MIF −173G/C polymorphism might increase the risk of cancer.

There are several limitations in our meta-analysis. First, publication bias exists in the current meta-analysis. If the future studies find that MIF polymorphism was not associated with cancer risk, then publication bias might cause false outcomes. Second, there were some studies lacking in necessary data to calculate ORs under different genetic models. Although we had tried our best to communicate with the first and corresponding authors, some were unable to reply. Third, the patients included in the meta-analysis were limited. It was difficult for us to perform subgroup analyses and obtain specific results. Additionally, only papers published in English or Chinese were included in our meta-analysis, and eligible studies written in other languages that could have fulfilled our study criterion were not included.

Conclusion

Our meta-analysis concluded that MIF −173G/C polymorphism might increase the risk of cancer. Given the above limitations, more studies are needed to confirm the association between MIF polymorphism and the risk of cancer.

Preferred reporting items for systematic reviews and meta-analyses (PRISMA) checklist

Notes: Data from Moher D, Liberati A, Tetzlaff J, Altman DG. The PRIS MA Group (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRIS MA Statement. PLoS Med. 6(6):e1000097.1 For more information, visit: www.prisma-statement.org.

Meta-analysis on genetic association studies checklist

Abbreviations: Para, paragraph; SNP, single nucleotide polymorphisms.

Forest plot of MIF –173G/C polymorphism and cancer risk in heterozygote comparison.

Abbreviations: MIF, migration inhibitory factor; CI, confidence interval.

Forest plot of MIF –173G/C polymorphism and cancer risk in recessive model.

Abbreviations: MIF, migration inhibitory factor; CI, confidence interval.

Forest plot of MIF –173G/C polymorphism and cancer risk in homozygote comparison.

Abbreviations: MIF, migration inhibitory factor; CI, confidence interval.

Forest plot of MIF –173G/C polymorphism and cancer risk in allelic model.

Abbreviations: MIF, migration inhibitory factor; CI, confidence interval.

Table S1

A summary of ORs for the overall and subgroup analyses of MIF polymorphism and cancer risk

Subgroups	Dominant model (ORs)	95% CI	P-value	Recessive model (ORs)	95% CI	P-value	Allelic model (ORs)	95% CI	P-value
Overall	1.57	1.1–2.24	0.01	0.98	0.67–1.45	0.93	1.32	1.00–1.74	0.05
Prostate cancer	3.34	2.24–4.97	<0.001	–	–	–	2.94	1.91–4.54	<0.001
Other cancer	1.2	0.9–1.59	0.21	0.98	0.67–1.45	0.93	1.12	0.92–1.36	0.27
Solid cancer	1.78	1.04–3.04	0.04	1.04	0.64–1.69	0.88	1.44	0.94–2.22	0.1
Non-solid cancer	1.27	1.03–1.56	0.03	0.81	0.40–1.66	0.57	1.17	0.98–1.40	0.07
Asian	1.41	0.97–2.06	0.07	0.98	0.67–1.45	0.93	1.32	0.96–1.81	0.1
Caucasian	2.13	0.78–5.81	0.14	–	–	–	1.34	0.67–2.71	0.41
HB	1.8	1.06–3.04	0.03	0.8	0.45–1.44	0.46	1.67	0.90–3.12	0.1
PB	1.49	0.93–2.37	0.1	1.06	0.64–1.75	0.82	1.15	0.87–1.52	0.32
Subgroups	Homozygote comparison (ORs)	95% CI	P-value	Heterozygote comparison (ORs)	95% CI	P-value
Overall	1.02	0.64–1.63	0.93	1.38	1.01–1.87	0.04
Prostate cancer	–	–	–	2.39	1.65–3.47	<0.001
Other cancer	1.02	0.64–1.63	0.93	1.23	0.90–1.68	0.19
Solid cancer	1.05	0.56–2.00	0.87	1.44	0.88–2.35	0.15
Non-solid cancer	0.9	0.47–1.75	0.76	1.32	1.06–1.63	0.01
Asian	1.02	0.64–1.63	0.93	1.4	0.97–2.01	0.07
Caucasian	–	–	–	1.23	0.77–1.98	0.23
HB	0.88	0.50–1.56	0.67	1.75	1.22–2.51	0.002
PB	1.08	0.56–2.10	0.82	1.2	0.81–1.79	0.35

Abbreviations: ORs, odds ratios; MIF, migration inhibitory factor; CI, confidence interval; HB, hospital-based; PB, population-based.

Table S2

The influence of individual study on ORs in dominant model

Study omitted	Year	OR	95% CI	P-value	Heterogeneity
					I2	P-value
None		1.57	1.10–2.24	0.01	87	P<0.001
Ramireddy et al2Leukemia	2014	1.60	1.07–2.39	0.02	88	P<0.001
Wu et al3	2011	1.55	1.05–2.27	0.03	88	P<0.001
Ziino et al4	2005	1.65	1.12–2.43	0.01	88	P<0.001
Razzaghi et al5	2012	1.54	1.06–2.24	0.02	88	P<0.001
Ramireddy et al6CRC	2014	1.60	1.07–2.37	0.02	88	P<0.001
Meyer-Siegler et al7	2007	1.40	1.01–1.93	0.04	83	P<0.001
Yuan et al8	2012	1.75	1.31–2.35	0.0002	77	P<0.001
Arisawa et al9	2007	1.61	1.07–2.42	0.02	88	P<0.001
Xue et al10	2010	1.62	1.07–2.44	0.02	88	P<0.001
Ding et al11	2009	1.44	1.03–2.03	0.04	84	P<0.001

Abbreviations: OR, odds ratio; CI, confidence interval.