Figures
Abstract
Background
Conflicting results were reported on the association between the TNF-α -308G/A polymorphism and idiopathic recurrent miscarriage (IRM). Though three meta-analyses have been conducted on this topic, the conclusions were contradictory, and the results may be unreliable as certain crucial conditions were neglected.
Method
A complete search was conducted in PubMed, Cochrane Library, and Embase, other sources like Google Scholar, ClinicalTrial.gov and reference lists of relevant articles were also retrieved. All candidate articles were accessed and screened using specific inclusion and exclusion criteria. Statistical analyses were performed on data extracted from eligible studies using the STATA 12.0 software and the TSA 0.9 beta software.
Results
Eventually, 12 case-control studies from 11 publications (with 1,807 cases and 2,012 controls) were included in this meta-analysis, and no evidence of any significant association was found in the overall analyses between the TNF-α -308G/A polymorphism and IRM risk. However, significant association was shown in Asian population (four studies from three publications) in the dominant model (AA + GA vs. GG), the allelic model (A vs. G), and the heterozygote model (GA vs. GG).
Citation: Dong J, Li J, Zhou G, Peng Z, Li J, Lin S, et al. (2016) No Association between TNF-α -308G/A Polymorphism and Idiopathic Recurrent Miscarriage: A Systematic Review with Meta-Analysis and Trial Sequential Analysis. PLoS ONE 11(11): e0166892. https://doi.org/10.1371/journal.pone.0166892
Editor: Jefferson Terry, BC Children's Hospital, CANADA
Received: July 12, 2016; Accepted: November 4, 2016; Published: November 28, 2016
Copyright: © 2016 Dong et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: The authors received no specific funding for this study.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Spontaneous miscarriage, which afflicts 10% to 20% of pregnant women [1–4], is a distressing experience and a common complication in early pregnancy [5, 6]. To make matters worse, 1% to 5% of women will suffer two or more consecutive, unexplained pregnancy losses with the same partner prior to the 20th week of gestation [7, 8]. This is called idiopathic recurrent miscarriage (IRM) [9]. IRM is related to a variety of causes [10], among which the disturbance of the maternal homeostatic balance between the Th1 and Th2 cytokine system, is best studied [11–13]. This balance is maintained by a series of cytokines [14]. It has been reported that Th1 cytokines are detrimental and associated with IRM, whereas Th2 cytokines are not [15, 16].
As a pro-inflammatory Th1 cytokine, tumor necrosis factor alpha (TNF-α) is mapped in chromosome 6p21.3 and mainly secreted by mono-nuclear phagocytes, lymphocytes, and natural killer (NK) cells [17]. Some researches have demonstrated that TNF-α is implicated in the development of IRM [18–21], possibly by inducing the apoptosis of trophoblasts and promoting the expression of apoptotic genes in the human fetal membrane [22, 23]. Moreover, the production of TNF-α is mainly controlled by genes, whereas mutations of these genes could result in changes of TNF-α level, especially in the promoter region [24, 25]. Therefore, polymorphisms in this region may be associated with IRM risk. And a bunch of studies have been performed to evaluate the association between TNF-α promoter polymorphisms and IRM risk [26–50].
Among all the genetic variants in this region, TNF-α -308G/A (rs1800629) is most studied [28–50]. However, the results of these studies are inconsistent and often conflicting. Although one meta-analysis of 7 studies in 2009 [48], another one of 12 studies in 2012 [49], and the third one of 10 studies in 2016 [50] have been conducted, their conclusions were conflicting and unreliable due to the inclusion of studies deviating significantly from Hardy-Weinberg equilibrium (HWE) [39, 40], and studies without sufficient data to calculate HWE [44–47]. Meanwhile, a missing study [32] in the previous meta-analyses and a couple of new studies with different results [38, 41, 42] were found by us during the investigation. Therefore, we conducted this study to obtain more concrete and conclusive conclusions concerning the correlation between the TNF-α -308G/A polymorphism and IRM through a comprehensive and robust meta-analysis.
Materials and Methods
The present meta-analysis was conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidance (S1 Table). No review protocol was registered for this study.
Identification of eligible trials
Relevant articles were identified by a comprehensive search of the following electronic databases through July 2016: PubMed, Cochrane Library, Embase. The search terms included the synonyms of miscarriage, tumor necrosis factor and polymorphism (S1 File). The SNP number (rs1800629) was also searched in combination with the synonyms of miscarriage. In addition, Google Scholar, ClinicalTrial.gov and reference lists of relevant articles were also screened by two authors independently to collect the randomized controlled trials (RCTs) published.
Inclusion and exclusion criteria
For all the relevant literature, the following inclusion criteria were adopted: 1) case-control designed studies or retrospective cohort studies with clear inclusion criteria; 2) data on allele and genotype frequencies provided; and 3) information on DNA genotyping method and characteristics of cases and controls included. Studies without genotype data or with duplicate data were excluded. Letters, case reports, editorials, review articles, conference abstracts, and animal studies were also excluded. Eligible studies were selected by the same two authors independently by screening the title, abstract, and full article based on the above criteria. Disputes were solved by consultation.
Data extraction
From all eligible studies, the following data were extracted: last name of the first author, publication date, country, ethnicity, mean age and source of the cases and controls, total sample size, genotype frequencies, and genotyping method. For each study, the HWE of the control group was computed from the genotype frequencies extracted above, and studies with p <0.05 were considered as significantly deviating from the HWE and would be excluded from this meta-analysis. If a study had subgroups, each subgroup would be listed as a separate study. Two authors completed the whole process independently. If differences existed, data would be rechecked independently by the two authors. Further discrepancies would be referred to a third author. To obtain necessary missing data, authors were contacted via e-mail.
Quality assessment
The quality of each study was assessed using the assessment scale adapted from Peng et al. [51] for the present meta-analysis (Table 1). Each study was scored and labelled as either low quality (score ≤6) or high quality (score >6) based on items such as the definition of IRM adopted, representativeness of controls, description of genotyping method, mean age of cases, and total sample size. The quality assessment was performed by two authors independently, and disagreements were settled by consultation.
Statistical analysis
Based on the genotype frequencies of cases and controls in each study, we conducted a series of overall meta-analyses using the following five genetic models: the homozygote model (AA vs. GG), the heterozygote model (GA vs. GG), the recessive model (AA vs. GA + GG), the dominant model (AA + GA vs. GG), and the allelic model (A vs. G). Then, using the odd ratio (OR) and confidence interval (CI) produced, we evaluated the association between the above genetic models and IRM risk. The overall significance of the association was calculated by a paired z-test, and a p value < 0.05 was considered significant. Heterogeneity among studies was computed by the Q statistic and the I2 statistic. For each study, either the fixed-effects model or the random-effects model was used, based on the PQ value. If the PQ value was >0.1, the former was adopted; otherwise, the latter was chosen. To investigate the influence of primary characteristics and explore the source of heterogeneity, we conducted a series of subgroup analyses. In addition, we conducted a sensitive analysis to test the stability of the overall results by sequentially taking out one study each time, a cumulative meta-analysis to portray the shift of the association over time by adding studies one by one based on publication date, and a trial sequential analysis (TSA) to minimize the risk of type I errors. Furthermore, we performed Galbraith plot to facilitate the examination of heterogeneous studies. For the evaluation of publication bias, Egger’s regression test were performed. Funnel plots and Egger’s publication-bias plots were also generated in the process. All analyses were conducted using STATA software version 12.0 and TSA software version 0.9 beta. Two-tailed p values <0.05 were considered as statistically significant.
Results
Literature selection
The initial search generated 162 relevant records, of which 32 were duplicates. After reviewing the abstracts of the remaining 130 records, 104 records were ruled out as irrelevant articles, reviews, letters or case-reports. The full texts of the left 26 potential publications were obtained and reviewed. Among them, three publications without sufficient data [44–47], one with duplicated data [43], and four out of HWE [39–42] were excluded. Eventually, 12 studies from 11 publications [28–38] were included in the meta-analysis. Fig 1 illustrates the process of search and selection. S2 File details the excluded articles and the reasons for their exclusion, as well as the original data obtained from the author via e-mail. No genome-wide association studies (GWAS) was found on this topic.
Characteristics of included studies
Table 2 summarizes the primary characteristics of the 12 studies finally included in the analysis. As for ethnicity, there are four studies from three publications [34, 36, 38] conducted in Asians, whereas eight studies [28–33, 35, 37] in Caucasians. All articles are in English, except one in Spanish with English abstract [32].
Meta-analysis results
Table 3 presents the primary results of all five genetic models in this comprehensive meta-analysis. No significant association was detected in the overall meta-analysis. As there were one studies [31] with zero AA phenotype in both cases and controls, the overall meta-analysis was performed with only 11 studies in the homozygous model (AA vs. GG) and the recessive model (AA vs. GA + GG). The subgroup meta-analyses showed significant associations in Asian subjects in the dominant model, the allele model, and the heterozygote model between TNF-α -308G/A and IRM risk (Fig 2). Similar results were found in the hospital-based-control group and the total-sample-size <150 group. No significant associations were found in any of the subgroups classified by the definition of IRM, the score or the Galbraith plot.
Forest plots for the association between TNF-α -308G/A Polymorphism and IRM risk classified by ethnicity in dominant model (A) and allelic model (B).
Publication bias
The results of Egger’s test confirmed that no significant publication bias existed in our meta-analysis (S3 File). Furthermore, Funnel plot (Fig 3), Egger’s publication-bias plot (S3 File) of the 12 studies demonstrated no sign of significant publication bias.
Heterogeneity analysis
I2 and PQ values showed significant heterogeneity among the 12 studies in 3 genetic models (AG vs. GG; AA + GA vs. GG; A vs. G), whereas 2 genetic models (AA vs. GG; AA vs. AG + GG) with 11 studies available for analysis demonstrated little heterogeneity (Table 3). Galbraith plot (S1 Fig) of the included 12 studies confirmed the existence of significant heterogeneity and illustrated 4 studies [33, 35–37] were the outliners. All the subgroup analyses showed a decline in heterogeneity in at least one subgroup, except the subgroups classified by the definition of IRM (Table 3).
Sensitivity analysis
In the sensitivity analysis, the pooled standardized mean difference (SMD) and 95% CIs were not significantly affected, no matter which study was taken out. (S2 Fig).
Cumulative meta-analysis
In the cumulative meta-analysis, no significant association between TNF-α -308G/A and IRM ever appeared over time (S3 Fig).
Trial sequential analysis
Repeated tests for significance upon new trials by meta-analyses may incur type I error [52]. To evaluate and minimize it, TSA was employed using software version 0.9 beta [53]. TSA combines traditional meta-analysis with information size calculation, and methods to adjust the significance according to the quantified strength of evidence and the impact of multiplicity for the repeated tests on accumulating trial data. In the present analysis, TSA was performed in dominant model with a two-tailed alpha of 0.05, beta of 0.20, and a relative risk increase of 10%. And a constant value correction of 0.5 in the no event trials was applied. The result of TSA demonstrates that neither the traditional significance boundaries nor the α-spending boundaries is crossed by the cumulative z-curve (dominant model) (Fig 4).
A two-sided graph is plotted by TSA where the blue etched lines represent conventional significance boundaries, the blue line indicates the cumulative Z-score, and the red lines shows the α -spending boundary and the required information size.
Discussion
The present meta-analysis enrolled 12 studies from 11 publications with 1,807 cases and 2,012 controls to assess the correlation between TNF-α -308G/A polymorphism and IRM risk. In addition, this meta-analysis presented the first cumulative meta-analysis and the first trial sequential analysis on this topic. In the cumulative meta-analysis, we found no trend of association, and more stable CIs with the accumulation of studies based on publication dates (S3 Fig). The trial sequential analysis shows a parallel cumulative z-curve to both the conventional boundaries and the α-spending boundaries (Fig 4), which indicates and confirms the inexistence of association between TNF-α -308G/A and IRM (S4 Fig). Sensitivity analysis also demonstrated that the overall results were reliable and robust, with no single influential study. All in all, the overall analyses found no significant associations in all five genetic models, which is quite opposite to the results of two recent meta-analyses in 2012 and 2016 [49, 50], whereas consistent with the meta-analysis in 2009 [48] and two reviews about the effect of polymorphisms of TNF-α[54, 55].
Since the main difference between the present meta-analysis and the previous meta-analyses [48–50], primarily lies in the exclusion and inclusion of studies without sufficient data to calculate HWE [44–47], and studies deviating from HWE [39–41] (Table 4). After looking into the previous meta-analyses with great care, we noticed that some data could not be found in the original manuscripts [31, 44–46] were listed out and analyzed in the meta-analyses in 2009 [48] and 2012 [49], and some studies without sufficient data to calculate HWE [28, 31, 45, 46] were included and analyzed in the meta-analysis in 2016 [50] (Table 4). The possibility exists that the suspicious data in the meta-analysis in 2012 may be obtained from authors directly. Hence, an additional meta-analysis was conducted with these suspected data, and the primary results remain unchanged (S4 File). Therefore, we may conclude that the associations found in previous meta-analyses were skewed due to the studies inconsistent with HWE. After all, departure from HWE can indicate systematic errors in genotyping, and data generated under this condition were unreliable and may significantly affects the conclusions of meta-analysis, which is the reason why HWE was ranked as an essential and routine item of the scrutinizing procedure in population-based genetic association meta-analyses (S2 Table, Item 9) [56].
What is more, we conducted several stratified analyses to trace the possible sources of the heterogeneity and found several points noteworthy. First, as for the definition of IRM, no significant difference was shown between the ≥2 miscarriages group and the ≥3 miscarriages group (Table 3), indicating that the different definition of IRM is not the possible cause of heterogeneity. Similar result was also found by Lee et al [38], who performed stratified analysis according to the number of consecutive spontaneous abortions in the study. In practice, most clinicians generally work to the less rigorous ≥2 miscarriages definition, probably because patients will be extremely worried under this condition and it is doctor’s responsibility to address their problems. Second, in addition to the significant association, a lower heterogeneity was also seen in the Asian populations compared to the Caucasian populations. On one hand, it is possible that the association between TNF-α -308G/A polymorphism and IRM risk is of ethnic specificity. However, this result should be interpreted with care, since the association found in the Asian population was supported by four studies from only three publications, which is far from sufficient. On the other hand, the low heterogeneity may be due to the limited studies themselves. Third, eight insiders and four outliers discovered in Galbraith plot were classified and analyzed, and a remarkable decrease of heterogeneity were demonstrated among the 8 insiders in the dominant model (I2 33.9) and the allelic model (I2 29.1), indicating the four outliners may be the cause of the heterogeneity (Table 3). Fourth, there are signs of possible sample-selection bias. As mentioned above, one study [34] with zero AA genotype frequency in both cases and controls, is not included in the overall meta-analyses of both the homozygous model (AA vs. GG) and the recessive model (AA vs. GA + GG). And a significant decrease in heterogeneity is demonstrated in both model (with I2 4.7 and 0, respectively). One possible explanation is that either the cases or the controls in this study are not representative. Another indication is that the heterogeneity of the total-sample-size >150 group was significantly higher than that of its counterpart group.
More importantly, during this meta-analysis, we summarized several meaningful points that may be helpful to further studies. First, the inclusion and exclusion criteria must be specific and comprehensive and should be strictly carried out. Otherwise, confounding factors will inevitably be mixed in. Second, more studies on Asian populations are needed. Third, more attention should be paid to the selection of the control group, which is always neglected and carried out without strictly following criteria. Four studies [39–42] deviating from HWE and one study with 0 AA genotype frequency in both cases and controls [31], are good representations of this issue.
There are some limitations in the present meta-analysis, which should be noted. First, misclassification bias and selection bias may be incurred due to unified diagnostic criteria of IRM and various sources of controls. Second, we failed to evaluate the gene-gene and gene-environment associations due to lack of the original data. Third, heterogeneity in several subgroups remains high in the subgroup analyses. Last but not least, the limited number of studies in Asian population may restrict the statistical power of the association.
In summary, no association between the TNF-α -308G/A promoter polymorphism and IRM was found in the present meta-analysis. The correlation found in Asian population needs confirmation from more studies.
Conclusions
The present meta-analysis demonstrated no association between TNF-α -308G/A polymorphism and IRM risk, and the association found in the previous meta-analyses may result from the inclusion of studies inconsistent with HWE. Significant association demonstrated in Asian subjects in the subgroup analyses, should be interpreted with caution due to limited studies. Further rigorously-designed large-scale studies on Asian population are needed to confirm this conclusion.
Supporting Information
S4 Fig. Trial sequential analysis result summary (dominant model).
https://doi.org/10.1371/journal.pone.0166892.s004
(TIF)
S2 File. Data obtained via email & Excluded studies categorized by reasons.
https://doi.org/10.1371/journal.pone.0166892.s006
(PDF)
S3 File. Results of Publication Bias Test & Figures.
https://doi.org/10.1371/journal.pone.0166892.s007
(PDF)
S4 File. Additional analyses with dubious data.
https://doi.org/10.1371/journal.pone.0166892.s008
(PDF)
S5 File. Additional analyses with the Chinese article.
https://doi.org/10.1371/journal.pone.0166892.s009
(PDF)
S2 Table. Meta-analysis on genetic association studies form.
https://doi.org/10.1371/journal.pone.0166892.s011
(DOCX)
S3 Table. Inclusion criteria of all the studies included.
https://doi.org/10.1371/journal.pone.0166892.s012
(DOCX)
Author Contributions
- Conceptualization: SMD ZP JSD.
- Data curation: HHL CLW JJL.
- Formal analysis: JSD JWL.
- Investigation: GCZ JWL.
- Methodology: GCZ JJL.
- Resources: SZL XLL.
- Software: ZP XLL.
- Supervision: JSD SMD.
- Validation: ZP JJL.
- Writing – original draft: ZP JSD JWL.
- Writing – review & editing: SMD YJH ZP.
References
- 1. Wilcox AJ, Weinberg CR, O'Connor JF, Baird DD, Schlatterer JP, Canfield RE, et al. Incidence of early loss of pregnancy. New England Journal of Medicine. 1988;319(4):189–94. pmid:3393170
- 2. Miller J, Williamson E, Glue J, Gordon Y, Grudzinskas J, Sykes A. (1980) Fetal loss after implantation: a prospective study. The Lancet 316: 554–556.
- 3. Andersen A- MN, Wohlfahrt J, Christens P, Olsen J, Melbye M (2000) Maternal age and fetal loss: population based register linkage study. Bmj 320: 1708–1712. pmid:10864550
- 4. de La Rochebrochard E, Thonneau P (2002) Paternal age and maternal age are risk factors for miscarriage; results of a multicentre European study. Human Reproduction 17: 1649–1656. pmid:12042293
- 5. Friedman T, Gath D (1989) The psychiatric consequences of spontaneous abortion. The British Journal of Psychiatry 155: 810–813. pmid:2620207
- 6. Carrington B, Sacks G, Regan L (2005) Recurrent miscarriage: pathophysiology and outcome. Current Opinion in Obstetrics and Gynecology 17: 591–597. pmid:16258340
- 7. Stirrat GM (1990) Recurrent miscarriage I: definition and epidemiology. The Lancet 336: 673–675.
- 8. Rai R, Regan L (2006) Recurrent miscarriage. The Lancet 368: 601–611.
- 9. Brigham S, Conlon C, Farquharson R (1999) A longitudinal study of pregnancy outcome following idiopathic recurrent miscarriage. Human Reproduction 14: 2868–2871. pmid:10548638
- 10. Wilcox AJ, Weinberg CR, Baird DD. Risk factors for early pregnancy loss. Epidemiology. 1990;1(5):382–5. pmid:2078614
- 11. Raghupathy R, Kalinka J. Cytokine imbalance in pregnancy complications and its modulation. Front Biosci. 2008;13(1):985–94.
- 12. Makhseed M, Raghupathy R, Azizieh F, Omu A, Al-Shamali E, Ashkanani L. (2001) Th1 and Th2 cytokine profiles in recurrent aborters with successful pregnancy and with subsequent abortions. Human Reproduction 16: 2219–2226. pmid:11574519
- 13.
Raghupathy R. Pregnancy: success and failure within the Th1/Th2/Th3 paradigm; 2001. Elsevier. pp. 219–227.
- 14. Saito S (2000) Cytokine network at the feto-maternal interface. Journal of reproductive immunology 47: 87–103. pmid:10924744
- 15.
Kwak-Kim J, Gilman-Sachs A, Kim C (2005) T helper 1 and 2 immune responses in relationship to pregnancy, nonpregnancy, recurrent spontaneous abortions and infertility of repeated implantation failures. Immunology of Gametes and Embryo Implantation: Karger Publishers. pp. 64–79.
- 16. Wilczyński JR (2005) Th1/Th2 cytokines balance—yin and yang of reproductive immunology. European Journal of Obstetrics & Gynecology and Reproductive Biology 122: 136–143.
- 17. Locksley RM, Killeen N, Lenardo MJ (2001) The TNF and TNF receptor superfamilies: integrating mammalian biology. Cell 104: 487–501. pmid:11239407
- 18. Mallmann P, Mallmann R, Krebs D. Determination of tumor necrosis factor alpha (TNFα) and Interleukin 2 (IL2) in women with idiopathic recurrent miscarriage. Archives of gynecology and obstetrics. 1991;249(2):73–8. pmid:1953054
- 19. Berman J, Girardi G, Salmon JE. TNF-α is a critical effector and a target for therapy in antiphospholipid antibody-induced pregnancy loss. The Journal of Immunology. 2005;174(1):485–90. pmid:15611274
- 20. Vitoratos N, Papadias C, Economou E, Makrakis E, Panoulis C, Creatsas G. Elevated circulating IL-1β and TNF-alpha, and unaltered IL-6 in first-trimester pregnancies complicated by threatened abortion with an adverse outcome. Mediators of inflammation. 2006;2006.
- 21. Saini V, Arora S, Yadav A, Bhattacharjee J. Cytokines in recurrent pregnancy loss. Clinica chimica acta. 2011;412(9):702–8.
- 22. Yui J, Garcia-Lloret M, Wegmann Tea, Guilbert L. Cytotoxicity of tumour necrosis factor-alpha and gamma-interferon against primary human placental trophoblasts. Placenta. 1994;15(8):819–35. pmid:7886023
- 23. Haider S, Knöfler M. Human tumour necrosis factor: physiological and pathological roles in placenta and endometrium. Placenta. 2009;30(2):111–23. pmid:19027157
- 24. Wilson AG, Symons JA, McDowell TL, McDevitt HO, Duff GW. Effects of a polymorphism in the human tumor necrosis factor α promoter on transcriptional activation. Proceedings of the National Academy of Sciences. 1997;94(7):3195–9.
- 25. Hajeer AH, Hutchinson IV. TNF-α gene polymorphism: clinical and biological implications. Microscopy research and technique. 2000;50(3):216–28. pmid:10891887
- 26. Choi YK, Kwak-Kim J. Cytokine gene polymorphisms in recurrent spontaneous abortions: a comprehensive review. Am J Reprod Immunol. 2008;60(2):91–110. Epub 2008/06/25. pmid:18573127.
- 27. Walia GK, Mukhopadhyay R, Saraswathy K, Puri M, Chahal S. Immuno-molecular etiology of recurrent pregnancy loss and the anthropological perspective. International Journal of Human Genetics. 2008;8(1/2):227.
- 28. Babbage SJ, Arkwright PD, Vince GS, Perrey C, Pravica V, Quenby S, et al. Cytokine promoter gene polymorphisms and idiopathic recurrent pregnancy loss. J Reprod Immunol. 2001;51(1):21–7. Epub 2001/07/05. pmid:11438378.
- 29. Reid JG, Simpson NA, Walker RG, Economidou O, Shillito J, Gooi HC, et al. The carriage of pro-inflammatory cytokine gene polymorphisms in recurrent pregnancy loss. Am J Reprod Immunol. 2001;45(1):35–40. Epub 2001/02/24. pmid:11211945.
- 30. Pietrowski D, Bettendorf H, Keck C, Burkle B, Unfried G, Riener EK, et al. Lack of association of TNFalpha gene polymorphisms and recurrent pregnancy loss in Caucasian women. J Reprod Immunol. 2004;61(1):51–8. Epub 2004/03/19. pmid:15027477.
- 31. Kamali-Sarvestani E, Zolghadri J, Gharesi-Fard B, Sarvari J. Cytokine gene polymorphisms and susceptibility to recurrent pregnancy loss in Iranian women. J Reprod Immunol. 2005;65(2):171–8. Epub 2005/04/07. pmid:15811521.
- 32. Quintero-Ramos A, Valdez-Velazquez LL, Hernandez G, Baltazar LM, Padilla-Gutierrez JR, Valle Y, et al. [Assessment of five thrombophilic genetic polymorphisms among couples with habitual abortion]. Gac Med Mex. 2006;142(2):95–8. Epub 2006/05/23. pmid:16711541.
- 33. Zammiti W, Mtiraoui N, Finan RR, Almawi WY, Mahjoub T. Tumor necrosis factor alpha and lymphotoxin alpha haplotypes in idiopathic recurrent pregnancy loss. Fertil Steril. 2009;91(5):1903–8. Epub 2008/04/09. pmid:18394614.
- 34. Liu C, Wang J, Zhou S, Wang B, Ma X. Association between -238 but not -308 polymorphism of Tumor necrosis factor alpha (TNF-alpha)v and unexplained recurrent spontaneous abortion (URSA) in Chinese population. Reprod Biol Endocrinol. 2010;8:114. Epub 2010/10/06. pmid:20920206; PubMed Central PMCID: PMCPmc2956720.
- 35. Palmirotta R, La Farina F, Ferroni P, Ludovici G, Nigro C, Savonarola A, et al. TNFA gene promoter polymorphisms and susceptibility to recurrent pregnancy loss in Italian women. Reprod Sci. 2010;17(7):659–66. Epub 2010/04/15. pmid:20388617.
- 36. Gupta R, Prakash S, Parveen F, Agrawal S. Association of CTLA-4 and TNF-alpha polymorphism with recurrent miscarriage among North Indian women. Cytokine. 2012;60(2):456–62. Epub 2012/06/26. pmid:22727980.
- 37. Alkhuriji AF, Alhimaidi AR, Babay ZA, Wary AS. The relationship between cytokine gene polymorphism and unexplained recurrent spontaneous abortion in Saudi females. Saudi Med J. 2013;34(5):484–9. Epub 2013/05/17. pmid:23677264.
- 38. Lee BE, Jeon YJ, Shin JE, Kim JH, Choi DH, Jung YW, et al. Tumor necrosis factor-alpha gene polymorphisms in Korean patients with recurrent spontaneous abortion. Reprod Sci. 2013;20(4):408–17. Epub 2012/12/04. pmid:23202728; PubMed Central PMCID: PMCPmc4077515.
- 39. Finan RR, Al-Irhayim Z, Mustafa FE, Al-Zaman I, Mohammed FA, Al-Khateeb GM, et al. Tumor necrosis factor-alpha polymorphisms in women with idiopathic recurrent miscarriage. J Reprod Immunol. 2010;84(2):186–92. Epub 2010/01/29. pmid:20106534.
- 40. Kaur A, Kaur A. Recurrent pregnancy loss: TNF-alpha and IL-10 polymorphisms. J Hum Reprod Sci. 2011;4(2):91–4. Epub 2011/11/09. pmid:22064760; PubMed Central PMCID: PMCPmc3205540.
- 41. Liu RX, Wang Y, Wen LH. Relationship between cytokine gene polymorphisms and recurrent spontaneous abortion. Int J Clin Exp Med. 2015;8(6):9786–92. Epub 2015/08/27. pmid:26309657; PubMed Central PMCID: PMCPmc4538128.
- 42. Sudhir N, Beri A, Kaur A. Association of tumor necrosis factor-alpha 308G/A polymorphism with recurrent miscarriages in women. Journal of Human Reproductive Sciences. 2016;9(2):86. pmid:27382232
- 43. Zammiti W, Mtiraoui N, Khairi H, Gris JC, Almawi WY, Mahjoub T. Associations between tumor necrosis factor-alpha and lymphotoxin-alpha polymorphisms and idiopathic recurrent miscarriage. Reproduction. 2008;135(3):397–403. Epub 2008/02/27. pmid:18299433.
- 44. Baxter N, Sumiya M, Cheng S, Erlich H, Regan L, Simons A, et al. Recurrent miscarriage and variant alleles of mannose binding lectin, tumour necrosis factor and lymphotoxin alpha genes. Clin Exp Immunol. 2001;126(3):529–34. Epub 2001/12/12. pmid:11737072; PubMed Central PMCID: PMCPmc1906238.
- 45. Daher S, Shulzhenko N, Morgun A, Mattar R, Rampim GF, Camano L, et al. Associations between cytokine gene polymorphisms and recurrent pregnancy loss. J Reprod Immunol. 2003;58(1):69–77. Epub 2003/03/01. pmid:12609526.
- 46. Prigoshin N, Tambutti M, Larriba J, Gogorza S, Testa R. Cytokine gene polymorphisms in recurrent pregnancy loss of unknown cause. Am J Reprod Immunol. 2004;52(1):36–41. Epub 2004/06/25. pmid:15214940.
- 47. Bompeixe EP, Carvalho Santos PS, Vargas RG, von Linsingen R, Zeck SC, Wowk PF, et al. HLA class II polymorphisms and recurrent spontaneous abortion in a Southern Brazilian cohort. Int J Immunogenet. 2013;40(3):186–91. Epub 2012/09/04. pmid:22938381.
- 48. Medica I, Ostojic S, Pereza N, Kastrin A, Peterlin B. Association between genetic polymorphisms in cytokine genes and recurrent miscarriage—a meta-analysis. Reprod Biomed Online. 2009;19(3):406–14. Epub 2009/09/26. pmid:19778488.
- 49. Zhang B, Liu T, Wang Z. Association of tumor necrosis factor-α gene promoter polymorphisms (-308G/A,-238G/A) with recurrent spontaneous abortion: a meta-analysis. Human immunology. 2012;73(5):574–9. pmid:22369788
- 50. Li H-H, Xu X-H, Tong J, Zhang K-Y, Zhang C, Chen Z-J. Association of TNF-α genetic polymorphisms with recurrent pregnancy loss risk: a systematic review and meta-analysis. Reproductive Biology and Endocrinology. 2016;14(1):1.
- 51. Peng Z, Lv X, Sun Y, Dai S. Association of Interleukin-10-1082A/G Polymorphism with Idiopathic Recurrent Miscarriage: A Systematic Review and Meta-Analysis. Am J Reprod Immunol. 2016;75(2):162–71. Epub 2015/12/20. pmid:26682645.
- 52. Borm GF, Donders AR. Updating meta-analyses leads to larger type I errors than publication bias. J Clin Epidemiol. 2009;62(8):825–30.e10. Epub 2009/01/13. pmid:19136233.
- 53. Kulinskaya E, Wood J. Trial sequential methods for meta-analysis. Res Synth Methods. 2014;5(3):212–20. Epub 2015/06/09. pmid:26052847.
- 54. Mekinian A, Tamouza R, Pavy S, Gestermann N, Ittah M, Mariette X, et al. Functional study of TNF-α promoter polymorphisms: literature review and meta-analysis. European cytokine network. 2011;22(2):88–102. pmid:21768061
- 55. Bayley J, Ottenhoff T, Verweij C. Is there a future for TNF promoter polymorphisms? Genes and immunity. 2004;5(5):315–29. pmid:14973548
- 56. Namipashaki A, Razaghi-Moghadam Z, Ansari-Pour N. The Essentiality of Reporting Hardy-Weinberg Equilibrium Calculations in Population-Based Genetic Association Studies. Cell J. 2015;17(2):187–92. Epub 2015/07/23. pmid:26199897; PubMed Central PMCID: PMCPmc4503832.