Angiotensinogen Gene M235T and T174M Polymorphisms and Susceptibility of Pre-Eclampsia: A Meta-Analysis

Authors


Dr. Rong Lin, Hainan Medical College, 3 Xueyuan Road, Haikou, Hainan Province 571101, China. Tel: 86-898-66893739; Fax: 86-898-31350713; E-mail: xianronglin@gmail.com

Summary

There are controversies in reports on the association of the angiotensinogen (AGT) gene polymorphisms with the risk of developing pre-eclampsia (PE). We performed a meta-analysis to examine the association between the AGT polymorphisms and PE risk: M235T (31 studies involving 2555 patients and 6114 controls) and T174M (six studies involving 681 patients and 2076 controls). For the M235T polymorphism, the TT genotype increased the PE risk as compared to the MM genotype (odds ratio 1.61, 95% confidence intervals 1.22–2.14, P= 0.001). When stratified by ethnicity, the TT genotype remained significantly associated with higher PE risk in Caucasians and Mongolians but not in Africans. Similar results were also obtained under all three genetic models of the M235T polymorphism. For the T174M polymorphism, no significant association was found in the comparisons (MT vs. TT and MM vs. TT) and under any genetic models. The analysis excluding the highly significant Hardy–Weinberg equilibrium-violating studies and sensitivity analysis further strengthened the validity of these associations. No publication bias was observed in this study. This meta-analysis demonstrates that the AGT M235T polymorphism is significantly associated with PE whereas the T174M polymorphism is not.

Introduction

Pre-eclampsia (PE), identified clinically by maternal hypertension and proteinuria occurring after the 20th week of gestation, is a leading cause of maternal and perinatal mortality and morbidity. It is believed to be a multifactorial disorder with a strong genetic component (Roberts & Cooper, 2001; Sibai et al., 2005; Kobashi, 2006). Angiotensinogen (AGT) has been suggested to be implicated in PE development and the AGT gene may be a susceptibility gene of PE (Takimoto et al., 1996).

AGT encodes angiotensinogen, the precursor of angiotensin II, which is a potent vasoconstrictor, a major determinant of salt and water homeostasis and a growth factor. AGT is mainly expressed in the liver, but it is also expressed in many other tissues including the placenta. Until now, many studies have investigated the association between the AGT gene polymorphisms and susceptibility of PE. And most of them focused on two common nonsynonymous polymorphisms (M235T and T174M), both in AGT exon 2. The M235T variant (rs699, also described as M268T or T704C) encodes threonine instead of methionine at residue 235. The T174M variant (rs4762, also described as T207M or C521T) encodes methionine instead of threonine at residue 174.

In 1993, Ward et al. first found that the 235T allele was significantly associated with an increased PE risk in Caucasians and Japanese. But later attempts to replicate this finding have yielded inconsistent results. In UK families affected by PE, the M235T polymorphism showed no association with PE (GOPEC Consortium, 2005). Several population case-control studies also indicated that there was no significant association of the M235T variant with PE (Guo et al., 1997; Morgan et al., 1999; Curnow et al., 2000). A few studies even suggested that the 235T allele was associated with a reduced PE risk (Aggarwal et al., 2010). The studies of the T174M polymorphism with PE risk also provided contradictory results. Lévesque et al. (2004) reported that the 174M allele was associated with PE, but this finding was unrepeated by Choi et al. (2004). The UK family study also showed no association of the T174M polymorphism with PE (GOPEC Consortium, 2005).

Meta-analysis offers a powerful means of overcoming problems such as small sample sizes and the inadequate statistical powers of genetic studies on complex disorders. Therefore, in this study, we conducted a meta-analysis of available studies to evaluate the association of AGT M235T and T174M polymorphisms with PE risk.

Methods

Search Strategy

We searched the PubMed, Human Genome Epidemiology Network (HuGENet), Embase, ISI Web of Science, China National Knowledge Infrastructure (CNKI), Wanfang, Weipu and Chinese Biomedical (CBM) databases (last search February 6, 2012) without language restriction to identify studies investigating the association between the AGT gene polymorphisms and susceptibility of PE. Search terms included combinations of terms such as “angiotensinogen,”“AGT,”“gestational hypertension,”“hypertension in pregnancy,”“pre-eclampsia,”“preeclampsia,”“eclampsia,”“pregnancy-induced hypertension,”“blood pressure in pregnancy,”“transient hypertension of pregnancy” and their synonyms. The exact search is available on request from the authors. We perused the reference lists of all retrieved articles and relevant reviews.

In this study, we only performed a meta-analysis to examine the association between the AGT M235T and T174M polymorphisms and PE risk. We did not consider other AGT polymorphisms because the available evidence on them is limited; the AGT promoter polymorphism G-6A (rs5051), for example, is thought to modulate ATG gene expression.

Study Selection Criteria

PE, the second category of hypertension during pregnancy, is the endpoint of this study. PE was defined as a new-onset hypertension (at least 140/90 mmHg in a pregnant woman who was normotensive before 20 weeks’ gestation) accompanied by proteinuria [≥300 mg protein in a 24 h urine collection, which is usually correlated with ≥30 mg/dl (≥1+ reading on dipstick) in a random urine determination with no evidence of urinary tract infection]. Eligible studies were those that used case-control, nested case-control or cohort designs and validated genotyping methods to investigate the frequency of the two polymorphisms in unrelated PE patients and unaffected individuals. We did not consider editorials, narrative reviews and letters to the editor or other articles not reporting primary research results. Family-based studies were excluded because their design and analysis is different from that of population-association studies.

Data Extraction

The following information was extracted from each study: first author's name, year of publication, country, population evaluated, mean age or age range for cases and controls, definition and number of cases and controls, allele and genotype distribution in case patients and controls (where data were not given, they were calculated from the raw data of the case and control groups), consistency of genotype frequencies with Hardy–Weinberg equilibrium (HWE) (calculated).

Literature search, data extraction and study quality assessment were performed independently by two authors (R. L. and Y-P. L.) with disagreements resolved by consensus. We also compared the extracted data with those in two previous meta-analyses of the M235T polymorphism (Medica et al., 2007; Zafarmand et al., 2008). If inconsistent, we referred to the original literature.

Statistical Analyses

Crude odds ratios (ORs) with 95% confidence intervals (CIs) were used to access the strength of association between the AGT polymorphisms and PE. The pooled ORs were performed for two comparisons of different genotypes (heterozygotes vs. major allele homozygotes, and minor allele homozygotes vs. major allele homozygotes), and three genetic models (dominant, additive and recessive). Stratified analyses were also performed by ethnicities.

A χ2 -based Q-test was performed to check the between-study heterogeneity, which is considered to be significant for P < 0.10, as well as I2 for estimation of inconsistency in meta-analyses. I2 takes a value of 0–100% (I2= 0–25%, no heterogeneity; I2= 25–50%, moderate heterogeneity; I2= 50–75%, large heterogeneity; I2= 75–100%, extreme heterogeneity). If the result of the heterogeneity test was P > 0.1, ORs were pooled according to the fixed-effects model (the Mantel–Haenszel model). Otherwise the random-effects model (the DerSimonian–Laird model) was used. Meta-regression analysis was conducted to explore heterogeneity. This meta-regression analysis examined the effect of certain variables, including HWE in controls (cases or all subjects), genotyping method, the number of cases (controls or all subjects), mean age of cases (or controls) and ethnicity.

The potential publication bias was estimated using Begg's test and Egger's linear regression test by visual inspection of the funnel plot, and P < 0.05 was considered representative of statistically significant publication bias. To access the stability of the meta-analysis’ results and case definition impact on the pooled estimates, one-way sensitivity analyses were performed. All statistical tests were preformed with the software STATA version 11.0 (Stata Corporation, College station, TX, USA). Genetic differences between populations for the M235T and T174M polymorphisms were estimated using unbiased FST following Weir & Hill (2002).

Results

Eligible Studies

A total of 54 papers were identified as potentially relevant and were screened for inclusion. Of these papers, 26 were subsequently excluded: nine reported on overlapping populations, data of two could not be extracted, two were meta-analyses, one was a family-based study, 11 considered other endpoints and one did not exclude PE patients superimposed on pre-existing hypertension in the PE cases. In four papers for the M235T polymorphism, the results were based on two different populations (Ward et al., 1993; Guo et al., 1997; Curnow et al., 2000; Jenkins et al., 2008); we considered these populations separately in the meta-analysis. Finally, 31 (in 27 articles involving 2555 patients and 6114 controls) and six studies (in six articles involving 681 patients and 2076 controls), respectively, were considered for the meta-analysis on the AGT M235T and T174M polymorphisms (Table 1). One study of the M235T polymorphism (Curnow et al., 2000) was included in the meta-analysis only under the additive model because only allele counts were available from it. Wang et al. (2006) miswrote the 235M allele for the 235T allele in Table 2 of their report and we corrected it in this study.

Table 1.  Characteristics of case-control studies included in a meta-analysis of the association between the AGT gene polymorphisms and pre-eclampsia.
First authorYearCountryEthnicityMean age ± SD (year)Gestation at delivery ± SD (weeks)No. of eligible subjectsPolymorphism(s) investigatedGenotyping method
CasesControlsCasesControlsCasesControls
  1. AGT, angiotensinogen; HWE, Hardy–Weinberg equilibrium; RFLP, restriction fragment length polymorphisms; ASOH, allele-specific oligonucleotide hybridization; ASA, allele-specific amplification; MS-PCR, mutagenically separated polymerase chain reaction.

  2. aMedian (interquartile range).

  3. bMedian (range).

  4. cMean (range).

Ward1993USACaucasian    45571M235TASOH
Guo1997AustraliaCaucasian    10681M235TASA, RFLP(SfaNI)
Morgan1999UKCaucasian28.1 ± 5.927.6 ± 4.3  4384M235TASOH
Curnow2000USACaucasian    17195M235TRFLP(Tth111 I)
Procopciuc2002RomaniaCaucasian25.31 ± 4.5728.83 ± 4.81  136M235TRFLP(Tth111 I)
Bouba2003GreeceCaucasian31 (21–45)a29 (17–48)a34.32 ± 3.638.96 ±1.241102M235TRFLP(Tth111 I)
Zhang2003USACaucasian25.2 ± 6.325.6 ± 6.033.5 ± 3.839.1 ± 2.060400M235TMS-PCR
Lévesque2004CanadaCaucasian26 ± 526.5 ± 5.036.7 ± 4.038.5 ± 3.3180667M235T, T174MASA
Tempfer2004AustraliaCaucasian29 (23–42)b29 (21–43)b38 (29–40)b40 (36–41)b2424M235TASOH
Benedetto2007ItalyCaucasian30 ± 430 ± 434 ± 439 ± 2120103M235TRFLP (HindII)
Jenkins2008USACaucasian28.1 ± 5.824.0 ± 5.235 ± 3.939.5 ±1.9193288M235TRFLP(Tth111 I)
Knyrim2008GermanyCaucasian30 (15–40)b 35.5 (25–41)b 67100M235T, T174MSSCP
Ward1993USAJapanese    1880M235TASOH
Guo1997AustraliaChinese    7248M235TASA, RFLP(SfaNI)
Romi1997JapanIndonesian    20100M235TRFLP(Tth111 I)
Kobashi1999JapanJapanese 29.3 ± 5.9 39.1 ± 2.090381M235TASA, ASOH, RFLP(Tth111 I)
Suzuki1999JapanJapanese30.44 ± 4.9030.36 ± 3.88  9280M235TASOH
Li1999ChinaChinese27.6(22–37)c 35 (28–40)c 5890M235TRFLP(Tth111 I)
Choi2004KoreaKorean30.2 ± 4.530.9 ± 4.736.2 ± 3.539.7 ± 1.09098M235T, T174MM235T: RFLP(Tth111 I), T174M: RFLP(NCo I)
Kim2004South KoreaKorean30.6 ± 5.730.8 ± 3.336.2 ± 4.439.0 ± 2.1104114M235TRFLP(Tth111 I)
Huang2007ChinaChinese29±429±328-3628-3658102M235TRFLP(Tth111 I)
Song2007ChinaChinese30.5±3.728.7±3.235.32±3.638.96±1.24545M235TRFLP(Tth111 I)
Jiang2008ChinaChinese    5570T174MRFLP(Nco I)
Curnow2000USAAfrican    10130M235TRFLP(Tth111 I)
Roberts2004South AfricaAfrican 25 38.8391338M235TASA
Hillermann2005South AfricaAfrican21(14-31)b29(18-43)b30(20-36)b39(37-44)b5050M235TRFLP(AspI)
Wang2006USAAfrican29.0±7.228.0±6.535.4±4.038.0±3.51251040M235T, T174MTaqman
Jenkins2008USAAfrican21.3±6.120.8±3.936 ±3.439.2 ±2.730271M235TRFLP(Tth111 I)
Bashford2001USAMexican or Central American Hispanic25(21-31)a53(23-30)a37.4(35-39)a39.9(39-41)a8753M235TASOH
Dissanayake2009UKSinhalese27.0±5.327.1±5.135(32-38)a39(38-40)a175171M235T, T174MASOH
Aggarwal2010IndiaNorth Indian25.7±3.826.3±4.033.2±2.735.9±2.7120118M235TRFLP(Tth111 I)
Aggarwal2011IndiaNorth Indian25(18-40)b25.8(18-45)b34±538±3200200M235TRFLP(Tth111 I)
Total       27166400  
Table 2.  Meta-analysis of association between AGT M235T polymorphism and pre-eclampsia risk.
 All studiesStudies excluding the study of Zhang et al.I2
Pooled OR (95% CI) P OR I 2 P Q Pooled OR (95% CI) P OR I 2 P Q
  1. P OR and PQ refer to the significance levels of the odds ratio and Q-test of heterogeneity, respectively.

  2. P values, which materially alter when compared with the analysis in all studies, are shown in bold.

  3. All pooled ORs were derived from random-effects models except for cells marked with F (fixed-effects model)

Caucasian         
 MT versus MM1.50 (1.17–1.93)0.00223.5%0.2201.37 (1.11–1.69)F0.0030.0%0.56223.5%
 TT versus MM1.93 (1.18–3.14)0.00970.1%<0.0011.49 (1.16–1.92)F0.00246.6%0.05123.5%
 Dominant1.63 (1.20–2.22)0.00252.9%0.0201.41 (1.16–1.71)F0.00120.0% 0.25932.9%
 Additive1.36 (1.06–1.73)0.01571.5%<0.0011.24 (1.02–1.51)0.03049.8%0.03021.7%
 Recessive1.47 (1.02–2.13)0.04064.2%0.0021.29 (0.96–1.73) 0.09137.4% 0.11026.8%
Mongolian         
 MT versus MM1.30 (0.87–1.94)0.2080.0%0.9091.38 (0.93–2.06)F0.1090.0%0.9090.0%
 TT versus MM1.80 (1.18–2.76)0.0070.0%0.7611.92 (1.27–2.91)F0.0020.0%0.7610.0%
 Dominant1.54 (1.05–2.26)0.0270.0%0.9351.62 (1.11–2.36)F0.0120.0%0.9350.0%
 Additive1.44(1.10-1.90)0.00953.6%0.0221.44(1.10-1.90)0.00953.6%0.0220.0%
 Recessive1.53(1.00-2.32)0.04864.4%0.0031.53(1.00-2.32)0.04864.4%0.0030.0%
African         
 MT versus MM0.70(0.26-1.89)0.4820.0%0.7110.67(0.27-1.67)F0.3940.0%0.7110.0%
 TT versus MM0.92(0.35-2.40)0.8650.0%0.6480.91(0.38-2.17)F0.8290.0%0.6480.0%
 Dominant0.86(0.33-2.24)0.7570.0%0.6740.84(0.35-2.00)F0.6980.0%0.6740.0%
 Additive1.16(0.91-1.49)0.2350.0%0.8071.16(0.91-1.49)0.2350.0%0.8070.0%
 Recessive1.22(0.92-1.61)0.1610.0%0.7641.22(0.92-1.61)0.1610.0%0.7640.0%
Overall         
 MT versus MM1.20(0.97-1.50)0.09631.8%0.0531.14(0.97-1.34)F0.10319.2% 0.18312.6%
 TT versus MM1.61(1.22-2.14)0.00145.1%0.0051.43(1.18-1.72)F<0.00117.7% 0.20327.4%
 Dominant1.32(1.05-1.64)0.01539.9%0.0151.23(1.05-1.42)F0.00820.9% 0.16219.0%
 Additive1.26(1.08-1.46)0.00365.9%<0.0011.21(1.05-1.38)0.00856.4%<0.0019.5%
 Recessive1.41(1.10-1.79)0.00670.2%<0.0011.34(1.06-1.69)0.01466.0%<0.0014.2%

Ward et al. (1993), Romi et al. (1997), Procopciuc et al. (2002) and Kim et al. (2004) did not mention the case definitions they had used. Li et al. (1999) recruited pregnant women with severe PE [PE with a blood pressure of ≥160/110 mmHg, and/or more than 2000 mg/24 h (or 2+ dipstick) of proteinuria] as cases. In the study of Zhang et al. (2003), PE was defined as the development of new-onset hypertension of at least 140/90 with postpartum resolution and either more than 500 mg of protein in 24 h or a new 3+ dipstick without infection in a pregnant woman with a singleton pregnancy. Choi et al. (2004) defined PE based on hypertension (BP greater than systolic BP 140 mmHg or diastolic BP 90 mmHg) and more than 1000 mg/l (or 2+ dipstick) of proteinuria in random urine determinations after the 20th week of pregnancy. In the study of Roberts et al. (2004), PE was defined as a blood pressure of at least 140/90 mmHg on two separate occasions 6 h apart, and at least 1+ proteinuria on dipstick after 34 weeks of gestation, while early onset PE was the development of hypertension and proteinuria for the first time prior to the 34th week of gestation. In all other studies, the cases were well-defined following the criteria as described earlier. We performed the main analysis and a sensitivity analysis on the basis of case definitions. Controls were normotensive pregnant women.

In most of the studies, the genotype frequencies were in agreement with HWE expectation. Under the assumption of HWE, genotype frequencies in unrelated controls from a general population should be randomly distributed on the basis of the observed allelic frequencies. However, deviations from HWE were discovered in controls of four studies (Ward et al., 1993; Bashford et al., 2001; Tempfer et al., 2004; Aggarwal et al., 2011) for the M235T polymorphism (Supplementary Table S1) and two for the T174M polymorphism (Lévesque et al., 2004; Knyrim et al., 2008) (Table S2). But these studies should not be excluded because there are no other grounds for doubting the quality of these studies. Therefore, in this study, we preformed analyses including and excluding the two highly significant (P < 0.01) HWE-violating studies (Tempfer et al., 2004; Aggarwal et al., 2011).

M235T Polymorphism and Susceptibility of Pre-Eclampsia

There were 12 studies in Caucasians (Ward et al., 1993; Guo et al., 1997; Morgan et al., 1999; Curnow et al., 2000; Procopciuc et al., 2002; Bouba et al., 2003; Zhang et al., 2003; Lévesque et al., 2004; Tempfer et al., 2004; Benedetto et al., 2007; Jenkins et al., 2008; Knyrim et al., 2008), 10 in Mongolians (Ward et al., 1993; Guo et al., 1997; Romi et al., 1997; Kobashi et al., 1999; Li et al., 1999; Suzuki et al., 1999; Choi et al., 2004; Kim et al., 2004; Huang et al., 2007; Song et al., 2007), five in Africans (Curnow et al., 2000; Roberts et al., 2004; Hillermann et al., 2005; Wang et al., 2006; Jenkins et al., 2008) and four in other ethnic populations (Bashford et al., 2001; Dissanayake et al., 2009; Aggarwal et al., 2010; Aggarwal et al., 2011) which we put together in a miscellaneous group (Table S1). There were significant differences in the frequency of 235T allele between Caucasians (0.424), Mongolians (0.711) and Africans (0.869) (all pairwise P < 10−70). Great differences at the M235T polymorphism existed between Caucasians and Mongolians (FST= 0.151), and between Caucasians and Africans (FST= 0.340) whereas difference between Mongolians and Africans was small (FST= 0.074).

In the total population, compared with the MM homozygotes, the TT genotype increased the PE risk (OR 1.61, 95% CI 1.22–2.14, P= 0.001) whereas the MT genotype did not (P= 0.096) (Table 2 and Figs S1–S2). Moderate heterogeneity was found in both comparisons (MT vs. MM, I2= 31.8%, P= 0.053; TT vs. MM, I2= 45.1%, P= 0.005). In the stratified analysis by ethnicity, both the MT and TT genotypes increased the PE risk in Caucasians (MT vs. MM, OR 1.50, 95% CI 1.17–1.93, P= 0.002; TT vs. MM, OR 1.93, 95% CI 1.18–3.14, P= 0.009) but not in Africans. In Mongolians, the TT genotype increased the PE risk (OR 1.80, 95% CI 1.18–2.76, P= 0.007), whereas the MT genotype did not. The results of meta-regression showed that the ethnic background explained 81.92% of the between-study heterogeneity in the comparison between MT versus MM genotype (P= 0.001) but explained little in the comparison between TT versus MM genotype (P= 0.093). All findings remained similar after excluding the highly significant HWE-violating studies except that the MT genotype, compared with the MM homozygotes, increased the PE risk in the total population (OR 1.31, 95% CI 1.11–1.54, P= 0.001) with little heterogeneity (I2= 0.0%, P= 0.513) (Figs S3–S4).

Under the dominant model, a higher risk for development of PE was found for the 235T allele carriers in Caucasians (OR 1.63, 95% CI 1.20–2.22, P= 0.002), Mongolians (OR 1.54, 95% CI 1.05–2.26, P= 0.027) and the total population (OR 1.32, 95% CI 1.05–1.64, P= 0.015) but not in Africans (Table 2). The between-study heterogeneity was moderate among all studies (I2= 39.9%, P= 0.015). The results of meta-regression showed that the ethnic background explained 56.67% of the between-study heterogeneity (P= 0.002). After excluding the highly significant HWE-violating studies, similar results were observed except that the heterogeneity among all studies became marginal (I2= 26.2%, P= 0.106).

Under the additive model, a higher risk for development of PE was seen for the 235T allele in Caucasians (OR 1.36, 95% CI 1.06–1.73, P= 0.015), Mongolians (OR 1.44, 95% CI 1.10–1.90, P= 0.009) and the total population (OR 1.26, 95% CI 1.08–1.46, P= 0.003) but not in Africans (Table 2). The between-study heterogeneity was large among all studies (I2= 65.9%, P < 0.001). The results of meta-regression showed that none of the factors investigated in this study was a source of the heterogeneity. All observations remained similar after excluding the highly significant HWE-violating studies (data not shown).

Under the recessive model, a significant association was found in Caucasians (OR 1.47, 95% CI 1.02–2.13, P= 0.040), Mongolians (OR 1.53, 95% CI 1.00–2.32, P= 0.048) and the total population (OR 1.41, 95% CI 1.10–1.79, P= 0.006) but not in Africans (Table 2). The between-study heterogeneity was large among all studies (I2= 70.2%, P < 0.001). The results of meta-regression showed that none of the factors investigated in this study was a source of the between-study heterogeneity. After excluding the highly significant HWE-violating studies, all observations remained similar (data not shown).

To summarize, in the comparison between TT versus MM genotype and under all three inheritance models, (i) significant association between the AGT M235T polymorphism and PE risk was observed in Caucasians, Mongolians and the total population but not in Africans; and (ii) significant heterogeneity among all studies existed (Table 2). In the comparison between MT versus MM genotype, (i) significant association between the AGT M235T polymorphism and PE risk was observed only in Caucasians, but not in Mongolians, Africans or the total population; and (ii) significant heterogeneity among all studies also existed. The ethnic background might explain part of the heterogeneity in the comparison between MT versus MM genotype and under the dominant model.

T174M Polymorphism and Susceptibility of Pre-Eclampsia

Among six studies of the T174M polymorphism, two were conducted in Caucasians (Lévesque et al., 2004; Knyrim et al., 2008), two in Mongolians (Choi et al., 2004; Jiang & Fu, 2008), one in Africans (Wang et al., 2006) and one in Sinhalese (Dissanayake et al., 2009). The frequency of 174M allele in Africans (0.063) was significantly lower than that in Caucasians (0.106) (P= 4.33×10−7) and Sinhalese (0.120) (P= 7.97×10−7), but similar to that in Mongolians (0.085) (P= 0.057). Unlike the M235T polymorphism, the differences at the T174M polymorphism between these four ethnic groups were small (all pairwise FST < 0.05).

No significant association was found between the T174M polymorphism and risk of PE in the comparisons (MT vs. TT, and MM vs. TT) (Figs S5–S6) and under any of the three inheritance models (data not shown). No significant between-study heterogeneity was found except in the comparison between MT versus TT genotype (I2= 48.3%, P= 0.085).

Publication Bias and Sensitivity Analyses

The results of Begg's and Egger's tests indicated no obvious evidence of publication bias in each polymorphism (data not shown). A single study involved in the meta-analysis was deleted each time to reflect the influence of the individual data set to the pooled ORs. For the M235T polymorphism, the corresponding pooled ORs were not materially altered in Caucasians, Mongolians, Africans and the total population in the comparison between MT versus MM genotype except in the total population after exclusion of the study of Aggarwal et al. (2011) (OR 1.32, 1.12–1.55, P= 0.001), which was consistent with the result after excluding the highly significant HWE-violating studies, or the study of Bashford et al. (2001) (OR 1.25, 1.00–1.55, P= 0.048). The study of Aggarwal et al. (2011) was a highly significant HWE-violating study. Two studies [Zhang et al. (2003) and Aggarwal et al. 2011)] might be the main cause of heterogeneity. After exclusion of the study of Zhang et al. (2003), the heterogeneity among all studies was no longer significant (I2= 19.2%) and the ethnic background explained 100% of the heterogeneity among all studies (P= 0.001). After exclusion of the study of Aggarwal et al. (2011), the heterogeneities among all studies no longer existed (I2= 0.0%).

In the comparison between TT versus MM genotype of the M235T polymorphism, the corresponding pooled ORs were not materially altered in Caucasians, Mongolians, Africans and the total population. One study (Zhang et al., 2003) might be the main cause of heterogeneity. After exclusion of this study, the heterogeneity among all studies was no longer significant (I2= 17.7%, P= 0.203) (Table 2) and the ethnic background explained little of the heterogeneity among all studies (P= 0.160).

Under the dominant model for the M235T polymorphism, the corresponding pooled ORs were not materially altered in Caucasians, Mongolians, Africans and the total population. The same two studies [Zhang et al. (2003) and Aggarwal et al. (2011)] might be the main cause of heterogeneity. After exclusion of the study of Zhang et al. (2003) or Aggarwal et al. (2011), the heterogeneity was no longer significant (I2= 20.9% and 23.8%, respectively). The ethnic background explained 96.80% and 30.64%, respectively, of the heterogeneity among all studies (P < 0.05). After exclusion of the studies of Zhang et al. (2003) and Aggarwal et al. (2011), the heterogeneity among all studies no longer existed (I2= 0.0%), but still reached positive associations in Caucasians, Mongolians and the total population (data not shown).

Under the additive model for the M235T polymorphism, the corresponding pooled ORs were not materially altered in Caucasians, Mongolians, Africans and the total population. After exclusion of the study of Zhang et al. (2003), the change of the heterogeneity was relatively large (I2 changed from 71.5% to 49.8%) in Caucasians (Table 2); and none of the factors investigated in this study were the sources of the heterogeneity among all studies. After exclusion of the study of Kobashi et al. (1999) which was conducted in Mongolians, the heterogeneity was no longer significant (I2 changed from 53.6% to 12.1%) in Mongolians. The results suggested that the studies by Kobashi et al. (1999) and Zhang et al. (2003) might contribute to the heterogeneity under the additive model. However, after exclusion of these two studies, unexplained heterogeneity remained among all studies (I2= 46.8%, P= 0.003).

Under the recessive model for the M235T polymorphism, positive association in Caucasians was lost after deleting the study of Ward et al. (1993), Guo et al. (1997), Bouba et al. (2003), Zhang et al. (2003) or Jenkins et al. (2008), and positive association in Mongolians was lost after deleting the study of Ward et al. (1993), Guo et al. (1997), Kobashi et al. (1999), Li et al. (1999), Kim et al. (2004) or Song et al. (2007). After exclusion of the study of Zhang et al. (2003), the heterogeneity changed greatly (I2 changed from 64.2% to 37.4%) in Caucasians (Table 2); and none of the factors investigated in this study were the source of the heterogeneity among all studies. After exclusion of the study of Kobashi et al. (1999), the heterogeneity also changed greatly (I2 changed from 64.4% to 38.2%) in Mongolians. The results suggested that the studies by Kobashi et al. (1999) and Zhang et al. (2003) might also contribute to the heterogeneity under the recessive model. However, after exclusion of these two studies, unexplained heterogeneity remained among all studies (I2= 59.3%, P < 0.001).

To summarize, for the M235T polymorphism, the recessive model had more heterogeneity and lower stability and then was rejected as a poor fit to the observed data. Under the other two models and in the comparisons (MT vs. MM, and TT vs. MM), few pooled ORs were materially altered in the sensitivity analyses, indicating their robustness, although the case definitions in some studies were not mentioned or were not well-matched to the case definition described earlier. The study of Zhang et al. (2003) might be the main cause of heterogeneity in both comparisons and under all three inheritance models. However, after exclusion of this study, the corresponding pooled ORs were not materially altered except in Caucasians under the recessive model.

In Caucasians, the dominant model with lowest I2 and P-value for the significance of the OR (POR) (Table 2) fitted the data best. In Mongolians, the association had lowest POR but had large heterogeneity (I2= 53.6%, P= 0.022) under additive model, and therefore the dominant model with lowest I2 (I2= 0.0%, P= 0.935) fitted the data best. In the total population, compared to the additive model, the dominant model showed higher POR (0.015 vs. 0.003) but lower I2 (39.9% vs. 65.9%); moreover, after exclusion of the study of Zhang et al. (2003), the heterogeneity was no longer significant under dominant model but not under additive model. Therefore, the dominant model fitted the data best in the total population. Overall, the best-fit genetic disease model for the M235T polymorphism might be a dominant model.

Discussion

In the present meta-analysis, the M235T variant was associated with the development of PE whereas the T174M variant was not. For the M235T polymorphism, the TT genotype compared to the MM genotype was associated with an increased risk of PE. After stratification by ethnicity, this association remained significant in Caucasians and Mongolians but not in Africans. Similar results were observed under dominant, additive and recessive genetic models.

In a meta-analysis of the M235T polymorphism (Staessen et al., 1999), the TT genotype compared to the MM genotype was associated with an increased risk of hypertension in Caucasians but not in Asians or Africans. However, two latter meta-analyses of the M235T polymorphism indicated that the TT genotype compared to the MM genotype was also associated with an increased risk of hypertension in Asians (Sethi et al., 2003) and in Han Chinese (Ji et al., 2010). Of note, the former meta-analysis included seven studies in Asians (n= 1995) and six studies in Africans (n= 1108) (Staessen et al., 1999) .The latter two meta-analyses included 12 studies in Asians (n= 5181) and six studies in Africans (n= 1170) (Sethi et al., 2003), and 31 studies in Han Chinese (n= 8317) (Ji et al., 2010), respectively. It suggested that the lack of statistical significance in Asians or Africans in the former meta-analysis might be due to lower numbers of studies. We acknowledge that one limitation of the present meta-analysis of the M235T polymorphism is that it only included five studies in Africans (n= 2305). PE and essential hypertension share some inherited features but not all. Further meta-analysis warrants more studies of the M235T polymorphism with PE risk in Africans.

The term “African” generally refers to a very heterogeneous group in terms of human genetics. This may be another reason for no statistical significance in Africans. Of five studies in Africans in the present meta-analysis, only one (Roberts et al., 2004) was conducted in genetic stringency only indigenous Africans. Three studies (Curnow et al., 2000; Wang et al., 2006; Jenkins et al., 2008) were performed in African Americans (admixture of African and European populations). Hillermann et al. (2005) conducted their study in the South African population who harboured genes that are derived from varying degrees of admixture between indigenous groups and immigrants from Europe and the East.

A meta-analysis involving 34 studies (n= 11,079) demonstrated that the T174M polymorphism was associated with essential hypertension in the total and Asian populations but not in the Caucasian population. We acknowledge that another limitation of the present analysis is that it only included six studies involving 681 PE patients and 2076 controls for the T174M polymorphism. PE and essential hypertension only share some inherited features (Chappell & Morgan, 2006). We are not sure that the effect of the T174M polymorphism on PE is similar to that on essential hypertension. To reach a definitive conclusion, further studies based on larger sample size are still needed to assess the effect of the T174M polymorphism on PE.

Significant heterogeneity existed among all studies of the M235T polymorphism. The sensitivity analyses showed that the study of Zhang et al. (2003) might contribute to the heterogeneity. All PE patients recruited in the study of Zhang et al. (2003) delivered intrauterine growth restriction (IUGR) fetuses. This study also recruited 174 pregnant women with IUGR fetuses and demonstrated that pregnant women with higher 235T allele frequency were at higher risk of having an IUGR baby, although no other studies assessed this maternal effect on IUGR. If this effect is true, the frequency of the T allele, MT and TT genotypes of the M235T polymorphism in PE patients with IUGR fetuses might be higher than PE patients without IUGR fetuses. That might be why the study of Zhang et al. (2003) could cause the between-study heterogeneity.

Our results showed that the highly significant HWE-violating studies only materially altered the significance levels of the OR in the total population in the comparison between MT versus MM genotype of the M235T polymorphism. The results of one-way sensitivity analyses showed that, of the two highly significant HWE-violating studies, the study of Aggarwal et al. (2011) was the main factor which materially altered the significance levels of the OR in the total population in the comparison between MT versus MM genotype. The results of one-way sensitivity analyses also showed that the study of Aggarwal et al. (2011) might also contribute to the heterogeneity.

Great differences at the M235T polymorphism existed between Caucasians and Mongolians, and between Caucasians and Africans. The heterogeneity among studies of the M235T polymorphism was almost fully explained by the ethnic background in the comparison between MT versus MM genotype and under the dominant model after deleting the study of Zhang et al. (2003). However, after exclusion of the studies of Zhang et al. (2003) and Aggarwal et al. (2011), the heterogeneity among all studies no longer existed (I2= 0.0%) in the comparison between MT versus MM genotype and under the dominant model, suggesting that the studies of Zhang et al. (2003) and Aggarwal et al. (2011) might be the primary source of heterogeneity rather than the ethnic background.

Compared with the two previous meta-analyses of the M235T polymorphism with PE risk [15 studies involving 1128 patients and 2276 controls (Medica et al., 2007), and 17 studies with 1446 patients and 3829 controls (Zafarmand et al., 2008), respectively], this study is larger (31 studies involving 2555 patients and 6114 controls), with almost twice as many cases as the earlier meta-analyses. In addition, we assessed not only the association between the M235T polymorphism and PE risk but also the association between the T174M polymorphism and PE. Furthermore, we explored differences at the polymorphisms between ethnic groups and potential sources of heterogeneity across studies whereas Medica et al. (2007) did not. Zafarmand et al. (2008) only included six studies in Mongolians and did not observe that the M235T variant was significantly associated with PE in Mongolians in the ethnicity-based subgroup analyses. They included three studies in Africans and pooled them into the miscellaneous group. They also concluded that ethnic background was a significant source of the heterogeneity due to lower numbers of studies. Comparably, this study included 10 studies in Mongolians and five in Africans and found that the M235T variant was significantly associated with PE in Mongolians and ethnic background might not a significant source of the heterogeneity.

In conclusion, a significant association was detected between the AGT M235T polymorphism and PE in Caucasians, Mongolians and the total population, but not in Africans. The T174M variant was not associated with PE. Further large well-designed studies are still needed to determine the effects of AGT M235T and T174M polymorphisms on PE, especially in Africans.

Acknowledgements

This work was supported by grants from the Research Start-Up Fund in Hainan Medical College, the Young Scientists Fund of the National Natural Science Foundation of China (Grant No. 31100904), Program for Young Excellent Talents in Tongji University (2009KJ105) and Program Shanghai Municipal Health Bureau (2009151).

Ancillary