Evaluation of Association Studies and Meta-Analyses of eNOS Polymorphisms in Type 2 Diabetes Mellitus Risk

Background: Numerous studies reported the associations between endothelial nitric oxide synthase (eNOS) polymorphisms (4b/a VNTR (rs869109213), G894T (rs1799983) and T786C (rs2070744)) and type 2 diabetes mellitus (T2DM) risk. However, the conclusions were incongruent. Moreover, since no published meta-analyses were performed, a key issue regarding false-positive results needs to be addressed. Furthermore, four new articles have been published on these issues. Therefore, an updated meta-analysis was conducted to further explore these associations. Objectives: To investigate the association between eNOS 4b/a, G894T and T786C polymorphisms and T2DM risk. Methods: Studies were searched by using the PubMed, China National Knowledge Infrastructure (CNKI), Medline, Embase, International Statistical Institute (ISI) and the China Wanfang databases. Odds ratios (ORs) with 95% confidence intervals (CIs) were used to evaluate the associations using five genetic models. Furthermore, the false-positive report probability (FPRP), Bayesian false discovery probability (BFDP), and the Venice criteria were employed to assess the credibility of statistically significant associations. Results: Overall, the eNOS 4b/a polymorphism was associated with a significantly decreased T2DM risk in Asians (bb vs. aa: OR = 0.44, 95% CI = 0.23–0.84; ab + bb vs. aa: OR = 0.45, 95% CI = 0.24–0.86; bb vs. aa + ab: OR = 0.73, 95% CI = 0.59–0.91; b vs. a: OR = 0.71, 95% CI = 0.57–0.88); the eNOS G894T polymorphism was associated with a significantly increased T2DM risk in Asians (GT vs. GG: OR = 1.52, 95% CI = 1.15–2.01; GT + TT vs. GG: OR = 1.52, 95% CI = 1.15–2.01; T vs. G: OR = 1.39, 95% CI = 1.09–1.76); the eNOS T786C polymorphism was associated with a significantly increased T2DM risk in Indian (TC vs. TT: OR = 1.93, 95% CI = 1.27–2.94; TC + CC vs. TT: OR = 2.06, 95%CI = 1.26–3.36; C vs. T: OR = 1.90, 95%CI = 1.17–3.08). However, when a sensitivity analysis was performed after excluding low quality and Hardy–Weinberg Disequilibrium (HWD) studies, no significant association was found for the eNOS G894T polymorphism. After credibility assessment, we identified “less-credible positive results” for the statistically significant associations in the current meta-analysis. Conclusion: In conclusion, this article suggests that all substantial relationships between eNOS 4b/a, G894T, and T786C polymorphisms and T2DM risk are most likely due to false positive results rather than real connections or biological variables.


INTRODUCTION
Type 2 diabetes mellitus (T2DM), which is defined by chronic hyperglycemia caused by insulin resistance as well as multiple related micro-vascular and macro-vascular complications, is one of the most common chronic illnesses at home and abroad. Over the last 3 decades, the global prevalence of diabetes mellitus has more than quadrupled, making it one of the most serious global health issues (Moore et al., 2009). At the same time, it is reported that the incidence of T2DM is increasing at an alarming rate. Diabetes is anticipated to impact 702 million people by 2045, which means one in every eleven people will be affected, and huge amounts of money will be required globally to cure diabetes and manage its complications (https://diabetesatlas.org/en/). However, the pathogenesis of T2DM remains unclear and may be related to diet and exercise, obesity, geography, genetic susceptibility, environment, etc. Furthermore, there is abundant evidence that genetic predisposition plays a significant role in the etiology of T2DM (Ferland-McCollough et al., 2010). It has been reported that there is an important genetic predisposition for T2DM (Papazafiropoulou et al., 2017). At the same time, over 100 T2DM risk loci have been identified to date, although the molecular pathways of risk genes are unclear (Gaulton, 2017). In conclusion, genetic factors play an essential impact in the occurrence and development of T2DM.
Nitric oxide (NO) is a ubiquitous vasoactive substance, whose main function is to protect vascular endothelial cells from damage (Larsen et al., 2012). Endothelial dysfunction due to reduce in NO levels is an important mechanism for the development of T2DM. One of the essential enzymes in the process of NO generation is endothelial nitric oxide synthase (eNOS), which is encoded by the eNOS gene on chromosome 7q35-7q36 (Jamwal and Sharma, 2018). It has been observed that eNOS malfunction can cause a nitric oxide production problem, which can contribute to the development of characteristic T2DM aberrant metabolic phenotypes such as reduced glucose tolerance and insulin resistance (Li et al., 2002;Tsutsui et al., 2006). Therefore, eNOS polymorphisms may biologically be an ideal genetic marker for T2DM in biology.
Many eNOS gene polymorphisms have been identified in recent years, of which 4b/a, G894T, and T786C are the most investigated polymorphisms in T2DM (Wang et al., 1997;Veldman et al., 2002), although their associations remain controversial and equivocal. Several relevant meta-analyses have been performed to evaluate the correlations of T2DM with eNOS gene polymorphisms (Jia et al., 2013;Zhang et al., 2017;Dong et al., 2018), with conflicting results. And previously published meta-analyses did not evaluate the quality of the literature, nor did they evaluate the positive results to identify multiple comparisons. As a result, an updated meta-analysis was conducted to further investigate the possible association between eNOS genetic variants (4b/a, G894T, and T786C) and T2DM risk. This analysis included more papers and credible findings than previous meta-analyses (Jia et al., 2013;Zhang et al., 2017;Dong et al., 2018).

Search Strategy
The current study was performed according to the guidelines of the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) group (Moher et al., 2009). The literature was searched using PubMed, China National Knowledge Infrastructure (CNKI), Medline, Embase, ISI (International Statistical Institute) and the China Wanfang databases. The following search strategies were applied: (eNOS OR endothelial nitric oxide synthase OR nitric oxide synthase type III OR NOS3) AND (polymorphism OR variant OR mutation OR genotype OR allele) AND (diabetes OR mellitus OR diabetes mellitus OR DM). The literature search was updated to 15 March 2022. Furthermore, the reference lists of previously published metaanalyses were carefully reviewed to identify additional eligible studies (Jia et al., 2013;Zhang et al., 2017;Dong et al., 2018).

Selection Criteria
Inclusion criteria were as listed below: 1) case-control or cohort studies; 2) described the association between the eNOS 4b/a, G894T and T786C polymorphisms and risk of T2DM; 3) provided sufficient genotype data or the odds ratio (OR) with 95% confidence intervals (CI) in the selected literature. Exclusion criteria were as listed below: 1) duplicate genotype data; 2) studies with no available data; 3) meta-analyses of case reports, abstracts, reviews and letters.

Data Extraction and Quality Score Assessment
Two investigators independently extracted the data and crossexamined it, trying to resolve differences through discussion. If no consensus was reached after discussion, the third author would be invited to extract the data again for final review and confirmation. Moreover, the original authors could be contacted via e-mail if necessary. Races were divided into "Caucasians," "Asians," "Indians," and "Africans." "Mixed populations" was defined if race was not stated or the sample size of several races cannot be separated in original study.
The Newcastle-Ottawa Scale (NOS) (Wells et al., 2014) was applied by the two investigators to independently assess the quality of all included research. These scales are influenced by three factors: selection (four points), comparability (two points), and exposure (three points). Hardy-Weinberg Equilibrium (HWE) was employed to conduct a quality assessment on the basis of NOS (one point). The overall score varied from zero (worst) to ten (highest), with seven points or more as high quality.

Trial Sequential Analysis
Meta-analyses could increase the power and accuracy of evaluating intervention effects and are regarded as good evidence when available studies are used. However, misleading conclusions may be generated owing to random mistakes if the sample size is very small. Therefore, TSA was carried out to decrease random mistakes and predict the required information size (RIS) in this study (Brok et al., 2008;Thorlund et al., 2011). TSA was performed with the help of TSA 0.9 software (Copenhagen Trial Unit, Centre for Clinical Intervention Research, Copenhagen). The random effect model was used in this work. Alpha (type I error) and beta (type II error) were given as 0.05 and 0.2, respectively. The accruing information size (AIS) was used to determine information size, and the OR value was used to determine the combined effect amount. Based on O'Brien-Fleming-spending functions, a TSA employs trial sequential monitoring boundaries. In addition, the relative risk reduction (RRR) is set at 15% (Kulinskaya and Wood, 2014). If the cumulative Z-curve passes the monitoring border, the RIS line, or enters the futility region, strong evidence for our study may well be affirmed. Otherwise, additional research is required (Wetterslev et al., 2009).

Statistical Analysis
Potential associations between the eNOS genetic polymorphisms (4b/a VNTR, G894T and T786C) and T2DM risk were expressed by ORs and corresponding 95% CIs. Five genetic models were used for comparison: hybrid, homozygous, dominant, recessive and allele model. Chi-square-based Q-test and I 2 value were employed in assessing the Heterogeneity. When P was less than 0.10 and/or I 2 was greater than 50% , the random-effects model (Mantel and Haenszel, 1959) was adopted because of the significant heterogeneity. On the contrary, the fixed-effects model (DerSimonian and Laird, 2015) was adopted. In addition, the source of heterogeneity was explored by metaregression analysis. Subgroups were conducted by race, type of control, age and gender. Three methods were applied for sensitivity analyses: 1) excluded one study in turn; 2) eliminated low-quality and medium-quality or Hardy-Weinberg Disequilibrium (HWD) studies; 3) kept only high-quality and HWE studies. Furthermore, HWE was assessed using the Chi-square goodness-of-fit test. p > 0.05 was defined as HWE, otherwise as HWD in the control group. Begg's funnel plot (Begg and Mazumdar, 1994) and Egger's test (Egger et al., 1997) were applied to evaluate publication bias. If there were publication bias, the number of missing studies would be estimated and supplemented using a nonparametric "trim and fill" method (Dual and Tweedie, 2000). The false-positive report probability (FPRP) (Wacholder et al., 2004), Bayesian False Discovery Probability (BFDP) (Wakefield, 2007), and the Venice criteria (Ioannidis et al., 2008) were used to evaluate the credibility of statistically significant associations. Stata 12.0 software (STATA Corporation, College Station, TX) was applied to calculate all statistical analyses.

Study Characteristics
Initially, 984 articles were retrieved from PubMed, CNKI, Medline, Embase, ISI and the China Wan-fang databases. We excluded 321 papers by carefully evaluating titles and abstracts. Moreover, 16 were excluded due to duplication and invalidation of data, and 19 were excluded due to inadequate controls. Finally, 66 articles with 68 studies were eligible for inclusion in our meta-analysis ( Table 1). The detailed investigation process is shown in Figure 1. A total of 68 studies ( Figure 1) met our inclusion criteria (involving 15,988 T2DM cases and 25,452 controls), of which 36 studies reported the eNOS 4b/a (8,553 cases and 6,613 controls), 44 studies investigated the eNOS G894T (10,722 cases and 21,256 controls), and 13 studies reported the eNOS T786C (4,676 cases and 3,842 controls), as shown in Tables 2-4. Furthermore, twenty, thirty-seven, six, two and three studies were conducted to investigate Caucasians, Asians, Indians, Africans, and mixed groups, respectively. In addition, the eNOS 4b/a had 19 highquality studies and 17 low-quality studies, the eNOS G894T had 18 high-quality studies and 26 low-quality studies, and the eNOS T786C had five high-quality studies and eight low-quality studies. Moreover, the complete features, scores, HWE and the genotype frequencies of the selected literature were shown in Tables 2-4. Furthermore, Table 5 showed the results of the detailed quality scores for the included articles according to the NOS.

Heterogeneity and Sensitivity Analyses
Several possible causes of variation were discovered, including ethnicity, gender, sample size, age, quality score, type of controls and HWE. Therefore, a meta-regression analysis was used to identify causes of heterogeneity. For the eNOS G894T, no covariate was found as a possible cause of between-study variation. A meta-regression analysis revealed that HWE (ab vs. aa: p = 0.045) was the source of heterogeneity between the eNOS 4b/a polymorphism and the risk of T2DM. At the same time, the quality score (TC vs. TT: p = 0.045; TC + CC vs. TT: p = 0.042; C vs. T: p = 0.041) and HWE (CC vs. TT: p = 0.029; CC vs. TC + TT: p = 0.041) were the sources of heterogeneity between the eNOS T786C polymorphism and the risk of T2DM. Three methods were employed for sensitivity analyses in this meta-analysis. Firstly, results did not alter when a single study was removed each time. Second, when HWD studies were omitted, Asians were found to have a significantly lower risk of eNOS 4b/a polymorphism and T2DM in the overall analysis ( Table 6). For the eNOS G894T polymorphism, significantly increased T2DM risk was only observed in Asians and healthy population when we retained high-quality and HWE studies in the control group (Table 7). For the eNOS T786C polymorphism, a significant association was also discovered in the healthy population when we only included high-quality and HWE studies in the control group (Table 8).

Publication Bias
Begg's funnel plot and Egger's test revealed only publication bias between the eNOS G894T polymorphism and T2DM risk (GT vs. GG: p = 0.002; GT + TT vs. GG: p = 0.003; T against G: p = 0.014, Table 7). Then, publication bias was adjusted using the nonparametric "trim and fill" method. And we need to add 13, 11, and 10 articles in the future for GT vs. GG, GT + TT vs. GG, and T vs. G models, respectively ( Figure 5). In the overall analysis, the findings for GT vs. GG, GT + TT vs. GG, and T vs. G models did not change (data not shown), demonstrating that more research cannot alter the merger outcomes.

TSA Results
The TSA of the dominant model for the eNOS 4b/a and T786C polymorphisms revealed that the cumulative z-curve passed both the RIS line and the TSA threshold, indicating that no more evidence was required to confirm the conclusion. However, multiple comparisons and other confounding factors, we believe, can still increase the occurrence of false positive errors, so credibility analysis is still required for the eNOS 4b/a and T786C polymorphisms. The cumulative Z-curve of the dominant model for the eNOS G894T polymorphism did not surpass the TSA threshold, and the total number of cases and controls was smaller than the RIS, according to the TSA. Therefore, more trials were still required to confirm the association between eNOS G894T polymorphism and T2DM risk. Figure 6 displays the above results.

Credibility of the Identified Genetic Associations
The credibility of this meta-analysis was assessed using the FPRP, BFDP, and Venice criteria. Associations meeting the following criteria were regarded to be of high credibility [29]: 1) statistically significant associations were observed in at least two of the genetic models; 2) FPRP <0.2 and BFDP <0.8; 3) I 2 < 50%; and 4) statistical power >80%. All other major findings were viewed as "less credible results". All statistically significant associations were deemed "less credible" in this study.

DISCUSSION
T2DM is a polygenic genetic disease, which is also greatly influenced by environmental factors. And it is the outcome of the combined action of numerous genes and environmental factors. Several studies have shown that diabetes is the most important risk factor for        Frontiers in Genetics | www.frontiersin.org June 2022 | Volume 13 | Article 887415 mortality and disability caused by cardiovascular and cerebrovascular illnesses, according to several research. However, the molecular mechanism of the genetics of T2DM has not been elucidated. Much significant evidence indicates that the eNOS polymorphisms have been considered as potential genetic factors for T2DM. Numerous eNOS polymorphisms have been reported, and their relationship with various disorders has been studied, including coronary artery disease, myocardial infarction, coronary spasm, hypertension, end-stage renal disease (ESRD), and T2DM. Previous research has focused on three eNOS polymorphisms: the intronic 427-bp repeat (4b/a) in the promoter region; the G894T (Glu298Asp) missense mutation in exon seven and the T786C single nucleotide polymorphism. T786C inhibits eNOS transcription, G894T inhibits eNOS activity, and 4b/a inhibits plasma NO concentrations, which may be a reflection of eNOS activity. Many researchers have sought to investigate the potential relationship between eNOS polymorphisms and T2DM risk. Regrettably, no credible evidence is available, which might be attributable to a variety of factors such as small sample numbers, ethnic and geographical disparities. As a result, meta-analysis is an effective method to conquer these flaws.
Overall, the eNOS 4b/a was connected with a substantially lower the risk of T2DM in Asians; the eNOS G894T was connected with a significantly higher risk of T2DM in Asians, however, it had no significant effect on the risk of T2DM in Caucasians. the eNOS T786C was connected with a significantly higher risk of T2DM in Indians. However, after omitting low-quality and HWD studies, we observed that eNOS 4b/a polymorphism substantially lowered T2DM risk in the entire population while eNOS T786C polymorphism considerably raised T2DM risk in the whole population. However, after omitting low-quality and HWD studies, we observed that eNOS 4b/a polymorphism substantially lowered T2DM risk in the entire population while eNOS T786C polymorphism considerably raised T2DM risk in the whole population. The current study used many subgroups and distinct genetic models, which resulted in multiple comparisons, so the pooled p value must be corrected. FPRP has been described as a proper method for assessing the likelihood of significant outcomes in molecular epidemiology investigations using multiple hypothesis testing. Furthermore, Wakefield suggested a more accurate Bayesian metric of false detection in genetic epidemiology investigations in 2007. Many factors may lead to errors and biases, such as genotyping errors and phenotypic misclassification, of which statistical power was a significant factor. A substantial amount of evidence (statistical power>80%) can achieve a higher degree of statistical significance or reduce the false-discovery rate. As a result, in this study, we used the FPRP test, BFDP test and the Venice criteria to evaluate false discovery. All the statistically significant connections were less-credible in the current meta-analysis after assessing credibility. Our meta-analysis has also revealed heterogeneity. According to the results of the meta-regression study, the quality score and HWE were the sources of heterogeneity. Furthermore, bias and mistakes were widespread in several HWD studies with low quality and small sample size, making the conclusion of these original studies untrustworthy,   particularly in molecular epidemiology studies. And small sample studies with positive results may be easier to accept since they are likely to produce false-positive results because their research is less rigorous and frequently of poor quality. The asymmetry of the funnel plot was created by a study of low-quality small samples. Therefore, we added high-quality and HWE to evaluate sensitivity analyses in control studies. We hypothesized that the single and combined effects of the eNOS 4b/a, G894T, and T786C polymorphisms were linked with T2DM risk in all races based on the biochemical features outlined for these genes. Nevertheless, when we applied the FPRP, BFDP test, and Venice criteria to assess the credibility of this meta-analysis, all statistically significant relationships were declared "less credible" (greater heterogeneity, FPRP >0.2, BRDP >0.8, and lower statistic power). Therefore, these results indicated that much larger sample size was needed to study the potential gene-gene interactions.
A total of three previously published meta-analyses investigated the relationship between the eNOS 4b/a, G894T, and T786C polymorphisms and the risk of T2DM. There was a clear mismatch in the categorization of ethnic groupings between the previous related metaanalyses and the current meta-analysis. Furthermore, the sample size in this study was substantially greater. A total of 66 articles were included in this study, of which 36 articles reported the eNOS 4b/a (8,553 cases and 6,613 controls), 44 articles investigated the eNOS G894T (10,722 cases and 21,256 controls), and 13 articles reported the eNOS T786C (4,676 cases and 3,842 controls). Five genetic models were compared separately in this study. However, Dong et al., Zhang et al. and Jia et al. applied four genetic models. In addition, when we used the FPRP, BFDP test, and Venice criteria to assess the credibility of the previous meta-analyses, all statistically significant relationships were deemed "less credible." As a result, their findings may be unreliable.
Compared with the previous meta-analysis, the new metaanalysis had several advantages: 1) credibility was investigated using FPRP, BFDP test and Venice criteria; 2) the quality of the eligible research was evaluated; 3) The sample size was larger and the data collected was more detailed than the previous meta-analyses; 4) more subgroup analyses were performed according to the type of control, matching and quality score; 5) TSA was carried out to decrease random mistakes. However, there are still some potential limitations in the current meta-analysis. First, the current meta-analysis included only published research, although positive outcomes are known to be published more frequently than negative ones. Second, T2DM is a complex multi-genetic disorder, and the link between an individual SNP and T2DM risk is relatively weak. However, we did not retrieve the corresponding data on the combined impacts of gene-gene and gene-environment. Third, the relationship between the eNOS polymorphisms and the risk of T2DM complications has not been  investigated. Therefore, the current meta-analysis has a large sample size and a sufficiently large subgroup to help confirm our findings.  To summarize, our study reveals that all substantial relationships between the eNOS 4b/a, G894T, and T786C polymorphisms and T2DM risk are most likely due to false-positive results rather than real connections or biological variables. larger-scale epidemiological studies on this topic should be conducted in the future to confirm or disprove our findings.