This meta-analysis was performed according to the statements in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting standard [14]. Systematic literature search was performed in PubMed, Embase, Web of Science, Medline, Scopus, WanFang datebase, Vip datebase, and CNKI database up to October 30, 2020. Various combinations of terms used for searching were (“endothelial nitric oxide synthase” OR “nitric oxide synthase type III” OR “eNOS” OR “NOS3”) AND (“polymorphism” OR ‘‘variant” OR “mutation”) AND (“hypertension” OR “high blood pressure”). Moreover, we also retrieved and scrutinized related articles from the reference lists of literatures to replenish literatures that had not been identified in the initial search. A detailed form of the search strategy used in datebases was displayed in Additional file 1: Table S1.

Quality of the included studies was evaluated using the Newcastle–Ottawa scale (NOS) [15] that has a “star” rating system consisting of selection, comparability, and exposure. The highest score of this rating system is 9 points. Moreover, the data extraction and quality assessment were performed by two investigators (Jikang Shi and Yanbo Guo) independently, and conflicts were resolved by discussing with the third investigator (Sainan Liu) if the results of two investigators didn’t reach an agreement.

HWE was evaluated for control groups of each study using Goodness of fit Chi-square test, and P < 0.05 was considered as a significant deviation from HWE. The associations between eNOS rs1799983 polymorphisms and hypertension in this meta-analysis were measured based on five different genetic models including six comparisons: allelic model (T vs G), codominant model (GT vs GG and TT vs GG), dominant model (GT + TT vs GG), recessive model (TT vs GG + GT), overdominant model (GT vs GG + TT). Odds ratios (OR) and 95% confidence intervals (95% CI) were used to assess the strength of association between eNOS rs1799983 polymorphisms and hypertension. Q-statistic and I²-statistic were used to evaluate heterogeneity, random-effect models (DerSimonian and Laird methods) were used when heterogeneity existed (I² ≥ 50% considered heterogeneity existed in between-study in this meta-analysis); otherwise, fixed-effect models (Mantel and Haenszel methods) were used. Subgroup analyses were performed by region, ethnicity, and HWE to detect main sources of heterogeneity and observe differences of the association in different groups. Sensitivity analysis was conducted to evaluate stability of our results by omitting each study at each time. Publication bias was estimated using funnel plots, and quantified by the Egger’s tests (P < 0.05 considered statistically significant publication bias) [16]. All data management and statistical analyses were performed using R Studio (Version 1.1.383) (RStudio, Inc., MA, USA) for Windows.

The risk of random error in traditional meta-analysis may increase because of the dispersed data and repeated significance testing [17, 18]. TSA was used to reduce the risk of type I error by adjusting threshold for statistical significance and to evaluate the required information size (RIS) and statistical reliability [19]. In our meta-analysis, trial sequential analysis software (TSA, version 0.9; Copenhagen Trial Unit, Copenhagen, Denmark, 2011) were performed, and additional studies were not needed when Z-curve crossed the trial sequential monitoring boundary or RIS has reached; otherwise, further studies were needed.

A total of 60 eligible articles involving 14,185 cases and 13,407 controls were finally selected after strict screening on the basis of the inclusion and exclusion criteria, the protocol of literature search and selection is shown in Fig. 1, and the main characteristics and genotype distribution of the eligible studies are listed in Table 1.

There were significant heterogeneities between eNOS rs1799983 polymorphism and hypertension in the five different genetic models, and thus random-effects model was used for all analyses. We found significant association between eNOS rs1799983 polymorphism and the risk of hypertension under any genetic model (T vs G: OR 1.44, 95% CI 1.26–1.63; GT vs GG: OR 1.34, 95% CI 1.18–1.52; TT vs GG: OR 1.80, 95% CI 1.41–2.31; GT + TT vs GG: OR 1.42, 95% CI 1.25–1.63; TT vs GG + GT: OR 1.68, 95% CI 1.35–2.08; GT vs GG + TT: OR 1.24, 95% CI 1.11–1.40) (Fig. 2).

We performed subgroup analysis by region and ethnicity because gene polymorphism may be associated with variations in region and ethnicity. For region, there is only difference for the association between eNOS rs1799983 polymorphism and hypertension under overdominant model, when GT was compared with GG + TT, the association with risk of hypertension was identified in China (OR 1.29; 95% CI 1.12–1.49), and the association between eNOS rs1799983 polymorphism with risk of hypertension was found in any region under other genetic models. With regard to ethnicity, we found the association between eNOS rs1799983 polymorphism with risk of hypertension was significant in Asian population under all genetic models (T vs G: OR 1.42, 95% CI 1.27–1.58; GT vs GG: OR 1.37, 95% CI 1.21–1.54; TT vs GG: OR 1.64, 95% CI 1.35–2.00; GT + TT vs GG: OR 1.43, 95% CI 1.27–1.61; TT vs GG + GT: OR 1.56, 95% CI 1.29–1.88; GT vs GG + TT: OR 1.31, 95% CI 1.15–1.48); however, with respect to contrast of TT versus GG and TT versus GG + GT, the genotype TT was associated with the increased risk of hypertension not only in Asian population but also in other population (OR 2.07, 95% CI 1.05–4.08 and OR 1.87, 95% CI 1.07–3.25, respectively) (Table 2).

To examine the influence of individual study on the overall results, sensitivity analysis was performed by excluding a single study at each time in our meta-analysis. The results of sensitivity analysis showed that the corresponding pooled ORs and 95% CIs under any model of inheritance were not substantially altered after excluding any single study, suggesting that results of our meta-analysis were relatively stable and credible (Additional file 2: Figure S1).

Publication bias was evaluated by funnel plots and quantified by Egger’s tests. The funnel plots for recessive model (TT vs GG + GT) seemed symmetrical, and the results of Egger’s tests showed that there was no publication bias (P = 0.102); however, the funnel plots were asymmetrical in other genetic models for the association between eNOS rs1799983 polymorphism with hypertension, and the results of Egger’s tests showed that there were publication bias (T vs G: P = 0.026; GT vs GG: P = 0.023; TT vs GG: P = 0.032; GT + TT vs GG: P = 0.011; GT vs GG + TT: P = 0.038) (Additional file 3: Figure S2).

For the association between eNOS rs1799983 polymorphism with hypertension under codominant model (GT vs GG), codominant model (TT vs GG), and dominant model (GT + TT vs GG), the Z-curve crossed trial sequential monitoring boundary, although the sample size did not reach the RIS (Fig. 3B–D). However, for the association between eNOS rs1799983 polymorphism with hypertension under allelic model (T vs G), recessive model (TT vs GG + GT), and overdominant model (GT vs GG + TT), the Z-curve crossed trial sequential monitoring boundary, and the sample sizes were also more than the RIS (Fig. 3A, E, F). Therefore, concrete evidence indicates that further studies are not necessary for the association between eNOS rs1799983 polymorphism with hypertension.