Association of the endothelial nitric oxide synthase (eNOS) 4a/b polymorphism with the risk of incident diabetic retinopathy in patients with type 2 diabetes mellitus: a systematic review and updated meta-analysis

Abstract

Objective

To conduct a systematic review and updated meta-analysis on the potential association between endothelial nitric oxide synthase (eNOS) 4a/b polymorphism and the risk of developing diabetic retinopathy (DR) in patients with type 2 diabetes mellitus (T2DM) and to identify possible clinical biomarkers for early screening of DR.

Materials and methods

A meta-analysis based on case-control or cross-sectional studies was conducted to examine the correlation between eNOS 4a/b polymorphism and DR. Pooled odds ratio (OR) and 95% confidence interval (CI) were used to estimate the association strength.

Results

We included 19 studies, covering 7838 subjects. An association was observed in Caucasians (allelic model: OR = 1.273, 95% CI: 1.006–1.610, p = .045; recessive model: OR = 0.575, 95% CI: 0.371–0.892, p = .014; dominant model: OR = 1.268, 95% CI: 1.052–1.528, p = .013; homozygote model: OR = 1.833, 95% CI: 1.176–2.856, p = .007). Moreover, population-based studies have indicated an association between eNOS 4a/b polymorphism and DR susceptibility.

Conclusions

The present study showed that intron 4a allele of eNOS 4a/b is a risk factor for DR in Caucasians with T2DM. Thus, eNOS 4a/b may be used as a biomarker for the early screening and diagnosis of DR in Caucasian T2DM patients.

KEY MESSAGES

1. Endothelial nitric oxide synthase 4a/b gene polymorphism is not associated with the risk of developing diabetic retinopathy in the overall population, Asians, or Chinese Han patients with type 2 diabetes. However, 4a is a risk factor for the development of diabetic retinopathy in Caucasians.

2. Endothelial nitric oxide synthase 4a/b gene polymorphism is not associated with the type of diabetic retinopathy.

Keywords:

• Diabetic retinopathy
• endothelial nitric oxide synthase
• intron 4a/b
• polymorphism
• meta-analysis

View correction statement:

Correction

Introduction

Diabetic retinopathy (DR) is one of the leading causes of irreversible visual impairment in adults and is projected to affect 160.5 million adults by 2045 [1–3]. Although the pathogenesis of DR is not sufficiently understood, the concept that hereditary variables influence the onset and progression of DR is generally recognized [4].

Damage to the inner blood-retinal barrier is a typical pathological change in DR, and endothelial dysfunction has been shown to be involved in its pathogenesis [5–7]. Endothelial-derived nitric oxide (NO) synthesized by endothelial nitric oxide synthase (eNOS), a vital endogenous vasodilator for maintaining the normal physiological state of the microvasculature, plays a positive role in vascular protection, steady-state vasodilation and blood flow regulation [8–11]; however, abnormal NO utilization plays a deleterious role in ocular haemodynamics. Oxidative stress is the primary mechanism underlying endothelial cell dysfunction under hyperglycaemic conditions [12,13]. In DR, ischaemia and hypoxia can induce increased expression of superoxide anion (O2−), which can react with NO and produce the toxic oxidant peroxynitrite, leading to loss of the anti-atherosclerotic effect of NO [14]. Furthermore, although high glucose levels can upregulate eNOS expression and enhance NO production [15,16], excessive production of NO accelerates oxidative stress, which in turn impairs endothelial cell and tissue function by increasing DNA damage, triggering lipid peroxidation and depleting glutathione levels [9,17–20]. Diabetic individuals with DR have been reported to have much higher plasma NO levels than those without DR and healthy controls [9].

Gene polymorphisms are alterations in DNA sequences among individuals that may account for differences in phenotypes and occasionally in susceptibility to certain diseases. Recent research has shown that eNOS gene polymorphisms may affect eNOS activity, which in turn triggers a discrepancy in NO production, thereby inducing the development and progression of DR [1,21–24]. Studying the correlation between eNOS gene polymorphisms and DR has dual significance, as it may be both a susceptibility gene leading to the development of DR and a valid biological marker for the early detection and management of DR. Human eNOS is located on chromosome 7q35-36. It spans 21 kb, including 26 exons and 25 introns, and is transcribed, translated and modified to produce eNOS with a relative molecular mass of 135 × 10³ and 1203 amino acid sequences [25,26]. Studies have corroborated that the polymorphism of the 27-base pair (bp) variable number of tandem repeats (VNTR) located in intron 4 of eNOS gene is closely related to NO levels [27]. Two alleles of this gene are ‘a’ and ‘b’: the ‘a’ allele represents the mutant type, with four repeats, and the ‘b’ allele represents the wild type, with five repeats [28,29]. Consequently, numerous studies have attempted to investigate the association between eNOS 4a/b polymorphisms and DR risk.

Although it is theoretically believed that the eNOS 4a/b gene polymorphism is related to the occurrence of DR, and previous studies have observed that 4a was associated with the susceptibility to DR in Tunisians, Caucasians, Africans and Asian Indians, studies have also proposed that 4a is not related to the risk of DR in Caucasians and Asians. Therefore, the association between eNOS 4a/b polymorphisms and the incidence of DR in different ethnic groups remains controversial. A meta-analysis is a robust statistical tool that provides more reliable results than a single study and explains inconsistent conclusions. Meta-analyses have been conducted to address this gap, in the absence of different ethnic studies. Nevertheless, prior meta-analyses [29–32] have been limited by the number of original studies, especially among Caucasians and Tunisians. Studies on Caucasians and Tunisians have been added in the last five years, and the results of these studies warrant updating.

Therefore, to provide more reliable research evidence for this issue, we updated the prior meta-analyses with a systematic review and meta-analysis, comparing ethnicity stratification, DR typing and control source stratification to evaluate the impact of reported eNOS 4a/b gene polymorphisms on DR among T2DM patients.

Materials and methods

The guidelines stated by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) were implemented in the present systematic review and meta-analysis, and the PRISMA guideline checklist is available in Supplementary Material S1.

Search strategy

We conducted a comprehensive search of PubMed, EMBASE, and Web of Science databases to identify all relevant studies published between January 1999 and March 2023 related to the association between eNOS 4a/b polymorphisms and the risk of incident DR in patients with type 2 diabetes. References to the original studies, reviews, systematic evaluations and meta-analyses included in our preliminary review were manually screened to eliminate the possibility of missing relevant studies. Unpublished studies were excluded. The most recent renewal was conducted on 4 March 2023. A search approach was provided (e.g. EMBASE).

#1 Diabetic retinopathy

#2 Diabetic retinopathies OR retinopathies, diabetic OR retinopathy, diabetes

#3 #1 OR #2

#4 Endothelial nitric oxide synthase

#5 Nitric oxide synthase, type iii OR endothelial nitric oxide synthase OR eNOS enzyme OR ecNOS enzyme OR endothelial constitutive nitric oxide synthase

#6 #4 OR #5

#7 Polymorphism

#8 Polymorphisms, genetic OR genetic polymorphisms OR genetic polymorphism OR gene polymorphisms OR gene polymorphism OR polymorphisms, gene OR polymorphism, gene OR polymorphisms (genetics) OR polymorphism (genetics)

#9 #7 OR #8

#10 #3 AND #6 AND #9

Inclusion and exclusion criteria

Studies that satisfied all of the following criteria were included:

(1) Studies on DR and eNOS 4a/b polymorphisms. (2) Case-control studies or cross-sectional studies. (3) The case group recruited patients with type 2 diabetes and DR, whereas the control group selected patients with a similar duration of type 2 diabetes but without retinopathy. (4) Allele or genotype counts or frequencies for the case and control groups were calculated from study data. (5) The outcome predictors of the study were odds ratio (OR) and 95% confidence interval (CI). (6) DR Diagnosis based on structural and functional evidence of diabetic retinal damage.

Studies conforming to any of these criteria were excluded:

(1) Studies in which complete data were unavailable. (2) Meta-analysis, systematic review, case report, review, abstracts, conference proceedings, editorials and animal experimentation. (3) Duplicate publications. (4) Published studies that are not in English or Chinese. (5) The Newcastle–Ottawa Scale (NOS) scored less than 7.

Data extraction and quality assessment

Two authors (Fan and Shi) separately collected and reviewed the following information according to the inclusion and exclusion criteria: first author, publication year, nationality, ethnicity, publication source, control group source, study design, genotyping methods, DR typing, case and control group sample sizes, allelic and genotypic counts and frequencies (Tables 1 and 2). If there were any discrepancies, a third reviewer (Zhang) assessed the study independently and a consensus decision was made. Additionally, we applied the chi-square test to calculate the probability value (p value) of the Hardy–Weinberg equilibrium (HWE) test based on genotypic frequencies in the control group. Furthermore, two authors (Zhang and Fan) separately evaluated the methodological quality of all included studies using the Newcastle–Ottawa Scale (NOS), which was adapted for evaluating case-control, cross-sectional and cohort studies [33]. The NOS consists of three aspects: selection of cases and controls (0–4 points), comparability between cases and controls (0–2 points) and exposure in cases and controls (0–3 points) [34]. The highest number of stars accessible in a study was 9; the quality of a study was regarded as high if it obtained 7–9 stars, medium if it obtained 4–6 stars and low if it obtained 0–3 stars [35].

Table 1. The characteristics of included studies.

Author	Publication year	Publication source	Country	Ethnicity	Control group source	Study design	Genotyping methods	Type of DR	NOS score
Midani et al. [24]	2019	PubMed	Tunisia	Tunisian	HB	CCS	PCR-RFLP	NA	8
Dejan et al. [15]	2018	PubMed	Slovenia	Caucasian	HB	CCS	PCR	NA	8
Narne, P. et al. [16]	2016	PubMed	India	Asian	PB	CCS	PCR	NA	7
She et al. [11]	2015	PubMed	China	Asian	PB	CSS	PCR	PDR/NPDR	7
Cheema et al. [10]	2012	PubMed	India	Asian	PB	CCS	PCR	PDR/NPDR	7
Cilenšek et al. [36]	2012	Web of Science	Slovenia	Caucasian	HB	CCS	PCR	NA	8
Santos et al. [37]	2012	PubMed	Brazil	Caucasian	PB	CCS	PCR	PDR/NPDR	7
Mehrab-Mohseni et al. [8]	2011	PubMed	Iran	Asian	HB	CSS	PCR	NA	8
Yue et al. [40]	2011	Web of Science	China	Asian	HB	CCS	PCR	NA	8
Li et al. [23]	2010	Web of Science	China	Asian	HB	CCS	PCR	PDR/NPDR	8
Ezzidi et al. [38]	2008	PubMed	Tunisia	Tunisian	HB	CCS	PCR	NA	8
Petrovic et al. [39]	2008	PubMed	Slovenia	Caucasian	HB	CSS	PCR-RFLP	NA	8
Uthra et al. [19]	2007	PubMed	India	Asian	PB	CCS	PCR	PDR/NPDR	7
Chen et al. [18]	2007	PubMed	Ghana and Nigeria	African	PB	CSS	PCR	NA	7
Suganthalakshmi et al. [4]	2006	PubMed	India	Asian	PB	CCS	PCR-RFLP	NA	7
de Syllos et al. [22]	2006	PubMed	Brazil	Caucasian	HB	CCS	PCR-RFLP	NA	8
Zhang et al. [41]	2005	Web of Science	China	Asian	PB	CCS	PCR	PDR/NPDR	7
Sun et al. [42]	2004	Web of Science	China	Asian	HB	CCS	PCR-RFLP	PDR/NPDR	7
Awata et al. [28]	2004	PubMed	Japan	Asian	PB	CCS	PCR	PDR/NPDR	7

Note: PB: population-based; HB: hospital-based; CCS: case-control study; CSS: cross-sectional study; PCR: polymerase chain reaction; PCR-RFLP: polymerase chain reaction-restriction fragment length polymorphism; PDR: proliferative diabetic retinopathy; NPDR: non-proliferative diabetic retinopathy; NA: not available; NOS: Newcastle-Ottawa Scale.

Table 2. Genotype and allele frequencies in the case and control groups.

Study	Publication year	Case/control	Genotype frequencies						Allele frequencies				p Value for HWE
			Cases(N)			Controls(N)			Cases(N)		Controls(N)
			aa	ba	bb	aa	ba	bb	a	b	a	b
Midani et al. [24]	2019	127/110	17	52	58	19	23	68	86	168	61	159	.000***
Dejan et al. [15]	2018	270/256	16	77	177	7	50	199	109	431	64	448	.086
Narne et al. [16]	2016	149/162	2	52	95	3	55	104	56	242	61	263	.159
She et al. [11]	2015	130/148	0	28	102	1	32	115	28	232	34	262	.441
Cheema et al. [10]	2012	842/604	87	344	411	66	270	268	518	1166	402	806	.871
Cilenšek et al. [36]	2012	172/405	15	51	106	13	114	278	81	263	140	670	.754
Santos et al. [37]	2012	407/184	12	111	284	5	53	126	135	679	63	305	.838
Mehrab-Mohseni et al. [8]	2011	41/72	1	9	31	5	20	47	11	71	30	114	.180
Yue et al. [40]	2011	154/167	2	27	125	4	45	118	31	277	53	281	.905
Li et al. [23]	2010	87/79	5	13	69	7	11	61	23	151	25	133	.000***
Ezzidi et al. [38]	2008	383/489	17	115	251	33	180	276	149	617	246	732	.621
Petrovic et al. [39]	2008	283/143	16	91	176	7	44	92	123	443	58	228	.563
Uthra et al. [19]	2007	187/188	7	54	126	1	57	130	68	306	59	317	.045
Chen.et al. [18]	2007	68/301	10	17	41	48	136	117	37	99	232	370	.423
Suganthalakshmi et al. [4]	2006	120/90	7	27	86	3	18	69	41	199	24	156	.202
de Syllos et al. [22]	2006	56/114	1	16	39	3	27	84	18	94	33	195	.643
Zhang et al. [41]	2005	119/203	0	11	108	2	31	170	11	227	35	371	.662
Sun et al. [42]	2004	147/155	3	36	108	1	21	133	42	252	23	287	.864
Awata et al. [28]	2004	117/109	1	25	91	3	24	82	27	207	30	188	.450

Notes: Case: diabetic retinopathy; Control: diabetic without retinopathy; HWE: Hardy–Weinberg equilibrium, p<.05 was considered significant; ***: p<.001.

Bold indicates the P value was less than 0.05 and a significant association was found between eNOS (4a/4b) and the risk of incident DR.

Statistical analysis

We evaluated the gene frequencies of allelic (a vs. b), recessive (ab + bb vs. aa), dominant (aa + ab vs. bb), additive (aa + bb vs. ab), homozygous (aa vs. bb) and heterozygous (ab vs. bb) models of the eNOS 4a/b polymorphism in the case and control groups. Summary odds ratios (ORs) and 95% confidence intervals (95% CIs) were calculated to analyse differences between the case and control groups in the allelic and genetic models. Between-study heterogeneity was examined using the chi-square test based on the Q statistic and I² indicator. Significant heterogeneity was perceived when p < .1 for the Q statistic and I² > 50% for the I² indicator [32]. The fixed effects model using the Mantel–Haenszel method was selected when there was no heterogeneity among the included studies. Alternatively, the random effects model using the DerSimonian and Laird method was used. Sensitivity analyses were performed by excluding one study at a time and recalculating the summarized OR values to assess studies with notable effects on interstudy heterogeneity. Meta-regression analysis was used to test the sources of heterogeneity. Subgroup analysis was performed to assess whether different races, DR types and control sources were the origin of heterogeneity. Publication bias was evaluated using Begg’s funnel plots and Egger’s tests. Asymmetric funnel plots or Egger’s test with p values less than .05 revealed proof of publication bias. All statistical analyses were performed using STATA software (version 14.0; Stata, College Station, TX, USA), and two-sided p values less than .05 was accepted as statistically significant.

Results

Characteristics of the studies

A total of 121 studies were retrieved, including 26 from PubMed, 20 from EMBASE and 75 from Web of Science. First, 40 repetitive studies were eliminated; 60 studies with irrelevant content and inconsistent research methods were excluded by reading the titles and abstracts, and two studies were removed by reading the full text. Finally, we included 19 eligible studies linked to eNOS 4a/b polymorphism and DR in patients with T2DM. Figure 1 depicts the complete process of inclusion and exclusion of literature for this study. Among these, 14 papers [4,8,10,15,16,18,19,22,24,28,36–39] were published in English, and the remaining five [11,23,40–42] were published in Chinese. There were 3859 cases and 3979 controls in the present study, including 11 studies on Asians (2006 cases, 1977 controls), five studies on Caucasians (1188 cases, 1102 controls), two studies on Tunisians (510 cases, 599 controls) and one study on Africans (68 cases, 301 controls). Five studies investigated the Han Chinese ethnic group (637 cases, 752 controls). Eight articles divided DR into proliferative diabetic retinopathy (PDR) (678 cases, 1670 controls) and non-proliferative diabetic retinopathy (NPDR) (1358 cases, 1670 controls). Ten studies used hospital-based (HB) control group sources (1720 cases, 1990 controls), and nine studies used population-based (PB) control group sources (2139 cases, 1989 controls). Sixteen studies in the control group were eligible for HWE, in addition to three studies [19,23,24]. Deleting three studies did not materially alter the statistical results; therefore, we retained these papers.

Figure 1. Flowchart showing studies of the selection process.

Meta-analysis results

Table 3 summarizes the results of the overall population and ethnicity-based subgroup analyses of the risk association between the eNOS 4a/b polymorphism and DR in patients with T2DM. In the overall population, no genetic model found a risk relationship between eNOS 4a/b polymorphism and DR. However, subgroup analyses based on the race of the study population showed that the 4a/b polymorphism in Caucasians was significantly associated with the occurrence of DR (allelic model: OR = 1.273, 95% CI: 1.006–1.610, p = .045; recessive model: OR = 0.575, 95% CI: 0.371–0.892, p = .014; dominant model: OR = 1.268, 95% CI: 1.052–1.528, p = 0.013; homozygote model: OR = 1.833, 95% CI: 1.176–2.856, p = .007) (Figure 2). In contrast, no significant difference was detected in the relationship between the eNOS 4a/b polymorphism and DR in Asian, Tunisian and Han Chinese individuals. Subgroup analysis for control group sources demonstrated statistically significant relationships between the eNOS 4a/b polymorphism and population-based studies: dominant model: OR = 0.858, 95% CI: 0.749–0.983, p = .027; additive model: OR = 1.168, 95% CI: 1.016–1.341, p = .029; heterozygote model: OR = 0.847, 95% CI: 0.735–0.977, p = .022 (Table 4, Figure 3). In addition, no association between eNOS 4a/b polymorphisms and DR was found in the subgroup analysis of the type of DR (Table 5).

Figure 2. Forest plot for the association between diabetic retinopathy risk and eNOS 4b/a polymorphism for allelic model (a), recessive model (b), dominant model (c), homozygote model (d) in the ethnicity-based subgroup analysis. The red diamond represents the summary or value. Gray squares indicate the or in each study. The horizontal line symbolizes 95% CI. The vertical dashed line means the overall pooled or value. The vertical solid line is at the null value (or = 1.0).

Figure 3. Forest plot for the association between diabetic retinopathy risk and eNOS 4b/a polymorphism for dominant model (a), additive model (b) and heterozygote model (c) in the HB-PB group. The red diamond represents the summary or value. Gray squares indicate the or in each study. The horizontal line symbolizes 95% CI. The vertical dashed line means the overall pooled or value. The vertical solid line is at the null value (or = 1.0).

Table 3. Results of overall population and ethnicity-based subgroup analysis of eNOS (4b/4a) polymorphism and DR risk in T2DM patients.

Variables		N	References	Case/Control	Genetic model	Fixed-effects model		Random-effects model		Heterogeneity		Egger’s Test (p)
						OR (95%CI)	p value	OR (95%CI)	p value	PH	I^2 (%)
Overall					Allelic model (a versus b)
	Total	19	[4,8,10,11,15,16,18,19,22–24,28,36–42]	3859/3979		0.973(0.897–1.056)	0.517	0.995(0.849–1.167)	0.953	0.000	67.6%	0.750
	HWE	16	[4,8,10,11,15,16,18,22,28,36–42]	3458/3602		0.952(0.874–1.038)	0.265	0.972(0.812–1.163)	0.754	0.000	70.9%	0.802
Ethnicity
Asians	Total	11	[4,8,10,11,16,19,23,28,40–42]	2093/1977		0.925(0.826–1.035)	0.173	0.934(0.766–1.138)	0.496	0.025	51.2%	–
	HWE	9	[4,8,10,11,16,28,40–42]	1819/1710		0.9067(0.804–1.022)	0.108	0.911(0.722–1.151)	0.436	0.018	56.7%	–
Caucasians	Total/HWE	5	[15,22,36,37,39]	1188/1102		1.284(1.094–1.506)	0.002	1.273(1.006–1.610)	0.045	0.084	51.4%	–
Tunisians	Total	2	[24,38]	510/599		0.841(0.691–1.024)	0.085	0.958(0.523–1.755)	0.891	0.008	85.9%	–
Han Chinese	Total	5	[11,23,40–42]	637/752		1.130(0.888–1.438)	0.322	1.145(0.704–1.862)	0.585	0.005	73.3%	–
Overall					Recessive model (4a4b + 4b4b versus 4a4a)
	Total	19	[4,8,10,11,15,16,18,19,22–24,28,36–42]	3859/3979		0.990(0.809–1.210)	0.918	0.969(0.738–1.271)	0.818	0.198	21.1%	0.969
	HWE	16	[4,8,10,11,15,16,18,22,28,36–42]	3458/3602		0.981(0.791–0.217)	0.863	0.944(0.712–1.253)	0.692	0.259	17.0%	0.767
Ethnicity
Asians	Total	11	[4,8,10,11,16,19,23,28,40–42]	2093/1977		1.056(0.795–1.404)	0.706	1.079(0.804–1.446)	0.613	0.524	0.0%	–
	HWE	9	[4,8,10,11,16,28,40–42]	1819/1710		1.107(0.820–1.494)	0.507	1.098(0.809–1.490)	0.550	0.761	0.0%	–
Caucasians	Total/HWE	5	[15,22,36,37,39]	1188/1102		0.575(0.371–0.892)	0.014	0.568(0.363–0.889)	0.013	0.388	3.3%	–
Tunisians	Total	2	[24,38]	510/599		1.470(0.930–2.325)	0.099	1.468(0.928–2.323)	0.101	0.764	0.0%	–
Han Chinese	Total	5	[11,23,40–42]	637/752		0.710(0.323–1.562)	0.394	0.694(0.302–1.594)	0.389	0.697	0.0%	–
Overall					Dominant model (4a4a + 4a4b versus 4b4b)
	Total	19	[4,8,10,11,15,16,18,19,22–24,28,36–42]	3859/3979		0.960(0.870–1.059)	0.410	0.988(0.818–1.192)	0.896	0.000	67.6%	0.619
	HWE	16	[4,8,10,11,15,16,18,22,28,36–42]	3458/3602		0.928(0.836–1.029)	0.156	0.946(0.772–1.160)	0.597	0.000	68.7%	0.767
Ethnicity
Asians	Total	11	[4,8,10,11,16,19,23,28,40–42]	2093/1977		0.903(0.788–1.036)	0.145	0.925(0.749–1.142)	0.468	0.057	44.1%	–
	HWE	9	[4,8,10,11,16,28,40–42]	1819/1710		0.885(0.764–1.025)	0.104	0.908(0.703–1.171)	0.456	0.029	53.3%	–
Caucasians	Total/HWE	5	[15,22,36,37,39]	1188/1102		1.268(1.052–1.528)	0.013	1.264(0.991–1.614)	0.060	0.163	38.8%	–
Tunisians	Total	2	[24,38]	510/599		0.860(0.676–1.094)	0.220	1.118(0.404–3.092)	0.830	0.001	91.7%	–
Han Chinese	Total	5	[11,23,40–42]	637/752		1.111(0.852–1.449)	0.436	1.121(0.664–1.891)	0.669	0.006	72.0%	–
Overall					Additive model (4a4a + 4b4b versus 4a4b)
	Total	19	[4,8,10,11,15,16,18,19,22–24,28,36–42]	3859/3979		1.047(0.946–1.159)	0.371	1.008(0.845–1.202)	0.932	0.000	59.9%	0.440
	HWE	16	[4,8,10,11,15,16,18,22,28,36–42]	3458/3602		1.088(0.977–1.210)	0.124	1.062(0.892–1.265)	0.501	0.005	54.2%	0.713
Ethnicity
Asians	Total	11	[4,8,10,11,16,19,23,28,40–42]	2093/1977		1.097(0.954–1.261)	0.195	1.073(0.892–1.291)	0.453	0.199	25.7%	–
	HWE	9	[4,8,10,11,16,28,40–42]	1819/1710		1.105(0.952–1.283)	0.189	1.076(0.856–1.352)	0.529	0.103	39.7%	–
Caucasians	Total/HWE	5	[15,22,36,37,39]	1188/1102		0.865(0.712–1.050)	0.143	0.865(0.703–1.064)	0.170	0.345	10.8%	–
Tunisians	Total	2	[24,38]	510/599		1.046(0.814–1.344)	0.724	0.738(0.213–2.562)	0.633	0.000	93.3%	–
Han Chinese	Total	5	[11,23,40–42]	637/752		0.934(0.710–1.231)	0.629	0.939(0.570–1.547)	0.805	0.018	66.6%	–
Overall					Homozygote model (4a4a versus 4b4b)
	Total	19	[4,8,10,11,15,16,18,19,22–24,28,36–42]	3859/3979		0.974(0.793–1.197)	0.804	1.021(0.743–1.403)	0.897	0.064	35.4%	0.828
	HWE	16	[4,8,10,11,15,16,18,22,28,36–42]	3458/3602		0.946(0.759–1.179)	0.620	1.000(0.702–1.424)	1.000	0.068	37.1%	0.932
Ethnicity
Asians	Total	11	[4,8,10,11,16,19,23,28,40–42]	2093/1977		0.889(0.663–1.193)	0.433	0.867(0.639–1.174)	0.356	0.456	0.0%	–
	HWE	9	[4,8,10,11,16,28,40–42]	1819/1710		0.839(0.615–1.145)	0.269	0.844(0.614–1.160)	0.297	0.689	0.0%	–
Caucasians	Total/HWE	5	[15,22,36,37,39]	1188/1102		1.833(1.176–2.856)	0.007	1.827(1.127–2.961)	0.014	0.331	13.1%	–
Tunisians	Total	2	[24,38]	510/599		0.722(0.453–1.150)	0.170	0.742(0.408–1.352)	0.330	0.209	36.8%	–
Han Chinese	Total	5	[11,23,40–42]	637/752		1.420(0.646–3.120)	0.382	1.460(0.634–3.366)	0.374	0.611	0.0%	–
Overall					Heterozygote model (4a4b versus 4b4b)
	Total	19	[4,8,10,11,15,16,18,19,22–24,28,36–42]	3859/3979		0.952(0.858–1.055)	0.347	0.990(0.820–1.196)	0.920	0.000	64.3%	0.464
	HWE	16	[4,8,10,11,15,16,18,22,28,36–42]	3458/3602		0.915(0.821–1.021)	0.112	0.938(0.773–1.137)	0.513	0.001	61.3%	0.735
Ethnicity
Asians	Total	11	[4,8,10,11,16,19,23,28,40–42]	2093/1977		0.904(0.784–1.043)	0.165	0.928(0.763–1.129)	0.457	0.146	31.6%	–
	HWE	9	[4,8,10,11,16,28,40–42]	1819/1710		0.892(0.765–1.039)	0.142	0.919(0.723–1.169)	0.493	0.073	44.3%	–
Caucasians	Total/HWE	5	[15,22,36,37,39]	1188/1102		1.198(0.985–1.458)	0.070	1.199(0.960–1.497)	0.110	0.288	20.0%	–
Tunisians	Total	2	[24,38]	510/599		0.912(0.707–1.178)	0.482	1.328(0.362–4.875)	0.669	0.000	93.4%	–
Han Chinese	Total	5	[11,23,40–42]	637/752		1.079(0.819–1.421)	0.590	1.075(0.645–1.793)	0.780	0.014	68.1%	–

The boldface: The p value was less than 0.05 and a significant association was found between eNOS (4a/4b) and the risk of incident DR. N: number of studies; OR: odd ratio; 95%CI: 95% confidence intervals; HWE: Hardy–Weinberg equilibrium.

Table 4. Result of source of control subgroup analysis of eNOS (4b/4a) polymorphism and DR risk in T2DM patients.

Variables	N	References	Case/Control	Genetic model	Fixed-effects model		Random-effects model		Heterogeneity
					OR (95%CI)	p value	OR (95%CI)	p value	PH	I^2 (%)
Source of control				Allelic model
HB	10	[8,15,22–24,36,38–40,42]	1720/1990		1.066(0.945–1.201)	0.297	1.086(0.824–1.431)	0.558	0.000	77.9%
PB	9	[4,10,11,16,18,19,28,37,41]	2139/1989		0.901(0.806–1.006)	0.064	0.905(0.778–1.053)	0.197	0.200	27.4%
				Recessive model
HB	10	[8,15,22–24,36,38–40,42]	1720/1990		0.961(0.714–1.294)	0.796	0.941(0.593–1.492)	0.794	0.060	45.0%
PB	9	[4,10,11,16,18,19,28,37,41]	2139/1989		1.014(0.772–1.331)	0.922	1.034(0.783–1.366)	0.813	0.613	0.0%
				Dominant model
HB	10	[8,15,22–24,36,38–40,42]	1720/1990		1.084(0.940–1.249)	0.267	1.131(0.825–1.549)	0.445	0.000	76.3%
PB	9	[4,10,11,16,18,19,28,37,41]	2139/1989		0.858(0.749–0.983)	0.027	0.859(0.711–1.039)	0.118	0.136	35.2%
				Additive model
HB	10	[8,15,22–24,36,38–40,42]	1720/1990		0.925(0.797–1.073)	0.301	0.863(0.639–1.164)	0.334	0.000	70.8%
PB	9	[4,10,11,16,18,19,28,37,41]	2139/1989		1.168(1.016–1.341)	0.029	1.162(0.985–1.371)	0.075	0.298	16.2%
				Homozygote model
HB	10	[8,15,22–24,36,38–40,42]	1720/1990		1.097(0.812–1.482)	0.547	1.120(0.677–1.854)	0.659	0.027	52.1%
PB	9	[4,10,11,16,18,19,28,37,41]	2139/1989		0.879(0.663–1.164)	0.367	0.860(0.644–1.148)	0.307	0.464	0.0%
				Heterozygote model
HB	10	[8,15,22–24,36,38–40,42]	1720/1990		1.084(0.933–1.260)	0.293	1.164(0.848–1.599)	0.348	0.000	73.4%
PB	9	[4,10,11,16,18,19,28,37,41]	2139/1989		0.847(0.735–0.977)	0.022	0.850(0.702–1.028)	0.094	0.172	30.8%

The boldface: The p value was less than 0.05 and a significant association was found between eNOS (4a/4b) and the risk of incident DR. PB: population-based; HB: hospital-based; N: number of studies; OR: odd ratio; 95%CI: 95% confidence intervals.

Table 5. Result of type of DR subgroup analysis of eNOS (4b/4a) polymorphism and DR risk in T2DM patients.

Variables	N	References	Case/control	Genetic model	Fixed-effects model		Random-effects model		Heterogeneity
					OR (95%CI)	p Value	OR (95%CI)	p Value	PH	I^2 (%)
Type of DR				Allelic model
PDR	8	[10,11,19,23,28,37,41,42]	678/1670		0.822(0.697–0.970)	.021	1.036(0.738–1.454)	.837	0.001	70.6%
NPDR	8	[10,11,19,23,28,37,41,42]	1358/1670		1.001(0.879–1.140)	.990	0.970(0.792–1.188)	.770	0.072	44.5%
PDR	8	[10,11,19,23,28,37,41,42]	678/1670	Recessive model	1.427(0.922–2.211)	.111	1.086(0.556–2.119)	.809	0.133	35.7%
NPDR	8	[10,11,19,23,28,37,41,42]	1358/1670		0.881(0.643–1.205)	.448	0.873(0.633–1.203)	.406	0.700	0.0%
PDR	8	[10,11,19,23,28,37,41,42]	678/1670	Dominant model	0.822(0.675–1.001)	.052	1.022(0.699–1.493)	.912	0.002	67.3%
NPDR	8	[10,11,19,23,28,37,41,42]	1358/1670		0.970(0.826–1.139)	.712	0.953(0.763–1.191)	.672	0.125	36.7%
PDR	8	[10,11,19,23,28,37,41,42]	678/1670	Additive model	1.125(0.918–1.379)	.257	0.982(0.716–1.346)	.908	0.045	49.4%
NPDR	8	[10,11,19,23,28,37,41,42]	1358/1670		1.066(0.905–1.255)	.445	1.065(0.866–1.311)	.549	0.208	26.5%
PDR	8	[10,11,19,23,28,37,41,42]	678/1670	Homozygote model	0.611(0.392–0.951)	.029	0.947(0.432–2.076)	.891	0.042	50.2%
NPDR	8	[10,11,19,23,28,37,41,42]	1358/1670		1.123(0.809–1.559)	.487	1.134(0.810–1.588)	.463	0.667	0.0%
PDR	8	[10,11,19,23,28,37,41,42]	678/1670	Heterozygote model	0.848(0.690–1.042)	.116	1.017(0.710–1.456)	.928	0.011	59.8%
NPDR	8	[10,11,19,23,28,37,41,42]	1358/1670		0.953(0.806–1.127)	.572	0.949(0.765–1.177)	.634	0.193	28.2%

Note: PDR: Proliferative diabetic retinopathy; NPDR: Non-proliferative diabetic retinopathy; N: number of studies; OR: odd ratio; 95%CI: 95% confidence interval.

Assessment of sensitivity analysis and potential publication biases

Concerning the study quality assessment, each of the eligible studies was appraised with 7 or 8 stars, suggesting that all 19 studies were of high quality. Regarding the HWE fit test, the control groups in three studies were not satisfied. Consequently, in the sensitivity analysis, the above-mentioned three studies were omitted and the pooled ORs were recalculated. The results of the final analysis did not materially change. We then deleted the included studies one at a time and summarized the OR values. The results did not alter significantly (Supplementary Material, Figure S2). Furthermore, Begg’s funnel plot was essentially symmetrical (Figure 4), and Egger’s test indicated that the p value for the overall study was larger than .05 (Table 3). This indicates that publication bias is unlikely to influence the research results. In conclusion, the results of our study are statistically reliable and stable.

Figure 4. Begg’s funnel plots between diabetic retinopathy risk and eNOS 4b/a polymorphisms for allelic model (a), recessive model (b), dominant model (c), additive model (d), homozygote model (e) and heterozygote model (f) in the overall study. The Middle solid vertical line represents the logarithm of the overall summary or value, and the two slanted dotted lines show the limit of the 95% confidence interval. Each point signals an independent study.

Discussion

DR, an eye disease that causes vision loss in adults, has become a global public health concern [43]. eNOS is distributed in vascular endothelial cells and is a pivotal enzyme that mediates the production of NO, a potent vasodilator [44]. The eNOS gene has been found to contain several polymorphisms; the 4a/b polymorphism may affect NO levels by interfering with the shearing efficiency of mRNA during transcription or by interlocking with other functional sites, thus affecting the occurrence of diabetic retinopathy [45]. Significantly higher plasma NO levels in patients with DR can exacerbate endothelial dysfunction, endangering the retinal vascular system by increasing vascular permeability, oxidative stress and leukocyte proliferation [9,30,46]. Endothelial dysfunction of the small retinal arteries and accelerated progression of retinopathy were observed in mice with the eNOS gene knockout, indicating that eNOS dysfunction is a significant contributor to the pathogenesis of DR [47]. Furthermore, previous studies have demonstrated that eNOS dysfunction is intimately correlated with diabetes progression and that maintaining appropriate eNOS function is critical for alleviating diabetic vasculopathy [48,49]. Therefore, it is essential to explore whether eNOS gene polymorphisms affect the risk of incident DR. Several original articles and meta-analyses on the association between eNOS 4a/b gene polymorphisms and DR susceptibility have been published, but the results were discordant and highly contentious. Consequently, to further understand the role of eNOS 4a/b gene polymorphisms in DR, we performed a systematic review and updated the meta-analysis of all the original papers retrieved in this field.

The current study comprised 19 studies (7838 subjects), all of which had an NOS score of 7 or 8. The results showed no significant risk association of the eNOS4a/b polymorphism with the development of DR in the overall population, which was consistent with previous research by Ma [31], Yu [29] and Dong [30]. Subgroup analysis by ethnicity revealed that the 4a allele of the 4a/b polymorphism in the eNOS gene was a risk factor for the occurrence of DR in Caucasians, suggesting that 4a may be used as a biomarker for early prediction and screening of DR in Caucasian patients with T2DM. This was inconsistent with the results of Zhao, Ma, Dong and Yu’s meta-analysis. For the two included studies on Tunisians, although their results reported that the eNOS 4a/b polymorphism was significantly associated with the occurrence of DR in Tunisian patients with type 2 diabetes, the subgroup analysis of this study showed no relevance between the eNOS 4a/b gene polymorphism and DR in Tunisians. This could be explained by the fact that the sample sizes for the two studies differed considerably and that the study of Midani et al. [24] deviated from HWE, which caused significant heterogeneity in the subgroup analysis. Therefore, the risk interaction between the eNOS 4a/b polymorphism and DR in the Tunisian population could not be determined. In addition, no significant association between eNOS 4a/b polymorphisms and DR was observed in Asians, which is consistent with the findings of Ma, Yu and Dong. There was no statistically significant difference in the subgroup analysis of Han Chinese, suggesting that the eNOS 4a/b polymorphism may contribute little to the occurrence of DR in Chinese patients with type 2 diabetes. A single African study was included in our meta-analysis [18]. Although this study noted that 4a was a protective factor against the occurrence of DR in African populations, there were not enough data to conduct a subgroup analysis. Therefore, the relationship between eNOS 4a/b polymorphism and DR in Africans cannot be explained in the present study. In case-control studies, controls from PB are well represented, selection bias is small, and bias from differences in residential settings can be reduced. Our meta-analysis showed that the eNOS 4a/b polymorphism is associated with the risk of incident DR in PB studies. In subsequent studies, population controls were used to validate our results. Moreover, no association with eNOS 4a/b gene polymorphism was observed in either the PDR or NPDR groups. This may be related in part to the more significant heterogeneity of the subgroup analysis and publication bias.

In 2012, Zhao et al. undertook a meta-analysis to explore the potential role of eNOS 4a/b polymorphisms in DR susceptibility and concluded that the 4a allele is a protective factor affecting the development of DR. The 2014 meta-analyses of the eNOS 4a/b polymorphism and DR by Ma et al. and Yu et al. and the 2018 meta-analysis of the eNOS gene versus diabetes and its vascular complications by Dong et al. showed no risk association between eNOS 4a/b polymorphisms and DR in the overall population, which is consistent with the results of our study. However, in a subgroup study of ethnographic stratification, Yu et al. argued that the eNOS 4a/b polymorphism may reduce the danger of occurring DR in African patients with type 2 diabetes. There are four possible explanations for the differences between current and previous meta-analyses. Firstly, the original study on the association of eNOS 4a/b polymorphisms with DR in patients with T2DM has been updated to 2021; therefore, our sample sizes were larger, and the included studies were more comprehensive. Secondly, none of the previous studies fully analysed the allelic model and the other five gene models (recessive, dominant, additive, homozygote and heterozygote models); however, we analysed these six models in both subgroup analyses and the overall population. Thirdly, the studies by Zhao, Ma, Yu and Dong all differed in their classification of Tunisian ethnicity, with Zhao classifying them as mixed, Ma classifying them as Asian, Yu classifying them as African and Dong classifying them as Caucasian, which caused inconsistencies in the results of the ethnically based subgroup analysis. After reviewing the relevant literature [50–54], we determined that there was no precise criterion for the classification of Tunisian ethnicity; thus, we conducted a separate subgroup analysis of the two Tunisian studies included in this study. Lastly, Zhao, Ma and Yu reported differences in data extraction for the control group in the original study by Mehrab-Mohseni et al. [8] Zhao extracted data on patients with type 2 diabetes without DR but with diabetic nephropathy, diabetic neuropathy, or without microvascular complications as a control group. Yu extracted data from the T2DM subject group as the control, while the control group data extracted by Ma did not match the data of any group from the original study. Based on the selection criteria for the control group in the case-control study, we concluded that the control group in this original study should be the diabetic group without any complications; therefore, we extracted the complication-free group as the control group. In summary, we performed a systematic review and updated meta-analysis that included more comprehensive original studies, analysis of genetic models and a more scientific classification of races and data extraction to compensate for the shortcomings of previous studies to further explore the risk association between eNOS 4a/b and DR in T2DM patients.

Heterogeneity was observed in the overall study and subgroup analyses of some genetic models. Meta-regression analysis suggested that different ethnic groups, control group origin and type of diabetic retinopathy were not the main sources of heterogeneity in this study. As DR is a complex disease caused by the interaction of multiple factors, we speculated that it might also be related to the living environment, lifestyle and family history of diabetes. Assessment of publication bias was performed using Begg’s funnel plots and Egger’s test. The funnel plots of the overall study were symmetrical and Egger’s test p > .05, indicating that the statistical results of our updated meta-analysis were credible. Furthermore, the outcome of the sensitivity analysis demonstrated that after removing any study, the change in the pooled OR value was not significant, implying that the findings of this study were dependable and stable.

In addition to the inherent limitations of individual studies, our analysis has several limitations. Firstly, although we gathered all relevant original papers, there were few studies on Tunisians and Africans. Secondly, since there was no research in any other language, except English and Chinese, other valid results could be omitted from this study. Third, the genetic and pathogenetic mechanisms of DR are complex and unclear and gene–gene interactions, gene–environment interactions, and the lifestyle of the subjects may all have an impact on the results of this study. Therefore, the association between eNOS 4a/b polymorphism and the incidence of DR still needs to be confirmed by studies in different human populations and rigorous study designs.

Conclusions

In conclusion, the results of our updated systematic review and meta-analysis revealed evidence at the 4a allele of the eNOS4a/b gene probably increases the risk incidence of DR in Caucasian T2DM patients. Furthermore, studies based on population controls showed that the occurrence of DR may be related to the eNOS4a/b polymorphism. Further research is still required to verify our findings and explore the biological significance of eNOS 4a/b in type 2 DR patients of various ethnicities.

Ethics statement

All of our analyses were based on published studies and did not include the original overview of the clinical trials. Therefore, ethical approval and informed consent are not required.

Author contributions

Yaling Ma designed and conceived the study. Xin Fan and Yushan Shi searched the database and extracted the data. Kaiyun Zhang, Xin Fan and Yushan Shi analysed the data. Yushan Shi and Xin Fan drafted the manuscript of the paper. Kaiyun Zhang and Yaling Ma revised the manuscript. All the authors approved the final manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

All data analysed and generated by the Institute are included in this document or supplementary materials and are available from the corresponding author upon reasonable request.

Correction Statement

This article was originally published with errors, which have now been corrected in the online version. Please see Correction (http://dx.doi.org/10.1080/07853890.2023.2245611)

Additional information

Funding

This work was supported by the Natural Science Foundation of Ningxia Province [No. 2022AAC03474].