2024 Volume 71 Issue 9 Pages 839-849
This umbrella review was conducted aiming to assess the association between genetic variations and the development of diabetic retinopathy (DR) by collecting and evaluating available systematic reviews and meta-analysis results. We evaluated the methodological quality using the Measurement Tool to Assess Systematic Reviews (AMSTAR) 2.0, estimated the summary effect size by using the random effects model and calculated the 95% prediction intervals (PIs). Evidence from the included meta-analyses was graded according to established criteria as follows: convincing, highly suggestive, suggestive, weak, or not significant. This umbrella review included 32 meta-analyses of 52 candidate SNPs. The 12 selected meta-analyses were rated as “high,” 2 studies were rated as “moderate,” 11 studies were graded as “low,” and the remaining 7 studies were graded as “critically low” in terms of methodological quality. Carriers of specific genotypes and alleles of the transcription Factor 7-like 2 C/T (TCF7L2 C/T) polymorphism (rs7903146, p < 0.001) might be more susceptible to the occurrence of DR in the homozygous and recessive models, and these associations were supported by “convincing” evidence. Significant associations were also found between interleukin-6 (IL-6) -174 G/C (rs1800795; p < 0.05) or vascular endothelial growth factor (VEGF) polymorphisms (rs2010963, rs699947, rs1570360, rs2010963, rs699947, rs2146323; all p values <0.05) and DR risk, but these associations were supported by “weak” evidence. The TCF7L2 C/T variant could be identified as a definitive genetic risk factor for the development and progression of DR. Data from additional in-depth studies are needed to establish robust evidence for the associations between polymorphisms of IL-6 or VEGF and DR.
Diabetic retinopathy (DR), one of the complications of diabetes mellitus (DM), is a microvascular disease that causes damage to retinal capillaries and secondary visual impairment and is thus mainly characterized by persistent retinal lesions due to glucose metabolism dysfunction [1]. DR causes severe damage to visual function, such as blurred and decreased vision, and is a major cause of sight loss worldwide [2]. As the population continues to age and obesity rates continue to increase sharply, the global prevalence of DM, as well as DR, is also increasing. The global prevalence of DR reached 103 billion in 2020, and this number is projected to increase to 161 million by 2045 [3]. According to the International Diabetes Federation in 2020, which reported the prevalence of diabetic retinopathy (DR) in all regions of the world, DR is prevalent among individuals with diabetes in Africa (35.9%), North America and the Caribbean (33.3%), and the Middle East and North Africa (32.9%) [3]. Research has indicated that approximately 9.60 million people in the US were living with DR in 2021, accounting for 26.4% of individuals with DM [4], and prevalence among the Asian population was approximately 28.0% [5].
The pathological mechanisms of DR are complex and multifactorial. A recent systematic review showed that neuroinflammation and degeneration, vascular degeneration, and glial activation can be regarded as the major aetiologies of DR [6]. Oxidative stress can also contribute to the occurrence of DR by increasing the flux of the polyol pathway and hexosamine pathway, hyperactivating protein kinase C (PKC) isoforms, stimulating the accumulation of advanced glycation end products (AGEs), and finally resulting in hyperglycaemia, which may further trigger the accumulation of ROS and damage structural and functional alterations in the retina [7]. Inflammatory factors and vascular endothelial growth factor (VEGF) may also be involved in the development of DR [8]. In addition, epigenetic modifications, mainly DNA methylation, histone posttranslational modifications and noncoding RNAs, can modulate the expression of specific genes, and genetic variations play a special role in the development of DR by modulating inflammatory and immunological responses [9]. However, the exact mechanism of interaction between heredity and environmental risk factors in DR susceptibility remains poorly elucidated.
To date, by qualitative and/or quantitative data synthesis, numerous recent meta-analyses have been conducted to assess the association of DR risk with genetic variants, such as polymorphisms of VEGF [10], tumour necrosis factor-α (TNF-α) [11], interleukin-1 (IL-1), IL-18 [11], and IL-6 [12]. However, most of these studies have focused on single nucleotide polymorphisms (SNPs) of candidate genes that might be related to the occurrence of DR, and the results of these meta-analyses are somewhat elusive. Moreover, the associations identified by a single meta-analysis may not be accurate due to the influence of heterogeneity and systematic bias. An umbrella review will be helpful in distinguishing consistent from contradictory findings and elucidating the strength of existing evidence from published systematic reviews and meta-analyses on a specific research topic [13].
As such, we conducted an umbrella review study of all the published systematic reviews or meta-analyses investigating associations between genetic variations and DR risk with the aims of comprehensively synthesizing the findings and accurately assessing the strength of evidence that has been previously reported in meta-analyses.
An umbrella review is commonly used to collect and evaluate information from available systematic reviews and meta-analyses and then propose recommendations for practice or research. This umbrella review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and Meta-analyses of Observational Studies in Epidemiology (MOOSE) guidelines. The protocol of the current umbrella review has been registered with the International Prospective Register of Systematic Reviews (PROSPERO) database 2020 under the registration number CRD42023449493.
2.1 Literature search strategy and eligibility criteriaThe eligible studies were selected from PubMed, EMBASE and the Cochrane Library of Systematic Reviews for the collection of diverse systematic meta-analyses and reviews from inception until August 17, 2023 without language restrictions. The search strategy and free terms are displayed in Supporting Information Table S1. After excluding duplicates, two reviewers (HSF and FYH) independently retrieved and abstracted the full texts of potentially eligible articles. The results were limited to systematic reviews and meta-analyses with a search filter, and the reference lists of eligible studies were also searched for additional citations. We first independently screened the titles and abstracts of each potential study and then conducted full-text screening for potential inclusion. Any discrepancies were resolved through discussion with the rest of the team.
The eligible studies included meta-analyses of four types: i) single nucleotide polymorphism (SNP)-based systematic reviews and gene meta-analyses that evaluated observational studies investigating the association between SNPs and DR risk; ii) case–control studies or genome–wide association studies (GWASs) that reported the number of cases and controls and cohort studies that described the number of cases and participants; and iii) meta-analyses that provided either effect estimates of individual studies or definite information on the calculation of these estimates. We excluded systematic reviews or meta-analyses if i) they were cross-sectional studies, letters to editors, narrative reviews without a formal search of the literature, or conference abstracts; ii) they included studies whose subjects were not human; or iii) they did not provide information about the search strategy, inclusion criteria, or quality assessment methods. To avoid overlap of studies, for reviews or meta-analyses that used a similar study design and shed light on the same SNPs, only the most published study with the largest number of included studies was retained.
2.2 Data extraction and quality assessmentTwo reviewers independently carried out the data extraction: one investigator (FYH) extracted the data, which were then checked by a second investigator (SY). The following characteristics were extracted from each eligible study: first author name, publication year, study design, number of included studies, number of participants, gene name, number of examined SNPs, sample size (DR case/control), genotype and allele counts, study-specific risk estimates (odds ratios [2]), corresponding 95% confidence intervals (CIs), p values and I2 for heterogeneity.
The methodological quality of the selected reviews was assessed using the Measurement Tool to Assess Systematic Reviews version 2 (AMSTAR-2) [14], which comprises 16 items with great interrater reliability and usability. Based on the rules of the AMSTAR-2, the overall confidence of the included reviews was rated as “high,” “moderate,” “low” or “critically low” according to the number of critical flaws or noncritical weaknesses of the items. Among the critical domains are i) registering the protocol before beginning the review; ii) conducting a comprehensive literature search; iii) justifying the exclusion of individual studies; iv) adequately assessing the risk of bias in the studies included in the review; v) performing a meta-analysis using appropriate statistical methods; vi) accounting for potential bias when interpreting the results; and vii) evaluating the presence and effects of publication bias.
2.3 Data synthesis and analysis 2.3.1 Assessment of pooled effects and heterogeneityThe summary effect size and its CI were measured by using the random-effects DerSimonian and Laird (DL) method [15]. The prediction intervals (PIs) and 95% CIs were calculated to assess the between-study effect and certainty of the association if a new study addressed that same association, with a statistically significant p value threshold of <0.05 for the pooled effect estimates. Using the I2 statistic, we detected heterogeneity between studies by determining the proportion of total variation that was not caused by chance, with an I2 value greater than 50% or 75% considered to represent large or very large heterogeneity, respectively.
2.3.2 Evaluation of the quality of evidenceEgger’s regression asymmetry test was used to test for potential publication bias, with an Egger’s p value of <0.10 suggesting the presence of publication bias. The combination of Egger’s p < 0.10 with the effect size of the largest study of the meta-analysis being lower than the random effect estimates was deemed sufficient to support the findings of small-study analyses. An excess significance test was carried out to evaluate the extent to which there was an overabundance of formally significant findings in the published literature (p < 0.05) [16]. The greater the difference between the observed (O) and expected (E) number of studies, the greater the degree of excess significance. The significance threshold of the excess significance bias was set at p < 0.10, in combination with O > E.
The credibility of the evidence was determined based on the following criteria [17]: 1) a statistically significant p value of less than 10–6; 2) >1,000 cases; 3) the largest component study reported a significant result (p < 0.05); 4) no large heterogeneity (I2 < 50%); 5) 95% prediction interval that excluded the null value; 6) null result of small-study effect (p > 0.10); and 7) nonsignificant result of excess significance bias (p > 0.10). We then classified the evidence from the reviews or meta-analyses of observational studies into 4 classes: convincing (class I evidence met all criteria), highly suggestive (class II evidence met 1 to 3 of these criteria), suggestive (class III evidence met criteria 1 or 2), weak (class IV evidence met criteria 3), and no significant evidence (p value ≥0.05 under random effects).
These parameters were processed based on five commonly used genetic models: allelic model (mutant allele vs. wild-type allele), dominant model (mutant homozygote + heterozygote vs. wild-type homozygote), heterozygote model (heterozygote vs. wild-type homozygote), homozygote model (mutant homozygote vs. wild-type homozygote), and recessive model (mutant homozygote vs. wild-type homozygote + heterozygote). All the statistical analyses were performed using STATA 13.0 (Stata Corp LP, College Station, TX, USA).
After a systematic literature search, a total of 567 records were retrieved. After screening the titles and abstracts, duplicates and irrelevant publications were removed. A total of 32 meta-analyses fully met our criteria and were included in the present umbrella review. Fig. 1 summarizes the study selection process and the reasons that meta-analyses were included or excluded.
The flow chart of literature search and study selection
The characteristics of the included meta-analyses are detailed in Table 1. All the selected meta-analyses were published between 2011 and 2022 and were conducted based on case-control studies, and the pooled effect was measured by calculating the odds ratios (ORs). The number of included studies in each meta-analysis ranged from 2 to 40, and the corresponding number of total participants ranged from 346 to 10,168. These studies considered the associations between 52 SNPs in 24 types of candidate genes and DR risk. Of these studies, 12 were related to VEGF [10, 18] and 10 to IL (including IL-6: 4 studies; IL-10: n = 3; IL-4 and IL-8: n = 2) [12, 19-21]; four each of receptor for advanced glycation end products (RAGE) [20, 21], TNF [17, 22, 23], and vitamin D receptor (VDR) [24]; three each of aldose reductase (AR, AKR1B1 and ALR) [25-27], erythropoietin (EPO) [28], nitric oxide synthase 3 (NOS3) [29], and transforming growth factor-β ( TGF-β) [17, 30]; two each of serine protease inhibitor-1 (SERPINE1) or plasminogen activator inhibitor-1 (PAI-1) [31, 32], peroxisome proliferator-activated receptor γ2 (PPAR-γ2) [33] and uncoupling proteins (UCP) [34]; and the remaining genes were all from 1 study.
Main characteristics of included systematic reviews or meta-analyses that focused on the role of gene polymorphisms in diabetic retinopathy
| Authors | Year | Study design | No. of included studies (total participants) | Gene category | Effect metrics | AMSTAR 2 |
|---|---|---|---|---|---|---|
| Zhang Y | 2022 | Case-control study | 5 (9,860) | TCF7L2 rs7903146 C/T | OR | High |
| Zhang YQ | 2022 | Case-control study | 11(5,903) | TLR4 Asp299Gly | OR | High |
| Sesti LFC | 2022 | Case-control study | 11(6,791) | EPO rs1617640 | OR | Critically low |
| 8 (4,856) | EPO rs507392 | |||||
| 8 (4,864) | EPO rs551238 | |||||
| Chen J | 2021 | Case-control study | 10 (2,898) | SERPINE1 rs1799889 | OR | High |
| Jafarzadeh F | 2022 | Case-control study | 2 (595) | IFN-γ rs2430561 | OR | Critically low |
| 3 (1,333) | TGF-β rs1800469 | |||||
| 5 (2,096) | IL10 rs1800896 | |||||
| 3 (734) | IL-6 rs1800795 | |||||
| 4 (1,997) | TNFα rs1800629 | |||||
| 2 (432) | IL-4 rs2243250 | |||||
| 2 (1,427) | IL-8 rs4073 | |||||
| Liu X | 2021 | Case-control study | 5 (5,337) | UCP1 rs1800592 | OR | Low |
| 5 (4,488) | UCP2 rs659366 | |||||
| Li XF | 2021 | Case-control study | 9 (5,113) | PPAR-γ2 rs1801282 | OR | High |
| 3 (1,757) | PPAR-γ2 rs3856806 | |||||
| Dastgheib SA | 2020 | Case-control study | 11 (2,942) | PAI-1 4G5G | OR | Critically low |
| Hu L (NPDR) | 2021 | Case-control study | 9 (2,316) | VEGF rs2010963 | OR | Moderate |
| 5 (1,071) | VEGF rs3025039 | |||||
| 4 (1,415) | VEGF rs833061 | |||||
| 4 (1,486) | VEGF rs1570360 | |||||
| 2 (346) | VEGF rs699947 | |||||
| Hu L (PDR) | 12 (3,557) | VEGF rs2010963 | ||||
| 5 (1,009) | VEGF rs3025039 | |||||
| 5 (1,852) | VEGF rs833061 | |||||
| 3 (1,605) | VEGF rs1570360 | |||||
| 2 (931) | VEGF rs13207351 | |||||
| 3 (866) | VEGF rs699947 | |||||
| Gao W | 2021 | Case-control study | 9 (4,281) | TNF α -308 G/A | OR | High |
| 2 (1,188) | TNF α -238 G/A | |||||
| Ulhaq ZS | 2020 | Case-control study | 3 (1,121) | IL6 -174 G/C | OR | Low |
| Liu M | 2020 | Case-control study | 7 (2,948) | TNF α -308G/A | OR | Critically low |
| Sun X | 2020 | Case-control study | 2 (915) | IL6 rs1800795 | OR | Critically low |
| 3 (865) | IL6 rs1800796 | |||||
| 4 (2,364) | IL10 rs1800896 | |||||
| Lin S | 2020 | Case-control study | 19 (8,752) | AR-C (-106) T | OR | Moderate |
| Song N | 2019 | Case-control study | 4 (1,001) | VDR ApaI | OR | Low |
| 5 (625) | VDR BsmI | |||||
| 6 (1,413) | VDR FokI | |||||
| 2 (581) | VDR TaqI | |||||
| Xie Z | 2018 | Case-control study | 9 (3,192) | ICAM-1 rs5498 | OR | High |
| Shu Y | 2018 | Case-control study | 2 (1,340) | IL10 rs1800896 | OR | Low |
| Cao M | 2018 | Case-control study | 18 (8,364) | AKR1B1 rs759853 | OR | High |
| Tao D | 2017 | Case-control study | 5 (1,790) | RAGE -374T/A | OR | Low |
| Huang L | 2017 | Case-control study | 8 (2,286) | MnSOD Val16Ala | OR | High |
| Li GY | 2017 | Case-control study | 4 (5,004) | ALDH2 rs671 | OR | Low |
| Luo S | 2016 | Case-control study | 40 (10,168) | ACE I/D | OR | High |
| Yu W | 2016 | Case-control study | 5 (3,120) | RAGE -429T/C | OR | High |
| 6 (2,866) | RAGE -374T/A | |||||
| 6 (3,436) | RAGE Gly82Ser | |||||
| Wang W | 2016 | Case-control study | 4 (6,118) | MCP-1 2518 A/G | OR | High |
| Song ZD | 2016 | Case-control study | 5 (1,057) | ALR C-106T | OR | Low |
| Zeng Y | 2015 | Case-control study | 4 (1,307) | VEGF rs2146323 | OR | Low |
| Gong JY | 2015 | Case-control study | 5 (1,078) | ITGA2 BgII | OR | Low |
| Ma ZJ | 2014 | Case-control study | 15 (6,593) | eNOS -4b/a | OR | High |
| Liu L | 2014 | Case-control study | 2 (608) | TGF-beta1 +869T/C(L10P) | OR | Critically low |
| 3 (1,101) | TGF-beta1 -509 C/T | |||||
| Wang J | 2013 | Case-control study | 2 (451) | PON1 Q192R | OR | Critically low |
| Zhao S | 2012 | Case-control study | 8 (3,084) | NOS3 4b/a | OR | Low |
| 4 (1,898) | NOS3 T-786C | |||||
| 7 (2,819) | NOS3 G894T | |||||
| Tian C | 2011 | Case-control study | 3 (994) | SOD2 C47T (rs4880) | OR | Low |
Abbreviations: OR, odd ratio; AMSTAR 2, A MeaSurement Tool to Assess systematic Reviews-2.
A random effects model was used to reanalyse these 299 associations in the 32 included meta-analyses. The summary random effects model estimates ranged from 0.20 to 3.11. Using p < 0.05 as the threshold for statistical significance, 63 (21.1%) random effects estimates reached significance. Of these significant associations, 35 (11.7%) showed an increased OR between various gene polymorphisms and DR risk, whereas 28 (9.4%) indicated a decreased risk for DR (Fig. 2).
Forest plot and assessment of cumulative evidence of associations between genetic variants and risk of diabetic retinopathy
Twenty-seven (72%) had an I2 statistic >50% (Fig. 2). There was moderate to high heterogeneity (I2 = 50%–75%) in 22 meta-analyses and high heterogeneity (I2 > 75%) in 5 meta-analyses. However, the I2 statistics of 299 studies revealed that among 12 genes, including TCF7L2, EPO, IL-8, UCP2, VEGF, TNFα, IL-10, VDR, ICAM-1, RAGE, TGF-β1 and PON1, 157 genes (52.5%) exhibited low heterogeneity (I2 < 50%), 88 genes (12 of them, 29.4%) exhibited considerable heterogeneity (I2 = 50–75%), and 50 genes (16.7%) exhibited considerable heterogeneity (I2 > 75%). To further demonstrate heterogeneity among studies, we further calculated 95% PIs and found 59 (19.7%) associations excluding null results (Table S2).
3.4 Small-Study Effects and excess significance biasAn Egger’s test p value <0.10 was found in four meta-analyses, including the rs1800796 polymorphism of IL-6 (dominant model), the FokI polymorphism of VDR (recessive model), the rs2010963 polymorphism of VEGF (dominant model), and the Bg II polymorphism of ITGA2 (recessive model) (Table S2). Nevertheless, none of the 32 meta-analyses had associations that met the criteria for a small-study effect under the five genetic models (Egger’s test p < 0.10 and the effect size of the largest study estimate was smaller than the random effect size). Thirty-three (11.04%) associations had evidence of excess significance bias for several genes, including EPO, UCP2, VEGF, AR, VDR, ICAM-1, IL-10, AKR1B1, MnSOD, ACE, RAGE, MCP-1, ALR, ITGA2, eNOS, and TGF-β1 (O > E and p for significant study <0.1) (Table S2).
3.5 Assessment of methodological quality and certainty of evidenceAll items of the AMSTAR 2.0 tool were assessed for each meta-analysis. The 12 selected meta-analyses were rated as “high” methodological quality, and 11 studies were graded as “low” methodological quality; another 2 studies were rated as “moderate” methodological quality, and the methodological quality of 7 studies was graded as “critically low.” Among the seven critical items of the AMSTAR 2.0, in addition to items 4, 7 and 11, item 2 was missing in 11 studies (34.4%), item 9 was missing in 1 study (3.1%), item 13 was missing in 8 studies (25.0%), and item 15 was missing in 5 studies (15.6%). The details of the quality assessment can be found in Table 1 and Table S3.
All included meta-analyses explored the associations between different SNPs and DR risk under five different genetic models: allelic model, dominant model, heterozygous model, homozygous model, and recessive model. According to the criteria for the level of evidence, we identified 299 associations in these 32 meta-analyses (Table S2). We found that only two associations presented convincing evidence and indicated a positive association between DR risk and the rs7903146 polymorphism of TCF7L2 (homozygous model: OR = 2.09 (1.62, 2.70), p < 0.001; recessive model: OR = 2.11 (1.75, 2.56), p < 0.001) [22], and 68 associations were supported by weak evidence.
Based on the prior and observed scores, 299 associations were investigated as genetic factors for DR. The study of rs7903146 C/T in TCF7L2 was graded as high quality, and the homozygous and recessive models of TCF7L2 (rs7903146) were both supported by convincing evidence for a contribution to the pathogenesis of DR. However, the association of the IL6 gene -174 G/C polymorphism with DR was supported by only weak evidence in the heterozygote model, and its methodological quality was rated as “Critically low.” The presence of 13 genetic mutations in seven genes regarded as potential risk factors for DR was supported by weak evidence. In contrast, nine genetic variants of the corresponding nine genes also showed weak evidence for a negative association with DR.
As presented in Fig. 2, prior meta-analyses have provided supporting evidence regarding the positive association between genetic polymorphisms and DR risk. For instance, a study by Hu et al. [10] suggested that the rs2010963 polymorphism of VEGF was significantly associated with an increased risk of nonproliferative DR in the Asian population (dominant model: OR = 1.29 [95% CI, 1.04–1.60]) and an increased risk of proliferative DR in the whole population (dominant model: OR = 1.20 [95% CI, 1.03–1.41]). They also found that the rs833061 and rs699947 polymorphisms of VEGF were both associated with increased proliferative DR risk in the Asian population, whereas the rs699947 polymorphism was only associated with increased nonproliferative DR in the whole population [10]. A study by Zeng et al. [16] consistently confirmed the positive association between the VEGF rs2146323 polymorphism and DR risk (dominant model: OR = 1.38 [95% CI, 1.10–1.72]). Similarly, the study by Jafarzadeh et al. [17] showed “weak” evidence of rs1800795 in the IL-6 gene in individuals with DR (homozygous model: OR = 1.68 [95% CI, 1.01–2.80]; recessive model: OR = 1.66 [95% CI, 1.14–2.43]). In addition, 229 associations presented no significant associations. The grading details of the evidence are shown in Table S2.
TCF7L2 is a member of the T-cell factor/lymphoid enhancer binding factor family (TCF/LEF), and it is one of the most frequently-investigated genes because of its strong genetic association with type 2 diabetes mellitus (T2DM) [23]. This gene not only plays fundamental developmental and metabolic roles in adipose tissue, but also regulates adipocyte size, adipocyte endocrine function, and glucose metabolism [23, 24]. Notably, the rs7903146-T allele is one of the TCF7L2 variants with the greatest risk for T2DM [25]. TCF7L2 gene variants may confer a genetic risk for T2DM and its complications. A meta-analysis based on GWAS confirmed the positive association between rs7903146 (TCF7L2) and DR risk (OR = 1.30, p < 0.001) [26]. A meta-analysis by Zhang et al. involving 12,982 individuals from 6 original studies demonstrated the contribution of the rs7903146 C/T polymorphism of TCF7L2 to the development of DR, particularly among the Caucasian population [22]. However, the functional significance and role of rare mutations in the TCF7L2 gene in DR pathophysiology in diverse ethnic groups are unknown. A study by Gambio et al. reported a correlation between rs7903146-T in TCF7L2 and hyperglycaemia and B-cell dysfunction [27]. VEGF is a target gene of TCF7L2 [28], which is a key component of the Wnt signalling pathway [29]. TCF7L2 mutation may trigger pathological retinal neovascularization through endoplasmic reticulum stress-dependent VEGF upregulation [29]. Furthermore, the contribution of TCF7L2 variability to DR risk may vary across genetic models. For instance, in the study by Ciccacci et al., positive associations between variant genotypes of rs7903146 in TCF7L2 and DR risk were found in heterozygous and homozygous models [30]. This additive effect might be partly explained by the fact that the heterozygosity of rs7903146 confers risk to the pancreatic islets by cis regulating and modifying local chromatin structure, and the variant alters enhancer activity [31]. In addition, individuals with two variant alleles may suffer from a greater risk of DR than heterozygotes. In this study, we further reanalyzed and categorized the strength of evidence and found that the homozygous and recessive models of the rs7903146 polymorphism of TCF7L2 associated with increased DR risk were supported by convincing evidence with a p value <0.001, but this association was supported by only weak evidence under the allelic model with high between-study heterogeneity (81.9%). Thus, it is supposed that the rs7903146 polymorphism of TCF7L2 may confer DR risk.
IL-6 is a multifunctional pleiotropic cytokine that can be involved in many inflammatory responses stimulated by different types of stimuli, particularly responses related to insulin-resistant states [32]. A recent study in India revealed that the average serum IL-6 concentration was significantly greater in the DR group than in the control group without DR (p < 0.001), and they also identified a significant correlation between the serum IL-6 concentration and the severity of DR (p = 0.001) [33]. Regarding the regulation of IL-6 gene expression, polymorphisms resulting in approximately 150 SNPs have been detected in both the 5' and 3' flanking regions of the IL-6 gene. In particular, the polymorphism -174G/C, located in the upstream regulatory region of the gene, is an upstream regulatory polymorphism that affects the production of IL-6. IL-6 polymorphisms may affect the outcomes of several diseases, and a previous study reported that the G→C polymorphism at position -174 of IL-6 may affect IL-6 transcription and be associated with systemic onset of insulin sensitivity [34]. A study including 422 patients suggested that patients with T2DM + proliferative DR (PDR) were more likely to carry a higher frequency of the IL-6 -174 C allele than those with only T2DM (OR = 1.35 [95% CI 1.03–1.78]) [35]. However, Weger et al. even reported that, compared with reference groups with genotypes of GG/GC, the prevalence of the CC genotype of IL-6 -174 was associated with a decreased risk of retinal artery occlusion (OR = 0.50 [95% CI 0.28–0.89]) [36], which is one of the common underlying mechanisms associated with DR [37]. According to the included meta-analyses, Ulhaq et al. [12] reported that the IL-6 -174 G/C polymorphism seems to be a potential genetic risk factor for PDR. In this umbrella review, -174G/C, which is a polymorphism of IL-6, only showed high significance in the heterozygous model. More importantly, this association was supported by only weak evidence.
In particular, the bioactivity of IL-6 could be affected by SNP mutations. We noticed that IL-6 (-174 G/C, rs1800795) and IL-6 (-572 C/G, rs1800796) showed conflicting association results, with the former being positively associated with DR susceptibility and the latter being negatively related. Similarly, a meta-analysis consistently revealed contradictory findings between two SNPs, with the “C” allele of IL-6 rs1800795 being positively associated with the risk of diabetic nephropathy (DN) and the “G” allele of IL-6 rs1800796 being negatively associated with the risk of DN [38]. However, recent evidence regarding the effects of these two IL-6 SNPs on diseases remains inconclusive. Although a meta-analysis provided evidence supporting the associations between the IL-6 rs1800795 polymorphism and the risk of multiple diseases, including coronary artery diseases, inflammatory diseases, ischaemic stroke and rheumatoid arthritis [39], another meta-analysis reported no association between the rs1800795 polymorphism in the IL-6 gene and the risk of microvascular complications of T2DM [40]. Hence, more high-quality studies with highly credible evidence are warranted to determine the effect of the IL-6 -174 G/C and 572 C/G polymorphisms on DR.
VEGF is a well-recognized angiogenic factor in the eye that is an antiapoptotic, endothelial cell survival factor, a vasodilator, and a promoter of the migration of endothelial cells [41]. Like the proinflammatory cytokines mentioned above, VEGF plays a critical role in the development of microvascular impairments and the degeneration of the blood-retinal barrier [8]. In addition to hypoxia and inflammatory mediators, the expression of VEGF within retinal microvascular endothelial cells is upregulated by high glucose-mediated oxidative stress [41, 42]. Moreover, the overexpression of VEGF could contribute to neovascularization, resulting in PDR in the advanced stage of DR [8]. Based on our findings, an elevated plasma VEGF level was observed in diabetic Han Chinese patients with the polymorphism -634 C/G (rs2010963) of VEGF, and these carriers may have a greater genetic risk of DR [43]. Another study also indicated that the A allele at rs699947 of VEGF was positively associated with the risk of DR (OR = 1.84 [95% CI 1.28–2.66]) [44]. Among the included meta-analyses, a positive association between the VEGF rs2146323 polymorphism and DR risk was also demonstrated in a meta-analysis by Zeng et al. (dominant OR = 1.38 [95% CI 1.10–1.72]; codominant OR = 1.37 [95% CI 1.08–1.74]) [18]. Another meta-analysis by Wang et al. [10] consistently supported the association between the VEGF-2578 C/A (rs699947) polymorphism and increased DR risk (ORs ranged from 1.26 to 1.67 across models). In the present study, we indeed found that four SNPs (rs2010963, rs699947, rs1570360, and rs2146323) were significantly associated with increased DR risk. However, the studies investigating the associations between VEGF polymorphisms and DR risk were of low critical quality according to the AMSTAR 2.0 and were supported by only weak or nonsignificant evidence.
There are several limitations in the present umbrella review. First, all selected meta-analyses for genetic polymorphisms were case–control studies, and genetic associations from original studies not reported in meta-analyses were not evaluated; thus, this is indicative of a greater inherent potential for bias. Second, the strength of the evidence was not assessed for most genetic associations due to the limited sample size. Third, we did not take the grey literature into consideration, which may have resulted in underestimates of the results. Finally, in the present study, our search found that almost all studies that did not perform meta-analysis were less than two original studies (such as rs6214 and rs5742632), and these SNPs are very fragmented and rarely reported in the original studies, so we believe that these genetic loci have little impact on our current comprehensive analysis. Therefore, this study did not include individual studies that did not undergo meta-analysis.
Our assessment maps the status of evidence on the associations between genetic mutations and DR risk. Notably, the findings of this study imply that the rs7903146 C/T polymorphism in TCF7L2 may be involved in the development of DR, and this association is supported by highly convincing evidence. However, considering the weak evidence and heterogeneity across studies, the associations between IL6 genetic variants (-174 G/C and -572 C/G) or VEGF polymorphisms and DR risk need to be further demonstrated by substantial evidence.
Not applicable.
Informed Consent StatementNot applicable.
Competing interestsNone of the authors have any potential conflicts of interest associated with this research.
Authors’ contributionsZ.F.F. designed the study. H.S.F. and F.Y.H. did the literature search and screening. F.Y.H. extracted data which were then checked by S.Y., H.S.F., F.Y.H. and S.Y. conducted the data analyses. H.S.F., F.Y.H., X.Y. and Z.F.F. made the figures and tables, and drafted manuscript. All authors participated in the interpretation of results, and critically revised the manuscript. All authors approved the final version of the manuscript for publication. X.Y. and Z.F.F. were the co-corresponding authors who took responsibility for the integrity of the data and the accuracy of the data analysis.
FundingThis study is supported by Basic Research Project of Shenzhen Science and Technology Plan, China (JCYJ20210324125004013 and JCYJ20210324125202006) and Guangdong Basic and Applied Basic Research Foundation (2021A1515012428).
Availability of data and materialsAll data included in this umbrella review were extracted from publicly available systematic reviews.
