Aqueous Humor Mediator and Cytokine Aberrations in Diabetic Retinopathy and Diabetic Macular Edema: A Systematic Review and Meta-Analysis

Purpose To evaluate the relationship between the aqueous humor levels of VEGF, TNF-α, IL-10, IL-6, IL-12, MCP-1, and IP-10 with DR/DME. Methods PubMed, Web of Science, Embase, China National Knowledge Infrastructure (CNKI), and Wanfang databases were searched up to October 2018. Systematic review and meta-analysis were conducted. Results 18 studies comprising 362 cases with DR (100 with DME) and 620 controls without DR were included in this meta-analysis. There was a significant association between VEGF levels in the aqueous humor and DR (standardized mean difference (SMD) 1.94 (95% CI 1.05-2.83)) and DME (1.07 (0.71, 1.42)). Furthermore, a significant correlation was observed between levels of IL-6 and DR (3.53 (0.37, 6.69)), and similarly correlation with DME (1.26 (0.30, 2.21)). The relationship between MCP-1 and DR and DME was significant, in which the SMD was (0.49 (0.09, 0.89)) and (1.49 (0.78, 2.20)), respectively. However, IL-12, IP-10, and TNF-α had no correlation with DR and DME, whereas there was a significant relationship between IL-8 and DME (1.68 (0.97, 2.40)). Conclusion Elevated levels of VEGF, IL-6, and MCP-1 in the aqueous humor were associated with the risk for the presence of DR, and levels of VEGF, IL-6, IL-8, and MCP-1 were associated with the risk of DME. Furthermore, these biomarkers may be used as potential predictors or therapeutic targets for DR/DME.


Introduction
DR is one of the most common microcomplications of diabetes and has become the major cause of decreased vision and blindness in adults aged 20-74 years [1]. Recently, the disturbance of inflammatory reaction may play an important role in the numerous researches of the complex pathogenesis of DR [2]. Previous studies have shown that deregulation of immune responses associated with diabetes can induce high expression of various mediators resulting in the development of DR [3]. Furthermore, analysis of intraocular humors obtained from DR patients has indicated that some of the mediators (cytokines or chemokines) may be responsible for the pathogenesis of DR. The immune and inflammatory factors, e.g., VEGF, TNF-α, IL-6, IL-8, and MCP-1, have been observed elevated in both aqueous humor and vitreous fluids in patients with DR [4,5]. Hence, the altered concentrations of various cytokines regarding these mechanisms may serve as important biomarkers to assess early detection or treatment of DR.
Diabetic macular edema (DME) is also one of the most serious causes of visual disability and blindness and affects approximately 14% of patients with diabetes [6]. The pathophysiology of DME is multifactorial, complicated, and partly still unknown. Several mediators, including VEGF, as well as other inflammatory cytokines, such as IL-6, IL-8, TNF-α, IP-10, and MCP-1, are among the possible responsible mechanisms or are involved in the development of DME [7][8][9]. However, the exact cytokines in the aqueous humor regarding the pathogenesis of DME remain unknown.
Taken together, the association between levels of inflammatory cytokines in the aqueous humor and DR is still controversial [10,11]. Moreover, there have been no systematic review and meta-analysis to evaluate the available evidence on the association of the aqueous humor levels of inflammatory cytokines with the risk of DR or DME. Herein, we perform this systematic review and meta-analysis to investigate the relationships between these biomolecule levels in the aqueous humor and the risk of DR and DME.

The Study Identification and Search Strategy.
We identified the relevant evidences of mediators or cytokines in the aqueous humor among patients with type 2 diabetes mellitus (T2DM) by systematically searching PubMed, Embase, Web of Science, China National Knowledge Infrastructure (CNKI), and Wanfang databases, from inception to October 2018. The following key words were used to perform searches in the databases mentioned above: (inflammation, inflammatory markers, inflammatory biomarkers, inflammatory mediators, inflammatory cytokines) or (vascular endothelial growth factor, VEGF, tumor necrosis factor, TNF-α, interleukin 6, IL-6, interleukin 8, IL-8, interleukin 10, IL-10, monocyte chemoattractant protein-1, MCP-1), and (diabetic retinopathy, retinopathy, DR) or (diabetic macular oedema, diabetic macular edema), and (aqueous humor, aqueous humour), without year restriction. Moreover, an extensive hand search was performed and the additional relevant studies were further reviewed in reference lists. The selection process was according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram [12].

Inclusion and Exclusion
Criteria. The inclusion criteria were as follows: (1) the study should report the correlation between the aqueous humor levels of mediators or cytokines (VEGF, TNF-α, IL-6, IL-8, IL-12, IP-10, and MCP-1) and DR/DME; (2) it should be written in English or Chinese with the full text available; (3) patients with any age, gender, region or race were considered.
Exclusion criteria were predefined as follows: (1) literature reviews, case reports, and cell lines or animal studies; (2) duplication: same studies came from different databases; (3) no sufficient data to perform meta-analysis; (4) no DM group or no health control group.

Data Extraction. Two reviewers (Lei Liu and Jingyang
Wu) independently reviewed each included study. The disagreements on eligibility during the reviewing were discussed and resolved by the third reviewer (Song Yue). The following data were extracted by three reviewers (Yifan Zhong, Kaibo Yang, and Song Yue): the first author, location, year of publication, number of subjects in DR, DME, and control groups, definition of DR, cytokine measurements, and name of    cytokines in detail. We defined that diabetic patients without retinopathy and/or matched healthy persons constituted the controls and patients with DR or DME were the cases.

Assessment of Methodology
Quality. The quality of the included studies was assessed using the Newcastle-Ottawa scale (NOS) [13]. A quality score of more than or equal to seven on the nine-point NOS was considered to be high quality for included studies.

Statistical
Analysis. The weighted standardized mean difference (SMD) was measured with 95% confidence interval (CI) using random-effects method. The subgroup analyses were conducted based on location, DR definition, and cytokine measurements. I 2 tests were used to evaluate the statistical heterogeneity among the included studies, and the heterogeneity was considered statistical when P < 0:05 and I 2 ≥ 50%. Additionally, sensitive analysis was also performed to evaluate the influences of individual studies on the final effect. The funnel plots were performed to evaluate the potential publication bias. This meta-analysis was performed using the Stata 11.0 statistical software (Stata Corp., College Station, TX). A two-tailed P less than 0.05 was considered as significant difference.

3.2.
Characteristics of the Studies. 18 case-control studies involving 362 cases with DR including 100 cases with DME and 620 controls with type 2 diabetes but without DR were included in the current meta-analysis. The characteristics of the included studies were presented in Table 1 and Table S1. All cited references measured cytokines in different ways (e.g., Luminex xMAP suspension array, ELISA, and multiplex bead immunoassay). As shown in Table 2, according to the NOS checklist, 10 studies with scores ≥ 7 stars were considered high quality, and the remaining eight studies were medium quality for 6 stars.  The results present that there was a significant difference (SMD 1.94, 95% CI (1.05, 2.83)) but with heterogeneity in the aqueous humor level of VEGF between the DR and control groups. Further, the publication bias was not significant (Figure 2(b)). Bending with significant heterogeneity, it was necessary to perform subgroup and sensitivity analyses. Location, DR definition, and cytokine measurements may be the main causes of heterogeneity, but heterogeneity was still significant (I 2 = 93:4%, 93.1%, and 97.8%, respectively) after the subgroup analysis (Table 3). A sensitivity analysis was used to evaluate the stability and reliability of the results ( Figure 3). After removing the studies [11,14,21] that were contributing the most to the heterogeneity, the results did not change substantially (SMD 1.61, 95% CI (0.89, 2.33), P < 0:001, I 2 = 85:8%). There were significant differences in the aqueous humor level of VEGF between the DME and controls with SMD 1.07 (95% CI (0.71, 1.42), P < 0:001, Figures 2(c) and 2(d)) and no significant heterogeneity (I 2 = 43:6%, P = 0:114). Moreover, the funnel plot showed that publication bias was also not significant.

3.3.2.
Analysis of IL-6. There were 3 studies that reported data on the relationship of IL-6 levels and DR among participants with T2DM. The results present that there were significant differences between DR and control groups (SMD 3.53, 95% CI (0.37, 6.69), P = 0:028), but with significant heterogeneity (Figure 4(a)). However, the publication bias was not significant (Figure 4(b)).
The funnel plot showed that publication bias was not significant ( Figure 5(d)).

Analysis of IL-8.
The pooled SMD was 0.38 (95% CI (-0.05-0.81), P = 0:087) for the association between IL-8 levels in the aqueous humor and risk for DR (Table 4). Moreover, the funnel plot showed that publication bias was not significant.  On the basis of the five studies, the overall SMD of DME was 1.68 (95% CI (0.97-2.40), P < 0:001) in IL-8 levels ( Figure 6(a)). However, there was a significant heterogeneity among the five studies (I 2 = 81:0%, P < 0:001). Dividing into subgroups according to countries, definition of DR, and cytokine measurements, respectively, heterogeneity was still evident (Table 4). After removing the study by Noma et al. [17], which was contributing to the significant heterogeneity, the heterogeneity was not significant (I 2 = 0:0%, P = 0:525). The funnel plot showed that publication bias was not significant (Figure 6(b)).

Analysis of IL-12.
Three original studies were included in the current meta-analysis for the association of IL-12 levels with the risk of DR. Finally, the overall SMD of DR was 0.44 (95% CI (-0.81-1.70), P = 0:488) in IL-12 levels ( Table 4). The publication bias was not significant.
3.3.7. Analysis of TNF-α. There were 4 studies which involved the aqueous humor levels of TNF-α and its risk for DR. Table 4 also shows that the pooled SMD of TNF-α for DR was 0.51 (95% CI (-0.04-1.06), P = 0:067) and without significant publication bias.

Discussion
To the best of our knowledge, the present meta-analysis evaluates the influence of the aqueous humor concentration of inflammatory cytokines (IL-6 among ten studies, IL-8 among ten studies, IL-10 among six studies, IL-12 among seven studies, TNF-α among six studies, MCP-1 among nine studies, and VEGF among thirteen studies) on the risk of DR or DME. In the current meta-analysis, we found that elevated levels of mediators or cytokines (VEGF, IL-6, and MCP-1) in the aqueous humor were strongly associated with DR, while the levels of VEGF, IL-6, IL-8, and MCP-1 were strongly associated with DME. Moreover, our meta-analysis presents strong points: the literature review was updated to comprise the latest evidence; a set of inclusion and exclusion criteria were applied for study selection; and outcomes were robust in the sensitivity or subgroup analysis. In the current meta-analysis, there was a significant association between the aqueous humor VEGF levels and the risk for both DR and DME, and similar outcomes were also found in previous studies by Wu et al. [15] and Endo et al. [25]. VEGF is an endothelial cell mitogen, which can induce increases in vascular permeability and angiogenesis, enhance collateral vessel formation, and increase the permeability of the microvasculature [29]. Moreover, previous studies have pointed out that the VEGF level was significantly correlated with the severity of DR [30]. Recently, anti-VEGF drug intraventricular injection had been commonly used for the treatment of DR and DME [31,32]. However, there are some complications associated with intravitreal injection, such as increased intraocular pressure (IOP) [33], endophthalmitis [34], and geographic atrophy [35]. Our results confirmed that VEGF levels in the aqueous humor have a significant correlation with both DR and DME. Therefore, the dosage form of VEGF may be changed to avoid the above complications and achieve the same therapeutic effect.
Another interesting finding of this meta-analysis is that levels of IL-6 in the aqueous humor have a significant relationship with DR and DME. It is well known that continuous proinflammatory responses and neovascularization are related to the occurrence and progression of DR. IL-6 is a proinflammatory cytokine and is a key factor in host defense against environmental stress such as inflammation, infection, and injury [36]. Moreover, IL-6 can also increase vascular permeability and angiogenesis by inducing VEGF expression [37]. Furthermore, Arjamaa et al. have documented that increased levels of IL-6 in the vitreous significantly correlated with the activity of neovascularization [38]. Furthermore, IL-6 could not only enhance inflammation responses of DR or DME but also contribute to neovascularization and promote the reaction with VEGF at the same time.
MCP-1 is also a proinflammatory cytokine, which can induce monocyte and macrophage infiltration into tissues and trigger their transmigration to the sites of inflammation produced by tissue hypoxia, infection, or macrophage injury [39][40][41][42]. Several studies reported that vitreous or aqueous levels of MCP-1 were higher in the eyes with DR compared with normal controls [7,9]. They suggested that MCP-1 may play as a mediator in capillary occlusion in DR through the activation and adhesion of leukocytes and macrophages to the endothelium. Moreover, vitreous MCP-1 levels were found significantly correlated with the degree of proliferative membrane in the eyes with proliferative diabetic retinopathy (PDR), suggesting that MCP-1 may play an important role in the development of PDR [43]. In the current meta-analysis, due to the little information on the subtype of DR, we did not perform a meta-analysis on a different type of DR. Fur-ther studies are still needed to evaluate the association regarding cytokines in the aqueous humor and its risk for the severity of DR.  Moreover, previous studies have reported that the aqueous levels of MCP-1 were significantly higher in DME patients than in controls [8,14]. Furthermore, the aqueous humor level of MCP-1 was significantly correlated with central macular thickness (CMT), suggesting that MCP-1 might play a role in the development of DME. MCP-1 is a potent eosinophil chemotactic cytokine and could recruit the monocytes to the sites of vascular injury [44]. Meanwhile, the monocytes and macrophages which accumulate on the wall of the blood vessel increase vascular permeability that potentiate DME [45]. In addition, the activity of MCP-1 seems to be closely related to VEGF because MCP-1 could mediate the gene expression of VEGF-A [46].
Indeed, IL-8 is the prototype of a CXC chemokine, which has been recognized as a potent chemoattractant and an activator of neutrophils and T lymphocytes [41,47]. It is induced by hypoxia in vascular endothelial cells and plays an important role in promoting angiogenesis and tumor metastasis [41]. Once, some scholars reported that the levels of IL-8 in the vitreous were higher in the eyes with active PDR compared with quiescent PDR and they suggested that IL-8 might play an important part in the pathogenic process of PDR [48]. Elner et al. [49] also pointed out that the levels of IL-8 in the vitreous were significantly increased in active PDR. In our meta-analysis, we did not find any relationship between IL-8 and DR, maybe because we did not classify DR by severity. Moreover, researchers reported that IL-8 directly stimulates VEGF expression and the autocrine activation of VEGF receptor-(VEGFR-) 2 in vascular endothelial cells [19]. Kaneda et al. [50] reported that IL-8 appears to be involved in angiogenesis, endothelial cell binding and regeneration, endothelial wound healing, and vascular remodeling, presumably together with VEGF. These may be the reason why the aqueous humor levels of IL-8 increased significantly in the patients with DME.
Although some studies revealed that there are potential associations between the aqueous levels of IL-12, IP-10, or TNF-α and DR [15,24], other outcomes regarding those associations are controversial [14,20,26,27]. In our metaanalysis, there was no significant correlation between these cytokines and DR. Due to the lack of sufficient data, the association between cytokines (IL-12, IP-10, and TNF-α) and DME was not evaluated. Hence, further researches are needed for the evidence-based consequence.
The strengths of the current study include the comprehensive exploration of the evidence on the association between the cytokines in the aqueous humor and the risk of both DR and DME. However, there were some limitations in our study. First, there was a lack of studies available on the severity of DR for comparisons. Hence, our metaanalysis could only incorporate studies regarding any DR and controls. Second, we cannot obtain powerful outcomes adjusting potential confounders between the levels of cytokines and DR although we have done the subgroup analysis based on whether included studies had differences between case and control group or not. Last but not the least, there were insufficient studies for some cytokines such as IL-12, IP-10, and TNF-α to provide enough evidences to demonstrate the association between the biomarkers and risk for both DR and DME. Hence, large-sample, longitudinal, population-based studies are needed to validate our findings in the future.

Conclusion
In conclusion, we reviewed the literatures and conducted a current comprehensive meta-analysis to evaluate the association between the aqueous humor levels of VEGF, TNF-α, and inflammatory cytokines and risk for both DR and DME. Our findings indicated that lower mediator or cytokine levels in the aqueous humor may work best to attenuate DR and DME risk. This meta-analysis was the first comprehensive quantitative assessment of the aqueous humor levels of VEGF, IL-6, and MCP-1 on DR, which suggested that they may be used as biomarkers of the occurrence or development of DR, and reminded a new treatment besides intravitreal injection.

Data Availability
The data used to support the findings of this study are included within the article.

Disclosure
The funders had no involvement in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.