Literature search and study characteristics
Initially, a total of 414 articles were identified to be relevant to the search terms from several different databases. After a systematical literature search according to the inclusion and exclusion criteria, 11 case-control studies from 7 published articles with a total of 2215 DN cases and 2001 DM controls investigating the genetic effects of EMLO1 rs741301 polymorphism and IL–8 rs4073 polymorphism on DN susceptibility, were recruited for this meta-analysis[7, 11, 12, 15–18]. A flow diagram in Figure 1 displayed the selection process with more particulars.
The main characteristics of included studies were summarized in Table 1. Of the 11 studies included, 9 studies from 6 selected articles containing 1879 cases and 1654 controls were performed to detect EMLO1 rs741301 polymorphism[11, 12, 15–18]; 5 studies (661 cases and 674 controls) in 2 published articles were used to examine IL–8 rs4073 polymorphism[7, 17]. In terms of EMLO1 rs741301 polymorphism, 5 studies were performed in Asians[11, 15, 17], 4 studies was undertaken in Caucasians according to ethnicity[12, 16–18]. As for IL–8 rs4073 polymorphism, there were 2 studies carried out of Asian ethnicity[17] and 3 studies conducted of Caucasian ethnicity[7, 17]. Genotype distributions and HWE examination results were exhibited in Table 2. The genotypes of EMLO1 rs741301 polymorphism in the control groups were in agreement with HWE excepted to 1 article[11]. For IL–8 rs4073 polymorphism, HWE deviation existed in control of 1 eligible study[7].
Individual polymorphism meta-analysis and subgroup analysis
The relationship between EMLO1 rs741301 polymorphism or IL–8 rs4073 variant and DN risk were shown in Table 3.
For EMLO1 rs741301 polymorphism, the susceptibility for DN did not show significant difference under the following genetic models: recessive model (GG vs. GA +AA: OR = 1.141, 95% CI = 0.941–1.385, P = 0.179); dominant model (GG+GA vs. AA: OR = 0.999, 95% CI = 0.874–1.142, P = 0.990); codominant model (GA vs. AA: OR = 0.968, 95% CI = 0.840–1.115, P = 0.650); homozygote model (GG vs. AA: OR = 1.120, 95% CI = 0.911–1.378, P = 0.283) and allele model (G vs. A: OR = 1.033, 95% CI = 0.938–1.139, P = 0.506). For studies assessing the relationship between EMLO1 gene polymorphism and DN risk in different ethnicity, significant increase in DN risk was identified in Asians under the recessive model (GG vs. GA +AA: OR = 1.840, 95% CI = 1.338–2.529, P = 0.000), the homozygote model (GG vs. AA: OR = 1.834, 95% CI = 1.309–2.569, P = 0.000) and the allele model (G vs. A: OR = 1.222, 95% CI = 1.053–1.417, P = 0.008) (Figure 2); no significant association was found among Asians in other genetic models (dominant model, GG+GA vs. AA: OR = 1.121, 95% CI = 0.915–1.372, P = 0.271 and codominant model, GA vs. AA: OR = 0.997, 95% CI = 0.805–1.234, P = 0.977). Available data did not suggest an association between this polymorphism and DN risk in Caucasians under all genetic models.
As for IL–8 rs4073 polymorphism, pooled effect among all studies suggested a significant association between the IL–8 rs4073 polymorphism and DN risk in dominant model (AA+AT vs. TT: OR = 1.450, 95% CI = 1.166–1.802, P = 0.001); codominant model (AT vs. TT: OR = 1.420, 95% CI = 1.129–1.786, P = 0.003); homozygote model (AA vs. TT: OR = 1.553, 95% CI = 1.094–2.203, P = 0.014) and allele model (A vs. T: OR = 1.291, 95% CI = 1.102–1.512, P = 0.002); but not in recessive model (AA vs. AT +TT: OR = 1.272, 95% CI = 0.921–1.757, P = 0.144) (Figure 3). When stratified population by the basis of ethnicity, a positive correlation between this gene polymorphism and DN risk among Caucasians was found under the dominant model (AA+AT vs. TT: OR = 1.770, 95% CI = 1.354–2.315, P = 0.000), codominant model (AT vs. TT: OR = 1.733, 95% CI = 1.304–2.302, P = 0.000), homozygote model (AA vs. TT: OR = 1.939, 95% CI = 1.263–2.976, P = 0.002) and allele model (A vs. T: OR = 1.494, 95% CI = 1.227–1.820, P = 0.000); but not in recessive model (AA vs. AT+TT: OR = 1.419, 95% CI = 0.958–2.104, P = 0.081) (Figure 4). However, IL–8 rs4073 variant was not found to be associated with the risk of DN in Asians under all the genetic models.
Sensitivity analysis
In order to explore the stability of the crude results, sensitivity analysis was performed by sequentially omitting each study.
For EMLO1 rs741301 polymorphism, the result showed that the Bodhini’s study influenced the stability of the pooled ORs. Therefore, a systematical meta-analysis was performed after deleting this study. The main result of the relationship between EMLO1 rs741301 polymorphism and diabetic nephropathy risk are presented in Table 4. Overall, the GG genotype of EMLO1 rs741301 polymorphism was a significantly increased risk factor for the development of DN (recessive model, GG vs. GA +AA: OR = 1.496, 95% CI = 1.164–1.922, P = 0.002; homozygote model, GG vs. AA: OR = 1.531, 95% CI = 1.172–1.999, P = 0.002 and allele model, G vs. A: OR = 1.172, 95% CI = 1.039–1.321, P = 0.010). After categorizing subjects into different subgroups on the basis of ethnicity, the results in Asians remained significant. Exclusion of the sensitivity analysis-deviated studies did not meaningfully change the pooled estimates in Asians.
As for IL–8 rs4073 polymorphism, the first study of Ahluwalia’s altered the stability of the pooled ORs. After omitting this study, the results showed that a positive correlation was found between IL–8 rs4073 polymorphism and DN in the overall population under the dominant model (AA+AT vs. TT: OR = 1.492, 95% CI = 1.132–1.966, P = 0.005) and codominant model (AT vs. TT: OR = 1.622, 95% CI = 1.213–2.168, P = 0.001). The subgroup analysis by ethnicity was not performed due to the limited data provided in Caucasians and Asians.
Publication bias
Publication bias was evaluated by Begg’s funnel plot and Egger’s linear regression test. All the shapes of the funnel plots were found to be symmetrical (Figure 5), indicating that there was a lack of publication bias for the association of EMLO1 rs741301 variant and IL–8 rs4073 polymorphism in all the genetic models (Table 5).