Study searching and their basic information
Using various combinations of key terms, a total of 262 article titles were garnered by a document search using the PubMed (222 titles) and Wanfang (40 titles) databases. As shown in Figure 1, 178 articles were excluded after screening the Abstract sections of the manuscripts. The full texts were then evaluated, and 49 additional articles were excluded due to duplication (7), meta-analysis or systematic analysis (26), clinical trial (10), randomized controlled trial (6). Finally, 35 different articles [26, 30-57, 65-69] were included in our meta-analysis, including 38 case-control studies about HTRA1 gene rs11200638 polymorphism and AMD risk (Table 1) and 27 case-control studies about HTRA1 gene rs11200638 polymorphism and wet or dry AMD risk (Table 2). Five case-control studies [65-69] were not consistent with HWE in control groups. To make our analysis to more strict, we deleted above five studies, so there were about 33 case-control studies (8101 cases and 7215 controls) for the whole AMD [26, 30-57], and 22 case-control studies for wet or dry (3938 cases and 4427 controls) studies [26, 30, 31, 33, 34, 40, 41, 43, 44, 46-48, 51, 53-56]. The frequency of the A allele from case group was found to be higher in control individuals (54.2% vs. 36.5%) (Figure 2). There were 19 case-control studies of Asian population, and 14 from Caucasian population; source of control in 22 case-control studies were from HB, and 11 were from PB; 17 case-control studies were about wet AMD disease, and 5 were about dry disease. Finally, we checked the Minor Allele Frequency (MAF) reported for the five main worldwide populations in the 1000 Genomes Browser (https://www.ncbi.nlm.nih.gov/snp/rs11200638): Global (0.290); East Asian (EAS=0.411); European (EUR=0.194); African (AFR=0.257); American (AMR=0.250); and South Asian (SAS=0.340) (Figure 3). The genotyping methods included polymerase chain reaction-restrictive fragment length polymorphism and matrix-assisted laser desorption/ionization time-of-flight, sequencing, real-time PCR, and TaqMan.
Quantitative Synthesis
Total analysis
Results of the overall meta-analysis were suggestive of increasing associations between this polymorphism and AMD susceptibility in all five genetic models (for example: AA vs. GG: OR = 4.26, 95CI% = 3.19-6.70, P < 0.001) (Table 3). In order to make this study more convincing and reliable, we detected five studies, which were not according with HWE, finally, we tested the 32 case-control studies. Also significantly increasing correlations were observed in whole genetic models (for example: A-allele vs. G-allele: OR = 2.36, 95%CI = 1.98-2.81, P < 0.001; AA+AG vs. GG: OR = 2.53, 95%CI = 2.04-3.14, P < 0.001) (Figure 4) (Table 3).
Subgroup analysis
Coming up, we all know that the frequency of A-allele in different races was not the same, so we tried to analysis the relationships by ethnicity subgroups in further, which indicated an incremental statistically association between this polymorphism and both in Asians (A-allele vs. G-allele: OR = 2.51, 95% CI = 2.22-2.83, Pheterogeneity < 0.001, P < 0.001, ,Figure 4; AA vs. AG+GG: OR = 3.69, 95% CI = 3.17-4.29, Pheterogeneity = 0.013, P < 0.001) and Caucasian populations (dominant model, OR = 2.15, 95% CI = 1.46-3.17, Pheterogeneity < 0.001; heterozygote comparison, OR = 1.82, 95% CI = 1.34–2.48, Pheterogeneity < 0.001; allelic comparison, OR = 2.17, 95% CI = 1.53-3.10, Pheterogeneity < 0.001, P < 0.001, Figure 4) (Table 3). In addition, regular analysis by source of control, also significantly increased risks were detected for this SNP both in PB and HB studies (AG vs. GG: OR = 1.70, 95% CI = 1.33-2.16, Pheterogeneity < 0.001, P < 0.001 for HB; AG vs. GG: OR = 2.16, 95% CI = 1.84-2.53, Pheterogeneity = 0.021, P < 0.001 for PB) (Table 3) (Figure 5). AMD have different types and stages, the different of clinical presentation for dry and wet AMD is completely different, so we firmly believed that the correlations existed should be evaluated separately, significant positive associations were found both for dry (such as AA+AG vs. GG: OR = 2.73, 95% CI = 2.13-3.51, Pheterogeneity = 0.498, P < 0.001, Figure 6A) and wet AMD (for example in AA+AG vs. GG: OR = 3.40, 95% CI = 2.90-3.99, Pheterogeneity = 0.073, P < 0.001, Figure 6B). Finally, different genotype methods were applied in included studies, we tried to in each method, whether associations may exist in our analysis, we found some positive results in some methods (such as in AA vs. GG model: OR = 7.00, 95% CI = 5.84-8.39, Pheterogeneity = 0.677 about sequencing, Figure 7A; OR = 7.52, 95% CI = 2.05-27.68, Pheterogeneity < 0.001, P = 0.002 about TaqMan; OR = 4.30, 95% CI = 2.51-7.35, Pheterogeneity = 0.073, Figure 7B) (Table 3).
Bias Diagnosis for publication and sensitivity Analysis
The publication bias was evaluated by both Begg’s funnel plot and Egger’s test. At beginning, the shape of the funnel plots seemed asymmetrical in allele comparison for rs11200638 by Begg’s test, suggesting no publication bias was existed. Then, Egger’s test was applied to provide statistical evidence of funnel plot symmetry. As a result, no obvious evidence of publication bias was observed (A-allele vs. G-allele, t = 0.89, P = 0.38 for Egger’s test; z = 0.85, P = 0.396 for Begg’s test, Supplementary Figure 1A,B)(Table 4).
To delete studies which may influence the power and stability of whole study, we applied the sensitive analysis, finally, no sensitive case-control studies were found (Supplementary Figure 2).
Gene-gene network diagram and interaction of online website
String online server indicated that HTRA1 gene interacts with numerous genes. The network of gene-gene interaction has been illustrated in Figure 8.