Literature search and selection
The literature search obtained 217 publications. Of them, 103 duplicates were removed, and a further 86 were removed after assessing for relevance by screening titles and abstracts. After reading the remaining 28 full-text articles, 14 studies that did not satisfy the eligibility criteria were rejected. Finally, 14 articles were included in the present meta-analysis. The flowchart of literature selection process was illustrated in Fig. 2.
Table 1 summarized the major characteristics of all included studies. A total of 14 studies were included in this work [15-28]. Among them, 12 studies analyzed the rs13118928 polymorphism, and 9 studies investigated the rs1828591 polymorphism. The publication years of the studies ranged from 2013 to 2021. There were 10 studies on Asians and 4 studies on Caucasians. Of note, the studies by Xie et al  and Xu et al , consisted three and two independent cohorts respectively. For rs13118928, 3 studies did not conform to HWE. Regarding NOS, 12 studies were deemed as excellent quality (with 7 to 8 scores), and the rest 2 studies were in fair quality (with 6 scores).
Table 2 summarized the outcomes of the meta-analysis and subgroup analysis. The effect sizes were estimated based on the allelic, homozygous, heterozygous, dominant, and recessive models.
Rs13118928 polymorphism and COPD
Eleven published articles with 15 studies reported the association of rs13118928 polymorphism and COPD risk. Large heterogeneity was observed in five genetic models, and the random-effects model was used to analyze the data. The combined results supported statistical differences between the rs13118928 polymorphism and COPD: A vs. G, OR=1.18, 95CI%=[1.07-1.30], P=0.0006; AA vs. GG, OR=1.56, 95CI%=[1.22-2.00], P=0.0004; AG vs. GG, OR=1.28, 95CI%=[1.05-1.55], P=0.01; AA+AG vs. GG, OR=1.36, 95CI%=[1.12-1.65], P=0.002; AA vs. AG+GG, OR=1.18, 95CI%=[1.05-1.33], P=0.006. Forest plot of A vs. G in rs13118928 was illustrated in Fig 3.
Subgroup analysis based upon ethnicity revealed that the heterogeneity mainly originated from the Asians, in which the random-effects model was adopted. The pooled data indicated that rs13118928 polymorphism contributed to an increased risk to COPD in both Asians and Caucasians.
Rs1828591 polymorphism and COPD
Nine studies with ten cohorts examine the rs1828591 polymorphism and COPD susceptibility. Considerable between-study heterogeneity was found in homozygous, heterozygous, and dominant models, where the random-effects model was employed. The merged outcomes suggested rs1828591 polymorphism was correlated with COPD: A vs. G, OR=1.12, 95CI%=[1.05-1.19], P=0.0003; AA vs. GG, OR=1.27, 95CI%=[1.04-1.56], P=0.02; AG vs. GG, OR=1.25, 95CI%=[1.03-1.51], P=0.02; AA+AG vs. GG, OR=1.26, 95CI%=[1.04-1.53], P=0.02; AA vs. AG+GG, OR=1.10, 95CI%=[1.01-1.19], P=0.03. Forest plot of A vs. G in rs1828591 was illustrated in Fig 4.
Subgroup analysis by ethnicity revealed the heterogeneity mainly existed in the Asians, in which the random-effects model was adopted in four genetic models. The pooled data indicated that rs1828591 polymorphism was associated with an elevated susceptibility to COPD in both Caucasians but not Asians.
Sensitivity analyses and publication bias
After the removal the studies deviating from HWE, the corresponding pooled ORs and 95CIs were not significantly altered. Therefore, they were kept in the final data combination. Sensitivity analyses did not provide reverse outcomes regarding the rs13118928 and rs1828591 polymorphisms by removing each study at a time. Therefore, the sensitivity analysis confirmed that the results were reliable and robust. Forest plots did not show significant evidence of publication bias of the included studies. Labbe graph of sensitivity analyses and Funnel plot of publication bias are displayed in Fig 5 and Fig 6.