Quality assessment
The quality of included studies and risk of bias were assessed by two reviewers using Newcastle Ottawa Scale (NOS, a design-specific scale –case–control vs. cohort study) independently (14). Three items were contained in NOS, including selection of study subjects, comparability, exposure assessment and outcome. Eight items were used to identify the quality of study. In exposure/outcome assessment, each item could be given one star, and comparability a maximum of two stars. As such, the total score ranges from zero to nine stars (15). Study inclusion was not affected by quality score, but considered in sensitivity analysis and results interpretation.