2 | Material and Methods

2.1 | Searches strategy and selection criteria

Our systematic review was performed and reported in accordance with Meta-analysis Of Observational Studies in Epidemiology (MOOSE)20 and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)21guidelines
We performed a systematic search in PubMed/Medline (1950 to Apr 2020), EMBASE (1980 to Apr 2020), and Cochrane Central Register of Controlled Trials (1950 to Apr 2020) for association between BCG vaccination and Allergic disease by using relevant keywords including asthma, eczema, rhinitis, BCG vaccines and other synonyms. The search method is provided in appendix 1. We restricted the search to studies published in English language and we screened bibliographies of relevant review articles to ensure that all relevant studies were included.
Studies were first selected on the basis of their titles and abstracts by two independent investigators. Then they retrieved full texts and performed further screening when studies were deemed eligible. Studies had to be either cohort or cross-sectional with participants data included. Disagreements were resolved by discussion and, if necessary, in consultation with a third, senior investigator.

2.2 | Data extraction and quality assessment

Two authors extracted data independently using a standard data extraction form. The following baseline characteristics were extracted from the included studies: first author, year of publication, study design, location in which the study was performed, number of included participants and allergic disease. Studies were excluded when participant data was not integrated.
Quality of all included trials was assessed by two authors independently by using the Cochrane Collaboration risk of bias tool.22 This tool evaluated biases from seven domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and others. The risk of bias in each domain was judged as low, high, or unclear. The overall risk of bias in a study was classified as low if all domains had low risk; as high if one or more domains had high risk, or as unclear otherwise.12 Based on these standards, we classified the studies into the following three grades: A, high quality and low risk of bias; B, moderate quality and moderate risk of bias and C, low quality and high risk of bias. Disagreements between the reviewers were resolved by discussion with involvement from a third senior investigator if necessary.

2.3 | Data analysis

We used STATA (version 12.0) to perform the data analysis, Odds ratios (ORs) and their associated 95% confidence intervals (CI) were used to assess the strength of association between BCG vaccination in early life and the risk of getting allergic disease. Statistical tests were judged statistically significant if the two-side P value was less than 0.0523. I² statistic were used for investigating heterogeneity and if I² value was greater than 50%, it implied statistical heterogeneity. We used random-effects modeling to perform the meta-analysis if significant heterogeneity were performed and if not, we used fixed-effects modeling.
If heterogeneity existed, we performed a subgroup analysis to investigate whether the heterogeneity was related to the participant’s race. Different ethnic background in different continents may be considered potentially important to heterogeneity because of living habit diversity. Funnel plots were used for displaying the publication bias graphically, both specifically and officially with Egger’s test.