Data synthesis
Meta-analyses were performed using STATA version 13 software (Stata Corp, College Station, TX, USA). Heterogeneity among studies was assessed using the χ2 test and I2statistic; an I2 > 50% or P < 0.05 for the Q-statistic indicated significant heterogeneity[15]. The DerSimonian and Laird random-effects model was used when studies were heterogeneous; otherwise, the Mantel-Haenszel fixed-effects model was used[16]. When possible, adjusted effect estimates (odds ratio [OR], relative risk, and hazard risk) of outcome measures from exposure to a neuraminidase inhibitor were extracted, along with standard errors. Associations between maternal neuraminidase-inhibitor exposure during pregnancy and adverse neonatal outcomes were estimated using ORs and corresponding 95% confidence intervals (CIs), generated by comparing cases and controls. Publication bias was evaluated using Begg’s test[16, 17]. A P-value < 0.05 was considered to indicate significance.