Statistical analysis
Hazard ratios (HR) or Mantel–Haenszel (MH) risk ratio (RR) models where appropriate and 95% confidence intervals (CI) were used to summarize data across treatment arms. The Cochran Q-test and the Higgins I-squared (I2) statistic were used for heterogeneity testing; in cases of heterogeneity (defined as I2>25%), random effects models of DerSimonian and Laird were used, otherwise (I2 < 25%) fixed effects models were used. To address publication bias, we used the funnel plots. If any bias was observed, further bias quantification was measured using the Begg-Mazumdar test, Egger test, and the Duval and Tweedie´s trim and fill test.
Descriptive statistics are presented as means and standard deviations (SD) for continuous variables or number of cases (n) and percentages (%) for dichotomous and categorical variables. Number needed to treat (NNT) was calculated. Statistical analysis was performed in line with recommendations from the Cochrane Collaboration and the PRISMA guidelines, using Review Manager (RevMan) [Computer program]. Version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014, and the Comprehensive Meta-Analysis Software version 2.0 (Biostat, Inc).