Data analysis
We performed a meta-analysis, using a random effects model, if at least three studies reported on the outcome of interest. Odds ratios (OR) and 95% confident intervals (CI) comparing women with liver cirrhosis to controls were calculated. A p-value <0.05 was considered statistically significant. We used the I2 for testing statistical heterogeneity. 16 Sensitivity analyses were performed when the I2 test showed evidence of high heterogeneity, corresponding to I2>75%. We performed sensitivity analyses by removing contributing papers from the analysis that were thought to be responsible for heterogeneity based on deviating study design, case definition, study population and etiology of liver cirrhosis. We used The Cochrane Manager Reviewer (REVMAN 5) 17 for the statistical analysis.
To analyze decreases of maternal mortality and variceal hemorrhage we calculated the odds ratio of the total events that occurred, after dichotomizing the odds ratios from studies conducted before and after the median of total study period of all included studies. Also, we made time sequenced plots to visualize the decreases of maternal mortality and variceal hemorrhage. We used the median date of inclusion of individual studies as the value on the X-axis and the incidence rates as value on the Y-axis.