Data Analysis
For studies with consecutive enrolment and where data was presented on the planned mode and actual mode of delivery for all subjects, we calculated the median TOLAC rate and median VBAC rate across studies, using frequency weights for study sample size. The primary clinical outcome assessed was uterine rupture. Secondary outcomes were maternal mortality and perinatal (stillbirth and neonatal) mortality. We estimated the overall rate of these adverse outcomes across all births for which this data was available to generate estimates across all available studies, also weighted for study sample size. We further compared uterine rupture rate, maternal mortality and perinatal mortality (stillbirth and neonatal mortality) between TOLAC and ERCS in studies where comparison groups were available and calculated the pooled odds ratios of adverse outcomes with 95% confidence intervals. For this comparison, a random-effects model was used to combine the studies while accounting for heterogeneity. We used forest plots to graphically summarise the pooled results. We performed sensitivity analyses using studies with a MINORS score of ≥16.
We performed subgroup analyses to evaluate TOLAC and VBAC rates over time and compared rates in studies published before and after January 1st, 1996 using the rank-sum test to assess for differences by region and changing practice patterns and trends. We choose 1996, as a comparison date as the last published metanalysis of TOLAC in SSA included studies up to 1995.10 This year also corresponds to a period when a marked decline in the rates of TOLAC and VBAC was seen in the other regions of the world due to emerging data on risks of uterine rupture following trial of labour.13
All statistical calculations were performed with STATA, version 13 (StataCorp, College Station, TX USA).