Data synthesis
Data were summarised as risk ratio (RR) with 95% confidence interval (CI) for dichotomous
outcomes, and standardised or weighted mean difference with 95% CI for continuous outcomes
using the random effects model. We assessed statistical heterogeneity between trials by
using the I2 statistic. We undertook subgroup analysis planned a priori to explore whether the
effect on the outcome would vary according to the type of intervention, and Body Mass Index
(BMI), and risk status of the participants for GDM. The subgroup difference was evaluated
using Chi squared test. When more than one intervention was compared to standard care in a
study, we chose the combined intervention over the individual diet or nutritional supplement
for the pooled analysis. We used random effects model for meta-analysis. Sensitivity analysis
was undertaken by substituting the individual intervention instead of the combined method to
assess for any change in the summary estimates of effects. We used Harbord’s modified test to
assess for publication bias [17] and potential small study effect. All analyses were performed
with Review Manager (RevMan version 5.2) and Stata software (version 11).