Objective
Randomised controlled trials of diet and physical activity (PA) interventions in pregnancy consistently demonstrate little impact on gestational weight gain (GWG) and pregnancy outcomes (1, 2), although meta-analyses report some benefit (3, 4). Our aim was to evaluate how the quality of included trials impacted treatment effect estimates.
Study design
We conducted a systematic review of dietary and/or PA interventions for pregnant women with a body mass index >18.5kg/m2. Studies were assessed for risk of bias and methodological features impacting reliability. Trials were classified based on the level of potential bias (no/negligible, minimal, moderate or substantial risk of bias) and intervention type (diet and/or PA). Outcomes included GWG; gestational diabetes mellitus (GDM); pre-eclampsia; caesarean birth; and birth weight measures. A sequence of meta-analyses were performed for each outcome, based on intervention type and level of potential bias in the effect estimate.
Results
We identified 128 eligible studies. The most robust estimate from a combined diet and PA behavioural intervention, with only studies at negligible risk of bias, was a difference in GWG of 1.10kg (95% CI -1.62 to -0.58; 17,755 women). There was no evidence of an effect on any clinical outcomes.
Conclusions
These findings highlight discrepancies produced by the indiscriminate inclusion of studies with methodological flaws in previous systematic reviews. A continued focus on GWG as an outcome and advocating for the implementation of antenatal dietary and PA interventions is ineffective, and may be harmful, in the absence of beneficial clinical outcomes. Our findings do not support the regular weighing of pregnant women, a practice which may distract from the consideration of other strategies to improve maternal health.