Background Gestational diabetes (GDM) is hyperglycaemia first identified in pregnancy. GDM is one of the most common medical conditions and associated with increased risk of adverse perinatal outcomes. Cohort studies have shown graded linear increases in several important perinatal outcomes across the whole glucose distribution in women without existing diabetes or GDM, with no clear threshold where risk increases substantially. The International Association of Diabetes in Pregnancy Study Groups (IADPSG) used a recent study’s results to define threshold criteria for GDM diagnosis. These criteria aim to reduce obesity risk by identifying infants who are large at birth or with high levels of adiposity. Our aim was to determine the extent to which this study’s findings and the IADPSG thresholds generalise across a range of perinatal outcomes and populations.

Methods We conducted a systematic review and meta-analyses. Databases including MEDLINE were searched until October 2014. Studies comprised pregnant women with oral glucose tolerance or challenge testing (OGTT or OGCT), without pre-existing or gestational diabetes

Results Twenty-eight studies were included. GDM was diagnosed with either the 75 g or 100 g OGTT and a variety of glucose thresholds. Meta-analyses showed positive associations for all glucose exposures with Caesarean section (CS), induction of labour, large for gestational age, macrosomia and shoulder dystocia. The magnitude of associations were stronger for fasting, compared with post-load glucose measurements (e.g. CS fasting glucose: OR/1 mmol/L 1.59 (CI 1.49–1.70), CS 75 g two-hour post-glucose: OR/1 mmol/L 1.11 (CI 0.93–1.31)). Fasting glucose was inversely associated with preterm delivery (OR/1 mmol/L 0.77 (CI 0.62–0.96). Outcomes associations were generally monotonic, suggesting linear associations across the distribution with no clear threshold. I² statistic for heterogeneity for all associations was very low or 0.

Conclusion The positive linear associations of fasting and post-load glucose with most adverse perinatal outcomes suggest no clear threshold where risk increases substantially, which means that thresholds to diagnose GDM must to some extent be arbitrary. Generally, the magnitude of glucose associations across studies were similar and the lack of heterogeneity provides assurance that our estimates are appropriate.

Because there is a graded linear association between glucose and risk of adverse perinatal outcomes, risk of these outcomes may be reduced if glucose thresholds are lowered, however there are no trials using these new thresholds and no robust evidence that longer-term obesity risk would be improved. Trials evaluating different approaches to the detection of GDM with longer-term follow up are therefore required.