Table 3.

Different approaches used in triangulation to determine the effect of maternal circulating pregnancy glucose on birthweight

Approach	Brief description of approach	Key sources of bias and directiona	Duration/timing of exposurea
Multivariable regression (European)55	Multivariable regression in 6008 European mother-offspring pairs. Adjusted for offspring sex and gestational age	Residual confounding by maternal socioeconomic position, age, parity and adiposity, which would result in exaggeration of any true causal effect	Fasting glucose assessed at single time point (24–28 weeks of gestation). For cumulative effect we assumed this is from then until birth (i.e. the last 12–16 weeks)
Cross-cohort comparison56	Comparing multivariable regression between 750 Pakistani origin and 607 White British origin mother-offspring pairs	Because of differences in the associations of SEP with fasting glucose between the two populations [mean difference in glucose by maternal education in Pakistani 0.00 mmol/l (-0.03, 0.04) and in White British 0.04 (0.02, 0.06); by receipt of income support in Pakistani -0.25 (-0.44, -0.07) and in White British 0.10 (0.01, 0.21)], if an association in White British were due to SEP confounding, we would expect a weaker association in Pakistani women	Fasting glucose assessed at single time point (26–28 weeks of gestation). For cumulative effect we assumed this is from then until birth (i.e. last 12–14 weeks)
MR55	Use of a weighted allele score of genetic variants known to be robustly associated with fasting glucose as an IV in 11 493 European mother-offspring pairs	Methods, including sensitivity analyses, were undertaken in the paper to explore the possibility of bias due to: (i) weak instruments; and (ii) violation of the exclusion restriction criteria. On the basis of these we concluded that these results may be somewhat biased towards the null as a result of adjusting for offspring genetic variants (see Supplementary text)	Assumed this approach tests fasting glucose across the whole of pregnancy
IV of intermediate in RCT57	958 women with mild gestational diabetes mellitus randomized to dietary advice, glucose monitoring and insulin treatment if necessary or usual care. We calculated an IV ratio estimate (difference in birthweight by randomised group ÷ difference in glucose by randomised group)	Glucose was not monitored in the control arm and we brought the baseline value forward. Maternal fasting glucose levels increase in the second and third trimesters of pregnancy. As a result the denominator of the IV ratio estimate (i.e. difference in fasting glucose by randomized group) is likely to have been an underestimate of the true difference and the IV estimate of the effect of fasting glucose on birthweight an exaggeration of any real causal effect. Potential violation of the exclusion restriction criteria by dietary advice possibly result in changes to other risk factors that influence birthweight independently of glucose. The likely direction of effect of this bias is unclear (see Supplementary text). Overall, we decided that likely combined bias for this approach would be to exaggerate a positive effect of glucose on birth weight (Supplementary text)	Randomization and fasting glucose assessed at ∼ 29 weeks of gestation. For a cumulative effect we assume differences were present for the last 11 weeks of pregnancy

^aIn Supplementary text (available at IJE online) we provide full details of how we assessed a range of potential key sources of bias and their likely direction; here we describe the ones that we concluded were the key sources.

SEP, socioeconomic position.