Statistical analysis
The differences in characteristics between excluded and included women were compared using chi-square tests for categorical variables and independent t-tests for continuous variables. The associations of the change in interpregnancy BMI with subsequent pregnancy outcomes in the second pregnancy were examined using modified Poisson regression models to estimate relative risks (RR) and 95% confidence intervals (CI).17 The change in interpregnancy BMI was included as a categorical exposure (BMI stable, loss, moderate gain, and excess gain), with stable BMI used as the reference group, as conventionally used by other studies.10 The models were adjusted for maternal age (continuous), ethnicity (categorical), gestational-age-adjusted BMI at 12 weeks in the first pregnancy (continuous), interpregnancy interval (continuous) and pregnancy outcomes in the first pregnancy (categorical).
As the effect of interpregnancy change may differ by maternal BMI at the beginning of the first pregnancy, we performed post-hoc analysis to examine whether there was any effect modification by weight status <23 vs ≥23 kg/m2 at 12 weeks during the first pregnancy on any observed association. These models included categorical interpregnancy BMI change, weight status (effect modifier), the interaction terms between categorical interpregnancy BMI change weight status (3 degrees of freedom), and potential confounders as the independent variables. The results were stratified by weight status.
Sensitivity analyses were performed using a similar modified Poisson regression to analyse the associations of the crude (unstandardised for gestational age) change in interpregnancy BMI with pregnancy outcomes, with confounder adjustment. These analyses were restricted to those with measures before or at 12 weeks gestation for both pregnancies. Statistical analyses were performed using Stata 16 (Stata, College Station, TX, USA).