Fig. 4

Comparison of A adjusted multivariable regression and main Mendelian randomisation estimates and B mutually adjusted multivariable regression estimates and paternal negative control (exposure, paternal body mass index)—for the association of maternal body mass index with binary outcomes (Part 2). Paternal BMI was used as a negative control exposure to explore the potential presence, direction, and magnitude of bias in multivariable estimates for associations of maternal BMI with outcomes.. Results are expressed as odds ratios per SD unit of maternal BMI and paternal BMI for ‘Multivariable regression’ and ‘Paternal negative control’, respectively. Multivariable regression results were adjusted for paternal BMI, maternal age, parity, education, smoking during pregnancy, alcohol use during pregnancy, and offspring sex where available. Paternal negative control results were adjusted for maternal BMI, paternal age, number of children (ALSPAC only), paternal education, paternal smoking, paternal alcohol use, and offspring sex. BMI, body mass index; NICU, neonatal intensive care unit