“…We ran two models including the basic model (model 1) that was adjusted for age and sex, and the fully adjusted model (model 2) that was further adjusted for ethnicity (white or others), TDI, education (degree or no degree), BMI, physical activity (<500 or ≥500 MET-min/week), smoking status (never, former or current), alcohol consumption (never, former or current), vegetable and fruit consumptions (<2.0, 2.0–3.9 or ≥4.0 pieces/day or tablespoons/day), maternal smoking (yes or no), and following variables which were reported as present or absent, including breastfed as a baby, part of multiple birth, aspirin use, non-aspirin NSAID use, vitamin supplements use, mineral and other dietary supplements use, hypertension, diabetes, dyslipidemia. These factors were included because were well-known to be associated with the risk of developing CVD ( 13 , 14 ), or commonly included in multivariate models for the outcomes ( 15 , 16 ). Furthermore, with birth weight as a continuous exposure variable, we used restricted cubic splines with five knots at 5th, 27.5th, 50th, 72.5th, and 95th percentiles to evaluate the potential non-linear effect of birth weight on CVD incidence in the fully adjusted model ( 17 ).…”