Figure 2. Associations of tobacco smoking with mammographic density phenotypes compared to non-smokers, overall and stratified by BMI category, menopausal status, and use of menopausal hormone therapy.

All models were adjusted for tobacco use, age, age², BMI, BMI², BMI³, late adolescent BMI, age at menarche, age at first birth, parity, menopausal status, menopausal hormone therapy (MHT) use, first-degree relative with breast cancer, physical activity, and image batch. Effects were estimated using separate linear regression models of the square-root transformed phenotype in the Hologic and GE cohorts, and combined using restricted maximum likelihood (REML) random effects meta-analysis. Coefficients (β) and 95% confidence intervals (CI) were back-transformed to the original scale. PPD = packs per day.