Table 4.

Multiple regression models investigating factors associated to BMI Changes

Model	All Women in Sample n=73,525	Underweight	Normal Weight	Overweight	Class I Obesity	Class II Obesity	Class III Obesity
	N=72291	N=850	N=29,849	N=24,640	N=10,891	N=3,854	N=2,207
Antidepressants	0.17 - 0.0001	0.43 - 0.043	0.12 - 0.0001	0.23 - 0.0001	0.28 - 0.0001	0.27 - 0.002	0.63 - 0.073
Antipsychotics	0.3 - 0.32	NA	0.57 - 0.199	0.23 - 0.501	0.44 - 0.645	0.39 - 0.685	1.93 - 0.57
Betablockers	0.02 - 0.311	0.18 - 0.378	0.02 - 0.47	0.04 - 0.159	0.04 - 0.357	0.01 - 0.866	−0.26 - 0.451
OTC Insulin	0.44 - 0.0001	−1.11 - 0.274	0.34 - 0.008	0.38 - 0.0001	0.34 - 0.008	0.79 - 0.0001	1.29 - 0.04
Glucocorticosteroids	−0.13 - 0.017	−0.66 - 0.07	0 - 0.993	−0.18 - 0.015	−0.31 - 0.007	−0.12 - 0.608	−0.18 - 0.847

Each row represents a model which main regressor is a weight promoting drug class, as each column represent a stratus by BMI class (except for the first column that represents all sample). All models are adjusted for baseline health conditions as diabetes, hypertension and hypercholesterolemia, physical activity, sleep medications, smoking and diet.

Values represent regression coefficients along with p-values (β - p value). Regression coefficients show the average change in BMI observed moving from non-user to intermittent user to persistent user relatively to the specific drug. Regression coefficients of covariates are shown in table S6.