Table 2.

Hazard ratios for presence of elevated depressive symptoms and antidepressant use with respect to incident diabetes risk

	Unadjusted Cox PH model	Approach 1a	Approach 2a	Approach 3a	Approach 4a
	Cox PH Model (baseline antidepressant medication use and baseline presence of elevated depressive symptoms)	Cox PH Models (baseline values of all predictors)	Cox Models (time varying antidepressant medication use, presence of elevated depressive symptoms, and BMI; baseline values for other covariates)	Cox Models (time varying antidepressant medication use, presence of elevated depressive symptoms, and BMI; adjusted for propensity score and baseline values for other covariates)	Marginal Structural Models (time varying antidepressant medication use, presence of elevated depressive symptoms, and BMI; baseline values for other covariates)
Predictor	HR (95 % CI)	HR (95 % CI)	HR (95 % CI)	HR (95 % CI)	HR (95 % CI)
	Observational Cohort (OS) (N = 68,169)
Antidepressant medication use up to year 3	1.26	1.19	1.31	1.32	1.35
	(1.12–1.41)	(1.06–1.35)	(1.18–1.46)	(1.19–1.47)	(1.21–1.51)
Presence of elevated depressive symptoms	1.34	1.11	1.12	1.09	1.10
	(1.23–1.45)	(1.02–1.21)	(1.03–1.23)	(1.00–1.19)	(1.00–1.20)
	Clinical Trial (CT) (N = 52,326)
Antidepressant medication use up to year 3	1.17	1.14	1.26	1.25	1.27
	(1.04–1.31)	(1.01–1.30)	(1.12–1.41)	(1.12–1.40)	(1.13–1.43)
Presence of elevated depressive symptoms	1.31	1.13	1.12	1.09	1.10
	(1.21–1.42)	(1.04–1.23)	(1.03–1.22)	(0.996–1.18)	(1.00–1.20)

^aModels adjusted for baseline values of age, ethnicity, education, minutes of recreational physical activity per week, total energy intake, hormone therapy use, family history of diabetes and smoking status