Table 3. Association between change in sleep duration from pre- to post diagnosis and mortality (N=1949).

	Model 1		Model 2		Model 3
Change in sleep duration (deaths/N)	HR	(95% CI)	HR	(95% CI)	HR	(95% CI)
All-cause deaths (381 deaths)
No change (123/824)	1 (reference)
Decrease (104/477)	1.29	(0.99–1.68)	1.26	(0.96–1.64)	1.26	(0.97–1.65)
Increase (144/648)	1.41	(1.10–1.81)	1.41	(1.10–1.82)	1.35	(1.04–1.74)
Breast cancer deaths (132 deaths)
No change (49/824)	1 (reference)
Decrease (28/477)	0.90	(0.56–1.45)	0.87	(0.54–1.41)	0.89	(0.55–1.45)
Increase (55/648)	1.30	(0.85–1.98)	1.29	(0.84–1.98)	1.29	(0.84–2.00)
Non-breast cancer deaths (249 deaths)
No change (84/824)	1 (reference)
Decrease (76/477)	1.49	(1.08–2.04)	1.43	(1.03–1.97)	1.42	(1.02–1.96)
Increase (89/648)	1.44	(1.06–1.97)	1.46	(1.07–2.00)	1.32	(0.96–1.82)

Abbreviations: BMI=body mass index; CI=confidence intervals; HR=hazard ratio; OC=oral contraceptive; PMH=postmenopausal hormone.

Model 1: year of diagnosis, age at diagnosis, time since diagnosis, cancer stage, surgery, chemotherapy, radiation therapy, hormone therapy, prevalent diabetes or heart disease, missing indicators for oncologic treatments, sleep duration pre-diagnosis.

Model 2: Model 1+age, marital status, education level, income, OC use, number of pregnancies, family history of breast cancer, menopausal status, PMH use.

Model 3: Model 2+BMI, alcohol consumption, smoking, caffeine, calories intake, physical activity.

Sample includes women who completed either the 1986 & 2000 or 2002 & 2008 sleep duration assessments and who received their breast cancer in between these two time points. Bold characters indicate statistically significant results. Italic characters indicate the reference group.