Table 3.

Overview of studies on association between breast density and breast cancer-specific survival

Study	Country	Setting	Study design	Breast cancer deaths (N)	Density measurement	Outcome (univariable or minimally adjusted)	Outcome (multivariable)
Olsen [20]	Denmark	Screening interval 2y, age 50-69y, 1991–2001	Cohort. Inclusion of interval and screen-detected cancer deaths.	Fatty: 37 Mixed/dense: 53	Dichotomized measure with Fatty breasts and Mixed/Dense breasts	Dense vs. Fatty Case fatality RR 0.53 (0.34–0.82)a, Mortality: RR 1.78 (95%CI 1.17–2.72)
Olsson [21]	Sweden	Screening interval 18 to 24 m, age 50-69y, 1991–1996	Cohort. Inclusion of non-symptomatic and symptomatic cancer deaths.	Fatty: 7 Moderate: 37 Dense: 32	Modification of the BI-RADS classification with Fatty, Moderate, and Dense breasts	Dense vs. Fatty All: HR 1.92 (0.85–4.35) Non-symptomatic: HR 1.47 (0.41–5.27) Symptomatic: HR 2.33 (0.80–6.80)	Dense vs. Fatty All: HR 2.63 (1.10–6.30)b Non-symptomatic: HR 2.04 (0.49–8.49) Symptomatic: HR 3.40 (1.06–10.90)
Porter [22]	England	Screening interval 3y, age 50-64y, since 1987	Cohort. Inclusion of interval and screen-detected cancer deaths.	BI-RADS 1: 186 BI-RADS 2: 126 BI-RADS 3: 373 BI-RADS 4: 74	BI-RADS breast density	Dense vs. Fatty No significant difference all cancer deaths (p = 0.12), screen-detected (p = 0.75), or interval (p = 0.82)
van Gils [23]	The Netherlands	Screening interval 2y, age 50-69y, 1977–1994	Cohort. Inclusion of screen-detected cancer deaths.	26	Visual assessment fatty (≤ 25%) and dense (> 25%)	Dense vs. Fatty 10-year survival dense 73% vs. fatty 83% (survival p = 0.07) in first rounds, later rounds 10-year survival dense 89% vs. fatty 96% (no significant difference, p = 0.48)
Eriksson [24]	Sweden	Screening interval 24 m, age 50-74y, 1993–1995	Cohort. Inclusion of interval and screen-detected cancer deaths.	Dense: 345 Nondense: 1055	Cumulus dichotomized into nondense (< 25%) and dense (≥ 25%)	Interval vs. Screen-detected All: HR 3.50 (2.25–5.44)c Dense: HR 3.00 (1.26–7.17) Nondense: HR 3.62 (2.17–6.06)	Interval vs. Screen-detected All: HR 1.69 (1.03–2.76)d Dense: HR 1.26 (0.47–3.38) Nondense: HR 1.76 (1.01–3.09)
Gierach [25]	US	Age > 30y, 1996–2005	Cohort. Inclusion of screen-detected, interval, other screen-detected, clinically detected, and other cancer deaths.	BI-RADS 1: 72 BI-RADS 2: 346 BI-RADS 3: 387 BI-RADS 4: 84	BI-RADS breast density	Extremely dense vs. Scattered All: HR 0.83 (0.65–1.07)e	Extremely dense vs. Scattered All: HR 0.92 (0.71–1.19)f
Maskarinec [26]	US	1993–1996	Case-control.	Low density: 12 High density: 15	Cumulus dichotomized into low (< 35%) and high (> 35%)		High vs. Low percent density All: HR 0.83 (0.36–1.92)g
Chiu [27]	Sweden	Age 45-59y, 1977–2004	Cohort. Mammographically detected, interval cancer, and refuser cases.	Dense: 28 Nondense: 99	Tabár dichotomized into dense (IV and V) and nondense (I-III)	Dense vs. Fatty All: HR 1.41 (0.92–2.14)	Dense vs. Fatty All: HR 1.75 (0.99–3.10)h
Zhang [28]	US	1994–2008	Cohort.		BI-RADS breast density dichotomized into dense (3 and 4) and fatty (1 and 2)		Dense vs. Fatty All: HR 0.91 (p-value = 0.1242)i

In brackets behind the HRs are the 95% confidence intervals

^a Case fatality was based on breast cancer death, including all deaths was 0.60. This involves age-adjusted rate ratios

^b Adjusted for age at diagnosis, tumour size, axillary lymph node involvement, Nottingham grade, oestrogen receptor, progesterone receptor, diagnosis year, HRT use, BMI, and mode of detection (if applicable)

^c Adjusted for age, BMI, and hormone replacement therapy (HRT)

^d Adjusted for age, BMI, HRT, and lymph node metastases

^e Adjusted for registry site, age at diagnosis, and diagnosis year

^f Adjusted for registry site, age at diagnosis, diagnosis year, BMI, mode of detection, surgery/radiation, chemotherapy, and income

^g Adjusted for age at diagnosis, ethnicity, obesity, disease stage, and radiation treatment

^h Adjusted for age, tumour size, node status, grade, and BMI

ⁱ Adjusted for age, ethnicity, and tumour stage