Table 3 Association between breast age gap and future breast cancer using Cox proportional hazards regression models in Inhouse and EMBED datasets

From: Mammo-AGE: deep learning estimation of breast age from mammograms

Breast AGE

Years, mean ± SD

N

Events

Inc.

Model 1 (Unadj.)

Model 2 (CA-ACR-adj.)

Model 3 (RF-adj.)

HR (95% CI)

P value

HR (95% CI)

P value

HR (95% CI)

P value

Inhouse BC events (10Years)

0.51 ± 4.55

10,392

3656

35.2

-

-

-

-

-

-

Age gap, per one age (Years)

-

-

-

-

1.013 (1.006–1.020)

<0.001

1.016 (1.009–1.023)

<0.001

1.014 (1.007-1.022)

<0.001

Low-risk group (Gap <0.36)

−3.08 ± 2.70

5196

1722

33.1

Reference

-

Reference

-

Reference

-

High-risk group (Gap > 0.36)

4.10 ± 2.88

5196

1934

37.2

1.164 (1.091–1.242)

<0.001

1.157 (1.084–1.235)

<0.001

1.140 (1.067–1.217)

<0.001

EMBED BC events (5Years)

0.08 ± 5.025

58,983

1029

3.49

-

-

-

-

-

-

Age gap, per one age (Years)

-

-

-

-

1.022 (1.001–1.034)

<0.001

1.020 (1.008–1.033)

0.002

1.020 (1.008–1.033)

0.002

Low-risk group (Gap < -0.01)

−3.75 ± 2.85

29,492

463

3.14

Reference

-

Reference

-

Reference

-

High-risk group (Gap > -0.01)

3.88 ± 3.06

29,491

566

3.84

1.225 (1.083–1.385)

0.001

1.152 (1.071–1.305)

0.026

1.151 (1.016–1.304)

0.027

  1. P values are calculated using two-sided Wald tests.
  2. Inc incidence per 1000 person-years, CI confidence interval, BC breast cancer, HR hazard ratio, Unadj. HR unadjusted HR, CA-ACR-adj. HR HR adjusted HR on chronological age (CA) and breast density (ACR), RF-adj. HR (Inhouse) HR adjusted HR on more risk factors for inhouse dataset, including CA, ACR, Race, Gene mutation, Menarche, Menopausal status, and Manufacturer of mammograms, RF-adj. HR (EMBED) HR adjusted HR on more risk factors for EMBED dataset, including CA, ACR, Ethnicity, and Manufacturer of mammograms, Breast AGE deep learning-based breast biological age.
Back to article page