Skip to main content

Table 3 Associations between identified dense area features and IC versus SC based on multivariate logistic regression modelling

From: Novel mammographic image features differentiate between interval and screen-detected breast cancer: a case-case study

 

Odds ratios (95 % Cl) for interval versus screen-detected cancer, estimated by logistic regression modelling

Primary cohort

Validation cohort

Covariate

Model 1 n = 1403

Model 2 n = 1314

Model 3 n = 1312

Model 3 n = 1182

Percent density

1.37 (1.22 to 1.53)

1.23 (1.08 to 1.41)

0.98 (0,81 to 1.17)

1.04 (0.86 to 1.24)

BMI

-

0.84 (0.73 to 0.98)

0.88 (0.76 to 1.02)

0.98 (0.84 to 1.15)

HRT use at diagnosis

-

1.57 (1.17 to 2.11)

1.53 (1.14 to 2.07)

1.32 (0.99 to 1.77)

Age at diagnosis

-

0.98 (0.87 to 1.12)

1.01 (0.89 to 1.15)

0.90 (0.77 to 1.05)

F40 skewness of the intensity gradient

-

-

1.32 (1.12 to 1.56)

1.21 (1.04 to 1.41)

F21 eccentricity

-

-

1.20 (1.04 to 1.39)

1.17 (0.98 to 1.39)

  1. Feature values calculated based on the dense area of the mammogram as identified by Otsu's method, then Box-Cox transformed and standardized. Odds ratios are estimated as per-standard deviation change in the underlying covariate. Validation cohort is an older breast cancer cohort with similar covariate definitions as the primary cohort
  2. 1C interval breast cancer, SC screen-detected breast cancer, 95 % CI 95 % confidence interval, BMI body mass index, HRT hormone replacement therapy