This poster is published under an
open license. Please read the
disclaimer for further details.
Keywords:
Breast, Mammography, Observer performance, Comparative studies, Cancer, Epidemiology, Image registration
Authors:
R. R. Winkel1, M. von Euler-Chelpin1, M. Nielsen2, M. Bachmann Nielsen1, W. Uldall1, P. Diao1, I. Vejborg1; 1Copenhagen/DK, 2Copenhagen /DK
DOI:
10.1594/ecr2015/C-1037
Methods and materials
This retrospective nested case-control study included 122 cases and 262 age-matched controls based on all 14,736 women with negative screening mammograms from a population-based screening service in Copenhagen in 2007 (follow-up period 2007-2011).
Use of screening data and tumour-related information was approved by the Danish Data Inspection Agency (2013-41-1604).
Film-based craniocaudal (CC) and mediolateral oblique (MLO) images (right and left) from the screening in 2007 were digitized.
The randomized mammograms were classified independently by two MDs (a senior breast radiologist and a resident in radiology) according to two visual classifications—the BI-RADS density classification (4th edition,
2003; Figure 1) ) [5] and the Tabár classification on parenchymal patterns (Figure 2) [4],
[9]—as well as a computerized interactive threshold technique measuring area-based percent mammographic density (denoted PMD; Figure 3) [10].
Readings were done without knowledge of the woman’s age at screening or (later) cancer status.
The right and the left breasts were read independently and considered independent measurements.
Readings by the three different methodologies were completed at different stages in a MatLab scoring-database,
where the readers were blinded from evaluations by the other classifications in order to reduce artificial agreement between the methods.
Statistics:
Inter-observer consistency was investigated on both a multiple-category scale and on a high/low-risk scale based on all independently scored right and left mammograms.
High-risk density was defined as: BI-RADS: D3 and D4,
Tabár: PIV and PV and the upper two quartiles of PMD (within density range corresponding to the BI-RADS classification).
Kappa statistics and the Intraclass Correlation Coefficient (ICC) (which is also equivalent to the weighted kappa) were used to evaluate inter-observer agreement.
The association between mammographic density/parenchymal pattern and breast cancer risk was estimated using logistic regression to calculate odds ratios (OR) adjusted for age.
The higher density groups were compared individually with the lowest density group for each method (reference categories: BIRADS: D1; Tabár: PII; PMD: the lowest quartile).
In addition,
dichotomized high and low-risk groups were compared.
Exact two-sided P-values and 95% confidence intervals (95% CI) have been listed and results were considered statistically significant with P-values ≤ 0.05.