Mammographic Texture Resemblance generalizes as an independent risk factor of breast cancer

Konstantin Chernoff, S G Christopher, Gopal Karemore, N Karssemeijer, Mads Nielsen, C M Vachon

Abstract

PURPOSE
Breast density has been established as a risk factor of breast cancer in numerous studies. Mammographic Texture Resemblance (MTR) has shown to be a density independent risk factor, but only on a single study. We examine if the statistics of the texture recorded in one study generalize as an independent risk factor in an unrelated cohort.

METHOD AND MATERIALS
The statistics of texture were recorded in digitalized film-mammograms of one 4-year prospective study (S1, Dutch screening program) of 245 breast cancers and 250 matched controls. From an independent cohort study (S2, Mayo Mammography Health Study cohort) 226 incident breast cancer cases diagnosed through 2008 and 442 matched controls (on age) were used for scoring screening digitized mammograms that were ascertained years prior to diagnosis 1993-2006. Mammographic percent density (PD), using Cumulus, and other major risk factors were ascertained in S2. Finally S2 was MTR scored based on textures from S1 and S2 in a leave-two-out fashion.
Scores on S2 were related to future breast cancer incidence by AUC and analyses of quartiles adjusted for BMI, menopause status, and postmenopausal hormone (PMH) use. A combined density and MTR model was also evaluated.

RESULTS
The MTR scores on S1 showed in a cross validation fashion significant capability to separate cancers from controls (AUC=0.63±0.02, p<0.001). This persisted after adjustment for mammographic density. Age of S1 was 58.0±5.7 years while age in S2 was younger with higher variation: 55.2±10.5 years. No significant difference in BMI, menopause status, or PMH was found between cases and matched controls in S2. S2 showed an AUC of 0.633, 0.613, and 0.600 based on respectively percentage density, MTR scores trained on S2, and MTR scores trained on S1 respectively. Adjusted for PD, MTR scores of S2 trained on S1 showed OR in quartiles of ref; 1.10 (0.64-1.89); 0.93 (0.52-1.68); 1.96 (1.19-3.23) respectively and a combined AUC of 0.654.

CONCLUSION
The heterogeneities that are recorded to characterize breast cancer risk in S1 were also a density independent risk factor in S2. Hence the textural patterns that indicated elevated risk persisted under differences in x-ray technology, population demographics, follow up time, and geography.
OriginalsprogEngelsk
Publikationsdato2011
StatusUdgivet - 2011
Begivenhed5th International Workshop on Breast Densitometry and Breast Cancer Risk Assessment - San Francisco, USA
Varighed: 9 jun. 2011 → …

Konference

Konference5th International Workshop on Breast Densitometry and Breast Cancer Risk Assessment
Land/OmrådeUSA
BySan Francisco
Periode09/06/2011 → …

Citationsformater