RSNA 2018

Abstract Archives of the RSNA, 2018


SSE01-03

Using Quantitative Breast Density Analysis to Predict Interval Cancers and Node Positive Cancers in Pursuit of Improved Screening Protocols

Monday, Nov. 26 3:20PM - 3:30PM Room: E451B



Participants
Elizabeth S. Burnside, MD,MPH, Madison, WI (Presenter) Dr. Burnside has a research grant from Hologic
Lucy M. Warren, PhD, Guildford, United Kingdom (Abstract Co-Author) Nothing to Disclose
Louise S. Wilkinson, MBBCh,FRCR, London, United Kingdom (Abstract Co-Author) Nothing to Disclose
Kenneth C. Young, PhD, Guildford, United Kingdom (Abstract Co-Author) Nothing to Disclose
Jonathan Myles, London, United Kingdom (Abstract Co-Author) Nothing to Disclose
Stephen W. Duffy, London, United Kingdom (Abstract Co-Author) Nothing to Disclose

PURPOSE

This study investigates whether quantitative breast density can predict interval cancers and node positive screen detected cancers in order to serve as a biomarker to consider more aggressive screening to improve early detection.

METHOD AND MATERIALS

We conducted a case-control study of 1204 women drawn from the U.K. NHS Breast Screening Program aged 50-74 including 599 cases (comprising 302 screen detected cancers, 297 interval cancers; 239 node positive, 360 node negative) and 605 controls. Each woman had prior digital mammograms and 70% had unprocessed images. A radiologist assessed breast density using a visual analog scale (VAS) from 0 to 100 and BI-RADS 5th Edition density categories. Volpara software (V1.5.1) calculated fibroglandular volume (FGV) and volumetric density grade (VDG) on unprocessed images. Logistic regression determined whether the breast density measures could predict mode of detection (screen detected or interval); node-negative cancers; and node-positive cancers, all vs. controls.

RESULTS

FGV predicted both screen-detected (p<0.01) and interval cancers (p<0.01) compared to controls. VDG, VAS and BI-RADS predicted interval cancers (all p<0.01) but not screen-detected cancers (p=0.16, p=0.18, p=0.46 resp.). FGV demonstrated impressive risk stratification with an age-adjusted relative risk (RR) of the 4th quartile compared to the 1st quartile of 3.7 overall, 2.8 for screen detected, and 5.3 for interval cancers. VDG also had notable risk stratification with an age-adjusted RR of 3.6 for interval cancers (Table). FGV predicted node-negative cancers as compared to controls (p<0.01) while BI-RADS, VAS, and VDG did not (p=0.07, p=0.09, and p=0.47 resp.). FGV, BI-RADS, and VDG predicted node-positive cancers (all p<0.01) while VAS did not (p=0.14).

CONCLUSION

FGV predicts interval, screen detected, node-positive and node-negative cancers compared to controls and provides remarkable stratification the RR of interval cancers. BI-RADS and VDG predict interval and node positive cancers. VAS only predicts interval cancers. The quantitative and automated nature of FGV and VDG and notable risk stratification based on RR indicates that these variables may be promising biomarkers.

CLINICAL RELEVANCE/APPLICATION

By predicting mode of detection and nodal status, FGV may be a biomarker for more intensive screening. By predicting interval cancers, BI-RADS, VAS, and VDG may act as supplementary biomarkers.