Abstract P2-02-10: Evaluation of apparent diffusion coefficient to predict grade, micro-invasion and invasion in DCIS

Abstract

Abstract PURPOSE: To evaluate the role of apparent diffusion coefficient (ADC) values in distinguishing DCIS grades and identifying the presence of micro-invasive/invasive disease. METHODS AND MATERIALS: REB approved study with informed consent obtained from 70 women (age 36-84) scheduled for core-biopsy with results of 71 non-invasive/high-risk breast lesions. All patients underwent surgery and were assessed pre-operatively using diffusion weighted (DWI)-MRI. Lesion size, morphology and ADC values were recorded. The Kruskal Wallis or one-way ANOVA test and Pearson correlation coefficient were used to study the association between ADC values and the analyzed MRI lesion characteristics. Logistic regression analysis was used to evaluate the ability of ADC values to predict the presence of invasion. RESULTS: Of 71 cases, 45.1% were imaged on a 3T magnet and 54.9% on 1.5 T. Final pathology demonstrated invasive cancer in 26.8%, micro-invasion in 18.3% and pure DCIS in 59.2%. On 3T, mean ADC value was 1.20 ×10-3 mm2/s ± 0.48 (SD) (range, 0.47 - 1.78 ×10-3 mm2/s) for non-high grade DCIS, 1.23 × 10-3 mm2/s ± 0.40 (SD) (range, 0.26 - 1.77 ×10-3 mm2/s) for high-grade DCIS, and 1.15 × 10-3 mm2/s ± 0.45 (SD) (range, 0.26 - 1.75 ×10-3 mm2/s) for invasive/microinvasive disease. On 1.5T, mean ADC value was 1.04 × 10-3 mm2/s ± 0.41 (SD) (range, 0.15 - 1.85 ×10-3 mm2/s) for non-high grade DCIS, 1.01 × 10-3 mm2/s ± 0.37 (SD) (range, 0.06 - 1.76 ×10-3 mm2/s) for high-grade DCIS, and 1.11 × 10-3 mm2/s ± 0.30 (SD) (range, 0.64 - 1.76 ×10-3 mm2/s) for invasive/microinvasive disease. Based on logistic regression analysis, mean ADC value was not a significant predictor for invasiveness using 1.5 T [OR = 2.6 (95% CI (0.409, 17.12)), p = 0.3] or 3T [OR = 0.4 (95% CI (0.076, 2.399)), p = 0.3] CONCLUSION: Mean ADC acquired using a 1.5T or 3T MRI was unable to predict high-grade or invasive disease in biopsy-proven DCIS lesions. Further work is exploring voxel-based approaches that may better appreciate tumor heterogeneity and identify sub-regions of tumor with these higher risk features. Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P2-02-10.