Beyond mammography : an evaluation of complementary modalities in breast imaging

Breast cancer is the main cause of cancer death among women worldwide and the goal of mammography screening is to reduce breast cancer-specific mortality. The reduction of the sensitivity of mammography for detecting cancer among women with dense breasts requires the use of complementary methods for this subset of women. Three of the projects in this thesis examine the performance of such complementary methods and a fourth study investigates the association between the biomarker BPE (background parenchymal enhancement) and risk factors for breast cancer. In study 1, we prospectively compared the sensitivity and specificity of Automated Breast Volume Scanner (ABVS) with handheld ultrasound for detection of breast cancer among women with a suspicious mammographic finding who were recalled after attending the population-based mammography screening program. We performed both methods on 113 women and found 26 malignant lesions. Analysis was performed in two categories: breasts with a suspicious screening mammography and breasts with a negative screening mammography. In the first category (n=118) the sensitivity of both methods was 88% (p=1.0), the specificity of handheld ultrasound was 93.5 % and ABVS was 89.2%. The difference in specificity was not statistically significant (p=0.29). For breasts without a suspicious mammographic finding, the sensitivity of handheld ultrasound and ABVS was 100% (p=1.0), the specificity was 100% and 94.1% respectively. The difference in specificity was statistically significant (p=0.03). In summary, ABVS has similar sensitivity to handheld ultrasound, but lower specificity in breasts with a negative mammogram. In study 2, we explored the incremental cancer detection rate when adding a threedimensional infrared imaging (3DIRI) score to screening mammography among women with dense breasts (Volpara volumetric density >6 % on the previous mammography examination) who attended the population-based mammography screening program. Women with a negative mammogram and positive 3DIRI score were triaged for a DCEMRI examination to verify the presence of cancer. Of 1727 participants, 7 women had a mammography-detected breast cancer. Among women with a negative mammogram and a positive infrared imaging (n=219), an additional 6 cancers in 5 women were detected on MRI resulting in an incremental cancer detection rate of 22.5 per 1000. Among women with a negative mammography and infrared examination, one woman was diagnosed with breast cancer during the two-year follow-up. The study does not provide information on the proportion of cancers that might have been detected had MRI been performed among women with a negative mammogram and 3DIRI score. Consequently, this study does not shed light on the diagnostic accuracy of infrared imaging or whether using an infrared risk score is the optimal method for identifying women who would benefit from additional imaging modalities. In study 3, we used MRI examinations of study 2 among women without breast cancer (n=214) to explore the association between BPE at DCE-MRI and a large array of risk factors for breast cancer. Thanks to the Karma database, we had unique access to data from self-reporting questionnaires on risk factors. BPE and mammographic density were assessed visually by three radiologists and BPE was further dichotomized into low and high. We created categorical variables for other risk factors. We calculated the univariable associations between BPE and each risk factor and fitted an adjusted logistic regression model. In the adjusted model, we found a negative association with age (p=0.002), and a positive association with BMI (p=0.03). There was a statistically significant association with systemic progesterone (p=0.03) but since only five participants used progesterone preparations, the result is uncertain. Although the likelihood for high BPE increased with increase in mammographic density, the association was not statistically significant (p=0.23). We were able to confirm earlier findings that BPE is associated with age, BMI and progesterone, but we could not find an association with other risk factors for breast cancer. In study 4, we compared the diagnostic accuracy, reading-time, and inter-rater agreement of an abbreviated protocol (aMRI) to the routine full protocol (fMRI) of contrast-enhanced breast MRI. The MRI examinations were performed before biopsy and among women who were not part of a surveillance program due to an increased familial risk of breast cancer. Analysis was performed on a per breast basis. Aggregated across three readers, the sensitivity and specificity were 93.0% and 91.7% for aMRI, and 92.0% and 94.3% for the fMRI. Using a generalized estimating equations approach to compare the two protocols, the difference in sensitivity was not statistically significant (p=0.840), and the difference in specificity was significant (p=0.003). There was a statistically significant difference in average reading time of 67 seconds for aMRI and 126 seconds for the fMRI (p= 0.000). The inter-rater agreement was 0.79 for aMRI and 0.83 for fMRI. We were able to demonstrate that the abbreviated protocol has similar sensitivity to the full protocol even if MRI is performed before biopsy and the images lack telltale signs of malignancy. In conclusion, this thesis provides new knowledge about the biomarker BPE, broadens our knowledge on the diagnostic accuracy of two different imaging modalities and highlights the importance of good study design for diagnostic accuracy studies