Purpose: Screening of women at intermediate risk of breast cancer (BC) requires yearly mammography and US; screening of high-risk women requires additional yearly MRI. Our goal is to verify the subjective interpretation of BC risks against a validated risk-estimation model (IBIS). Thus, to obtain more effective breast prevention strategies.
Methods and Materials: We retrospectively reviewed 471 clinically medium- to high-risk women who underwent screening mammography from August 2014 to August 2016. Because of presence of a BC gene mutation (33) or a lack of records (7), 40 women were excluded. Of the 431 remaining patients, 267 (62%) and 33 (14%) received additional screening US and MRI, respectively. Data from each included patient necessary to calculate the IBIS-BC-risk were inserted into the latest version (8.0b) of the risk- evaluation tool.
Results: The median age is 51 years and the median IBIS risk score is 23%. There were 6 (14%) breast cancer cases discovered through the screening. The personal lifetime risks are distributed as follows: <17% in 111, 17%-20% in 46, 20%-25% in 99, 25-30% in 71 and ≥30% in 104 women.
Conclusion: One fourth of the included patients showed a low IBIS-BC risk (<17%), lifetime risk of BC was clinically overestimated: they should not be screened more intensely. Considering an IBIS score of 25% as reference for breast MRI, 40% of our study group should receive breast MRI, instead of the 14% actually. More effective breast prevention strategies could be obtained by an objectification of the BC risk assessment.
Purpose: This study evaluated the consistency and discriminatory power of short-term breast cancer risk models with and without biopsy history (BxHx) within a general screening population.
Methods and Materials: All screen-detected breast cancer cases among digitally screened women 40-75 years (2009-2015) within a population-based breast screening program and 3 age- and screen year-matched controls were sampled. Clinical risk factors, fully automated percent mammographic density (PMD), and breast volume assessments were obtained for 1,593 cases and 5,003 controls and used to derive patient-specific risk estimates from a series of logistic regression models. Predictive performance was assessed using area under receiver operator characteristic (AUROC), and agreement between models for assigning women to low (<90th percentile)- versus high (≥ 90th percentile)-risk groups was assessed using weighted kappa.
Results: Predictive performance of the multivariate models varied substantially (AUROC: 0.547-0.655). A reduced model with PMD, breast volume, age, family history, and BxHx performed equivalently to the full model that additionally included menopausal status and HRT use (AUROC=0.655 and 0.656, respectively); removing BxHx from the reduced model decreased performance (AUROC= 0.591). Agreement between predicted probabilities of the full versus reduced model classified into low versus high risk was almost perfect (kappa=0.982).
Conclusion: A short-term risk model incorporating PMD, breast volume, family history, BxHx, and age may provide a practical solution for risk stratification within a screening population without the need to collect other clinical risk factors that are prone to recall bias and are not always available.
Purpose: To quantify the reproducibility and measurement error of volumetric breast density in mammograms.
Methods and Materials: We analysed data from the Predicting Risk of Cancer at Screening Study (PROCAS), a large cohort study to estimate risk of breast cancer in women attending routine breast screening. We used data from post-menopausal women who had a repeat mammographic view taken within 28 days of their screen on entry to PROCAS; all mammograms were taken on GE Senographe Essential systems. Volumetric percentage breast density for each mammographic view was assessed using Volpara V1.5.4 and log transformed. Measurement error was assessed by calculating the coefficient of variation and reproducibility assessed using the intraclass correlation coefficient (ICC).
Results: 519 repeat views from 419 women were eligible for analysis. The mean age of women was 60 years (range 46-73 years) and the median number of days between images was 18 days (interquartile range 14-21 days). The ICC (95% confidence interval [CI]) was 0.88 (0.84, 0.91), 0.89 (0.85, 0.92), 0.89 (0.84, 0.93) and 0.93 (0.90, 0.95) for the left and right craniocaudal (CC) and mediolateral oblique (MLO) views respectively. The coefficient of variation was 16.7% (15.1, 18.8), 16.4% (14.8, 18.5), 15.8% (13.9, 18.4) and 13.8% (95% CI: 12.1, 16.1) for LCC, RCC, LMLO and RMLO views, respectively.
Conclusion: This study found good to excellent reproducibility and a measurement error of 14-17% in assessing volumetric percentage breast density. This should be taken into account when interpreting changes in breast density between successive measurements.
Purpose: To identify the various radiological and clinical factors associated with disagreement between automated volumetric breast density measurement (VBDM) and visual assessment by radiologists in assessment of breast density.
Methods and Materials: Three thousand and forty-three women who underwent screening and diagnostic mammography from August 2016 and February 2017 were included. The agreement in breast density between visual assessment by the radiologist based on 5th BI-RADS and VBDM (Volpara Version 3.1) were compared using a weighted kappa (k) value. The factors including patient age, the purpose of mammography, presence of mass, microcalcifications, macrocalcifications, asymmetry or architectural distortion, a difference in bilateral breast density and BI-RADS final assessment were evaluated using univariate and multivariate analyses.
Results: Among 3043 women, 873 (28.7%) showed disagreement. The agreements between visual assessment by radiologist and VBDM were substantial (weighted k value = 0.674). Univariate analysis showed patient age (p <0.001), purpose of mammography (p = 0.026) and a difference in bilateral density (p < 0.001) as factors contributing to disagreement. In multivariate analysis, patient age (p = 0.003), presence of mass (p = 0.016) and a difference in bilateral density (p < 0.001) were contributing factors for disagreement.
Conclusion: There is substantial agreement in breast density evaluation between VBDM and visual assessment by radiologists. Disagreement between VBDM and visual assessment was related to patient age and a difference in bilateral breast density.
Purpose: This study evaluated the predictive performance of short-term interval breast cancer risk models within a general screening population.
Methods and Materials: This case control study was performed among digitally screened women aged 40-75 (2009-2015) within a provincial breast screening program in Canada. The sample included all 132 interval breast cancer cases and 885 controls. Interval breast cancer was defined as breast cancer diagnosed after a negative screening examination or after an abnormal screening examination with negative work-up but before the next regularly scheduled screening examination. Data on clinical risk factors including age, breast volume (as a surrogate for BMI), first degree family history, history of breast biopsy, menopausal status, and HRT use were obtained for all subjects. Percent mammographic density (PMD) and breast volume assessments were obtained via automated software (Densitas Inc.). Logistic regression models were used to derive patient-specific risk estimates. Predictive performance was assessed using the area under receiver operator characteristic (AUROC).
Results: There was no difference in the average age of cases and controls at 55 and 56 years respectively. The model with PMD alone outperformed a model with all other clinical risk factors combined (AUROC=0.679 vs 0.614, respectively). Adding PMD to a model with all the other clinical risk factors increased the AUROC to 0.716.
Conclusion: Percent mammographic density was the most significant predictor of interval detected cancers and was a stronger predictor of interval detected cancers than all other clinical risk factors combined. Model performance needs to be validated using a separate dataset.
Purpose: To investigate differences in parenchymal pattern (PP) of breast tissue on screening mammograms and evaluate association with screen-detected or interval cancer and node status.
Methods and Materials: This case-control study of 1204 women age 50-74 included 302 screen-detected and 297 interval cancers (239 node positive, 360 node negative) and 605 controls. Three readers classified PP grade on prior mammograms from smooth (PP1) to nodular (PP5). Volpara software calculated fibroglandular volume (FGV) and volumetric breast density (VBD) on unprocessed images. Intraclass correlation (ICC) compared readers’ PP grades. Trend analysis was performed after one-way ANOVA to test for linear trends between increased PP and mean VBD, and mean FGV. Conditional logistic regression determined whether PP could predict mode of detection (screen detected or interval), and node status (positive or negative).
Results: There was good correlation between readers for PP grade (ICC 0.736, 95% CI 0.713, 0.757). Mean VBD did not differ significantly with increased PP grade. Mean FGV linearly increased with PP grade (p<0.0001). The relative risk (RR) of cancer in PP5 vs. PP1 was 2.2 (p<0.01) for screen detected cancer and 2.5 (p<0.01) for interval cancers. In PP5 vs. PP1, the odds ratio (OR) of node positive cancer vs. controls was 1.8 (non-significant, p=0.16) and for node negative cancer was 3.0 (p<0.01).
Conclusion: Visual assessment of PP is reproducible. PP grade increases with FGV and age. RR of screen-detected, interval, and node negative cancer significantly increases with PP grade, though the RR of node positive cancer does not.
Purpose: Evaluate the association between Body Mass Index (BMI) and a subsequent breast cancer, among women with a prior negative recall after additional imaging and ultrasound only (false-positive without biopsy) in BreastScreen Norway.
Methods and Materials: BreastScreen Norway is a population based screening program that invites all Norwegian women to biennial independently double-read two-view mammography in birth cohorts corresponding to age 50-69 years. 3,201,915 screens (97% negative and 3% resulting in a recall) were performed among 763,470 women in the study period 1995-2016. We included 55,625 women who experienced a recall concluded negative after additional imaging and ultrasound only. We followed women for invasive breast cancer (ipsilateral and/or contralateral), from false-positive until end of follow-up (31.12.2016, date of death, emigration or diagnosis of breast cancer, whichever came first). Cox regression was used to study how BMI was associated with subsequent breast cancer among women with a prior false-positive without biopsy, adjusted for age.
Results: Compared to women of normal weight (BMI 18.5-25), underweight women (BMI <18.5) had a similar long-term risk of subsequent breast cancer (HR=1.07, 95% CI: 0.63-1.83) after a prior false-positive without biopsy. We observed a statistically significant increased long-term risk of subsequent breast cancer (HR=1.25, 95% CI: 1.09-1.43 and HR=1.43; 95% CI: 1.19-1.71) among overweight women (BMI 25-30) and obese women (BMI ≥30).
Conclusion: Among women with a prior false-positive without biopsy in BreastScreen Norway, overweight and obese women have a 25% and 43%, respectively, increased long-term risk of a subsequent breast cancer compared to women of normal weight.
Purpose: The objective of our study was to assess the performance of the new imaging techniques: full-field digital mammography (FFDM), 3D mammography and automated breast ultrasound (ABUS) in the detection of breast cancer.
Methods and Materials: Two radiologists independently evaluated a total of 127 ABUS acquisitions, the FFDM and 3D mammograms of women with dense breast tissue. During the study no clinical information or patient history was provided to the readers. The results were compared to the gold standard: histopathology for biopsied lesions, HHUS-handheld ultrasound for typically benign lesions (cysts) and follow-up for benign appearing lesions unchanged for at least 2 years.
Results: Nineteen breast cancers were proved by biopsy. For FFDM alone the sensibility was 72,7%, the specificity and the positive predictive value was 100% and the negative predictive value was 92,3%. By completing the FFDM with ABUS the sensibility increased to 78,3%, the specificity was 91,8%, the positive predictive value was 69,2% and the negative predictive value was 94,7%. For 3D mammography the sensibility was 100%, the specificity 86,4%, the positive predictive value was 75% and the negative predictive value was 100%.
Conclusion: In screening, ABUS added to FFDM compared with FFDM alone, improved reader’s detection of breast cancers in women with dense breast tissue, but did not exceed 3D mammography. As a diagnostic method, ABUS associated with FFDM outperformed 3D mammography.
Purpose: To compare the added value of automated breast ultrasound (ABUS) versus breast tomosynthesis (BT) as diagnostic tools in the work-up of screening detected positive findings in dense breasts recalled from the national screening program.
Methods and Materials: After ethics committee approval, and patients' consent, 242 women with dense breasts who underwent screening mammography, and were recalled for suspected positive findings were enrolled in the study. Positive findings included focal asymmetry, mass, distortion, or microcalcifications. All patients underwent both BT and ABUS by two independent breast radiologists and a BIRADS score was given for each modality. BT was performed in CC and MLO views. ABUS images were acquired in anteroposterior, lateral and medial views. Images were interpreted in the coronal 3D view using the survey mode, followed by the transverse and sagittal reconstructed images. Results were compared to pathology and follow-up of negative/typically benign findings
Results: Sensitivity, specificity, PPV, NPV, LR positive, LR negative and accuracy of ABUS were 92, 98, 92, 98, 44, 0.08 and 97, respectively, and in BT were 92, 92, 76, 98,12, 0.09 and 92. Agreement by Kappa was 0.896. ABUS and BT both agreed on TP in 43 cases out of 51 proved cancers. There were 4 FN cases in each modality. There were 4 FP by ABUS and 15 by BT. Biopsy was avoided by ABUS in 187 cases and by BT in 176 cases.
Conclusion: ABUS has shown a higher accuracy than BT. Its main limitations are microcalcifications and the retroareolar region.
Purpose: There is currently no standard of care for screening asymptomatic men who are at increased risk for developing breast cancer. The purpose of this study is to investigate the utility of mammography for breast cancer screening in men at increased risk for breast cancer.
Methods and Materials: In this HIPAA compliant IRB-approved single-institution study, mammography, pathology and clinical records on 827 men who underwent mammography between 09/2011 and 07/2018 were analysed via the electronic medical record. 664 of these men presented with masses, pain or nipple discharge and were excluded. The remaining 163 asymptomatic men with family and/or personal history of breast cancer and/or known breast cancer mutation underwent screening mammography and are the subject of this analysis.
Results: 163 asymptomatic men 24-87 years (median 66 years) underwent 806 screening mammograms. 125/163 (76%) had a personal history and 72/163 (44%) had a family history of breast cancer. 24/163 (15%) had known mutations: 4/24 (17%) BRCA1 and 20/24 (83%) BRCA2. 792/806 (98%) of screening mammograms were negative (BI-RADS 1 or 2); 10/806 (1.2%) were BI-RADS 3, all of which were subsequently downgraded to BI-RADS 2. 4/806 (0.4%) mammograms were suspicious (BI-RADS 4/5) and all were malignant (mean size 6.25mm, range 4-7mm). There were three T1 and one T4 breast cancer. The estimated cancer detection rate was 4.9/1,000 examinations.
Conclusion: Screening mammography in high-risk men yielded a cancer detection rate of 4.9/1000 examinations.