Contrast-enhanced mammography for surveillance in women with a personal history of breast cancer.

PURPOSE: Women with a personal history of breast cancer have an increased risk of subsequent breast malignancy and may benefit from more sensitive surveillance than conventional mammography (MG). We previously reported outcomes for first surveillance episode using contrast-enhanced mammography (CEM), demonstrating higher sensitivity and comparable specificity to MG. We now report CEM performance for subsequent surveillance. METHODS: A retrospective study of 1,190 women in an Australian hospital setting undergoing annual surveillance following initial surveillance CEM between June 2016 and December 2022. Outcome measures were recall rate, cancer detection rate, contribution of contrast to recalls, false positive rate, interval cancer rate and characteristics of surveillance detected and interval cancers. RESULTS: 2,592 incident surveillance episodes were analysed, of which 93% involved contrast-based imaging. Of 116 (4.5%) recall episodes, 40/116 (34%) recalls were malignant (27 invasive; 13 ductal carcinoma in situ), totalling 15.4 cancers per 1000 surveillance episodes. 55/116 (47%) recalls were contrast-directed including 17/40 (43%) true positive recalls. Tumour features were similar for contrast-directed recalls and other diagnoses. 8/9 (89%) of contrast-directed invasive recalls were Grade 2-3, and 5/9 (56%) were triple negative breast cancers. There were two symptomatic interval cancers (0.8 per 1000 surveillance episodes, program sensitivity 96%). CONCLUSION: Routine use of CEM in surveillance of women with PHBC led to an increase in the detection of clinically significant malignant lesions, with a low interval cancer rate compared to previous published series. Compared to mammographic surveillance, contrast-enhanced mammography increases the sensitivity of surveillance programs for women with PHBC.

Contrast enhanced mammography in breast cancer surveillance.

PURPOSE: Mammography (MG) is the standard imaging in surveillance of women with a personal history of breast cancer or DCIS (PHBC), supplemented with ultrasound. Contrast Enhanced Mammography (CEM) has higher sensitivity than MG and US. We report the performance of CEM compared with MG ± US. METHODS: A retrospective study of patients undergoing their first surveillance CEM in an Australian hospital setting between June 2006 and October 2020. Cases where a patient was recalled for assessment were identified, recording radiology, pathology and treatment details. Blinded re-reading of recalled cases was performed to determine the contribution of contrast. Use of surveillance US across the board was assessed for the period. RESULTS: 73/1191 (6.1%) patients were recalled. 35 (48%) were true positives (TP), with 26 invasive cancers and 9 cases of DCIS, while 38 (52%) were false positive (FP) with a positive predictive value (PPV) 47.9%. 32/73 were recalled due to MG findings, while 41/73 were only recalled due to Contrast. 14/73 had 'minimal signs' with a lesion identifiable on MG with knowledge of the contrast finding, while 27/73 were visible only with contrast. 41% (17/41) recalled due to contrast were TP. Contrast-only TPs were found with low and high mammographic density (MD). Screening breast US reduced by 55% in the year after CEM was implemented. CONCLUSION: Compared to MG, CEM as a single surveillance modality for those with PHBC has higher sensitivity and comparable specificity, identifying additional malignant lesions that are clinically significant. Investigation of interval cancer and subsequent round outcomes is warranted.

The impact of the Covid-19 pandemic on breast cancer early detection and screening.

The COVID-19 pandemic affects mortality and morbidity, with disruptions expected to continue for some time, with access to timely cancer-related services a concern. For breast cancer, early detection and treatment is key to improved survival and longer-term quality of life. Health services generally have been strained and in many settings with population breast mammography screening, efforts to diagnose and treat breast cancers earlier have been paused or have had reduced capacity. The resulting delays to diagnosis and treatment may lead to more intensive treatment requirements and, potentially, increased mortality. Modelled evaluations can support responses to the pandemic by estimating short- and long-term outcomes for various scenarios. Multiple calibrated and validated models exist for breast cancer screening, and some have been applied in 2020 to estimate the impact of breast screening disruptions and compare options for recovery, in a range of international settings. On behalf of the Covid and Cancer Modelling Consortium (CCGMC) Working Group 2 (Breast Cancer), we summarize and provide examples of such in a range of settings internationally, and propose priorities for future modelling exercises. International expert collaborations from the CCGMC Working Group 2 (Breast Cancer) will conduct analyses and modelling studies needed to inform key stakeholders recovery efforts in order to mitigate the impact of the pandemic on early diagnosis and treatment of breast cancer.

Atypical ductal hyperplasia is a multipotent precursor of breast carcinoma.

The current model for breast cancer progression proposes independent 'low grade (LG)-like' and 'high grade (HG)-like' pathways but lacks a known precursor to HG cancer. We applied low-coverage whole-genome sequencing to atypical ductal hyperplasia (ADH) with and without carcinoma to shed light on breast cancer progression. Fourteen out of twenty isolated ADH cases harboured at least one copy number alteration (CNA), but had fewer aberrations than LG or HG ductal carcinoma in situ (DCIS). ADH carried more HG-like CNA than LG DCIS (e.g. 8q gain). Correspondingly, 64% (7/11) of ADH cases with synchronous HG carcinoma were clonally related, similar to LG carcinoma (67%, 6/9). This study represents a significant shift in our understanding of breast cancer progression, with ADH as a common precursor lesion to the independent 'low grade-like' and 'high grade-like' pathways. These data suggest that ADH can be a precursor of HG breast cancer and that LG and HG carcinomas can evolve from a similar ancestor lesion. We propose that although LG DCIS may be committed to a LG molecular pathway, ADH may remain multipotent, progressing to either LG or HG carcinoma. This multipotent nature suggests that some ADH cases could be more clinically significant than LG DCIS, requiring biomarkers for personalising management. Copyright © 2019 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Molecular comparison of interval and screen-detected breast cancers.

Breast cancer (BC) diagnosed after a negative mammogram but prior to the next screening episode is termed an 'interval BC' (IBC). Understanding the molecular differences between IBC and screen-detected BCs (SDBC) could improve mammographic screening and management options. Therefore, we assessed both germline and somatic genomic aberrations in a prospective cohort. Utilising the Lifepool cohort of >54 000 women attending mammographic screening programs, 930 BC cases with screening status were identified (726 SDBC and 204 IBC). Clinico-pathological and family history information were recorded. Germline and tumour DNA were collected where available and sequenced for BC predisposition and driver gene mutations. Compared to SDBC, IBCs were significantly associated with a younger age at diagnosis and tumour characteristics associated with worse prognosis. Germline DNA assessment of BC cases that developed post-enrolment (276 SDBCs and 77 IBCs) for pathogenic mutations in 12 hereditary BC predisposition genes identified 8 carriers (2.27%). The germline mutation frequency was higher in IBC versus SDBC, although not statistically significant (3.90% versus 1.81%, p = 0.174). Comparing somatic genetic features of IBC and SDBC matched for grade, histological subtype and hormone receptor revealed no significant differences, with the exception of higher homologous recombination deficiency scores in IBC, and copy number changes on chromosome Xq in triple negative SDBCs. Our data demonstrates that while IBCs are clinically more aggressive than SDBC, when matched for confounding clinico-pathological features they do not represent a unique molecular class of invasive BC, but could be a consequence of timing of tumour initiation and mammographic screening. Copyright © 2019 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Treatment Intensity Differences After Early-Stage Breast Cancer (ESBC) Diagnosis Depending on Participation in a Screening Program.

BACKGROUND: While population mammographic screening identifies early-stage breast cancers (ESBCs; ductal carcinoma in situ [DCIS] and invasive disease stages 1-3A), commentaries suggest that harms from overdiagnosis and overtreatment may outweigh the benefits. Apparent benefits to patients with screen-detected cancers may be due to selection bias from exclusion of interval cancers (ICs). Treatment intensity is rarely discussed, with an assumption that all ESBCs are treated similarly. We hypothesized that women diagnosed while in a screening program would receive less-intense treatment than those never or not recently screened (NRS). METHODS: This was a retrospective analysis of all women aged 50-69 years managed for ESBC (invasive or DCIS) during the period 2007-2013 within a single service, comparing treatment according to screening status. Data on demographics, detection, pathology, and treatment were derived from hospital, cancer registry, and screening service records. RESULTS: Overall, 622 patients were active screeners (AS) at diagnosis (569 screen-detected and 53 ICs) and 169 patients were NRS. AS cancers were smaller (17 mm vs. 26 mm, p < 0.0001), less likely to involve nodes (26% vs. 48%, p < 0.0001), and lower grade. For invasive cancer, NRS patients were more likely to be recommended for mastectomies [35% vs. 16%; risk ratio(RR) 2.2, p < 0.0001], axillary dissection (43% vs. 19%; RR 2.3, p < 0.0001), adjuvant chemotherapy (65% vs. 37%; RR 1.7, p < 0.0001), and postmastectomy radiotherapy (58% vs. 39%; RR 1.5, p = 0.04). CONCLUSION: Participants in population screening diagnosed with ESBC receive substantially less-intense treatment than non-participants. Differences persist when potential overdiagnosis is taken into account; these differences should be factored into debates around mammographic screening.

Application of Machine Learning Models to Evaluate Hypoglycemia Risk in Type 2 Diabetes.

INTRODUCTION: To identify predictors of hypoglycemia and five other clinical and economic outcomes among treated patients with type 2 diabetes (T2D) using machine learning and structured data from a large, geographically diverse administrative claims database. METHODS: A retrospective cohort study design was applied to Optum Clinformatics claims data indexed on first antidiabetic prescription date. A hypothesis-free, Bayesian machine learning analytics platform (GNS Healthcare REFS™: Reverse Engineering and Forward Simulation) was used to build ensembles of generalized linear models to predict six outcomes defined in patients' 1-year post-index claims history, including hypoglycemia, antidiabetic class persistence, glycated hemoglobin (HbA1c) target attainment, HbA1c change, T2D-related inpatient admissions, and T2D-related medical costs. A unified set of 388 variables defined in patients' 1-year pre-index claims history constituted the set of predictors for all REFS models. RESULTS: The derivation cohort comprised 453,487 patients with a T2D diagnosis between 2014 and 2017. Patients with comorbid conditions had the highest risk of hypoglycemia, including those with prior hypoglycemia (odds ratio [OR] = 25.61) and anemia (OR = 1.29). Other identified risk factors included insulin (OR = 2.84) and sulfonylurea use (OR = 1.80). Biguanide use (OR = 0.75), high blood glucose (> 125 mg/dL vs. < 100 mg/dL, OR = 0.47; 100-125 mg/dL vs. < 100 mg/dL, OR = 0.53), and missing blood glucose test (OR = 0.40) were associated with reduced risk of hypoglycemia. Area under the curve (AUC) of the hypoglycemia model in held-out testing data was 0.77. Patients in the top 15% of predicted hypoglycemia risk constituted 50% of observed hypoglycemic events, 26% of T2D-related inpatient admissions, and 24% of all T2D-related medical costs. CONCLUSIONS: Machine learning models built within high-dimensional, real-world data can predict patients at risk of clinical outcomes with a high degree of accuracy, while uncovering important factors associated with outcomes that can guide clinical practice. Targeted interventions towards these patients may help reduce hypoglycemia risk and thereby favorably impact associated economic outcomes relevant to key stakeholders.

Evaluation of machine learning methodology for the prediction of healthcare resource utilization and healthcare costs in patients with critical limb ischemia-is preventive and personalized approach on the horizon?

BACKGROUND: Critical limb ischemia (CLI) is a severe stage of peripheral arterial disease and has a substantial disease and economic burden not only to patients and families, but also to the society and healthcare systems. We aim to develop a personalized prediction model that utilizes baseline patient characteristics prior to CLI diagnosis to predict subsequent 1-year all-cause hospitalizations and total annual healthcare cost, using a novel Bayesian machine learning platform, Reverse Engineering Forward Simulation™ (REFS™), to support a paradigm shift from reactive healthcare to Predictive Preventive and Personalized Medicine (PPPM)-driven healthcare. METHODS: Patients ≥ 50 years with CLI plus clinical activity for a 6-month pre-index and a 12-month post-index period or death during the post-index period were included in this retrospective cohort of the linked Optum-Humedica databases. REFS™ built an ensemble of 256 predictive models to identify predictors of all-cause hospitalizations and total annual all-cause healthcare costs during the 12-month post-index interval. RESULTS: The mean age of 3189 eligible patients was 71.9 years. The most common CLI-related comorbidities were hypertension (79.5%), dyslipidemia (61.4%), coronary atherosclerosis and other heart disease (42.3%), and type 2 diabetes (39.2%). Post-index CLI-related healthcare utilization included inpatient services (14.6%) and ≥ 1 outpatient visits (32.1%). Median annual all-cause and CLI-related costs per patient were $30,514 and $2196, respectively. REFS™ identified diagnosis of skin and subcutaneous tissue infections, cellulitis and abscess, use of nonselective beta-blockers, other aftercare, and osteoarthritis as high confidence predictors of all-cause hospitalizations. The leading predictors for total all-cause costs included region of residence and comorbid health conditions including other diseases of kidney and ureters, blindness of vision defects, chronic ulcer of skin, and chronic ulcer of leg or foot. CONCLUSIONS: REFS™ identified baseline predictors of subsequent healthcare resource utilization and costs in CLI patients. Machine learning and model-based, data-driven medicine may complement physicians' evidence-based medical services. These findings also support the PPPM framework that a paradigm shift from post-diagnosis disease care to early management of comorbidities and targeted prevention is warranted to deliver a cost-effective medical services and desirable healthcare economy.

The TP53 mutation rate differs in breast cancers that arise in women with high or low mammographic density.

Mammographic density (MD) influences breast cancer risk, but how this is mediated is unknown. Molecular differences between breast cancers arising in the context of the lowest and highest quintiles of mammographic density may identify the mechanism through which MD drives breast cancer development. Women diagnosed with invasive or in situ breast cancer where MD measurement was also available (n = 842) were identified from the Lifepool cohort of >54,000 women participating in population-based mammographic screening. This group included 142 carcinomas in the lowest quintile of MD and 119 carcinomas in the highest quintile. Clinico-pathological and family history information were recorded. Tumor DNA was collected where available (n = 56) and sequenced for breast cancer predisposition and driver gene mutations, including copy number alterations. Compared to carcinomas from low-MD breasts, those from high-MD breasts were significantly associated with a younger age at diagnosis and features associated with poor prognosis. Low- and high-MD carcinomas matched for grade, histological subtype, and hormone receptor status were compared for somatic genetic features. Low-MD carcinomas had a significantly increased frequency of TP53 mutations, higher homologous recombination deficiency, higher fraction of the genome altered, and more copy number gains on chromosome 1q and losses on 17p. While high-MD carcinomas showed enrichment of tumor-infiltrating lymphocytes in the stroma. The data demonstrate that when tumors were matched for confounding clinico-pathological features, a proportion in the lowest quintile of MD appear biologically distinct, reflective of microenvironment differences between the lowest and highest quintiles of MD.

The genetic architecture of breast papillary lesions as a predictor of progression to carcinoma.

Intraductal papillomas (IDP) are challenging breast findings because of their variable risk of progression to malignancy. The molecular events driving IDP development and genomic features of malignant progression are poorly understood. In this study, genome-wide CNA and/or targeted mutation analysis was performed on 44 cases of IDP, of which 20 cases had coexisting ductal carcinoma in situ (DCIS), papillary DCIS or invasive ductal carcinoma (IDC). CNA were rare in pure IDP, but 69% carried an activating PIK3CA mutation. Among the synchronous IDP cases, 55% (11/20) were clonally related to the synchronous DCIS and/or IDC, only one of which had papillary histology. In contrast to pure IDP, PIK3CA mutations were absent from clonal cases. CNAs in any of chromosomes 1, 16 or 11 were significantly enriched in clonal IDP lesions compared to pure and non-clonal IDP. The observation that 55% of IDP are clonal to DCIS/IDC indicates that IDP can be a direct precursor for breast carcinoma, not limited to the papillary type. The absence of PIK3CA mutations and presence of CNAs in IDP could be used clinically to identify patients at high risk of progression to carcinoma.

Factors influencing reoperation following breast-conserving surgery.

BACKGROUND: Reoperation rates after breast-conserving surgery are highly variable and the best techniques for optimizing margin clearance are being evaluated. The aim was to identify the reoperation rate at our centre and identify influential factors, including a change in guidelines on margin recommendations and the introduction of in-theatre specimen X-ray. METHODS: A retrospective review of medical records was undertaken to identify 562 patients who underwent breast conservation at The Royal Melbourne Hospital and Royal Women's Hospital between 2013 and 2015. All cases that underwent subsequent re-excision or total mastectomy were captured and factors influencing margin excision recorded. RESULTS: Reoperation was undertaken in 19.5% of patients (110; 86 re-excisions and 24 total mastectomies). There was a reduction in reoperation rate from 25% to 17% (P = 0.01) with adoption of the margin guidelines in 2014, but no significant reduction with the introduction of in-theatre specimen X-ray in 2015 (21% versus 16%, P = 0.14). On multivariate analysis, factors that significantly influenced reoperation rates were the presence of multifocality on mammogram (odds ratio (OR): 5.3, 95% confidence interval (CI): 1.6-16.7, P < 0.01); lesion size on mammogram (OR: 2.2 per 10 mm, 95% CI: 1.4-3.6, P < 0.01); smaller excision specimen weight (OR: 0.5 per 25 g of resection, 95% CI: 0.3-0.8, P < 0.01); and pure ductal carcinoma in situ on final pathology (OR: 5.9, 95% CI: 1.9-16.7, P < 0.01). CONCLUSION: Optimizing reoperation rates following breast-conserving surgery remains a surgical challenge, particularly in patients with in situ or multifocal disease. Adoption of international margin guidelines reduced reoperation rates at our centre; however, introduction of intraoperative specimen X-ray had no influence.

Uptake of adjuvant breast cancer treatments recommended by multi-disciplinary meetings.

BACKGROUND: Adjuvant therapy for breast cancer is routinely discussed and recommended in multi-disciplinary meetings (MDMs). Current literature explores how treatments received by patients differ from national guidelines; however, it does not explore whether treatment is concordant with MDMs. This study provides an Australian perspective on the uptake of MDM recommendations and reasons for non-concordance. METHODS: A retrospective cohort study of patients with breast cancer presented at The Royal Melbourne Hospital MDM in 2010 and 2014 to investigate the concordance between MDM recommendations and treatment received. RESULTS: The study group comprised 441 patients (161 from 2010 and 280 from 2014). A total of 375 patients were included in the analyses. Overall, 82% of patients had perfect concordance between recommended and received treatment for all modes of adjuvant therapy. Concordance to endocrine therapy was higher for invasive cancers than ductal carcinoma in situ (97% versus 81%, P < 0.0001). Concordance to radiotherapy was high and did not differ according to type of cancer or surgery (ranging from 88 to 91%). Concordance to chemotherapy recommendations was high overall (92%) and did not vary with nodal status. Women aged over 65 years were least likely to be recommended for adjuvant therapy but most likely to concordant with the recommendation. CONCLUSIONS: Uptake of MDM-recommended treatments is high. There is a minority of patients in whom MDM recommendations are not followed, highlighting that there are extra steps between recommendations at an MDM and decisions with patients. More attention to this issue is appropriate, and the reasons for non-concordance warrant further study.