Artificial intelligence (AI) to enhance breast cancer screening: protocol for population-based cohort study of cancer detection

IntroductionArtificial intelligence (AI) algorithms for interpreting mammograms have the potential to improve the effectiveness of population breast cancer screening programmes if they can detect cancers, including interval cancers, without contributing substantially to overdiagnosis. Studies suggesting that AI has comparable or greater accuracy than radiologists commonly employ ‘enriched’ datasets in which cancer prevalence is higher than in population screening. Routine screening outcome metrics (cancer detection and recall rates) cannot be estimated from these datasets, and accuracy estimates may be subject to spectrum bias which limits generalisabilty to real-world screening. We aim to address these limitations by comparing the accuracy of AI and radiologists in a cohort of consecutive of women attending a real-world population breast cancer screening programme.Methods and analysisA retrospective, consecutive cohort of digital mammography screens from 109 000 distinct women was assembled from BreastScreen WA (BSWA), Western Australia’s biennial population screening programme, from November 2016 to December 2017. The cohort includes 761 screen-detected and 235 interval cancers. Descriptive characteristics and results of radiologist double-reading will be extracted from BSWA outcomes data collection. Mammograms will be reinterpreted by a commercial AI algorithm (DeepHealth). AI accuracy will be compared with that of radiologist single-reading based on the difference in the area under the receiver operating characteristic curve. Cancer detection and recall rates for combined AI–radiologist reading will be estimated by pairing the first radiologist read per screen with the AI algorithm, and compared with estimates for radiologist double-reading.Ethics and disseminationThis study has ethical approval from the Women and Newborn Health Service Ethics Committee (EC00350) and the Curtin University Human Research Ethics Committee (HRE2020-0316). Findings will be published in peer-reviewed journals and presented at national and international conferences. Results will also be disseminated to stakeholders in Australian breast cancer screening programmes and policy makers in population screening.

965The estimated impact of improved breast screening tests targeted at women with dense breasts

Background There is increasing interest in risk-based breast cancer screening, including interventions to improve outcomes for women with mammographically dense breasts. Methods Policy1-Breast is a continuous-time, multiple-cohort micro-simulation whole-population model which incorporates breast cancer risk, life-course breast density, menopause, hormone therapy use and screening participation. Outcomes include cancer diagnoses and characteristics (invasive/DCIS, tumour size, grade), mode of detection (screen-detected/interval/other) and mortality (breast cancer and other cause). Policy1-Breast validates well against key observed clinical outcomes in Australia. We estimate changes in outcomes within and outside the BreastScreen program upon the introduction of a hypothetical screening test with improved sensitivity for women with dense breasts. Results We estimate that introducing in year 2020 a screening test for women in the highest quintile of breast density at age 50 that halves the masking effect of their breast density would, by 2026-2030, increase diagnosis rates of population-level invasive cancers ( 10%) and screen-detected cancers (20%) and decrease rates of interval cancers (17%) and community-detected cancers (6%). Conclusions Targeted screening tests with improved sensitivity for women with dense breasts are expected to markedly reduce interval cancers and other cancers diagnosed outside the BreastScreen program, while increasing all cancer diagnoses due to increased rates of screen-detected cancers. Key messages Specialised breast cancer screening tests directed at women with very high breast density are expected to reduce interval cancers and increase overall cancer diagnoses. Population simulation models such as Policy1-Breast can complement trial evidence by evaluating a range of scenarios and estimating short and long-term implications.

Breast Density and Risk of Interval Cancers: The Effect of Annual Versus Biennial Screening Mammography Policies in Canada

Regular screening mammography reduces breast cancer mortality. However, in women with dense breasts, the performance of screening mammography is reduced, which is reflected in higher interval cancer rates (ICR). In Canada, population-based screening mammography programs generally screen women biennially; however, some provinces and territories offer annual mammography for women with dense breast tissue routinely and/or on recommendation of the radiologist. This study compared the ICRs in those breast screening programs with a policy of annual vs. those with biennial screening for women with dense breasts. Among 148,575 women with dense breasts screened between 2008 to 2010, there were 288 invasive interval breast cancers; screening programs with policies offering annual screening for women with dense breasts had fewer interval cancers 63/70,814 (ICR 0.89/1000, 95% CI: 0.67-1.11) compared with those with policies of usual biennial screening 225/77,761 (ICR 1.45 /1000 (annualized), 95% CI: 1.19-1.72) i.e. 63% higher (p = 0.0016). In screening programs where radiologists’ screening recommendations were able to be analyzed, a total of 76,103 women were screened, with 87 interval cancers; the ICR was lower for recommended annual (65/69,650, ICR 0.93/1000, 95% CI: 0.71, 1.16) versus recommended biennial screening (22/6,453, ICR 1.70/1000 (annualized), 95%CI: 0.70, 2.71)(p = 0.0605). Screening program policies of annual as compared with biennial screening in women with dense breasts had the greatest impact on reducing interval cancer rates. We review our results in the context of current dense breast notification in Canada.

Positive impact of a faecal-based screening programme on colorectal cancer mortality risk

Introduction The effectiveness of colorectal cancer (CRC) screening programs is directly related to participation and the number of interval CRCs. The objective was to analyse specific-mortality in a cohort of individuals invited to a CRC screening program according to type of CRC diagnosis (screen-detected cancers, interval cancers, and cancers among the non-uptake group). Material and methods Retrospective cohort that included invitees aged 50–69 years of a CRC screening program (target population of 85,000 people) in Catalonia (Spain) from 2000–2015 with mortality follow-up until 2020. A screen-detected CRC was a cancer diagnosed after a positive faecal occult blood test (guaiac or immunochemical); an interval cancer was a cancer diagnosed after a negative test result and before the next invitation to the program (≤24 months); a non-uptake cancer was a cancer in subjects who declined screening. Results A total of 624 people were diagnosed with CRC (n = 265 screen-detected, n = 103 interval cancers, n = 256 non-uptake). In the multivariate analysis, we observed a 74% increase in mortality rate in the group with interval CRC compared to screen-detected CRC adjusted for age, sex, location and stage (HR: 1.74%, 95% CI:1.08–2.82, P = 0.02). These differences were found even when we restricted for advanced-cancers participants. In the stratified analysis for type of faecal occult blood test, a lower mortality rate was only observed among FIT screen-detected CRCs. Conclusion CRC screening with the FIT was associated with a significant reduction in CRC mortality.

Quantitative breast density analysis to predict interval and node-positive cancers in pursuit of improved screening protocols: a case–control study

Background This study investigates whether quantitative breast density (BD) serves as an imaging biomarker for more intensive breast cancer screening by predicting interval, and node-positive cancers. Methods This case–control study of 1204 women aged 47–73 includes 599 cancer cases (302 screen-detected, 297 interval; 239 node-positive, 360 node-negative) and 605 controls. Automated BD software calculated fibroglandular volume (FGV), volumetric breast density (VBD) and density grade (DG). A radiologist assessed BD using a visual analogue scale (VAS) from 0 to 100. Logistic regression and area under the receiver operating characteristic curves (AUC) determined whether BD could predict mode of detection (screen-detected or interval); node-negative cancers; node-positive cancers, and all cancers vs. controls. Results FGV, VBD, VAS, and DG all discriminated interval cancers (all p < 0.01) from controls. Only FGV-quartile discriminated screen-detected cancers (p < 0.01). Based on AUC, FGV discriminated all cancer types better than VBD or VAS. FGV showed a significantly greater discrimination of interval cancers, AUC = 0.65, than of screen-detected cancers, AUC = 0.61 (p < 0.01) as did VBD (0.63 and 0.53, respectively, p < 0.001). Conclusion FGV, VBD, VAS and DG discriminate interval cancers from controls, reflecting some masking risk. Only FGV discriminates screen-detected cancers perhaps adding a unique component of breast cancer risk.

Fibulin-2 expression associates with vascular invasion and patient survival in breast cancer

Stromal elastosis is related to good prognosis in breast cancer and fibulin-2 helps to stabilize elastic fibers in basement membranes. Here, we examined the level of perivascular fibulin-2 expression in relation to elastosis content, vascular invasion, molecular subtypes, tumour detection mode, and patient prognosis in breast cancer. We performed a population based retrospective study of invasive breast cancers from the Norwegian Breast Screening Program (Vestfold County, 2004–2009) including 200 screen-detected and 82 interval cancers. Perivascular fibulin-2 staining was semi-quantitatively graded based on immunohistochemistry (1–3) and dichotomized as high expression (grade 2–3) and low expression (grade 1). Elastosis content was graded on a 4-tiered scale and dichotomized as high (score 3) and low (score 0–2) expression, whereas lymphatic (LVI) and blood vessel invasion (BVI) were recorded as absent or present by immunohistochemistry. High perivascular fibulin-2 expression was strongly related to stromal elastosis (p<0.001), and inversely associated with BVI and LVI (p<0.001 for both). High fibulin-2 was associated with luminal breast cancer subgroups (p<0.001) and inversely with interval cancers compared with screen-detected tumours (p<0.001). By univariate analysis, low perivascular fibulin-2 was associated with reduced recurrence-free survival (p = 0.002) and disease specific survival (p = 0.019). Low perivascular fibulin-2 expression was strongly related to vascular invasion, low stromal elastosis, non-luminal breast cancer subtypes, interval presentation, and adverse prognosis.

Future Prospects of Colorectal Cancer Screening: Characterizing Interval Cancers

Tumors that are not detected by screening tests are known as interval cancers and are diagnosed clinically after a negative result in the screening episode but before the next screening invitation. Clinical characteristics associated with interval colorectal cancers have been studied, but few molecular data are available that describe interval colorectal cancers. A better understanding of the clinical and biological characteristics associated with interval colorectal cancer may provide new insights into how to prevent this disease more effectively. This review aimed to summarize the current literature concerning interval colorectal cancer and its epidemiological, clinical, and molecular features.