Personalized Breast Cancer Screening and the Role of Breast Density

Other than accepting the invitations to speak here today, I have no other conflicts to the clear. Now you've seen these slides a couple of times before. An yes, my talk will focus on personalized screening. However, in reality, if we are thinking about personalized screening, it will most likely affect almost all aspects of the screening process. So before going into detail on personalized screening and tomosynthesis and breast density, just let me very briefly explain what it is that we would like to do with personalized screening. So the key question there is, will we do better in screening if we tailor our screening strategies to a woman's individual risk of breast cancer? Currently we are offering age based screening limited to a certain age group, but we know that there are many more risk factors for breast cancer. So can we use this information to the benefit of the women that we are inviting now in order to think about personalized screening? There are a number of requirements. Obviously if you think about risk, we would need a way of predicting the risk for an individual. A woman in this population, so we need a breast cancer risk prediction model now. Once we've established risk, the the the next question would be what can we offer this woman? So can we design screening strategies that will work for the women in the different risk categories? Ultimately the program still needs to be affordable so cost effectiveness also comes into play. And last but certainly not least, we need to understand the acceptability of this new approach. Working risk based. Rather than age based, and so is that acceptable to women to professionals and to society. Now in the interest of time I will not be discussing the last two requirements, but I will be discussing the 1st two in a little more detail and so thinking about breast cancer risk prediction. Let's assume we would like to group women in, let's say lower than average risk middle. So average risk or higher than average risk. And there are a number of breast cancer risk prediction models that are. Already available and a lot of work has been done over the last years. To extend these risk prediction models with new information. Adding, for instance mammographic density but also the genetic variants called snips. However, none of these models were actually developed for use in screening, so it's important to validate that these models are actually also helpful in stratifying women in the average risk groups, and I think more recently, there's also a growing interest in looking at more short-term risk prediction. Rather than just the long term risk prediction for which these models were originally established, the ultimate aim is still for these models to improve their descriminatory power because we do want to make sure that we are able to stratify this screening population into meaningful risk groups. Now, let's say that we are successful in establishing this risk prediction model and allocating women to the correct risk categories. Then we still need to understand what the optimal strategy will be to offer to these women in screening, and this we could address by looking at a variation of combinations. Of course, there needs to be a point in time. Then when you start assessing risk, so there must be some kind of a baseline screening examination, and you might actually wonder at what age you would like that to take place. And what risk factors you would like to take to and into account when assessing risk? Now thinking about strategies we can think about, starting ages, stopping ages, screening interval. But we can also think about imaging modalities in addition to mammography, so maybe combinations might work better for women who are at higher than average risk and maybe in the future. We will not just limit ourselves to imaging, but other tests like biomarkers may come into play. So going back to the title of my talk personalized breast screening. Assuming that tomosynthesis will be the future screening test and the role of breast density, how do we bring that old together? And when I was preparing and reviewing the literature, I tried to ask myself some questions which I will now endeavour to at least give you some kind of an answer and that is. So do we know how tomosynthesis performs across breast answer to categories? Do we know what the relationship of breast density probably measured using automated measure with bra? Scan service because we know it's a risk factor for breast cancer, but also with masking. So what are these? What does it do to the interval Kansas? And if we think about additional imaging modalities, assuming that tomosynthesis is going to be the primary screening test, where where is the role for additional image Ng then? Now to start off, if we look at the performance of digital mammography across density categories, I just selected these two studies to show you that for digital mammography it's fairly obvious that there is a clear pattern across the density categories. If you look at the recall rate, it will go up at density. Detection rate will go up, and interval cancers will go up and you can see that it actually looks quite similar. Now. I was really hoping to show you a table like this also for tomosynthesis. I did want to focus on European screening programs to make it a generalizable to our screening. Today I was trying to focus on really comparing tomosynthesis to digital remote mammography and not to the combination. And then there isn't that much information available yet and what we have are actually quite small numbers. Mostly first round results. So if I try to piece together the table, we do still see a pattern. But you can see that some of these studies have just reported on recall RAID. Some have just reported on detection and interval cancers are obviously still missing. A Sofia has just shown there a limited number of studies undergoing difficult to find any results across density categories. Yet because the numbers are so small, All in all, if you just look at this and look at the general pattern, it does seem to indicate that yes. Also internal synthesis. You will see trends across the density categories. You can see some of actually grouped the non lens and the dance categories. So what do we know already about breast density? And let's say the risk of breast cancer and masking? Well, at least studies so far have indicated that if you measure volumetric density on digital mammography and on tomosynthesis there appeared to be well correlated. So that's very promising. However, I think it's also fair to say that there are differences between these methods. And of course, if we use density measurements to place a woman in a risk category. We do want to make sure she's in the right category, so it's important to understand why the methods differ and what impact it may potentially have on the strategy that a woman would be offered. When we think about personalized screening. Now, of course, we're interested in associations with screen detected an interval cancers when density measures are applied to tomosynthesis. I've not been able to find much yet, and also other recent studies have actually brought this up as a limitation that it's difficult to find these studies where you have. Density assessed on tomosynthesis and looking at the risk for screen detected in interval Kansas. Additional imaging metallurgies is fuming. Tomosynthesis is the screening test well in many of the studies out there. Now, tomosynthesis is still considered an additional test, so there are studies that focus on adding, for instance, MRI. The dense trial or ultrasound for instance, is down for women with dense breast and negative mammography and there will be reports of incremental detection rates. But what are we going to see when tomosynthesis becomes the primary screening test an? How will then the incremental cancer detection rates look when you compare ultrasound added to Tomo or MRI added to tomo negative examinations? And so if we think about the future of screening with two machines is an the role of breast density. That's basically saying that in this figure from a recent publication by Arellano, this screening mammogram would be replaced by DBT, and there are a number of areas where you can see that density comes into play, and all of these will then be measurements made on tomosynthesis, and that we need to understand better in order to figure out what the best screening strategy for the women attending screening will ultimately be. And because we have these automated measurements, it would be possible to act on it. For instance, even on the spot when the woman has had the examination, be it not a mammography exam, but it's almost interested exam. But I think we do need to think very clearly what is the aim of assessing the density? Are we going to use it for risk prediction? Are we going to decide on supplemental image Ng based on breast density alone? Or do we need other risk factors? That we need to take into account to think about supplemental imaging, and that may or may not make it possible to decide on the spot what to offer to this woman. And, of course, the health care system and the organization of your program will also make a possible or not to actually offer something on the spot. If I think of the Dutch program where we have largely mobile units, it would be very difficult to make an offer to a woman based on the density measurement. Under spot, because in the mobile we wouldn't have anything else but mammography or tomosynthesis to offer. So taken together, if we look at the future, then personalized screening is meant to optimize the benefit harm ratio of breast cancer screening. It still needs to be efficient. It still needs to be acceptable. It will affect a lot of these expert aspects of the screening program, but you can see that there also still a lot of question marks there. I think the one thing we will know for sure is that it will become. I more complex program with lot more variation and many more options than we currently know and many of these options may be affected by measurements of breast density. Thank you.

25 Conclusion: What should the future of breast cancer Annals of Internal Medicine What should the future of breast cancer screening with What is required? Personalized screening Design personalized screening strategies Q2: Breast density - risk / masking Breast cancer risk prediction model The future of screening with DBT QI: FFDM performance - density QI: FFDBT performance - density QI: DBT performance - density Q3: Additional imaging modalities - DBT Automated and Clinical Breast Imaging Repotting and Data System screening with digital breast tomosynthesis look like? digital breast tomosynthesis look like? and the role of breast density Density Measures Predict Risk for Screen-Detected and Interval Personalized breast cancer Personalized screening DBnaåizedbreast cancer DBT and the role of breast density Studies so far indicate that volumetric density Several models available: Risk assessment — when? 1) How does DBT perform across breast density Cancers DBT is still considered an additional modality TO-BE Focus on European screening programmes Nothing to declare in DM and DBT are well correlated Spaini screen ib • Recall rate (960) Recall rate (%o) • Recall rate (%o) • Recall rate (96) Recall rate (B/00) Recall rate (Oho0) Recall rate (9/60) Recall rate (0/60) • e.g. Gaill, Tyrer-Cuzick2, BOADlCEA3 — adding density + SNPs categories? • Recall rate Scientific Method 25 22 21 25 22 42 36 A Study • Baseline screening examination Breast cancer risk prediction model A breast cancer risk prediction model Reduction in mortality Attendance rate • Recall rate (960) • Recall rate 27.2 21.7 89.6 Conclusion MO: Christopher G. MS: P. Lin MS: PhD: MO: Christopher G. MS: P. MS: PhD: Middle STORM-22 ? Screening strategies ? Acceptability • Detection rate (9600) • Detection rate (0/0) • Detection rate (0/00) 2.8 6.8 4.0 64.8 work now 58.1 15.9 15.6 now • Combination of risk factors of breast density breast density breast cancer Recall rate (Oho) 45.3 86.6 73.6 89.6 63.9 EUROPEAN SOCIETY OF RADIOLOGY Studies focus on adding e.g. MRI (DENSE) or breast Focus on comparing DBT to FFDM John A. Shepherd. PhD; Kathleen R. MO; L and M. PhD ? Benefit-harm Personalized SIEMENS .. SIEMENS 2) What is the relationship of automated breast • Interval cancer rate (0/60) Are being adapted for use in screening Differences exist between automated 6.9 15.2 15.9 24.0 2.9 27.9 Design personalized screening strategies Detection rate ('60) • Detection rate Detection rate (960) Detection rate ('/60) • Recall rate (960) 12.9 16.0 12.9 13.9 alblcld Mammography breast cancer of breast density ONE ultrasound (ASTOUND) for women with Personalized screening Radiation dose Imaging density measures on DBT with breast cancer risk DOES Nor DOES NOV Direct Malmö trial methods which can impact risk assessment as • growing interest in short-term risk prediction models DOES screening with ALL ? Screening strategies Risk-based density S u Starting and stopping ages / screening interval DBT and the role • Recall rate • Recall rate (96) dense breasts and negative mammography Limited information available, small numbers Low High Establish cost-effectiveness (screen-detected cancers) and masking (interval DBT and role Cuc* RISK Netherlands2 M i re illegroeders Efficiency / Acceptability Limitation: Neither automated nor clinical Bl- well as the choice of screening strategy CARE c ARE STRATEGY (c breast (c. g. breast center of breast density Workflow, reading CARE • Recall rate (9/60) 6.0 4.0 63.9 15.2 30.7 01.7 24.0 Cancers cancers)? Healthineers Detection performance & reporting Aim: improve discriminatory power • Imaging modalities (alone or in combination) • Detection rate 4.0 6.8 2.19 RADS density was assessed on tomosynthesis, Netherlands Verona Understand acceptability to women and society ? Screening strategies ? Screening strategies Mostly first round results Can we interpret e.g. incremental detection • Interval cancer rate (0/60) 308.7 08.7 30.7 124.9 15.2 4.0 Detection rate ('60) Risk prediction and / or supplemental imaging 62.9 12.9 Complexity • i.e. the ability to differentiate between high- and low-risk Study associations with screen-detected and Will screening tailored to a woman's individual risk of breast Personalized breast cancer screening with DBT 3) What is the role for additional imaging modalities • Density alone or in combination Cordoba New screening tests (e.g. biomarkers) rates when DBT replaces FFDM? groups in screening interval cancers with density measures on DBT • Detection rate Detection rate 63.9 16.6 15.7 85.6 5.7 45.7 98 4.5 4.0 6.0 45.0 when DBT is the screening test? and the role of breast density? cancer become the new screening paradigm? DBT • Health care system and organization of the programme Health care system and organization of the programme Dutch Expert Centre for Screening, @ Siemens Healthcare GmbH, 2019 Radboud umc J Med 2019 Ann Med 2018 Arienoet 2019 1 Med 2018: 2 van veen al, 3 Lee et al, Gen Med 2018 2019 Med 2018: 2 van veen al, 3 Lee et Gen Med 2018 2018; 2 2016; 3 The 2015; Radboud u mc Radboud u m C 2 Wanders Treat 2017 4 La Radiol 5 caumo al. Radiol Radiol 2018 4 La 5 caumo al. Radiol 2018 SIEMENS .. SIEMENS ... Healthineers • Healthineers