PEPconnect

Conclusion from the Malmö Breast Tomosynthesis Screening Trial – a concept for breast screening

This webinar by Sophia Zackrisson (Lund University, Malmö/Sweden) gives an overview about the Malmö Breast Screening Trial that focuses on DBT screening performance, effectiveness and reduced reading time. Prof. Zackrisson concludes that 1-view tomosynthesis with minimum compression increases cancer detection and can be feasible in screening.

Due to high data volume of clinical image files, restrictions in the bandwidth may compromise image quality depicted in this webinar.

Alright, so let's see what we have done in Malma lately. I guess some of you have heard this before, so thank you very much for the kind introduction. Mr chairman. Today I'm going to briefly in like 10 or 12 minutes give you the conclusions of our trial. The Mama breast tomosynthesis screening trial. Quite apparent disclosures, I would say at this Siemens and Bayer meeting I would like to give a big thank you to my collaborators, coauthors and also the funding resources that made this trial possible. And not the least, the group of the research group, but also the all the nurses technologist. Doctors at the breast center in Mama who did a great job to do this trial. So I mean your introduction was really a good one because we always have to think about you, know the pros and cons in when it comes to screening in population based screening we actually screening healthy women and those has always been an issue. That's been a lot discussed over the years. Just to look, you pointed out. But still we there are other factors in this principles of screening that we should consider when we're talking about new screening approaches. These are just some of the principles of screening that are listed according to Wilson and Jungner already in 1968 by WH Oh. We should have a fast and safe but accurate test in screening, and it should also be cost effective. That's why also is very important to really assess the efficacy of DVT in screening and not just accept that it's a better mammography because there's a lot of factors that you have to take into account. And when we designed our trial, there were already some trials that were designed and an ongoing, and they wanted to investigate the combination of digital mammography, plus DBT in some form. But we wanted to go even more simple. And this is just the publication that we have published in October last year where you have our final results and you can find all the detailed information which I don't really have time to give you here. This is just to show you how we were thinking when we were designing our trial with just one view tivity for screening. So in this case here you see quite an evident spiculated tumor in the breast, but it's only visible on the mammogram on the CC view, where it's surrounded by fatty tissue while it's hidden by overlapping tissue on the envelope view. And of course, it's visible on the ultrasound, but this illustrates the fact that why you need several views when you do. The two D image Ng when you do digital mammography, you need different angles to actually prove that there is something in the breast, or that there is nothing in the breast. When we got our first prototype in MoMA in 2006, we started trying it out on difficult or almost occult cases on mammography, and we soon realized that in quite a few cases, we could actually see the tumor in just one of the views where it was invisible or not. At least very little visible on the 2D mammogram. So this is the same patient, and you can easily see the spiculated tumor on the low view of the DBT. So this gave us the idea that you might actually define. With just one view for screening, because if you want to do screening, you should have something that is fast. You shouldn't have two extensive screening imaging protocols, not just because of those, but also because of reading time and also the acquisition time. So This is why the aim of our trial is to study the accuracy onview DVT versus classic to view DM in population based screening. So how did we do? Well, we got all the clearance to do the trial, of course, and this is some of the technical background data of our trial. We use the Momma inspiration unit. And as you know this is a wide angle system it has. Quite a long spanning time, but since we only do one view it's we didn't think it was a problem. One thing here is that we decided to go for reduced compression for the DBT examination for the women, because we had seen in our earlier studies that if you don't compress the breast as much for DBT, you actually get a better depth resolution because you get more slices out of the breast. Of course, you still have to keep the breast in place and firmly compressed, and of course to keep the. Radiation dose you know reasonable, but we have also proven that the extra pressure add at the end of a mammography examination mainly causes pain. It actually does not really compress the breast more. It doesn't really make the best thinner. So I would say that you have a very little effect on the dose by not compressing so much. So in screening in Sweden we screen women already from the age of 40 up to 74 with a couple of different screening intervals that you see up here. So we just had a random sample of the women who were to be invited to the screening, and Alma and those women got an invitation to participate in the trial. We had a participation rate of 68%. So we needed 15,000 women to have enough power in our trial to see if we have enough or higher detection at the reasonable recall rate in this population. So 14,848 women were finally included in this trial. And they did their regular screening mammogram at one occasion, and then at the same location they actually did the DVT in just the one emolo view at the same unit. But then we separated the images into two reading arms, but this means that each woman is our own control for this trial is not a randomized trial, but the reading ARM DBT was separated for reading ARM DM and we had independent double reading and scoring in both arms and also consensus meetings in both arms. And these are the primary outcome measures that the trial was designed to measure. Sensitivity, specificity, cancer detection, recall rate, and PPV. And of course MTV. If you find that interesting as well. So what did we find? Well, when it comes to compression and those we managed to reduce the compression force by up to 40% on average with DVT compared to mammography, we didn't do really any formal. You know investigation whether how much the women appreciated this or not because it's was not possible in the screening setting, but we got a lot of spontaneous comments about it. Was such a relief. Not having such a painful mammography. So this might have implications on future attendance. For some women. When it comes to dose, you see here that if you only do one view DVT, of course you might even arrive at the lower dose compared to two view mammography. So we had an average of 15% lower average Glendora dose compared to 2D mammography. So it sort of comes with the package, I would say. So let's have a look at the outcome measures you see here. The sensitivity in the DBT reading arm compared to the reading ARM DM. And as you see, there's a much higher sensitivity 81% compared to 60% of reading on DM, so this is a huge difference. Of course you might think that this is a bit low for screening mammography. It usually run like 70 or 80, but you have to keep in mind that this is in relation to the fact that you found extra cancers with DBT that were not found with the. Why in this trial the sensitivity became relatively lower? Having look at the specificity, you can see we have a very already very high specificity with two D mammography in Sweden. 98% that's hard to beat. We were a little bit inferior with the DBT, 'cause you don't only seek answers, better you see a lot of other things better with DBT in the breast that we had to recall. Come back to that a little bit later in my talk. I'll just make sure I'm keeping that I'm here. Yeah, I'm good. So this means that we had a higher cancer detection rate, almost nine cancers per thousand women screen compared to 6.5 with mammography. And having a look at the recall rate we had recall rate with three point 6% with digital resume thesis compared to 2.5 with digital mammography. And as you see, we had a very low recall rate, just like the specificities indicating here. So there was really no other way to go then up and I would just be interesting to see how many of you in your settings that have recall rates below 3 hands up. Some, but the majority actually have usually much higher recall rates in jail or general, so I wouldn't say that recall rate is an issue in this setting, and the PPD was quite similar, 24 versus 26, and then the negative predictive value was really, really high in both arms. So I just want to point out that we did have a little different reading steps in the trial. So in the reading ARM DBT we first for the readers we showed first the emolo views of the DBT only, and they rate it and score that and 40 of the DBT only detected cancers where detected in that step. But then we actually added the current view of the mammogram just to see if we had any effect of finding more calcifications or helping the radiologist to find something more. It was only one of the counselors that was detected in that reading step, and this was before the era of synthetic mammograms. We didn't have them available. That's why we we did this image combination and the third step was just an addition of the priors and one extra counsel was found in that step. The majority of the councils into trial were actually found in both reading arms, both by DVT and mammography, but some of them were also only detected, or at least yeah, found as positive in the DM arm. That might be a reader variability, but also just a random effect. Having a look at how they distribute over breast density, the DVT detected only detected cancers. They were distributed overall density categories. Interestingly enough, you might expect that you only have them in the higher density categories, but as you can see here, the extra height of the pink bars are the tomosynthesis only detected cancers and you have them across all different Pirates categories. So tomosynthesis is actually good for all women. So, just to summarize, the cancer detection rate, there is an incremental increase by two extra cancers per thousand women screen with DVT. It means if you put it in relative terms of 34% increase with one view Tomo compared to mammography. And again the recall rate was a bit higher, but it's still well below the European recommendations and much much lower than a lot in a lot of countries. I would say so recall rates in my opinion would probably also come down if you would do repeated screenings with. T and you have a learning effect and you would have less false positives. If you have a comparison with prior DVT as well. So our trial shows increased detection and the interesting thing is, no matter how you actually combine images, how many views you have of the tomororw, the memo we seem to arrive at very similar results and this is just to show you that there are some trials who had the full protocols of both modalities and also some trials that use two view DBT. Plus the synthetic mammograms and they all show increased detection to some degree. Around 30 to 4% thirty to 40% or even more and we arrive at the same result with just one view wide angle DBT. And just a quick look at what kind of cancers you find with DVT. There were no large differences when it came to age at diagnosis, tumor size, or. Lymph node involvement between the reading arms. But we did find more lobular cancers with DVT compared to the. But the majority of those extra councils were invasive. Almost 90% of them were invasive. We had no excess detection of DCIS with DBT only. And this is just an example of a lobular cancer grade two which was not detected by the readers in the DM arm. It's actually not visible, but it's clearly visible on the DBT. And this is just another look at how we can divide it into lumenal versus non luminal cancers. And you can see that you find quite the same kind of cancers with tomosynthesis as to do with mammography. There mainly lumenal cancers. So this is but still you do find some extra triple negative cancers in her two positive cancers. And this is such a case. Only visible on the DBT. But still, this is just descriptive data, but it's we can conclude that we have a similar subtype distribution as in studies with deemed detectors cancer screening, so we seem to find more of the same with DVT. It's hard to catch those really aggressive cancers also with DVT. One important question. I'm soon arriving at my final slide is are we reducing interval cancers after screening with DBT? That's a very important question because that would be a way to work around the fact if we have an A mortality effect or not. There have been reports from a couple of the other trials, but they all non significance and we have investigated the interval cancer rates in the trial which was 1.5 for thousand screens. We have a big control group of the. All the concurrent screenings during the same. 100,000, you can see that was higher, so there seems to be a reduction, but it's not significant in. This really shows that the trials are not designed to. You, as requested about decreased. Interval cancer rates. So we need to pull this data together and that's what's happening in this proposed protocol for me to analysis. So in a couple of years you probably have some interesting data coming with from all the different trials trying to address this important question. So I think we will come back to this picture later. I we without trial, we can actually address quite a few of those points that we will discuss. But I would like to conclude that it seems feasible to screen breast cancer with using one view DBT with reduced compression and it could be one strategy for the future. Thank you very much.

SIEMENS .. SIEMENS Population based screening program Malmö, Sweden* THE LANCET Healthineers Subscribe Claim R 52 year old woman with 12 mm triple ne Oncology tive breast cancer detected in DBT reading arm only Ages 40-74 yrs Cancers by BI-RADS Breast Density Primary outcomes and screening performance measures (95% Cl) Different Image Protocols — Similar Results Discussion Cancer Characteristics Results Invasive Cancers: Luminal vs non-luminal Protocol for meta-analysis Recall Rate Interval Cancer Rates - prospective trials Cancers and Reading Steps Cancer Characteristics Compression and Dose Cancer Detection Rate Disclosures Primary Outcome Measures Ethical Approval Attendance: Image Acquisition Background Attendance: Screening interval: 18 months 40-55 yrs, 24 months 55-74 yrs View of future screc Conclusion Malmö Breast Tomosynthesis Screening Trial, MBTST mortality TO be continued... reduced compression Reading arm DBT only Reading arm DMT Reading arm DBT Reading arm DMT Reading arm DBT only Reading arm DM Reading arm DMT only P•value P-value programs: Conclusion from the Skåne University Hospital Breast Clinic, DBT reading arm only All DM detected cancers All cancers One-view breast tomosynthesis versus two-view mammography in the Ma IC contrial/1000 screens IC controal/1000 screens Comparison group Reading arm DBT only Reading arm DM Both reading arms -View View -view Reading arm DMT only Reading arm DBT Reading arm DBT only Reading arm DM • No large differences in mean age at diagnosis, tumor size, grade or lymph node Spensitivity • Siemens Mammomat Inspiration (DBT and DM) • On average 40% reduced compression force with DBT vs DM • Principles of screening Dept of Imaging and Functional M Dustier MSc, PhD • Trial approved by ethical committee at Lund University, (Dnr 2009/770) • 3.6 % in Reading arm DBT • 3.6 % in Reading arm DB T • Speaker's fees and travel support from Siemens Healthineers I Andersson, MD PhD 2.1 (51/24,301) 2.0 (118/59,877) Oslo Previous DM Breast Tomosynthesis Screening Trial (MBTST): a prospective, popu al 2014, 2019 Spensitivity Sensitivity n=40 n=89 n=42 Random sample 2010-15 (74-2-86-9) (974-2-186-9) (54-2-768-9) (974-9-986-3) (34-2-36-9) n=89 n=42 (52-3-68-0) (52-2-786-9) (52-3-28-0) (52-3-680) involvement between reading arms 81-1 981-1 60-4 Medicine • Specificity • Descriptive data Malmö Breast Tomc Malmö Breast Tomosgnthesis H Förnvik MSc, PhD student acceptable for target population Bernardi et al 2016 • 25 low-dose projections 25 low-dose projections K Lång, MD, PhD screening rounds screening cohort screening strategy N invited= 21 ,691 Ciatto et al 2013 al 2018 Sophia based, diagnostic accuracy study 2.2/1000 ( 1.2-3.2, p

  • tomosynthesis
  • breast screening
  • FFDM
  • mammography
  • screening guideline
  • dense breast
  • risk
  • Malmö trail
  • Malmö study
  • wide-angle tomosynthesis
  • dose
  • image quality