A standalone breast screening modality
Breast cancer screening with mammography is one of the most widely implemented screening programmes in Europe. However, the choice of using mammography as the screening method has lately been challenged. The sensitivity of mammography is known to be limited mainly due to overlapping tissue-obscuring tumours, which is pronounced in women with dense breasts.1,2
Breast tomosynthesis (BT) is a development of the mammographic technique, giving the possibility to reduce the negative effect of overlapping tissue. This is achieved by the X-ray tube moving in an angular range over the breast, acquiring multiple low-dose projections. These projections are reconstructed into an image volume parallel to the detector plane consisting of a stack of thin slices, typically 1mm. BT images can be obtained using the same views as in conventional mammography, but are usually confined to two views: craniocaudal (CC) and mediolateral oblique (MLO).
Even if the role of BT in clinical practice has not yet been established, it is evident that it is a valuable clinical tool compared with mammography. It has higher accuracy3 and better tumour visualisation, which can provide improved preoperative staging, for example, tumour size and multifocality,4 and by discarding ‘pseudo-lesions’ generated by the superimposition of tissues in mammography. Considering the known limitations of mammography, the role of BT as a screening modality has been of particular interest lately and was also the reason for the initiation of the Malmö Breast Tomosynthesis Screening Trial (MBTST).
The MBTST is a prospective population-based screening trial aiming to compare the performance of one-view BT to two-view digital mammography (DM) in breast cancer screening. The trial started in 2010, and in 2015 the planned inclusion of 15,000 women was achieved. Women eligible for the screening programme in the city of Malmö, 40–74 years old, were randomly invited to participate in the trial. The trial was designed as a one-arm study where participating women underwent both BT (MLO) and DM (CC + MLO) in the same session using a Siemens Mammomat Inspiration system, which is one of the BT systems with the widest acquisition angles (50°). The BT images were acquired with reduced compression force (about 40% reduction). Because BT was performed only in one view, the radiation dose was lower for BT compared with two-view DM (1.6mGy vs. 2.4mGy).
The BT and DM images were read in two independent reading arms (double blinded reading), by a total of five breast radiologists. Each reading arm had three reading steps (Figure 1). In each step, the radiologist assigned a 5-level score: normal (1); benign finding (2); non-specific finding with low probability of malignancy (3); findings suspicious of malignancy (4); and, highly suspicious of malignancy (5). If the case received a positive score (3 or higher) in any of the reading steps, it was assigned to an arbitration meeting where at least two readers re-evaluated the images and decided whether to recall the woman for further work-up, irrespective of the score on the other modality.
Analysing the first half of the trial
So far, results from the first half of the study population (7500 women) have been analysed in regard to cancer detection rates, recall rates and positive predictive values (PPV) for the two reading arms. The ground truth was established by the pathological report of biopsies or after surgery and through record linkage with the South Swedish Cancer Registry with at least one-year follow up.5 McNemar’s test was used to analyse the paired binary data for recall and cancer detection rates. The cancers were also characterised in terms of radiographic appearance and histological type and grade. Furthermore, in a 2016 paper, false positives (FPs) have been assessed for the same population.6 A FP was defined as a recalled woman who was considered disease-free after assessment and at least three-year follow-up. The FPs were assessed with several parameters including FP recall rate after arbitration, the radiographic finding leading to recall and the outcome of the work-up. Differences in the characteristics of FPs recalled on BT alone and DM alone were analysed using Chi-X2 and Fischer’s exact test.
Summing up the first results
An overview of the first half of the study population and the performance of BT and DM are presented in Figure 2. In summary, 68 of the 7500 women had a screen-detected breast cancer. These cancers were detected on both BT and DM (n=46), on BT alone (n=21) and on DM alone (n=1). Notably, all cancer cases received a positive score in the first reading step in the separate reading arms, that is, one-view BT alone and two-view DM alone, respectively. This resulted in a detection rate of 8.9/1000 screens (95% CI 6.9–11.3) for one-view BT as a stand-alone modality and 6.3/1000 screens (95% CI 4.6–8.3) for two-view DM (p<0.0001), which represents a 43% increase in the cancer detection rate. The 21 additional cancers detected on BT, of which 17 were invasive, did not differ significantly from the DM-detected cancers. However, these cancers tended to be more of histologic grade 1, lymph node-negative, and of smaller size than DM-detected cancers, suggesting a downstaging. A cancer detected on BT alone is exemplified in Figure 3. Contrary to what was expected, the additional cancers were found not only in women with dense breast, but also in women with non-dense breasts, which suggests that the overall lesion conspicuity increases with BT. Still, more biological data are needed in order to draw any conclusions which the upcoming analysis of the whole MBTST population hopefully will provide.
Figure 3: A cancer detected on BT alone. A 66-year-old asymptomatic woman recalled for findings suspected only on BT. A 15-mm invasive ductal carcinoma, histological grade 1 and lymph node-negative, was diagnosed at histological examination. Reproduced from European Radiology with permission.
The recall rate after arbitration was 3.8% (95% CI 3.3–4.2) for BT and 2.6% (95% CI 2.3–3.0) for DM (p<0.0001). The increase in recall rate for BT was a result of the increased cancer detection but also in the slight increase of FPs. However, the recall rate for BT was still low and well within the European guidelines. The FP recall rate was 1.7% for BT alone (n=131), 0.9% for DM alone (n=69), and 1.1% for FPs recalled on both modalities (n=81). Importantly, the FP rate for BT alone was halved after the initial phase of the trial (1.5 years), stabilising at 1.5%. Hence, a clear learning curve with increased specificity for BT was observed. Furthermore, BT seemed to be especially sensitive for stellate radiographic patterns resulting in more cancers as well as FPs, where the latter included radial scars and post-operative scar tissue.
The results in context
To date there are three prospective population-based screening trials that have evaluated the use of BT in screening. Besides the MBTST, the Oslo Tomosynthesis Screening Trial (OTST) and the Screening with Tomosynthesis or Mammography (STORM) trial have presented interim and final results, respectively.7,8 However, OTST and STORM share a fundamental different design compared with the MBTST, because they both evaluate the addition of two-view BT to two-view DM (so called combination mode) versus two-view DM. Hence, MBTST is the only study that evaluates BT in only one-view as well as BT as a standalone modality. The idea behind the MBTST design was to find a rational screening method, which besides increased diagnostic performance should be fairly fast. In spite of the different approaches of these three trials, similar significant increase in cancer detection rate of 30–40% has been observed. The OTST and the MBTST showed similar increase in recall rate, which might be expected with very low baseline recall rates. The STORM trial, however, had lower recall rates for the combination mode compared with DM. One should keep in mind the different cut-off levels for a recall, especially when comparing results from the US trials. Several retrospective screening trials in the US have shown a significant reduction in recall rates using the combination mode versus conventional DM. These results cannot unconditionally be translated to a European setting because the baseline recall rates, in general, are much higher in the US than in Europe.
One of the main challenges with using BT in a community-based screening programme will be the increased reading time, which will add to the radiologists’ workload and demand more resources. Even if only one-view BT is applied, the reading time will most likely be doubled. Further research is needed to find efficient reading strategies and/or image presentations. If CAD for BT will be further developed, changing the double-reading protocol to single reading combined with CAD or using CAD to eliminate clearly normal images could be a practical solution.
The MBTST was not designed to assess screening efficacy in terms of effect on breast cancer mortality. In order to do that, a very large randomised controlled trial with a long follow-up would be needed. Instead, as part of the forthcoming analyses after the completion of the MBTST, we will use interval cancers as a surrogate endpoint for screening efficacy. Furthermore, with more tumour biological data the additionally detected cancers will be analysed using clinically established parameters and oncogenetic profiling, trying to answer in what extent BT will add to overdiagnosis. Finally, a cost–benefit analysis of BT screening will be performed, because a screening modality should not only be simple, safe and sensitive, but also affordable.
Implications for screening practice
The clinical implication of the MBTST might be considerable because it indicates that one-view BT, with reduced compression force, could replace DM in breast cancer screening. Upcoming analyses on the whole MBTST population will see whether this holds true.
1 Laming D, Warren R. Improving the detection of cancer in the screening of mammograms. J Med Screen 2000;7:24–30.
2 Kolb TM, Lichy J, Newhouse JH. Comparison of the performance of screening mammography, physical examination, and breast US and evaluation of factors that influence them: an analysis of 27,825 patient evaluations. Radiology 2002;225:165–75.
3 Houssami N, Skaane P. Overview of the evidence on digital breast tomosynthesis in breast cancer detection. Breast 2013;22(2):101–8.
4 Förnvik D et al. Breast tomosynthesis: Accuracy of tumor measurement compared with digital mammography and ultrasonography. Acta Radiol 2010;51:240–7.
5 Lång K et al. Performance of one-view breast tomosynthesis as a stand-alone breast cancer screening modality: results from the Malmö Breast Tomosynthesis Screening Trial, a population-based study. Eur Radiol 2015;26(1):184–90.
6 Lång K et al. False positives in breast cancer screening with one-view breast tomosynthesis: An analysis of findings leading to recall, work-up and biopsy rates in the Malmö Breast Tomosynthesis Screening Trial. Eur Radiol 2016;26(11):3899–907.
7 Ciatto S et al. Integration of 3D digital mammography with tomosynthesis for population breast-cancer screening (STORM): a prospective comparison study. Lancet Oncol 2013;14:583–9.
8 Skaane P et al. Prospective trial comparing full-field digital mammography (FFDM) versus combined FFDM and tomosynthesis in a population-based screening programme using independent double reading with arbitration. Eur Radiol 2013;23:2061–71.