Keywords breast - digital radiography - screening - breast cancer - digital breast tomosynthesis
- digital mammography
Introduction
Previous randomized controlled trials (RCT) on mammography screening have found that
breast cancer mortality was reduced by 20–25% among women between ages 50 and 69 who
were invited to screenings [1 ]. These effects have been confirmed by major systematic reviews of prospective observational
studies [2 ]
[3 ].
The benefits of breast cancer screening programs result from detecting tumors at an
early stage, when therapy is often less invasive and the prognosis is significantly
better [4 ]
[5 ]
[6 ]. Within this group of early-detected, non-metastatic breast cancers, the prognosis
depends mainly on intrinsic factors, especially histological grading [7 ]
[8 ]. For example, the detection of grade 2 or 3 breast cancer reduces breast cancer-related
mortality more than that of grade 1 breast cancer (risk reduction 0.68 or 0.65 vs.
0.94) [9 ].
From the perspective of radiology, digital breast tomosynthesis (DBT) reduces tumor
masking by overlying breast tissue compared to digital mammography (DM) [10 ]. The randomized controlled TOSYMA trial (TOmosynthesis plus SYnthesized MAmmography
Screening Study) embedded in the German Mammography Screening Program (MSP) demonstrated
in women of ages 50 to 70 years that the detection rate for invasive breast cancer
is significantly higher with screening using DBT plus synthesized mammography (DBT+SM)
(71/10,000 screened women) compared to DM (48/10,000) [11 ]. A subsequent subanalysis showed that with DBT+SM, the rate of detecting breast
cancer in UICC stage I (tumor size up to 20 mm without regional metastasis or distant
metastasis) was significantly increased, and this increase was mainly due to the detection
of stage I carcinomas of grades 2 or 3 (+12.3 per 10,000 compared to DM) [12 ].
To sharpen the focus to diagnose such breast cancers, knowledge and expertise of the
leading mammographic tumor signs are important. For this reason, this exploratory
subanalysis of the TOSYMA study compares the frequency of the individual mammographic
tumor signs – and their combinations – that enabled detection of stage I breast cancer
in the screening group with DBT+SM and the group with DM, and examines whether the
morphological criteria differ in light of the histological grade.
Materials and Methods
Study design
Phase 1 of the TOSYMA study was conducted from July 2018 to December 2020 in 17 screening
units in the German states of North Rhine-Westphalia and Lower Saxony. In the process,
a total of 99,689 women were randomized 1:1 to the test arm (DBT+SM) or the control
arm (DM). The study protocol was approved by the responsible ethics committee and
rated positively by two other ethics committees. All study participants provided their
written consent. The study protocol, the results of the first primary endpoint – including
some secondary endpoints – and several exploratory sub-analyses have already been
published [11 ]
[12 ]
[13 ]
[14 ]
[15 ]
[16 ].
Study participants
Women ages 50 to 69 were invited by letter every two years to participate in the German
MSP, temporarily up to the age of 70 due to a COVID-19 regulation. In the catchment
areas of the study centers, a personal invitation to the TOMSYA study was added to
the regular invitation letter. Women who had been diagnosed with breast cancer up
to five years previously or who had had a mammogram within the last 12 months were
not eligible. Breast implants or repeated TOSYMA participation were specific exclusion
criteria of the study [13 ]
[15 ].
Preparing the screening examination
The opportunity to participate in the study was offered at 21 locations (Northwest
Lower Saxony (Wilhelmshaven), Hanover, North Lower Saxony (Stade), Central Lower Saxony
(Vechta), Northeast Lower Saxony (Lüneburg), Duisburg, Krefeld/Mönchengladbach/Viersen,
Wuppertal/Solingen (Bergisches Land/Mettmann district), Aachen-Düren-Heinsberg, Cologne
rechtsrheinisch (Bergisch Gladbach), Münster-South/Coesfeld, Bottrop, Gelsenkirchen,
Recklinghausen, Minden-Lübbecke/Herford, Bielefeld/Gütersloh, Hamm/Unna/Märkischer
Kreis (Schwerte), Höxter, Paderborn, Soest (Lippstadt), Münster-North/Warendorf).
Seven different mammography device manufacturers were used to prepare the DBT+SM or
DM examination: Amulet Innovality (Fujifilm Cooperation, Tokyo, Japan; n=10,075 examinations),
Class Tomo (IMS Giotto, Sasso Marconi, Italy; n=7,970), Lorad Selenia 3Dimensions
(Hologic, Malborough, USA; n=10,955), Lorad Selenia Dimensions (Hologic, Malborough,
USA; n=40,645), MAMMOMAT Inspiration (Siemens Healthineers, Erlangen, Germany; n=6,759),
MAMMOMAT Relevation (Siemens Healthineers, Erlangen, Germany; n=12,917), Senographe
Essential (GE Healthcare, Chicago, USA; n=10,237).
In both study arms, the examinations included the cranio-caudal and the medio-lateral-oblique
projection for each breast. In the test arm, in addition to the synthesized mammograms
(SM), stacked slices of ≤ 1 mm thickness were reconstructed to view the findings [11 ]
[13 ]
[15 ].
Mammographic findings
As in the ongoing MSP, an independent double reading was performed in both study arms
by the same certified physicians. A total of 83 experienced readers participated,
each of whom had previous screening experience of at least two years with at least
5,000 screening examinations per year. Before the start of the TOSYMA study, DBT training
was conducted for all study physicians by the Reference Center for Mammography in
Münster.
In case of abnormalities, the results were discussed with the physician responsible
for the program in the consensus conference in order to decide on further diagnostic
assessment. During the consensus conference, mammographic abnormalities that prompted
a recall were documented digitally and used in the current subanalysis.
The diagnostic assessment of study participants did not differ from the established
MSP procedure and included, depending on the findings, a clinical examination and,
if necessary, further mammographic projections (e.g. magnification mammograms or DBT),
sonography, MR scans, or minimally invasive diagnostic procedures.
All screening data were stored centrally and electronically in the MaSc documentation
system (KV-IT GmbH, Dortmund, Germany) [15 ].
Histological assessment
The 32 pathologists involved, all with at least 15 years of experience, made at least
100 histopathological diagnoses annually as part of the MSP, and they attended mandatory
training every two years, in addition to self-assessment procedures. The study training
focused on the Nottingham grading system for invasive carcinomas, based on a semiquantitative
assessment (1 to 3) of glandular differentiation, nuclear pleomorphism, and the number
of mitoses per square millimeter (G1: ScoreΣ 3–5, G2: ScoreΣ 6–7, G3: ScoreΣ 8–9)
[4 ]
[8 ]
[12 ]
[17 ].
Study data
The stages of breast cancer were classified based on postoperative histological findings
or, in case of neoadjuvant treatment, clinical imaging according to the TNM classification
of the Union for International Cancer Control (UICC) [18 ]. In multifocal or multicentric carcinomas, the largest tumor diameter was used for
staging [18 ]; in bilateral breast cancer, the side with the highest TNM classification was used.
UICC stage I is defined as a tumor size ≤ 20 mm (T1), no regional metastases (N0,
N1mi) and no distant metastases (M0). Women with stage I breast cancer were divided
into two groups, those with histological grade 1 and those with grades 2 or 3. In
case of multifocality or multicentricity, the highest grade was used.
Previously published results from the TOSYMA study reported the detection rates of
invasive breast cancer (iCDR) per 10,000 examined women stratified by stage and histological
grade based on the aforementioned grouping [12 ]. This group also forms the basis of this subanalysis. Of the 99,689 randomized women,
66 women who had not undergone any study examination and 10 women who repeatedly participated
in the TOSYMA study were excluded from the analysis set. In addition, we excluded
women with missing data regarding screening results, breast density, and UICC stage.
Accordingly, the current subanalysis included 49,462 women in the DBT+SM arm and 49,669
women in the DM arm [12 ].
Tumor signs
The frequencies of tumor signs documented during the consensus conference and leading
to diagnostic assessment with detection of stage I invasive breast cancer after independent
double reading were collected from the screening software. These were: masses, microcalcifications,
architectural distortions, asymmetries, densities, or their combinations [19 ].
Statistical analysis
The descriptive subanalysis included the assessment of tumor signs for screening-detected
stage I breast cancer at the time of the consensus conference, which the callback
for diagnostic assessment was based on. For these categorical variables, the absolute
frequencies were determined for each study arm overall and stratified by histological
grade, i.e. grade 1 versus grade 2 or 3. These were determined for each mammographic
tumor sign or their combinations as detection rates (DR) per 10,000 women examined.
Subsequently, the DR differences between the two study arms were calculated.
Results
Stage I breast cancer was found in 255 women in the DBT+SM group (DR 51.6 per 10,000)
and in 149 women in the DM arm (30.0/10,000). The most common individual mammographic
tumor sign leading to the diagnosis in both study arms was masses (DBT+SM: 20.2 per
10,000 vs. DM: 16.1 per 10,000; difference +4.1 per 10,000), less frequently, it was
based on the morphology of microcalcification (DBT+SM: 6.1/10,000 vs. DM: 3.0/10,000;
difference +3.1/10,000) or the morphology of architectural distortion (6.1 vs. 2.4/10,000;
difference +3.7/10,000). For all tumor signs considered above, DR was higher with
DBT+SM than with DM, the largest DR difference between the two study arms was achieved
with the combination of mammographic tumor signs: it was 12.4/10,000 higher with DBT+SM
(18.0/10,000) than with DM (5.6/10,000). In both study arms, asymmetries and densities
were rare as tumor signs for stage I ([Table 1 ], [Fig. 1 ], [Fig. 2 ]).
Table 1 Frequencies of mammographic tumor signs of stage I breast cancer with corresponding
detection rates of the randomized controlled screening study TOSYMA.
Mammographic tumor signs*
DM
(DR per 10,000)
DBT+ SM
(DR per 10,000)
Difference
Difference in
DM: Digital mammography.
DBT+SM: Digital breast-tomosynthesis plus synthetic mammography
*Mammographic tumor signs based on the ACR BI-RADS Atlas, Breast Imaging Reporting
and Data System [19 ].
Examinations included in the DM arm: 49,669
Examinations included in the DBT+SM arm: 49,462
DR: Detection rate per 10,000 examinations
Mass
80 (16.1)
100 (20.2)
20
+4.1
Combination
28 (5.6)
89 (18.0)
61
+12.4
Microcalcification
15 (3.0)
30 (6.1)
15
+3.1
Architectural distortion
12 (2.4)
30 (6.1)
18
+3.7
Asymmetry
3 (0.6)
0 (0)
−3
−0.6
Density
9 (1.8)
2 (0.4)
−7
−1.4
No data
2 (0.4)
4 (0.8)
Total
149 (30.0)
255 (51.6)
106
+21.6
Fig. 1 Part of a single layer of a screening examination with digital breast tomosynthesis
of a right cranio-caudal view of the lateral quadrants. Marked with a circle is an
architectural distortion measuring 15 mm, which has centrally amorphous microcalcifications.
Histology: Invasive breast carcinoma, pT1b (8mm), pN0, cM0.
Fig. 2 Part of the magnification view right cranio-caudal of the lateral quadrants, which
was obtained as part of the assessment. Amorphous microcalcifications are marked with
a circle, morphologically depicted comparable to the single layer in Figure 1. Magnifying
mammography does not reveal the architectural distortion shown by digital breast tomosynthesis.
Histology: Invasive breast carcinoma, pT1b (8mm), pN0, cM0.
In both study arms, fewer stage I grade 1 carcinomas (DBT+SM) were detected: 18.0/10,000
vs. DM: 8.7/10,000; difference +9.3/10,000) compared to grade 2 or 3 (DBT+SM: 33.6/10,000
vs. DM: 21.3/10,000; difference: +12.3/10,000) ([Table 2 ], [Table 3 ]).
Table 2 Frequencies of mammographic tumor signs of stage I breast cancer histological grade
1 with corresponding detection rates of the randomized controlled screening study
TOSYMA.
Mammographic tumor signs*
DM
DBT+ SM
Grade 1
Difference
Difference in
DM: Digital mammography
DBT+SM: Digital breast-tomosynthesis plus synthetic mammography
*Mammographic tumor signs based on the ACR BI-RADS Atlas, Breast Imaging Reporting
and Data System [19 ].
Examinations included in the DM arm: 49,669
Examinations included in the DBT+SM arm: 49,462
DR: Detection rate per 10,000 examinations
Mass
25 (5.0)
32 (6.5)
7
+1.5
Combination
9 (1.8)
39 (7.9)
30
+6.1
Microcalcification
2 (0.4)
6 (1.2)
4
+0.8
Architectural distortion
4 (0.8)
12 (2.4)
8
+1.6
Asymmetry
1 (0.2)
0 (0)
−1
−0.2
Density
2 (0.4)
0 (0)
−2
−0.4
No data
0 (0)
0 (0)
Total
43 (8.7)
89 (18.0)
46
+9.3
Table 3 Frequencies of mammographic tumor signs of stage I breast cancer histological grade
2 or 3 with corresponding detection rates of the randomized controlled screening study
TOSYMA.
Mammographic tumor signs*
DM
DBT+ SM
Difference
Difference in detection rate DBT+SM – DM per 10,000
DM: Digital mammography
DBT+SM: Digital breast-tomosynthesis plus synthetic mammography
*Mammographic tumor signs based on the ACR BI-RADS Atlas, Breast Imaging Reporting
and Data System [19 ].
Examinations included in the DM arm: 49,669
Examinations included in the DBT+SM arm: 49,462
DR: Detection rate per 10,000 examinations
Mass
55 (11.1)
68 (13.7)
13
+2.6
Combination
19 (3.8)
50 (10.1)
31
+6.3
Microcalcification
13 (2.6)
24 (4.9)
11
+2.3
Architectural distortion
8 (1.6)
18 (3.6)
10
+2.0
Asymmetry
2 (0.4)
0 (0)
−2
−0.4
Density
7 (1.4)
2 (0.4)
−5
−1.0
No data
2 (0.4)
4 (0.8)
Total
106 (21.3)
166 (33.6)
60
+12.3
UICC I grade 1 carcinomas were detected significantly more frequently with combinations
of tumor signs using DBT+SM than with DM, the DR difference was +6.1/10,000. Among
the individual tumor signs, the difference of DR DBT+SM minus DM was +1.5/10,000 for
masses, +0.8/10,000 for microcalcifications, and +1.6/10,000 for architectural distortions
([Table 2 ]).
UICC stage I breast cancer grades 2 or 3 were detected more frequently with combinations
of tumor signs with DBT+SM (DR 10.1/10,000) than with DM (3.8/10,000), corresponding
to a DR difference of +6.3/10,000. Here again, masses were the leading individual
tumor sign (DBT+SM: 13.7/10,000 vs. DM: 11.1/10,000); however, the DR difference DBT+SM
minus DM was smaller (+2.6/10,000) compared to combinations of tumor signs. The DR
differences due to microcalcifications (4.9 vs. 2.6/10,000; difference +2.3/10,000)
and due to architectural distortions (3.6 vs. 1.6/10,000; difference +2.0/10,000)
were of similar magnitudes ([Table 3 ]).
Discussion
As the first RCT, the multicenter TOSYMA study showed that in a national, population-based
mammography screening program, the invasive detection rate with DBT+SM is higher than
with DM [11 ]. A supplementary subanalysis also showed that the DBT+SM screening particularly
increased the detection of stage I breast cancer at grades 2 or 3 [12 ]. Increased detection of breast cancers of these grades at early tumor stages may
have a potentiating effect on reducing breast cancer mortality [9 ]. Therefore, an imaging characterization of the mammographic tumor signs leading
to the detection of these carcinomas in screening with DBT+SM or DM is of particular
interest.
In addition to the previously published grading-related detection rates of UICC I
carcinomas [12 ], the present subanalysis of the TOSYMA study presents the associated mammographic
tumor signs that led to diagnostic assessment after independent, qualified double
reading of the screening mammography with resulting consensus conference. The documented
tumor signs were based on the BIRADS Atlas, 5th edition, in accordance with study
training, and referred to the level of the examination [19 ]. The additional characterizations from the diagnostic assessment were not included
in the analysis.
Masses are the most common abnormality of breast cancer in DBT screening [20 ]. With DBT+SM and DM, masses were the most common tumor sign of stage I carcinomas
in the TOSYMA study. Consistent with the literature, DBT achieves higher detection
for breast cancers up to 20 mm than DM [21 ]. In the TOSYMA study, DBT+SM showed a higher detection compared to DM of +4.1/10,000
in stage I, which was more pronounced in histological grades 2 or 3 at +2.6/10,000
than in grade 1 (+1.5/10,000). The lesion-related detection rate in the DBT+SM arm
of 13.7/10,000 corresponded to 41% (68 of 166) of the total stage I grade 2 or 3 breast
cancers detected. By reducing masking, DBT can help to assess margins more accurately
and can more clearly identify highly suspicious morphologies, such as spiculation,
compared to DM [10 ]
[20 ]. Although screening-detected spiculated lesions are described as having a favorable
prognosis [22 ]
[23 ], DBT+SM does not dominate when it comes to detecting grade 1 breast cancers, but
does so instead with prognostically more significant grades [9 ]
[22 ].
Microcalcifications also resulted in diagnoses of grade 2 or 3 breast cancer more
frequently with DBT+SM than with DM. At the same time, the difference in detection
rates after DBT+SM screening was greater for grade 2 and 3 tumors (+2.3/10,000) than
for grade 1 (+0.8/10,000). In the DBT+SM arm, approx. 14% (24/166) of UICC stage I
grade 2 or 3 breast cancers were detected via microcalcifications (DM approx. 12%
(13/106)). The reason could be the increase in contrast of microcalcifications – despite
varying technical resolution limits [10 ]. In line with this finding, the detection of ductal carcinoma in situ with DBT+SM
versus DM did not show marked inferiority in the TOSYMA study (DBT+SM: 12/10,000;
DM: 13/10,000) [11 ]. Breast cancers smaller than 15 mm with microcalcification have an increased breast
cancer mortality risk [24 ]. Screening with DBT+SM could therefore slightly increase the diagnosis frequency
of such breast cancers with a less favorable prognosis compared to DM. The equipment
technology used does not support the need for double exposure (DBT+DM) for microcalcification
imaging in screening.
Regarding the isolated architectural distortion, DBT+SM also led to a higher detection
rate of grade 1 as well as grades 2 or 3 breast cancer compared to DM, but the DR
differences between the study arms were very similar for both grade strata (+0.16
and +0.20/10,000, respectively). Accordingly, the architectural distortion in the
diagnosis of stage I breast cancer is not strikingly typical for stage 1 breast cancer.
DBT+SM may allow a somewhat clearer perception of subtle architectural distortions
associated with stage I breast cancer, detecting 11% (18/166) of UICC stage I grade
2 or 3 breast cancers compared with approx. 7.5% (8/106) with DM. The number of UICC
I stage 1 breast cancers detected by architectural distortions in screening was lower
in both arms than in grades 2 or 3 (DBT+SM: 12 vs 18, DM: 4 vs. 8). However, a higher
rate of benign findings is described for DBT compared to DM, such as radial scars,
fibrosis, and sclerosing adenosis [25 ].
The greatest superiority of DBT+SM over DM was found in the combination of individual
tumor signs (grade 1: +6.1/10,000, grade 2 or 3: +6.3/10,000). By reducing superposition
effects, sensitivity could be increased by perceiving a combination of suspicious
signs. For example, DBT can more accurately identify tissue thickening than DM, the
combination of which with microcalcifications increases the probability of malignancy
[26 ]. Of UICC stage I breast cancer grade 2 or 3, the combination of tumor signs with
DBT+SM screening led to the detection of approx. 30% of all tumors (50/166). The perception
of combined tumor signs by readers – or in the future also by the use of artificial
intelligence – seems to be very important to identify as an abnormality in screening
[19 ].
It should be emphasized that the increased tumor detection in the DBT+SM study arm
is not due to an increased recall rate for diagnostic assessment (recall rates DBT+SM:
4.9% vs. DM: 5.1%), but was instead associated with a higher positive predictive value
(PPV 1 DBT+SM: 17.2%, DM: 12.3%) [11 ].
TOSYMA is the largest randomized, controlled screening trial to date to study DBT+SM
versus DM with almost 100,000 study participants. It enables additional exploratory
analyses based on successful randomization. The pragmatic approach has a high degree
of external validity and also demonstrates practical feasibility, particularly due
to the inclusion of numerous screening units and equipment technologies. All diagnosticians
were experienced and did not differ between the study arms or between the study and
routine screening [15 ].
The study also had limitations. TOSYMA analyzed only one round of screening. So differences
between study arms may be influenced by an initial prevalence screening effect with
DBT+SM. In addition, there may be a learning curve in the interpretation of tomosynthesis.
The comparison with previous examinations of the DBT+SM arm may have had a complicating
influence on the findings, since no DBT+SM preliminary examinations were performed
in the screening program [15 ]. Furthermore, the TOSYMA study did not include any sub-differentiations regarding
mass margins, calcification morphologies, or calcification distributions. Due to the
large number of possible combinations of radiological tumor signs and the resulting
limited number of cases, a differentiated, comparative presentation within this stratification
group was not carried out.
Clinical relevance
In DBT+SM screening, the detection rate for UICC I breast cancers is higher than in
DM: this is due to single tumor signs, such as masses, microcalcifications, and architectural
distortions, as well as their combinations.
The detection rates for UICC I grade 2 or 3 breast cancers are higher in both study
arms than those for grade 1 cancers: the increased detection rate for grades 2 or
3 using DBT+SM is based about half on the combination of tumor signs.
A systematic evaluation of DBT+SM screening examinations has the potential to enhance
intended screening effects through the increased detection of various single and,
particularly, combined mammographic tumor signs.
