A randomised trial of screening with digital breast tomosynthesis plus conventional digital 2D mammography versus 2D mammography alone in younger higher risk women.
Summary (3 min read)
- In recent years digital breast tomosynthesis (DBT) has become established as a useful adjunct to, and in some centres a replacement for, conventional full-field digital mammography (FFDM).
- Imaging the breast with DBT as a stack of contiguous thin (usually 1 mm thick) slices allows separation of overlying opacities, thus making cancers more obvious and resolving composite densities.
- Several large trials together with in-service observational studies have demonstrated a significantly higher sensitivity of combined FFDM and DBT compared to FFDM alone together with the potential to reduce the false positive rate .
- False positive rates in the UK National Health Service Breast Screening Programme are more common in women undergoing a first screen than those undergoing repeat (incident screening) (7.9% v. 3.0% ), as the lack of a prior mammogram for comparison can increase uncertainty about benign glandular densities.
- These younger women are often anxious because of their increased breast cancer risk , and measures which potentially reduce the need for recall in this group are worthy of investigation.
- The primary study was a prospective parallel group study of asymptomatic women undergoing two rounds of incident screening a year apart, conducted in two large UK breast screening centres.
- The study utilised a crossover design with 1:1 block randomisation to either conventional 2-dimensional full-field digital mammography alone (hereafter referred to as 2D mammography) or 2D mammography plus digital breast tomosynthesis (referred to as 2D+DBT), stratified by centre.
- The crossover at the second screen was to give equitable access to DBT and to reduce disparities in the radiation dose which participants received within the study.
- Recruitment of these women was terminated in the second year of the trial due to futility.
- The trial was approved by the UK National Research Ethics Service (North West - Greater Manchester West committee).
- Asymptomatic women aged 40 to 49 years previously assessed in a family history clinic (FHC) as being at moderate or high risk of breast cancer as defined by NICE (the National Institute for Health and Care Excellence) .
- The authors have previously demonstrated an expected survival benefit of screening in this group .
- Some previously screened women were scheduled only for mammography due to clinic capacity limitations.
- Eligible women were identified from NBSS and sent an invitation letter and participant information leaflet with the mammogram appointment.
- Recruitment in Centre A was from April 2014 to August 2015 and in Centre B from September 2014 to November 2015.
- Randomisation was performed after written informed consent using a commercial webbased randomisation service (www.sealedenvelope.com).
- The participant’s study number 8 was the sole identifier.
- Randomisation was with random permuted blocks within the strata with stratification by study site.
- The outcome of the randomisation was displayed on the web page and was manually recorded on the clinic list.
- The date, participant’s age at entry, trial centre, screen type (prevalent or incident), whether the woman was also undergoing annual screening MRI and the assigned first examination for randomised incident women were entered.
- Mammography was performed on Hologic Selenia Dimensions machines.
- Those undergoing incident screening were randomised to undergo either 2D mammography alone or 2D+DBT as the first examination.
- All readers underwent prior training in DBT image interpretation, either through participation in the TOMMY trial  or through attendance at an NHSBSPapproved DBT course.
- Where there was disagreement in the final opinion, a consensus discussion took place between the readers, with arbitration by a third reader if necessary.
- The findings at assessment were recorded, together with whether needle biopsy was performed.
- The primary outcome measure for participants undergoing incident screening at trial entry was the recall rate for further assessment with 2D mammography and 2D+DBT respectively.
- For participants undergoing prevalent screening at trial entry, the planned primary outcome measure was the recall rate for further assessment.
- Planned secondary outcome measures included cancer detection rates and individual reader recall rates.
- For prevalent screens it was intended to examine the number of women whose 2D mammograms were regarded as abnormal (recorded prior to viewing the DBT images) relative to the overall decision based on viewing the 2D+DBT images together.
- These were based on historic data from Centre A. Incident screening: A 7% recall rate was anticipated for incident screening with FFDM alone, reducing to 4.9% with the addition of DBT.
- For 90% power to detect a difference of this magnitude as significant using McNemar or conditional logistic regression estimates (5% level, two-sided testing) would require 1305 women and 2610 screens.
- An 11.5% recall rate was anticipated for prevalent screening with FFDM alone, reducing to 8% with the addition of DBT.
- These numbers were insufficient for a comparative trial with adequate power, hence a single arm study of combined 2D+DBT was performed.
- A study size of 528 would give approximately 80% power for a one-sample test of these recall rates with an expected confidence interval of less than 2.3% in either direction.
- A generalised estimating equation regression model with an exchangeable correlation structure was used to compare recall and biopsy rates between 2D only and 2D+DBT examination strategies.
- This analysis accounted appropriately for the crossover design of the study, namely, each woman was examined using either 2D only or 2D+DBT at the first of two incident screening rounds and by the alternative examination strategy at the second incident screening round.
- Differences between sites (Centres A and B) and the effect of the order of the examination strategy were also assessed.
- Analysis was by intention to treat (i.e. by randomised examination strategy allocation).
- A total of 1227 women were recruited and randomised, 608 to 2D mammography only at first screen and 619 to 2D+DBT.
- Two women were randomised to 2D+DBT but incorrectly underwent 2D mammography only as the first study examination (both had 2D+DBT at the second screen).
- Table 1 shows the number of women screened, the number of cancers detected and the number of false positive recalls in each arm.
- There was no difference between the groups for either the overall recall rates or the false positive recall rates (2D 2.8% and 2.4%, 2D+DBT 2.7% and 2.2%; Table 2).
- The corresponding numbers for Centre B are 1 after 2D and 2 after 2D+DBT.
- Table 4 shows the number of women recalled for further assessment.
- None of these women were diagnosed with breast cancer.
- No further analysis has been performed due to the low numbers.
- There is now a substantial body of published literature which demonstrates that digital breast tomosynthesis has a higher sensitivity for cancer detection than conventional full field digital mammography alone      .
- Many of the studies that have shown a reduced recall rate with DBT had a higher recall rate with 2D mammography than ours       and thus had a larger proportion of women who could potentially benefit.
- This was despite the fact that all readers were trained in DBT and almost all had experience of using it routinely in screening assessment prior to the start of the study.
- This ‘learning curve’ effect did not have a significant impact upon the false positive recall rate in their study due to the routine use of consensus discussion of 17 cases where the readers disagreed, together with arbitration by a third reader when necessary.
- The potential remains for DBT to reduce the recall rate at prevalent screening, but this question has yet to be answered.
CONFLICT OF INTEREST STATEMENT
- The authors confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome.
- 14/428 (3.3%; 2.0% - 5.4%) 22/438 (5.0%; 3.3% - 7.5%) p=0.20 Second study screen.
- 12/401 (3.0%; 1.7% - 5.2%) 8/401 (2.0%; 1.0% - 3.9%) p=0.36.
Did you find this useful? Give us your feedback
"A randomised trial of screening wit..." refers methods in this paper
...Published studies of screening DBT have utilised various reading methods, including double reading by two radiologists , single reading with computedaided detection (CAD) [16,26] and, in a multicentre study of over 450,000 women, presumably either single reading or a mixture of methods ....
Related Papers (5)
Julianne S. Greenberg, Marcia C. Javitt +3 more
Samantha P. Zuckerman, Emily F. Conant +6 more
Francesca Caumo, Giovanna Romanucci +5 more
Emily F. Conant, Samantha P. Zuckerman +7 more
Per Skaane, Sofie Sebuødegård +5 more