June 25, 2020, summary by Cécile Bour, MD
G. Welsch, professor at the Center for Surgery and Public Health Brigham and Women's Hospital proposes an analysis of breast cancer screening by MRI in an educational video, using data from the "DENSE" trial (results published in 2019), with the reservations (see comment at the end of the article) that were expressed after the publication of the results in the NEJM.
What is this trial?
This is a randomized trial, of good quality.
Explanation of the method on the video at 1:12 video.
The researchers divide 40,373 women aged between 50 and 75 years of age, with extremely dense breast tissue and negative results on the initial screening mammogram, into two groups: in the " supplementary MRI" group or in the "screening mammogram only" group; more precisely 8,061 women in the "MRI invitation" group and 32,312 women in the "mammogram only" group.
A control mammogram is then performed for both groups after two years in order to compare the results on the number of cancers found.
The main finding was the difference between the groups in terms of incidence of interval cancers over a two-year screening period.
Additional MRI screening appears to be associated with fewer interval cancers compared to mammography alone in women with extremely dense breast tissue.
Specifically, the researchers found that the interval cancer rate was 2.5 per 1,000 screenings in 4,783 women in the MRI invitation group compared to 5 per 1,324 women in the mammography alone group.
However, in the analysis of these results, one important element is missing, according to G.Welsch: the exhaustive counting of all cancers in both groups.
The fact that there are more cancers in the screening-MRI group compared to the non-MRI group suggests that there are more detections with MRI. But this finding may also suggest something else, supporting the theory of variable kinetics of breast cancers: all cancers do not have same expression pattern, some of them possibly regress.
Regression of cancers?
We remind that this is a randomized trial. Since women are randomly assigned to one or the other group, it is expected that after two years we should have the same overall cancer rate in both groups.Video at 2:03
Cancers not anticipated by screening in women from the group without MRI would be necessarily expected to express themselves at the end of two years on the mammogram performed for both groups at the end of the study.
What is actually observed?
- No cancer found at the initial mammogram in either group.
- Globally more cancers found in the MRI group
- Fewer interval cancers in the MRI group
- At the end of the two-year control mammogram for end of the study, more interval cancers in the non-MRI group, as they were not anticipated by MRI.
- Fewer mammographic cancers at the end of two years for the MRI group since a fraction is anticipated on MRI.
===> In total, we get an excess of 5.4 cancers found in the MRI group (Video 3:34)
G.Welsch explains: in a randomized trial with two random distribution groups, what is expected is that at the end of the study we will have a similar 'total cancer' rate, since cancers not anticipated by MRI in the non-MRI group are logically detected later on the control mammogram done at the end of two years.
The picture illustrates the group of women screened with additional MRI on the left and the control group without MRI on the right. We see in the red column all the cancers detected by MRI resulting in an excess of detection; the green squares indicate the interval cancers which are more present in the group without MRI as they were not anticipated by this examination; the yellow squares symbolize the cancers seen two years later on the mammogram at the end of the trial which are more numerous in the women without the anticipatory MRI. Nevertheless, a comparison of the 'total-cancers' of the two groups clearly shows an excess of cancers for the group with additional MRI.
Now, what happened to those 5.4 excess cancers not found in the non-MRI group, are they cancers that will appear later?
Or have they disappeared?
In general, the explanation put forward is the hyper-slow growth of these cancers, which do not appear during mammography at 2 years of age, not progressing or very very slow progressing. This would mean that more than half of the cancers found by MRI are hyper-slow-growing cancers so that they are undetectable on mammography. Reminder at 3/30 of the video: we find in the trial 9.8 cancers in the MRI group, 5.4 excess cancers /9.8 total cancers in the MRI group = 0.5
The alternative explanation exists, and it is the cancerous regression, namely the disappearance of these cancers not found on the mammogram for control after two years. They simply disappear. (Editor's note: this hypothesis also emerges from the 2008 Oslo study, where the group of women screened every two years had a 22% excess of cancers detected over the group of women not screened, the two groups being compared after 6 years with a mammogram performed for each group.)
However, if all cancers were inexorably progressing and were expected to manifest themselves, the same number of cancers should be found at the end of the 6-year observation period in regularly screened women and in the non-screened women, in whom the cancers not detected by previous screenings should then be seen on the mammogram at the end of the study, at the end of 6 years. If this is not the case and there is an excess of cancers in the screened group, it is likely that some cancers that did not occur in the unscreened women have disappeared in the meantime.
Cancer regression is observed for other forms of cancer: kidney cancer (1/4 of cancerous lesions regress) of the thyroid (1/3 of cancerous lesions regress). So why not for breast cancer???
The benefit/risk balance
So let's consider the risk/benefit balance of increased MRI monitoring:
Overall, the results tends to suggest that the risks associated with increased MRI surveillance of dense breasts outweigh the benefits.
Good news and bad news
There are good news when it comes to breast cancer. Since the 90s breast cancer mortality has dropped by 40% (i.e. already before the arrival of national screening campaigns, editor's note).
This is an important reduction, according to G.Welsch who reminds us that this drop in mortality is due to therapeutic advances, and not to screening.
The bad news according to him is that screening is the target of a technological weapons race, since the first analog mammography, then mammography, the arrival of 3D (tomosynthesis) and now the advent of MRI, all aimed at finding more cancers. For what benefit?
Conclusion of the author
The challenge is not to find more and more cancers inducing unnecessary over-treatment, but to detect those cancers that are important to find because they are a threat to a patient's life.