-
Gedanken experiment- what would provide the better screening mammography outcomes?
What would provide the highest accuracy in terms of PPV and NPV, all other variables equal:
Film screen mammography with independent double reading by separate readers
or
digital tomosynthesis with a single reader?
-
16 Replies
-
It’s not 2008 anymore.
My money is on DBT. Eventually DBT+AI verified by a rad. Can’t wait.
-
I hope CMS in there ultimate wisdom keeps up with inflation and time/equipment requirements for DBT + AI, especially now every 40 year female and up will be having mammograms.
-
DBT hands down. Even without AI. Our best and worst breast imagers improved their audit stats for CBR and CDR after implementation. No way double reading makes up for that.
-
Most breast cancers are actionable in hindsight on mammography, digital or screening. We all know this and it has been documented in published studies. I hope the promised new AI is of actual benefit compared to the old.
Hard to believe and counterintuitive, but the sensitivity of film screen mammography in the Breast Cancer Detection Demonstration Project (BCDDP) of the 1980s was around 80%.
DMIST sensitivity was about the same. Even now, its not that much different or better with all the advances and MQSA.
-
A large percentage of breast cancers on are actionable in hindsight on mammography, digital or screening. We all know this and it has been documented in published studies. I hope the promised new AI is of actual benefit compared to the current commercial software, which is terrible.
Hard to believe and counterintuitive, but the sensitivity of film screen mammography in the Breast Cancer Detection Demonstration Project (BCDDP) of the 1980s was around 80%.
DMIST sensitivity was about the same. Even now with tomo, its not that much different or better.-
The variables being discussed so far are not the most relevant ones. The biggest advantage of DBT is its ability to detect small invasive cancers all with having a lower CBR. This will give it a significant advantage when assessing mortality rate benefits, especially with the USPSTF assessing “harm” of false positives.
How was the sensitivity measured back in the 80’s? There were obviously many unknown cancers that did not declare themselves until much later. No way to know the proper denominator. What was the average size of those cancers they thought they had 80% sensitivity for?-
Old AJR discussion letter:
[link=https://www.ajronline.org/doi/pdf/10.2214/ajr.154.2.2105038?src=recsys]https://www.ajronline.org/doi/pdf/10.2214/ajr.154.2.2105038?src=recsys[/link] -
Old letter in AJR discussing sensitivity in the BCDDP:
[link=https://www.ajronline.org/doi/pdf/10.2214/ajr.154.2.2105038?src=recsys]https://www.ajronline.org/doi/pdf/10.2214/ajr.154.2.2105038?src=recsys[/link]-
Yes, they were discussing the problem of the statistics and how to truly calculate sensitivity back then. The concept of incident vs prevalent cancers is mentioned as well. Any sensitivity claim in mammography at time has to understand there was no method of finding very small invasive cancers the way we can now with US and MR. The denominator was well underestimated.
Film screen mammography was sensitive enough to improve survival by at least 30% in experienced hands and it wasnt nearly as good as tomo is now, even if the studies were double read.
Remember, we were being told by the USPSTF that the harm of mammo outweighs the survival benefit in women between 40-50. The decrease in CBR with DBT is enough to change that calculus.
DBT FTW.-
It says the sensitivity of mammography in the BCDDP was estimated at around about 80%, when interval cancers were included from the data.
That is about what the sensitivity is now. One would think the sensitivity would have been much greater with digital mammography during DMIST, but it wasnt. With tomo, the published sensitivities seem to be similar to slightly better spending on the study, but not as much improved as one would expect comparing to the 1980s. -
says the sensitivity of mammography in the BCDDP was estimated at around about 80%, when interval cancers were included from the data.
One would think the sensitivity would have been much greater with digital mammography during DMIST, but it wasnt. With tomo, the published sensitivities seem to be similar to slightly better depending on the study, but not as much improved as one would expect comparing to the 1980s. -
The letter in AJR says the sensitivity of mammography in the BCDDP was estimated at around about 80%, when interval cancers were included from the data.
One would think the sensitivity of digital,mammography would have been much greater than film screen mammography during DMIST, but it wasnt. With tomo, the published sensitivities seem to be similar to slightly better depending on the study, but not as much improved as one would expect comparing to the 1980s. -
The letter in AJR says the sensitivity of mammography in the BCDDP was estimated at around about 80%, when the interval cancers from the BCDDP were included from the data.
Most thought the sensitivity of digital mammography was going to be much greater than film screen mammography during DMIST, but it turned out it wasnt. With tomo, the published sensitivities seem to be similar to slightly better depending on the study, but not as much improved as one would expect comparing to the 1980s. We have apparently nudged the sensitivity only incrementally since the 1980s with new technology (and MQSA), but the main problems are the inherent limitation of the modality itself and the perceptual limitations of the readers.
-
I think the best estimate for sensitivity of both Mammo and US was the ACRs 6666 trial – that used both. The gold standard – which is critical to this discussion – was MR AND 3 year follow up. As I recall both mammography and US had a sensitivity of around 52% (when I get a chance, I will look up the real numbers to be more precise – but they were around 50%)
Anyone have an update to these?
-
Exactly Phil. The gold standard is critical.
Sensitivity is the wrong metric when comparing the effectiveness of breast cancer screening tools. -
[link=https://cancerstatisticscenter.cancer.org/#!/cancer-site/Breast]https://cancerstatisticsc…/#!/cancer-site/Breast[/link]
Since breast cancer screening was standardized (1990-ish), deaths from breast cancer have been falling. I don’t see a convincing change since the 2000 first FDA approval of an FFDM unit. It’s the kind of curve I’d expect. And if someone decides we need to bring back film/screen, I’ll go find some other career before I’ll deal with dark rooms and chemistry again.
-
-
-
-
-
-
-
-
-
