I want to do a TWEETORIAL on a controversial topic...
This is the first of a MULTIPART SERIES on Mammographic screening
This one will summarize one important paper in 2014
I start here because, this one is a dagger
TIme for TWEETORIAL PART 2 on mammographic screening.
A MULTI-part series on this topic
Today, we we will review this 2012 paper by
Archie Bleyer, a fellow OHSU faculty member, and.... H. Gilbert Welch
Sit down, It's a doozy.
As promised time for PART 3 TWEETORIAL on mammographic screening, which will cover this PROVOCATIVE paper by H Gilbert Welch at Dartmouth
Again, the purpose of this series is just to teach, rising above the usual debates
In Part 2, I took you through the population data, as mapped by Archie Bleyer and Gil Welch
In this THREAD, I will explain how Welch and Frankel estimate the probability that a woman who HAD A BREAST CANCER found by mammographic screening HAD HER LIFE SAVED (avoided dying of breast cancer) by that screening.
But, baseline: What's your guess?
Before I explain the paper, quick background. Studies show that the public has been MISLEAD about the benefit of screening. If you survey women, you find they believe that 10 yrs of mammograms is the difference btw the Right and Left
But, BEST CASE SCENARIO, the benefit is this
Right (no screening) vs. Left (screening)
I say best case because this assumes that breast cancer deaths avoided are not lost via off target deaths. Not the INTENTIONAL use of 39 OR 40 in the graphic
To make their estimate Welch and Frankel do something simple
First, they calculate what is the probability per 100k women of having a breast cancer found by mammography
It's pretty simple, its the 10 year probability of having a breast cancer (for a 50 year old woman) multipled by the percent of cancers found by mammography
So 1910 per 100k
Next they calculate the probability a death with or without mammography
Here they use 20 yrs (a conservative assumption aka favors screening
So they know 990/100 k deaths in the current world
They assumed it would be 20% higher without mammograms.
So mammograms avert 250 deaths per 100k
You gotta note that these are all very favorable assumptions for mammography. As shown in the first TWEETORIAL, there is huge uncertainty whether this 20% is true, it is, as I say, a BEST CASE Scenario
So using the best case scenario, one can make the calculation easily
Absolute 20 year reduction in death (optimistic) divided by real diagnoses from screening = 250/1910 =.....
20% is an optimistic estimate of cause specific death reduction, and we might be better off looking more at this range, here you see the prob. a woman who HAD CANCER found on screening had her life saved, across more plausible estimates
So Welch and Frankel end with this JUGGERNAUT of a statement
And remember this is BEST CASE SCENARIO
Why does this really matter?
Well, it is human nature to assume that HAD IT NOT BEEN FOR <whatever you did in your life> you would be <in some very bad state of being>
This is the same cognitive error that keeps residency work hours barbaric because HAD IT NOT BEEN FOR <my long hours> I would be <a bad doctor>
And pretty much every other entrenched, unsupported status quo thing you can imagine
With cancer screening, Welch and Frankel are wise to note the power of the anecdote seems to trump all data and all reason, they write:
Critics can view this paper as TOO PESSIMISTIC or OPTIMISTIC-- I am fairly sure it is too optimistic, but the major virtue of the paper is that it tries to get us to start using our best thinking to understand a question rather than resort to simple, false sentiments
Sadly, this is a lesson we desperately need in other topics, for instance cancer exceptional or super responders, where one must interrogate whether the outcome is truly due to the drug or was it due to the biology/ person