Profile picture
Vinay Prasad MD MPH @VPrasadMDMPH
, 84 tweets, 30 min read Read on Twitter
Going to slowly tweet out 15 of the most interesting things I read, listened to and saw in 2018. With oncology/ health policy/ evidence themes

Starting with Same Dataset, Different Interpretations by @BrianNosek nature.com/news/crowdsour…
The point in this paper is if you give researchers the (a) exact same data set and ask the (b) exact same question you get a variety of answers.

Some point out that *some* of the point estimates are consistent....
Of course that is true, but other teams are reporting qualitatively different conclusions: some say no evidence of bias, others say bias

And, imagine a second what happens if...
1. You free researchers to build their own dataset or select there own data
2. You pick a question where there is a broader range of preconceived notions (e.g. does the IVC filter help? Is saturated fat good or bad?)
3 You don't have analytic teams review each other
4. You allow publication in any of 100k journals
5. You let anyone put out a press release without any accountability
6. You place no restrictions on analysis plan
7. There can be far more than 29 teams
8. Pre-registration is <100%
You see-- what the 'real world' looks like is just MUCH MORE flexibility.
Almost surely this will translate into greater VARIATION
Some questions using retrospective data with strong preconceived notions are likely little more than self fulfilling prophecy
This is supported by work by Chirag Patel on vibration of effects. Another great paper.
#2 Bad Blood by @JohnCarreyrou
I think I read the entire book on 2 flights from PDX to Europe
It was that riveting
I think the simple way to interpret the book is that this is an isolated case of terrible fraud and cover-up, and I completely agree that this is an EXCEPTIONAL case. But I think some of the core failures are far more common
You see the company promised that *by offering their technology* we would get lots of info for people to use and therefore leverage improved health outcomes. They had a grandiose vision, many shared.
Shown here
But at every step there is some validation needed to prove that hope is met with reality.
Theranos failed early and hard with the very first step.
But many other companies and services also present grandiose futuristic views of medicine... but few people out there consider the proper validation needed for these companies
Doesn't matter if you fail early due to fraud or you fail late due to hubris in both cases people are either (a) harmed or (b) not benefitted, and you are equally a failure/ seller of snake oil IMHO

Read my longer take here:
#3 two papers showing progression free survival is a poor surrogate for Quality of Life.
onlinelibrary.wiley.com/doi/abs/10.100…
and
jamanetwork.com/journals/jamai…
In cancer medicine, we are in punchdrunk love with surrogate endpoints (typically changes in tumor size on CT scans)
We love to use these for regulatory approval (2/3 of all our drugs)
Previously, we showed that the vast majority of these surrogates had a lousy correlation with overall survival
jamanetwork.com/journals/jamai…
In the largest update to date, @AlysonHaslam confirmed this was the case
drive.google.com/file/d/16NYOFB…
Well, if PFS does not predict overall survival, what if it predicts QoL is the natural next question. In 2018, Both kovics and @oncology_bg found that it had a lousy correlation with quality of life.
A decade ago @EAEisenhauer and Chris Booth asked if PFS is meaningful or simply measurable?
ascopubs.org/doi/full/10.12…
In 2018, we can say that, on the whole, for the most part, PFS poorly correlates with OS, and poorly correlates with HR QoL. You can measure it and approve more drugs, but there is remarkable uncertainty if you make cancer patients better off.
#4 This lecture by JPA Ioannidis on nutritional epidemiology.
In part because he rips this awful field--- and it is awful, with endless press releases of low credibility-- into tiny bits but also because of what follows.
In the Q&A, there is a fierce argument between Ioannidis and Willett, who is probably one of the largest beneficiaries of low credibility nutritional science papers
Tweeps will know that I don't like when strong prouncements are made based on lousy data, such as this
That was the first time a reporter called me to cover a #TWEETORIAL
cbc.ca/news/health/se…
#5 The Letters to the Editor in the @TheLancet re #ORBITA
If you wrote in to fault the authors, I am sorry for what happened to you...
This one must have left a mark.
thelancet.com/journals/lance…
Gems like this
And this! LOL
Ouch ouch ouch....
Hahaha
Ooofff
Zing...

oh boy, I almost feel bad for those who wrote in
Btw none of the letter writers have spent the last year trying to change my slide showing all the data on the other hand.. All the positive, sham controlled trials of stenting chronic stable angina... here is the full list!
and @barttels2 is right, the topic is one I follow with interest as @adamcifu and I speculate about it in our 2015 book Ending Medical Reversal

The evidence q is sham trials for subjective endpoints & mechanical devices?
ANS = YES, ALWAYS!!
#6 Chris Booth and Alan Detsky in @NatRevClinOncol
Why do patients get treatments that are minimally effective? (At best, oh, and super toxic and costly)
nature.com/articles/s4157…
In this stellar commentary they address patient factors, doctor factors and industry factors, among others
They ask the tough question, as to whether this is right?
I had so much to say about this paper, that I spend 20 minutes of episode 1.15 of @Plenary_Session talking about it.
Check it out
soundcloud.com/plenarysession…
On a related note, I notice on Twitter, some celebrate cancer drugs proportionate to the # approved this year, or proportionate to the number of 0s on their consulting checks.
That's not the right measure of drugs
It is how much OS and QoL they provide
Avg Rx is marginal
#7 Demand cancer drugs that truly help patients
A tour de force essay that says something I cannot believe we debate
Cancer is deadly
For something deadly, we need therapies that make it less deadly
We don't just need CT scans to look different if patients don't live longer
With his essay, Dr. Agarwal joins other leaders in oncology, who have asked for more from our drugs
nature.com/articles/d4158…
A few, rare, leading voices
Once again @EAEisenhauer & Booth
Our regulators are failing us on this issue, here is why:
drive.google.com/file/d/1PiqKUz…
In the latest failure, they accept non-randomized data for AML ineligible for induction. That has a median OS of ~8months... are you kidding me
RCT is mandatory here
soundcloud.com/plenarysession…
Ironically, Rick Pazdur (FDA) himself said that for this disease and setting in an ODAC circa 2009!
#8 @OtisBrawley on the fall of WTC and potential risk of cancer. An example of when the editorial is better (much better) than the papers themselves.

2 paired papers linked first responders with incr. myeloma and all cancer
Brawley works methodically. First, he disarms you.
He is sympathetic to the goal of the work
He reminds us that in science honor means seeking and saying what is true, not merely what one wishes to believe or what might be convenient.
He quickly finds limitations with the choice of control group in Study 1
He reminds us that strength of the association is a key factor in considering truth from fiction, and more importantly identifying useful links from noisy ones.
Ooh, he gives a sweet little lecture about overdiagnosis
Seems like some people could probably use that lecture...
His questions rip holes in the second paper
And he ends with a meditation on the limits of science and cancer attribution.

And in one sentence puts more wisdom out there than half the journals I read.
#9 @KimmelmanJ and @JAMAInternalMed mapping the trials landscape of pregabalin or lyrica

This is a major & important paper mapping clinical trials.
What do you need to know?
First, when drug companies get their drug approved for 1 purpose, doctors can use it for off label purposes. So, you might imagine, once a company get's their drug on the market, they have a much lower bar
Instead of persuading regulators, all they need to do is persuade the average doctor. And unlike regulators at the FDA, you can wine and dine the average doctor, you can pay the 'thought leaders' and KOLs for 'consulting' and who knows? Maybe some might endorse off label use!
But doctors and KOLs like a veneer of evidence. They may not understand all its intricacies, but they want you to be able to hand them some uncontrolled, inconclusive, garbage seeding trial-- or Phase 2, to justify the off label use. So that is what @KimmelmanJ studies
He notices that although Lyrica has very limited FDA approvals, there is a glut of seeding (inconclusive) studies that generate some interest in the drug for off label purposes
There are far more of these exploratory trials than confirmatory ones! And remember, even garbage exploratory trials count as industry 'R&D' on SEC reports
But it isn't real R&D---- it is the hijacking of Evidence Based Medicine. Marketing masquerading as R&D.
Does not deserve any tax breaks IMHO.
We studied the off-label recs of 1 major cancer organization in 2018 (that mandates CMS to pay for these drugs), and similarly found many recommendations beyond @FDAOncology approvals based on WEAK data if anything at all! bmj.com/content/360/bm…
And, just a week later, a paper in JAMA shows that these low quality seeding trials really do translate into increased prescribing
jamanetwork.com/journals/jama/…
10. @venkmurthy and Kim Eagle's commentary on the 100 million dollar, NHLBI hailmary pass to show stenting chronic stable angina can reduce MI and death

What do you need to know?
In the aftermath of COURAGE, the field sought to once again test whether some strategy--this time selecting pts with myocardial ischemia and stenting them-- could demonstrate improvements in hard outcomes. @ProfDFrancis called this search for significance Desperate!
@venkmurthy and @keaglemd review in depth the protocol amendments made to ISCHEMIA trial midstream. These changes turned a trial with a small chance of providing useful information into one that will forever be tainted by design, incapable of answering a useful question
1. They changed the primary endpoint. Since the trial was not sham controlled, switching to endpoints that depending on or are contigent on doctor or patient beliefs introduces bias.
2. They elaborate on this point. Lack of blinding affects how doctors make decisions, which can rack up endpoints, whether or not patients are better off
3. They conducted a global study and gave away free stents... this may affect behavior in all sorts of hard to predict ways
4. They broadened the definition of ISCHEMIA in a way that introduces challenges
Why did this happen? Why did ISCHEMIA recruit poorly? The eternal question since CAST.
Because doctors in the USA and elsewhere already think they know the answer and are unwilling to randomize their patients, and.... they are making money hand over fist from the status quo.
More on that last point with #11.....
11. 2018 was the year in which the pervasive personal payments from pharma to doctors that cripple biomedicine, and distort so much of the evidence & interpretation, was made front page news by @charlesornstein and @katie_thomas
nytimes.com/2018/09/08/hea…
While we hope doctors judge medical evidence impartially, a wealth of psychological literature and empirical data from medicine, show the presence of these financial ties is a consistent bias
That bias points only one direction: towards more drugs, given earlier, continuous, regardless of whether survival is improved, no matter the side effects or opportunity cost. That bias is always more, newer, and negative results can be p-hacked to positive ones
All drugs are cost effective as long as the industry writes the analysis and gets the expert to ghost author it. This bias is the most corrosive force in biomedicine.
And for decades, the most we ever did about this, was ask for voluntary disclosure, but the @nytimes showed that over and over doctors disregarded it.
Non disclosure is common, after all, only an 'unsuccessful expert' has no financial ties -- and this isn't a joke. Financial payments from drug companies become seen as the hallmark of success. Trainees tell me how they wish they made as much as Dr. XXX or YY or Z
I call these 'closed financial loops' in an article in the hastings center and note that when they happen in goverment they are called corruption, but in medicine its just a footnote
researchgate.net/profile/Kevin_…
Also on this topic, comes a great article in Science by @cpiller on the cozy relationship between FDA and industry

Regulatory capture means the FDA comes to see the industry, and not the public as it's client
science.sciencemag.org/content/361/63…
Piller finds a similar result as @jungleland and I, that the most common post employment FDA vocation is working for or consulting for the industry you were just tasked with regulating
bmj.com/content/354/bm…
The most realistic next step to improve conflict of interest is to engage in this modest proposal by @VincentRK and I

This is a compromise solution-- it is a step forward-- it could be embraced tomorrow
nature.com/articles/bcj20…
I keep telling my colleagues. We will want to take the lead and self regulate this matter as a profession. Otherwise, outsiders will step in and do it for us. Only a matter of time.
Missing some Tweet in this thread?
You can try to force a refresh.

Like this thread? Get email updates or save it to PDF!

Subscribe to Vinay Prasad MD MPH
Profile picture

Get real-time email alerts when new unrolls are available from this author!

This content may be removed anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member and get exclusive features!

Premium member ($30.00/year)

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!