David Nunan Profile picture
Director MSc in EBHC Teaching & Education @UniofOxford @CebmOxford; Visiting Prof of EBM Petrópolis Med School; Leads @catalogofbias; Uncertainty Nerd #MedEd
Philip Dawid Profile picture C. Niederstadt, MD, MPH Profile picture Dr Javier Bueno Antequera 😃 Profile picture 3 subscribed
Feb 29 26 tweets 9 min read
Lots of takes on this. Many are what people have anyway regardless of this paper, such is the general diet/nutrition discourse, and particularly around UPF.

For those of you interested in what an actual evidence-based approach to this paper looks like, buckle up🧵

1/ In my view it is a good example of pervasive issue in the way most medical/health research is interpreted, and this even includes the authors themselves.

What’s the issue?

IGNORING UNCERTAINTY!

2/
Sep 22, 2022 17 tweets 5 min read
"It's going to protect YOU..."
"It will reduce YOUR risk of a heart attack, cancer, diabetes etc".

We see this all the time when it comes to medical treatments and health interventions.

I'm going to show why 99.9% of the time this type of phrasing/framing ("YOU/R") is wrong🧵 When people hear these phrases with "YOU" / "YOUR" in them, it's most likely they will perceive the benefits & risks of treatment/intervention in relation to their own personal benefit/risk. The thing is, we do not, & cannot, know your own personal benefit/risk.
Feb 8, 2022 9 tweets 3 min read
ON MISINFORMATION

Here's the thing. You cannot police it in a fair, reproducible & representative way.

The reason being is the lines between misinformation and genuine uncertainty soon become blurred.
Trust me, I tried.

(ps. does this tweet count as misinformation) I mean, look at Twitters own criteria.
"...widely available, authoritative sources'

Trouble is these sources are dealing mainly with evidence/information in the Green zone.
Apr 22, 2021 15 tweets 4 min read
So here's the thing with this. I posted a tweet out of frustration of seeing an organisation promote evidence from a commentary piece as if it were proof a question had been answered. No uncertainty. Proof.
/1 The same organisation didn't mention a systematic review attempting to answer the same question (which happened to show uncertainty). Yes, the organisation was aware of the review's existence.
This signals, to me, an agenda. It's also a good example of the state of things.
/2
Feb 5, 2021 46 tweets 20 min read
Our latest publication revisits a well-known problem: reporting of relative effect estimates without absolute effects in journal publications of clinical trials:

ebm.bmj.com/content/early/…

🧵on this issue
("nah, show me the findings": ) First an intro to the problem. A practical example is probably best here. Take a look at the image. How many more people are at increased risk bowel cancer? Image
Jan 2, 2021 11 tweets 4 min read
Lots of folk already commented on “1 or 2 dose”. One thing that comes up often is “efficacy”, followed by numbers like “80%”, 95%”. In many cases, it reads as if folk think this is how much YOUR chance of getting the virus is reduced by. That’s incorrect. These figures are actually the relative risk reduction (RRR) of infection with the vaccine. Eg. 2000 people without Covid-19 - 1000 vaccinated (group 1), 1000 not vaccinated (group 2).
200 people (20%) in group 2 get Covid-19
10 people (1%) in group 1 get Covid-19
= 95% RRR
Dec 2, 2020 8 tweets 2 min read
Here's a thing. In an effort to not be part of the problem, I responded to a tweet regarding the definition(s), & its misinformed application, of a common statistical concept when used in medical/health research The response has been an interesting mix of 'Even the greatest statisticians make mistakes'; frustration over-simplifying complex concepts; suggested definitions unlikely to be adopted by non-statisticians

As for evidence of the impact of the misinformed take = nothing concrete
Dec 2, 2020 5 tweets 4 min read
@MaartenvSmeden @dailyzad Thanks Maarten. Agree - lots of examples where wrong interpretations of statistics (in general) = correct/incorrect concl.

I'm interested in your comment re: 'trust in the limits of the CI'. A common method for assessing confidence in an observed estimate may do exactly this! @MaartenvSmeden @dailyzad Method = GRADE (no doubt you familiar with). For those who do not - a brief outline:

It's the predominant in medical/health outcomes research. Assesses factors deemed important for 'confidence' in observed estimates, usually from a body of evidence (e.g. systematic review).
Sep 30, 2020 15 tweets 5 min read
Have to say I’m somewhat surprised by these pronouncements, and modus thereof (RR without AR etc), particular given the expertise.

The replies here on methods are important, especially this:


⚠️ of “more research” feel cursory.

#classiccovid Also not pre-registered and certainly not PRISMA compliant in pre-print version.

Hmmm....
Sep 23, 2020 5 tweets 3 min read
Rod Jackson is a Prof of Epidemiology in NZ. He also trained in EBM w the founding fathers. He's on twitter (just) @rod_rt
I've had the pleasure of working w him for >10 years, including just last week:

He's one of the best teachers of Epi/EBM I've met
/1
Here's an example of why.

Using the current COVID pandemic, Rod takes us through how epidemiology works in way we can all understand, showing us why in his view of data COVID is 10x deadlier than flu.
nzherald.co.nz/nz/news/articl…

There's more...
Sep 22, 2020 39 tweets 11 min read
You may have seen a mini-Twitter storm on COVID tests claiming they falsely tell people they have COVID when they do not have COVID

Other’s have done a good job correcting this

I want to show how we work what the results of a COVID test actually mean 🧵

/1 The first things we need to know are how common is COVID-19 and how accurate is the test we are using to identify if someone has or does not have COVID.

Let’s focus on the first question - how common is COVID-19?

/2
Sep 12, 2020 5 tweets 2 min read
Hi @AerLingus
I was going to make this a negative rant. But instead, I will try to help you improve your customer relations.
The email (see pic) you sent regarding your cancelling my flight leads to unnecessary frustration for those wishing to request a refund.
/1 Image You helpfully provide direct links to the necessary webpages should I wish to exchange my flight for a voucher (thereby retaining my funds).

Whilst you provide information regarding a full refund, you don't provide a useful direct link.

/2
Jul 1, 2020 4 tweets 1 min read
Writing an article on lifestyle medicine - positives, negatives, known knowns, known unknowns.

What specific things would you want to see included?

Any key supporting articles/evidence would be helpful.

Thanks. Thanks to all who have responded to this. Very helpful points and references.
Jun 17, 2020 24 tweets 8 min read
A nice opportunity for me to show the usefulness of this:


It may surprise folk to know that many modern-day medical interventions [including 'ahem' exercise) are far from miracles (see examples in pic). But NNT of 8 & 25 is right up there w the best.

/1 Image Borrowing* figures from:
(*herein lies one of the problems with press-released data raised by others:
)

We can use @d_spiegel tool to explore how we can communicate the #RECOVERYtrial findings
understandinguncertainty.org/node/233

/2
May 20, 2020 16 tweets 7 min read
So about the @GOVUK addition of loss of smell and taste to the list of COVID-19 symptoms for self-isolation.

Was it really asleep at the wheel on this one?

I reviewed the evidence. Here's what it showed 🧵

/1 1st - Gov's website states “We have been closely monitoring the emerging data and evidence on COVID-19 and, after thorough consideration, we are now confident enough to recommend this new measure.” There's no link to the “emerging data and evidence”

gov.uk/government/new…

/2
May 3, 2020 5 tweets 3 min read
Seen some interesting takes on "EBM" and its role in the current pandemic.

For me, EBM = the answer you give when patient(s)/public ask(s) "Why is/was this part of the decision?"

Disliking your answer(s) to that question doesn't = "EBM doesn't work"

And twitter.com/i/events/10347… On RCTs - this by @stephensenn on arguments for/against: google.co.uk/amp/s/errorsta…

“Show me how you’ll analyse something & I’ll tell you what allocations are exchangeable. If you refuse to choose one at random I’ll say, “why? Do you have some magical thinking you’d like to share?”
Mar 10, 2020 9 tweets 6 min read
There’s a saying about the numbers/data in epidemiology being the tears of loved one’s or something similar.
Anyone know it and who said it?
#epitwitter /1 Think I found it - "Health statistics represent people with the tears wiped off.” Austin Bradford Hill (of course!).

Remember this here.

I followed the @JohnHopkinsUn CSSE #COVID19 tracker - “A DAY IN THE LIFE OF COVID19”

9:42am (GMT): ImageImage
Mar 6, 2020 5 tweets 5 min read
Did I just hear at the @WHO briefing that they are awaiting studies on traditional chinese medicine for #covid19UK
@DrTedros
??

pscp.tv/w/1OdKrqQoBbPxX @WHO @DrTedros Also heard

180 publications (trials) to date
50 in Chinese language (being translated)

@TrialsTracker - are these being registered?

And wouldn’t this be the ideal time to run a prospective systematicc review/meta-analysis?
@WHO
#covid19
Feb 13, 2020 6 tweets 3 min read
Ok. Just spotted this in NEJM.

nejm.org/doi/full/10.10…

Any #EBMtwitter #epitwitter #statstwitter convos on it yet?

If not let’s start it here #RCT_RWE #RCT_RWE

https://t.co/wSKQTBH66A
Jan 20, 2020 10 tweets 9 min read
@carlheneghan mindfully preparing for his talk on why we need a UK Sunshine Act to help solve the problem of conflicts of interest in medicine @OxEBHC @CebmOxford @EBM_Live @OxPrimaryCare

#COI #EBM Image Starts with the story of Rofecoxib (Vioxx).
Presents the evidence that didn’t get into the NEJM because one of the study authors was a paid consultant on the drug company producing Vioxx.

#COI #EBM #EBMmanifesto Image
Oct 28, 2019 4 tweets 5 min read
Dear @BiochemSoc @PPPublishing

I have spent over 3 hrs posting comments on @PubPeer for 30 Pubmed records each citing a paper published in your journal that was found have duplicate data in a later paper by the same author group.

See example pubpeer.com/publications/0…

/1 @BiochemSoc @PPPublishing @PubPeer The paper w duplicate data is this one: ncbi.nlm.nih.gov/pubmed/17688420 published in your journal Clinical Science

Importantly, most, if not all of the articles citing this paper do so without acknowledging the known concerns with the data contained within.

/2