,
23 tweets,
5 min read
Read on Twitter
Very brief data interpretation tweetorial. <5mins.
The FFR threshold was discovered like this, in the landmark NEJM 1996 paper.
The FFR threshold was discovered like this, in the landmark NEJM 1996 paper.
The FFR threshold was drawn so that almost every patient with a positive ischaemia test has a positive FFR, and as few others as possible.
In this dataset, did it generally succeed in this aim?
In this dataset, did it generally succeed in this aim?
I think it was generally successful in this dataset, somehow.
Indeed there is even a citation of a slew of papers to add an unimpeachable air of convincingness. Who dares to argue with the Reverend Bayes?
How many papers are cited to bolster the solidity of this reasoning?
Indeed there is even a citation of a slew of papers to add an unimpeachable air of convincingness. Who dares to argue with the Reverend Bayes?
How many papers are cited to bolster the solidity of this reasoning?
How many papers?
My good friend and fellow London-based ironist Narbeh Melikian wrote a cheeky paper:
Careful design:
Blinded analysis of perfusion scans. Kudos!
Blinded analysis of perfusion scans. Kudos!
Here is the results analysed as in NEJM 1996.
Each patient is categorised as FFR + or -, and Perfusion scan + or -.
Each patient is categorised as FFR + or -, and Perfusion scan + or -.
In patients with a positive FFR (below 0.80), what PROPORTION of myocardial perfusion imaging was positive?
In patients with a negative FFR (i.e. >0.80), what proportion of myocardial perfusion imaging was positive?
How closely do these numbers match the findings of NEJM 1996?
I think this suggests the NEJM 1996 results were an accident, or similarly unusual circumstance.
In ORBITA, 94-96% of patients had 1 or more positive ischaemia tests.
What proportion had a positive FFR?
In ORBITA, 94-96% of patients had 1 or more positive ischaemia tests.
What proportion had a positive FFR?
Nearly done!
What would you have concluded from Narbeh's data (Myocardial perfusion imaging versus FFR), once you found that only 75% of FFR positive people had positive perfusion scan, and 50% of the FFR negative people had positive perfusion scans?
What would you have concluded from Narbeh's data (Myocardial perfusion imaging versus FFR), once you found that only 75% of FFR positive people had positive perfusion scan, and 50% of the FFR negative people had positive perfusion scans?
Well, here is the conclusion of the paper.
Is this the conclusion you would draw?
Stop wasting your life on perfusion scans, they are shit.
FFR is brilliant.
Just do FFR.
Stop wasting your life on perfusion scans, they are shit.
FFR is brilliant.
Just do FFR.
I wish I could simply say that Narbeh Melikian is an idiot.
Unfortunately I happen to have known him for many years as we trained together. He was intelligent and logical and there is no reason to think he has blown a fuse upstairs in recent years.
Unfortunately I happen to have known him for many years as we trained together. He was intelligent and logical and there is no reason to think he has blown a fuse upstairs in recent years.
So why does the paper conclude something different from what a calm twitteratus such as yourself would conclude?
Clue:
Clue:
Tweetorial over! Comments welcome!
Money back if it took you >5min.
[Although I might write-in a 4 minute value if I don't like your timing measurement]
Money back if it took you >5min.
[Although I might write-in a 4 minute value if I don't like your timing measurement]
Points to consider.
Which study has more patients?
Which study describes blinding?
Which study has prespecified analysis methods?
So if they conflict, which study, by essentially the same authors, do you believe to be more valid?
Which study has more patients?
Which study describes blinding?
Which study has prespecified analysis methods?
So if they conflict, which study, by essentially the same authors, do you believe to be more valid?
Embarrassing erratum.
So I was wrong to say that the rates of perfusion scanning positivity are
50% and 75% in FFR negative and FFR positive patients.
It is 50% and ~66%.
The paper's conclusion is all the more hilarious, in light of this.
50% and 75% in FFR negative and FFR positive patients.
It is 50% and ~66%.
The paper's conclusion is all the more hilarious, in light of this.
Richard @richardbogle sent me Narbeh's cell phone number so I have texted him.
Poor guy was hopeful we cited him in an actual paper.
He's not on Twitter. So I sent him link. Ominously no reply.
Maybe I need to steer clear of him for a few more years ... 😄
Poor guy was hopeful we cited him in an actual paper.
He's not on Twitter. So I sent him link. Ominously no reply.
Maybe I need to steer clear of him for a few more years ... 😄
