Read on Twitter
@thelancet have published another paper to try to show that #hydroxychloroquine is "dangerous" (it's not).
Here's why I think it's fake https://www.thelancet.com/journals/lanrhe/article/PIIS2665-9913(20)30276-9/fulltext#articleInformation
Here's why I think it's fake https://www.thelancet.com/journals/lanrhe/article/PIIS2665-9913(20)30276-9/fulltext#articleInformation
First off, the authorship is a bit unusual for a clinical (medical) paper.
Almost all biostatisticians and bioinformatics people. These people understand numbers, modelling and code but have a limited grasp of medicine.
They are working a bioinformatics project...
Almost all biostatisticians and bioinformatics people. These people understand numbers, modelling and code but have a limited grasp of medicine.
They are working a bioinformatics project...
The project is the OHDSI-COVID-19 consortium.
It's on github if you want to take a look.
Not exactly the usual clinical repository of data. In fact, the data for this paper is not supplied - despite @thelancet being under intense scrutiny https://github.com/ohdsi-studies
It's on github if you want to take a look.
Not exactly the usual clinical repository of data. In fact, the data for this paper is not supplied - despite @thelancet being under intense scrutiny https://github.com/ohdsi-studies
So, let's take a dive. First up, the "covid 19 prediction model". Hmmmm... remember that #surgisphere had exactly the same model.
https://github.com/ohdsi-studies/Covid19PredictionStudies
https://github.com/ohdsi-studies/Covid19PredictionStudies
Oh hang on, what's the "study-a-thon"?
Well it was a virtual conference in March where hundreds of people from @OHDSI got together to exchange ideas and within 4 days had come up with a whole bunch of protocols for studies. Pretty fast eh?
Well it was a virtual conference in March where hundreds of people from @OHDSI got together to exchange ideas and within 4 days had come up with a whole bunch of protocols for studies. Pretty fast eh?
One of the central components of their research was a global database comprising healthcare databases from many different countries all amalgamated into one easy database (yes, we've been here before haven't we?)
[This time they left out Australia]
![One of the central components of their research was a global database comprising healthcare databases from many different countries all amalgamated into one easy database (yes, we've been here before haven't we?)[This time they left out Australia]](https://pbs.twimg.com/media/EgFrMVLUwAExFzZ.png "One of the central components of their research was a global database comprising healthcare databases from many different countries all amalgamated into one easy database (yes, we've been here before haven't we?)[This time they left out Australia]")
[This time they left out Australia]
Now, I'm not saying that this massive database is fake, like the (almost identical) #surgisphere data but there are a couple of red flags.
Take for example the @CPRD database. It's run out the UK and only has GP information. You would need to submit an application to access it..
Take for example the @CPRD database. It's run out the UK and only has GP information. You would need to submit an application to access it..
There were no suitable applications for this project this year. Check yourself
It's possible that they have used another protocol, but this is not usually allowed.
https://cprd.com/protocol-list
It's possible that they have used another protocol, but this is not usually allowed.
https://cprd.com/protocol-list
Another database is @IQVIA_global. Another huge corporation it seems, skimming data from health care providers but with no transparency.
They say they have the data but there is no way to check it.
Almost identical to #surgisphere
https://www.iqvia.com/about-us/iqvia-core
They say they have the data but there is no way to check it.
Almost identical to #surgisphere
https://www.iqvia.com/about-us/iqvia-core
Back to the authors. Who is behind them, and the @OHDSI?
Well, #bigpharma of course.
@JanssenUS @sanofi @Amgen @Merck as well as some interesting university collaborators...
https://www.ohdsi.org/who-we-are/collaborators/
Well, #bigpharma of course.
@JanssenUS @sanofi @Amgen @Merck as well as some interesting university collaborators...
https://www.ohdsi.org/who-we-are/collaborators/
Including the Gates foundation, Vanderbilt University and the David Geffen School of Medicine (yes there is one)
Janssen are running the show
Janssen are running the show
The conflict of interests page reads like a who's who of #BigPharma and it seems that most of the collaborators are actually Pharma scientists..
The lead author, Jenni Lane, is a trainee in Orthopaedics. Unusual for such a huge project (analysing millions of data records, apparently). She has worked earlier in the year on a similar project with the same group of pharma statisticians https://pubmed.ncbi.nlm.nih.gov/32511443/
In fact, the research coming out of this huge global project all seem to be centred around a small number of individuals. This paper is a bit of an oddity.
http://www.encepp.eu/encepp/links.htm?id=26519&resourceType=ResearchCentre
http://www.encepp.eu/encepp/links.htm?id=26519&resourceType=ResearchCentre
Massoud Toussi seems to be heavily involved and is listed as affiliated with IQVIA (the #surgisphere-like company)
Here is is researchgate profile and the ENEPP studies list (the Lancet study at the top)
Here is is researchgate profile and the ENEPP studies list (the Lancet study at the top)
Interestingly this paper was uploaded to @medrxiv on May 31st, which means that the project was seeded and huge data repository crunched in 8 weeks.
You can't access the data (even though the Lancet version suggests you can)
So, that's odd....
You can't access the data (even though the Lancet version suggests you can)
So, that's odd....
The links don't take us to a data repository. They take us to a repository for R code.
Not that many doctors speak R. In fact, most doctors don't know what it is.
Here is the repository https://github.com/ohdsi-studies/Covid19EstimationHydroxychloroquine
Not that many doctors speak R. In fact, most doctors don't know what it is.
Here is the repository https://github.com/ohdsi-studies/Covid19EstimationHydroxychloroquine
Now, don't get me wrong. It's possible that Jenny is one of the rare doctor that speaks R. But her OHDSI profile is pretty bare...
https://forums.ohdsi.org/u/jenniferlane
https://forums.ohdsi.org/u/jenniferlane
So I'd hazard a guess that James Weaver supplied all the R code.
He works for Janssen and submitted Megabytes of R code 2 weeks after the project was thought up at the "study-a-thon"
He works for Janssen and submitted Megabytes of R code 2 weeks after the project was thought up at the "study-a-thon"
Pretty good work. Anyway I've had a good look through the R code (yes, I speak R) and what I can't find is the pipe to the data repository.
There are some weird things in there though...
Like "Synthesized positive controls"
Gotta say, I haven't come across those before
There are some weird things in there though...
Like "Synthesized positive controls"
Gotta say, I haven't come across those before
Of course one downside of having bioinformatics people running your clinical study is not knowing clinical information.
Like that apixaban isn't an antiarrythmic agent (there are others in here too, like dabigatran)
Like that apixaban isn't an antiarrythmic agent (there are others in here too, like dabigatran)
....to be continued
