Thread coming on one way docs might add value to the #COVID19 crisis: critical appraisal.
I wrote a critique of a flawed study published in a major journal. The study reported on the assoc of anticoagulant use and death in COVID pts.
medscape.com/viewarticle/93…
I wrote a critique of a flawed study published in a major journal. The study reported on the assoc of anticoagulant use and death in COVID pts.
medscape.com/viewarticle/93…
It is an important topic b/c numerous studies have suggested a high rate of clot-related complications in acute #COVID19 illness. it makes sense to consider anti-coagulants...
But anticoagulants come with harms--bleeding. What's more, in non-COVID pts we know that there are actual clots that we are treating. Infection control issues sometimes prevents knowing whether a person w/ #COVID19 actually has clots.
The key findings from the observational study are in the screen shot. Patients given AC seemed to do better.
Here is the link to the research letter in JACC onlinejacc.org/content/early/…
Here is the link to the research letter in JACC onlinejacc.org/content/early/…
The authors added some qualifying words to their conclusions but their take was clear: see screenshot
Now it gets tricky. The second author on the paper is the highly-cited and highly-regarded Dr. Valentin Fuster, who is also editor-in-chief of the journal publishing the paper.
Dr Fuster is convinced AC is beneficial.
Dr Fuster is convinced AC is beneficial.
This had a major effect. The paper has an Altmetric score of 1247 -- which is huge. But the paper has MAJOR flaws --many of which were brought out on Twitter --within hours of publication
The major flaw is immortal time bias -- in which pts who get AC have to be alive to get it, and those who do not can die at any time during the study. @rwyeh has an amazing explanation using Cheetos here.
See also MGH team us19.campaign-archive.com/?u=ef98149bee3…
See also MGH team us19.campaign-archive.com/?u=ef98149bee3…
Two problems with making over-reaching conclusions from hugely biased observational studies:
1) You could be wrong --harm both pts and trust in science
2) You destroy equipoise --> makes it hard for thinking docs to individualize care and to randomize pts in proper trials.
1) You could be wrong --harm both pts and trust in science
2) You destroy equipoise --> makes it hard for thinking docs to individualize care and to randomize pts in proper trials.
Some argue that in the absence of certainty, we should use the data we have. But this paper and the press attention it received and its effects on AC protocols actually suggests that no data might be better.
This study also has me concerned about traditional peer review. Like @Richard56 once opined, Twitter can do better blogs.bmj.com/bmj/2011/01/26…
The problem w/ post-publication review is that it's hard to put the genie back in the bottle. Exhibit A -hydroxychloroquine
The problem w/ post-publication review is that it's hard to put the genie back in the bottle. Exhibit A -hydroxychloroquine
Second to last: many people helped me on this column. Some wanted to stay anonymous. It's scary being critical of a leader and a big journal. I'm not an academic and it makes me nervous.
None of this is personal. The critique is of ideas not people.
None of this is personal. The critique is of ideas not people.
Last: When the history of #COVID19 is written, I hope we re-learn the benefits of slow science and proper trials. You know, like we did about treating PVCs after heart attacks.
