Just published @TheLancet@twitter.com
The largest study of hydroxychloroquine shows a significant increase in death (~35%) and >2-fold increase of serious heart arrhythmias. ~96,000 patients, ~15,000 on HCQ or CQ from 671 hospitals, 6 continents.
This is a retrospective study, and the big question is: why were those people prescribed those drugs in the first place, and people of control group weren’t? This wasn’t a randomized study where you’d take people and randomly put one third of them on a drug, and the other two thirds on another one. No, in this study they took people that were prescribed the drugs and people that weren’t, and compared, without first investigating how different those two groups of people were.
My first guess would be that people who did get those drugs were much more severely ill, and doctors took chances with these drugs to save their lives. And there’s more mortality in that group not because of the drugs, but because those patients were more severe in the first place.
This is exactly why they teach basic statistics in medical schools, and yet you people fall for such papers.
@evgenykuznetsov.org Have you read the paper before calling me "you people"? I think they controlled it quite well and it makes sense in context from what I've heard from people who specialise in heart diseases in respect to both COVID-19 and chloroquine (which is well studied as both a malaria drug and an immuno suppressant).
Have you read the paper before calling me “you people”?
I have. And I’m really sorry it came out this way, I wasn’t attacking you personally; my note was a reply to original tweet and your repost, I didn’t mean to single you out.
Maybe it also came out a little too harsh. You see, I really get frustrated with misrepresentation of results published in medical papers (my occupational hazard, probably), and with current COVID-19 situation I get a lot of reasons for frustration.
They go into why some people were prescribed those drugs and others weren’t.
They actually don’t, at least not to the point of properly showing there was no bias in prescriptions. And lacking the randomized trial design they can’t realistically have done that, either. If you read the discussion part of the paper closely, they themselves point out that the observational design of the study has these fallacies. Here’s a direct quote that failed to get into any reposts of this paper I’ve seen: “Randomised clinical trials will be required before any conclusion can be reached regarding benefit or harm of these agents in COVID-19 patients” (the emphasis is mine).
As for your guess that the people on the drugs were more severely ill, I doubt that.
Yet that’s exactly what Baseline/SPO2 line in their Table 2 seems to imply.
Also: It would be extremely unethical to “take more chances” just because someone looks sicker to you.
Oh, quite the contrary. Where no clear and binding protocols are established (like a new infection no one really knows how to treat yet), every prescription is basically an attempt to balance estimated benefits and estimated risks. For every patient you have the chance the patient gets well without your prescription (A), the risk the patient dies without it (B), the chance the patient gets better because of your prescription (C) and the risk the patient gets worse because of it (D), and your job as a doctor is to figure out whether the drug is worth prescribing, while the classic “do no harm” maxim tells you which side is better to err on. In a patient more severely ill, A gets lower and B gets higher, obviously affecting the balance of the mental equasion you’re solving; thus, more severly ill patients are more likely to be prescribed the drug the doctor is less sure of.
This is a known issue with observational studies, and this is the exact reason why randomized studies were invented.
You have to start with the drugs as soon as possible (goes for antivirals like remdesivir too) so you have no time to see how someone develops.
True. Yet you still assess your patient and judge your prescription (see above). You still get biased. We still need a randomized study.
Again, I’m really sorry if my previous remark came out as a personal attack. I didn’t mean to.
@evgenykuznetsov.org Nah, it’s all good. Thank you for clarifying. You are probably right and obviously more knowledgable than me. 😅
It do still think they did the best they could in this situation. Randomised trials are obviously not something you’d be doing at that point in the general panic. But I do agree it’s needed. I will link to your explanation from my blog.
Per procrastinatum ad astra