I take every study with a massive grain of salt, [especially one this size], until solid meta-analysis is done, but this is more promising than I expected for hydroxychloroquine without zinc:
>Conclusion: in a cohort of 166 patients from 18 to 85 years hospitalised with COVID-19, hydroxychloroquine treatment with an 800mg added loading dose increased the cumulative mean survival according to the diagnosis of severity at the beginning by 1.4 to 1.8 times. This difference was statistically significant when the clinical picture at admission was mild.
In an earlier era, issues with the Santa Clara study would have been addressed during peer review—the process in which scientific work is assessed by other researchers before being published in a journal. But like many COVID-19 studies, this one was uploaded as a preprint—a paper that hasn’t yet run the peer-review gauntlet. Preprints allow scientists to share data quickly, and speed is vital in a pandemic: Several important studies were uploaded and discussed a full month before being published.
Preprints also allow questionable work to directly enter public discourse, but that problem is not unique to them. The first flawed paper on hydroxychloroquine and COVID-19 was published in a peer-reviewed journal, whose editor in chief is one of the study’s co-authors.[1]
That doesn't mean one can't read the preprint and critically evaluate it (I've done so for a few posted here on HN).
On this specific paper, I think tracking survival is not the right metric. Studies which evaluate the time to recovery, like the NIAID trial with remedisivir, are far more important in the current situation.
Sure. You can critically evaluate random Internet posts. They are still problematic when they enter public non-specialist discourse without peer review by those that possess the experience and context to properly evaluate them.
Agreed. It's disappointing to see observational study after observational study pumped out and still little-to-no actual causal evidence after three months.
The U of Minnesota trial (which is a true vs-placebo trial) reported on April 22nd that they needed at least 200 more people to get enough power for their study. According to their web page[1] the interim meeting of the safety board is expected for today.
> In the study of 368 patients, 97 patients who took hydroxychloroquine had a 27.8% death rate. The 158 patients who did not take the drug had an 11.4% death rate.
Personally, when I find two studies report opposite conclusions, I take a neutral position. Neutral meaning: lets give drugs to people who need them instead of using a lupus drug to treat flu symptoms.
“ Age distribution according to the severity at admission was homogeneous between HCQ and non- HCQ treatment groups when the clinical picture was mild (57,6 years HCQ – 58,4 years non-HCQ, p=0,865) or moderate (63,8 years HCQ – 70 years non-HCQ, p=0,269). ”
Those are some high p-values. Makes sense with a small control group of 43 people.
This paper can be put in the bin labeled “no statically significance”
They're meant to be high. This means "we didn't find a detectable difference in age distribution between the groups" and therefore that the results are (hopefully) not confounded by age.
This comment brings nothing to discussion, plus making any kind of definitive statement based on 'none statistically significant study' is plain rubbish.
Even if Trump finally turns out to be right, he announced his claims before seeing this (or any other) study. If you claim a thousand things eventually one or two will be true.
Trump often makes claims about things he wants to draw attention to. Given that HCQ is such a cheap, off patent drug, it stands to reason that he wanted to get it into the public consciousness so it would be looked at with more scrutiny.
Without him mentioning it, there isn't much incentive to study it.
>Conclusion: in a cohort of 166 patients from 18 to 85 years hospitalised with COVID-19, hydroxychloroquine treatment with an 800mg added loading dose increased the cumulative mean survival according to the diagnosis of severity at the beginning by 1.4 to 1.8 times. This difference was statistically significant when the clinical picture at admission was mild.