Cheap steroid may lower COVID-19 death rate—but experts urge caution

Researchers have yet to release data on the trial, but many are optimistic.

Read the whole story

 

[Deleted member 817175]

First we switched from peer-reviewed papers to pre-peer-review ones.

And now we're all the way down the chain to press releases based on unpublished data.

Progress!

It's just lovely to see the highest number of up-votes going to a sarcastic comment that implies this is some quack run trial trying to get it's name in the headlines.

Get your napkin out and do a quick calculation of the potential lives lost for each day spent going through the "correct" process.

Right, because life-and-death decisions should always be made on the basis of a press release.

There is a "correct process" for a reason.

 

[lucyfersam]

First we switched from peer-reviewed papers to pre-peer-review ones.

And now we're all the way down the chain to press releases based on unpublished data.

Progress!

It's just lovely to see the highest number of up-votes going to a sarcastic comment that implies this is some quack run trial trying to get it's name in the headlines.

Get your napkin out and do a quick calculation of the potential lives lost for each day spent going through the "correct" process.

Then release the data with the press release so it can be verified. A press release is not enough to go by to change medical best practices.

 

[Bolognesus]

Going back to the P-values, the 0.0003 means there's only a 0.03% chance that the reduced deaths on ventilated patients is a statistical fluke. That's really good. 0.14 on patients without respiratory support means there was a 14% chance that the increased deaths were just a fluke. That's big enough of a chance of not panning out that it wouldn't be appropriate to claim that the study showed an increase.

Edit: You do have to be a bit careful with P-values for a couple of reasons, that are best explained by someone much wittier than I am.

P-values don't include priors. You need Bayesian interference for that. https://xkcd.com/1132/ That's not really a problem here since Covid is novel and we don't really have prior data to contradict, plus there's a reasonable explanation why this drug works as it does.

If you try enough times, eventually you'll get a result with a low p-value. This is called p-hacking, and is covered by https://xkcd.com/882/ This isn't likely to be a problem here: even though this study is testing several different treatments, p=0.0003 is REALLY unlikely to be random.

Well, the p=0.0003 doesn't mean shit unless the methodology by which it was determined can be externally validated. Or would at least be available for cursory inspection. You can get really, really impressive P-values if you just faff about enough with either your methods, your data, or both.

IMO until they at least publish a decent pre-print which confirms good (enough) methodology,* no P value in the world means all that much. I was just responding to the question regarding relative mortality being higher in the moderately ill cohort.

Let's face it, conducting scientific research by press release does *not* engender confidence in the quality of the underlying data set, not with the eleventy-act shitshow we've seen on that front so far.

I agree P values in themselves do not consider prior plausibility. However, viewing P-values in context very much *should* consider prior plausibility; it's how you can generally weed out the one fluke positive with no biologically plausible backing (as you mentioned) or, more insidiously, the systemically wrong result that's based on errors in methodology, or, as so often seen in quackery, outright fraud or self-delusion.

The 'reasonable explanation why this drug works' is very much the prior plausibility / mechanism of action I was on about. Medicine and pharmacology inherently deal with data sets that are just too dirty not to apply such a filter.

I'll certainly agree I wasn't clear enough in my original post!

*it's a reputable group, I'd be willing to take their word for the validity of their dataset

((PS. also, re: being careful with P-values: if anything looking at the actual p=0.05 CIs on this press release presents an even rosier picture of the actual efficacy IMO, so no disagreement there either!)

 

[Happy Medium]

"...it is unacceptable to tout study results by press release without releasing the paper."

There needs to be more discipline. The masses do not understand research and the false alarms have caused harm to the effort to control the pandemic and the reputation of medical science.

Offset this against the lives that could be saved or lost by expediting or delaying release of the initial results. As someone who has been through this cycle published in peer reviewed journals this can often take months to get a formal write-up to publication standard and 6 months from acceptance into print.

The institution running the study is Oxford University and I work at one of the study sites. It's a big study that as far as I can see is being run properly.

Yeah that's not how it works. Just because the authors believe their results show efficacy doesn't give them the right to make blanket statements without publishing associated data to back it up! Because THEY MIGHT BE WRONG. You publish information so that other experts can critique your methods and results and judge for themselves whether they find the data compelling or not. Rarely are results so cut and dried that a single study definitively proves the efficacy of a treatment. Even with the best run studies there are almost always methodologic concerns that need discussion and analysis.

Just look at the literature debate on steroids in sepsis. There have been TONS of actually published papers and yet still it's a big question. So no, the authors need to put up or shut up, because their 2 page presser which completely lacks very significant details about study methodology doesn't give anybody the ability to trust them in any way.

 

[Borderliner]

First we switched from peer-reviewed papers to pre-peer-review ones.

And now we're all the way down the chain to press releases based on unpublished data.

Progress!

It's just lovely to see the highest number of up-votes going to a sarcastic comment that implies this is some quack run trial trying to get it's name in the headlines.

Get your napkin out and do a quick calculation of the potential lives lost for each day spent going through the "correct" process.

Right, because life-and-death decisions should always be made on the basis of a press release.

There is a "correct process" for a reason.

Nicely sanctimonious

"Given the public health importance of these results, we are now working to publish the full details as soon as possible."

I wonder if mechanisms for handling a potentially significant positive outcome mid trial were considered.

 

[Deleted member 817175]

First we switched from peer-reviewed papers to pre-peer-review ones.

And now we're all the way down the chain to press releases based on unpublished data.

Progress!

It's just lovely to see the highest number of up-votes going to a sarcastic comment that implies this is some quack run trial trying to get it's name in the headlines.

Get your napkin out and do a quick calculation of the potential lives lost for each day spent going through the "correct" process.

Right, because life-and-death decisions should always be made on the basis of a press release.

There is a "correct process" for a reason.

Nicely sanctimonious

"Given the public health importance of these results, we are now working to publish the full details as soon as possible."

I wonder if mechanisms for handling a potentially significant positive outcome mid trial were considered.

"It's okay, it's Oxford so they don't need to show their work" strikes me as a piss-poor method of evaluating scientific research.

 

[Nature Lover]

I think it is useful to keep in mind that this is not an antiviral treatment - if anything, it's the contrary, and may well lead to decreased / slower clearing of the virus.

So it's definitely not to be used as prophylactic or even milder to moderate disease - it's likely going to makes things worse.

It does appear to be useful in the way high dose steroids have always been useful - tamping down on a dangerous inflammatory response such as a Cytokine storm often observed in severe Covid-19. Indeed, in a substantial number of cases, apparently life-saving - first drug to reach that distinction for this disease (in contrast to Remdesivir, for instance, and much, much cheaper than the specific Il-6 inhibitors that also appear useful in the Cytokine storm).

As I pointed out in that recent FDA and hydroxychloroquine thread there is evidence that Chronic prednisone use, another shorter acting glucocorticoid (the more specific name as estrogens and testosterone are steroids) appears to increase the hospitalization rate in rheum patients, maybe by causing issues at the viral replication phase, not the primarily inflammatory phase. Timing is everything.

ed corticosteroids also appropriate. The commenter who noted that Trump will take for muscles either is misinformed or stretching the classification. Steroids include progestins, estrogens, anabolic steroids (male muscle builders) and glucocorticoids/corticosteroids.

 

[Borderliner]

First we switched from peer-reviewed papers to pre-peer-review ones.

And now we're all the way down the chain to press releases based on unpublished data.

Progress!

It's just lovely to see the highest number of up-votes going to a sarcastic comment that implies this is some quack run trial trying to get it's name in the headlines.

Get your napkin out and do a quick calculation of the potential lives lost for each day spent going through the "correct" process.

Right, because life-and-death decisions should always be made on the basis of a press release.

There is a "correct process" for a reason.

Nicely sanctimonious

"Given the public health importance of these results, we are now working to publish the full details as soon as possible."

I wonder if mechanisms for handling a potentially significant positive outcome mid trial were considered.

"It's okay, it's Oxford so they don't need to show their work" strikes me as a piss-poor method of evaluating scientific research.

How about you read the release...

 

[ColdWetDog]

The fascinating Internet Book of Critical Care has been suggesting dexamethasone as a treatment since mid-March based on extrapolation of early data from China. Not being a medic I'm not sure how constrained ICU staff are in what medication they try when all else fails with a novel disease, I'm curious to know whether this has already been in wide use in the same sort of Hail Mary fashion as chloroquine?

Here is the problem. Dexamethasone (and other corticosteroids) have been used since the beginning. By some people. At different times. With different drugs. And different doses. Some folks though they worked, others weren't sure. So doing a controlled study was critically important.

But...

Patient selection, timing and dosing (which I guess we know) is absolutely critical. The idea behind the immunosuppressants of whatever flavor is that you mitigate the post infectious damage done by the immune system. Which isn't a problem for every patient. But is a significant issue for patients who end up in hospital and is *probably* universal for anyone on a ventilator.

So details matter. They really matter. It is completely unprofessional and really unethical to announce the findings in this way, especially with the rah-rah statements by the authors.

Very, very poor form Oxford.

 

[Arkleseizure]

This just in: dexamethasone now $8000 a tablet.

"It's a fair price", say industry experts.

 

[andygates]

"...it is unacceptable to tout study results by press release without releasing the paper."

There needs to be more discipline. The masses do not understand research and the false alarms have caused harm to the effort to control the pandemic and the reputation of medical science.

Offset this against the lives that could be saved or lost by expediting or delaying release of the initial results. As someone who has been through this cycle published in peer reviewed journals this can often take months to get a formal write-up to publication standard and 6 months from acceptance into print.

The institution running the study is Oxford University and I work at one of the study sites. It's a big study that as far as I can see is being run properly.

We run a local chunk of it down here too; our whole research department has been diverted to covid work since March. Arsians commenters have been burned by bad preprints.

 

[Nature Lover]

The fascinating Internet Book of Critical Care has been suggesting dexamethasone as a treatment since mid-March based on extrapolation of early data from China. Not being a medic I'm not sure how constrained ICU staff are in what medication they try when all else fails with a novel disease, I'm curious to know whether this has already been in wide use in the same sort of Hail Mary fashion as chloroquine?

Here is the problem. Dexamethasone (and other corticosteroids) have been used since the beginning. By some people. At different times. With different drugs. And different doses. Some folks though they worked, others weren't sure. So doing a controlled study was critically important.

But...

Patient selection, timing and dosing (which I guess we know) is absolutely critical. The idea behind the immunosuppressants of whatever flavor is that you mitigate the post infectious damage done by the immune system. Which isn't a problem for every patient. But is a significant issue for patients who end up in hospital and is *probably* universal for anyone on a ventilator.

So details matter. They really matter. It is completely unprofessional and really unethical to announce the findings in this way, especially with the rah-rah statements by the authors.

Very, very poor form Oxford.

I agree. see my comment above.

 

[jdale]

"...it is unacceptable to tout study results by press release without releasing the paper."

There needs to be more discipline. The masses do not understand research and the false alarms have caused harm to the effort to control the pandemic and the reputation of medical science.

Offset this against the lives that could be saved or lost by expediting or delaying release of the initial results. As someone who has been through this cycle published in peer reviewed journals this can often take months to get a formal write-up to publication standard and 6 months from acceptance into print.

The institution running the study is Oxford University and I work at one of the study sites. It's a big study that as far as I can see is being run properly.

Getting a paper peer reviewed and into print takes a while. But there's no reason they couldn't have released a pre-print of the paper. That has become very common for covid-19 related studies. All the data is present, the methods are present, peer review is not finished but it permits the scientific community to start discussing the paper immediately.

That might delay their ability to release a press release, but at this stage we've been burned enough times that no one should put any faith in a press release. You certainly should not change your treatment plans based on a press release alone. That would be highly irresponsible.

 

[tucu]

The fascinating Internet Book of Critical Care has been suggesting dexamethasone as a treatment since mid-March based on extrapolation of early data from China. Not being a medic I'm not sure how constrained ICU staff are in what medication they try when all else fails with a novel disease, I'm curious to know whether this has already been in wide use in the same sort of Hail Mary fashion as chloroquine?

Here is the problem. Dexamethasone (and other corticosteroids) have been used since the beginning. By some people. At different times. With different drugs. And different doses. Some folks though they worked, others weren't sure. So doing a controlled study was critically important.

But...

Patient selection, timing and dosing (which I guess we know) is absolutely critical. The idea behind the immunosuppressants of whatever flavor is that you mitigate the post infectious damage done by the immune system. Which isn't a problem for every patient. But is a significant issue for patients who end up in hospital and is *probably* universal for anyone on a ventilator.

So details matter. They really matter. It is completely unprofessional and really unethical to announce the findings in this way, especially with the rah-rah statements by the authors.

Very, very poor form Oxford.

To be fair, this was not just the decision of a couple of Oxford researchers. This press release was a decision of the Steering Committee of the trial that includes people from multiple universities, the NHS and the UK govt.

 

[Borderliner]

The fascinating Internet Book of Critical Care has been suggesting dexamethasone as a treatment since mid-March based on extrapolation of early data from China. Not being a medic I'm not sure how constrained ICU staff are in what medication they try when all else fails with a novel disease, I'm curious to know whether this has already been in wide use in the same sort of Hail Mary fashion as chloroquine?

Here is the problem. Dexamethasone (and other corticosteroids) have been used since the beginning. By some people. At different times. With different drugs. And different doses. Some folks though they worked, others weren't sure. So doing a controlled study was critically important.

But...

Patient selection, timing and dosing (which I guess we know) is absolutely critical. The idea behind the immunosuppressants of whatever flavor is that you mitigate the post infectious damage done by the immune system. Which isn't a problem for every patient. But is a significant issue for patients who end up in hospital and is *probably* universal for anyone on a ventilator.

So details matter. They really matter. It is completely unprofessional and really unethical to announce the findings in this way, especially with the rah-rah statements by the authors.

Very, very poor form Oxford.

That's a very high horse you've got yourself on there.

It sure is a pity the Oxford team didn't publish anything about the the study protocol.

 

[ColdWetDog]

The fascinating Internet Book of Critical Care has been suggesting dexamethasone as a treatment since mid-March based on extrapolation of early data from China. Not being a medic I'm not sure how constrained ICU staff are in what medication they try when all else fails with a novel disease, I'm curious to know whether this has already been in wide use in the same sort of Hail Mary fashion as chloroquine?

Here is the problem. Dexamethasone (and other corticosteroids) have been used since the beginning. By some people. At different times. With different drugs. And different doses. Some folks though they worked, others weren't sure. So doing a controlled study was critically important.

But...

Patient selection, timing and dosing (which I guess we know) is absolutely critical. The idea behind the immunosuppressants of whatever flavor is that you mitigate the post infectious damage done by the immune system. Which isn't a problem for every patient. But is a significant issue for patients who end up in hospital and is *probably* universal for anyone on a ventilator.

So details matter. They really matter. It is completely unprofessional and really unethical to announce the findings in this way, especially with the rah-rah statements by the authors.

Very, very poor form Oxford.

To be fair, this was not just the decision of a couple of Oxford researchers. This press release was a decision of the Steering Committee of the trial that includes people from multiple universities, the NHS and the UK govt.

It usually takes either tequila, a computer or a committee to really fuck things up.

 

[jdale]

How does this compare with remdesivir? It's written as if this is the only effective treatment found, but I thought that was shown in also speeding recovery?

The data for remdesivir showed that it improved the recovery time (by four days) but the study was stopped early before it was possible to show whether it improved mortality.

I think it's very likely that it's beneficial to use both. Start using remdesivir earlier, and then add dexamethasone if the patient's condition worsens to the point where they need ventilation or oxygen.

There are probably going to be other treatments that also help and it's quite possible the optimal treatment will be some kind of cocktail. Clinical studies are still ongoing and we probably will also have shortages of drugs that force us to evaluate more alternatives (including remdesivir which is not in adequate supply).

Of course every treatment that proves out further demonstrates the importance of flattening the curve. Even if you were going to catch covid-19 either way, you're better off if you managed to delay until these treatments were known. And you're also better off if you continue to delay until there is adequate supply of these drugs.

 

[rmgoat]

If this is true (and I think it probably is), it's likely that timing when to use the drug is important because it likely mitigates one reaction that some people have to the virus: an overactive immune response. That reaction can be fatal.

A lot of people, most people, don't have that reaction. Giving them an immunosuppressant is, like the original guidance, a distinctly bad idea.

If idiots (Trumpers) take this prophylactically, it could make things substantially worse.

Or, I mean, depending on how you look at it, could make things substantially better. . .

I'll say one thing about Trump, I've never seen a politician try so hard to get their own voter base killed off right before an election.

According to the demographic data it's probably killing non-whites faster than whites, and urban folk faster than rural ones.

I don't think that's why Trump decided to throw caution to the wind here, but I'm sure that's occurred to a few of his supporters.

A few rallies will fix that. Lots of old white rural folks will rush to Tulsa to bask in his glory (and he bask in their adoration, which is what it's really about). According to some press coverage cases are now hitting more rural states where they all thought this was 'fake news'. Maybe when living on a farm where the nearest town is under 800 people and twenty miles away (like friends in WI) does not save you anymore, they will take it seriously.

 

[tucu]

The fascinating Internet Book of Critical Care has been suggesting dexamethasone as a treatment since mid-March based on extrapolation of early data from China. Not being a medic I'm not sure how constrained ICU staff are in what medication they try when all else fails with a novel disease, I'm curious to know whether this has already been in wide use in the same sort of Hail Mary fashion as chloroquine?

Here is the problem. Dexamethasone (and other corticosteroids) have been used since the beginning. By some people. At different times. With different drugs. And different doses. Some folks though they worked, others weren't sure. So doing a controlled study was critically important.

But...

Patient selection, timing and dosing (which I guess we know) is absolutely critical. The idea behind the immunosuppressants of whatever flavor is that you mitigate the post infectious damage done by the immune system. Which isn't a problem for every patient. But is a significant issue for patients who end up in hospital and is *probably* universal for anyone on a ventilator.

So details matter. They really matter. It is completely unprofessional and really unethical to announce the findings in this way, especially with the rah-rah statements by the authors.

Very, very poor form Oxford.

To be fair, this was not just the decision of a couple of Oxford researchers. This press release was a decision of the Steering Committee of the trial that includes people from multiple universities, the NHS and the UK govt.

It usually takes either tequila, a computer or a committee to really fuck things up.

Maybe, but if this is a fuck-up, it is a lot bigger than just Oxford. It involves the NIHR (that supervised the trial), the UK govt (that has now approved the drug for treating covid patients) and the WHO (that has congratulated the team after having access to "initial insights")

https://www.who.int/news-room/detail/16 ... 9-patients

 

[rmgoat]

Looking at how Japan has dealt with its outbreak, essentially using nothing but facemasks, you have to wonder what would have happened if all countries had followed a 100% facemask policy.

A look at the Wikipedia article on the Japanese government's response shows (unsurprisingly) there has been more to it than just facemasks.

Just one example:

On 30 March, (Tokyo Governor) Koike requested residents to refrain from nonessential outings for the next two weeks due to a continued increase in infections in Tokyo.

That's the 30th of March. Speaking to relatives in Japan, in practise on the ground it's been a 50% reduction in commuting in the Tokyo area, lots of pupils still in school (cram schools closed), and everyone wearing facemasks in all public places (not so unusual in Japan anyway). The calls to reduce social interactions were widely ignored, but a lot of restaurants closed (despite not necessarily having to). Back in April it was looking dicey, and the lack of preparedness of things like mass testing, hospital space, etc, was beginning to be of great concern.

However, the outbreak has simply fizzled (by comparison to lots of Western Europe). That same Wikipedia page lists them as having returned to normal more or less by the 21st May, and so far they've had less than 1,000 deaths nationally.

If you've ever been on the Tokyo metro in rush hour, you'll know that a 50% reduction in commuters does not result in a nice, comfy, 2meter margin between yourself and fellow passengers. It simply mean you've got only 1 armpit in front of your nose, not 2. For them to have more or less eliminated their outbreak whilst still packing people on to trains strongly suggests that the face masks people have been wearing were highly efficacious.

Why won't Americans wear masks?

1. They are hot and uncomfortable.
2. Trump refuses to wear one.
3. It's now a sign you are one of those edicated lirbuls.
4. I'm under 30 and invulnerable.
5. Those old people are almost dead anyway.
6. They are not to protect me, they are to protect other people FROM me, like I care that I may be infected but asymptomatic spreader (at a go Typhoid Mary)
7. I have rights and they Trump your rights.
8. I'm a tough independent murican and can do what I want.

Or any combination of the above.

 

[ChrisSD]

The fascinating Internet Book of Critical Care has been suggesting dexamethasone as a treatment since mid-March based on extrapolation of early data from China. Not being a medic I'm not sure how constrained ICU staff are in what medication they try when all else fails with a novel disease, I'm curious to know whether this has already been in wide use in the same sort of Hail Mary fashion as chloroquine?

Here is the problem. Dexamethasone (and other corticosteroids) have been used since the beginning. By some people. At different times. With different drugs. And different doses. Some folks though they worked, others weren't sure. So doing a controlled study was critically important.

But...

Patient selection, timing and dosing (which I guess we know) is absolutely critical. The idea behind the immunosuppressants of whatever flavor is that you mitigate the post infectious damage done by the immune system. Which isn't a problem for every patient. But is a significant issue for patients who end up in hospital and is *probably* universal for anyone on a ventilator.

So details matter. They really matter. It is completely unprofessional and really unethical to announce the findings in this way, especially with the rah-rah statements by the authors.

Very, very poor form Oxford.

To be fair, this was not just the decision of a couple of Oxford researchers. This press release was a decision of the Steering Committee of the trial that includes people from multiple universities, the NHS and the UK govt.

It usually takes either tequila, a computer or a committee to really fuck things up.

Maybe, but if this is a fuck-up, it is a lot bigger than just Oxford. It involves the NIHR (that supervised the trial), the UK govt (that has now approved the drug for treating covid patients) and the WHO (that has congratulated the team after having access to "initial insights")

https://www.who.int/news-room/detail/16 ... 9-patients

The UK government currently has a strong interest in heavily hyping anything that comes from British research. They need a distraction from all the failures in leadership. This is likely why the UK government have been so strongly promoting the results before even pre-prints are available. They're making it into political capital and using patriotism to avoid answering any hard questions.

I should hope the NIHR signed off on it. If they didn't then the trial shouldn't have gone ahead.

A "congrats" from the WHO is nice but pretty meaningless. They haven't seen the full methods or data either.

Don't get me wrong, the study may well lead to better outcomes. But we won't get there by denouncing anyone who dares to ask to see the actual science first. Mistakes happen, things are overlooked and thousand over things could be wrong without any maliciousness or gross incompetence. It's all very well merely trusting the word of respected individuals and institutions but science moved past that a long time ago.

 

[tucu]

The fascinating Internet Book of Critical Care has been suggesting dexamethasone as a treatment since mid-March based on extrapolation of early data from China. Not being a medic I'm not sure how constrained ICU staff are in what medication they try when all else fails with a novel disease, I'm curious to know whether this has already been in wide use in the same sort of Hail Mary fashion as chloroquine?

Here is the problem. Dexamethasone (and other corticosteroids) have been used since the beginning. By some people. At different times. With different drugs. And different doses. Some folks though they worked, others weren't sure. So doing a controlled study was critically important.

But...

Patient selection, timing and dosing (which I guess we know) is absolutely critical. The idea behind the immunosuppressants of whatever flavor is that you mitigate the post infectious damage done by the immune system. Which isn't a problem for every patient. But is a significant issue for patients who end up in hospital and is *probably* universal for anyone on a ventilator.

So details matter. They really matter. It is completely unprofessional and really unethical to announce the findings in this way, especially with the rah-rah statements by the authors.

Very, very poor form Oxford.

To be fair, this was not just the decision of a couple of Oxford researchers. This press release was a decision of the Steering Committee of the trial that includes people from multiple universities, the NHS and the UK govt.

It usually takes either tequila, a computer or a committee to really fuck things up.

Maybe, but if this is a fuck-up, it is a lot bigger than just Oxford. It involves the NIHR (that supervised the trial), the UK govt (that has now approved the drug for treating covid patients) and the WHO (that has congratulated the team after having access to "initial insights")

https://www.who.int/news-room/detail/16 ... 9-patients

The UK government currently has a strong interest in heavily hyping anything that comes from British research. They need a distraction from all the failures in leadership. This is likely why the UK government have been so strongly promoting the results before even pre-prints are available. They're making it into political capital and using patriotism to avoid answering any hard questions.

I should hope the NIHR signed off on it. If they didn't then the trial shouldn't have gone ahead.

A "congrats" from the WHO is nice but pretty meaningless. They haven't seen the full methods or data either.

Don't get me wrong, the study may well lead to better outcomes. But we won't get there by denouncing anyone who dares to ask to see the actual science first. Mistakes happen, things are overlooked and thousand over things could be wrong without any maliciousness or gross incompetence. It's all very well trusting the word of respected individuals and institutions but science moved past that a long time ago.

We have not seen the data, but the protocols are available in their site:

https://www.recoverytrial.net/files/rec ... -05-14.pdf
https://www.recoverytrial.net/files/rec ... _06_20.pdf
https://www.recoverytrial.net/files/rec ... s-v1-0.pdf

 

[rmgoat]

The fascinating Internet Book of Critical Care has been suggesting dexamethasone as a treatment since mid-March based on extrapolation of early data from China. Not being a medic I'm not sure how constrained ICU staff are in what medication they try when all else fails with a novel disease, I'm curious to know whether this has already been in wide use in the same sort of Hail Mary fashion as chloroquine?

Here is the problem. Dexamethasone (and other corticosteroids) have been used since the beginning. By some people. At different times. With different drugs. And different doses. Some folks though they worked, others weren't sure. So doing a controlled study was critically important.

But...

Patient selection, timing and dosing (which I guess we know) is absolutely critical. The idea behind the immunosuppressants of whatever flavor is that you mitigate the post infectious damage done by the immune system. Which isn't a problem for every patient. But is a significant issue for patients who end up in hospital and is *probably* universal for anyone on a ventilator.

So details matter. They really matter. It is completely unprofessional and really unethical to announce the findings in this way, especially with the rah-rah statements by the authors.

Very, very poor form Oxford.

To be fair, this was not just the decision of a couple of Oxford researchers. This press release was a decision of the Steering Committee of the trial that includes people from multiple universities, the NHS and the UK govt.

It usually takes either tequila, a computer or a committee to really fuck things up.

Channeling Mitch Ratcliffe ?

 

[BugblatterII]

Normally I agree with the zeitgeist on Ars but not this time

The scientists involved in the study are clearly convinced that it can save hundreds of lives per day. I've seen nothing about any real downsides to the drug. In their position would you wait for the paper to be completed and peer reviewed, with every day costing hundreds of lives that you believe could have been saved, or would you do what they did?

They still need to go through the rigorous process, and they will, but in the meantime they're saving thousands of lives or, worst case, curing some incidental eczema.

 

[mhalpern]

Steroids you say, I know a certain President that will be so buff soon!!

No steroids just numb the pain, they don't make you strong

 

[rjd185]

In an emergency situation I totally understand and approve of not waiting for the full publication process but releasing the data at a minimum should be a prerequisite for a press release. Ideally it should include a pre-preprint at least distributed privately to qualified researchers and clinicians describing their methodology.

The protocol/methodology and the analysis approach for the RECOVERY studies are already in the public domain freely available at https://www.recoverytrial.net/results and have been for a while now. The press release re dexamethasone includes the results of analysis (to the level of gross numbers and statistical assessments) and the commitment to publish data as soon as possible. As far as I could see there is no public link for the raw data yet nor a published date when that would happen (s6.7 of the protocol outlines the intended approach for publication with s5.4 noting retention periods).

Some comments here have also linked this to the withdrawal of the Lancet article covering the publication of harms due to hydroxychloroquine. That study was not part of the RECOVERY trial - the relevant hydroxychloroquine communication from RECOVERY trial is at https://www.recoverytrial.net/files/hcq ... al-002.pdf.

 

[rjd185]

The fascinating Internet Book of Critical Care has been suggesting dexamethasone as a treatment since mid-March based on extrapolation of early data from China. Not being a medic I'm not sure how constrained ICU staff are in what medication they try when all else fails with a novel disease, I'm curious to know whether this has already been in wide use in the same sort of Hail Mary fashion as chloroquine?

Here is the problem. Dexamethasone (and other corticosteroids) have been used since the beginning. By some people. At different times. With different drugs. And different doses. Some folks though they worked, others weren't sure. So doing a controlled study was critically important.

But...

Patient selection, timing and dosing (which I guess we know) is absolutely critical. The idea behind the immunosuppressants of whatever flavor is that you mitigate the post infectious damage done by the immune system. Which isn't a problem for every patient. But is a significant issue for patients who end up in hospital and is *probably* universal for anyone on a ventilator.

So details matter. They really matter. It is completely unprofessional and really unethical to announce the findings in this way, especially with the rah-rah statements by the authors.

Very, very poor form Oxford.

To be fair, this was not just the decision of a couple of Oxford researchers. This press release was a decision of the Steering Committee of the trial that includes people from multiple universities, the NHS and the UK govt.

It usually takes either tequila, a computer or a committee to really fuck things up.

Maybe, but if this is a fuck-up, it is a lot bigger than just Oxford. It involves the NIHR (that supervised the trial), the UK govt (that has now approved the drug for treating covid patients) and the WHO (that has congratulated the team after having access to "initial insights")

https://www.who.int/news-room/detail/16 ... 9-patients

The participants and funding for the trial are properly documented at https://www.recoverytrial.net/ and in the referenced trial registrations.

 

[rjd185]

"...it is unacceptable to tout study results by press release without releasing the paper."

There needs to be more discipline. The masses do not understand research and the false alarms have caused harm to the effort to control the pandemic and the reputation of medical science.

Offset this against the lives that could be saved or lost by expediting or delaying release of the initial results. As someone who has been through this cycle published in peer reviewed journals this can often take months to get a formal write-up to publication standard and 6 months from acceptance into print.

The institution running the study is Oxford University and I work at one of the study sites. It's a big study that as far as I can see is being run properly.

Yeah that's not how it works. Just because the authors believe their results show efficacy doesn't give them the right to make blanket statements without publishing associated data to back it up! Because THEY MIGHT BE WRONG. You publish information so that other experts can critique your methods and results and judge for themselves whether they find the data compelling or not. Rarely are results so cut and dried that a single study definitively proves the efficacy of a treatment. Even with the best run studies there are almost always methodologic concerns that need discussion and analysis.

Just look at the literature debate on steroids in sepsis. There have been TONS of actually published papers and yet still it's a big question. So no, the authors need to put up or shut up, because their 2 page presser which completely lacks very significant details about study methodology doesn't give anybody the ability to trust them in any way.

Surprising how many comments here are focused on the lack of publication of the methodology. The study design, methodology and analysis approach are on the web site for free public download to those who can be bothered to read and have been for weeks.

 

[rjd185]

Going back to the P-values, the 0.0003 means there's only a 0.03% chance that the reduced deaths on ventilated patients is a statistical fluke. That's really good. 0.14 on patients without respiratory support means there was a 14% chance that the increased deaths were just a fluke. That's big enough of a chance of not panning out that it wouldn't be appropriate to claim that the study showed an increase.

Edit: You do have to be a bit careful with P-values for a couple of reasons, that are best explained by someone much wittier than I am.

P-values don't include priors. You need Bayesian interference for that. https://xkcd.com/1132/ That's not really a problem here since Covid is novel and we don't really have prior data to contradict, plus there's a reasonable explanation why this drug works as it does.

If you try enough times, eventually you'll get a result with a low p-value. This is called p-hacking, and is covered by https://xkcd.com/882/ This isn't likely to be a problem here: even though this study is testing several different treatments, p=0.0003 is REALLY unlikely to be random.

Well, the p=0.0003 doesn't mean shit unless the methodology by which it was determined can be externally validated. Or would at least be available for cursory inspection. You can get really, really impressive P-values if you just faff about enough with either your methods, your data, or both.

IMO until they at least publish a decent pre-print which confirms good (enough) methodology,* no P value in the world means all that much. I was just responding to the question regarding relative mortality being higher in the moderately ill cohort.

Let's face it, conducting scientific research by press release does *not* engender confidence in the quality of the underlying data set, not with the eleventy-act shitshow we've seen on that front so far.

I agree P values in themselves do not consider prior plausibility. However, viewing P-values in context very much *should* consider prior plausibility; it's how you can generally weed out the one fluke positive with no biologically plausible backing (as you mentioned) or, more insidiously, the systemically wrong result that's based on errors in methodology, or, as so often seen in quackery, outright fraud or self-delusion.

The 'reasonable explanation why this drug works' is very much the prior plausibility / mechanism of action I was on about. Medicine and pharmacology inherently deal with data sets that are just too dirty not to apply such a filter.

I'll certainly agree I wasn't clear enough in my original post!

*it's a reputable group, I'd be willing to take their word for the validity of their dataset

((PS. also, re: being careful with P-values: if anything looking at the actual p=0.05 CIs on this press release presents an even rosier picture of the actual efficacy IMO, so no disagreement there either!)

If you are comfortable with taking their word for the validity of the data set, then is reading the published methodology and checking the referenced public trial registrations also an option?

 

[Borderliner]

Surprising how many comments here are focused on the lack of publication of the methodology. The study design, methodology and analysis approach are on the web site for free public download to those who can be bothered to read and have been for weeks.

Likely a strong correlation with "failed to follow the single important link in the article"

 

[jdale]

Here is the problem. Dexamethasone (and other corticosteroids) have been used since the beginning. By some people. At different times. With different drugs. And different doses. Some folks though they worked, others weren't sure. So doing a controlled study was critically important.

But...

Patient selection, timing and dosing (which I guess we know) is absolutely critical. The idea behind the immunosuppressants of whatever flavor is that you mitigate the post infectious damage done by the immune system. Which isn't a problem for every patient. But is a significant issue for patients who end up in hospital and is *probably* universal for anyone on a ventilator.

So details matter. They really matter. It is completely unprofessional and really unethical to announce the findings in this way, especially with the rah-rah statements by the authors.

Very, very poor form Oxford.

To be fair, this was not just the decision of a couple of Oxford researchers. This press release was a decision of the Steering Committee of the trial that includes people from multiple universities, the NHS and the UK govt.

It usually takes either tequila, a computer or a committee to really fuck things up.

Maybe, but if this is a fuck-up, it is a lot bigger than just Oxford. It involves the NIHR (that supervised the trial), the UK govt (that has now approved the drug for treating covid patients) and the WHO (that has congratulated the team after having access to "initial insights")

https://www.who.int/news-room/detail/16 ... 9-patients

The UK government currently has a strong interest in heavily hyping anything that comes from British research. They need a distraction from all the failures in leadership. This is likely why the UK government have been so strongly promoting the results before even pre-prints are available. They're making it into political capital and using patriotism to avoid answering any hard questions.

I should hope the NIHR signed off on it. If they didn't then the trial shouldn't have gone ahead.

A "congrats" from the WHO is nice but pretty meaningless. They haven't seen the full methods or data either.

Don't get me wrong, the study may well lead to better outcomes. But we won't get there by denouncing anyone who dares to ask to see the actual science first. Mistakes happen, things are overlooked and thousand over things could be wrong without any maliciousness or gross incompetence. It's all very well trusting the word of respected individuals and institutions but science moved past that a long time ago.

We have not seen the data, but the protocols are available in their site:

https://www.recoverytrial.net/files/rec ... -05-14.pdf
https://www.recoverytrial.net/files/rec ... _06_20.pdf
https://www.recoverytrial.net/files/rec ... s-v1-0.pdf

Ok, let's take a look at that protocol. This is from page 6 under Design Considerations:

The protocol is deliberately flexible so that it is suitable for a wide range of settings,
allowing:
* a broad range of patients to be enrolled in large numbers;
* randomisation between only those treatment arms that are both available at the
hospital and not believed by the enrolling doctor to be contraindicated (e.g. by
particular co-morbid conditions or concomitant medications);
* treatment arms to be added or removed according to the emerging evidence; and
* additional sub-studies may be added to provide more detailed information on side
effects or sub-categorisation of patient types but these are not the primary objective
and are not required for participation.

Nope. I can stop right there. The protocol does not tell me the number of patients, the number of treatment arms, any consequences of contraindications, etc.

The protocol also, necessarily, does not indicate how many patients withdrew from the study and whether that was the same between study arms, or why they might have withdrawn.

 

[password123]

Many good debates, but dang Science sure is hard.

 

[Deleted member 388703]

 

[wolfwood6]

Cheap steroid huh? Not going to be cheap for long...

 

[drmn43]

While important for scientific knowledge to disseminate quickly, we have to take caution prior to peer review - this process is in place for a reason. I am sure that this study will hold up to peer review ultimately, but, for example, one component that has not been discussed is how this trial was analyzed from an assignment standpoint.

These risk ratios may be derived from the intention to treat analysis or the as-treated or per-protocol cohorts. Although an as-treated analysis would tend reflect accurately the magnitude of benefit of the therapy itself, it may result in deviations of the initial randomization such that the samples are no longer matched, which can result in confounding. Particularly in a complex, large-scale trial like this, the the magnitude of benefit may be substantially different than how the results are quoted in this press release.

 

[aleph_nought]

Interesting that dex is used for cerebral edema and altitude sickness for Everest climbers. COVID-19 could be having similar effects in the bloodstream by affecting oxygen binding.

 

[Happy Medium]

Ok, let's take a look at that protocol. This is from page 6 under Design Considerations:

The protocol is deliberately flexible so that it is suitable for a wide range of settings,
allowing:
* a broad range of patients to be enrolled in large numbers;
* randomisation between only those treatment arms that are both available at the
hospital and not believed by the enrolling doctor to be contraindicated (e.g. by
particular co-morbid conditions or concomitant medications);
* treatment arms to be added or removed according to the emerging evidence; and
* additional sub-studies may be added to provide more detailed information on side
effects or sub-categorisation of patient types but these are not the primary objective
and are not required for participation.

Nope. I can stop right there. The protocol does not tell me the number of patients, the number of treatment arms, any consequences of contraindications, etc.

The protocol also, necessarily, does not indicate how many patients withdrew from the study and whether that was the same between study arms, or why they might have withdrawn.

Exactly. The approved protocol is generally what the study authors INTEND to happen. What it doesn't say anything about, which is critically important, is what ACTUALLY happened. It was an open-label study. Providers knew what was being given to whom, and because of that it's entirely possible that bias developed DURING the study.

The only way to get a sense of that is by being able to critique the allocation figure and arm demographics of the study, which are only provided with the actual study writeup. So once again, the study authors need to publish before they start spouting press releases.

 

[Fatesrider]

Yeah, I'm not a huge pro-chloroquine guy (don't really care one way or another), but that retracted "study" from a bunch of non-scientists touting how it doesn't work at all did NOT help with public trust about these topics.

https://www.statnews.com/2020/06/04/lan ... ria-drugs/

My understanding is that the promise of this drug is coming from the NIH which is not some fly-by-night group. Nevertheless full vetting is appropriate here.

If this was any other administration, I might have agreed with you.

But being good at chemistry and pharmacology (yes, I did that, too, back in the day), I have SOME inkling about the effects of HCQ.

Since I sling the lingo, I did some deep research into the cellular and systemic impacts of it. Nothing formal. Just using the data to synthesize a conclusion based on "gut" (there's a reason I do that I can't go into, but it's fairly effective for arriving at accurate conclusions).

It's a complicated interaction and depends largely on the chemistry of the individual (chemistry used pretty broadly here). Dummying it down horribly, people with the right chemistry will respond in a less negative fashion than those with the wrong chemistry. The less negative fashion response produces a very slight improvement in overall outcome because of the way the body responds to that negative response.

BUT...

If the negative reaction is stronger the outcome is catastrophic, because there's a paradoxical response from the immune system that causes more inflammation, vascular insufficiency, hypoixia and death.

So, with it, a little bit bad reaction results in a better outcome (not a lot, though) but a bigger bad reaction (and the "bigger" that triggers it varies a lot depending on the individual) can (and does) kill people more often.

I'd link the three or four dozen articles I've read over the last two months to get here, but that'd be overkill. Suffice to say there was some indication it might help. But on closer examination (even without the discredited study saying it doesn't work, which was wrong in a lot of cases) the chemistry, biology (which is what pharmacology is all about) and what published studies that have been peer reviewed all point to a conclusion that it's not going to help, and will probably do worse things.

A steroid...

I know a lot more about them, but they've not been THAT effective in helping up to this point (at least in the margins of what I've been looking at), though the inflammatory responses that lead to the worst outcomes SHOULD respond well to them. Apparently, they don't.

In that sense, biologically and physiologically, there may well be a steroid that works a whole hell of a lot better than HCQ would as an anti-inflammatory agent. In these cases, a little goes a longer way than a lot (that's what happens with HCQ, the inflammatory response goes haywire in bad reaction outcomes). The same happens with steroids (from what I read).

But we have a lot more anti-inflammatory agents out there that MIGHT work. HCQ isn't a good one. Steroids work, but might be "overkill" in a more literal sense than is comfortable. NSAIDS work on a different inflammatory mechanism (the study of which is a degree in and of itself).

So, long story short (too late, I know, sorry), steroids would seem to be the go-to option for dealing with the kind of inflammatory responses seen in crashing cases of COVID-19 victims. But it may be simply a case of having a lot of keys on a ring and only one or two will fit to keep the case from crashing.

Yeah, it's that weirdly picky about things. Lots of factors involved, since people are individuals and not identical in the first place.

Go deep dive into one of these things sometime for an eye-opening experience into what the researchers are coping with. I get the studies enough to understand the outcomes and results and why, but Jesus, trying to do that research myself would drive me to drink.

 

[Sangrea90]

Oh come on! This is ridiculous. Just another cheap attempt to sell something through the corona scare. I mean, the rates of new cases are dropping, and people are getting more and more gullible because they are starting to feel safe.
https://corontine.live/
Checking the local spread is always refreshing. I mean, it's still a real thing.

 

[ramza32]

"There were also patients who did not require any respiratory intervention and, of those, 13 percent died."

so, what is the cause of death for these patients group, blood clotting?