OT? Metformin may reduce long Covid

Considering the potential reduction of work force by long Covid, it’s possible that reducing long Covid might have Macroeconomic impact.

A common diabetes drug could lower long COVID risk by 40%

  • Researchers evaluated whether a common diabetes medication, metformin, can prevent long COVID.
  • They found that metformin can reduce long COVID diagnoses by 40%.
  • Further studies are needed to know if these findings apply to the general population since the study group consisted of overweight and obese people.

The study was a randomized phase 3 clinical trial. The researchers enrolled 1,126 patients aged between 30 and 85 years who had COVID-19 symptoms and a positive PCR or antigen test result for this viral infection. They were followed for 10 months.

While the patients were not hospitalized for COVID-19, they had either overweight or obesity, putting them at a higher risk for developing severe COVID. Overall, 8.3% of patients reported a long COVID diagnosis after 300 days of follow-up.

However, just 6.3% of patients that received metformin developed long COVID, compared to 10.4% who received a placebo.

Previous studies suggest that metformin has an antiviral effect by inhibiting its replication…
[end quote]

Metformin is an inexpensive drug with a very long history of safe use in diabetics. Metformin has also been shown to potentially reduce severe COVID and to stop [SARS-CoV-2 virus replication in a lab setting.

So far, there have been 6,183,075 hospitalizations for Covid. That doesn’t count many millions of Covid patients who didn’t go to the hospital. DH and I are almost the only people I know who haven’t had Covid and I don’t know anyone who was hospitalized.

If a simple 14-day treatment with metformin could keep 10% of Covid patients from getting long Covid, that would be HUGE. Goodrx.com has a 90 day supply of metformin for $13.72, no insurance required.

I agree that future studies should be done, but for myself it’s a no-brainer. If I got Covid I would immediately request Paxlovid ( the first-choice treatment for mild to moderate COVID in people with a higher risk of severe illness – worked great on an elderly friend) and then request 2 weeks of metformin. I’m marginally overweight and could make a good case for this.



The sample is meaningless.

A sample size over 1,000 is typical for U.S. clinical trials. The really big studies (hundreds of thousands) are from countries with socialized medicine and national databases.

The sample is NOT meaningless – or I would not have posted it.



There is nothing suspicious about the number of enrollees in this study. “Smaller” numbers are to be expected at this stage of information gathering in clinical studies.

Mind, a study arm that included folk who aren’t overweight/obese and with good body composition would be of more interest to me if it came to personal decision making with this drug. Smacks of BMI discrimination to exclude that demographic.


I think they focused on high BMI people because that demographic is at higher risk of severe Covid and complications. They were more likely to get a signal.


I know a sample size of 1000 is typical. It is meaningless. It is still published.

And so it should be. How else would it become part of the scientific archive? Publication is nothing more than opening up a hypothesis testing exercise to peer review…with opportunities for independent labs to refute or confirm the validity of the data. That’s the Scientific Method in action.


Absolutely correct.

Many, really most, of these things do not pan out.

That does not stop Ph.Ds from being true believers and setting out at all costs to prove it anyway. That is the risk the public has when people like Dr. Oz sell 500 berries per year. Most of what we hear are theories that do not hold water. Or that hold a tad of water. Or that hold water for only a few people.

adding statistically 1000 people in any of these studies is too small a sample.

The flipside with new medications is scary. The pharma companies in the first five years if a drug makes it to market the med can do a lot of damage most drugs are pulled from the market.

1 Like

That’s not true. For sure, most drugs in a company’s pipeline don’t make it to market. Also, a fair number do get withdrawn after they’re introduced but the figure is a long way from most.

1 Like

If you a population of say, 230,000,000 (that is, the size of the United States) and want a confidence level of say 95% (typical), and a confidence interval of say, ± 5% (also typical) to be statistically valid you need a sample size of…wait for it…384.

If you want a confidence level of 99% you need a sample of 666. This is an easy calculation to make, and you if like you can see for yourself that increasing the sample size above that amount doesn’t meaningfully increase the confidence level.

So for this type of study where they are looking to see if there is a signal, a sample size of 1000 is more than plenty fine.


It is true most drugs in the market for a short period of time are pulled because of side effects.

Skye yes there was a signaling. That does not mean it will hold water.

I think the main issue in situations like this isn’t so much the nature of the study design (we could argue all day about that for all the World as if we knew what we were talking about) …but the fact that it’s being announced in the media at an inappropriately early stage. Good old Science By Press Release. Preliminary data is just that and, as suggested, might just turn out to be another paper headed for publication in the Journal of Irreproducible Results.

The Covid era has seen a dramatic increase in publications being retracted after getting the banner headline treatment following press releases …and not just from the predatory, pay-to-publish rags, either. Come to that, even pre-prints have begun to appear. I stopped following Retraction Watch it became so depressing and I don’t recall these retractions ever getting 1/10th the publicity the gussied up press release did.

1 Like


My larger concern w our readers is the pharma industry. 1000 ssmples do not run long enough for side effects to be fully known. I’d rather err on the side y’all don’t believe it and avoid a lot of it. You can get hurt.

This is science you weren’t supposed to believe it

Adding as for covid it is an odder disease partivularly earlier on. It could attack in more thsn one way.

Cell my router is on the fritz

@syke6 you 230m pop 95%

Do you see the problem with that stat model?

Ai would never see it.

The problem there is no 1000 or 230million.

That in coding is a class problem.

The entire modrl is wrong. The class begins but is not at limited to DNA and RNA combinations and then all posible envirmental factors.

There was a signal. The data behind it is entirely missing.

Oh and yes labs in cambridge boston are in narrow and broader way deriving these larger datasets

I went back to the original article in The Lancet and that didn’t make much more sense than the summary.


Metformin is a long-acting drug that smooths insulin levels and reduces sugar levels for people with Type-2 diabetes. Diabetes and obesity were two of many factors found to trigger higher levels of acuity with COVID cases, up to death. Of course, there were and continue to be many examples of perfectly healthy individuals with zero “co-morbidity” factors who were also afflicted with acute COVID illness and / or subsequent long-haul COVID. (Remember, many people with now-debilitating long COVID often did not have terribly severe initial infections…)

The study itself didn’t identify a biological theory on why metformin was selected along with ivermectin or fluvoxamine for inclusion in the test. Were they thought to help reduce inflammation as the immune system attempted to fight a novel virus signature not previously seen? Were they thought to minimize damage to other systems while the body rid itself of the core COVID virus, thus reducing long-lasting aftereffects (while being useless after-the-fact)?

The study also only involved patients who a) took the course of drugs DURING their original acute COVID infection (whether hospitalized or at home) and b) were willing to submit to follow-up reviews remotely. The study kicked off in May 2021 and ended January 28, 2022 so patients with variants common outside of that time interval may have issues not reflected in the base of patients who enrolled in the study during that interval.

Consider the variations:

  • original “classic COVID” – December 2019
  • alpha – December 2020 in Britain
  • beta – December 2020 in South Africa
  • delta – late 2020 in India – biggest variant until omicron
  • omicron – November 2021 in Botswana / South Africa

From a calendar perspective, this study would have likely collected data from “delta” patients. At that point, there were already tens of thousands of patients of prior variants already experiencing long COVID issues.

While existing vaccines (and booster modifications) were effective at combatting these variants, the fact that these variants differed substantially in infection rates and severity of original infection illness suggests each variant was capable of unique damage to infected patients. Extrapolating results from a patient group with delta to the full population of infected patients with five variants seems dubious from a scientific standpoint.

Given such attention getting results for metformin helping to AVOID long COVID, it would be useful to know if any organization is executing a study to administer metformin to EXISTING long COVID patients to see if it helps mitigate / eliminate existing cases. One theory for long COVID is that long COVID patients have trace levels of the virus still in their system which escapes detection of tests but is still enough to trigger continued inappropriate responses from the immune system. If clinicians are just going to throw treatments against a wall in random combinations and measure results to see if something sticks for AVODING long COVID, they might as well try the same for “CURING” it.

Most doctors are still blowing off patients with long haul symptoms as they did with CFE patients. Recent statistics show 107,204,117 unique COVID cases in the US with 1,116,924 deaths. That (cases) - (deaths) delta would be 106,087,193 survivors but the same page shows a figure of 105,349,072 “recovered.” That implies 738,121 people “not recovered,” presumably because of lingering long-haul symptoms.

I think the number is WAY higher than that. Most GPs and most health care plans have zero clue about long COVID or organized programs for people to consult for help. And the range of impacts is so wide that “recovery” is not well defined. If you are bed-ridden, unable to perform physical work for more than five minutes, etc., sure – you have long COVID. What if you have brain fog? Random, uncontrolled racing heart rhythm? Shortness of breath. Chronic pain? Drastically increased allergy sensitivities? Out of 107 million total cases, I would bet there are at least 5 million Americans with long COVID ailments that have impaired their ability to work by between 10 and 30 percent. That might not seem acute but if you have to work on your feet all day, do critical think work as a medical / legal / technical professional, etc., those might as well be career ending impairments.

Early on in the COVID epidemic, I read something (probably in The Atlantic or The New Yorker) from a doctor on the front lines who said the world lost several months by focusing on treating COVID as a RESPIRATORY disease threat when it was apparent from some of the more shocking extended ICU cases that the root problem was inflammatory in nature. Once exposed to the novel virus, the immune system over-reacted and was doing things that impaired organs throughout the body – lungs for sure but heart, kidneys, brain, pancreas, etc. While we have drugs that have effectively halted transmission and infection, the fact that we still don’t understand the biology that COVID exploits to trigger so much damage is discouraging to say the least.




Well yes, the sample size is representative of the population as it currently exists. Not every combination that might theoretically exist.

However, if you understand math you would understand why the sample size doesn’t increase in proportion to the population size. So even if the population was 230 billion you still wouldn’t need a sample much larger than about 1000 (assuming the sample is representative, of course).

The statistics are the dead easy part of any study. The hard part is crafting the study such that the statistics are meaningful.

Yep the study is not meaningful because the model is simplistic.

So what should sample size be? Show your work.

It is not my work. Labs hire a special stats person to data mine dna in more reliable ways than a head count. The ph.d running the lab generally is not prepared to data mine

I’ve read a number of articles by statisticians who note that the important (and not always easy) part is using the proper statistics correctly.