Oh shit! What's a Latrinsorm to do when he finds out....

[drumpel]

Ivermectin is beneficial after all.


https://www.cureus.com/articles/8216...score-matching

[Parkbandit]

Ivermectin is beneficial after all.


https://www.cureus.com/articles/8216...score-matching

"It's horse medicine" - Probably Latrinsorm.

[Neveragain]

Ivermectin is beneficial after all.


https://www.cureus.com/articles/8216...score-matching

The ivermectin non-users were two times more likely to die of COVID-19 than ivermectin users in the overall population analysis.

[Shaps]

But Pfizer and YouTube said it doesn't work!!!

[dzulthu]

tl;dr garbage in, garbage out

Observational Study of 223,128 Subjects Using Propensity Score Matching

Do you know what a double blind study is, read the study, understand what response bias is, or read the credentials of the researchers who are behind the study? Eight out of nine of them have no other published research, and the ninth has been active in research for a grand total of one year. Light Google searching shows one to be a self-proclaimed data analyst but with little information on credentials.

Maybe there's something to it. They have nothing that can be used that would separate causation from correlation. If someone brought statistical analysis like this to me with a proposal for funding or anything other than demonstrating the ability to put together flow charts, I'd end the meeting.

Of the 223,128 citizens of Itajaí considered for the study, a total of 159,561 subjects were included in the analysis: 113,845 (71.3%) regular ivermectin users and 45,716 (23.3%) non-users. -- Why doesn't that add up to 100%? It may seem like a small thing, but indicators of sloppy analysis and documentation are red flags for any serious study.

In the absence of contraindications, ivermectin was offered as an optional treatment to be taken for two consecutive days every 15 days at a dose of 0.2 mg/kg/day -- this isn't how testing works. Offering something as an 'option' automatically introduces selection bias that makes it difficult to separate correlation and causation, unless you control for every single thing to isolate for impact. And this sample is way too small to enable that. You need to apply the actual treatment you intend to the test group and the control group.

Using PSM, two cohorts of 3,034 subjects suffering from COVID-19 infection were compared. -- this is equivalent to Elizabeth Holmes saying "a chemistry is done to analyze the blood." Earlier in the paper, they mention age, sex, and comorbidities. What about things like lifestyle, weight, and other risk factors for COVID?

After adjustment for residual variables -- this disqualifies any possibility of taking this article seriously. They said they will release the datasets after peer review, but that's not really what they care about, is it? It's to push out an academic paper so that people can do exactly what the OP is doing.

[dzulthu]

Reading up on Cureus a bit more, it's for crowdsourcing research and reviews. I'm not going to say it doesn't have its place, but it's not exactly the benchmark for credibility.

On the flip side, I also learned that the links for each of the authors takes you to their publication history only with Cureus. Some of them do in fact have more established research profiles. And some of them appear to only be active with minority-view COVID research.

[drumpel]

tl;dr garbage in, garbage out

Observational Study of 223,128 Subjects Using Propensity Score Matching

Of the 223,128 citizens of Itajaí considered for the study, a total of 159,561 subjects were included in the analysis: 113,845 (71.3%) regular ivermectin users and 45,716 (23.3%) non-users. -- Why doesn't that add up to 100%? It may seem like a small thing, but indicators of sloppy analysis and documentation are red flags for any serious study.

It's okay if you couldn't figure it out. There were other numbers in there, I know, it's a lot and hard to keep track of. I won't hold it against you....most likely another failure of the school system. But, don't just read the first few paragraphs, you'll find the info you're in search of further down. Here, I'll help you:


From the registry of the city population (223,128 inhabitants), subjects below 18 years old (61,583 subjects) were removed. Of the 161,545 subjects above 18 years old from the city of Itajaí, we removed the 1,984 COVID-19 cases that occurred before July 7, 2020, and 159,561 subjects remained. Subjects above 18 years old were considered those who were born before June 30, 2002.

A total of 147,223 subjects participated in the program of ivermectin prophylaxis used for COVID-19. Of these, 24,304 subjects were below 18 years old. Of the 122,919 ivermectin users above 18 years old, 8,346 were from other cities, and 728 had COVID-19 before July 7, 2020, although they used ivermectin afterward. In total, 113,845 subjects that participated in the program remained in the dataset. The 45,716 non-participants, remaining subjects among the 159,561 subjects, were considered as the ivermectin non-users.

After adjustment for residual variables -- this disqualifies any possibility of taking this article seriously. They said they will release the datasets after peer review, but that's not really what they care about, is it? It's to push out an academic paper so that people can do exactly what the OP is doing.

Raw data was provided. Everything will be released for peer review. Seems okay to me.

Big pharms - they keep their raw data, arrange the data they release to best fit their objectives (to make money and get the drug approved) in a peer review form and give this data to peer reviewers while they never hand out raw data.

[Vindicate]

tl;dr garbage in, garbage out

Observational Study of 223,128 Subjects Using Propensity Score Matching

Do you know what a double blind study is, read the study, understand what response bias is, or read the credentials of the researchers who are behind the study? Eight out of nine of them have no other published research, and the ninth has been active in research for a grand total of one year. Light Google searching shows one to be a self-proclaimed data analyst but with little information on credentials.

I don't disagree with what you are saying. That said, I think the takeaway is that all we have is fringe studies on anything that isn't Big Pharma approved. Everything else is crushed.

[dzulthu]

It's okay if you couldn't figure it out. There were other numbers in there, I know, it's a lot and hard to keep track of. I won't hold it against you....most likely another failure of the school system. But, don't just read the first few paragraphs, you'll find the info you're in search of further down. Here, I'll help you:


From the registry of the city population (223,128 inhabitants), subjects below 18 years old (61,583 subjects) were removed. Of the 161,545 subjects above 18 years old from the city of Itajaí, we removed the 1,984 COVID-19 cases that occurred before July 7, 2020, and 159,561 subjects remained. Subjects above 18 years old were considered those who were born before June 30, 2002.

A total of 147,223 subjects participated in the program of ivermectin prophylaxis used for COVID-19. Of these, 24,304 subjects were below 18 years old. Of the 122,919 ivermectin users above 18 years old, 8,346 were from other cities, and 728 had COVID-19 before July 7, 2020, although they used ivermectin afterward. In total, 113,845 subjects that participated in the program remained in the dataset. The 45,716 non-participants, remaining subjects among the 159,561 subjects, were considered as the ivermectin non-users.

.

I read the entire article. I do this for a living.

What are the numbers that would add up to 100%?

113,845/.713 = 159,670.4 -- checks out
45,716/.233 = 196,206.00

196,206 - 159,670.4 = 36,535.6

And the remaining 36.5K are ...?

[drumpel]

I read the entire article. I do this for a living.

What are the numbers that would add up to 100%?

113,845/.713 = 159,670.4 -- checks out
45,716/.233 = 196,206.00

196,206 - 159,670.4 = 36,535.6

And the remaining 36.5K are ...?

Well, 71.3 + 23.3 = 94.6, clearly that's not 100%.

Later in the report it shows 45,716 = 28.7% (multiple times)


113,845/.713 = 159,670.4
45,716/.287 = 159,289.2

Sadly, the percentage wasn't listed correctly in the first mention of 45,716 where it shows 23.3%.