W
The philosophy of medicine has come a long ways. You need the following things:
Large scale trial: Like 7-10,000+ of participants in the study with many events of the endpoints
Long duration: To generate that many events it take a long duration
Randomization: If you don't randomize people at the start it can give misleading answers, computers do it now to get extreme similar between drug vs. placebo groups at baseline
Double blinded: Both the participants and the clinicians cannot know whether they got the drug or the placebo
Then with statistics you can prove the probability. Say in the trial 500 people died of the disease in the vaccine group and 600 people died in the placebo group. You can calculate the chances that was random. And you get the p value and a confidence interval. You then do smaller studies and other big studies in different places with different conditions and see if you get the same result.
So say you get 1 big trial and it was a 1 in 10,000 chance that it was better than placebo 'by chance'. The more trials you do the more you refine how much better and which patient populations it is best in.
How does the FDA determine what we 'know'. The FDA wants to see a <1 in 10,000 chance that the drug beats placebo was by chance. And the FDA wants to see a >15% improvement in clinical endpoints.
So say someone is claiming they ate sawdust pills and their eczema cleared up. Did the sawdust cure their eczema, who knows. Because if you have 100 people with eczema some of them the eczema flairs up worse, others it gets better even if they do nothing. Or there are unknown other factors involved.
Now if 10,000 people who were randomized and taking sawdust for 1 year and in the sawdust group 30% of them had improvement in their eczema. While in the placebo group 15% of them had improvement in their eczema. Then we can calculate the odds and it would be sawdust works.
Thats what evidence based medicine is.
Large scale trial: Like 7-10,000+ of participants in the study with many events of the endpoints
Long duration: To generate that many events it take a long duration
Randomization: If you don't randomize people at the start it can give misleading answers, computers do it now to get extreme similar between drug vs. placebo groups at baseline
Double blinded: Both the participants and the clinicians cannot know whether they got the drug or the placebo
Then with statistics you can prove the probability. Say in the trial 500 people died of the disease in the vaccine group and 600 people died in the placebo group. You can calculate the chances that was random. And you get the p value and a confidence interval. You then do smaller studies and other big studies in different places with different conditions and see if you get the same result.
So say you get 1 big trial and it was a 1 in 10,000 chance that it was better than placebo 'by chance'. The more trials you do the more you refine how much better and which patient populations it is best in.
How does the FDA determine what we 'know'. The FDA wants to see a <1 in 10,000 chance that the drug beats placebo was by chance. And the FDA wants to see a >15% improvement in clinical endpoints.
So say someone is claiming they ate sawdust pills and their eczema cleared up. Did the sawdust cure their eczema, who knows. Because if you have 100 people with eczema some of them the eczema flairs up worse, others it gets better even if they do nothing. Or there are unknown other factors involved.
Now if 10,000 people who were randomized and taking sawdust for 1 year and in the sawdust group 30% of them had improvement in their eczema. While in the placebo group 15% of them had improvement in their eczema. Then we can calculate the odds and it would be sawdust works.
Thats what evidence based medicine is.
