Quote:
Originally Posted by OrangeBlossomBaby
11 patients makes for an absolutely guaranteed, solid, confirmed, bonafide, verified, reliable prediction that this cocktail will -not- work 100%, AND that it comes with risk of death (since 1 had to discontinue because the drug was giving him deadly side effect).
When you start with the hypothesis that it's 100% effective, then a single failure to be effective - crumbles that hypothesis and renders it categorically false.
|
Be cautious not to over-interpret findings in either direction. The original study from Marseille did report that 100% of six patients given both hydroxychloroquine and Zithromax cleared of detectible virus, both by culture and PCR. However the author made no claim that it would work 100% on all patients. Most studies report a confidence interval or range. That is - given the result we found, what is the true result in a larger study likely to show. These authors did not calculate an interval. They just reported their data.
Think of it this way. You are going to flip a coin. We know that it should be 1/2 heads, 1/2 tails. But you are coin naïve and don't know that. So you flip it six times and get all heads. Does that prove that all coins have heads on both sides? Or that no matter what else, the head side always lands up even if there are other options? Another group tries to replicate. It tosses a coin 10 times. It gets 8 tails and 2 heads. That doesn't prove your 6 results were wrong only that your study had too few examples to establish the scientific proof to answer "What are the odds when you flip a coin". The only thing the second study did is show that any conclusion made on the first 6 coin flips was wrong as the test size was inadequate.
When you hear about meta analysis studies, those are when the work of other groups is combined to get a better picture because we have more numbers. In our coin case we can say we have 6 + 10 = 16 flips if we combine data. And our result is amazingly 8 heads and 8 tails. It doesn't usually work out that well. But even with that result it is way to early to say "coin flips will result in 1/2 heads" because clearly if the real result is 60% to 40% after only 16 flips we could easily have 1/2, 1/2. If this were a study we should report our result as 50% heads with a confidence interval.
This is a long explanation but it is important to understand why some are cautious about drawing final conclusions from small data sets. However, your statement that the second study established that use of HCZ and Zith is NOT 100% effective is accurate. It still could be 99% and give the results we have so far. It could be 0 % and give the results we have so far. How could it be 0 when there are patients who got the meds and cleared. It surely must be higher than zero!
No, because you cannot show that those patients who cleared did so because of the drug. That is why you need a control group as I have stressed. You have to show how many patients become test negative with no treatment before you can establish your intervention had an effect on the outcome
Example:
If without any treatment 1/2 of Covid patients become test negative by 5 days [their body clears the virus itself] and 1/2 are still positive by 5 days then I do foot massage on 10 patients with Covid and publish my result that after only 5 days of massage 1/2 of those patients are now Covid free, my data is true but any conclusion such as Foot Massage has been shown to cure 50% of patients is false. Foot massage caused no improvement at all.
It is very difficult for people not to conclude that results must be caused by prior actions. The phrase post hoc ergo propter hoc gets thrown about in that situation. My favorite example is that you can show that having the sun warm the water at the beach causes drowning. The data are clear. As the water gets warmer more people drown. You see through that because you know there is some other factor I am not taking into consideration. But that is not clear if you don't already have knowledge about the topic and can't see the hocus pocus of the statistical manipulation.