Thread by @FlavioCadegiani

1/26 Kintor announces the first partial analysis of the study with proxalutamide in outpatients.

The results? POSITIVE (very positive if we consider the current scenario)

2/26 Why do you consider the results positive if Kintor said they "didn't reach statistical significance"?
Due to the reduction in overall hospitalization rates, I was hoping that the study would even be interrupted because it wouldn’t be able to show the differences we found.

3/26 1. A. This is the study with outpatients, which evaluated whether proxalutamide prevented hospitalizations. There were very few hospitalizations among patients who were included in the research.

4/26 With very few hospitalizations, which was how the research was supposed to evaluate the efficacy of proxalutamide, it is VIRTUALLY IMPOSSIBLE to achieve a difference that we call as 'statististically significa t'. What is that?

5/26 B. It's the likelihood that a difference in results is due to actual efficacy, rather than by chance, that determines whether a drug is effective.

6/26 We say that a drug showed 'statistical significance' when the chance of the differences of being due to the efficacy of the drug is higher than 95%, that is, the chance of the results to have occurred by chance is below 5%.

7/26 C. So, in the case of the study, it is neither the percentage nor the number of hospital admissions that would determine the efficacy of proxalutamide.

8/26 D. I'll give you a hypothetical yet practical example:

- 100 patients were assigned to the proxalutamide group and 100 patients were assigned to the placebo group.

9/26 - There was 1 hospitalization among 100 patients in the proxalutamide group and 5 hospitalizations among 100 patients in the placebo group.

10/26 - There was, therefore, an 80% reduction in hospitalization rate. There were 5 times more hospitalizations in the placebo group than in the proxalutamide group.

11/26 - Still, the p is 0.14, which means that the chance of the result being by chance is 14%, much higher than the 5% allowed.

- Here, even with all this difference, I can only affirm that there was 'no statistically significant difference'.

12/26 - Now I think it makes sense what I said about the issue that when there are very few hospitalizations, or low hospitalization rate, it is extremely difficult to prove efficacy.

13/26 E. There is no way to generate statistical significance if you have no events (in this case, hospitalizations). It is almost impossible, from a statistical/mathematical point of view.

F. And that's what happened.

14/26 2. A. On the contrary, I was surprised that the study was not interrupted, and allowed the study of proxalutamide in hospitalized patients to be continued only.

15/26 B. The study was not interrupted because IT IS VERY LIKELY that even with a very small number of hospitalizations, PROXALUTAMIDE HAS ALREADY SHOWN EFFICACY

16/26 C. If it had not shown efficacy, the study would already have been interrupted due to a criteria called futility, that occurs when the drug does not show a tendency towards efficacy.

17/26 This would prevent including more patients in the study unnecessarily and would also save further financial efforts.

18/26 D. It is important, therefore, to make it very clear that if there was no efficacy, THEY WOULD HAVE ALREADY DISCONTINUED the study.

19/26 3. A. Since very few patients are hospitalized in the omicron variant (thankfully!), it will still be very difficult to generate a statistical difference from now on, even if there are large differences in terms of percentage.

20/26 B. That's why I don't see that the study can make a statistical difference ever. How to conduct a study in this context using hospitalizations as the only outcome?

21/26 C. Those drugs that have already been tested and showed efficacy had higher hospitalization rates (just like our previous studies). It is important that everyone knows and understands this.

22/26 4. By the way, in the study of hospitalized patients we published on Saturday the 25th (Christmas gift!), we already drew attention to the fact that differences in responses between studies on the same drug would not only be possible, but also expected.

23/26 5. This does not weaken the androgenic theory at all. On the contrary, not having interrupted the study only strengthens it.

24/26 6. Still, if proxalutamide had not worked, that would be was it called non-reproducibility of studies in science.

25/26 This is super common to occur in science, and it's even more common in COVID-19. So much so that drugs such as remdesevir went from more than 75% reduction in serious events to practically zero reduction, depending on the study.

26/26 7. As a scientist, and, above all, as a doctor, I don't have a preferred drug. It can be the drug a, b, c… to me, what really matters is to save lives.