"Statistics isn't magic"
agreed.
statistics are used all the time to measure small differences, and what you are comparing in these trials is the natural variability between all subjects vs. the difference between treatment groups. that is, if there is 5% natural variability among all members in the experiment and 10% difference between the control and treatment groups, then you can statistically come to a conclusion on the probability that the difference was not due to chance and at what confidence level.

you have to assume you are sampling from a normally distributed population, that most candidates are within the normal range of variability and not skewed at the extremes.

considering that many trials include people from all over the country/world, one can assume you are sampling from a normal distribution. most people will be average and a few many standard deviations different.


there is research showing tobaco reduces the probability of getting pd, there are no biomarkers, statisticians just survey the data. and yes, i think a 20%neuroprotective benefit could be detected over time without biomarkers.