We use casual inference, a type of math that is basically allowing us to establish causal relationships based on actions we could have taken but didn't, and give everyone in the trial the drug. We can then compare what "would have happened" with what actually happened.