Why are we interested in proportion? In (bio)statistics, sometimes I see that the proportion is of interest. I am looking for justification, why we do that. For example, in this article, "key secondary end points were the proportion of participants achieving weight loss $\geq$ 5%".
In this example, there is no difference between weight loss 12% and weight loss 5.7%. Both of them are counted equally when comparing to 5%.
My question is in general, why are researchers (sometimes) interested in the proportion? Why do they ignore absolute values and concern relative values? Is there any general justification?
Edit: In the example, my question is translated as:
Why are we not interested in the difference between 12% and 5.7%. Why do we take 5% into account. Why do we have to introduce a "threshold"?
 A: Some things in live are digital. Either a user clicks on your advertisement or she does not. You are either admitted to study medicine or you are not. You pass the exam or not. Probands finish a study or drop out early.
These things are largely binary and should be modelled as binary.
Sometimes you do not have enough data to model something not so binary in a more suitable way - Corona-Virus Mortalitiy is a huge topic right now. So some people appeared to be healthy and died from it and some were bound to die soon but did only short time earlier from corona virus and ohters were infected and ill with corona virus but there is a assumption they died form a heart attack that has little to do with it. It is very hard in many single cases to tell -- impossible to judge in large scales like for a whole country. So you may decide to only compute in a binary model based on the data you can access, not on data you would prefer to have but cannot get.
Also statistics is not only for specialists to compute but also for audiences to communicate data with. It may sometimes be acceptable to compute something with a model, that is not the best model but the models whose results are best communicated with decision makers.
Edit:
Extending on my last point of statistics sometimes have to be communicable. The paper you cite is a good example for that. It points out that

one of the efficacy thresholds for U.S. Food and Drug Administration approval of obesity therapeutics is ≥5% placebo (PBO)-subtracted reduction in body weight after 1 year of treatment

So obviously people at the FDA will assume, that loosing 4.9% is good for the obese as well but they are in need to determine strict rules to make decisions somewhat transparent and consistent and predictable.
