The tools that we use to quantify medical phenomena are like a chainsaw cutting up a large tree trunk. When we use it on a tree trunk, it does a pretty good job and helps cut up the trunk into manageable pieces.  This is like when we set up a basic study to assess whether a given chemo agent works on a particular cancer.  The problem is that when we start to ask more specific questions about its efficacy (such as what subtypes of the cancer, what extent of the cancer, what complications it’s associated with, how it works in patients with differing health statuses etc etc),  the tool becomes increasingly cruder and less useful, to the point where it is like using a chainsaw to slice a strawberry.  The problem is that the instruments and tools we use to quantify the detailed complex medical phenomena are simply too crude and lacking in nuance to be able to measure most of the phenomena meaningfully.  The margin of error begins to exceed the actual measurement. We cannot hope to really measure most things with the primitive tools and instruments. No matter how hard we pretend that we can.