Same question. I would assume this is partially based on replicability, or doing the study over and getting the same results. Recently, there was an attempt to replicate major studies in psychology and only about 36% were replicated. I think something similar was in medicine.