Drugs newly developed for precision medicine, like all other drugs, must prove their efficacy and effectiveness prior to market approval. For this purpose, the “gold standard,” i.e. a randomized controlled trial (RCT), is usually conducted because it is supposedly able to derive causalities compared to other study designs.^[4] In the course of RCTs, correlations, i.e. dependencies between intake of the drug and effect, can be calculated. This step ultimately provides the first indications of whether a drug works as expected or not. However, these RCTs often have high inclusion criteria, so that new drugs are usually tested on people who do not correspond 100% to the patients that are going to be treated later. For this reason, the responder behavior, i.e. the (side-)effects of the drug on the study participant, sometimes differs greatly from the real patient. 

Another problem of RCTs is randomization, which is the decision that is made by chance (randomization) which study participant receives the drug and which participant receives a placebo. Causality suffers from randomization because any prior knowledge, e.g. about responder behavior, must be ignored for the purpose of randomization.^[1] This can lead to bias in the results. Nevertheless, among all available study designs, RCTs are the ones that can provide any evidence at all on causality.^[5] Thus, the argument of diminished causality due to randomization is given but not necessarily relevant.

On the basis of the evidence obtained in this way, physicians later decide which patient receives which drug. Studies have shown that the predictive accuracy (level of evidence) of the ten most prescribed drugs is only between 5 and 33%.^[1] In that perspective, it would be equivalent to a one-size-fits-most approach, which has little to do with precision medicine.^[1] 

So, to make precision medicine truly marketable and ubiquitously available, the level of evidence must be improved and thus the predictive accuracy.