Abstract
Currently, most evidence assessments in guidelines or health technology assessments (HTAs) rely on the assumption that a randomized controlled trial (RCT) is always the best source of evidence. However, if the outcome in a control group is certain, e.g. death within a short time with an almost 100% chance, or if an event can only occur in the treatment group, there is no need for a randomized control group; the evidence cannot be improved by a control group, nor by an RCT design. If a cause-effect relationship is certain ("primary or direct evidence"), a therapeutic effect can be diluted in the population of an RCT by cross-over, etc. This can lead to serious misinterpretations of the effect. While experts such as the GRADE group or Cochrane institutes recommend using all available evidence, the leading approach in many guidelines and HTAs is assessing "the best available trials", i.e. RCTs. But since RCTs only deliver probabilities of cause-effect relationships, it is not appropriate to demand RCTs for certain effects. A control group can only diminish the net value of a treatment since the outcome in the control group is subtracted from the outcome in the treatment group. Therefore, under identical circumstances, an RCT will always show lower effect rates compared to a single arm study of the same quality, for desired as well as for adverse effects. Considering these inconsistencies in evidence-based medicine interpretation, the evidence pyramid with RCTs at the top is not always a reliable indicator for the best quality of evidence.