Discussion: Some fields in health care have no tradition in performing RCTs. This is especially the case in fields where the applied interventions are not well defined or poorly standardised and where the intervention is patient taylored. In these fields N=1 trials have become very popular. It seems to be a scientific way to choose the best of two or more interventions for a patient and it is easily integrated in health care. It is sometimes suggested that by pooling the results of these N=1 trials a RCT may be mimicked.

By comparing the differences in design and in application of N=1 trials and RCTs, it will be argued that meta-analyses of a number of N=1 trials cannot replace a randomised clinical trial. The main problems occur in diversity of the study populations and of interventions and, most of all, in expected publication bias of N=1 trials.