Commentary: On the levels of patient selection in registry-based randomized controlled trials.

Registry-based randomized controlled trials (RCTs) are presumed to include a less-selected patient population and thus to have enhanced generalizability compared to conventional RCTs. However, this view disregards the levels of patient selection in registry-based RCTs: the registry selection level and the trial selection level. At both levels, systematic selection can occur and generalizability can be diminished. Nevertheless, using a registry as a basis for recruitment, randomization, and data collection results in an advantage: the trial selection takes place within the registry framework, where baseline characteristics of non-enrolled patients are automatically documented as well. By comparing the baseline variables of the enrolled and non-enrolled patients, the trial selection can always be investigated, which gives a sound basis for discussing the generalizability to the registry population.