RESUMO
INTRODUCTION: The internal and external validity of studies is endangered by many factors, such as selection of subjects for inclusion. Selection bias itself is a major problem, but remains unmentioned and probably unexamined in the majority of published clinical trials in traumatology. AIM OF THE STUDY: The aim of this investigation was to detect effects of subject selection which occurred during our own prospective intervention study. The clinical trial compared subjects with whiplash injury who were either treated by early mobilization or immobilization (soft collar). MATERIAL AND METHODS: Source population, eligible subjects, study participants and final study participants were compared for differences on various items like age, gender and further sociodemographic as well as crash related factors and clinical findings. RESULTS: Between 21.08.1997 and 30.04.1999 a total of 732 patients was examined and treated after whiplash in our trauma department. The options for inclusion were met by 453 patients. While 346 escaped from the study, 107 agreed to participate. Of these another 39 patients dropped out of the study. Selection effects were detected on two different levels, leading to distinct statistical procedures from those proposed in the study protocol. CONCLUSIONS: Uncontrolled selection effects could undermine the interpretability of the results of clinical trials. Awareness of selection effects is mandatory regarding the applicability of these results to subjects, other than those in the group of the final study participants.