BACKGROUND: In population-based studies non-participation rates of about one third of the sample can be expected. The number of refusals may even be higher, if personal attendance of the subjects is requested. A different participation behaviour of the diseased and non-diseased may affect the prevalence estimation of a disease as well as the risk factor association. METHODS AND RESULTS: The European Vertebral Osteoporosis Study (EVOS) is an international, multicenter, cross-sectional survey in men and women aged 50 to 79 years. Within Europe 36 centres recruited 17,342 participants. The 8 German centres contribute about one forth of the data. In Germany subjects were recruited in several steps: random sample drawing from population registries, initial postal questionnaire, medical interview, lateral X-rays of the thoracic and lumbar spine. Depending on the extent of participation a variable amount of sociodemographic data and information on subjective health is available from the non-reachable, non-responders, and responders with incomplete and with complete examination. A comparison of the different participation groups showed, that especially old women were lost from the study. Compared to those, who only answered to an initial questionnaire, subjects, who were interviewed and X-rayed, more often suffered from back pain but reported a better functional capacity. This could be a hint on a selection of a population of the "worried well". CONCLUSION: On the whole the selection processes seem to have a minor influence on the outcome. The initial postal questionnaire turned out to be effective in collecting basic information from those who refused to attend a personal examination.
