Differences in the characteristics of responders and non-responders in a prevalence survey of vertebral osteoporosis. European Vertebral Osteoporosis Study Group.
O'Neill TW., Marsden D., Silman AJ.
Men and women aged 50 years and over were recruited for participation in a population-based prevalence survey of vertebral osteoporosis from 36 centres in 19 European countries. All subjects were invited to attend by letter of invitation for a "lifestyle" interview and lateral spinal radiograph. The aim of this analysis was to investigate how far those who agreed to attend were representative of the target population and thus whether any important level of non-response bias existed. To address this a second invitation was sent to all non-responders and, in 20 centres, a sample of ultimate non-responders was contacted by mail, telephone or home visit and given a shortened version of the lifestyle questionnaire. Compared with the sample of non-responders, responders might be considered less at risk from osteoporosis in that as a group they took more exercise and were less likely to be current smokers. Other factors suggested the contrary in that they consumed less calcium and were more likely to have suffered a previous fracture. Amongst responders, these factors appeared also to be related to the timing of response. Thus compared with delayed responders, those who participated after a first letter of invitation took more exercise, were less likely to be smokers and more likely to have suffered a previous fracture. However, in contrast to the results suggested by the non-response survey early responders consumed more calcium than late responders. The magnitude of the differences between responders and non-responders was small (less than 10% for most of the categorical variables) and the differences were not consistently in the direction of an increased or decreased risk of osteoporosis. Additionally the size or direction of these differences was not consistently influenced by the response rate based on classifying centres into those with a high, medium or low response rate. This suggests that in this multicentre study response bias probably does not have a major influence on the prevalence estimates of vertebral fracture. In epidemiological studies of osteoporosis comparison of the lifestyle differences between early and late responders provides useful information concerning response characteristics.