Interpretation of data in epidemiological cohort studies may be confounded if differences exist between non-participants and participants. In the Malmö part of the European Vertebral Osteoporosis Study (EVOS), which was designed to evaluate the prevalence of vertebral deformity in 50-80-year-old persons, we compared 74 men and 95 women who had been invited, but declined to participate with age- and gender-matched participants as regards alcohol abuse, previous fractures and subsequent mortality, factors known to affect the prevalence of vertebral deformity. We found more men with alcohol abuse, more men with a previous fragility fracture and a tendency to more men with a previous clinical vertebral fracture among the non-participants than in the male participants. In contrast, there were fewer female non-participants than female participants with a previous clinical vertebral fracture. The mortality rate during the decade after the baseline examination was higher among both male and female non-participants. The "true" prevalence of vertebral deformity in the whole male population at risk in Malmö seems to be underestimated in the EVOS study. In women, it is more difficult to estimate the combined result of the confounding factors. Conclusions based on the EVOS participants may not be applicable to the whole population at risk.