Bias from non-participation: a study of low-level lead exposure in children.

Epidemiological studies on neurobehavioural effects in childhood are highly sensitive to selection bias. A study on the significance of low-level lead exposure was performed among school children in the first grade in Aarhus in 1982-83. Deciduous teeth were collected and analysed for the circumpulpal dentin lead concentration. On the basis of the lead levels, a case and a control group were identified. Information on socioeconomical status and past medical history was obtained for all children. A "common sense" stratification of the reason for non-participation, showed bias in particular due to non-responders, while "refusers" caused little, if any bias. A non-participation of 7% significantly distorted the main result of the study in the direction of the null-hypothesis. Current studies of low-level lead exposure vary greatly with regard to non-participation and the extent of selection bias may contribute to understand the apparent diversity of findings of studies in this field.