Context: Monitoring of iodine fortification programs is required for early recognition of oversupply or ineffective supply with iodine. Potential effects of non-response studies that are aimed at monitoring such programs have not been analysed.
Objective: With the present analyses we investigated the potential role of selection bias for studies performed to monitor the iodine supply and prevalence of thyroid disorders.
Design: The prospective cohort Study of Health in Pomerania (SHIP).
Subjects: From 4310 baseline respondents, 3949 were eligible for follow-up. Due to nonresponse, 649 subjects were lost. The 3300 follow-up respondents were divided into early and late respondents according to their recruitment level.
Main outcome measure: We used the baseline SHIP population as a hypothesized source population and associated the information on response behaviour from the 5-year follow-up examinations with baseline socio-demographic and thyroid-related characteristics.
Results: We found no significant bias in the prevalence estimates of thyroid-related variables in the comparison between all respondents and the hypothesized source population after response maximization techniques had been applied. Prevalence estimates of several thyroid-related characteristics, however, were biased if we would have only investigated subjects who agreed to participate in our study after the first postal invitation. All differences in thyroid-related characteristics between early respondents and the hypothesized source population were attenuated when analyses were adjusted for socio-demographic variables.
Conclusions: We conclude that the risk of selection bias in prevalence studies in thyroid epidemiology can be diminished by extended efforts put into the invitation procedure. Analyses of risk factors for thyroid disorders should be controlled for determinants of non-response bias that also modulate the risk of thyroid disorders.
03 - 07 May 2008
European Society of Endocrinology