Purpose: To investigate the extent of selection bias due to exclusion of recent immigrants with respect to prevalence of antidiabetic drug use. Methods: Data on antidiabetic drug redemptions were obtained from Odense University Pharmacoepidemiologic Database for the period 1992-2003 covering a population of approximately 470 000. Ordinary prevalence estimates based on drug redemptions during a run-in period for non-immigrants were compared to estimates based on the waiting time distribution (WTD) approach. The WTD method was used to compare estimates for non-immigrants to estimates for the entire population. Results: When a run-in period of half a year was used and treatment cessation was accounted for, reasonable agreement between run-in estimates and WTD estimates was found, range of relative difference was -3.1% to 4.5%. Analyses using WTD method showed that prevalence among immigrants was approximately half that of non-immigrants in each of the years 1993-2003, resulting in a relative bias (non-immigrants vs. entire population) of approximately 1% for each year (range 0.73%; 1.36%). Stratifying on age and sex made the bias vanish, except for ages under 40. Conclusions: Selection bias due to immigration, when studying prevalence of use of antidiabetics, is small after controlling for age and sex. The WTD approach is useful for studying selection bias and should potentially replace the use of run-in periods for determining treatment status. Copyright © 2007 John Wiley & Sons, Ltd.

Selection Bias Waiting time distribution Antidiabetic medications Pharmacoepidemiology Prevalence

antidiabetic agent selection bias human epidemiology immigration middle aged diabetes mellitus pharmacoepidemiology data base priority journal controlled study Time Factors Aged Humans Hypoglycemic Agents Adolescent Infant, Newborn male female Aged, 80 and over Infant Child, Preschool population research prevalence Article major clinical study adult drug use Emigration and Immigration Child