A study of the effects of exposure misclassification due to the time-window design in pharmacoepidemiologic studies.

This paper considers the effects of the time-window design on the validity of risk estimates in record linkage studies. A time-window constitutes the number of exposure days assigned to each prescription, often fixed time-intervals. Prescription information was drawn from 36 Dutch pharmacies. Persons, assuming full compliance to the dosage regimen, used NSAIDs during 58% of the 30 day window time (31% with 90 day window). This proportion ranged from 51 to 81% for different NSAIDs; from 75% for elderly to 35% for younger persons. We observed with longer windows a substantive attenuation of incidence rates of peptic ulcer therapy. Simulations also showed that the assignment of equal windows to groups with different durations of drug use can bias risk comparisons, either away from the null or towards the null. We concluded that the choice of prescription time-windows can influence the estimates of exposure risks. Time-windows should cover the period with potential excess risk and be validated.