Bias resulting from using the card-back system to contact patients in an epidemiologic study

In a case-control study of thyroid cancer in women in western Washington, the card-back system was used to contact patients. Demographic characteristics and reproductive history of 124 cases who returned a card and 35 who did not were compared, and were found to differ with respect to various reproductive and demographic factors. Relative risk estimates for various thyroid cancer risk factors were calculated two ways: including as cases only those who returned a card, and including all cases; these estimates differed for some risk factors. The results suggest that the use of the card-back system may result in a low level of response which can lead to biased study results.