Methodology for Case-Control Studies of Breast Cancer

摘要

This project investigated whether the use of random digit dialing (RDD) samplingfor obtaining a control group in a breast cancer case-control study incurred any substantial biases, compared to the use of area probability sampling. Data from a case-control study on metropolitan Atlanta women aged 20-54 years were analyzed, with two independently obtained control groups (area and RDD). The two sample control groups both agreed with sample Census data in making inference to the larger population. In unweighted analyses the two control groups differed somewhat on race, with a larger percentage of Blacks in the area sample. However, after adjustment for age and race, the two control groups were remarkably similar on 41 variables related to breast cancer risk and general health status. Two notable features of RDD sampling, compared to area sampling, were a significantly smaller enumeration or screening response rate and a significant under-identification of households which contained any women aged 20-54 years. Although this study provides no compelling evidence for caution regarding RDD sampling, an awareness of its generally higher nonresponse at all stages of public contact is recommended.