Selection of control groups by using a commercial database and random digit dialing

Identifying a control group when cases come from a specialized hospital is a challenge for epidemiologists. The authors compared controls recruited by using a commercial database with those recruited by random digit dialing in the context of a hospital-based case-control study of ovarian cancer. This part of the study was conducted in 1997-1998 among women aged 18 years or older who resided in the New York metropolitan area. A mailing list owner grouped cases into "lifestyle" clusters based on US zip+4 postal code microneighborhoods and generated a random sample of potential controls with the same distribution across the clusters. Controls recruited from the commercial database (n = 82) and from random digit dialing (n = 90) were similar in age and race. Women from the commercial database had somewhat more education and higher incomes and were more similar to the cases on these measures. The control groups resembled each other closely in terms of oral contraceptive use, nulliparity, and religion and differed from the cases on these measures. Response rates were similar for the two groups. Only 28% of the cases were included on the mailing list, indicating that it did not reflect the source population of the cases. Use of a commercial database provided a control group whose socioeconomic factors were similar to those of cases at a lower cost than when random digit dialing was used but did not result in a higher response rate.