Invited Commentary: Birth Certificates--A Best Control Scenario?

The majority of conditions that affect children under the age of 15 years (e.g., birth defects, diabetes, cancer) are sufficiently rare to warrant the case-control study approach. Furthermore, most pediatric studies require enrollment from multiple centers, registries, or hospitals to gain a sufficient sample size and thus can span large geographic areas. While selection of cases is usually straightforward, special care must be given to selection of controls to ensure that they represent the underlying population from which the cases arose. As Wacholder et al. (1–3) nicely described in a 1992 series of articles addressing control selection issues, an underlying cohort or study base can be primary or secondary. The primary base is the hallmark of a population-based casecontrol study in which all cases in a defined population are ascertained both geographically and temporally. When it is difficult to ascertain all cases in an area, a secondary base is chosen. In this instance, the base is defined as the cases, and the controls are identified as persons who would have become study cases had they developed the disease during the time period. For example, if one were to select all children diagnosed with cancer from a particular hospital, the proper control group would be those children who would have gone to that hospital had they developed cancer. The difficulty of defining this group will vary with the complexity of hospital referral patterns, which is especially true for childhood cancer because individual hospitals may have expertise related to specific diagnoses. In this commentary, we briefly examine the potential control group options for studies of childhood cancer etiology.