Back to Annual Meeting

Back to Annual Meeting

APHA Scientific Session and Event Listing

Anne Jurek¹, George Maldonado², Kirsten Ness¹, Wendy Leisenring³, and Ann Mertens¹. (1) Pediatrics, University of Minnesota, MMC 715, 420 Delaware Street SE, Minneapolis, MN 55455, 612-624-1639, jurek@epi.umn.edu, (2) Environmental Health Sciences, University of Minnesota School of Public Health, MMC 807, 420 Delaware Street SE, Minneapolis, MN 55455, (3) Clinical Statistics, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. N., D5-360, PO Box 19024, Seattle, WA 98109

When designing an epidemiologic study, it is important to choose a comparison group (CG) (i.e., control group) that is a good substitute for the counterfactual experience of the target (Maldonado & Greenland, Int J Epidemiol 2002;31:422-9). Presently, a large cohort study of childhood cancer survivors uses siblings as the CG. Study reviewers wondered how the siblings and general US population differed. This raised the question, “Are siblings the best choice for a CG in this cohort?” This led to the more general question, “What characteristics of a CG make the best substitute for the counterfactual experience?” We borrow a standard stratum-specific assumption from stratified analysis, and assume that a good CG is one which has an unstratified joint distribution of important risk factors that mirrors that of the target. We developed a simple method, using this assumption, to choose among potential CGs and illustrate it with health outcome data from the cohort study. Three CGs were considered: siblings of cancer survivors, participants in the Behavioral Risk Factor Surveillance System (BRFSS), and participants in the National Health Interview Survey (NHIS). Using lung fibrosis as an example, the cancer-survivor siblings had the joint smoking-gender distribution that best approximated that of the survivors. On average over the 6 joint-distribution strata (3 smoking by 2 gender categories), the siblings differed from the survivors by 2.44%, compared to 6.68% for BRFSS participants and 6.80% for NHIS participants. These results show that cancer-survivor siblings are the best CG for the outcome lung fibrosis, given the assumptions.

Learning Objectives:

• Gain an awareness of applying counterfactual theory to the practice of epidemiology.
• Describe an approach to selecting a comparison group for an epidemiologic study.
• Understand how the analysis of each health outcome is affected by the selection of a comparison group.

Keywords: Epidemiology, Methodology

Presenting author's disclosure statement:

Any relevant financial relationships? No