Abstract:
|
Nested case-control studies are often used in large epidemiology studies as it provides us an efficient means to study associations between a risk factor and a disease outcome at greatly reduced cost. A small subset of controls are sampled from the prospective full cohort for each case. The information for the risk factor (like a biomarker) is only needed to be collected on a much smaller cohort including the selected controls and the cases. However, choice of control sampling strategies can affect study power and possibly even bias risk estimates, and hence needs to be considered carefully. The well-characterized Women's Healthy Eating and Living (WHEL) cohort of over 3000 breast cancer survivors provides us a unique opportunity to investigate the impact of a variety of control sampling strategies on inference in nested case-control studies. In this paper, we conducted an in depth study to evaluate three case-control matching schemes. The hazard ratio estimates for associations between biomarkers of interest and breast cancer death will be estimated based on the nested case control studies and compared with results from the full cohort models, which are deemed to be the gold standard for this analysis. The analytic approach for estimating absolute risk in the nested case-control setting will also be described. Especially, we will describe the methods for the situation of the survival study with left truncation. The results from simulation biomarkers and a real biomarker (e.g. CRP) assayed on the WHEL cohort will be presented.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.