Editors' ChoiceStatistical Analysis of Genetic Data

Making the Most of Genetic Data from Association Studies

See allHide authors and affiliations

Science Translational Medicine  10 Mar 2010:
Vol. 2, Issue 22, pp. 22ec39
DOI: 10.1126/scitranslmed.3001025

In studies of genetic and environmental correlates of disease, data are collected either from many families with some affected members or from large samples of individuals with disease and unrelated controls without disease. Until now, most studies collected only one type of data. However, given the evolution of genetic studies over the past 15 years, many large databases—covering conditions from diabetes to colorectal cancer—now include data from both families and unrelated cases and controls. The most efficient use of the existing information involves combining the data from the two different types of studies, which is a statistically challenging task.

Recently, Zheng et al. proposed a new statistical approach to combining such data and showed that by using this method, one is more likely to detect associations when they exist. This is because the estimates of association found by using their methodology are both less biased and have smaller variance as compared with the associations found by other methods. Their approach has the further advantage of providing a measure of association from a family study that can easily be combined with the measure of association from a case-control study. Using this new approach, they identified associations between smoking, body mass index, and colorectal cancer using the Colorectal Cancer Family Registry, a resource created by an international consortium that contains both family- and case-unrelated control data. These associations were statistically insignificant when each type of data was studied separately. Thus, analyzing all available data sets has the potential to identify new genetic and environmental associations with disease.

Y. Zheng et al., On combining family-based and population-based case-control data in association studies. Biometrics 16 February 2010 (doi: 10.1111/j.1541-0420.2010.01393.x). [Abstract]

Navigate This Article