Robust methods for detecting familial aggregation of a quantitative trait in matched case–control family studies

Abstract

Assessing familial aggregation of a disease or its underlying quantitative traits is often undertaken as the first step in the investigation of possible genetic causes. When some major confounding variables are known and difficult to be quantified, the matched case–control family design provides an opportunity to eliminate biased results. In such a design, cases and matched controls are ascertained first, with subsequent recruitment of other members in their families. For the study of complex diseases, many continuously distributed quantitative traits or biomedical evaluations are of primary clinical and health significance, and distributions of these continuous outcomes are frequently skewed or non-normal. A non-normal distributed outcome may lead some standard statistical methods to suffer from loss of substantial power. To deal with the problem, in this study, we thus propose a rank-based test for detecting familial aggregation of a quantitative trait with the use of a within-cluster resampling process. According to our simulation studies, the proposed test expresses qualified and robust power performance. Specifically, the proposed test is slightly less powerful than the generalized estimating equations approach if the trait is normally distributed, and it is apparently more powerful if the trait distribution is essentially skewed or heavy-tailed. A user-friendly R-script and an executable file to perform the proposed test are available online to allow its implementation on ordinary research.

Keywords:

• case–control family studies
• familial aggregation
• matched design
• proband sampling
• within-cluster resampling

Acknowledgements

The authors would like to thank two anonymous referees and the Editor for their helpful comments that improved the manuscript. This research was supported by NSC grant 99-2118-M-468-001, Taiwan (JYW).

Web resource

An executable file and R-script designed for detecting familial aggregation of a quantitative trait with 1:m _i matched case–control family data are available at http://dns2.asia.edu.tw/ jjwang/program.htm. The numbers of control families within strata and the numbers of relatives in different families are allowed to be varied. An artificial example data set and brief manuals can also be found in the website.