Multiple genetic tests for susceptibility to smoking do not outperform simple family history

Abstract

Aims: To evaluate the utility of using predictive genetic screening of the population for susceptibility to smoking.

Methods: The results of meta-analyses of genetic association studies of smoking behaviour were used to create simulated data sets using Monte Carlo methods. The ability of the genetic tests to screen for smoking was assessed using receiver operator characteristic curve analysis. The result was compared to prediction using simple family history information. To identify the circumstances in which predictive genetic testing would potentially justify screening we simulated tests using larger numbers of alleles (10, 15 and 20) that varied in prevalence from 10 to 50% and in strength of association [relative risks (RRs) of 1.2-2.1].

Results: A test based on the RRs and prevalence of five susceptibility alleles derived from meta-analyses of genetic association studies of smoking performed similarly to chance and no better than the prediction based on simple family history. Increasing the number of alleles from five to 20 improved the predictive ability of genetic screening only modestly when using genes with the effect sizes reported to date.

Conclusions: This panel of genetic tests would be unsuitable for population screening. This situation is unlikely to be improved upon by screening based on more genetic tests. Given the similarity with associations found for other polygenic conditions, our results also suggest that using multiple genes to screen the general population for genetic susceptibility to polygenic disorders will be of limited utility.