Testing common genetic variants for breast cancer risk is not clinically useful

Testing common genetic variants for breast cancer risk is not clinically useful
22 March 2010 | By Dr Caroline Wright | Research article

The recent harvest of common genetic susceptibility variants uncovered by genome-wide association studies has raised hopes of personalised medicine and the use of individual genetic risk prediction to prevent disease. However, a mounting number of studies have shown that an individual’s genetic risk based on common single nucleotide polymorphisms (SNPs) does not substantially improve risk prediction based on traditional (non-genetic) risk factors (see previous news cardiovascular disease and diabetes).

A new study evaluating the utility of using common genetic variants in breast cancer risk models, published in the New England Journal of Medicine, has come to much the same conclusion [Wacholder S. et al. NEJM (2010) 362:986-93]. Combining four prospective cohort studies and a case-control study allowed nearly 6,000 cases and 6,000 controls to be followed for up to 15 years. Each individual’s risk of breast cancer was retrospectively calculated based on ten SNPs and four clinical risk factors which are part of the standard Gail model (number of first degree relatives with breast cancer, age at menarche, age at first live birth and number of previous breast biopsies), all of which individually have fairly small relative risks. Numerous risk models were constructed and compared in terms of their ability to discriminate between those who developed breast cancer from those who didn’t, using the area under the ROC curve (AUC, which varies from 0.5-1.0, where a value of 1.0 indicates a perfect prediction). The purely genetic model (AUC=0.59) performed as well as the Gail model (AUC=0.58), but the combined risk model had the best performance with an AUC of 0.62. Although this represents a modest improvement in performance, the incremental benefit is unlikely to justify the increased cost of genotyping for individuals. In addition, almost half the cases ended up in the same quintile of risk irrespective of whether the model included genetic variants. The authors therefore concluded that “even with the addition of these common variants, breast-cancer risk models are not yet able to identify women at a reduced or elevated risk in a clinically useful way.”

Comment: This conclusion may be somewhat pessimistic, as highlighted in the accompanying editorial [Devilee P & Rookus M.A. NEJM (2010) 362:1043-5]. Although an AUC of around 0.6 is certainly too low for accurate clinical prediction, the authors have “implicitly dismissed” the pure Gail model for clinical risk prediction, because it performs similarly well (or as poorly) as the purely genetic model. Importantly, however, these models differ substantially in terms of their feasibility for implementation; the ‘assay’ required for the Gail model – a simple questionnaire – is substantially cheaper, easier and faster than genotyping.

Although this study adds to the mounting evidence that common genetic variants are unlikely to be useful for risk prediction at the individual level, there are still plenty of reasons for optimism. Firstly, these common genetic risk variants reveal important information about the underlying biology of complex diseases, which may ultimately lead to better treatments. Secondly, genotyping these common risk variants could still be useful at a population level for targeting preventative interventions such as screening (see previous news). And finally, future studies using deep or whole genome sequencing methods may uncover more susceptibility variants, including rare variants with a substantially higher relative risk, that may improve risk prediction. For the moment, however, there is still very little evidence that personalised risk prediction – such as that offered to consumers by a wave of new companies (see previous news) – is worth the extra cost and logistical challenge of implementing genotyping in clinical practice.
http://www.phgfoundation.org/news/5309/


PHG Foundation is the trading name of the Foundation for Genomics and Population Health, a charitable company registered in England and Wales
Company Number: 5823194, Charity Number: 1118664