PHG Foundation | Uncertainty in disease risk estimates

Uncertainty in disease risk estimates
11 December 2009 | By Dr Caroline Wright | Research article

It has been widely suggested that findings from genome-wide association studies could be combined with population data in order to produce estimates of absolute risk for numerous common complex diseases (see previous news). This information could either be used to stratify populations into risk categories in order to better target preventative interventions such as screening (see previous news), or to estimate an individual’s risk of particular diseases, as is currently offered by numerous direct-to-consumer (DTC) genomics companies (see previous news).

Whilst it is generally acknowledged that these risk estimates have a large uncertainty associated with them, the size of that uncertainty is rarely well communicated. This problem has recently been addressed in two different ways. The first offers a quantitative approach to the problem, by attempting to quantify the size of the uncertainty associated with risk prediction using breast cancer as an example [Yang Q. et al. Am J Hum Genet (2009) 85:786-800]. Researchers found more than a 3-fold difference in the risk estimates using just a single SNP (in the FGFR2 gene), and a range of 6-21% for the absolute lifetime risk of breast cancer in women carrying five high-risk SNPs. The major contributor to this uncertainty was differences in the underlying population incidence rates of the disease; smaller contributions are also made by varying genotype frequencies, different estimates of individual effect sizes and alternative assumptions about interaction between genes.

A rather different approach to this problem has been taken by one of the latest entrants to the DTC market, Pathway Genomics. Instead of providing customers with figures for their absolute or relative risks, the company communicates risk by using a series of colours depending on the potential significance of the result; for example, green denotes below average risk, beige indicates average risk, and yellow through red signify various levels of above average risk (see review at Bio-IT World). Whilst this may be somewhat disappointing to some consumers, who are now accustomed to receiving quantitative relative and absolute risk estimates from many of the other providers, it has the advantage of explicitly acknowledging the inherent uncertainty of genomic risk estimates, and may also go some way to addressing the problem of frequently changing risk profiles as more associations are found (see previous news).

Comment: The large uncertainty associated with these risk estimates will come as no surprise to epidemiologists, who are accurately aware of numerous sources of error and bias in all health statistics. Nonetheless, it is likely that many people could be mislead by providing a single numerical risk estimate, with no indication of confidence in that value. It will therefore be interesting to see whether other risk prediction models (including those offered by other DTC genomics companies) will adopt one of these strategies to better communicate the inherent uncertainty in their predictions.
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
http://www.phgfoundation.org/pages/contact.htm

abrir aquí:
PHG Foundation | Uncertainty in disease risk estimates