Genetic Risk Prediction — Are We There Yet?

(from Peter Kraft, Ph.D., and David J. Hunter, M.B., B.S., Sc.D., M.P.H. - New England Medical Journal)

a major goal of the Human Genome Project was to facilitate the identification of inherited genetic variants that increase or decrease the risk of complex diseases. The completion of the International HapMap Project and the development of new methods for genotyping individual DNA samples at 500,000 or more loci have led to a wave of discoveries through genomewide association studies. These analyses have identified common genetic variants that are associated with the risk of more than 40 diseases and human phenotypes. Several companies have begun offering direct-to-consumer testing that uses the same single-nucleotide polymorphism chips that are used in genomewide association studies. These companies claim that such testing should be made available to consumers who are interested in their personal level of risk for the relevant diseases. Now, “risk tests” for specific diseases such as breast cancer are also being marketed to physicians and consumers. 

The availability of highly predictive and reasonably affordable tests of genetic predisposition to important diseases would have major clinical, social, and economic ramifications. But the great majority of the newly identified riskmarker alleles confer very small relative risks, ranging from 1.1 to 1.5, even though such analyses meet stringent statistical criteria (i.e., the identification of associations with disease that have very small P values and hence are unlikely to be false positives). However, even when alleles that are associated with a modest increase in risk are combined, they generally have low discriminatory and predictive ability. 

One argument in favor of using the available genetic predictors is that some information must be better than no information, and we should not let the perfect be the enemy of the good by refusing to make use of our knowledge until it is more complete. Why not begin testing for common genetic variants whose associations with susceptibility to disease have been established? The answer lies in the stability of the current risk estimates. 

The good news is that pooling the results of multiple genomewide association studies has led to increased statistical power and the discovery of many new loci linked to small increases in the risks of major diseases and phenotypes. For example, pooling efforts have led to the identification of more than 16 new loci associated with diabetes and more than 30 loci linked to Crohn’s disease. From a scientific perspective, we would like to know roughly how many risk loci remain to be discovered. From a clinical and policy perspective, we would like to know the extent to which the available associations are useful for measuring risk, both in absolute terms and relative to the expanded set of associations that are likely to be discovered in the next few years. 

Although we know of many more risk loci than we did 2 years ago, there are probably many more associations that are yet to be discovered for these complex diseases. Less common variants in the prevalence range of 0.5 to 5.0% also remain to be discovered. Estimates of risk based on established locus associations are therefore likely to change substantially in the next few years. But this is only one of the factors determining whether knowledge of genetic risk is beneficial. The clinical value of a genetic test also depends on its sensitivity, specificity, and positive and negative predictive values; the costs and benefits of interventions; and the availability of data linking specific variants to improved clinical outcomes. In particular, although we will be better able to distinguish subtle differences in risk as we discover more risk loci, most people will still be at or near the median level of risk. As a result, for less-common diseases (with a prevalence of 1% or less), the positive predictive value of a genetic test will almost always be low. 

We are still too early in the cycle of discovery for most tests that are based on newly discovered associations to provide stable estimates of genetic risk for many diseases. Although the major findings are highly unlikely to be false positives, the identified variants do not contribute more than a small fraction of the inherited predisposition. Once risk estimates are more stable, the usefulness of genetic screening will need to be considered for each disease, and recommendations about potential interventions will need to be made for persons whose predicted risk exceeds some threshold. Appropriate guidelines are urgently needed to help physicians advise patients who are considering this form of genetic testing as to how to interpret, and when to act on, the results as they become more stable.

(sent by M Cesco Gaspere)