The common disease-common variant (CDCV) hypothesis predicts that common disease-causing alleles, or variants, exist in all human populations that contribute to common diseases. Although each variant will necessarily only have a small effect on disease susceptibility (i.e. a low associated relative risk), numerous researchers have suggested that useful diagnostic or predictive tests may nonetheless be possible by combining multiple variants. To this end, the cumulative effect of common risk variants has been demonstrated for prostate cancer and cardiovascular disease (see previous news), and a scheme for improved targeting of breast cancer screening using common genetic markers has been outlined (see previous news).
More recently, researchers have investigated combining 18 common single nucleotide polymorphisms (SNPs) to predict the risk of type 2 diabetes [Meigs JB et al. (2008) NEJM 359(21):2208]. The study examined the extent to which a genotype score (i.e. the total number of risk variants possessed by an individual) could discriminate the risk of diabetes when used either alone or in addition to other clinical risk factors. The clinical performance was measured using a ‘c-statistic’, which represents the discriminatory accuracy of a test, i.e. how well it can distinguish between individuals that will get the disease from those who will not; this value varies from 0.5 (i.e. a useless test, equivalent to guessing randomly) to 1.0 (i.e. a perfect test, where everyone is correctly categorised).
There are two very different ways to represent the results of this study. The first is in a fairly positive light. There was a 12% increase in the risk of diabetes per risk allele, resulting in a substantial overall relative risk of 2.6 between people with the highest and lowest genotype scores. Adjusted only for sex, the genotype score produced a c-statistic of 0.581, a statistically significant improvement over a value of just 0.534 without the genetic information, resulting in improved discrimination and reclassification of 4.1% of people primarily from low risk categories into higher-risk ones. Given the increasingly high prevalence of type 2 diabetes (around 4% according to Diabetes UK), as well as the availability of highly effective, safe and well validated preventative measures, this would appear to be potentially a very valuable test.
However, the second representation is rather more discouraging. When family history or clinical risk factors were also used, all of which are frequently documented during adulthood, the genotype score did not significantly improve discrimination. The best validated model – based on sex, family history, age, body-mass index, fasting plasma glucose level, systolic blood pressure, high-density lipoprotein cholesterol level and triglyceride level – already offers a high discrimination power of 0.900, and the addition of the genotype score only resulted in a small and non-statistically significant increase of the c-statistic to 0.901. Here, the addition of genetic information to standard risk factors is essentially clinical useless.
Comment: This study highlights the importance examining the incremental benefits of genetic susceptibility testing within a particular clinical scenario – what might at first look like a very useful test on its own may in fact offer limited or no value when combined with traditional risk factors. Consideration of genetic tests within a specific clinical context is therefore crucial to realistic evaluation.
However, there is hope on the horizon. Firstly, more genetic susceptibility variants will doubtless be discovered, which may ultimately improve the discriminatory power of the genotype score. Secondly, it is currently unclear whether the ability to test people very early, before clinical risk factors have appeared, could be extremely useful for targeting screening and prevention programmes. Finally, even if ultimately genetic testing does prove to be ineffective for predicting or diagnosing disease, the discovery and validation of the genetic basis of disease will undoubtedly improve our understanding of disease aetiology, and may ultimately lead to the development of novel treatments.