- Ensure that companies provide high-quality genetic data to consumers, along with simple and accurate descriptions of the predictive value of the tests performed;
- Ensure that information about testing companies is placed in public databases that allow consumers to make a thoroughly informed choice between genetic testing providers; and
- Not prevent customers from gaining free access to any part of their own genetic information if they want it, or position clinicians as the absolute gate-keepers to such information.
Professor Peter Donnelly, who led the massive Wellcome Trust Case Control Consortium study of the genetics of common diseases, was more unambiguous in his support for the DTC industry:
Donnelly ... spoke positively about DCTs and suggested that they might be the best way to ensure that technology develops to a point where it becomes useful for public healthcare. They would also be beneficial for the small number of individuals who had a high risk of developing a disease due to the additive effects of having several low risk gene markers. He thought that DCTs were the first step to a service that would eventually be incorporated into routine clinical practice.
While most individuals would have an average risk for most diseases almost
everyone would be at very high risk for some diseases. Professor Donnelly
estimated that 95 percent of people would be in the top five percent of genetic risk
for at least one disease, 40 percent of people would be in the top one percent of
genetic risk for at least one disease and five percent of people would be in the top
0.1 percent of genetic risk for at least one disease.
[...] It may therefore be useful to think of genomic tests, including
those sold "direct to consumers", as a tool for individuals to identify the diseases
for which they had the highest genetic risk, based on current knowledge.
- The development of a voluntary code of practice for DTC genetic testing companies, including a requirement for companies to place information about their laboratory accreditation and testing standards in the public domain, and guidance regarding the provision of pre- and post-test counselling;
- The creation of an online registry hosted by the Department of Health that lists this information about DTC genetic testing companies, and also describes "the extent to which the DNA sequence variants used ... have been validated in genome-wide association studies, and shown in prospective trials to have utility for predictive genetic testing."
The first Times article you cite, states (with my emphasis):
In the past few years, the costs of reading DNA have fallen so sharply that many scientists predict it will be possible to sequence any individualâs entire genetic code for less than Â£1,000 within a year or two. Research has also revealed hundreds of genetic variations that affect peopleâs risk of disease, or their response to medicines. The committee, chaired by Lord Patel, said that while it would be several years before this information allows the accurate prediction of disease, several medical applications are already practical or not far away.
While the examples given (dividing type 2 diabetes into sub-diseases, and measuring the known warfarin dose response SNP) are sensible, it is not clear that "accurate prediction of disease" is possible, let alone achievable.
For example, two recent papers, using rather similar data, come to completely opposing conclusions:
Clayton DG (2009) Prediction and Interaction in Complex Disease Genetics: Experience in Type 1 Diabetes. PLoS Genet 5(7): e1000540 concludes that, for most diseases, the discovery of all disease susceptability loci would not give great predictive power.
Lu Q, Obuchowski N, Won S, Zhu X, Elston RC (2009) Using the Optimal Robust Receiver Operating Characteristic (ROC) Curve for Predictive Genetic Tests. Biometrics. 2009 Jun 8 on the other hand conclude that:
With the discovery of more disease risk variants, and eventually their interactions, we might be able to form a predictive genetic test [for type 1 diabetes] that could be implemented in clinical use.
I know which side I'm on - discovering you have twice the risk of something rare is not very interesting: see rule 12 of the excellent A Worrier's Guide to Risk, and an interpretation of an 18-fold increase in risk caused by the common rs9272346 AA genotype from SNPedia's entry for anonymous caucasian NA07022, picked up from an earlier post - but you could also check out the NEJM Perspective piece :
Great post Daniel -- thanks also for linking.
With respect to the comments of our Executive Director Dr Ron Zimmern, I should clarify that the key point is that there is no objection to companies selling "totally useless information" in the form of pointless or unvalidated genetic susceptibility tests - PROVIDED that consumers are able to easily access information that reveals test performance and clinical meaning (if any) of such tests.
Our evidence given to the House of Lords inquiry therefore called for evidence relating to test performance (or lack of it) to be transparent and publicly available, whilst proposing that "funders of health services, and clinicians, should be discouraged from using tests that are not backed by appropriate clinical evidence" (see http://www.phgfoundation.org/news/4705/).
Without really thinking about this, I had assumed Donnelly's calculations were the fruit of intensive analysis of the rich data source that is the WTCCC:
While most individuals would have an average risk for most diseases almost everyone would be at very high risk for some diseases. Professor Donnelly estimated that 95 percent of people would be in the top five percent of genetic risk for at least one disease, 40 percent of people would be in the top one percent of genetic risk for at least one disease and five percent of people would be in the top 0.1 percent of genetic risk for at least one disease.
[...] It may therefore be useful to think of genomic tests, including those sold "direct to consumers", as a tool for individuals to identify the diseases for which they had the highest genetic risk, based on current knowledge.
Not a bit of it - a colleague points out that it is simple probability - in a distribution, someone must be at the top, someone at the bottom, and "if genetic risks were independent and we consider 50 diseases, the percent of people in the top 5% in at least one disease would be
100 * (1 - 0.95^50)
which is 92.3%. For the top 1% we get 39.5% and for the top 0.1% we get 4.9%."
Which makes me think the second paragraph quoted is not a logical conclusion of the first, but a fairly nuanced position statement - that the predictability of genetic tests may be becoming good enough to be worth knowing about.
However, not everyone present bought into this: the Discussion that followed Donnelly's presentation (p107) begins:
One participant gave the view that discussions on clinical management run the risk of being hijacked by genetic enthusiasts who overemphasised the importance of genetic factors and genetic testing. Other conventional and possibly more important factors in disease aetiology and management, such as lifestyle, social factors and family history could therefore be disregarded.
That Donnelly's position is nuanced is shown by his second conclusion, which can be seen as a refutation of Steve Jones' challenge as to the purpose of GWAS - although the session being quoted was held on 19th March 2008, about a year before the Steve Jones story kicked off:
while there was ongoing uncertainty regarding the clinical utility of genomic tests for disease prediction, the most important outcome of this new research may be in advancing understanding of the molecular causes of disease development, which was already providing new leads for ways in which to prevent and treat common diseases.
Great points, and very useful links - I'd strongly second the Kraft paper as an excellent and worthwhile read.
There's definitely a danger that the predictive value of genetic tests will be over-hyped, and I think we can all agree that integrating these tests with both family history and environmental risk factors will be essential for generating meaningful predictions. Telling someone that their SNPs suggest a lower-than-average risk of lung cancer is rather pointless if the person is also a heavy smoker with a strong family history of cancer. Obviously this makes risk predictions more difficult, though.
As for Donnelly's pre-emption of Jones' critique (which was really just an unattributed echo of David Goldstein's NEJM opinion piece); I suspect GWAS researchers started carefully honing this defense as soon as the WTCCC results made it clear that GWAS would find lots of risk variants, but provide fairly trivial clinical predictive power. The "but we found new pathways!" defense does have some genuine truth to it, but it's only really employed because the "but we can now predict disease risk!" defense obviously doesn't hold any water...