Making sense of changing risk predictions from personal genomics

Mihaescu, R., van Hoek, M., Sijbrands, E., Uitterlinden, A., Witteman, J., Hofman, A., van Duijn, C., & Janssens, A. (2009). Evaluation of risk prediction updates from commercial genome-wide scans Genetics in Medicine, 11 (8), 588-594 DOI: 10.1097/GIM.0b013e3181b13a4f


Caroline Wright from the Public Health Genomics Foundation has a concise post describing the results from a recent paper in Genetic Medicine. The paper evaluates the probability that personal genomics customers will find that their predicted risk of a common disease changes significantly over time as their genetic data are updated, using data on known type 2 diabetes risk variants as a case study. 

As you might expect, when hypothetical customers started with information from only one risk variant and then had information from other risk variants and from non-genetic predictors (age, sex, body mass index) added in, their predicted risk often changed from above- to below-average and vice versa. A similar result was predicted by Peter Kraft and David Hunter in New England Journal of Medicine perspective piece they wrote back in April.
You can spin this as a negative - by arguing that such shifting risk predictions will confuse customers and undermine the public's view of the reliability of genetic risk prediction - but really such a result is a logical consequence of slow improvements to any new and imperfect risk prediction system. 
Personal genomics customers should of course treat their current risk predictions with caution - for most diseases, current risk predictions are barely better than noise - and bear in mind that their profiles will alter over the next few years as increases in scientific understanding of the genetic basis of disease improves, and as their predictions become steadily more accurate.
Wright provides a balanced review of the implications of the article, and finishes with a paragraph worth quoting verbatim:
However, far from supporting calls to forbid such tests being available DTC, this highlights the need for transparency in the provision of information.Companies offering genome-wide risk prediction services should ensure that their customers understand that, whilst the measurement of the DNA sequenceitself (the assay) will remain constant, the interpretation of the result (the test) is likely to change as the science develops.
Amen. As I have consistently argued here on Genetic Future, the solution to the alleged dangers of personal genomics is not banning the public from having access to their own genome (as Germany has recently done), or forcing them to seek permission from their doctor before peering into their own DNA. Instead, the answer is to ensure that companies provide at least the minimal information required for customers to make an informed decision about the utility of risk predictions, and to punish those companies who fail to do so with public censure and (if absolutely necessary) regulatory action.

More like this

Fan Liu, Kate van Duijn, Johannes R. Vingerling, Albert Hofman, André G. Uitterlinden, A. Cecile J.W. Janssens, Manfred Kayser (2009). Eye color and the prediction of complex phenotypes from genotypes Current Biology, 19 (5) DOI: 10.1016/j.cub.2009.01.027 In a recent post I noted that genetic…
Yurii S Aulchenko, Maksim V Struchalin, Nadezhda M Belonogova, Tatiana I Axenovich, Michael N Weedon, Albert Hofman, Andre G Uitterlinden, Manfred Kayser, Ben A Oostra, Cornelia M van Duijn, A Cecile J W Janssens, Pavel M Borodin (2009). Predicting human height by Victorian and genomic methods…
Well, it's a little late, but I finally have a list of what I see as some of the major trends that will play out in the human genomics field in 2009 - both in terms of research outcomes, and shifts in the rapidly-evolving consumer genomics industry. For genetics-savvy readers a lot of these…
The latest issue of the New England Journal of Medicine has four excellent and thought-provoking articles on the recent revolution in the genetics of common disease and its implications for personalised medicine and personal genomics. Razib and Misha Angrist have already commented, and there's…

We need to decide here. Are these tools for medicine or not?

Today I say they are not great tools. Maybe in 30 years when we have further data. My point is that we should not be hinting or inferring that they do (Currently) have great predictive power in risk prediction. That is one of the problems.

The second problem is that these Non-clinical tools put the consumer/patient at risk of disclosure of information which could be (IN the future) very predictive.

The third problem is that these companies make current genetic testing tools seem less valid (In the eyes of many pragmatic doctors) This leads to an increased skepticism of what is a good clinical tool, BRCA/HNPCC etc.....

Regulation is one way to say "This is a good Tool, this is not"

A physician's mind works just the same when it comes to regulation......they say "This is a good tool, this is not" merely based on the regulation of these tools, which, let's face it, there is little if any being had here or even being created.....

Gatekeepers only work, if the gatekeeper knows enough about the field to make informed choices.

We all saw a terrible case of this going awry with Michael Jackson's Propofol induced death.......

So this is not to place physicians as the end all be all, but until these companies are willing to state that their tests have limited ability NOW, then I am unwilling to give them a pass.......

Sure, let them play up what "Might be" But, playing up what currently isn't is dishonest.

-Steve

In addition, if you have given us permission to share your Personal Information by clicking on the appropriate box or button on the Site, we may occasionally make your Personal Information available to our parent or other affiliated companies

McCulloch accuses Steig et al. of appropriating his âfindingâ that Steig et al. did not account for autocorrelation when calculating the significance of trends. While the published version of the paper didnât include such a correction, it is obvious that the authors were aware of the need to do so, since in the text of the paper it is stated that this correction was made. The corrected calculations were done using well-known methods, the details of which are available in myriad statistics textbooks and journal articles. There can therefore be no claim on Dr. McCullochâs part of any originality either for the idea of making such a correction, nor for the methods for doing so, all of which were discussed in the original paper. Had Dr. McCulloch been the first person to make Steig et al. aware of the error in the paper, or had he written directly to Nature at any time prior to the submission of the Corrigendum, it would have been appropriate to acknowledge him and the authors would have been happy to do so. Lest there be any confusion about this, we note that, as discussed in the Corrigendum, the error has no impact on the main conclusions in the paper.