May 18, 2012

How useful is it to sequence everybody?

Like everyone else, public health insurances only have a limited amount of money, and therefore constantly need to make decisions which treatments to fund, and which are too expensive. That way, some treatments, although they have a health benefit, do not get funded, because they cost too much, and the money instead goes to treatments with a higher utility.

This applies to drugs, but also to diagnostic tests. From the point of view of insurers, genome sequencing is just another such test. The question then is, is it the best way to spend health insurance money?

A new door may be a better way to spend your money

In a recent publication in the journal Science Translational Medicine, Nicholas Roberts and coworkers partially address this question. They estimate the potential of whole genome sequencing to provide clinically useful information on the disease risk of people who are currently healthy.

Their result: Even in the best of circumstances, for most people sequencing will be uninformative about their risk for the majority of common diseases.

However, they also find that there is also a good chance that sequencing will provide useful information about at least one disease. This means that if you are particularly concerned about getting a specific disease, sequencing will probably not tell you anything new. On the other hand, it might well warn you about your risk of developing a disease you never even thought of before. But even then, the predictive value is likely to be quite limited.

This is interesting, but as the authors acknowledge, it comes with some caveats. Firstly, it covers only common diseases. For diseases that are caused by single genes, sequencing is much more likely to give useful results. Also, it assumes that people are sequenced after their birth, whilst some of the value of sequencing lies with prenatal diagnosis. It also ignores the possibility of sequencing only those people who are most likely to benefit, such as those that already show certain symptoms.

In summary, Nicholas Roberts and his colleagues pour cold water on the idea that sequencing everyone is going to have a large impact on how disease risks for common diseases are assessed for asymptomatic patients. On the more important question of whether sequencing everyone will be worth the cost in clinical practice, the jury is still out.

In next week's post, I will attempt to answer the related question of what the utility of sequencing needs to be in order to offset its cost.


  1. I hope in next week's column you will explain the value to any specific individual in being sequencing. Because without individual benefit, the case for populations screening is pointless.

    Unless a person is a known potential carrier of a genetic disease or is part of a trial to test the hypothesis that a disease (and treatment) differentiates by genomic mutation (BRAC), where is the evidence that we know enough about the relationships of genes to justify individual mapping? If, as seems likely, depression is a result of multiple genomic factors PLUS environmental triggers, what do we do differently? And who is to say that genomics are the building blocks--this is a reductionist approach that presumes a lot of knowledge that is accepted but largely unproven?

    I have explored a number of these issues in the context of drug discovery and biological complexity at

    1. Hi Steven, thank you for your comment and the link to your very relevant blog.

      The paper by Roberts et al. looks into the theoretically possible maximum utility of whole genome sequencing, considering the proportion of disease risk that is of genetic origin. They do that by looking at identical twins. As you say, most likely the actual utility of sequencing for individual patients will be lower.

  2. Two variables in the equation are constantly changing. The cost variable continually decreases. The predictive value variable continues to increase as more work is done. The cost-benefit equation will need to be recalculated periodically.

    1. Hi Goeff,

      Thank you for your comment, which describes the problem more clearly than I have done.

      In my post next week I'll look into what you call the cost variable in more detail.

      As the predictive value variable, Roberts et al. estimate the maximum value that this variable can assume in the future, considering what we know about disease heritability.

  3. See critique of this paper by Luke Jostins: