Wednesday, July 17, 2013

The deep epistemological problem in 'personalized genomic medicine' . . . that nobody wants to acknowledge

Two recent studies, both published in The Lancet, one from Denmark, and reported in this BBC story, and the other from England and reported by Gina Kolata in The New York Times, clearly illustrate one of the most important issues in personalized genomic medicine. The BBC writes, "People born in 1915 scored higher in cognitive tests in their 90s compared with those born a decade earlier, according to the study in The Lancet."  People now living into their 90s are experiencing a substantially higher quality of life than in the recent past, or that could have been predicted in any reliable way.  They are experiencing much less loss of mental function as well.  And the story in the NY Times says dementia rates are dropping as 'predicted'.  Senility, clearly, is not inherent in the human genome.

Also on the BBC site there's a story reporting that air pollution is harmful for people with at-risk hearts.  This is tragic for those people, and had been suspected but not specifically predicted.  The point is not just that we should pay heed to the quality of our air, but that we can't really predict where there will be more, or less air pollution, nor pollution by what mix of agents, and so on.   Yet, if there are genomic factors affecting heart vulnerability (regardless of whether there are other genomic factors related to how we respond to airborne pollution), we cannot reliably estimate the risk associated with those genomic factors.

These stories are interesting, given that we are being promised, with few and often rather hidden caveats, that if you just let investigators sequence your genome, they'll be able to predict your future disease risks (and companies and various advocates of genomics-everywhere, promise that other personal traits like academic ability, musical or athletic ability, or tendency to abuse drugs or commit violent crimes etc., will also be predictable from your DNA sequence).

The longevity study is great news for those of us in the dotage range!  But it has much deeper meaning when it comes to the promises being made by the genomics industry.  The aging experiences of two cohorts were very different, but surely their genotypes were not.  Thus, the genotypes of neither cohort, those born in 1905 or those born in 1915, could have been used to predict either healthy or less healthy aging.  That is, neither result was predictable from genes.  That the result was 'predicted' as described in the NY Times story doesn't mean that it was or could have been predicted in any precise way, for the reasons we discuss here, and of course this has nothing to do with specific genetically based predictions, nor can it, in any useful way.

This is what it means to point out that environment, whatever that includes, is not just a trivial variable to be regressed out in terms of genotype effects.  It also shows the hollowness of the rationale that epidemiologists often use to justify big genomics studies, that they just want to be able to regress out genotype effects so as to identify the more important environmental effects.  That assumes, inherently, that genotype-based risks are stable and well-estimated.

Clearly, and to an important extent, most predictions based on GWAS and other related omics approaches, cannot be taken seriously except for very clear-cut strong effects--most of which do not and did not require massive genome-wide studies to identify.

Since environments--physical and lifestyle, etc.--cannot be predicted, not even in principle, this is why we  repeatedly say that the promises used to justify much that is going on today in genomics and biomedical genetics is more to satisfy the investigators than to deliver the promised benefits.  The point that environmental effects are estimated retrospectively (based on today's GWAS subjects' past history), but personalized medicine is about prospective (future) risk, and that future risk cannot be predicted when environments are important.

Every week there are studies showing these points.  This is not mysterious, nor new, nor technically subtle.  But the problem is fundamental.  So is it being conveniently ignored by those who want to continue with current approaches?  Is it wrong to question their underlying motives?

1 comment:

Ken Weiss said...

A new example of our point arrives today in the NY Times, about a government report showing a narrowing of the health gap between US black and white populations. Regardless of issues about population definition itself, the amount of such change cannot have been accurately predicted in specific ways and ideas about differences in response to treatment (as, say, drugs being marketed as race-specific) is likely to change or disappear.

Since we have no crystal ball into the future, we should be more circumspect about our promises.